refactor: address review feedback

- Use pydantic Field(ge=0, le=300) instead of custom validators
- Extract shared wait logic to execution_utils.py
- Use asyncio.wait_for for proper timeout handling
- Remove duplicated code in agent_output.py and run_agent.py

Note: Direct DB access will need adjustment for #12057 compatibility

.agents/skills

@@ -1 +0,0 @@
-../.claude/skills

.claude/settings.json

@@ -1,10 +0,0 @@
-{
-  "permissions": {
-    "allowedTools": [
-      "Read", "Grep", "Glob",
-      "Bash(ls:*)", "Bash(cat:*)", "Bash(grep:*)", "Bash(find:*)",
-      "Bash(git status:*)", "Bash(git diff:*)", "Bash(git log:*)", "Bash(git worktree:*)",
-      "Bash(tmux:*)", "Bash(sleep:*)", "Bash(branchlet:*)"
-    ]
-  }
-}

.claude/skills/open-pr/SKILL.md

@@ -1,106 +0,0 @@
----
-name: open-pr
-description: Open a pull request with proper PR template, test coverage, and review workflow. Guides agents through creating a PR that follows repo conventions, ensures existing behaviors aren't broken, covers new behaviors with tests, and handles review via bot when local testing isn't possible. TRIGGER when user asks to "open a PR", "create a PR", "make a PR", "submit a PR", "open pull request", "push and create PR", or any variation of opening/submitting a pull request.
-user-invocable: true
-args: "[base-branch] — optional target branch (defaults to dev)."
-metadata:
-  author: autogpt-team
-  version: "1.0.0"
----
-
-# Open a Pull Request
-
-## Step 1: Pre-flight checks
-
-Before opening the PR:
-
-1. Ensure all changes are committed
-2. Ensure the branch is pushed to the remote (`git push -u origin <branch>`)
-3. Run linters/formatters across the whole repo (not just changed files) and commit any fixes
-
-## Step 2: Test coverage
-
-**This is critical.** Before opening the PR, verify:
-
-### Existing behavior is not broken
-- Identify which modules/components your changes touch
-- Run the existing test suites for those areas
-- If tests fail, fix them before opening the PR — do not open a PR with known regressions
-
-### New behavior has test coverage
-- Every new feature, endpoint, or behavior change needs tests
-- If you added a new block, add tests for that block
-- If you changed API behavior, add or update API tests
-- If you changed frontend behavior, verify it doesn't break existing flows
-
-If you cannot run the full test suite locally, note which tests you ran and which you couldn't in the test plan.
-
-## Step 3: Create the PR using the repo template
-
-Read the canonical PR template at `.github/PULL_REQUEST_TEMPLATE.md` and use it **verbatim** as your PR body:
-
-1. Read the template: `cat .github/PULL_REQUEST_TEMPLATE.md`
-2. Preserve the exact section titles and formatting, including:
-   - `### Why / What / How`
-   - `### Changes 🏗️`
-   - `### Checklist 📋`
-3. Replace HTML comment prompts (`<!-- ... -->`) with actual content; do not leave them in
-4. **Do not pre-check boxes** — leave all checkboxes as `- [ ]` until each step is actually completed
-5. Do not alter the template structure, rename sections, or remove any checklist items
-
-**PR title must use conventional commit format** (e.g., `feat(backend): add new block`, `fix(frontend): resolve routing bug`, `dx(skills): update PR workflow`). See CLAUDE.md for the full list of scopes.
-
-Use `gh pr create` with the base branch (defaults to `dev` if no `[base-branch]` was provided). Use `--body-file` to avoid shell interpretation of backticks and special characters:
-
-```bash
-BASE_BRANCH="${BASE_BRANCH:-dev}"
-PR_BODY=$(mktemp)
-cat > "$PR_BODY" << 'PREOF'
-<filled-in template from .github/PULL_REQUEST_TEMPLATE.md>
-PREOF
-gh pr create --base "$BASE_BRANCH" --title "<type>(scope): short description" --body-file "$PR_BODY"
-rm "$PR_BODY"
-```
-
-## Step 4: Review workflow
-
-### If you have a workspace that allows testing (docker, running backend, etc.)
-- Run `/pr-test` to do E2E manual testing of the PR using docker compose, agent-browser, and API calls. This is the most thorough way to validate your changes before review.
-- After testing, run `/pr-review` to self-review the PR for correctness, security, code quality, and testing gaps before requesting human review.
-
-### If you do NOT have a workspace that allows testing
-This is common for agents running in worktrees without a full stack. In this case:
-
-1. Run `/pr-review` locally to catch obvious issues before pushing
-2. **Comment `/review` on the PR** after creating it to trigger the review bot
-3. **Poll for the review** rather than blindly waiting — check for new review comments every 30 seconds using `gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/reviews --paginate` and the GraphQL inline threads query. The bot typically responds within 30 minutes, but polling lets the agent react as soon as it arrives.
-4. Do NOT proceed or merge until the bot review comes back
-5. Address any issues the bot raises — use `/pr-address` which has a full polling loop with CI + comment tracking
-
-```bash
-# After creating the PR:
-PR_NUMBER=$(gh pr view --json number -q .number)
-gh pr comment "$PR_NUMBER" --body "/review"
-# Then use /pr-address to poll for and address the review when it arrives
-```
-
-## Step 5: Address review feedback
-
-Once the review bot or human reviewers leave comments:
-- Run `/pr-address` to address review comments. It will loop until CI is green and all comments are resolved.
-- Do not merge without human approval.
-
-## Related skills
-
-| Skill | When to use |
-|---|---|
-| `/pr-test` | E2E testing with docker compose, agent-browser, API calls — use when you have a running workspace |
-| `/pr-review` | Review for correctness, security, code quality — use before requesting human review |
-| `/pr-address` | Address reviewer comments and loop until CI green — use after reviews come in |
-
-## Step 6: Post-creation
-
-After the PR is created and review is triggered:
-- Share the PR URL with the user
-- If waiting on the review bot, let the user know the expected wait time (~30 min)
-- Do not merge without human approval

.claude/skills/orchestrate/SKILL.md

@@ -1,709 +0,0 @@
----
-name: orchestrate
-description: "Meta-agent supervisor that manages a fleet of Claude Code agents running in tmux windows. Auto-discovers spare worktrees, spawns agents, monitors state, kicks idle agents, approves safe confirmations, and recycles worktrees when done. TRIGGER when user asks to supervise agents, run parallel tasks, manage worktrees, check agent status, or orchestrate parallel work."
-user-invocable: true
-argument-hint: "any free text — e.g. 'start 3 agents on X Y Z', 'show status', 'add task: implement feature A', 'stop', 'how many are free?'"
-metadata:
-  author: autogpt-team
-  version: "6.0.0"
----
-
-# Orchestrate — Agent Fleet Supervisor
-
-One tmux session, N windows — each window is one agent working in its own worktree. Speak naturally; Claude maps your intent to the right scripts.
-
-## Scripts
-
-```bash
-SKILLS_DIR=$(git rev-parse --show-toplevel)/.claude/skills/orchestrate/scripts
-STATE_FILE=~/.claude/orchestrator-state.json
-```
-
-| Script | Purpose |
-|---|---|
-| `find-spare.sh [REPO_ROOT]` | List free worktrees — one `PATH BRANCH` per line |
-| `spawn-agent.sh SESSION PATH SPARE NEW_BRANCH OBJECTIVE [PR_NUMBER] [STEPS...]` | Create window + checkout branch + launch claude + send task. **Stdout: `SESSION:WIN` only** |
-| `recycle-agent.sh WINDOW PATH SPARE_BRANCH` | Kill window + restore spare branch |
-| `run-loop.sh` | **Mechanical babysitter** — idle restart + dialog approval + recycle on ORCHESTRATOR:DONE + supervisor health check + all-done notification |
-| `verify-complete.sh WINDOW` | Verify PR is done: checkpoints ✓ + 0 unresolved threads + CI green + no fresh CHANGES_REQUESTED. Repo auto-derived from state file `.repo` or git remote. |
-| `notify.sh MESSAGE` | Send notification via Discord webhook (env `DISCORD_WEBHOOK_URL` or state `.discord_webhook`), macOS notification center, and stdout |
-| `capacity.sh [REPO_ROOT]` | Print available + in-use worktrees |
-| `status.sh` | Print fleet status + live pane commands |
-| `poll-cycle.sh` | One monitoring cycle — classifies panes, tracks checkpoints, returns JSON action array |
-| `classify-pane.sh WINDOW` | Classify one pane state |
-
-## Supervision model
-
-```
-Orchestrating Claude (this Claude session — IS the supervisor)
-  └── Reads pane output, checks CI, intervenes with targeted guidance
-        run-loop.sh (separate tmux window, every 30s)
-          └── Mechanical only: idle restart, dialog approval, recycle on ORCHESTRATOR:DONE
-```
-
-**You (the orchestrating Claude)** are the supervisor. After spawning agents, stay in this conversation and actively monitor: poll each agent's pane every 2-3 minutes, check CI, nudge stalled agents, and verify completions. Do not spawn a separate supervisor Claude window — it loses context, is hard to observe, and compounds context compression problems.
-
-**run-loop.sh** is the mechanical layer — zero tokens, handles things that need no judgment: restart crashed agents, press Enter on dialogs, recycle completed worktrees (only after `verify-complete.sh` passes).
-
-## Checkpoint protocol
-
-Agents output checkpoints as they complete each required step:
-
-```
-CHECKPOINT:<step-name>
-```
-
-Required steps are passed as args to `spawn-agent.sh` (e.g. `pr-address pr-test`). `run-loop.sh` will not recycle a window until all required checkpoints are found in the pane output. If `verify-complete.sh` fails, the agent is re-briefed automatically.
-
-## Worktree lifecycle
-
-```text
-spare/N branch  →  spawn-agent.sh (--session-id UUID)  →  window + feat/branch + claude running
-                                                                 ↓
-                                               CHECKPOINT:<step> (as steps complete)
-                                                                 ↓
-                                                        ORCHESTRATOR:DONE
-                                                                 ↓
-                                    verify-complete.sh: checkpoints ✓ + 0 threads + CI green + no fresh CHANGES_REQUESTED
-                                                                 ↓
-                                              state → "done", notify, window KEPT OPEN
-                                                                 ↓
-                              user/orchestrator explicitly requests recycle
-                                                                 ↓
-                                         recycle-agent.sh → spare/N (free again)
-```
-
-**Windows are never auto-killed.** The worktree stays on its branch, the session stays alive. The agent is done working but the window, git state, and Claude session are all preserved until you choose to recycle.
-
-**To resume a done or crashed session:**
-```bash
-# Resume by stored session ID (preferred — exact session, full context)
-claude --resume SESSION_ID --permission-mode bypassPermissions
-
-# Or resume most recent session in that worktree directory
-cd /path/to/worktree && claude --continue --permission-mode bypassPermissions
-```
-
-**To manually recycle when ready:**
-```bash
-bash ~/.claude/orchestrator/scripts/recycle-agent.sh SESSION:WIN WORKTREE_PATH spare/N
-# Then update state:
-jq --arg w "SESSION:WIN" '.agents |= map(if .window == $w then .state = "recycled" else . end)' \
-  ~/.claude/orchestrator-state.json > /tmp/orch.tmp && mv /tmp/orch.tmp ~/.claude/orchestrator-state.json
-```
-
-## State file (`~/.claude/orchestrator-state.json`)
-
-Never committed to git. You maintain this file directly using `jq` + atomic writes (`.tmp` → `mv`).
-
-```json
-{
-  "active": true,
-  "tmux_session": "autogpt1",
-  "idle_threshold_seconds": 300,
-  "loop_window": "autogpt1:5",
-  "repo": "Significant-Gravitas/AutoGPT",
-  "discord_webhook": "https://discord.com/api/webhooks/...",
-  "last_poll_at": 0,
-  "agents": [
-    {
-      "window": "autogpt1:3",
-      "worktree": "AutoGPT6",
-      "worktree_path": "/path/to/AutoGPT6",
-      "spare_branch": "spare/6",
-      "branch": "feat/my-feature",
-      "objective": "Implement X and open a PR",
-      "pr_number": "12345",
-      "session_id": "550e8400-e29b-41d4-a716-446655440000",
-      "steps": ["pr-address", "pr-test"],
-      "checkpoints": ["pr-address"],
-      "state": "running",
-      "last_output_hash": "",
-      "last_seen_at": 0,
-      "spawned_at": 0,
-      "idle_since": 0,
-      "revision_count": 0,
-      "last_rebriefed_at": 0
-    }
-  ]
-}
-```
-
-Top-level optional fields:
-- `repo` — GitHub `owner/repo` for CI/thread checks. Auto-derived from git remote if omitted.
-- `discord_webhook` — Discord webhook URL for completion notifications. Also reads `DISCORD_WEBHOOK_URL` env var.
-
-Per-agent fields:
-- `session_id` — UUID passed to `claude --session-id` at spawn; use with `claude --resume UUID` to restore exact session context after a crash or window close.
-- `last_rebriefed_at` — Unix timestamp of last re-brief; enforces 5-min cooldown to prevent spam.
-
-Agent states: `running` | `idle` | `stuck` | `waiting_approval` | `complete` | `done` | `escalated`
-
-`done` means verified complete — window is still open, session still alive, worktree still on task branch. Not recycled yet.
-
-## Serial /pr-test rule
-
-`/pr-test` and `/pr-test --fix` run local Docker + integration tests that use shared ports, a shared database, and shared build caches. **Running two `/pr-test` jobs simultaneously will cause port conflicts and database corruption.**
-
-**Rule: only one `/pr-test` runs at a time. The orchestrator serializes them.**
-
-You (the orchestrating Claude) own the test queue:
-1. Agents do `pr-review` and `pr-address` in parallel — that's safe (they only push code and reply to GitHub).
-2. When a PR needs local testing, add it to your mental queue — don't give agents a `pr-test` step.
-3. Run `/pr-test https://github.com/OWNER/REPO/pull/PR_NUMBER --fix` yourself, sequentially.
-4. Feed results back to the relevant agent via `tmux send-keys`:
-   ```bash
-   tmux send-keys -t SESSION:WIN "Local tests for PR #N: <paste failure output or 'all passed'>. Fix any failures and push, then output ORCHESTRATOR:DONE."
-   sleep 0.3
-   tmux send-keys -t SESSION:WIN Enter
-   ```
-5. Wait for CI to confirm green before marking the agent done.
-
-If multiple PRs need testing at the same time, pick the one furthest along (fewest pending CI checks) and test it first. Only start the next test after the previous one completes.
-
-## Session restore (tested and confirmed)
-
-Agent sessions are saved to disk. To restore a closed or crashed session:
-
-```bash
-# If session_id is in state (preferred):
-NEW_WIN=$(tmux new-window -t SESSION -n WORKTREE_NAME -P -F '#{window_index}')
-tmux send-keys -t "SESSION:${NEW_WIN}" "cd /path/to/worktree && claude --resume SESSION_ID --permission-mode bypassPermissions" Enter
-
-# If no session_id (use --continue for most recent session in that directory):
-tmux send-keys -t "SESSION:${NEW_WIN}" "cd /path/to/worktree && claude --continue --permission-mode bypassPermissions" Enter
-```
-
-`--continue` restores the full conversation history including all tool calls, file edits, and context. The agent resumes exactly where it left off. After restoring, update the window address in the state file:
-
-```bash
-jq --arg old "SESSION:OLD_WIN" --arg new "SESSION:NEW_WIN" \
-  '(.agents[] | select(.window == $old)).window = $new' \
-  ~/.claude/orchestrator-state.json > /tmp/orch.tmp && mv /tmp/orch.tmp ~/.claude/orchestrator-state.json
-```
-
-## Intent → action mapping
-
-Match the user's message to one of these intents:
-
-| The user says something like… | What to do |
-|---|---|
-| "status", "what's running", "show agents" | Run `status.sh` + `capacity.sh`, show output |
-| "how many free", "capacity", "available worktrees" | Run `capacity.sh`, show output |
-| "start N agents on X, Y, Z" or "run these tasks: …" | See **Spawning agents** below |
-| "add task: …", "add one more agent for …" | See **Adding an agent** below |
-| "stop", "shut down", "pause the fleet" | See **Stopping** below |
-| "poll", "check now", "run a cycle" | Run `poll-cycle.sh`, process actions |
-| "recycle window X", "free up autogpt3" | Run `recycle-agent.sh` directly |
-
-When the intent is ambiguous, show capacity first and ask what tasks to run.
-
-## Spawning agents
-
-### 1. Resolve tmux session
-
-```bash
-tmux list-sessions -F "#{session_name}: #{session_windows} windows" 2>/dev/null
-```
-
-Use an existing session. **Never create a tmux session from within Claude** — it becomes a child of Claude's process and dies when the session ends. If no session exists, tell the user to run `tmux new-session -d -s autogpt1` in their terminal first, then re-invoke `/orchestrate`.
-
-### 2. Show available capacity
-
-```bash
-bash $SKILLS_DIR/capacity.sh $(git rev-parse --show-toplevel)
-```
-
-### 3. Collect tasks from the user
-
-For each task, gather:
-- **objective** — what to do (e.g. "implement feature X and open a PR")
-- **branch name** — e.g. `feat/my-feature` (derive from objective if not given)
-- **pr_number** — GitHub PR number if working on an existing PR (for verification)
-- **steps** — required checkpoint names in order (e.g. `pr-address pr-test`) — derive from objective
-
-Ask for `idle_threshold_seconds` only if the user mentions it (default: 300).
-
-Never ask the user to specify a worktree — auto-assign from `find-spare.sh`.
-
-### 4. Spawn one agent per task
-
-```bash
-# Get ordered list of spare worktrees
-SPARE_LIST=$(bash $SKILLS_DIR/find-spare.sh $(git rev-parse --show-toplevel))
-
-# For each task, take the next spare line:
-WORKTREE_PATH=$(echo "$SPARE_LINE" | awk '{print $1}')
-SPARE_BRANCH=$(echo "$SPARE_LINE" | awk '{print $2}')
-
-# With PR number and required steps:
-WINDOW=$(bash $SKILLS_DIR/spawn-agent.sh "$SESSION" "$WORKTREE_PATH" "$SPARE_BRANCH" "$NEW_BRANCH" "$OBJECTIVE" "$PR_NUMBER" "pr-address" "pr-test")
-
-# Without PR (new work):
-WINDOW=$(bash $SKILLS_DIR/spawn-agent.sh "$SESSION" "$WORKTREE_PATH" "$SPARE_BRANCH" "$NEW_BRANCH" "$OBJECTIVE")
-```
-
-Build an agent record and append it to the state file. If the state file doesn't exist yet, initialize it:
-
-```bash
-# Derive repo from git remote (used by verify-complete.sh + supervisor)
-REPO=$(git remote get-url origin 2>/dev/null | sed 's|.*github\.com[:/]||; s|\.git$||' || echo "")
-
-jq -n \
-  --arg session "$SESSION" \
-  --arg repo "$REPO" \
-  --argjson threshold 300 \
-  '{active:true, tmux_session:$session, idle_threshold_seconds:$threshold,
-    repo:$repo, loop_window:null, supervisor_window:null, last_poll_at:0, agents:[]}' \
-  > ~/.claude/orchestrator-state.json
-```
-
-Optionally add a Discord webhook for completion notifications:
-```bash
-jq --arg hook "$DISCORD_WEBHOOK_URL" '.discord_webhook = $hook' ~/.claude/orchestrator-state.json \
-  > /tmp/orch.tmp && mv /tmp/orch.tmp ~/.claude/orchestrator-state.json
-```
-
-`spawn-agent.sh` writes the initial agent record (window, worktree_path, branch, objective, state, etc.) to the state file automatically — **do not append the record again after calling it.** The record already exists and `pr_number`/`steps` are patched in by the script itself.
-
-### 5. Start the mechanical babysitter
-
-```bash
-LOOP_WIN=$(tmux new-window -t "$SESSION" -n "orchestrator" -P -F '#{window_index}')
-LOOP_WINDOW="${SESSION}:${LOOP_WIN}"
-tmux send-keys -t "$LOOP_WINDOW" "bash $SKILLS_DIR/run-loop.sh" Enter
-
-jq --arg w "$LOOP_WINDOW" '.loop_window = $w' ~/.claude/orchestrator-state.json \
-  > /tmp/orch.tmp && mv /tmp/orch.tmp ~/.claude/orchestrator-state.json
-```
-
-### 6. Begin supervising directly in this conversation
-
-You are the supervisor. After spawning, immediately start your first poll loop (see **Supervisor duties** below) and continue every 2-3 minutes. Do NOT spawn a separate supervisor Claude window.
-
-## Adding an agent
-
-Find the next spare worktree, then spawn and append to state — same as steps 2–4 above but for a single task. If no spare worktrees are available, tell the user.
-
-## Supervisor duties (YOUR job, every 2-3 min in this conversation)
-
-You are the supervisor. Run this poll loop directly in your Claude session — not in a separate window.
-
-### Poll loop mechanism
-
-You are reactive — you only act when a tool completes or the user sends a message. To create a self-sustaining poll loop without user involvement:
-
-1. Start each poll with `run_in_background: true` + a sleep before the work:
-   ```bash
-   sleep 120 && tmux capture-pane -t autogpt1:0 -p -S -200 | tail -40
-   # + similar for each active window
-   ```
-2. When the background job notifies you, read the pane output and take action.
-3. Immediately schedule the next background poll — this keeps the loop alive.
-4. Stop scheduling when all agents are done/escalated.
-
-**Never tell the user "I'll poll every 2-3 minutes"** — that does nothing without a trigger. Start the background job instead.
-
-### Each poll: what to check
-
-```bash
-# 1. Read state
-cat ~/.claude/orchestrator-state.json | jq '.agents[] | {window, worktree, branch, state, pr_number, checkpoints}'
-
-# 2. For each running/stuck/idle agent, capture pane
-tmux capture-pane -t SESSION:WIN -p -S -200 | tail -60
-```
-
-For each agent, decide:
-
-| What you see | Action |
-|---|---|
-| Spinner / tools running | Do nothing — agent is working |
-| Idle `❯` prompt, no `ORCHESTRATOR:DONE` | Stalled — send specific nudge with objective from state |
-| Stuck in error loop | Send targeted fix with exact error + solution |
-| Waiting for input / question | Answer and unblock via `tmux send-keys` |
-| CI red | `gh pr checks PR_NUMBER --repo REPO` → tell agent exactly what's failing |
-| GitHub abuse rate limit error | Nudge: "Wait 60 seconds then continue posting replies with sleep 3 between each" |
-| Context compacted / agent lost | Send recovery: `cat ~/.claude/orchestrator-state.json | jq '.agents[] | select(.window=="WIN")'` + `gh pr view PR_NUMBER --json title,body` |
-| `ORCHESTRATOR:DONE` in output | Query GraphQL for actual unresolved count. If >0, re-brief. If 0, run `verify-complete.sh` |
-
-**Poll all windows from state, not from memory.** Before each poll, run:
-```bash
-jq -r '.agents[] | select(.state | test("running|idle|stuck|waiting_approval|pending_evaluation")) | .window' ~/.claude/orchestrator-state.json
-```
-and capture every window listed. If you manually added a window outside spawn-agent.sh, ensure it's in the state file first.
-
-### RUNNING count includes waiting_approval agents
-
-The `RUNNING` count from run-loop.sh includes agents in `waiting_approval` state (they match the regex `running|stuck|waiting_approval|idle`). This means a fleet that is only `waiting_approval` still shows RUNNING > 0 in the log — it does **not** mean agents are actively working.
-
-When you see `RUNNING > 0` in the run-loop log but suspect agents are actually blocked, check state directly:
-```bash
-jq '.agents[] | {window, state, worktree}' ~/.claude/orchestrator-state.json
-```
-A count of `running=1 waiting=1` in the log actually means one agent is waiting for approval — the orchestrator should check and approve, not wait.
-
-### State file staleness recovery
-
-The state file is written by scripts but can drift from reality when windows are closed, sessions expire, or the orchestrator restarts across conversations.
-
-**Signs of stale state:**
-- `loop_window` points to a window that no longer exists in the tmux session
-- An agent's `state` is `running` but tmux window is closed or shows a shell prompt (not claude)
-- `last_seen_at` is hours old but state still says `running`
-
-**Recovery steps:**
-
-1. **Verify actual tmux windows:**
-```bash
-tmux list-windows -t SESSION -F '#{window_index}: #{window_name} (#{pane_current_command})'
-```
-
-2. **Cross-reference with state file:**
-```bash
-jq -r '.agents[] | "\(.window) \(.state) \(.worktree)"' ~/.claude/orchestrator-state.json
-```
-
-3. **Fix stale entries:**
-```bash
-# Agent window closed — mark idle so run-loop.sh will restart it
-jq --arg w "SESSION:WIN" '(.agents[] | select(.window==$w)).state = "idle"' \
-  ~/.claude/orchestrator-state.json > /tmp/orch.tmp && mv /tmp/orch.tmp ~/.claude/orchestrator-state.json
-
-# loop_window gone — kill the stale reference, then restart run-loop.sh
-jq '.loop_window = null' ~/.claude/orchestrator-state.json > /tmp/orch.tmp && mv /tmp/orch.tmp ~/.claude/orchestrator-state.json
-LOOP_WIN=$(tmux new-window -t "$SESSION" -n "orchestrator" -P -F '#{window_index}')
-LOOP_WINDOW="${SESSION}:${LOOP_WIN}"
-tmux send-keys -t "$LOOP_WINDOW" "bash $SKILLS_DIR/run-loop.sh" Enter
-jq --arg w "$LOOP_WINDOW" '.loop_window = $w' ~/.claude/orchestrator-state.json \
-  > /tmp/orch.tmp && mv /tmp/orch.tmp ~/.claude/orchestrator-state.json
-```
-
-4. **After any state repair, re-run `status.sh` to confirm coherence before resuming supervision.**
-
-### Strict ORCHESTRATOR:DONE gate
-
-`verify-complete.sh` handles the main checks automatically (checkpoints, threads, CI green, spawned_at, and CHANGES_REQUESTED). Run it:
-
-**CHANGES_REQUESTED staleness rule**: a `CHANGES_REQUESTED` review only blocks if it was submitted *after* the latest commit. If the latest commit postdates the review, the review is considered stale (feedback already addressed) and does not block. This avoids false negatives when a bot reviewer hasn't re-reviewed after the agent's fixing commits.
-
-```bash
-SKILLS_DIR=~/.claude/orchestrator/scripts
-bash $SKILLS_DIR/verify-complete.sh SESSION:WIN
-```
-
-If it passes → run-loop.sh will recycle the window automatically. No manual action needed.
-If it fails → re-brief the agent with the failure reason. Never manually mark state `done` to bypass this.
-
-### Re-brief a stalled agent
-
-**Before sending any nudge, verify the pane is at an idle ❯ prompt.** Sending text into a still-processing pane produces stuck `[Pasted text +N lines]` that the agent never sees.
-
-Check:
-```bash
-tmux capture-pane -t SESSION:WIN -p 2>/dev/null | tail -5
-```
-If the last line shows a spinner (✳✽✢✶·), `Running…`, or no `❯` — wait 10–15s and check again before sending.
-
-```bash
-OBJ=$(jq -r --arg w SESSION:WIN '.agents[] | select(.window==$w) | .objective' ~/.claude/orchestrator-state.json)
-PR=$(jq -r --arg w SESSION:WIN '.agents[] | select(.window==$w) | .pr_number' ~/.claude/orchestrator-state.json)
-tmux send-keys -t SESSION:WIN "You appear stalled. Your objective: $OBJ. Check: gh pr view $PR --json title,body,headRefName to reorient."
-sleep 0.3
-tmux send-keys -t SESSION:WIN Enter
-```
-
-If `image_path` is set on the agent record, include: "Re-read context at IMAGE_PATH with the Read tool."
-
-## Self-recovery protocol (agents)
-
-spawn-agent.sh automatically includes this instruction in every objective:
-
-> If your context compacts and you lose track of what to do, run:
-> `cat ~/.claude/orchestrator-state.json | jq '.agents[] | select(.window=="SESSION:WIN")'`
-> and `gh pr view PR_NUMBER --json title,body,headRefName` to reorient.
-> Output each completed step as `CHECKPOINT:<step-name>` on its own line.
-
-## Passing images and screenshots to agents
-
-`tmux send-keys` is text-only — you cannot paste a raw image into a pane. To give an agent visual context (screenshots, diagrams, mockups):
-
-1. **Save the image to a temp file** with a stable path:
-   ```bash
-   # If the user drags in a screenshot or you receive a file path:
-   IMAGE_PATH="/tmp/orchestrator-context-$(date +%s).png"
-   cp "$USER_PROVIDED_PATH" "$IMAGE_PATH"
-   ```
-
-2. **Reference the path in the objective string**:
-   ```bash
-   OBJECTIVE="Implement the layout shown in /tmp/orchestrator-context-1234567890.png. Read that image first with the Read tool to understand the design."
-   ```
-
-3. The agent uses its `Read` tool to view the image at startup — Claude Code agents are multimodal and can read image files directly.
-
-**Rule**: always use `/tmp/orchestrator-context-<timestamp>.png` as the naming convention so the supervisor knows what to look for if it needs to re-brief an agent with the same image.
-
----
-
-## Orchestrator final evaluation (YOU decide, not the script)
-
-`verify-complete.sh` is a gate — it blocks premature marking. But it cannot tell you if the work is actually good. That is YOUR job.
-
-When run-loop marks an agent `pending_evaluation` and you're notified, do all of these before marking done:
-
-### 1. Run /pr-test (required, serialized, use TodoWrite to queue)
-
-`/pr-test` is the only reliable confirmation that the objective is actually met. Run it yourself, not the agent.
-
-**When multiple PRs reach `pending_evaluation` at the same time, use TodoWrite to queue them:**
-```
-- [ ] /pr-test https://github.com/Significant-Gravitas/AutoGPT/pull/NNNN — <feature description>
-- [ ] /pr-test https://github.com/Significant-Gravitas/AutoGPT/pull/MMMM — <feature description>
-```
-Run one at a time. Check off as you go.
-
-```
-/pr-test https://github.com/Significant-Gravitas/AutoGPT/pull/PR_NUMBER
-```
-
-**/pr-test can be lazy** — if it gives vague output, re-run with full context:
-
-```
-/pr-test https://github.com/OWNER/REPO/pull/PR_NUMBER
-Context: This PR implements <objective from state file>. Key files: <list>.
-Please verify: <specific behaviors to check>.
-```
-
-Only one `/pr-test` at a time — they share ports and DB.
-
-### /pr-test result evaluation
-
-**PARTIAL on any headline feature scenario is an immediate blocker.** Do not approve, do not mark done, do not let the agent output `ORCHESTRATOR:DONE`.
-
-| `/pr-test` result | Action |
-|---|---|
-| All headline scenarios **PASS** | Proceed to evaluation step 2 |
-| Any headline scenario **PARTIAL** | Re-brief the agent immediately — see below |
-| Any headline scenario **FAIL** | Re-brief the agent immediately |
-
-**What PARTIAL means**: the feature is only partly working. Example: the Apply button never appeared, or the AI returned no action blocks. The agent addressed part of the objective but not all of it.
-
-**When any headline scenario is PARTIAL or FAIL:**
-
-1. Do NOT mark the agent done or accept `ORCHESTRATOR:DONE`
-2. Re-brief the agent with the specific scenario that failed and what was missing:
-   ```bash
-   tmux send-keys -t SESSION:WIN "PARTIAL result on /pr-test — S5 (Apply button) never appeared. The AI must output JSON action blocks for the Apply button to render. Fix this before re-running /pr-test."
-   sleep 0.3
-   tmux send-keys -t SESSION:WIN Enter
-   ```
-3. Set state back to `running`:
-   ```bash
-   jq --arg w "SESSION:WIN" '(.agents[] | select(.window == $w)).state = "running"' \
-     ~/.claude/orchestrator-state.json > /tmp/orch.tmp && mv /tmp/orch.tmp ~/.claude/orchestrator-state.json
-   ```
-4. Wait for new `ORCHESTRATOR:DONE`, then re-run `/pr-test` from scratch
-
-**Rule: only ALL-PASS qualifies for approval.** A mix of PASS + PARTIAL is a failure.
-
-> **Why this matters**: A PR was once wrongly approved with S5 PARTIAL — the AI never output JSON action blocks so the Apply button never appeared. The fix was already in the agent's reach but slipped through because PARTIAL was not treated as blocking.
-
-### 2. Do your own evaluation
-
-1. **Read the PR diff and objective** — does the code actually implement what was asked? Is anything obviously missing or half-done?
-2. **Read the resolved threads** — were comments addressed with real fixes, or just dismissed/resolved without changes?
-3. **Check CI run names** — any suspicious retries that shouldn't have passed?
-4. **Check the PR description** — title, summary, test plan complete?
-
-### 3. Decide
-
-- `/pr-test` all scenarios PASS + evaluation looks good → mark `done` in state, tell the user the PR is ready, ask if window should be closed
-- `/pr-test` any scenario PARTIAL or FAIL → re-brief the agent with the specific failing scenario, set state back to `running` (see `/pr-test result evaluation` above)
-- Evaluation finds gaps even with all PASS → re-brief the agent with specific gaps, set state back to `running`
-
-**Never mark done based purely on script output.** You hold the full objective context; the script does not.
-
-```bash
-# Mark done after your positive evaluation:
-jq --arg w "SESSION:WIN" '(.agents[] | select(.window == $w)).state = "done"' \
-  ~/.claude/orchestrator-state.json > /tmp/orch.tmp && mv /tmp/orch.tmp ~/.claude/orchestrator-state.json
-```
-
-## When to stop the fleet
-
-Stop the fleet (`active = false`) when **all** of the following are true:
-
-| Check | How to verify |
-|---|---|
-| All agents are `done` or `escalated` | `jq '[.agents[] | select(.state | test("running\|stuck\|idle\|waiting_approval"))] | length' ~/.claude/orchestrator-state.json` == 0 |
-| All PRs have 0 unresolved review threads | GraphQL `isResolved` check per PR |
-| All PRs have green CI **on a run triggered after the agent's last push** | `gh run list --branch BRANCH --limit 1` timestamp > `spawned_at` in state |
-| No fresh CHANGES_REQUESTED (after latest commit) | `verify-complete.sh` checks this — stale pre-commit reviews are ignored |
-| No agents are `escalated` without human review | If any are escalated, surface to user first |
-
-**Do NOT stop just because agents output `ORCHESTRATOR:DONE`.** That is a signal to verify, not a signal to stop.
-
-**Do stop** if the user explicitly says "stop", "shut down", or "kill everything", even with agents still running.
-
-```bash
-# Graceful stop
-jq '.active = false' ~/.claude/orchestrator-state.json > /tmp/orch.tmp \
-  && mv /tmp/orch.tmp ~/.claude/orchestrator-state.json
-
-LOOP_WINDOW=$(jq -r '.loop_window // ""' ~/.claude/orchestrator-state.json)
-[ -n "$LOOP_WINDOW" ] && tmux kill-window -t "$LOOP_WINDOW" 2>/dev/null || true
-```
-
-Does **not** recycle running worktrees — agents may still be mid-task. Run `capacity.sh` to see what's still in progress.
-
-## tmux send-keys pattern
-
-**Always split long messages into text + Enter as two separate calls with a sleep between them.** If sent as one call (`"text" Enter`), Enter can fire before the full string is buffered into Claude's input — leaving the message stuck as `[Pasted text +N lines]` unsent.
-
-```bash
-# CORRECT — text then Enter separately
-tmux send-keys -t "$WINDOW" "your long message here"
-sleep 0.3
-tmux send-keys -t "$WINDOW" Enter
-
-# WRONG — Enter may fire before text is buffered
-tmux send-keys -t "$WINDOW" "your long message here" Enter
-```
-
-Short single-character sends (`y`, `Down`, empty Enter for dialog approval) are safe to combine since they have no buffering lag.
-
----
-
-## Protected worktrees
-
-Some worktrees must **never** be used as spare worktrees for agent tasks because they host files critical to the orchestrator itself:
-
-| Worktree | Protected branch | Why |
-|---|---|---|
-| `AutoGPT1` | `dx/orchestrate-skill` | Hosts the orchestrate skill scripts. `recycle-agent.sh` would check out `spare/1`, wiping `.claude/skills/` and breaking all subsequent `spawn-agent.sh` calls. |
-
-**Rule**: when selecting spare worktrees via `find-spare.sh`, skip any worktree whose CURRENT branch matches a protected branch. If you accidentally spawn an agent in a protected worktree, do not let `recycle-agent.sh` run on it — manually restore the branch after the agent finishes.
-
-When `dx/orchestrate-skill` is merged into `dev`, `AutoGPT1` becomes a normal spare again.
-
----
-
-## Thread resolution integrity (critical)
-
-**Agents MUST NOT resolve review threads via GraphQL unless a real code fix has been committed and pushed first.**
-
-This is the most common failure mode: agents call `resolveReviewThread` to make unresolved counts drop without actually fixing anything. This produces a false "done" signal that gets past verify-complete.sh.
-
-**The only valid resolution sequence:**
-1. Read the thread and understand what it's asking
-2. Make the actual code change
-3. `git commit` and `git push`
-4. Reply to the thread with the commit SHA (e.g. "Fixed in `abc1234`")
-5. THEN call `resolveReviewThread`
-
-**The supervisor must verify actual thread counts via GraphQL** — never trust an agent's claim of "0 unresolved." After any agent's ORCHESTRATOR:DONE, always run:
-
-```bash
-# Step 1: get total count
-TOTAL=$(gh api graphql -f query='{ repository(owner: "OWNER", name: "REPO") { pullRequest(number: PR) { reviewThreads { totalCount } } } }' \
-  | jq '.data.repository.pullRequest.reviewThreads.totalCount')
-echo "Total threads: $TOTAL"
-
-# Step 2: paginate all pages and count unresolved
-CURSOR=""; UNRESOLVED=0
-while true; do
-  AFTER=${CURSOR:+", after: \"$CURSOR\""}
-  PAGE=$(gh api graphql -f query="{ repository(owner: \"OWNER\", name: \"REPO\") { pullRequest(number: PR) { reviewThreads(first: 100${AFTER}) { pageInfo { hasNextPage endCursor } nodes { isResolved } } } } }")
-  UNRESOLVED=$(( UNRESOLVED + $(echo "$PAGE" | jq '[.data.repository.pullRequest.reviewThreads.nodes[] | select(.isResolved==false)] | length') ))
-  HAS_NEXT=$(echo "$PAGE" | jq -r '.data.repository.pullRequest.reviewThreads.pageInfo.hasNextPage')
-  CURSOR=$(echo "$PAGE" | jq -r '.data.repository.pullRequest.reviewThreads.pageInfo.endCursor')
-  [ "$HAS_NEXT" = "false" ] && break
-done
-echo "Unresolved: $UNRESOLVED"
-```
-
-If unresolved > 0, the agent is NOT done — re-brief with the actual count and the rule.
-
-**Include this in every agent objective:**
-> IMPORTANT: Do NOT resolve any review thread via GraphQL unless the code fix is committed and pushed first. Fix the code → commit → push → reply with SHA → then resolve. Never resolve without a real commit. "Accepted" or "Acknowledged" replies are NOT resolutions — only real commits qualify.
-
-### Detecting fake resolutions
-
-When an agent claims "0 unresolved threads", query GitHub GraphQL yourself and also inspect how each thread was resolved. A resolved thread whose last comment is `"Acknowledged"`, `"Same as above"`, `"Accepted trade-off"`, or `"Deferred"` — with no commit SHA — is a fake resolution.
-
-To spot these, paginate all pages and collect resolved threads with missing SHA links:
-```bash
-# Paginate all pages — first:100 misses threads beyond page 1 on large PRs
-CURSOR=""; FAKE_RESOLUTIONS="[]"
-while true; do
-  AFTER=${CURSOR:+", after: \"$CURSOR\""}
-  PAGE=$(gh api graphql -f query="
-  {
-    repository(owner: \"Significant-Gravitas\", name: \"AutoGPT\") {
-      pullRequest(number: PR_NUMBER) {
-        reviewThreads(first: 100${AFTER}) {
-          pageInfo { hasNextPage endCursor }
-          nodes {
-            isResolved
-            comments(last: 1) {
-              nodes { body author { login } }
-            }
-          }
-        }
-      }
-    }
-  }")
-  PAGE_FAKES=$(echo "$PAGE" | jq '[.data.repository.pullRequest.reviewThreads.nodes[]
-    | select(.isResolved == true)
-    | {body: .comments.nodes[0].body[:120], author: .comments.nodes[0].author.login}
-    | select(.body | test("Fixed in|Removed in|Addressed in") | not)]')
-  FAKE_RESOLUTIONS=$(echo "$FAKE_RESOLUTIONS $PAGE_FAKES" | jq -s 'add')
-  HAS_NEXT=$(echo "$PAGE" | jq -r '.data.repository.pullRequest.reviewThreads.pageInfo.hasNextPage')
-  CURSOR=$(echo "$PAGE" | jq -r '.data.repository.pullRequest.reviewThreads.pageInfo.endCursor')
-  [ "$HAS_NEXT" = "false" ] && break
-done
-echo "$FAKE_RESOLUTIONS"
-```
-Any resolved thread whose last comment does NOT contain `"Fixed in"`, `"Removed in"`, or `"Addressed in"` (with a commit link) should be investigated — either the agent falsely resolved it, or it was a genuine false positive that needs explanation.
-
-## GitHub abuse rate limits
-
-Two distinct rate limits exist with different recovery times:
-
-| Error | HTTP status | Cause | Recovery |
-|---|---|---|---|
-| `{"code":"abuse"}` in body | 403 | Secondary rate limit — too many write operations (comments, mutations) in a short window | Wait **2–3 minutes**. 60s is often not enough. |
-| `API rate limit exceeded` | 429 | Primary rate limit — too many read calls per hour | Wait until `X-RateLimit-Reset` timestamp |
-
-**Prevention:** Agents must add `sleep 3` between individual thread reply API calls. For >20 unresolved threads, increase to `sleep 5`.
-
-If you see a 403 `abuse` error from an agent's pane:
-1. Nudge the agent: `"You hit a GitHub secondary rate limit (403). Stop all API writes. Wait 2 minutes, then resume with sleep 3 between each thread reply."`
-2. Do NOT nudge again during the 2-minute wait — a second nudge restarts the clock.
-
-Add this to agent briefings when there are >20 unresolved threads:
-> Post replies with `sleep 3` between each reply. If you hit a 403 abuse error, wait 2 minutes (not 60s — secondary limits take longer to clear) then continue.
-
-## Key rules
-
-1. **Scripts do all the heavy lifting** — don't reimplement their logic inline in this file
-2. **Never ask the user to pick a worktree** — auto-assign from `find-spare.sh` output
-3. **Never restart a running agent** — only restart on `idle` kicks (foreground is a shell)
-4. **Auto-dismiss settings dialogs** — if "Enter to confirm" appears, send Down+Enter
-5. **Always `--permission-mode bypassPermissions`** on every spawn
-6. **Escalate after 3 kicks** — mark `escalated`, surface to user
-7. **Atomic state writes** — always write to `.tmp` then `mv`
-8. **Never approve destructive commands** outside the worktree scope — when in doubt, escalate
-9. **Never recycle without verification** — `verify-complete.sh` must pass before recycling
-10. **No TASK.md files** — commit risk; use state file + `gh pr view` for agent context persistence
-11. **Re-brief stalled agents** — read objective from state file + `gh pr view`, send via tmux
-12. **ORCHESTRATOR:DONE is a signal to verify, not to accept** — always run `verify-complete.sh` and check CI run timestamp before recycling
-13. **Protected worktrees** — never use the worktree hosting the skill scripts as a spare
-14. **Images via file path** — save screenshots to `/tmp/orchestrator-context-<ts>.png`, pass path in objective; agents read with the `Read` tool
-15. **Split send-keys** — always separate text and Enter with `sleep 0.3` between calls for long strings
-16. **Poll ALL windows from state file** — never hardcode window count. Derive active windows dynamically: `jq -r '.agents[] | select(.state | test("running|idle|stuck")) | .window' ~/.claude/orchestrator-state.json`. If you added a window mid-session outside spawn-agent.sh, add it to the state file immediately.
-20. **Orchestrator handles its own approvals** — when spawning a subagent to make edits (SKILL.md, scripts, config), review the diff yourself and approve/reject without surfacing it to the user. The user should never have to open a file to check the orchestrator's work. Use the Agent tool with `subagent_type: general-purpose` for drafting, then verify the result yourself before considering the task done.
-17. **Update state file on re-task** — whenever an agent is re-tasked mid-session (objective changes, new PR assigned), update the state file record immediately so objectives stay accurate for re-briefing after compaction.
-18. **No GraphQL resolveReviewThread without a commit** — see Thread resolution integrity above. This is rule #1 for pr-address work.
-19. **Verify thread counts yourself** — after any agent claims "0 unresolved threads", query GitHub GraphQL directly before accepting. Never trust the agent's self-report.

.claude/skills/orchestrate/scripts/capacity.sh

@@ -1,43 +0,0 @@
-#!/usr/bin/env bash
-# capacity.sh — show fleet capacity: available spare worktrees + in-use agents
-#
-# Usage: capacity.sh [REPO_ROOT]
-#   REPO_ROOT defaults to the root worktree of the current git repo.
-#
-# Reads: ~/.claude/orchestrator-state.json (skipped if missing or corrupt)
-
-set -euo pipefail
-
-SCRIPTS_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-STATE_FILE="${ORCHESTRATOR_STATE_FILE:-$HOME/.claude/orchestrator-state.json}"
-REPO_ROOT="${1:-$(git rev-parse --show-toplevel 2>/dev/null || echo "")}"
-
-echo "=== Available (spare) worktrees ==="
-if [ -n "$REPO_ROOT" ]; then
-  SPARE=$("$SCRIPTS_DIR/find-spare.sh" "$REPO_ROOT" 2>/dev/null || echo "")
-else
-  SPARE=$("$SCRIPTS_DIR/find-spare.sh" 2>/dev/null || echo "")
-fi
-
-if [ -z "$SPARE" ]; then
-  echo "  (none)"
-else
-  while IFS= read -r line; do
-    [ -z "$line" ] && continue
-    echo "  ✓ $line"
-  done <<< "$SPARE"
-fi
-
-echo ""
-echo "=== In-use worktrees ==="
-if [ -f "$STATE_FILE" ] && jq -e '.' "$STATE_FILE" >/dev/null 2>&1; then
-  IN_USE=$(jq -r '.agents[] | select(.state != "done") | "  [\(.state)] \(.worktree_path) → \(.branch)"' \
-    "$STATE_FILE" 2>/dev/null || echo "")
-  if [ -n "$IN_USE" ]; then
-    echo "$IN_USE"
-  else
-    echo "  (none)"
-  fi
-else
-  echo "  (no active state file)"
-fi

.claude/skills/orchestrate/scripts/classify-pane.sh

@@ -1,85 +0,0 @@
-#!/usr/bin/env bash
-# classify-pane.sh — Classify the current state of a tmux pane
-#
-# Usage: classify-pane.sh <tmux-target>
-#   tmux-target: e.g. "work:0", "work:1.0"
-#
-# Output (stdout): JSON object:
-#   { "state": "running|idle|waiting_approval|complete", "reason": "...", "pane_cmd": "..." }
-#
-# Exit codes: 0=ok, 1=error (invalid target or tmux window not found)
-
-set -euo pipefail
-
-TARGET="${1:-}"
-
-if [ -z "$TARGET" ]; then
-  echo '{"state":"error","reason":"no target provided","pane_cmd":""}'
-  exit 1
-fi
-
-# Validate tmux target format: session:window or session:window.pane
-if ! [[ "$TARGET" =~ ^[a-zA-Z0-9_.-]+:[a-zA-Z0-9_.-]+(\.[0-9]+)?$ ]]; then
-  echo '{"state":"error","reason":"invalid tmux target format","pane_cmd":""}'
-  exit 1
-fi
-
-# Check session exists (use %%:* to extract session name from session:window)
-if ! tmux list-windows -t "${TARGET%%:*}" &>/dev/null 2>&1; then
-  echo '{"state":"error","reason":"tmux target not found","pane_cmd":""}'
-  exit 1
-fi
-
-# Get the current foreground command in the pane
-PANE_CMD=$(tmux display-message -t "$TARGET" -p '#{pane_current_command}' 2>/dev/null || echo "unknown")
-
-# Capture and strip ANSI codes (use perl for cross-platform compatibility — BSD sed lacks \x1b support)
-RAW=$(tmux capture-pane -t "$TARGET" -p -S -50 2>/dev/null || echo "")
-CLEAN=$(echo "$RAW" | perl -pe 's/\x1b\[[0-9;]*[a-zA-Z]//g; s/\x1b\(B//g; s/\x1b\[\?[0-9]*[hl]//g; s/\r//g' \
-  | grep -v '^[[:space:]]*$' || true)
-
-# --- Check: explicit completion marker ---
-# Must be on its own line (not buried in the objective text sent at spawn time).
-if echo "$CLEAN" | grep -qE "^[[:space:]]*ORCHESTRATOR:DONE[[:space:]]*$"; then
-  jq -n --arg cmd "$PANE_CMD" '{"state":"complete","reason":"ORCHESTRATOR:DONE marker found","pane_cmd":$cmd}'
-  exit 0
-fi
-
-# --- Check: Claude Code approval prompt patterns ---
-LAST_40=$(echo "$CLEAN" | tail -40)
-APPROVAL_PATTERNS=(
-  "Do you want to proceed"
-  "Do you want to make this"
-  "\\[y/n\\]"
-  "\\[Y/n\\]"
-  "\\[n/Y\\]"
-  "Proceed\\?"
-  "Allow this command"
-  "Run bash command"
-  "Allow bash"
-  "Would you like"
-  "Press enter to continue"
-  "Esc to cancel"
-)
-for pattern in "${APPROVAL_PATTERNS[@]}"; do
-  if echo "$LAST_40" | grep -qiE "$pattern"; then
-    jq -n --arg pattern "$pattern" --arg cmd "$PANE_CMD" \
-      '{"state":"waiting_approval","reason":"approval pattern: \($pattern)","pane_cmd":$cmd}'
-    exit 0
-  fi
-done
-
-# --- Check: shell prompt (claude has exited) ---
-# If the foreground process is a shell (not claude/node), the agent has exited
-case "$PANE_CMD" in
-  zsh|bash|fish|sh|dash|tcsh|ksh)
-    jq -n --arg cmd "$PANE_CMD" \
-      '{"state":"idle","reason":"agent exited — shell prompt active","pane_cmd":$cmd}'
-    exit 0
-    ;;
-esac
-
-# Agent is still running (claude/node/python is the foreground process)
-jq -n --arg cmd "$PANE_CMD" \
-  '{"state":"running","reason":"foreground process: \($cmd)","pane_cmd":$cmd}'
-exit 0

.claude/skills/orchestrate/scripts/find-spare.sh

@@ -1,24 +0,0 @@
-#!/usr/bin/env bash
-# find-spare.sh — list worktrees on spare/N branches (free to use)
-#
-# Usage: find-spare.sh [REPO_ROOT]
-#   REPO_ROOT defaults to the root worktree containing the current git repo.
-#
-# Output (stdout): one line per available worktree: "PATH BRANCH"
-#   e.g.: /Users/me/Code/AutoGPT3 spare/3
-
-set -euo pipefail
-
-REPO_ROOT="${1:-$(git rev-parse --show-toplevel 2>/dev/null || echo "")}"
-if [ -z "$REPO_ROOT" ]; then
-  echo "Error: not inside a git repo and no REPO_ROOT provided" >&2
-  exit 1
-fi
-
-git -C "$REPO_ROOT" worktree list --porcelain \
-  | awk '
-      /^worktree / { path = substr($0, 10) }
-      /^branch /   { branch = substr($0, 8); print path " " branch }
-    ' \
-  | { grep -E " refs/heads/spare/[0-9]+$" || true; } \
-  | sed 's|refs/heads/||'

.claude/skills/orchestrate/scripts/notify.sh

@@ -1,40 +0,0 @@
-#!/usr/bin/env bash
-# notify.sh — send a fleet notification message
-#
-# Delivery order (first available wins):
-#   1. Discord webhook — DISCORD_WEBHOOK_URL env var OR state file .discord_webhook
-#   2. macOS notification center — osascript (silent fail if unavailable)
-#   3. Stdout only
-#
-# Usage: notify.sh MESSAGE
-# Exit: always 0 (notification failure must not abort the caller)
-
-MESSAGE="${1:-}"
-[ -z "$MESSAGE" ] && exit 0
-
-STATE_FILE="${ORCHESTRATOR_STATE_FILE:-$HOME/.claude/orchestrator-state.json}"
-
-# --- Resolve Discord webhook ---
-WEBHOOK="${DISCORD_WEBHOOK_URL:-}"
-if [ -z "$WEBHOOK" ] && [ -f "$STATE_FILE" ]; then
-  WEBHOOK=$(jq -r '.discord_webhook // ""' "$STATE_FILE" 2>/dev/null || echo "")
-fi
-
-# --- Discord delivery ---
-if [ -n "$WEBHOOK" ]; then
-  PAYLOAD=$(jq -n --arg msg "$MESSAGE" '{"content": $msg}')
-  curl -s -X POST "$WEBHOOK" \
-    -H "Content-Type: application/json" \
-    -d "$PAYLOAD" > /dev/null 2>&1 || true
-fi
-
-# --- macOS notification center (silent if not macOS or osascript missing) ---
-if command -v osascript &>/dev/null 2>&1; then
-  # Escape single quotes for AppleScript
-  SAFE_MSG=$(echo "$MESSAGE" | sed "s/'/\\\\'/g")
-  osascript -e "display notification \"${SAFE_MSG}\" with title \"Orchestrator\"" 2>/dev/null || true
-fi
-
-# Always print to stdout so run-loop.sh logs it
-echo "$MESSAGE"
-exit 0

.claude/skills/orchestrate/scripts/poll-cycle.sh

@@ -1,257 +0,0 @@
-#!/usr/bin/env bash
-# poll-cycle.sh — Single orchestrator poll cycle
-#
-# Reads ~/.claude/orchestrator-state.json, classifies each agent, updates state,
-# and outputs a JSON array of actions for Claude to take.
-#
-# Usage: poll-cycle.sh
-# Output (stdout): JSON array of action objects
-#   [{ "window": "work:0", "action": "kick|approve|none", "state": "...",
-#      "worktree": "...", "objective": "...", "reason": "..." }]
-#
-# The state file is updated in-place (atomic write via .tmp).
-
-set -euo pipefail
-
-STATE_FILE="${ORCHESTRATOR_STATE_FILE:-$HOME/.claude/orchestrator-state.json}"
-SCRIPTS_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-CLASSIFY="$SCRIPTS_DIR/classify-pane.sh"
-
-# Cross-platform md5: always outputs just the hex digest
-md5_hash() {
-  if command -v md5sum &>/dev/null; then
-    md5sum | awk '{print $1}'
-  else
-    md5 | awk '{print $NF}'
-  fi
-}
-
-# Clean up temp file on any exit (avoids stale .tmp if jq write fails)
-trap 'rm -f "${STATE_FILE}.tmp"' EXIT
-
-# Ensure state file exists
-if [ ! -f "$STATE_FILE" ]; then
-  echo '{"active":false,"agents":[]}' > "$STATE_FILE"
-fi
-
-# Validate JSON upfront before any jq reads that run under set -e.
-# A truncated/corrupt file (e.g. from a SIGKILL mid-write) would otherwise
-# abort the script at the ACTIVE read below without emitting any JSON output.
-if ! jq -e '.' "$STATE_FILE" >/dev/null 2>&1; then
-  echo "State file parse error — check $STATE_FILE" >&2
-  echo "[]"
-  exit 0
-fi
-
-ACTIVE=$(jq -r '.active // false' "$STATE_FILE")
-if [ "$ACTIVE" != "true" ]; then
-  echo "[]"
-  exit 0
-fi
-
-NOW=$(date +%s)
-IDLE_THRESHOLD=$(jq -r '.idle_threshold_seconds // 300' "$STATE_FILE")
-
-ACTIONS="[]"
-UPDATED_AGENTS="[]"
-
-# Read agents as newline-delimited JSON objects.
-# jq exits non-zero when .agents[] has no matches on an empty array, which is valid —
-# so we suppress that exit code and separately validate the file is well-formed JSON.
-if ! AGENTS_JSON=$(jq -e -c '.agents // empty | .[]' "$STATE_FILE" 2>/dev/null); then
-  if ! jq -e '.' "$STATE_FILE" > /dev/null 2>&1; then
-    echo "State file parse error — check $STATE_FILE" >&2
-  fi
-  echo "[]"
-  exit 0
-fi
-
-if [ -z "$AGENTS_JSON" ]; then
-  echo "[]"
-  exit 0
-fi
-
-while IFS= read -r agent; do
-  [ -z "$agent" ] && continue
-
-  # Use // "" defaults so a single malformed field doesn't abort the whole cycle
-  WINDOW=$(echo "$agent"   | jq -r '.window // ""')
-  WORKTREE=$(echo "$agent" | jq -r '.worktree // ""')
-  OBJECTIVE=$(echo "$agent"| jq -r '.objective // ""')
-  STATE=$(echo "$agent"    | jq -r '.state // "running"')
-  LAST_HASH=$(echo "$agent"| jq -r '.last_output_hash // ""')
-  IDLE_SINCE=$(echo "$agent"| jq -r '.idle_since // 0')
-  REVISION_COUNT=$(echo "$agent"| jq -r '.revision_count // 0')
-
-  # Validate window format to prevent tmux target injection.
-  # Allow session:window (numeric or named) and session:window.pane
-  if ! [[ "$WINDOW" =~ ^[a-zA-Z0-9_.-]+:[a-zA-Z0-9_.-]+(\.[0-9]+)?$ ]]; then
-    echo "Skipping agent with invalid window value: $WINDOW" >&2
-    UPDATED_AGENTS=$(echo "$UPDATED_AGENTS" | jq --argjson a "$agent" '. + [$a]')
-    continue
-  fi
-
-  # Pass-through terminal-state agents
-  if [[ "$STATE" == "done" || "$STATE" == "escalated" || "$STATE" == "complete" || "$STATE" == "pending_evaluation" ]]; then
-    UPDATED_AGENTS=$(echo "$UPDATED_AGENTS" | jq --argjson a "$agent" '. + [$a]')
-    continue
-  fi
-
-  # Classify pane.
-  # classify-pane.sh always emits JSON before exit (even on error), so using
-  # "|| echo '...'" would concatenate two JSON objects when it exits non-zero.
-  # Use "|| true" inside the substitution so set -euo pipefail does not abort
-  # the poll cycle when classify exits with a non-zero status code.
-  CLASSIFICATION=$("$CLASSIFY" "$WINDOW" 2>/dev/null || true)
-  [ -z "$CLASSIFICATION" ] && CLASSIFICATION='{"state":"error","reason":"classify failed","pane_cmd":"unknown"}'
-
-  PANE_STATE=$(echo "$CLASSIFICATION" | jq -r '.state')
-  PANE_REASON=$(echo "$CLASSIFICATION" | jq -r '.reason')
-
-  # Capture full pane output once — used for hash (stuck detection) and checkpoint parsing.
-  # Use -S -500 to get the last ~500 lines of scrollback so checkpoints aren't missed.
-  RAW=$(tmux capture-pane -t "$WINDOW" -p -S -500 2>/dev/null || echo "")
-
-  # --- Checkpoint tracking ---
-  # Parse any "CHECKPOINT:<step>" lines the agent has output and merge into state file.
-  # The agent writes these as it completes each required step so verify-complete.sh can gate recycling.
-  EXISTING_CPS=$(echo "$agent" | jq -c '.checkpoints // []')
-  NEW_CHECKPOINTS_JSON="$EXISTING_CPS"
-  if [ -n "$RAW" ]; then
-    FOUND_CPS=$(echo "$RAW" \
-      | grep -oE "CHECKPOINT:[a-zA-Z0-9_-]+" \
-      | sed 's/CHECKPOINT://' \
-      | sort -u \
-      | jq -R . | jq -s . 2>/dev/null || echo "[]")
-    NEW_CHECKPOINTS_JSON=$(jq -n \
-      --argjson existing "$EXISTING_CPS" \
-      --argjson found "$FOUND_CPS" \
-      '($existing + $found) | unique' 2>/dev/null || echo "$EXISTING_CPS")
-  fi
-
-  # Compute content hash for stuck-detection (only for running agents)
-  CURRENT_HASH=""
-  if [[ "$PANE_STATE" == "running" ]] && [ -n "$RAW" ]; then
-    CURRENT_HASH=$(echo "$RAW" | tail -20 | md5_hash)
-  fi
-
-  NEW_STATE="$STATE"
-  NEW_IDLE_SINCE="$IDLE_SINCE"
-  NEW_REVISION_COUNT="$REVISION_COUNT"
-  ACTION="none"
-  REASON="$PANE_REASON"
-
-  case "$PANE_STATE" in
-    complete)
-      # Agent output ORCHESTRATOR:DONE — mark pending_evaluation so orchestrator handles it.
-      # run-loop does NOT verify or notify; orchestrator's background poll picks this up.
-      NEW_STATE="pending_evaluation"
-      ACTION="complete"  # run-loop logs it but takes no action
-      ;;
-    waiting_approval)
-      NEW_STATE="waiting_approval"
-      ACTION="approve"
-      ;;
-    idle)
-      # Agent process has exited — needs restart
-      NEW_STATE="idle"
-      ACTION="kick"
-      REASON="agent exited (shell is foreground)"
-      NEW_REVISION_COUNT=$(( REVISION_COUNT + 1 ))
-      NEW_IDLE_SINCE=$NOW
-      if [ "$NEW_REVISION_COUNT" -ge 3 ]; then
-        NEW_STATE="escalated"
-        ACTION="none"
-        REASON="escalated after ${NEW_REVISION_COUNT} kicks — needs human attention"
-      fi
-      ;;
-    running)
-      # Clear idle_since only when transitioning from idle (agent was kicked and
-      # restarted). Do NOT reset for stuck — idle_since must persist across polls
-      # so STUCK_DURATION can accumulate and trigger escalation.
-      # Also update the local IDLE_SINCE so the hash-stability check below uses
-      # the reset value on this same poll, not the stale kick timestamp.
-      if [[ "$STATE" == "idle" ]]; then
-        NEW_IDLE_SINCE=0
-        IDLE_SINCE=0
-      fi
-      # Check if hash has been stable (agent may be stuck mid-task)
-      if [ -n "$CURRENT_HASH" ] && [ "$CURRENT_HASH" = "$LAST_HASH" ] && [ "$LAST_HASH" != "" ]; then
-        if [ "$IDLE_SINCE" = "0" ] || [ "$IDLE_SINCE" = "null" ]; then
-          NEW_IDLE_SINCE=$NOW
-        else
-          STUCK_DURATION=$(( NOW - IDLE_SINCE ))
-          if [ "$STUCK_DURATION" -gt "$IDLE_THRESHOLD" ]; then
-            NEW_REVISION_COUNT=$(( REVISION_COUNT + 1 ))
-            NEW_IDLE_SINCE=$NOW
-            if [ "$NEW_REVISION_COUNT" -ge 3 ]; then
-              NEW_STATE="escalated"
-              ACTION="none"
-              REASON="escalated after ${NEW_REVISION_COUNT} kicks — needs human attention"
-            else
-              NEW_STATE="stuck"
-              ACTION="kick"
-              REASON="output unchanged for ${STUCK_DURATION}s (threshold: ${IDLE_THRESHOLD}s)"
-            fi
-          fi
-        fi
-      else
-        # Only reset the idle timer when we have a valid hash comparison (pane
-        # capture succeeded). If CURRENT_HASH is empty (tmux capture-pane failed),
-        # preserve existing timers so stuck detection is not inadvertently reset.
-        if [ -n "$CURRENT_HASH" ]; then
-          NEW_STATE="running"
-          NEW_IDLE_SINCE=0
-        fi
-      fi
-      ;;
-    error)
-      REASON="classify error: $PANE_REASON"
-      ;;
-  esac
-
-  # Build updated agent record (ensure idle_since and revision_count are numeric)
-  # Use || true on each jq call so a malformed field skips this agent rather than
-  # aborting the entire poll cycle under set -e.
-  UPDATED_AGENT=$(echo "$agent" | jq \
-    --arg state "$NEW_STATE" \
-    --arg hash "$CURRENT_HASH" \
-    --argjson now "$NOW" \
-    --arg idle_since "$NEW_IDLE_SINCE" \
-    --arg revision_count "$NEW_REVISION_COUNT" \
-    --argjson checkpoints "$NEW_CHECKPOINTS_JSON" \
-    '.state = $state
-     | .last_output_hash = (if $hash == "" then .last_output_hash else $hash end)
-     | .last_seen_at = $now
-     | .idle_since = ($idle_since | tonumber)
-     | .revision_count = ($revision_count | tonumber)
-     | .checkpoints = $checkpoints' 2>/dev/null) || {
-    echo "Warning: failed to build updated agent for window $WINDOW — keeping original" >&2
-    UPDATED_AGENTS=$(echo "$UPDATED_AGENTS" | jq --argjson a "$agent" '. + [$a]')
-    continue
-  }
-
-  UPDATED_AGENTS=$(echo "$UPDATED_AGENTS" | jq --argjson a "$UPDATED_AGENT" '. + [$a]')
-
-  # Add action if needed
-  if [ "$ACTION" != "none" ]; then
-    ACTION_OBJ=$(jq -n \
-      --arg window "$WINDOW" \
-      --arg action "$ACTION" \
-      --arg state "$NEW_STATE" \
-      --arg worktree "$WORKTREE" \
-      --arg objective "$OBJECTIVE" \
-      --arg reason "$REASON" \
-      '{window:$window, action:$action, state:$state, worktree:$worktree, objective:$objective, reason:$reason}')
-    ACTIONS=$(echo "$ACTIONS" | jq --argjson a "$ACTION_OBJ" '. + [$a]')
-  fi
-
-done <<< "$AGENTS_JSON"
-
-# Atomic state file update
-jq --argjson agents "$UPDATED_AGENTS" \
-   --argjson now "$NOW" \
-   '.agents = $agents | .last_poll_at = $now' \
-   "$STATE_FILE" > "${STATE_FILE}.tmp" && mv "${STATE_FILE}.tmp" "$STATE_FILE"
-
-echo "$ACTIONS"

.claude/skills/orchestrate/scripts/recycle-agent.sh

@@ -1,32 +0,0 @@
-#!/usr/bin/env bash
-# recycle-agent.sh — kill a tmux window and restore the worktree to its spare branch
-#
-# Usage: recycle-agent.sh WINDOW WORKTREE_PATH SPARE_BRANCH
-#   WINDOW        — tmux target, e.g. autogpt1:3
-#   WORKTREE_PATH — absolute path to the git worktree
-#   SPARE_BRANCH  — branch to restore, e.g. spare/6
-#
-# Stdout: one status line
-
-set -euo pipefail
-
-if [ $# -lt 3 ]; then
-  echo "Usage: recycle-agent.sh WINDOW WORKTREE_PATH SPARE_BRANCH" >&2
-  exit 1
-fi
-
-WINDOW="$1"
-WORKTREE_PATH="$2"
-SPARE_BRANCH="$3"
-
-# Kill the tmux window (ignore error — may already be gone)
-tmux kill-window -t "$WINDOW" 2>/dev/null || true
-
-# Restore to spare branch: abort any in-progress operation, then clean
-git -C "$WORKTREE_PATH" rebase --abort 2>/dev/null || true
-git -C "$WORKTREE_PATH" merge --abort 2>/dev/null || true
-git -C "$WORKTREE_PATH" reset --hard HEAD 2>/dev/null
-git -C "$WORKTREE_PATH" clean -fd 2>/dev/null
-git -C "$WORKTREE_PATH" checkout "$SPARE_BRANCH"
-
-echo "Recycled: $(basename "$WORKTREE_PATH") → $SPARE_BRANCH (window $WINDOW closed)"

.claude/skills/orchestrate/scripts/run-loop.sh

@@ -1,215 +0,0 @@
-#!/usr/bin/env bash
-# run-loop.sh — Mechanical babysitter for the agent fleet (runs in its own tmux window)
-#
-# Handles ONLY two things that need no intelligence:
-#   idle    → restart claude using --resume SESSION_ID (or --continue) to restore context
-#   approve → auto-approve safe dialogs, press Enter on numbered-option dialogs
-#
-# Everything else — ORCHESTRATOR:DONE, verification, /pr-test, final evaluation,
-# marking done, deciding to close windows — is the orchestrating Claude's job.
-# poll-cycle.sh sets state to pending_evaluation when ORCHESTRATOR:DONE is detected;
-# the orchestrator's background poll loop handles it from there.
-#
-# Usage: run-loop.sh
-# Env:   POLL_INTERVAL (default: 30), ORCHESTRATOR_STATE_FILE
-
-set -euo pipefail
-
-# Copy scripts to a stable location outside the repo so they survive branch
-# checkouts (e.g. recycle-agent.sh switching spare/N back into this worktree
-# would wipe .claude/skills/orchestrate/scripts if the skill only exists on the
-# current branch).
-_ORIGIN_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-STABLE_SCRIPTS_DIR="$HOME/.claude/orchestrator/scripts"
-mkdir -p "$STABLE_SCRIPTS_DIR"
-cp "$_ORIGIN_DIR"/*.sh "$STABLE_SCRIPTS_DIR/"
-chmod +x "$STABLE_SCRIPTS_DIR"/*.sh
-SCRIPTS_DIR="$STABLE_SCRIPTS_DIR"
-
-STATE_FILE="${ORCHESTRATOR_STATE_FILE:-$HOME/.claude/orchestrator-state.json}"
-# Adaptive polling: starts at base interval, backs off up to POLL_IDLE_MAX when
-# no agents need attention, resets on any activity or waiting_approval state.
-POLL_INTERVAL="${POLL_INTERVAL:-30}"
-POLL_IDLE_MAX=${POLL_IDLE_MAX:-300}
-POLL_CURRENT=$POLL_INTERVAL
-
-# ---------------------------------------------------------------------------
-# update_state WINDOW FIELD VALUE
-# ---------------------------------------------------------------------------
-update_state() {
-  local window="$1" field="$2" value="$3"
-  jq --arg w "$window" --arg f "$field" --arg v "$value" \
-    '.agents |= map(if .window == $w then .[$f] = $v else . end)' \
-    "$STATE_FILE" > "${STATE_FILE}.tmp" && mv "${STATE_FILE}.tmp" "$STATE_FILE"
-}
-
-update_state_int() {
-  local window="$1" field="$2" value="$3"
-  jq --arg w "$window" --arg f "$field" --argjson v "$value" \
-    '.agents |= map(if .window == $w then .[$f] = $v else . end)' \
-    "$STATE_FILE" > "${STATE_FILE}.tmp" && mv "${STATE_FILE}.tmp" "$STATE_FILE"
-}
-
-agent_field() {
-  jq -r --arg w "$1" --arg f "$2" \
-    '.agents[] | select(.window == $w) | .[$f] // ""' \
-    "$STATE_FILE" 2>/dev/null
-}
-
-# ---------------------------------------------------------------------------
-# wait_for_prompt WINDOW — wait up to 60s for Claude's ❯ prompt
-# ---------------------------------------------------------------------------
-wait_for_prompt() {
-  local window="$1"
-  for i in $(seq 1 60); do
-    local cmd pane
-    cmd=$(tmux display-message -t "$window" -p '#{pane_current_command}' 2>/dev/null || echo "")
-    pane=$(tmux capture-pane -t "$window" -p 2>/dev/null || echo "")
-    if echo "$pane" | grep -q "Enter to confirm"; then
-      tmux send-keys -t "$window" Down Enter; sleep 2; continue
-    fi
-    [[ "$cmd" == "node" ]] && echo "$pane" | grep -q "❯" && return 0
-    sleep 1
-  done
-  return 1  # timed out
-}
-
-# ---------------------------------------------------------------------------
-# wait_for_claude_idle WINDOW — wait up to 30s for Claude to reach idle ❯ prompt
-# (no spinner or busy indicator visible in the last 3 lines of pane output)
-# Returns 0 when idle, 1 on timeout.
-# ---------------------------------------------------------------------------
-wait_for_claude_idle() {
-  local window="$1"
-  local timeout="${2:-30}"
-  local elapsed=0
-  while (( elapsed < timeout )); do
-    local cmd pane pane_tail
-    cmd=$(tmux display-message -t "$window" -p '#{pane_current_command}' 2>/dev/null || echo "")
-    pane=$(tmux capture-pane -t "$window" -p 2>/dev/null || echo "")
-    pane_tail=$(echo "$pane" | tail -3)
-    # Check full pane (not just tail) — 'Enter to confirm' dialog can scroll above last 3 lines.
-    # Do NOT reset elapsed — resetting allows an infinite loop if the dialog never clears.
-    if echo "$pane" | grep -q "Enter to confirm"; then
-      tmux send-keys -t "$window" Down Enter
-      sleep 2; (( elapsed += 2 )); continue
-    fi
-    # Must be running under node (Claude is live)
-    if [[ "$cmd" == "node" ]]; then
-      # Idle: ❯ prompt visible AND no spinner/busy text in last 3 lines
-      if echo "$pane_tail" | grep -q "❯" && \
-         ! echo "$pane_tail" | grep -qE '[✳✽✢✶·✻✼✿❋✤]|Running…|Compacting'; then
-        return 0
-      fi
-    fi
-    sleep 2
-    (( elapsed += 2 ))
-  done
-  return 1  # timed out
-}
-
-# ---------------------------------------------------------------------------
-# handle_kick WINDOW STATE — only for idle (crashed) agents, not stuck
-# ---------------------------------------------------------------------------
-handle_kick() {
-  local window="$1" state="$2"
-  [[ "$state" != "idle" ]] && return  # stuck agents handled by supervisor
-
-  local worktree_path session_id
-  worktree_path=$(agent_field "$window" "worktree_path")
-  session_id=$(agent_field "$window" "session_id")
-
-  echo "[$(date +%H:%M:%S)] KICK restart  $window — agent exited, resuming session"
-
-  # Wait for the shell prompt before typing — avoids sending into a still-draining pane
-  wait_for_claude_idle "$window" 30 \
-    || echo "[$(date +%H:%M:%S)] KICK WARNING  $window — pane still busy before resume, sending anyway"
-
-  # Resume the exact session so the agent retains full context — no need to re-send objective
-  if [ -n "$session_id" ]; then
-    tmux send-keys -t "$window" "cd '${worktree_path}' && claude --resume '${session_id}' --permission-mode bypassPermissions" Enter
-  else
-    tmux send-keys -t "$window" "cd '${worktree_path}' && claude --continue --permission-mode bypassPermissions" Enter
-  fi
-
-  wait_for_prompt "$window" || echo "[$(date +%H:%M:%S)] KICK WARNING  $window — timed out waiting for ❯"
-}
-
-# ---------------------------------------------------------------------------
-# handle_approve WINDOW — auto-approve dialogs that need no judgment
-# ---------------------------------------------------------------------------
-handle_approve() {
-  local window="$1"
-  local pane_tail
-  pane_tail=$(tmux capture-pane -t "$window" -p 2>/dev/null | tail -3 || echo "")
-
-  # Settings error dialog at startup
-  if echo "$pane_tail" | grep -q "Enter to confirm"; then
-    echo "[$(date +%H:%M:%S)] APPROVE dialog $window — settings error"
-    tmux send-keys -t "$window" Down Enter
-    return
-  fi
-
-  # Numbered-option dialog (e.g. "Do you want to make this edit?")
-  # ❯ is already on option 1 (Yes) — Enter confirms it
-  if echo "$pane_tail" | grep -qE "❯\s*1\." || echo "$pane_tail" | grep -q "Esc to cancel"; then
-    echo "[$(date +%H:%M:%S)] APPROVE edit   $window"
-    tmux send-keys -t "$window" "" Enter
-    return
-  fi
-
-  # y/n prompt for safe operations
-  if echo "$pane_tail" | grep -qiE "(^git |^npm |^pnpm |^poetry |^pytest|^docker |^make |^cargo |^pip |^yarn |curl .*(localhost|127\.0\.0\.1))"; then
-    echo "[$(date +%H:%M:%S)] APPROVE safe   $window"
-    tmux send-keys -t "$window" "y" Enter
-    return
-  fi
-
-  # Anything else — supervisor handles it, just log
-  echo "[$(date +%H:%M:%S)] APPROVE skip   $window — unknown dialog, supervisor will handle"
-}
-
-# ---------------------------------------------------------------------------
-# Main loop
-# ---------------------------------------------------------------------------
-echo "[$(date +%H:%M:%S)] run-loop started (mechanical only, poll ${POLL_INTERVAL}s→${POLL_IDLE_MAX}s adaptive)"
-echo "[$(date +%H:%M:%S)] Supervisor: orchestrating Claude session (not a separate window)"
-echo "---"
-
-while true; do
-  if ! jq -e '.active == true' "$STATE_FILE" >/dev/null 2>&1; then
-    echo "[$(date +%H:%M:%S)] active=false — exiting."
-    exit 0
-  fi
-
-  ACTIONS=$("$SCRIPTS_DIR/poll-cycle.sh" 2>/dev/null || echo "[]")
-  KICKED=0; DONE=0
-
-  while IFS= read -r action; do
-    [ -z "$action" ] && continue
-    WINDOW=$(echo "$action" | jq -r '.window // ""')
-    ACTION=$(echo "$action" | jq -r '.action // ""')
-    STATE=$(echo "$action"  | jq -r '.state // ""')
-
-    case "$ACTION" in
-      kick)     handle_kick "$WINDOW" "$STATE" || true; KICKED=$(( KICKED + 1 )) ;;
-      approve)  handle_approve "$WINDOW" || true ;;
-      complete) DONE=$(( DONE + 1 )) ;;  # poll-cycle already set state=pending_evaluation; orchestrator handles
-    esac
-  done < <(echo "$ACTIONS" | jq -c '.[]' 2>/dev/null || true)
-
-  RUNNING=$(jq '[.agents[] | select(.state | test("running|stuck|waiting_approval|idle"))] | length' \
-    "$STATE_FILE" 2>/dev/null || echo 0)
-
-  # Adaptive backoff: reset to base on activity or waiting_approval agents; back off when truly idle
-  WAITING=$(jq '[.agents[] | select(.state == "waiting_approval")] | length' "$STATE_FILE" 2>/dev/null || echo 0)
-  if (( KICKED > 0 || DONE > 0 || WAITING > 0 )); then
-    POLL_CURRENT=$POLL_INTERVAL
-  else
-    POLL_CURRENT=$(( POLL_CURRENT + POLL_CURRENT / 2 + 1 ))
-    (( POLL_CURRENT > POLL_IDLE_MAX )) && POLL_CURRENT=$POLL_IDLE_MAX
-  fi
-
-  echo "[$(date +%H:%M:%S)] Poll — ${RUNNING} running  ${KICKED} kicked  ${DONE} recycled  (next in ${POLL_CURRENT}s)"
-  sleep "$POLL_CURRENT"
-done

.claude/skills/orchestrate/scripts/spawn-agent.sh

@@ -1,129 +0,0 @@
-#!/usr/bin/env bash
-# spawn-agent.sh — create tmux window, checkout branch, launch claude, send task
-#
-# Usage: spawn-agent.sh SESSION WORKTREE_PATH SPARE_BRANCH NEW_BRANCH OBJECTIVE [PR_NUMBER] [STEPS...]
-#   SESSION       — tmux session name, e.g. autogpt1
-#   WORKTREE_PATH — absolute path to the git worktree
-#   SPARE_BRANCH  — spare branch being replaced, e.g. spare/6 (saved for recycle)
-#   NEW_BRANCH    — task branch to create, e.g. feat/my-feature
-#   OBJECTIVE     — task description sent to the agent
-#   PR_NUMBER     — (optional) GitHub PR number for completion verification
-#   STEPS...      — (optional) required checkpoint names, e.g. pr-address pr-test
-#
-# Stdout: SESSION:WINDOW_INDEX (nothing else — callers rely on this)
-# Exit non-zero on failure.
-
-set -euo pipefail
-
-if [ $# -lt 5 ]; then
-  echo "Usage: spawn-agent.sh SESSION WORKTREE_PATH SPARE_BRANCH NEW_BRANCH OBJECTIVE [PR_NUMBER] [STEPS...]" >&2
-  exit 1
-fi
-
-SESSION="$1"
-WORKTREE_PATH="$2"
-SPARE_BRANCH="$3"
-NEW_BRANCH="$4"
-OBJECTIVE="$5"
-PR_NUMBER="${6:-}"
-STEPS=("${@:7}")
-WORKTREE_NAME=$(basename "$WORKTREE_PATH")
-STATE_FILE="${ORCHESTRATOR_STATE_FILE:-$HOME/.claude/orchestrator-state.json}"
-
-# Generate a stable session ID so this agent's Claude session can always be resumed:
-#   claude --resume $SESSION_ID --permission-mode bypassPermissions
-SESSION_ID=$(uuidgen 2>/dev/null || python3 -c "import uuid; print(uuid.uuid4())")
-
-# Create (or switch to) the task branch
-git -C "$WORKTREE_PATH" checkout -b "$NEW_BRANCH" 2>/dev/null \
-  || git -C "$WORKTREE_PATH" checkout "$NEW_BRANCH"
-
-# Open a new named tmux window; capture its numeric index
-WIN_IDX=$(tmux new-window -t "$SESSION" -n "$WORKTREE_NAME" -P -F '#{window_index}')
-WINDOW="${SESSION}:${WIN_IDX}"
-
-# Append the initial agent record to the state file so subsequent jq updates find it.
-# This must happen before the pr_number/steps update below.
-if [ -f "$STATE_FILE" ]; then
-  NOW=$(date +%s)
-  jq --arg window "$WINDOW" \
-     --arg worktree "$WORKTREE_NAME" \
-     --arg worktree_path "$WORKTREE_PATH" \
-     --arg spare_branch "$SPARE_BRANCH" \
-     --arg branch "$NEW_BRANCH" \
-     --arg objective "$OBJECTIVE" \
-     --arg session_id "$SESSION_ID" \
-     --argjson now "$NOW" \
-     '.agents += [{
-       "window": $window,
-       "worktree": $worktree,
-       "worktree_path": $worktree_path,
-       "spare_branch": $spare_branch,
-       "branch": $branch,
-       "objective": $objective,
-       "session_id": $session_id,
-       "state": "running",
-       "checkpoints": [],
-       "last_output_hash": "",
-       "last_seen_at": $now,
-       "spawned_at": $now,
-       "idle_since": 0,
-       "revision_count": 0,
-       "last_rebriefed_at": 0
-     }]' "$STATE_FILE" > "${STATE_FILE}.tmp" && mv "${STATE_FILE}.tmp" "$STATE_FILE"
-fi
-
-# Store pr_number + steps in state file if provided (enables verify-complete.sh).
-# The agent record was appended above so the jq select now finds it.
-if [ -n "$PR_NUMBER" ] && [ -f "$STATE_FILE" ]; then
-  if [ "${#STEPS[@]}" -gt 0 ]; then
-    STEPS_JSON=$(printf '%s\n' "${STEPS[@]}" | jq -R . | jq -s .)
-  else
-    STEPS_JSON='[]'
-  fi
-  jq --arg w "$WINDOW" --arg pr "$PR_NUMBER" --argjson steps "$STEPS_JSON" \
-    '.agents |= map(if .window == $w then . + {pr_number: $pr, steps: $steps, checkpoints: []} else . end)' \
-    "$STATE_FILE" > "${STATE_FILE}.tmp" && mv "${STATE_FILE}.tmp" "$STATE_FILE"
-fi
-
-# Launch claude with a stable session ID so it can always be resumed after a crash:
-#   claude --resume SESSION_ID --permission-mode bypassPermissions
-tmux send-keys -t "$WINDOW" "cd '${WORKTREE_PATH}' && claude --permission-mode bypassPermissions --session-id '${SESSION_ID}'" Enter
-
-# wait_for_claude_idle — poll until the pane shows idle ❯ with no spinner in the last 3 lines.
-# Returns 0 when idle, 1 on timeout.
-_wait_idle() {
-  local window="$1" timeout="${2:-60}" elapsed=0
-  while (( elapsed < timeout )); do
-    local cmd pane_tail
-    cmd=$(tmux display-message -t "$window" -p '#{pane_current_command}' 2>/dev/null || echo "")
-    pane=$(tmux capture-pane -t "$window" -p 2>/dev/null || echo "")
-    pane_tail=$(echo "$pane" | tail -3)
-    # Check full pane (not just tail) — 'Enter to confirm' dialog can appear above the last 3 lines
-    if echo "$pane" | grep -q "Enter to confirm"; then
-      tmux send-keys -t "$window" Down Enter
-      sleep 2; (( elapsed += 2 )); continue
-    fi
-    if [[ "$cmd" == "node" ]] && \
-       echo "$pane_tail" | grep -q "❯" && \
-       ! echo "$pane_tail" | grep -qE '[✳✽✢✶·✻✼✿❋✤]|Running…|Compacting'; then
-      return 0
-    fi
-    sleep 2; (( elapsed += 2 ))
-  done
-  return 1
-}
-
-# Wait up to 60s for claude to be fully interactive and idle (❯ visible, no spinner).
-if ! _wait_idle "$WINDOW" 60; then
-  echo "[spawn-agent] WARNING: timed out waiting for idle ❯ prompt on $WINDOW — sending objective anyway" >&2
-fi
-
-# Send the task. Split text and Enter — if combined, Enter can fire before the string
-# is fully buffered, leaving the message stuck as "[Pasted text +N lines]" unsent.
-tmux send-keys -t "$WINDOW" "${OBJECTIVE} Output each completed step as CHECKPOINT:<step-name>. When ALL steps are done, output ORCHESTRATOR:DONE on its own line."
-sleep 0.3
-tmux send-keys -t "$WINDOW" Enter
-
-# Only output the window address — nothing else (callers parse this)
-echo "$WINDOW"

.claude/skills/orchestrate/scripts/status.sh

@@ -1,43 +0,0 @@
-#!/usr/bin/env bash
-# status.sh — print orchestrator status: state file summary + live tmux pane commands
-#
-# Usage: status.sh
-# Reads: ~/.claude/orchestrator-state.json
-
-set -euo pipefail
-
-STATE_FILE="${ORCHESTRATOR_STATE_FILE:-$HOME/.claude/orchestrator-state.json}"
-
-if [ ! -f "$STATE_FILE" ] || ! jq -e '.' "$STATE_FILE" >/dev/null 2>&1; then
-  echo "No orchestrator state found at $STATE_FILE"
-  exit 0
-fi
-
-# Header: active status, session, thresholds, last poll
-jq -r '
-  "=== Orchestrator [\(if .active then "RUNNING" else "STOPPED" end)] ===",
-  "Session: \(.tmux_session // "unknown")  |  Idle threshold: \(.idle_threshold_seconds // 300)s",
-  "Last poll: \(if (.last_poll_at // 0) == 0 then "never" else (.last_poll_at | strftime("%H:%M:%S")) end)",
-  ""
-' "$STATE_FILE"
-
-# Each agent: state, window, worktree/branch, truncated objective
-AGENT_COUNT=$(jq '.agents | length' "$STATE_FILE")
-if [ "$AGENT_COUNT" -eq 0 ]; then
-  echo "  (no agents registered)"
-else
-  jq -r '
-    .agents[] |
-    "  [\(.state | ascii_upcase)] \(.window)  \(.worktree)/\(.branch)",
-    "    \(.objective // "" | .[0:70])"
-  ' "$STATE_FILE"
-fi
-
-echo ""
-
-# Live pane_current_command for non-done agents
-while IFS= read -r WINDOW; do
-  [ -z "$WINDOW" ] && continue
-  CMD=$(tmux display-message -t "$WINDOW" -p '#{pane_current_command}' 2>/dev/null || echo "unreachable")
-  echo "  $WINDOW live: $CMD"
-done < <(jq -r '.agents[] | select(.state != "done") | .window' "$STATE_FILE" 2>/dev/null || true)

.claude/skills/orchestrate/scripts/verify-complete.sh

@@ -1,180 +0,0 @@
-#!/usr/bin/env bash
-# verify-complete.sh — verify a PR task is truly done before marking the agent done
-#
-# Check order matters:
-#   1. Checkpoints — did the agent do all required steps?
-#   2. CI complete — no pending (bots post comments AFTER their check runs, must wait)
-#   3. CI passing — no failures (agent must fix before done)
-#   4. spawned_at — a new CI run was triggered after agent spawned (proves real work)
-#   5. Unresolved threads — checked AFTER CI so bot-posted comments are included
-#   6. CHANGES_REQUESTED — checked AFTER CI so bot reviews are included
-#
-# Usage: verify-complete.sh WINDOW
-# Exit 0 = verified complete; exit 1 = not complete (stderr has reason)
-
-set -euo pipefail
-
-WINDOW="$1"
-STATE_FILE="${ORCHESTRATOR_STATE_FILE:-$HOME/.claude/orchestrator-state.json}"
-
-PR_NUMBER=$(jq -r --arg w "$WINDOW" '.agents[] | select(.window == $w) | .pr_number // ""' "$STATE_FILE" 2>/dev/null)
-STEPS=$(jq -r --arg w "$WINDOW" '.agents[] | select(.window == $w) | .steps // [] | .[]' "$STATE_FILE" 2>/dev/null || true)
-CHECKPOINTS=$(jq -r --arg w "$WINDOW" '.agents[] | select(.window == $w) | .checkpoints // [] | .[]' "$STATE_FILE" 2>/dev/null || true)
-WORKTREE_PATH=$(jq -r --arg w "$WINDOW" '.agents[] | select(.window == $w) | .worktree_path // ""' "$STATE_FILE" 2>/dev/null)
-BRANCH=$(jq -r --arg w "$WINDOW" '.agents[] | select(.window == $w) | .branch // ""' "$STATE_FILE" 2>/dev/null)
-SPAWNED_AT=$(jq -r --arg w "$WINDOW" '.agents[] | select(.window == $w) | .spawned_at // "0"' "$STATE_FILE" 2>/dev/null || echo "0")
-
-# No PR number = cannot verify
-if [ -z "$PR_NUMBER" ]; then
-  echo "NOT COMPLETE: no pr_number in state — set pr_number or mark done manually" >&2
-  exit 1
-fi
-
-# --- Check 1: all required steps are checkpointed ---
-MISSING=""
-while IFS= read -r step; do
-  [ -z "$step" ] && continue
-  if ! echo "$CHECKPOINTS" | grep -qFx "$step"; then
-    MISSING="$MISSING $step"
-  fi
-done <<< "$STEPS"
-
-if [ -n "$MISSING" ]; then
-  echo "NOT COMPLETE: missing checkpoints:$MISSING on PR #$PR_NUMBER" >&2
-  exit 1
-fi
-
-# Resolve repo for all GitHub checks below
-REPO=$(jq -r '.repo // ""' "$STATE_FILE" 2>/dev/null || echo "")
-if [ -z "$REPO" ] && [ -n "$WORKTREE_PATH" ] && [ -d "$WORKTREE_PATH" ]; then
-  REPO=$(git -C "$WORKTREE_PATH" remote get-url origin 2>/dev/null \
-    | sed 's|.*github\.com[:/]||; s|\.git$||' || echo "")
-fi
-
-if [ -z "$REPO" ]; then
-  echo "Warning: cannot resolve repo — skipping CI/thread checks" >&2
-  echo "VERIFIED: PR #$PR_NUMBER — checkpoints ✓ (CI/thread checks skipped — no repo)"
-  exit 0
-fi
-
-CI_BUCKETS=$(gh pr checks "$PR_NUMBER" --repo "$REPO" --json bucket 2>/dev/null || echo "[]")
-
-# --- Check 2: CI fully complete — no pending checks ---
-# Pending checks MUST finish before we check threads/reviews:
-# bots (Seer, Check PR Status, etc.) post comments and CHANGES_REQUESTED AFTER their CI check runs.
-PENDING=$(echo "$CI_BUCKETS" | jq '[.[] | select(.bucket == "pending")] | length' 2>/dev/null || echo "0")
-if [ "$PENDING" -gt 0 ]; then
-  PENDING_NAMES=$(gh pr checks "$PR_NUMBER" --repo "$REPO" --json bucket,name 2>/dev/null \
-    | jq -r '[.[] | select(.bucket == "pending") | .name] | join(", ")' 2>/dev/null || echo "unknown")
-  echo "NOT COMPLETE: $PENDING CI checks still pending on PR #$PR_NUMBER ($PENDING_NAMES)" >&2
-  exit 1
-fi
-
-# --- Check 3: CI passing — no failures ---
-FAILING=$(echo "$CI_BUCKETS" | jq '[.[] | select(.bucket == "fail")] | length' 2>/dev/null || echo "0")
-if [ "$FAILING" -gt 0 ]; then
-  FAILING_NAMES=$(gh pr checks "$PR_NUMBER" --repo "$REPO" --json bucket,name 2>/dev/null \
-    | jq -r '[.[] | select(.bucket == "fail") | .name] | join(", ")' 2>/dev/null || echo "unknown")
-  echo "NOT COMPLETE: $FAILING failing CI checks on PR #$PR_NUMBER ($FAILING_NAMES)" >&2
-  exit 1
-fi
-
-# --- Check 4: a new CI run was triggered AFTER the agent spawned ---
-if [ -n "$BRANCH" ] && [ "${SPAWNED_AT:-0}" -gt 0 ]; then
-  LATEST_RUN_AT=$(gh run list --repo "$REPO" --branch "$BRANCH" \
-    --json createdAt --limit 1 2>/dev/null | jq -r '.[0].createdAt // ""')
-  if [ -n "$LATEST_RUN_AT" ]; then
-    if date --version >/dev/null 2>&1; then
-      LATEST_RUN_EPOCH=$(date -d "$LATEST_RUN_AT" "+%s" 2>/dev/null || echo "0")
-    else
-      LATEST_RUN_EPOCH=$(TZ=UTC date -j -f "%Y-%m-%dT%H:%M:%SZ" "$LATEST_RUN_AT" "+%s" 2>/dev/null || echo "0")
-    fi
-    if [ "$LATEST_RUN_EPOCH" -le "$SPAWNED_AT" ]; then
-      echo "NOT COMPLETE: latest CI run on $BRANCH predates agent spawn — agent may not have pushed yet" >&2
-      exit 1
-    fi
-  fi
-fi
-
-OWNER=$(echo "$REPO" | cut -d/ -f1)
-REPONAME=$(echo "$REPO" | cut -d/ -f2)
-
-# --- Check 5: no unresolved review threads (checked AFTER CI — bots post after their check) ---
-UNRESOLVED=$(gh api graphql -f query="
-  { repository(owner: \"${OWNER}\", name: \"${REPONAME}\") {
-      pullRequest(number: ${PR_NUMBER}) {
-        reviewThreads(first: 50) { nodes { isResolved } }
-      }
-    }
-  }
-" --jq '[.data.repository.pullRequest.reviewThreads.nodes[] | select(.isResolved == false)] | length' 2>/dev/null || echo "0")
-
-if [ "$UNRESOLVED" -gt 0 ]; then
-  echo "NOT COMPLETE: $UNRESOLVED unresolved review threads on PR #$PR_NUMBER" >&2
-  exit 1
-fi
-
-# --- Check 6: no CHANGES_REQUESTED (checked AFTER CI — bots post reviews after their check) ---
-# A CHANGES_REQUESTED review is stale if the latest commit was pushed AFTER the review was submitted.
-# Stale reviews (pre-dating the fixing commits) should not block verification.
-#
-# Fetch commits and latestReviews in a single call and fail closed — if gh fails,
-# treat that as NOT COMPLETE rather than silently passing.
-# Use latestReviews (not reviews) so each reviewer's latest state is used — superseded
-# CHANGES_REQUESTED entries are automatically excluded when the reviewer later approved.
-# Note: we intentionally use committedDate (not PR updatedAt) because updatedAt changes on any
-# PR activity (bot comments, label changes) which would create false negatives.
-PR_REVIEW_METADATA=$(gh pr view "$PR_NUMBER" --repo "$REPO" \
-  --json commits,latestReviews 2>/dev/null) || {
-  echo "NOT COMPLETE: unable to fetch PR review metadata for PR #$PR_NUMBER" >&2
-  exit 1
-}
-
-LATEST_COMMIT_DATE=$(jq -r '.commits[-1].committedDate // ""' <<< "$PR_REVIEW_METADATA")
-CHANGES_REQUESTED_REVIEWS=$(jq '[.latestReviews[]? | select(.state == "CHANGES_REQUESTED")]' <<< "$PR_REVIEW_METADATA")
-
-BLOCKING_CHANGES_REQUESTED=0
-BLOCKING_REQUESTERS=""
-
-if [ -n "$LATEST_COMMIT_DATE" ] && [ "$(echo "$CHANGES_REQUESTED_REVIEWS" | jq length)" -gt 0 ]; then
-  if date --version >/dev/null 2>&1; then
-    LATEST_COMMIT_EPOCH=$(date -d "$LATEST_COMMIT_DATE" "+%s" 2>/dev/null || echo "0")
-  else
-    LATEST_COMMIT_EPOCH=$(TZ=UTC date -j -f "%Y-%m-%dT%H:%M:%SZ" "$LATEST_COMMIT_DATE" "+%s" 2>/dev/null || echo "0")
-  fi
-
-  while IFS= read -r review; do
-    [ -z "$review" ] && continue
-    REVIEW_DATE=$(echo "$review" | jq -r '.submittedAt // ""')
-    REVIEWER=$(echo "$review" | jq -r '.author.login // "unknown"')
-    if [ -z "$REVIEW_DATE" ]; then
-      # No submission date — treat as fresh (conservative: blocks verification)
-      BLOCKING_CHANGES_REQUESTED=$(( BLOCKING_CHANGES_REQUESTED + 1 ))
-      BLOCKING_REQUESTERS="${BLOCKING_REQUESTERS:+$BLOCKING_REQUESTERS, }${REVIEWER}"
-    else
-      if date --version >/dev/null 2>&1; then
-        REVIEW_EPOCH=$(date -d "$REVIEW_DATE" "+%s" 2>/dev/null || echo "0")
-      else
-        REVIEW_EPOCH=$(TZ=UTC date -j -f "%Y-%m-%dT%H:%M:%SZ" "$REVIEW_DATE" "+%s" 2>/dev/null || echo "0")
-      fi
-      if [ "$REVIEW_EPOCH" -gt "$LATEST_COMMIT_EPOCH" ]; then
-        # Review was submitted AFTER latest commit — still fresh, blocks verification
-        BLOCKING_CHANGES_REQUESTED=$(( BLOCKING_CHANGES_REQUESTED + 1 ))
-        BLOCKING_REQUESTERS="${BLOCKING_REQUESTERS:+$BLOCKING_REQUESTERS, }${REVIEWER}"
-      fi
-      # Review submitted BEFORE latest commit — stale, skip
-    fi
-  done <<< "$(echo "$CHANGES_REQUESTED_REVIEWS" | jq -c '.[]')"
-else
-  # No commit date or no changes_requested — check raw count as fallback
-  BLOCKING_CHANGES_REQUESTED=$(echo "$CHANGES_REQUESTED_REVIEWS" | jq length 2>/dev/null || echo "0")
-  BLOCKING_REQUESTERS=$(echo "$CHANGES_REQUESTED_REVIEWS" | jq -r '[.[].author.login] | join(", ")' 2>/dev/null || echo "unknown")
-fi
-
-if [ "$BLOCKING_CHANGES_REQUESTED" -gt 0 ]; then
-  echo "NOT COMPLETE: CHANGES_REQUESTED (after latest commit) from ${BLOCKING_REQUESTERS} on PR #$PR_NUMBER" >&2
-  exit 1
-fi
-
-echo "VERIFIED: PR #$PR_NUMBER — checkpoints ✓, CI complete + green, 0 unresolved threads, no CHANGES_REQUESTED"
-exit 0

.claude/skills/pr-address/SKILL.md

@@ -1,495 +0,0 @@
----
-name: pr-address
-description: Address PR review comments and loop until CI green and all comments resolved. TRIGGER when user asks to address comments, fix PR feedback, respond to reviewers, or babysit/monitor a PR.
-user-invocable: true
-argument-hint: "[PR number or URL] — if omitted, finds PR for current branch."
-metadata:
-  author: autogpt-team
-  version: "1.0.0"
----
-
-# PR Address
-
-## Find the PR
-
-```bash
-gh pr list --head $(git branch --show-current) --repo Significant-Gravitas/AutoGPT
-gh pr view {N}
-```
-
-## Read the PR description
-
-Understand the **Why / What / How** before addressing comments — you need context to make good fixes:
-
-```bash
-gh pr view {N} --json body --jq '.body'
-```
-
-> If GraphQL is rate-limited, `gh pr view` fails. See [GitHub rate limits](#github-rate-limits) for REST fallbacks.
-
-## Fetch comments (all sources)
-
-### 1. Inline review threads — GraphQL (primary source of actionable items)
-
-> ⚠️ **WARNING — PAGINATE ALL PAGES BEFORE ADDRESSING ANYTHING**
->
-> `reviewThreads(first: 100)` returns at most 100 threads per page AND returns threads **oldest-first**. On a PR with many review cycles (e.g. 373 threads), the oldest 100–200 threads are from past cycles and are **all already resolved**. Filtering client-side with `select(.isResolved == false)` on page 1 therefore yields **0 results** — even though pages 2–4 contain many unresolved threads from recent review cycles.
->
-> **This is the most common failure mode:** agent fetches page 1, sees 0 unresolved after filtering, stops pagination, reports "done" — while hundreds of unresolved threads sit on later pages.
->
-> One observed PR had 142 total threads: page 1 returned 0 unresolved (all old/resolved), while pages 2–3 had 111 unresolved. Another with 373 threads across 4 pages also had page 1 entirely resolved.
->
-> **The rule: ALWAYS paginate to `hasNextPage == false` regardless of the per-page unresolved count. Never stop early because a page returns 0 unresolved.**
-
-**Step 1 — Fetch total count and sanity-check the newest threads:**
-
-```bash
-# Get total count and the newest 100 threads (last: 100 returns newest-first)
-gh api graphql -f query='
-{
-  repository(owner: "Significant-Gravitas", name: "AutoGPT") {
-    pullRequest(number: {N}) {
-      reviewThreads { totalCount }
-      newest: reviewThreads(last: 100) {
-        nodes { isResolved }
-      }
-    }
-  }
-}' | jq '{ total: .data.repository.pullRequest.reviewThreads.totalCount, newest_unresolved: [.data.repository.pullRequest.newest.nodes[] | select(.isResolved == false)] | length }'
-```
-
-If `total > 100`, you have multiple pages — you **must** paginate all of them regardless of what `newest_unresolved` shows. The `last: 100` check is a sanity signal only; the full loop below is mandatory.
-
-**Step 2 — Collect all unresolved thread IDs across all pages:**
-
-```bash
-# Accumulate all unresolved threads — loop until hasNextPage == false
-CURSOR=""
-ALL_THREADS="[]"
-while true; do
-  AFTER=${CURSOR:+", after: \"$CURSOR\""}
-  PAGE=$(gh api graphql -f query="
-  {
-    repository(owner: \"Significant-Gravitas\", name: \"AutoGPT\") {
-      pullRequest(number: {N}) {
-        reviewThreads(first: 100${AFTER}) {
-          pageInfo { hasNextPage endCursor }
-          nodes {
-            id
-            isResolved
-            path
-            line
-            comments(last: 1) {
-              nodes { databaseId body author { login } }
-            }
-          }
-        }
-      }
-    }
-  }")
-  # Append unresolved nodes from this page
-  PAGE_THREADS=$(echo "$PAGE" | jq '[.data.repository.pullRequest.reviewThreads.nodes[] | select(.isResolved == false)]')
-  ALL_THREADS=$(echo "$ALL_THREADS $PAGE_THREADS" | jq -s 'add')
-  HAS_NEXT=$(echo "$PAGE" | jq -r '.data.repository.pullRequest.reviewThreads.pageInfo.hasNextPage')
-  CURSOR=$(echo "$PAGE" | jq -r '.data.repository.pullRequest.reviewThreads.pageInfo.endCursor')
-  [ "$HAS_NEXT" = "false" ] && break
-done
-
-# Reverse so newest threads (last pages) are addressed first — GitHub returns oldest-first
-# and the most recent review cycle's comments are the ones blocking approval.
-ALL_THREADS=$(echo "$ALL_THREADS" | jq 'reverse')
-
-echo "Total unresolved threads: $(echo "$ALL_THREADS" | jq 'length')"
-echo "$ALL_THREADS" | jq '[.[] | {id, path, line, body: .comments.nodes[0].body[:200]}]'
-```
-
-**Step 3 — Address every thread in `ALL_THREADS`, then resolve.**
-
-Only after this loop completes (all pages fetched, count confirmed) should you begin making fixes.
-
-> **Why reverse?** GraphQL returns threads oldest-first and exposes no `orderBy` option. A PR with 373 threads has ~4 pages; threads from the latest review cycle land on the last pages. Processing in reverse ensures the newest, most blocking comments are addressed first — the earlier pages mostly contain outdated threads from prior cycles.
-
-**Filter to unresolved threads only** — skip any thread where `isResolved: true`. `comments(last: 1)` returns the most recent comment in the thread — act on that; it reflects the reviewer's final ask. Use the thread `id` (Relay global ID) to track threads across polls.
-
-> If GraphQL is rate-limited, see [GitHub rate limits](#github-rate-limits) for the REST fallback (flat comment list — no thread grouping or `isResolved`).
-
-### 2. Top-level reviews — REST (MUST paginate)
-
-```bash
-gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/reviews --paginate
-```
-
-> **Already REST — unaffected by GraphQL rate limits or outages. Continue polling reviews normally even when GraphQL is exhausted.**
-
-**CRITICAL — always `--paginate`.** Reviews default to 30 per page. PRs can have 80–170+ reviews (mostly empty resolution events). Without pagination you miss reviews past position 30 — including `autogpt-reviewer`'s structured review which is typically posted after several CI runs and sits well beyond the first page.
-
-Two things to extract:
-- **Overall state**: look for `CHANGES_REQUESTED` or `APPROVED` reviews.
-- **Actionable feedback**: non-empty bodies only. Empty-body reviews are thread-resolution events — they indicate progress but have no feedback to act on.
-
-**Where each reviewer posts:**
-- `autogpt-reviewer` — posts detailed structured reviews ("Blockers", "Should Fix", "Nice to Have") as **top-level reviews**. Not present on every PR. Address ALL items.
-- `sentry[bot]` — posts bug predictions as **inline threads**. Fix real bugs, explain false positives.
-- `coderabbitai[bot]` — posts summaries as **top-level reviews** AND actionable items as **inline threads**. Address actionable items.
-- Human reviewers — can post in any source. Address ALL non-empty feedback.
-
-### 3. PR conversation comments — REST
-
-```bash
-gh api repos/Significant-Gravitas/AutoGPT/issues/{N}/comments --paginate
-```
-
-> **Already REST — unaffected by GraphQL rate limits.**
-
-Mostly contains: bot summaries (`coderabbitai[bot]`), CI/conflict detection (`github-actions[bot]`), and author status updates. Scan for non-empty messages from non-bot human reviewers that aren't the PR author — those are the ones that need a response.
-
-## For each unaddressed comment
-
-**CRITICAL: The only valid sequence is fix → commit → push → reply → resolve. Never resolve a thread without a real code commit.**
-
-Resolving a thread via `resolveReviewThread` without an actual fix is the most common failure mode — it makes unresolved counts drop without any real change, producing a false "done" signal. If the issue was genuinely a false positive (no code change needed), reply explaining why and then resolve. Otherwise:
-
-Address comments **one at a time**: fix → commit → push → inline reply → resolve.
-
-1. Read the referenced code, make the fix (or reply explaining why it's not needed)
-2. Commit and push the fix
-3. Reply **inline** (not as a new top-level comment) referencing the fixing commit — this is what resolves the conversation for bot reviewers (coderabbitai, sentry):
-
-Use a **markdown commit link** so GitHub renders it as a clickable reference. Always get the full SHA with `git rev-parse HEAD` **after** committing — never copy a SHA from a previous commit or hardcode one:
-
-```bash
-FULL_SHA=$(git rev-parse HEAD)
-gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/comments/{ID}/replies \
-  -f body="🤖 Fixed in [${FULL_SHA:0:9}](https://github.com/Significant-Gravitas/AutoGPT/commit/${FULL_SHA}): <description>"
-```
-
-| Comment type | How to reply |
-|---|---|
-| Inline review (`pulls/{N}/comments`) | `gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/comments/{ID}/replies -f body="🤖 Fixed in [abc1234](https://github.com/Significant-Gravitas/AutoGPT/commit/FULL_SHA): <description>"` |
-| Conversation (`issues/{N}/comments`) | `gh api repos/Significant-Gravitas/AutoGPT/issues/{N}/comments -f body="🤖 Fixed in [abc1234](https://github.com/Significant-Gravitas/AutoGPT/commit/FULL_SHA): <description>"` |
-
-### What counts as a valid resolution
-
-Only two situations justify calling `resolveReviewThread`:
-
-1. **Real code fix**: you changed the code, committed + pushed, and replied with the SHA. The commit diff must actually address the concern — not just touch the same file.
-2. **Genuine false positive**: the reviewer's concern does not apply to this code, and you can give a specific technical reason (e.g. "Not applicable — `sdk_cwd` is pre-validated by `_make_sdk_cwd()` which applies normpath + prefix assertion before reaching this point").
-
-**Anti-patterns that look resolved but aren't — never do these:**
-- `"Accepted, tracked as follow-up"` — a deferral, not a fix. The concern is still open. Do not resolve.
-- `"Acknowledged"` or `"Same as above"` — these are acknowledgements, not fixes. Do not resolve.
-- `"Fixed in abc1234"` where `abc1234` is a commit that doesn't actually change the flagged line/logic — dishonest. Verify `git show abc1234 -- path/to/file` changes the right thing before posting.
-- Resolving without replying — the reviewer never sees what happened.
-
-When in doubt: if a code change is needed, make it. A deferred issue means the thread stays open until the follow-up PR is merged.
-
-## Codecov coverage
-
-Codecov patch target is **80%** on changed lines. Checks are **informational** (not blocking) but should be green.
-
-### Running coverage locally
-
-**Backend** (from `autogpt_platform/backend/`):
-```bash
-poetry run pytest -s -vv --cov=backend --cov-branch --cov-report term-missing
-```
-
-**Frontend** (from `autogpt_platform/frontend/`):
-```bash
-pnpm vitest run --coverage
-```
-
-### When codecov/patch fails
-
-1. Find uncovered files: `git diff --name-only $(gh pr view --json baseRefName --jq '.baseRefName')...HEAD`
-2. For each uncovered file — extract inline logic to `helpers.ts`/`helpers.py` and test those (highest ROI). Colocate tests as `*_test.py` (backend) or `__tests__/*.test.ts` (frontend).
-3. Run coverage locally to verify, commit, push.
-
-## Format and commit
-
-After fixing, format the changed code:
-
-- **Backend** (from `autogpt_platform/backend/`): `poetry run format`
-- **Frontend** (from `autogpt_platform/frontend/`): `pnpm format && pnpm lint && pnpm types`
-
-If API routes changed, regenerate the frontend client:
-```bash
-cd autogpt_platform/backend && poetry run rest &
-REST_PID=$!
-trap "kill $REST_PID 2>/dev/null" EXIT
-WAIT=0; until curl -sf http://localhost:8006/health > /dev/null 2>&1; do sleep 1; WAIT=$((WAIT+1)); [ $WAIT -ge 60 ] && echo "Timed out" && exit 1; done
-cd ../frontend && pnpm generate:api:force
-kill $REST_PID 2>/dev/null; trap - EXIT
-```
-Never manually edit files in `src/app/api/__generated__/`.
-
-Then commit and **push immediately** — never batch commits without pushing. Each fix should be visible on GitHub right away so CI can start and reviewers can see progress.
-
-**Never push empty commits** (`git commit --allow-empty`) to re-trigger CI or bot checks. When a check fails, investigate the root cause (unchecked PR checklist, unaddressed review comments, code issues) and fix those directly. Empty commits add noise to git history.
-
-For backend commits in worktrees: `poetry run git commit` (pre-commit hooks).
-
-## Coverage
-
-Codecov enforces patch coverage on new/changed lines — new code you write must be tested. Before pushing, verify you haven't left new lines uncovered:
-
-```bash
-cd autogpt_platform/backend
-poetry run pytest --cov=. --cov-report=term-missing {path/to/changed/module}
-```
-
-Look for lines marked `miss` — those are uncovered. Add tests for any new code you wrote as part of addressing comments.
-
-**Rules:**
-- New code you add should have tests
-- Don't remove existing tests when fixing comments
-- If a reviewer asks you to delete code, also delete its tests, but verify coverage hasn't dropped on remaining lines
-
-## The loop
-
-```text
-address comments → format → commit → push
-→ wait for CI (while addressing new comments) → fix failures → push
-→ re-check comments after CI settles
-→ repeat until: all comments addressed AND CI green AND no new comments arriving
-```
-
-### Polling for CI + new comments
-
-After pushing, poll for **both** CI status and new comments in a single loop. Do not use `gh pr checks --watch` — it blocks the tool and prevents reacting to new comments while CI is running.
-
-> **Note:** `gh pr checks --watch --fail-fast` is tempting but it blocks the entire Bash tool call, meaning the agent cannot check for or address new comments until CI fully completes. Always poll manually instead.
-
-**Polling loop — repeat every 30 seconds:**
-
-1. Check CI status:
-```bash
-gh pr checks {N} --repo Significant-Gravitas/AutoGPT --json bucket,name,link
-```
-   Parse the results: if every check has `bucket` of `"pass"` or `"skipping"`, CI is green. If any has `"fail"`, CI has failed. Otherwise CI is still pending.
-
-2. Check for merge conflicts:
-```bash
-gh pr view {N} --repo Significant-Gravitas/AutoGPT --json mergeable --jq '.mergeable'
-```
-   If the result is `"CONFLICTING"`, the PR has a merge conflict — see "Resolving merge conflicts" below. If `"UNKNOWN"`, GitHub is still computing mergeability — wait and re-check next poll.
-
-3. Check for new/changed comments (all three sources):
-
-   **Inline threads** — re-run the GraphQL query from "Fetch comments". For each unresolved thread, record `{thread_id, last_comment_databaseId}` as your baseline. On each poll, action is needed if:
-   - A new thread `id` appears that wasn't in the baseline (new thread), OR
-   - An existing thread's `last_comment_databaseId` has changed (new reply on existing thread)
-
-   **Conversation comments:**
-   ```bash
-   gh api repos/Significant-Gravitas/AutoGPT/issues/{N}/comments --paginate
-   ```
-   Compare total count and newest `id` against baseline. Filter to non-empty, non-bot, non-author-update messages.
-
-   **Top-level reviews:**
-   ```bash
-   gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/reviews --paginate
-   ```
-   Watch for new non-empty reviews (`CHANGES_REQUESTED` or `COMMENTED` with body). Compare total count and newest `id` against baseline.
-
-4. **React in this precedence order (first match wins):**
-
-| What happened | Action |
-|---|---|
-| Merge conflict detected | See "Resolving merge conflicts" below. |
-| Mergeability is `UNKNOWN` | GitHub is still computing mergeability. Sleep 30 seconds, then restart polling from the top. |
-| New comments detected | Address them (fix → commit → push → reply). After pushing, re-fetch all comments to update your baseline, then restart this polling loop from the top (new commits invalidate CI status). |
-| CI failed (bucket == "fail") | Get failed check links: `gh pr checks {N} --repo Significant-Gravitas/AutoGPT --json bucket,link --jq '.[] \| select(.bucket == "fail") \| .link'`. Extract run ID from link (format: `.../actions/runs/<run-id>/job/...`), read logs with `gh run view <run-id> --repo Significant-Gravitas/AutoGPT --log-failed`. Fix → commit → push → restart polling. |
-| CI green + no new comments | **Do not exit immediately.** Bots (coderabbitai, sentry) often post reviews shortly after CI settles. Continue polling for **2 more cycles (60s)** after CI goes green. Only exit after 2 consecutive green+quiet polls. |
-| CI pending + no new comments | Sleep 30 seconds, then poll again. |
-
-**The loop ends when:** CI fully green + all comments addressed + **2 consecutive polls with no new comments after CI settled.**
-
-### Resolving merge conflicts
-
-1. Identify the PR's target branch and remote:
-```bash
-gh pr view {N} --repo Significant-Gravitas/AutoGPT --json baseRefName --jq '.baseRefName'
-git remote -v   # find the remote pointing to Significant-Gravitas/AutoGPT (typically 'upstream' in forks, 'origin' for direct contributors)
-```
-
-2. Pull the latest base branch with a 3-way merge:
-```bash
-git pull {base-remote} {base-branch} --no-rebase
-```
-
-3. Resolve conflicting files, then verify no conflict markers remain:
-```bash
-if grep -R -n -E '^(<<<<<<<|=======|>>>>>>>)' <conflicted-files>; then
-  echo "Unresolved conflict markers found — resolve before proceeding."
-  exit 1
-fi
-```
-
-4. Stage and push:
-```bash
-git add <conflicted-files>
-git commit -m "Resolve merge conflicts with {base-branch}"
-git push
-```
-
-5. Restart the polling loop from the top — new commits reset CI status.
-
-## GitHub rate limits
-
-Three distinct rate limits exist — they have different causes, error shapes, and recovery times:
-
-| Error | HTTP code | Cause | Recovery |
-|---|---|---|---|
-| `{"code":"abuse"}` | 403 | Secondary rate limit — too many write operations (comments, mutations) in a short window | Wait **2–3 minutes**. 60s is often not enough. |
-| `{"message":"API rate limit exceeded"}` | 429 | Primary REST rate limit — 5000 calls/hr per user | Wait until `X-RateLimit-Reset` header timestamp |
-| `GraphQL: API rate limit already exceeded for user ID ...` | 403 on stderr, `gh` exits 1 | **GraphQL-specific** per-user limit — distinct from REST's 5000/hr and from the abuse secondary limit. Trips faster than REST because point costs per query. | Wait until the GraphQL window resets (typically ~1 hour from the first call in the window). REST still works — use fallbacks below. |
-
-**Prevention:** Add `sleep 3` between individual thread reply API calls. When posting >20 replies, increase to `sleep 5`.
-
-### Detection
-
-The `gh` CLI surfaces the GraphQL limit on stderr with the exact string `GraphQL: API rate limit already exceeded for user ID <id>` and exits 1 — any `gh api graphql ...` **or** `gh pr view ...` call fails. Check current quota and reset time via the REST endpoint that reports GraphQL quota (this call is REST and still works whether GraphQL is rate-limited OR fully down):
-
-```bash
-gh api rate_limit --jq '.resources.graphql'   # { "limit": 5000, "used": 5000, "remaining": 0, "reset": 1729...}
-# Human-readable reset:
-gh api rate_limit --jq '.resources.graphql.reset' | xargs -I{} date -r {}
-```
-
-Retry when `remaining > 0`. If you need to proceed sooner, sleep 2–5 min and probe again — the limit is per user, not per machine, so other concurrent agents under the same token also consume it.
-
-### What keeps working
-
-When GraphQL is unavailable (rate-limited or outage):
-
-- **Keeps working (REST):** top-level reviews fetch, conversation comments fetch, all inline-comment replies, CI status (`gh pr checks`), and the `gh api rate_limit` probe.
-- **Degraded:** inline thread list — fall back to flat `/pulls/{N}/comments` REST, which drops thread grouping, `isResolved`, and Relay thread IDs. You still get comment bodies and the `databaseId` as `id`, enough to read and reply.
-- **Blocked:** `gh pr view`, the `resolveReviewThread` mutation, and any new `gh api graphql` queries — wait for the quota to reset.
-
-### Fall back to REST
-
-**PR metadata reads** — `gh pr view` uses GraphQL under the hood; use the REST pulls endpoint instead, which returns the full PR object:
-
-```bash
-gh api repos/Significant-Gravitas/AutoGPT/pulls/{N} --jq '.body'           # == --json body
-gh api repos/Significant-Gravitas/AutoGPT/pulls/{N} --jq '.base.ref'       # == --json baseRefName
-gh api repos/Significant-Gravitas/AutoGPT/pulls/{N} --jq '.mergeable'      # == --json mergeable
-```
-
-Note: REST `mergeable` returns `true|false|null`; GraphQL returns `MERGEABLE|CONFLICTING|UNKNOWN`. The `null` case maps to `UNKNOWN` — treat it the same (still computing; poll again).
-
-**Inline comments (flat list)** — no thread grouping or `isResolved`, but enough to read and reply:
-
-```bash
-gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/comments --paginate \
-  | jq '[.[] | {id, path, line, user: .user.login, body: .body[:200], in_reply_to_id}]'
-```
-
-Use this degraded mode to make progress on the fix → reply loop, then return to GraphQL for `resolveReviewThread` once the rate limit resets.
-
-**Replies** — already REST-native (`/pulls/{N}/comments/{ID}/replies`); no change needed, use the same command as the main flow.
-
-**`resolveReviewThread`** — **no REST equivalent**; GitHub does not expose a REST endpoint for thread resolution. Queue the thread IDs needing resolution, wait for the GraphQL limit to reset, then run the resolve mutations in a batch (with `sleep 3` between calls, per the secondary-limit guidance).
-
-### Recovery from secondary rate limit (403 abuse)
-
-1. Stop all API writes immediately
-2. Wait **2 minutes minimum** (not 60s — secondary limits are stricter)
-3. Resume with `sleep 3` between each call
-4. If 403 persists after 2 min, wait another 2 min before retrying
-
-Never batch all replies in a tight loop — always space them out.
-
-## Parallel thread resolution
-
-When a PR has more than 10 unresolved threads, addressing one commit per thread is slow. Use this strategy instead:
-
-### Group by file, batch per commit
-
-1. Sort `ALL_THREADS` by `path` — threads in the same file can share a single commit.
-2. Fix all threads in one file → `git commit` → `git push` → reply to **all** those threads with the same SHA → resolve them all.
-3. Move to the next file group and repeat.
-
-This reduces N commits to (number of files touched), which is usually 3–5 instead of 15–30.
-
-### Posting replies concurrently (for large batches)
-
-For truly independent thread groups (different files, no shared logic), you can post replies in parallel using background subshells — but always space out API writes:
-
-```bash
-# Post replies to a batch of threads concurrently, 3s apart
-(
-  sleep 3
-  gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/comments/{ID1}/replies \
-    -f body="🤖 Fixed in [${FULL_SHA:0:9}](https://github.com/Significant-Gravitas/AutoGPT/commit/${FULL_SHA}): ..."
-) &
-(
-  sleep 6
-  gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/comments/{ID2}/replies \
-    -f body="🤖 Fixed in [${FULL_SHA:0:9}](https://github.com/Significant-Gravitas/AutoGPT/commit/${FULL_SHA}): ..."
-) &
-wait  # wait for all background replies before resolving
-```
-
-Then resolve sequentially (GraphQL mutations):
-```bash
-for THREAD_ID in "$THREAD1" "$THREAD2" "$THREAD3"; do
-  gh api graphql -f query="mutation { resolveReviewThread(input: {threadId: \"${THREAD_ID}\"}) { thread { isResolved } } }"
-  sleep 3
-done
-```
-
-**Always sleep 3s between individual API writes** — GitHub's secondary rate limit (403) triggers on bursts of >20 writes. Increase to `sleep 5` when posting more than 20 replies in a batch.
-
-## Resolving threads via GraphQL
-
-Use `resolveReviewThread` **only after** the commit is pushed and the reply is posted:
-
-```bash
-gh api graphql -f query='mutation { resolveReviewThread(input: {threadId: "THREAD_ID"}) { thread { isResolved } } }'
-```
-
-**Never call this mutation before committing the fix.** The orchestrator will verify actual unresolved counts via GraphQL after you output `ORCHESTRATOR:DONE` — false resolutions will be caught and you will be re-briefed.
-
-> `resolveReviewThread` is GraphQL-only — no REST equivalent. If GraphQL is rate-limited, see [GitHub rate limits](#github-rate-limits) for the queue-and-retry flow.
-
-### Verify actual count before outputting ORCHESTRATOR:DONE
-
-Before claiming "0 unresolved threads", always query GitHub directly — don't rely on your own bookkeeping. Paginate all pages — a single `first: 100` query misses threads beyond page 1:
-
-```bash
-# Step 1: get total thread count
-gh api graphql -f query='
-{
-  repository(owner: "Significant-Gravitas", name: "AutoGPT") {
-    pullRequest(number: {N}) {
-      reviewThreads { totalCount }
-    }
-  }
-}' | jq '.data.repository.pullRequest.reviewThreads.totalCount'
-
-# Step 2: paginate all pages, count truly unresolved
-CURSOR=""; UNRESOLVED=0
-while true; do
-  AFTER=${CURSOR:+", after: \"$CURSOR\""}
-  PAGE=$(gh api graphql -f query="
-  {
-    repository(owner: \"Significant-Gravitas\", name: \"AutoGPT\") {
-      pullRequest(number: {N}) {
-        reviewThreads(first: 100${AFTER}) {
-          pageInfo { hasNextPage endCursor }
-          nodes { isResolved }
-        }
-      }
-    }
-  }")
-  UNRESOLVED=$(( UNRESOLVED + $(echo "$PAGE" | jq '[.data.repository.pullRequest.reviewThreads.nodes[] | select(.isResolved==false)] | length') ))
-  HAS_NEXT=$(echo "$PAGE" | jq -r '.data.repository.pullRequest.reviewThreads.pageInfo.hasNextPage')
-  CURSOR=$(echo "$PAGE" | jq -r '.data.repository.pullRequest.reviewThreads.pageInfo.endCursor')
-  [ "$HAS_NEXT" = "false" ] && break
-done
-echo "Unresolved threads: $UNRESOLVED"
-```
-
-Only output `ORCHESTRATOR:DONE` after this loop reports 0.

.claude/skills/pr-review/SKILL.md

@@ -1,86 +0,0 @@
----
-name: pr-review
-description: Review a PR for correctness, security, code quality, and testing issues. TRIGGER when user asks to review a PR, check PR quality, or give feedback on a PR.
-user-invocable: true
-args: "[PR number or URL] — if omitted, finds PR for current branch."
-metadata:
-  author: autogpt-team
-  version: "1.0.0"
----
-
-# PR Review
-
-## Find the PR
-
-```bash
-gh pr list --head $(git branch --show-current) --repo Significant-Gravitas/AutoGPT
-gh pr view {N}
-```
-
-## Read the PR description
-
-Before reading code, understand the **why**, **what**, and **how** from the PR description:
-
-```bash
-gh pr view {N} --json body --jq '.body'
-```
-
-Every PR should have a Why / What / How structure. If any of these are missing, note it as feedback.
-
-## Read the diff
-
-```bash
-gh pr diff {N}
-```
-
-## Fetch existing review comments
-
-Before posting anything, fetch existing inline comments to avoid duplicates:
-
-```bash
-gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/comments --paginate
-gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/reviews
-```
-
-## What to check
-
-**Description quality:** Does the PR description cover Why (motivation/problem), What (summary of changes), and How (approach/implementation details)? If any are missing, request them — you can't judge the approach without understanding the problem and intent.
-
-**Correctness:** logic errors, off-by-one, missing edge cases, race conditions (TOCTOU in file access, credit charging), error handling gaps, async correctness (missing `await`, unclosed resources).
-
-**Security:** input validation at boundaries, no injection (command, XSS, SQL), secrets not logged, file paths sanitized (`os.path.basename()` in error messages).
-
-**Code quality:** apply rules from backend/frontend CLAUDE.md files.
-
-**Architecture:** DRY, single responsibility, modular functions. `Security()` vs `Depends()` for FastAPI auth. `data:` for SSE events, `: comment` for heartbeats. `transaction=True` for Redis pipelines.
-
-**Testing:** edge cases covered, colocated `*_test.py` (backend) / `__tests__/` (frontend), mocks target where symbol is **used** not defined, `AsyncMock` for async.
-
-## Output format
-
-Every comment **must** be prefixed with `🤖` and a criticality badge:
-
-| Tier | Badge | Meaning |
-|---|---|---|
-| Blocker | `🔴 **Blocker**` | Must fix before merge |
-| Should Fix | `🟠 **Should Fix**` | Important improvement |
-| Nice to Have | `🟡 **Nice to Have**` | Minor suggestion |
-| Nit | `🔵 **Nit**` | Style / wording |
-
-Example: `🤖 🔴 **Blocker**: Missing error handling for X — suggest wrapping in try/except.`
-
-## Post inline comments
-
-For each finding, post an inline comment on the PR (do not just write a local report):
-
-```bash
-# Get the latest commit SHA for the PR
-COMMIT_SHA=$(gh api repos/Significant-Gravitas/AutoGPT/pulls/{N} --jq '.head.sha')
-
-# Post an inline comment on a specific file/line
-gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/comments \
-  -f body="🤖 🔴 **Blocker**: <description>" \
-  -f commit_id="$COMMIT_SHA" \
-  -f path="<file path>" \
-  -F line=<line number>
-```

.claude/skills/setup-repo/SKILL.md

@@ -1,195 +0,0 @@
----
-name: setup-repo
-description: Initialize a worktree-based repo layout for parallel development. Creates a main worktree, a reviews worktree for PR reviews, and N numbered work branches. Handles .env creation, dependency installation, and branchlet config. TRIGGER when user asks to set up the repo from scratch, initialize worktrees, bootstrap their dev environment, "setup repo", "setup worktrees", "initialize dev environment", "set up branches", or when a freshly cloned repo has no sibling worktrees.
-user-invocable: true
-args: "No arguments — interactive setup via prompts."
-metadata:
-  author: autogpt-team
-  version: "1.0.0"
----
-
-# Repository Setup
-
-This skill sets up a worktree-based development layout from a freshly cloned repo. It creates:
-- A **main** worktree (the primary checkout)
-- A **reviews** worktree (for PR reviews)
-- **N work branches** (branch1..branchN) for parallel development
-
-## Step 1: Identify the repo
-
-Determine the repo root and parent directory:
-
-```bash
-ROOT=$(git rev-parse --show-toplevel)
-REPO_NAME=$(basename "$ROOT")
-PARENT=$(dirname "$ROOT")
-```
-
-Detect if the repo is already inside a worktree layout by counting sibling worktrees (not just checking the directory name, which could be anything):
-
-```bash
-# Count worktrees that are siblings (live under $PARENT but aren't $ROOT itself)
-SIBLING_COUNT=$(git worktree list --porcelain 2>/dev/null | grep "^worktree " | grep -c "$PARENT/" || true)
-if [ "$SIBLING_COUNT" -gt 1 ]; then
-  echo "INFO: Existing worktree layout detected at $PARENT ($SIBLING_COUNT worktrees)"
-  # Use $ROOT as-is; skip renaming/restructuring
-else
-  echo "INFO: Fresh clone detected, proceeding with setup"
-fi
-```
-
-## Step 2: Ask the user questions
-
-Use AskUserQuestion to gather setup preferences:
-
-1. **How many parallel work branches do you need?** (Options: 4, 8, 16, or custom)
-   - These become `branch1` through `branchN`
-2. **Which branch should be the base?** (Options: origin/master, origin/dev, or custom)
-   - All work branches and reviews will start from this
-
-## Step 3: Fetch and set up branches
-
-```bash
-cd "$ROOT"
-git fetch origin
-
-# Create the reviews branch from base (skip if already exists)
-if git show-ref --verify --quiet refs/heads/reviews; then
-  echo "INFO: Branch 'reviews' already exists, skipping"
-else
-  git branch reviews <base-branch>
-fi
-
-# Create numbered work branches from base (skip if already exists)
-for i in $(seq 1 "$COUNT"); do
-  if git show-ref --verify --quiet "refs/heads/branch$i"; then
-    echo "INFO: Branch 'branch$i' already exists, skipping"
-  else
-    git branch "branch$i" <base-branch>
-  fi
-done
-```
-
-## Step 4: Create worktrees
-
-Create worktrees as siblings to the main checkout:
-
-```bash
-if [ -d "$PARENT/reviews" ]; then
-  echo "INFO: Worktree '$PARENT/reviews' already exists, skipping"
-else
-  git worktree add "$PARENT/reviews" reviews
-fi
-
-for i in $(seq 1 "$COUNT"); do
-  if [ -d "$PARENT/branch$i" ]; then
-    echo "INFO: Worktree '$PARENT/branch$i' already exists, skipping"
-  else
-    git worktree add "$PARENT/branch$i" "branch$i"
-  fi
-done
-```
-
-## Step 5: Set up environment files
-
-**Do NOT assume .env files exist.** For each worktree (including main if needed):
-
-1. Check if `.env` exists in the source worktree for each path
-2. If `.env` exists, copy it
-3. If only `.env.default` or `.env.example` exists, copy that as `.env`
-4. If neither exists, warn the user and list which env files are missing
-
-Env file locations to check (same as the `/worktree` skill — keep these in sync):
-- `autogpt_platform/.env`
-- `autogpt_platform/backend/.env`
-- `autogpt_platform/frontend/.env`
-
-> **Note:** This env copying logic intentionally mirrors the `/worktree` skill's approach. If you update the path list or fallback logic here, update `/worktree` as well.
-
-```bash
-SOURCE="$ROOT"
-WORKTREES="reviews"
-for i in $(seq 1 "$COUNT"); do WORKTREES="$WORKTREES branch$i"; done
-
-FOUND_ANY_ENV=0
-for wt in $WORKTREES; do
-  TARGET="$PARENT/$wt"
-  for envpath in autogpt_platform autogpt_platform/backend autogpt_platform/frontend; do
-    if [ -f "$SOURCE/$envpath/.env" ]; then
-      FOUND_ANY_ENV=1
-      cp "$SOURCE/$envpath/.env" "$TARGET/$envpath/.env"
-    elif [ -f "$SOURCE/$envpath/.env.default" ]; then
-      FOUND_ANY_ENV=1
-      cp "$SOURCE/$envpath/.env.default" "$TARGET/$envpath/.env"
-      echo "NOTE: $wt/$envpath/.env was created from .env.default — you may need to edit it"
-    elif [ -f "$SOURCE/$envpath/.env.example" ]; then
-      FOUND_ANY_ENV=1
-      cp "$SOURCE/$envpath/.env.example" "$TARGET/$envpath/.env"
-      echo "NOTE: $wt/$envpath/.env was created from .env.example — you may need to edit it"
-    else
-      echo "WARNING: No .env, .env.default, or .env.example found at $SOURCE/$envpath/"
-    fi
-  done
-done
-
-if [ "$FOUND_ANY_ENV" -eq 0 ]; then
-  echo "WARNING: No environment files or templates were found in the source worktree."
-  # Use AskUserQuestion to confirm: "Continue setup without env files?"
-  # If the user declines, stop here and let them set up .env files first.
-fi
-```
-
-## Step 6: Copy branchlet config
-
-Copy `.branchlet.json` from main to each worktree so branchlet can manage sub-worktrees:
-
-```bash
-if [ -f "$ROOT/.branchlet.json" ]; then
-  for wt in $WORKTREES; do
-    cp "$ROOT/.branchlet.json" "$PARENT/$wt/.branchlet.json"
-  done
-fi
-```
-
-## Step 7: Install dependencies
-
-Install deps in all worktrees. Run these sequentially per worktree:
-
-```bash
-for wt in $WORKTREES; do
-  TARGET="$PARENT/$wt"
-  echo "=== Installing deps for $wt ==="
-  (cd "$TARGET/autogpt_platform/autogpt_libs" && poetry install) &&
-  (cd "$TARGET/autogpt_platform/backend" && poetry install && poetry run prisma generate) &&
-  (cd "$TARGET/autogpt_platform/frontend" && pnpm install) &&
-  echo "=== Done: $wt ===" ||
-  echo "=== FAILED: $wt ==="
-done
-```
-
-This is slow. Run in background if possible and notify when complete.
-
-## Step 8: Verify and report
-
-After setup, verify and report to the user:
-
-```bash
-git worktree list
-```
-
-Summarize:
-- Number of worktrees created
-- Which env files were copied vs created from defaults vs missing
-- Any warnings or errors encountered
-
-## Final directory layout
-
-```
-parent/
-  main/              # Primary checkout (already exists)
-  reviews/           # PR review worktree
-  branch1/           # Work branch 1
-  branch2/           # Work branch 2
-  ...
-  branchN/           # Work branch N
-```

.claude/skills/worktree/SKILL.md

@@ -1,85 +0,0 @@
----
-name: worktree
-description: Set up a new git worktree for parallel development. Creates the worktree, copies .env files, installs dependencies, and generates Prisma client. TRIGGER when user asks to set up a worktree, work on a branch in isolation, or needs a separate environment for a branch or PR.
-user-invocable: true
-args: "[name] — optional worktree name (e.g., 'AutoGPT7'). If omitted, uses next available AutoGPT<N>."
-metadata:
-  author: autogpt-team
-  version: "3.0.0"
----
-
-# Worktree Setup
-
-## Create the worktree
-
-Derive paths from the git toplevel. If a name is provided as argument, use it. Otherwise, check `git worktree list` and pick the next `AutoGPT<N>`.
-
-```bash
-ROOT=$(git rev-parse --show-toplevel)
-PARENT=$(dirname "$ROOT")
-
-# From an existing branch
-git worktree add "$PARENT/<NAME>" <branch-name>
-
-# From a new branch off dev
-git worktree add -b <new-branch> "$PARENT/<NAME>" dev
-```
-
-## Copy environment files
-
-Copy `.env` from the root worktree. Falls back to `.env.default` if `.env` doesn't exist.
-
-```bash
-ROOT=$(git rev-parse --show-toplevel)
-TARGET="$(dirname "$ROOT")/<NAME>"
-
-for envpath in autogpt_platform/backend autogpt_platform/frontend autogpt_platform; do
-  if [ -f "$ROOT/$envpath/.env" ]; then
-    cp "$ROOT/$envpath/.env" "$TARGET/$envpath/.env"
-  elif [ -f "$ROOT/$envpath/.env.default" ]; then
-    cp "$ROOT/$envpath/.env.default" "$TARGET/$envpath/.env"
-  fi
-done
-```
-
-## Install dependencies
-
-```bash
-TARGET="$(dirname "$(git rev-parse --show-toplevel)")/<NAME>"
-cd "$TARGET/autogpt_platform/autogpt_libs" && poetry install
-cd "$TARGET/autogpt_platform/backend" && poetry install && poetry run prisma generate
-cd "$TARGET/autogpt_platform/frontend" && pnpm install
-```
-
-Replace `<NAME>` with the actual worktree name (e.g., `AutoGPT7`).
-
-## Running the app (optional)
-
-Backend uses ports: 8001, 8002, 8003, 8005, 8006, 8007, 8008. Free them first if needed:
-
-```bash
-TARGET="$(dirname "$(git rev-parse --show-toplevel)")/<NAME>"
-for port in 8001 8002 8003 8005 8006 8007 8008; do
-  lsof -ti :$port | xargs kill -9 2>/dev/null || true
-done
-cd "$TARGET/autogpt_platform/backend" && poetry run app
-```
-
-## CoPilot testing
-
-SDK mode spawns a Claude subprocess — won't work inside Claude Code. Set `CHAT_USE_CLAUDE_AGENT_SDK=false` in `backend/.env` to use baseline mode.
-
-## Cleanup
-
-```bash
-# Replace <NAME> with the actual worktree name (e.g., AutoGPT7)
-git worktree remove "$(dirname "$(git rev-parse --show-toplevel)")/<NAME>"
-```
-
-## Alternative: Branchlet (optional)
-
-If [branchlet](https://www.npmjs.com/package/branchlet) is installed:
-
-```bash
-branchlet create -n <name> -s <source-branch> -b <new-branch>
-```

.claude/skills/write-frontend-tests/SKILL.md

@@ -1,225 +0,0 @@
----
-name: write-frontend-tests
-description: "Analyze the current branch diff against dev, plan integration tests for changed frontend pages/components, and write them. TRIGGER when user asks to write frontend tests, add test coverage, or 'write tests for my changes'."
-user-invocable: true
-args: "[base branch] — defaults to dev. Optionally pass a specific base branch to diff against."
-metadata:
-  author: autogpt-team
-  version: "1.0.0"
----
-
-# Write Frontend Tests
-
-Analyze the current branch's frontend changes, plan integration tests, and write them.
-
-## References
-
-Before writing any tests, read the testing rules and conventions:
-
-- `autogpt_platform/frontend/TESTING.md` — testing strategy, file locations, examples
-- `autogpt_platform/frontend/src/tests/AGENTS.md` — detailed testing rules, MSW patterns, decision flowchart
-- `autogpt_platform/frontend/src/tests/integrations/test-utils.tsx` — custom render with providers
-- `autogpt_platform/frontend/src/tests/integrations/vitest.setup.tsx` — MSW server setup
-
-## Step 1: Identify changed frontend files
-
-```bash
-BASE_BRANCH="${ARGUMENTS:-dev}"
-cd autogpt_platform/frontend
-
-# Get changed frontend files (excluding generated, config, and test files)
-git diff "$BASE_BRANCH"...HEAD --name-only -- src/ \
-  | grep -v '__generated__' \
-  | grep -v '__tests__' \
-  | grep -v '\.test\.' \
-  | grep -v '\.stories\.' \
-  | grep -v '\.spec\.'
-```
-
-Also read the diff to understand what changed:
-
-```bash
-git diff "$BASE_BRANCH"...HEAD --stat -- src/
-git diff "$BASE_BRANCH"...HEAD -- src/ | head -500
-```
-
-## Step 2: Categorize changes and find test targets
-
-For each changed file, determine:
-
-1. **Is it a page?** (`page.tsx`) — these are the primary test targets
-2. **Is it a hook?** (`use*.ts`) — test via the page/component that uses it; avoid direct `renderHook()` tests unless it is a shared reusable hook with standalone business logic
-3. **Is it a component?** (`.tsx` in `components/`) — test via the parent page unless it's complex enough to warrant isolation
-4. **Is it a helper?** (`helpers.ts`, `utils.ts`) — unit test directly if pure logic
-
-**Priority order:**
-
-1. Pages with new/changed data fetching or user interactions
-2. Components with complex internal logic (modals, forms, wizards)
-3. Shared hooks with standalone business logic when UI-level coverage is impractical
-4. Pure helper functions
-
-Skip: styling-only changes, type-only changes, config changes.
-
-## Step 3: Check for existing tests
-
-For each test target, check if tests already exist:
-
-```bash
-# For a page at src/app/(platform)/library/page.tsx
-ls src/app/\(platform\)/library/__tests__/ 2>/dev/null
-
-# For a component at src/app/(platform)/library/components/AgentCard/AgentCard.tsx
-ls src/app/\(platform\)/library/components/AgentCard/__tests__/ 2>/dev/null
-```
-
-Note which targets have no tests (need new files) vs which have tests that need updating.
-
-## Step 4: Identify API endpoints used
-
-For each test target, find which API hooks are used:
-
-```bash
-# Find generated API hook imports in the changed files
-grep -rn 'from.*__generated__/endpoints' src/app/\(platform\)/library/
-grep -rn 'use[A-Z].*V[12]' src/app/\(platform\)/library/
-```
-
-For each API hook found, locate the corresponding MSW handler:
-
-```bash
-# If the page uses useGetV2ListLibraryAgents, find its MSW handlers
-grep -rn 'getGetV2ListLibraryAgents.*Handler' src/app/api/__generated__/endpoints/library/library.msw.ts
-```
-
-List every MSW handler you will need (200 for happy path, 4xx for error paths).
-
-## Step 5: Write the test plan
-
-Before writing code, output a plan as a numbered list:
-
-```
-Test plan for [branch name]:
-
-1. src/app/(platform)/library/__tests__/main.test.tsx (NEW)
-   - Renders page with agent list (MSW 200)
-   - Shows loading state
-   - Shows error state (MSW 422)
-   - Handles empty agent list
-
-2. src/app/(platform)/library/__tests__/search.test.tsx (NEW)
-   - Filters agents by search query
-   - Shows no results message
-   - Clears search
-
-3. src/app/(platform)/library/components/AgentCard/__tests__/AgentCard.test.tsx (UPDATE)
-   - Add test for new "duplicate" action
-```
-
-Present this plan to the user. Wait for confirmation before proceeding. If the user has feedback, adjust the plan.
-
-## Step 6: Write the tests
-
-For each test file in the plan, follow these conventions:
-
-### File structure
-
-```tsx
-import { render, screen, waitFor } from "@/tests/integrations/test-utils";
-import { server } from "@/mocks/mock-server";
-// Import MSW handlers for endpoints the page uses
-import {
-  getGetV2ListLibraryAgentsMockHandler200,
-  getGetV2ListLibraryAgentsMockHandler422,
-} from "@/app/api/__generated__/endpoints/library/library.msw";
-// Import the component under test
-import LibraryPage from "../page";
-
-describe("LibraryPage", () => {
-  test("renders agent list from API", async () => {
-    server.use(getGetV2ListLibraryAgentsMockHandler200());
-
-    render(<LibraryPage />);
-
-    expect(await screen.findByText(/my agents/i)).toBeDefined();
-  });
-
-  test("shows error state on API failure", async () => {
-    server.use(getGetV2ListLibraryAgentsMockHandler422());
-
-    render(<LibraryPage />);
-
-    expect(await screen.findByText(/error/i)).toBeDefined();
-  });
-});
-```
-
-### Rules
-
-- Use `render()` from `@/tests/integrations/test-utils` (NOT from `@testing-library/react` directly)
-- Use `server.use()` to set up MSW handlers BEFORE rendering
-- Use `findBy*` (async) for elements that appear after data fetching — NOT `getBy*`
-- Use `getBy*` only for elements that are immediately present in the DOM
-- Use `screen` queries — do NOT destructure from `render()`
-- Use `waitFor` when asserting side effects or state changes after interactions
-- Import `fireEvent` or `userEvent` from the test-utils for interactions
-- Do NOT mock internal hooks or functions — mock at the API boundary via MSW
-- Prefer Orval-generated MSW handlers and response builders over hand-built API response objects
-- Do NOT use `act()` manually — `render` and `fireEvent` handle it
-- Keep tests focused: one behavior per test
-- Use descriptive test names that read like sentences
-
-### Test location
-
-```
-# For pages: __tests__/ next to page.tsx
-src/app/(platform)/library/__tests__/main.test.tsx
-
-# For complex standalone components: __tests__/ inside component folder
-src/app/(platform)/library/components/AgentCard/__tests__/AgentCard.test.tsx
-
-# For pure helpers: co-located .test.ts
-src/app/(platform)/library/helpers.test.ts
-```
-
-### Custom MSW overrides
-
-When the auto-generated faker data is not enough, override with specific data:
-
-```tsx
-import { http, HttpResponse } from "msw";
-
-server.use(
-  http.get("http://localhost:3000/api/proxy/api/v2/library/agents", () => {
-    return HttpResponse.json({
-      agents: [{ id: "1", name: "Test Agent", description: "A test agent" }],
-      pagination: { total_items: 1, total_pages: 1, page: 1, page_size: 10 },
-    });
-  }),
-);
-```
-
-Use the proxy URL pattern: `http://localhost:3000/api/proxy/api/v{version}/{path}` — this matches the MSW base URL configured in `orval.config.ts`.
-
-## Step 7: Run and verify
-
-After writing all tests:
-
-```bash
-cd autogpt_platform/frontend
-pnpm test:unit --reporter=verbose
-```
-
-If tests fail:
-
-1. Read the error output carefully
-2. Fix the test (not the source code, unless there is a genuine bug)
-3. Re-run until all pass
-
-Then run the full checks:
-
-```bash
-pnpm format
-pnpm lint
-pnpm types
-```

.github/PULL_REQUEST_TEMPLATE.md

@@ -1,12 +1,8 @@
-### Why / What / How
-
-<!-- Why: Why does this PR exist? What problem does it solve, or what's broken/missing without it? -->
-<!-- What: What does this PR change? Summarize the changes at a high level. -->
-<!-- How: How does it work? Describe the approach, key implementation details, or architecture decisions. -->
+<!-- Clearly explain the need for these changes: -->
 
 ### Changes 🏗️
 
-<!-- List the key changes. Keep it higher level than the diff but specific enough to highlight what's new/modified. -->
+<!-- Concisely describe all of the changes made in this pull request: -->
 
 ### Checklist 📋
 

.github/workflows/classic-autogpt-ci.yml

@@ -6,19 +6,11 @@ on:
     paths:
       - '.github/workflows/classic-autogpt-ci.yml'
       - 'classic/original_autogpt/**'
-      - 'classic/direct_benchmark/**'
-      - 'classic/forge/**'
-      - 'classic/pyproject.toml'
-      - 'classic/poetry.lock'
   pull_request:
     branches: [ master, dev, release-* ]
     paths:
       - '.github/workflows/classic-autogpt-ci.yml'
       - 'classic/original_autogpt/**'
-      - 'classic/direct_benchmark/**'
-      - 'classic/forge/**'
-      - 'classic/pyproject.toml'
-      - 'classic/poetry.lock'
 
 concurrency:
   group: ${{ format('classic-autogpt-ci-{0}', github.head_ref && format('{0}-{1}', github.event_name, github.event.pull_request.number) || github.sha) }}
@@ -27,22 +19,47 @@ concurrency:
 defaults:
   run:
     shell: bash
-    working-directory: classic
+    working-directory: classic/original_autogpt
 
 jobs:
   test:
     permissions:
       contents: read
     timeout-minutes: 30
-    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.10"]
+        platform-os: [ubuntu, macos, macos-arm64, windows]
+    runs-on: ${{ matrix.platform-os != 'macos-arm64' && format('{0}-latest', matrix.platform-os) || 'macos-14' }}
 
     steps:
-      - name: Start MinIO service
+      # Quite slow on macOS (2~4 minutes to set up Docker)
+      # - name: Set up Docker (macOS)
+      #   if: runner.os == 'macOS'
+      #   uses: crazy-max/ghaction-setup-docker@v3
+
+      - name: Start MinIO service (Linux)
+        if: runner.os == 'Linux'
         working-directory: '.'
         run: |
           docker pull minio/minio:edge-cicd
           docker run -d -p 9000:9000 minio/minio:edge-cicd
 
+      - name: Start MinIO service (macOS)
+        if: runner.os == 'macOS'
+        working-directory: ${{ runner.temp }}
+        run: |
+          brew install minio/stable/minio
+          mkdir data
+          minio server ./data &
+
+      # No MinIO on Windows:
+      # - Windows doesn't support running Linux Docker containers
+      # - It doesn't seem possible to start background processes on Windows. They are
+      #   killed after the step returns.
+      #   See: https://github.com/actions/runner/issues/598#issuecomment-2011890429
+
       - name: Checkout repository
         uses: actions/checkout@v4
         with:
@@ -54,23 +71,41 @@ jobs:
           git config --global user.name "Auto-GPT-Bot"
           git config --global user.email "github-bot@agpt.co"
 
-      - name: Set up Python 3.12
+      - name: Set up Python ${{ matrix.python-version }}
         uses: actions/setup-python@v5
         with:
-          python-version: "3.12"
+          python-version: ${{ matrix.python-version }}
 
       - id: get_date
         name: Get date
         run: echo "date=$(date +'%Y-%m-%d')" >> $GITHUB_OUTPUT
 
       - name: Set up Python dependency cache
+        # On Windows, unpacking cached dependencies takes longer than just installing them
+        if: runner.os != 'Windows'
         uses: actions/cache@v4
         with:
-          path: ~/.cache/pypoetry
-          key: poetry-${{ runner.os }}-${{ hashFiles('classic/poetry.lock') }}
+          path: ${{ runner.os == 'macOS' && '~/Library/Caches/pypoetry' || '~/.cache/pypoetry' }}
+          key: poetry-${{ runner.os }}-${{ hashFiles('classic/original_autogpt/poetry.lock') }}
 
-      - name: Install Poetry
-        run: curl -sSL https://install.python-poetry.org | python3 -
+      - name: Install Poetry (Unix)
+        if: runner.os != 'Windows'
+        run: |
+          curl -sSL https://install.python-poetry.org | python3 -
+
+          if [ "${{ runner.os }}" = "macOS" ]; then
+            PATH="$HOME/.local/bin:$PATH"
+            echo "$HOME/.local/bin" >> $GITHUB_PATH
+          fi
+
+      - name: Install Poetry (Windows)
+        if: runner.os == 'Windows'
+        shell: pwsh
+        run: |
+          (Invoke-WebRequest -Uri https://install.python-poetry.org -UseBasicParsing).Content | python -
+
+          $env:PATH += ";$env:APPDATA\Python\Scripts"
+          echo "$env:APPDATA\Python\Scripts" >> $env:GITHUB_PATH
 
       - name: Install Python dependencies
         run: poetry install
@@ -81,13 +116,12 @@ jobs:
             --cov=autogpt --cov-branch --cov-report term-missing --cov-report xml \
             --numprocesses=logical --durations=10 \
             --junitxml=junit.xml -o junit_family=legacy \
-            original_autogpt/tests/unit original_autogpt/tests/integration
+            tests/unit tests/integration
         env:
           CI: true
           PLAIN_OUTPUT: True
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
-          S3_ENDPOINT_URL: http://127.0.0.1:9000
+          S3_ENDPOINT_URL: ${{ runner.os != 'Windows' && 'http://127.0.0.1:9000' || '' }}
           AWS_ACCESS_KEY_ID: minioadmin
           AWS_SECRET_ACCESS_KEY: minioadmin
 
@@ -101,11 +135,11 @@ jobs:
         uses: codecov/codecov-action@v5
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
-          flags: autogpt-agent
+          flags: autogpt-agent,${{ runner.os }}
 
       - name: Upload logs to artifact
         if: always()
         uses: actions/upload-artifact@v4
         with:
           name: test-logs
-          path: classic/logs/
+          path: classic/original_autogpt/logs/

.github/workflows/classic-autogpt-docker-ci.yml

@@ -148,7 +148,7 @@ jobs:
             --entrypoint poetry ${{ env.IMAGE_NAME }} run \
             pytest -v --cov=autogpt --cov-branch --cov-report term-missing \
             --numprocesses=4 --durations=10 \
-            original_autogpt/tests/unit original_autogpt/tests/integration 2>&1 | tee test_output.txt
+            tests/unit tests/integration 2>&1 | tee test_output.txt
 
           test_failure=${PIPESTATUS[0]}
 

.github/workflows/classic-autogpts-ci.yml

@@ -10,9 +10,10 @@ on:
       - '.github/workflows/classic-autogpts-ci.yml'
       - 'classic/original_autogpt/**'
       - 'classic/forge/**'
-      - 'classic/direct_benchmark/**'
-      - 'classic/pyproject.toml'
-      - 'classic/poetry.lock'
+      - 'classic/benchmark/**'
+      - 'classic/run'
+      - 'classic/cli.py'
+      - 'classic/setup.py'
       - '!**/*.md'
   pull_request:
     branches: [ master, dev, release-* ]
@@ -20,9 +21,10 @@ on:
       - '.github/workflows/classic-autogpts-ci.yml'
       - 'classic/original_autogpt/**'
       - 'classic/forge/**'
-      - 'classic/direct_benchmark/**'
-      - 'classic/pyproject.toml'
-      - 'classic/poetry.lock'
+      - 'classic/benchmark/**'
+      - 'classic/run'
+      - 'classic/cli.py'
+      - 'classic/setup.py'
       - '!**/*.md'
 
 defaults:
@@ -33,9 +35,13 @@ defaults:
 jobs:
   serve-agent-protocol:
     runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        agent-name: [ original_autogpt ]
+      fail-fast: false
     timeout-minutes: 20
     env:
-      min-python-version: '3.12'
+      min-python-version: '3.10'
     steps:
       - name: Checkout repository
         uses: actions/checkout@v4
@@ -49,22 +55,22 @@ jobs:
           python-version: ${{ env.min-python-version }}
 
       - name: Install Poetry
+        working-directory: ./classic/${{ matrix.agent-name }}/
         run: |
           curl -sSL https://install.python-poetry.org | python -
 
-      - name: Install dependencies
-        run: poetry install
-
-      - name: Run smoke tests with direct-benchmark
+      - name: Run regression tests
         run: |
-          poetry run direct-benchmark run \
-            --strategies one_shot \
-            --models claude \
-            --tests ReadFile,WriteFile \
-            --json
+          ./run agent start ${{ matrix.agent-name }}
+          cd ${{ matrix.agent-name }}
+          poetry run agbenchmark --mock --test=BasicRetrieval --test=Battleship --test=WebArenaTask_0
+          poetry run agbenchmark --test=WriteFile
         env:
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
+          AGENT_NAME: ${{ matrix.agent-name }}
           REQUESTS_CA_BUNDLE: /etc/ssl/certs/ca-certificates.crt
-          NONINTERACTIVE_MODE: "true"
-          CI: true
+          HELICONE_CACHE_ENABLED: false
+          HELICONE_PROPERTY_AGENT: ${{ matrix.agent-name }}
+          REPORTS_FOLDER: ${{ format('../../reports/{0}', matrix.agent-name) }}
+          TELEMETRY_ENVIRONMENT: autogpt-ci
+          TELEMETRY_OPT_IN: ${{ github.ref_name == 'master' }}

.github/workflows/classic-benchmark-ci.yml

@@ -1,24 +1,18 @@
-name: Classic - Direct Benchmark CI
+name: Classic - AGBenchmark CI
 
 on:
   push:
     branches: [ master, dev, ci-test* ]
     paths:
-      - 'classic/direct_benchmark/**'
-      - 'classic/original_autogpt/**'
-      - 'classic/forge/**'
+      - 'classic/benchmark/**'
+      - '!classic/benchmark/reports/**'
       - .github/workflows/classic-benchmark-ci.yml
-      - 'classic/pyproject.toml'
-      - 'classic/poetry.lock'
   pull_request:
     branches: [ master, dev, release-* ]
     paths:
-      - 'classic/direct_benchmark/**'
-      - 'classic/original_autogpt/**'
-      - 'classic/forge/**'
+      - 'classic/benchmark/**'
+      - '!classic/benchmark/reports/**'
       - .github/workflows/classic-benchmark-ci.yml
-      - 'classic/pyproject.toml'
-      - 'classic/poetry.lock'
 
 concurrency:
   group: ${{ format('benchmark-ci-{0}', github.head_ref && format('{0}-{1}', github.event_name, github.event.pull_request.number) || github.sha) }}
@@ -29,16 +23,23 @@ defaults:
     shell: bash
 
 env:
-  min-python-version: '3.12'
+  min-python-version: '3.10'
 
 jobs:
-  benchmark-tests:
-    runs-on: ubuntu-latest
+  test:
+    permissions:
+      contents: read
     timeout-minutes: 30
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.10"]
+        platform-os: [ubuntu, macos, macos-arm64, windows]
+    runs-on: ${{ matrix.platform-os != 'macos-arm64' && format('{0}-latest', matrix.platform-os) || 'macos-14' }}
     defaults:
       run:
         shell: bash
-        working-directory: classic
+        working-directory: classic/benchmark
     steps:
       - name: Checkout repository
         uses: actions/checkout@v4
@@ -46,88 +47,71 @@ jobs:
           fetch-depth: 0
           submodules: true
 
-      - name: Set up Python ${{ env.min-python-version }}
+      - name: Set up Python ${{ matrix.python-version }}
         uses: actions/setup-python@v5
         with:
-          python-version: ${{ env.min-python-version }}
+          python-version: ${{ matrix.python-version }}
 
       - name: Set up Python dependency cache
+        # On Windows, unpacking cached dependencies takes longer than just installing them
+        if: runner.os != 'Windows'
         uses: actions/cache@v4
         with:
-          path: ~/.cache/pypoetry
-          key: poetry-${{ runner.os }}-${{ hashFiles('classic/poetry.lock') }}
+          path: ${{ runner.os == 'macOS' && '~/Library/Caches/pypoetry' || '~/.cache/pypoetry' }}
+          key: poetry-${{ runner.os }}-${{ hashFiles('classic/benchmark/poetry.lock') }}
 
-      - name: Install Poetry
+      - name: Install Poetry (Unix)
+        if: runner.os != 'Windows'
         run: |
           curl -sSL https://install.python-poetry.org | python3 -
 
-      - name: Install dependencies
+          if [ "${{ runner.os }}" = "macOS" ]; then
+            PATH="$HOME/.local/bin:$PATH"
+            echo "$HOME/.local/bin" >> $GITHUB_PATH
+          fi
+
+      - name: Install Poetry (Windows)
+        if: runner.os == 'Windows'
+        shell: pwsh
+        run: |
+          (Invoke-WebRequest -Uri https://install.python-poetry.org -UseBasicParsing).Content | python -
+
+          $env:PATH += ";$env:APPDATA\Python\Scripts"
+          echo "$env:APPDATA\Python\Scripts" >> $env:GITHUB_PATH
+
+      - name: Install Python dependencies
         run: poetry install
 
-      - name: Run basic benchmark tests
+      - name: Run pytest with coverage
         run: |
-          echo "Testing ReadFile challenge with one_shot strategy..."
-          poetry run direct-benchmark run \
-            --fresh \
-            --strategies one_shot \
-            --models claude \
-            --tests ReadFile \
-            --json
-
-          echo "Testing WriteFile challenge..."
-          poetry run direct-benchmark run \
-            --fresh \
-            --strategies one_shot \
-            --models claude \
-            --tests WriteFile \
-            --json
+          poetry run pytest -vv \
+            --cov=agbenchmark --cov-branch --cov-report term-missing --cov-report xml \
+            --durations=10 \
+            --junitxml=junit.xml -o junit_family=legacy \
+            tests
         env:
           CI: true
-          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-          NONINTERACTIVE_MODE: "true"
 
-      - name: Test category filtering
-        run: |
-          echo "Testing coding category..."
-          poetry run direct-benchmark run \
-            --fresh \
-            --strategies one_shot \
-            --models claude \
-            --categories coding \
-            --tests ReadFile,WriteFile \
-            --json
-        env:
-          CI: true
-          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
-          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-          NONINTERACTIVE_MODE: "true"
+      - name: Upload test results to Codecov
+        if: ${{ !cancelled() }}  # Run even if tests fail
+        uses: codecov/test-results-action@v1
+        with:
+          token: ${{ secrets.CODECOV_TOKEN }}
 
-      - name: Test multiple strategies
-        run: |
-          echo "Testing multiple strategies..."
-          poetry run direct-benchmark run \
-            --fresh \
-            --strategies one_shot,plan_execute \
-            --models claude \
-            --tests ReadFile \
-            --parallel 2 \
-            --json
-        env:
-          CI: true
-          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
-          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-          NONINTERACTIVE_MODE: "true"
+      - name: Upload coverage reports to Codecov
+        uses: codecov/codecov-action@v5
+        with:
+          token: ${{ secrets.CODECOV_TOKEN }}
+          flags: agbenchmark,${{ runner.os }}
 
-  # Run regression tests on maintain challenges
-  regression-tests:
+  self-test-with-agent:
     runs-on: ubuntu-latest
-    timeout-minutes: 45
-    if: github.ref == 'refs/heads/master' || github.ref == 'refs/heads/dev'
-    defaults:
-      run:
-        shell: bash
-        working-directory: classic
+    strategy:
+      matrix:
+        agent-name: [forge]
+      fail-fast: false
+    timeout-minutes: 20
     steps:
       - name: Checkout repository
         uses: actions/checkout@v4
@@ -140,31 +124,53 @@ jobs:
         with:
           python-version: ${{ env.min-python-version }}
 
-      - name: Set up Python dependency cache
-        uses: actions/cache@v4
-        with:
-          path: ~/.cache/pypoetry
-          key: poetry-${{ runner.os }}-${{ hashFiles('classic/poetry.lock') }}
-
       - name: Install Poetry
         run: |
-          curl -sSL https://install.python-poetry.org | python3 -
-
-      - name: Install dependencies
-        run: poetry install
+          curl -sSL https://install.python-poetry.org | python -
 
       - name: Run regression tests
+        working-directory: classic
         run: |
-          echo "Running regression tests (previously beaten challenges)..."
-          poetry run direct-benchmark run \
-            --fresh \
-            --strategies one_shot \
-            --models claude \
-            --maintain \
-            --parallel 4 \
-            --json
+          ./run agent start ${{ matrix.agent-name }}
+          cd ${{ matrix.agent-name }}
+
+          set +e # Ignore non-zero exit codes and continue execution
+          echo "Running the following command: poetry run agbenchmark --maintain --mock"
+          poetry run agbenchmark --maintain --mock
+          EXIT_CODE=$?
+          set -e  # Stop ignoring non-zero exit codes
+          # Check if the exit code was 5, and if so, exit with 0 instead
+          if [ $EXIT_CODE -eq 5 ]; then
+            echo "regression_tests.json is empty."
+          fi
+
+          echo "Running the following command: poetry run agbenchmark --mock"
+          poetry run agbenchmark --mock
+
+          echo "Running the following command: poetry run agbenchmark --mock --category=data"
+          poetry run agbenchmark --mock --category=data
+
+          echo "Running the following command: poetry run agbenchmark --mock --category=coding"
+          poetry run agbenchmark --mock --category=coding
+
+          # echo "Running the following command: poetry run agbenchmark --test=WriteFile"
+          # poetry run agbenchmark --test=WriteFile
+          cd ../benchmark
+          poetry install
+          echo "Adding the BUILD_SKILL_TREE environment variable. This will attempt to add new elements in the skill tree. If new elements are added, the CI fails because they should have been pushed"
+          export BUILD_SKILL_TREE=true
+
+          # poetry run agbenchmark --mock
+
+          # CHANGED=$(git diff --name-only | grep -E '(agbenchmark/challenges)|(../classic/frontend/assets)') || echo "No diffs"
+          # if [ ! -z "$CHANGED" ]; then
+          #   echo "There are unstaged changes please run agbenchmark and commit those changes since they are needed."
+          #   echo "$CHANGED"
+          #   exit 1
+          # else
+          #   echo "No unstaged changes."
+          # fi
         env:
-          CI: true
-          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-          NONINTERACTIVE_MODE: "true"
+          TELEMETRY_ENVIRONMENT: autogpt-benchmark-ci
+          TELEMETRY_OPT_IN: ${{ github.ref_name == 'master' }}

.github/workflows/classic-forge-ci.yml

@@ -6,15 +6,13 @@ on:
     paths:
       - '.github/workflows/classic-forge-ci.yml'
       - 'classic/forge/**'
-      - 'classic/pyproject.toml'
-      - 'classic/poetry.lock'
+      - '!classic/forge/tests/vcr_cassettes'
   pull_request:
     branches: [ master, dev, release-* ]
     paths:
       - '.github/workflows/classic-forge-ci.yml'
       - 'classic/forge/**'
-      - 'classic/pyproject.toml'
-      - 'classic/poetry.lock'
+      - '!classic/forge/tests/vcr_cassettes'
 
 concurrency:
   group: ${{ format('forge-ci-{0}', github.head_ref && format('{0}-{1}', github.event_name, github.event.pull_request.number) || github.sha) }}
@@ -23,60 +21,131 @@ concurrency:
 defaults:
   run:
     shell: bash
-    working-directory: classic
+    working-directory: classic/forge
 
 jobs:
   test:
     permissions:
       contents: read
     timeout-minutes: 30
-    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.10"]
+        platform-os: [ubuntu, macos, macos-arm64, windows]
+    runs-on: ${{ matrix.platform-os != 'macos-arm64' && format('{0}-latest', matrix.platform-os) || 'macos-14' }}
 
     steps:
-      - name: Start MinIO service
+      # Quite slow on macOS (2~4 minutes to set up Docker)
+      # - name: Set up Docker (macOS)
+      #   if: runner.os == 'macOS'
+      #   uses: crazy-max/ghaction-setup-docker@v3
+
+      - name: Start MinIO service (Linux)
+        if: runner.os == 'Linux'
         working-directory: '.'
         run: |
           docker pull minio/minio:edge-cicd
           docker run -d -p 9000:9000 minio/minio:edge-cicd
 
+      - name: Start MinIO service (macOS)
+        if: runner.os == 'macOS'
+        working-directory: ${{ runner.temp }}
+        run: |
+          brew install minio/stable/minio
+          mkdir data
+          minio server ./data &
+
+      # No MinIO on Windows:
+      # - Windows doesn't support running Linux Docker containers
+      # - It doesn't seem possible to start background processes on Windows. They are
+      #   killed after the step returns.
+      #   See: https://github.com/actions/runner/issues/598#issuecomment-2011890429
+
       - name: Checkout repository
         uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          submodules: true
 
-      - name: Set up Python 3.12
+      - name: Checkout cassettes
+        if: ${{ startsWith(github.event_name, 'pull_request') }}
+        env:
+          PR_BASE: ${{ github.event.pull_request.base.ref }}
+          PR_BRANCH: ${{ github.event.pull_request.head.ref }}
+          PR_AUTHOR: ${{ github.event.pull_request.user.login }}
+        run: |
+          cassette_branch="${PR_AUTHOR}-${PR_BRANCH}"
+          cassette_base_branch="${PR_BASE}"
+          cd tests/vcr_cassettes
+
+          if ! git ls-remote --exit-code --heads origin $cassette_base_branch ; then
+            cassette_base_branch="master"
+          fi
+
+          if git ls-remote --exit-code --heads origin $cassette_branch ; then
+            git fetch origin $cassette_branch
+            git fetch origin $cassette_base_branch
+
+            git checkout $cassette_branch
+
+            # Pick non-conflicting cassette updates from the base branch
+            git merge --no-commit --strategy-option=ours origin/$cassette_base_branch
+            echo "Using cassettes from mirror branch '$cassette_branch'," \
+              "synced to upstream branch '$cassette_base_branch'."
+          else
+            git checkout -b $cassette_branch
+            echo "Branch '$cassette_branch' does not exist in cassette submodule." \
+              "Using cassettes from '$cassette_base_branch'."
+          fi
+
+      - name: Set up Python ${{ matrix.python-version }}
         uses: actions/setup-python@v5
         with:
-          python-version: "3.12"
+          python-version: ${{ matrix.python-version }}
 
       - name: Set up Python dependency cache
+        # On Windows, unpacking cached dependencies takes longer than just installing them
+        if: runner.os != 'Windows'
         uses: actions/cache@v4
         with:
-          path: ~/.cache/pypoetry
-          key: poetry-${{ runner.os }}-${{ hashFiles('classic/poetry.lock') }}
+          path: ${{ runner.os == 'macOS' && '~/Library/Caches/pypoetry' || '~/.cache/pypoetry' }}
+          key: poetry-${{ runner.os }}-${{ hashFiles('classic/forge/poetry.lock') }}
 
-      - name: Install Poetry
-        run: curl -sSL https://install.python-poetry.org | python3 -
+      - name: Install Poetry (Unix)
+        if: runner.os != 'Windows'
+        run: |
+          curl -sSL https://install.python-poetry.org | python3 -
+
+          if [ "${{ runner.os }}" = "macOS" ]; then
+            PATH="$HOME/.local/bin:$PATH"
+            echo "$HOME/.local/bin" >> $GITHUB_PATH
+          fi
+
+      - name: Install Poetry (Windows)
+        if: runner.os == 'Windows'
+        shell: pwsh
+        run: |
+          (Invoke-WebRequest -Uri https://install.python-poetry.org -UseBasicParsing).Content | python -
+
+          $env:PATH += ";$env:APPDATA\Python\Scripts"
+          echo "$env:APPDATA\Python\Scripts" >> $env:GITHUB_PATH
 
       - name: Install Python dependencies
         run: poetry install
 
-      - name: Install Playwright browsers
-        run: poetry run playwright install chromium
-
       - name: Run pytest with coverage
         run: |
           poetry run pytest -vv \
             --cov=forge --cov-branch --cov-report term-missing --cov-report xml \
             --durations=10 \
             --junitxml=junit.xml -o junit_family=legacy \
-            forge/forge forge/tests
+            forge
         env:
           CI: true
           PLAIN_OUTPUT: True
-          # API keys - tests that need these will skip if not available
-          # Secrets are not available to fork PRs (GitHub security feature)
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
-          S3_ENDPOINT_URL: http://127.0.0.1:9000
+          S3_ENDPOINT_URL: ${{ runner.os != 'Windows' && 'http://127.0.0.1:9000' || '' }}
           AWS_ACCESS_KEY_ID: minioadmin
           AWS_SECRET_ACCESS_KEY: minioadmin
 
@@ -90,11 +159,85 @@ jobs:
         uses: codecov/codecov-action@v5
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
-          flags: forge
+          flags: forge,${{ runner.os }}
+
+      - id: setup_git_auth
+        name: Set up git token authentication
+        # Cassettes may be pushed even when tests fail
+        if: success() || failure()
+        run: |
+          config_key="http.${{ github.server_url }}/.extraheader"
+          if [ "${{ runner.os }}" = 'macOS' ]; then
+            base64_pat=$(echo -n "pat:${{ secrets.PAT_REVIEW }}" | base64)
+          else
+            base64_pat=$(echo -n "pat:${{ secrets.PAT_REVIEW }}" | base64 -w0)
+          fi
+
+          git config "$config_key" \
+            "Authorization: Basic $base64_pat"
+
+          cd tests/vcr_cassettes
+          git config "$config_key" \
+            "Authorization: Basic $base64_pat"
+
+          echo "config_key=$config_key" >> $GITHUB_OUTPUT
+
+      - id: push_cassettes
+        name: Push updated cassettes
+        # For pull requests, push updated cassettes even when tests fail
+        if: github.event_name == 'push' || (! github.event.pull_request.head.repo.fork && (success() || failure()))
+        env:
+          PR_BRANCH: ${{ github.event.pull_request.head.ref }}
+          PR_AUTHOR: ${{ github.event.pull_request.user.login }}
+        run: |
+          if [ "${{ startsWith(github.event_name, 'pull_request') }}" = "true" ]; then
+            is_pull_request=true
+            cassette_branch="${PR_AUTHOR}-${PR_BRANCH}"
+          else
+            cassette_branch="${{ github.ref_name }}"
+          fi
+
+          cd tests/vcr_cassettes
+          # Commit & push changes to cassettes if any
+          if ! git diff --quiet; then
+            git add .
+            git commit -m "Auto-update cassettes"
+            git push origin HEAD:$cassette_branch
+            if [ ! $is_pull_request ]; then
+              cd ../..
+              git add tests/vcr_cassettes
+              git commit -m "Update cassette submodule"
+              git push origin HEAD:$cassette_branch
+            fi
+            echo "updated=true" >> $GITHUB_OUTPUT
+          else
+            echo "updated=false" >> $GITHUB_OUTPUT
+            echo "No cassette changes to commit"
+          fi
+
+      - name: Post Set up git token auth
+        if: steps.setup_git_auth.outcome == 'success'
+        run: |
+          git config --unset-all '${{ steps.setup_git_auth.outputs.config_key }}'
+          git submodule foreach git config --unset-all '${{ steps.setup_git_auth.outputs.config_key }}'
+
+      - name: Apply "behaviour change" label and comment on PR
+        if: ${{ startsWith(github.event_name, 'pull_request') }}
+        run: |
+          PR_NUMBER="${{ github.event.pull_request.number }}"
+          TOKEN="${{ secrets.PAT_REVIEW }}"
+          REPO="${{ github.repository }}"
+
+          if [[ "${{ steps.push_cassettes.outputs.updated }}" == "true" ]]; then
+            echo "Adding label and comment..."
+            echo $TOKEN | gh auth login --with-token
+            gh issue edit $PR_NUMBER --add-label "behaviour change"
+            gh issue comment $PR_NUMBER --body "You changed AutoGPT's behaviour on ${{ runner.os }}. The cassettes have been updated and will be merged to the submodule when this Pull Request gets merged."
+          fi
 
       - name: Upload logs to artifact
         if: always()
         uses: actions/upload-artifact@v4
         with:
           name: test-logs
-          path: classic/logs/
+          path: classic/forge/logs/

.github/workflows/classic-frontend-ci.yml

@@ -0,0 +1,60 @@
+name: Classic - Frontend CI/CD
+
+on:
+  push:
+    branches:
+      - master
+      - dev
+      - 'ci-test*' # This will match any branch that starts with "ci-test"
+    paths:
+      - 'classic/frontend/**'
+      - '.github/workflows/classic-frontend-ci.yml'
+  pull_request:
+    paths:
+      - 'classic/frontend/**'
+      - '.github/workflows/classic-frontend-ci.yml'
+
+jobs:
+  build:
+    permissions:
+      contents: write
+      pull-requests: write
+    runs-on: ubuntu-latest
+    env:
+      BUILD_BRANCH: ${{ format('classic-frontend-build/{0}', github.ref_name) }}
+
+    steps:
+      - name: Checkout Repo
+        uses: actions/checkout@v4
+
+      - name: Setup Flutter
+        uses: subosito/flutter-action@v2
+        with:
+          flutter-version: '3.13.2'
+
+      - name: Build Flutter to Web
+        run: |
+          cd classic/frontend
+          flutter build web --base-href /app/
+
+      # - name: Commit and Push to ${{ env.BUILD_BRANCH }}
+      #   if: github.event_name == 'push'
+      #   run: |
+      #     git config --local user.email "action@github.com"
+      #     git config --local user.name "GitHub Action"
+      #     git add classic/frontend/build/web
+      #     git checkout -B ${{ env.BUILD_BRANCH }}
+      #     git commit -m "Update frontend build to ${GITHUB_SHA:0:7}" -a
+      #     git push -f origin ${{ env.BUILD_BRANCH }}
+
+      - name: Create PR ${{ env.BUILD_BRANCH }} -> ${{ github.ref_name }}
+        if: github.event_name == 'push'
+        uses: peter-evans/create-pull-request@v8
+        with:
+          add-paths: classic/frontend/build/web
+          base: ${{ github.ref_name }}
+          branch: ${{ env.BUILD_BRANCH }}
+          delete-branch: true
+          title: "Update frontend build in `${{ github.ref_name }}`"
+          body: "This PR updates the frontend build based on commit ${{ github.sha }}."
+          commit-message: "Update frontend build based on commit ${{ github.sha }}"

.github/workflows/classic-python-checks.yml

@@ -7,9 +7,7 @@ on:
       - '.github/workflows/classic-python-checks-ci.yml'
       - 'classic/original_autogpt/**'
       - 'classic/forge/**'
-      - 'classic/direct_benchmark/**'
-      - 'classic/pyproject.toml'
-      - 'classic/poetry.lock'
+      - 'classic/benchmark/**'
       - '**.py'
       - '!classic/forge/tests/vcr_cassettes'
   pull_request:
@@ -18,9 +16,7 @@ on:
       - '.github/workflows/classic-python-checks-ci.yml'
       - 'classic/original_autogpt/**'
       - 'classic/forge/**'
-      - 'classic/direct_benchmark/**'
-      - 'classic/pyproject.toml'
-      - 'classic/poetry.lock'
+      - 'classic/benchmark/**'
       - '**.py'
       - '!classic/forge/tests/vcr_cassettes'
 
@@ -31,13 +27,44 @@ concurrency:
 defaults:
   run:
     shell: bash
-    working-directory: classic
 
 jobs:
+  get-changed-parts:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - id: changes-in
+        name: Determine affected subprojects
+        uses: dorny/paths-filter@v3
+        with:
+          filters: |
+            original_autogpt:
+              - classic/original_autogpt/autogpt/**
+              - classic/original_autogpt/tests/**
+              - classic/original_autogpt/poetry.lock
+            forge:
+              - classic/forge/forge/**
+              - classic/forge/tests/**
+              - classic/forge/poetry.lock
+            benchmark:
+              - classic/benchmark/agbenchmark/**
+              - classic/benchmark/tests/**
+              - classic/benchmark/poetry.lock
+    outputs:
+      changed-parts: ${{ steps.changes-in.outputs.changes }}
+
   lint:
+    needs: get-changed-parts
     runs-on: ubuntu-latest
     env:
-      min-python-version: "3.12"
+      min-python-version: "3.10"
+
+    strategy:
+      matrix:
+        sub-package: ${{ fromJson(needs.get-changed-parts.outputs.changed-parts) }}
+      fail-fast: false
 
     steps:
       - name: Checkout repository
@@ -54,31 +81,42 @@ jobs:
         uses: actions/cache@v4
         with:
           path: ~/.cache/pypoetry
-          key: ${{ runner.os }}-poetry-${{ hashFiles('classic/poetry.lock') }}
+          key: ${{ runner.os }}-poetry-${{ hashFiles(format('{0}/poetry.lock', matrix.sub-package)) }}
 
       - name: Install Poetry
         run: curl -sSL https://install.python-poetry.org | python3 -
 
+      # Install dependencies
+
       - name: Install Python dependencies
-        run: poetry install
+        run: poetry -C classic/${{ matrix.sub-package }} install
 
       # Lint
 
       - name: Lint (isort)
         run: poetry run isort --check .
+        working-directory: classic/${{ matrix.sub-package }}
 
       - name: Lint (Black)
         if: success() || failure()
         run: poetry run black --check .
+        working-directory: classic/${{ matrix.sub-package }}
 
       - name: Lint (Flake8)
         if: success() || failure()
         run: poetry run flake8 .
+        working-directory: classic/${{ matrix.sub-package }}
 
   types:
+    needs: get-changed-parts
     runs-on: ubuntu-latest
     env:
-      min-python-version: "3.12"
+      min-python-version: "3.10"
+
+    strategy:
+      matrix:
+        sub-package: ${{ fromJson(needs.get-changed-parts.outputs.changed-parts) }}
+      fail-fast: false
 
     steps:
       - name: Checkout repository
@@ -95,16 +133,19 @@ jobs:
         uses: actions/cache@v4
         with:
           path: ~/.cache/pypoetry
-          key: ${{ runner.os }}-poetry-${{ hashFiles('classic/poetry.lock') }}
+          key: ${{ runner.os }}-poetry-${{ hashFiles(format('{0}/poetry.lock', matrix.sub-package)) }}
 
       - name: Install Poetry
         run: curl -sSL https://install.python-poetry.org | python3 -
 
+      # Install dependencies
+
       - name: Install Python dependencies
-        run: poetry install
+        run: poetry -C classic/${{ matrix.sub-package }} install
 
       # Typecheck
 
       - name: Typecheck
         if: success() || failure()
         run: poetry run pyright
+        working-directory: classic/${{ matrix.sub-package }}

.github/workflows/platform-backend-ci.yml

@@ -5,14 +5,12 @@ on:
     branches: [master, dev, ci-test*]
     paths:
       - ".github/workflows/platform-backend-ci.yml"
-      - ".github/workflows/scripts/get_package_version_from_lockfile.py"
       - "autogpt_platform/backend/**"
       - "autogpt_platform/autogpt_libs/**"
   pull_request:
     branches: [master, dev, release-*]
     paths:
       - ".github/workflows/platform-backend-ci.yml"
-      - ".github/workflows/scripts/get_package_version_from_lockfile.py"
       - "autogpt_platform/backend/**"
       - "autogpt_platform/autogpt_libs/**"
   merge_group:
@@ -27,91 +25,10 @@ defaults:
     working-directory: autogpt_platform/backend
 
 jobs:
-  lint:
-    permissions:
-      contents: read
-    timeout-minutes: 10
-    runs-on: ubuntu-latest
-
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v6
-
-      - name: Set up Python 3.12
-        uses: actions/setup-python@v5
-        with:
-          python-version: "3.12"
-
-      - name: Set up Python dependency cache
-        uses: actions/cache@v5
-        with:
-          path: ~/.cache/pypoetry
-          key: poetry-${{ runner.os }}-py3.12-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
-
-      - name: Install Poetry
-        run: |
-          HEAD_POETRY_VERSION=$(python ../../.github/workflows/scripts/get_package_version_from_lockfile.py poetry)
-          echo "Using Poetry version ${HEAD_POETRY_VERSION}"
-          curl -sSL https://install.python-poetry.org | POETRY_VERSION=$HEAD_POETRY_VERSION python3 -
-
-      - name: Install Python dependencies
-        run: poetry install
-
-      - name: Run Linters
-        run: poetry run lint --skip-pyright
-
-    env:
-      CI: true
-      PLAIN_OUTPUT: True
-
-  type-check:
-    permissions:
-      contents: read
-    timeout-minutes: 10
-    strategy:
-      fail-fast: false
-      matrix:
-        python-version: ["3.11", "3.12", "3.13"]
-    runs-on: ubuntu-latest
-
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v6
-
-      - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v5
-        with:
-          python-version: ${{ matrix.python-version }}
-
-      - name: Set up Python dependency cache
-        uses: actions/cache@v5
-        with:
-          path: ~/.cache/pypoetry
-          key: poetry-${{ runner.os }}-py${{ matrix.python-version }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
-
-      - name: Install Poetry
-        run: |
-          HEAD_POETRY_VERSION=$(python ../../.github/workflows/scripts/get_package_version_from_lockfile.py poetry)
-          echo "Using Poetry version ${HEAD_POETRY_VERSION}"
-          curl -sSL https://install.python-poetry.org | POETRY_VERSION=$HEAD_POETRY_VERSION python3 -
-
-      - name: Install Python dependencies
-        run: poetry install
-
-      - name: Generate Prisma Client
-        run: poetry run prisma generate && poetry run gen-prisma-stub
-
-      - name: Run Pyright
-        run: poetry run pyright --pythonversion ${{ matrix.python-version }}
-
-    env:
-      CI: true
-      PLAIN_OUTPUT: True
-
   test:
     permissions:
       contents: read
-    timeout-minutes: 15
+    timeout-minutes: 30
     strategy:
       fail-fast: false
       matrix:
@@ -179,9 +96,9 @@ jobs:
         uses: actions/cache@v5
         with:
           path: ~/.cache/pypoetry
-          key: poetry-${{ runner.os }}-py${{ matrix.python-version }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
+          key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
 
-      - name: Install Poetry
+      - name: Install Poetry (Unix)
         run: |
           # Extract Poetry version from backend/poetry.lock
           HEAD_POETRY_VERSION=$(python ../../.github/workflows/scripts/get_package_version_from_lockfile.py poetry)
@@ -239,22 +156,22 @@ jobs:
           echo "Waiting for ClamAV daemon to start..."
           max_attempts=60
           attempt=0
-
+          
           until nc -z localhost 3310 || [ $attempt -eq $max_attempts ]; do
             echo "ClamAV is unavailable - sleeping (attempt $((attempt+1))/$max_attempts)"
             sleep 5
             attempt=$((attempt+1))
           done
-
+          
           if [ $attempt -eq $max_attempts ]; then
             echo "ClamAV failed to start after $((max_attempts*5)) seconds"
             echo "Checking ClamAV service logs..."
             docker logs $(docker ps -q --filter "ancestor=clamav/clamav-debian:latest") 2>&1 | tail -50 || echo "No ClamAV container found"
             exit 1
           fi
-
+          
           echo "ClamAV is ready!"
-
+          
           # Verify ClamAV is responsive
           echo "Testing ClamAV connection..."
           timeout 10 bash -c 'echo "PING" | nc localhost 3310' || {
@@ -269,15 +186,18 @@ jobs:
           DATABASE_URL: ${{ steps.supabase.outputs.DB_URL }}
           DIRECT_URL: ${{ steps.supabase.outputs.DB_URL }}
 
+      - id: lint
+        name: Run Linter
+        run: poetry run lint
+
       - name: Run pytest with coverage
         run: |
           if [[ "${{ runner.debug }}" == "1" ]]; then
-            poetry run pytest -s -vv -o log_cli=true -o log_cli_level=DEBUG \
-              --cov=backend --cov-branch --cov-report term-missing --cov-report xml
+            poetry run pytest -s -vv -o log_cli=true -o log_cli_level=DEBUG
           else
-            poetry run pytest -s -vv \
-              --cov=backend --cov-branch --cov-report term-missing --cov-report xml
+            poetry run pytest -s -vv
           fi
+        if: success() || (failure() && steps.lint.outcome == 'failure')
         env:
           LOG_LEVEL: ${{ runner.debug && 'DEBUG' || 'INFO' }}
           DATABASE_URL: ${{ steps.supabase.outputs.DB_URL }}
@@ -289,14 +209,6 @@ jobs:
           REDIS_PORT: "6379"
           ENCRYPTION_KEY: "dvziYgz0KSK8FENhju0ZYi8-fRTfAdlz6YLhdB_jhNw=" # DO NOT USE IN PRODUCTION!!
 
-      - name: Upload coverage reports to Codecov
-        if: ${{ !cancelled() }}
-        uses: codecov/codecov-action@v5
-        with:
-          token: ${{ secrets.CODECOV_TOKEN }}
-          flags: platform-backend
-          files: ./autogpt_platform/backend/coverage.xml
-
     env:
       CI: true
       PLAIN_OUTPUT: True
@@ -310,3 +222,9 @@ jobs:
       # the backend service, docker composes, and examples
       RABBITMQ_DEFAULT_USER: "rabbitmq_user_default"
       RABBITMQ_DEFAULT_PASS: "k0VMxyIJF9S35f3x2uaw5IWAl6Y536O7"
+
+      # - name: Upload coverage reports to Codecov
+      #   uses: codecov/codecov-action@v4
+      #   with:
+      #     token: ${{ secrets.CODECOV_TOKEN }}
+      #     flags: backend,${{ runner.os }}

.github/workflows/platform-frontend-ci.yml

@@ -120,6 +120,175 @@ jobs:
           token: ${{ secrets.GITHUB_TOKEN }}
           exitOnceUploaded: true
 
+  e2e_test:
+    name: end-to-end tests
+    runs-on: big-boi
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v6
+        with:
+          submodules: recursive
+
+      - name: Set up Platform - Copy default supabase .env
+        run: |
+          cp ../.env.default ../.env
+
+      - name: Set up Platform - Copy backend .env and set OpenAI API key
+        run: |
+          cp ../backend/.env.default ../backend/.env
+          echo "OPENAI_INTERNAL_API_KEY=${{ secrets.OPENAI_API_KEY }}" >> ../backend/.env
+        env:
+          # Used by E2E test data script to generate embeddings for approved store agents
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+
+      - name: Set up Platform - Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+        with:
+          driver: docker-container
+          driver-opts: network=host
+
+      - name: Set up Platform - Expose GHA cache to docker buildx CLI
+        uses: crazy-max/ghaction-github-runtime@v3
+
+      - name: Set up Platform - Build Docker images (with cache)
+        working-directory: autogpt_platform
+        run: |
+          pip install pyyaml
+
+          # Resolve extends and generate a flat compose file that bake can understand
+          docker compose -f docker-compose.yml config > docker-compose.resolved.yml
+
+          # Add cache configuration to the resolved compose file
+          python ../.github/workflows/scripts/docker-ci-fix-compose-build-cache.py \
+            --source docker-compose.resolved.yml \
+            --cache-from "type=gha" \
+            --cache-to "type=gha,mode=max" \
+            --backend-hash "${{ hashFiles('autogpt_platform/backend/Dockerfile', 'autogpt_platform/backend/poetry.lock', 'autogpt_platform/backend/backend') }}" \
+            --frontend-hash "${{ hashFiles('autogpt_platform/frontend/Dockerfile', 'autogpt_platform/frontend/pnpm-lock.yaml', 'autogpt_platform/frontend/src') }}" \
+            --git-ref "${{ github.ref }}"
+
+          # Build with bake using the resolved compose file (now includes cache config)
+          docker buildx bake --allow=fs.read=.. -f docker-compose.resolved.yml --load
+        env:
+          NEXT_PUBLIC_PW_TEST: true
+
+      - name: Set up tests - Cache E2E test data
+        id: e2e-data-cache
+        uses: actions/cache@v5
+        with:
+          path: /tmp/e2e_test_data.sql
+          key: e2e-test-data-${{ hashFiles('autogpt_platform/backend/test/e2e_test_data.py', 'autogpt_platform/backend/migrations/**', '.github/workflows/platform-frontend-ci.yml') }}
+
+      - name: Set up Platform - Start Supabase DB + Auth
+        run: |
+          docker compose -f ../docker-compose.resolved.yml up -d db auth --no-build
+          echo "Waiting for database to be ready..."
+          timeout 60 sh -c 'until docker compose -f ../docker-compose.resolved.yml exec -T db pg_isready -U postgres 2>/dev/null; do sleep 2; done'
+          echo "Waiting for auth service to be ready..."
+          timeout 60 sh -c 'until docker compose -f ../docker-compose.resolved.yml exec -T db psql -U postgres -d postgres -c "SELECT 1 FROM auth.users LIMIT 1" 2>/dev/null; do sleep 2; done' || echo "Auth schema check timeout, continuing..."
+
+      - name: Set up Platform - Run migrations
+        run: |
+          echo "Running migrations..."
+          docker compose -f ../docker-compose.resolved.yml run --rm migrate
+          echo "✅ Migrations completed"
+        env:
+          NEXT_PUBLIC_PW_TEST: true
+
+      - name: Set up tests - Load cached E2E test data
+        if: steps.e2e-data-cache.outputs.cache-hit == 'true'
+        run: |
+          echo "✅ Found cached E2E test data, restoring..."
+          {
+            echo "SET session_replication_role = 'replica';"
+            cat /tmp/e2e_test_data.sql
+            echo "SET session_replication_role = 'origin';"
+          } | docker compose -f ../docker-compose.resolved.yml exec -T db psql -U postgres -d postgres -b
+          # Refresh materialized views after restore
+          docker compose -f ../docker-compose.resolved.yml exec -T db \
+            psql -U postgres -d postgres -b -c "SET search_path TO platform; SELECT refresh_store_materialized_views();" || true
+
+          echo "✅ E2E test data restored from cache"
+
+      - name: Set up Platform - Start (all other services)
+        run: |
+          docker compose -f ../docker-compose.resolved.yml up -d --no-build
+          echo "Waiting for rest_server to be ready..."
+          timeout 60 sh -c 'until curl -f http://localhost:8006/health 2>/dev/null; do sleep 2; done' || echo "Rest server health check timeout, continuing..."
+        env:
+          NEXT_PUBLIC_PW_TEST: true
+
+      - name: Set up tests - Create E2E test data
+        if: steps.e2e-data-cache.outputs.cache-hit != 'true'
+        run: |
+          echo "Creating E2E test data..."
+          docker cp ../backend/test/e2e_test_data.py $(docker compose -f ../docker-compose.resolved.yml ps -q rest_server):/tmp/e2e_test_data.py
+          docker compose -f ../docker-compose.resolved.yml exec -T rest_server sh -c "cd /app/autogpt_platform && python /tmp/e2e_test_data.py" || {
+            echo "❌ E2E test data creation failed!"
+            docker compose -f ../docker-compose.resolved.yml logs --tail=50 rest_server
+            exit 1
+          }
+
+          # Dump auth.users + platform schema for cache (two separate dumps)
+          echo "Dumping database for cache..."
+          {
+            docker compose -f ../docker-compose.resolved.yml exec -T db \
+              pg_dump -U postgres --data-only --column-inserts \
+              --table='auth.users' postgres
+            docker compose -f ../docker-compose.resolved.yml exec -T db \
+              pg_dump -U postgres --data-only --column-inserts \
+              --schema=platform \
+              --exclude-table='platform._prisma_migrations' \
+              --exclude-table='platform.apscheduler_jobs' \
+              --exclude-table='platform.apscheduler_jobs_batched_notifications' \
+              postgres
+          } > /tmp/e2e_test_data.sql
+
+          echo "✅ Database dump created for caching ($(wc -l < /tmp/e2e_test_data.sql) lines)"
+
+      - name: Set up tests - Enable corepack
+        run: corepack enable
+
+      - name: Set up tests - Set up Node
+        uses: actions/setup-node@v6
+        with:
+          node-version: "22.18.0"
+          cache: "pnpm"
+          cache-dependency-path: autogpt_platform/frontend/pnpm-lock.yaml
+
+      - name: Set up tests - Install dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Set up tests - Install browser 'chromium'
+        run: pnpm playwright install --with-deps chromium
+
+      - name: Run Playwright tests
+        run: pnpm test:no-build
+        continue-on-error: false
+
+      - name: Upload Playwright report
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: playwright-report
+          path: playwright-report
+          if-no-files-found: ignore
+          retention-days: 3
+
+      - name: Upload Playwright test results
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: playwright-test-results
+          path: test-results
+          if-no-files-found: ignore
+          retention-days: 3
+
+      - name: Print Final Docker Compose logs
+        if: always()
+        run: docker compose -f ../docker-compose.resolved.yml logs
+
   integration_test:
     runs-on: ubuntu-latest
     needs: setup
@@ -148,11 +317,3 @@ jobs:
 
       - name: Run Integration Tests
         run: pnpm test:unit
-
-      - name: Upload coverage reports to Codecov
-        if: ${{ !cancelled() }}
-        uses: codecov/codecov-action@v5
-        with:
-          token: ${{ secrets.CODECOV_TOKEN }}
-          flags: platform-frontend
-          files: ./autogpt_platform/frontend/coverage/cobertura-coverage.xml

.github/workflows/platform-fullstack-ci.yml

@@ -1,18 +1,14 @@
-name: AutoGPT Platform - Full-stack CI
+name: AutoGPT Platform - Frontend CI
 
 on:
   push:
     branches: [master, dev]
     paths:
       - ".github/workflows/platform-fullstack-ci.yml"
-      - ".github/workflows/scripts/docker-ci-fix-compose-build-cache.py"
-      - ".github/workflows/scripts/get_package_version_from_lockfile.py"
       - "autogpt_platform/**"
   pull_request:
     paths:
       - ".github/workflows/platform-fullstack-ci.yml"
-      - ".github/workflows/scripts/docker-ci-fix-compose-build-cache.py"
-      - ".github/workflows/scripts/get_package_version_from_lockfile.py"
       - "autogpt_platform/**"
   merge_group:
 
@@ -28,28 +24,42 @@ defaults:
 jobs:
   setup:
     runs-on: ubuntu-latest
+    outputs:
+      cache-key: ${{ steps.cache-key.outputs.key }}
 
     steps:
       - name: Checkout repository
         uses: actions/checkout@v6
 
-      - name: Enable corepack
-        run: corepack enable
-
-      - name: Set up Node
+      - name: Set up Node.js
         uses: actions/setup-node@v6
         with:
           node-version: "22.18.0"
-          cache: "pnpm"
-          cache-dependency-path: autogpt_platform/frontend/pnpm-lock.yaml
 
-      - name: Install dependencies to populate cache
+      - name: Enable corepack
+        run: corepack enable
+
+      - name: Generate cache key
+        id: cache-key
+        run: echo "key=${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml', 'autogpt_platform/frontend/package.json') }}" >> $GITHUB_OUTPUT
+
+      - name: Cache dependencies
+        uses: actions/cache@v5
+        with:
+          path: ~/.pnpm-store
+          key: ${{ steps.cache-key.outputs.key }}
+          restore-keys: |
+            ${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml') }}
+            ${{ runner.os }}-pnpm-
+
+      - name: Install dependencies
         run: pnpm install --frozen-lockfile
 
-  check-api-types:
-    name: check API types
-    runs-on: ubuntu-latest
+  types:
+    runs-on: big-boi
     needs: setup
+    strategy:
+      fail-fast: false
 
     steps:
       - name: Checkout repository
@@ -57,288 +67,70 @@ jobs:
         with:
           submodules: recursive
 
-      # ------------------------ Backend setup ------------------------
-
-      - name: Set up Backend - Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: "3.12"
-
-      - name: Set up Backend - Install Poetry
-        working-directory: autogpt_platform/backend
-        run: |
-          POETRY_VERSION=$(python ../../.github/workflows/scripts/get_package_version_from_lockfile.py poetry)
-          echo "Installing Poetry version ${POETRY_VERSION}"
-          curl -sSL https://install.python-poetry.org | POETRY_VERSION=$POETRY_VERSION python3 -
-
-      - name: Set up Backend - Set up dependency cache
-        uses: actions/cache@v5
-        with:
-          path: ~/.cache/pypoetry
-          key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
-
-      - name: Set up Backend - Install dependencies
-        working-directory: autogpt_platform/backend
-        run: poetry install
-
-      - name: Set up Backend - Generate Prisma client
-        working-directory: autogpt_platform/backend
-        run: poetry run prisma generate && poetry run gen-prisma-stub
-
-      - name: Set up Frontend - Export OpenAPI schema from Backend
-        working-directory: autogpt_platform/backend
-        run: poetry run export-api-schema --output ../frontend/src/app/api/openapi.json
-
-      # ------------------------ Frontend setup ------------------------
-
-      - name: Set up Frontend - Enable corepack
-        run: corepack enable
-
-      - name: Set up Frontend - Set up Node
+      - name: Set up Node.js
         uses: actions/setup-node@v6
         with:
           node-version: "22.18.0"
-          cache: "pnpm"
-          cache-dependency-path: autogpt_platform/frontend/pnpm-lock.yaml
 
-      - name: Set up Frontend - Install dependencies
+      - name: Enable corepack
+        run: corepack enable
+
+      - name: Copy default supabase .env
+        run: |
+          cp ../.env.default ../.env
+
+      - name: Copy backend .env
+        run: |
+          cp ../backend/.env.default ../backend/.env
+
+      - name: Run docker compose
+        run: |
+          docker compose -f ../docker-compose.yml --profile local up -d deps_backend
+
+      - name: Restore dependencies cache
+        uses: actions/cache@v5
+        with:
+          path: ~/.pnpm-store
+          key: ${{ needs.setup.outputs.cache-key }}
+          restore-keys: |
+            ${{ runner.os }}-pnpm-
+
+      - name: Install dependencies
         run: pnpm install --frozen-lockfile
 
-      - name: Set up Frontend - Format OpenAPI schema
-        id: format-schema
-        run: pnpm prettier --write ./src/app/api/openapi.json
+      - name: Setup .env
+        run: cp .env.default .env
+
+      - name: Wait for services to be ready
+        run: |
+          echo "Waiting for rest_server to be ready..."
+          timeout 60 sh -c 'until curl -f http://localhost:8006/health 2>/dev/null; do sleep 2; done' || echo "Rest server health check timeout, continuing..."
+          echo "Waiting for database to be ready..."
+          timeout 60 sh -c 'until docker compose -f ../docker-compose.yml exec -T db pg_isready -U postgres 2>/dev/null; do sleep 2; done' || echo "Database ready check timeout, continuing..."
+
+      - name: Generate API queries
+        run: pnpm generate:api:force
 
       - name: Check for API schema changes
         run: |
           if ! git diff --exit-code src/app/api/openapi.json; then
             echo "❌ API schema changes detected in src/app/api/openapi.json"
             echo ""
-            echo "The openapi.json file has been modified after exporting the API schema."
+            echo "The openapi.json file has been modified after running 'pnpm generate:api-all'."
             echo "This usually means changes have been made in the BE endpoints without updating the Frontend."
             echo "The API schema is now out of sync with the Front-end queries."
             echo ""
             echo "To fix this:"
-            echo "\nIn the backend directory:"
-            echo "1. Run 'poetry run export-api-schema --output ../frontend/src/app/api/openapi.json'"
-            echo "\nIn the frontend directory:"
-            echo "2. Run 'pnpm prettier --write src/app/api/openapi.json'"
-            echo "3. Run 'pnpm generate:api'"
-            echo "4. Run 'pnpm types'"
-            echo "5. Fix any TypeScript errors that may have been introduced"
-            echo "6. Commit and push your changes"
+            echo "1. Pull the backend 'docker compose pull && docker compose up -d --build --force-recreate'"
+            echo "2. Run 'pnpm generate:api' locally"
+            echo "3. Run 'pnpm types' locally"
+            echo "4. Fix any TypeScript errors that may have been introduced"
+            echo "5. Commit and push your changes"
             echo ""
             exit 1
           else
             echo "✅ No API schema changes detected"
           fi
 
-      - name: Set up Frontend - Generate API client
-        id: generate-api-client
-        run: pnpm orval --config ./orval.config.ts
-        # Continue with type generation & check even if there are schema changes
-        if: success() || (steps.format-schema.outcome == 'success')
-
-      - name: Check for TypeScript errors
+      - name: Run Typescript checks
         run: pnpm types
-        if: success() || (steps.generate-api-client.outcome == 'success')
-
-  e2e_test:
-    name: end-to-end tests
-    runs-on: big-boi
-
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v6
-        with:
-          submodules: recursive
-
-      - name: Set up Platform - Copy default supabase .env
-        run: |
-          cp ../.env.default ../.env
-
-      - name: Set up Platform - Copy backend .env and set OpenAI API key
-        run: |
-          cp ../backend/.env.default ../backend/.env
-          echo "OPENAI_INTERNAL_API_KEY=${{ secrets.OPENAI_API_KEY }}" >> ../backend/.env
-          echo "SCHEDULER_STARTUP_EMBEDDING_BACKFILL=false" >> ../backend/.env
-        env:
-          # Used by E2E test data script to generate embeddings for approved store agents
-          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-
-      - name: Set up Platform - Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-        with:
-          driver: docker-container
-          driver-opts: network=host
-
-      - name: Set up Platform - Expose GHA cache to docker buildx CLI
-        uses: crazy-max/ghaction-github-runtime@v4
-
-      - name: Set up Platform - Build Docker images (with cache)
-        working-directory: autogpt_platform
-        run: |
-          pip install pyyaml
-
-          # Resolve extends and generate a flat compose file that bake can understand
-          export NEXT_PUBLIC_SOURCEMAPS NEXT_PUBLIC_PW_TEST
-          docker compose -f docker-compose.yml config > docker-compose.resolved.yml
-
-          # Ensure NEXT_PUBLIC_SOURCEMAPS is in resolved compose
-          # (docker compose config on some versions drops this arg)
-          if ! grep -q "NEXT_PUBLIC_SOURCEMAPS" docker-compose.resolved.yml; then
-            echo "Injecting NEXT_PUBLIC_SOURCEMAPS into resolved compose (docker compose config dropped it)"
-            sed -i '/NEXT_PUBLIC_PW_TEST/a\        NEXT_PUBLIC_SOURCEMAPS: "true"' docker-compose.resolved.yml
-          fi
-
-          # Add cache configuration to the resolved compose file
-          python ../.github/workflows/scripts/docker-ci-fix-compose-build-cache.py \
-            --source docker-compose.resolved.yml \
-            --cache-from "type=gha" \
-            --cache-to "type=gha,mode=max" \
-            --backend-hash "${{ hashFiles('autogpt_platform/backend/Dockerfile', 'autogpt_platform/backend/poetry.lock', 'autogpt_platform/backend/backend/**') }}" \
-            --frontend-hash "${{ hashFiles('autogpt_platform/frontend/Dockerfile', 'autogpt_platform/frontend/pnpm-lock.yaml', 'autogpt_platform/frontend/src/**') }}-sourcemaps" \
-            --git-ref "${{ github.ref }}"
-
-          # Build with bake using the resolved compose file (now includes cache config)
-          docker buildx bake --allow=fs.read=.. -f docker-compose.resolved.yml --load
-        env:
-          NEXT_PUBLIC_PW_TEST: true
-          NEXT_PUBLIC_SOURCEMAPS: true
-
-      - name: Set up tests - Cache E2E test data
-        id: e2e-data-cache
-        uses: actions/cache@v5
-        with:
-          path: /tmp/e2e_test_data.sql
-          key: e2e-test-data-${{ hashFiles('autogpt_platform/backend/test/e2e_test_data.py', 'autogpt_platform/backend/migrations/**', '.github/workflows/platform-fullstack-ci.yml') }}
-
-      - name: Set up Platform - Start Supabase DB + Auth
-        run: |
-          docker compose -f ../docker-compose.resolved.yml up -d db auth --no-build
-          echo "Waiting for database to be ready..."
-          timeout 60 sh -c 'until docker compose -f ../docker-compose.resolved.yml exec -T db pg_isready -U postgres 2>/dev/null; do sleep 2; done'
-          echo "Waiting for auth service to be ready..."
-          timeout 60 sh -c 'until docker compose -f ../docker-compose.resolved.yml exec -T db psql -U postgres -d postgres -c "SELECT 1 FROM auth.users LIMIT 1" 2>/dev/null; do sleep 2; done' || echo "Auth schema check timeout, continuing..."
-
-      - name: Set up Platform - Run migrations
-        run: |
-          echo "Running migrations..."
-          docker compose -f ../docker-compose.resolved.yml run --rm migrate
-          echo "✅ Migrations completed"
-        env:
-          NEXT_PUBLIC_PW_TEST: true
-
-      - name: Set up tests - Load cached E2E test data
-        if: steps.e2e-data-cache.outputs.cache-hit == 'true'
-        run: |
-          echo "✅ Found cached E2E test data, restoring..."
-          {
-            echo "SET session_replication_role = 'replica';"
-            cat /tmp/e2e_test_data.sql
-            echo "SET session_replication_role = 'origin';"
-          } | docker compose -f ../docker-compose.resolved.yml exec -T db psql -U postgres -d postgres -b
-          # Refresh materialized views after restore
-          docker compose -f ../docker-compose.resolved.yml exec -T db \
-            psql -U postgres -d postgres -b -c "SET search_path TO platform; SELECT refresh_store_materialized_views();" || true
-
-          echo "✅ E2E test data restored from cache"
-
-      - name: Set up Platform - Start (all other services)
-        run: |
-          docker compose -f ../docker-compose.resolved.yml up -d --no-build
-          echo "Waiting for rest_server to be ready..."
-          timeout 60 sh -c 'until curl -f http://localhost:8006/health 2>/dev/null; do sleep 2; done' || echo "Rest server health check timeout, continuing..."
-        env:
-          NEXT_PUBLIC_PW_TEST: true
-
-      - name: Set up tests - Create E2E test data
-        if: steps.e2e-data-cache.outputs.cache-hit != 'true'
-        run: |
-          echo "Creating E2E test data..."
-          docker cp ../backend/test/e2e_test_data.py $(docker compose -f ../docker-compose.resolved.yml ps -q rest_server):/tmp/e2e_test_data.py
-          docker compose -f ../docker-compose.resolved.yml exec -T rest_server sh -c "cd /app/autogpt_platform && python /tmp/e2e_test_data.py" || {
-            echo "❌ E2E test data creation failed!"
-            docker compose -f ../docker-compose.resolved.yml logs --tail=50 rest_server
-            exit 1
-          }
-
-          # Dump auth.users + platform schema for cache (two separate dumps)
-          echo "Dumping database for cache..."
-          {
-            docker compose -f ../docker-compose.resolved.yml exec -T db \
-              pg_dump -U postgres --data-only --column-inserts \
-              --table='auth.users' postgres
-            docker compose -f ../docker-compose.resolved.yml exec -T db \
-              pg_dump -U postgres --data-only --column-inserts \
-              --schema=platform \
-              --exclude-table='platform._prisma_migrations' \
-              --exclude-table='platform.apscheduler_jobs' \
-              --exclude-table='platform.apscheduler_jobs_batched_notifications' \
-              postgres
-          } > /tmp/e2e_test_data.sql
-
-          echo "✅ Database dump created for caching ($(wc -l < /tmp/e2e_test_data.sql) lines)"
-
-      - name: Set up tests - Enable corepack
-        run: corepack enable
-
-      - name: Set up tests - Set up Node
-        uses: actions/setup-node@v6
-        with:
-          node-version: "22.18.0"
-          cache: "pnpm"
-          cache-dependency-path: autogpt_platform/frontend/pnpm-lock.yaml
-
-      - name: Set up tests - Cache Playwright browsers
-        uses: actions/cache@v5
-        with:
-          path: ~/.cache/ms-playwright
-          key: playwright-${{ runner.os }}-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml') }}
-          restore-keys: |
-            playwright-${{ runner.os }}-
-
-      - name: Copy source maps from Docker for E2E coverage
-        run: |
-          FRONTEND_CONTAINER=$(docker compose -f ../docker-compose.resolved.yml ps -q frontend)
-          docker cp "$FRONTEND_CONTAINER":/app/.next/static .next-static-coverage
-
-      - name: Set up tests - Install dependencies
-        run: pnpm install --frozen-lockfile
-
-      - name: Set up tests - Install browser 'chromium'
-        run: pnpm playwright install --with-deps chromium
-
-      - name: Run Playwright E2E suite
-        run: pnpm test:e2e:no-build
-        continue-on-error: false
-
-      - name: Upload E2E coverage to Codecov
-        if: ${{ !cancelled() }}
-        uses: codecov/codecov-action@v5
-        with:
-          token: ${{ secrets.CODECOV_TOKEN }}
-          flags: platform-frontend-e2e
-          files: ./autogpt_platform/frontend/coverage/e2e/cobertura-coverage.xml
-          disable_search: true
-
-      - name: Upload Playwright report
-        if: always()
-        uses: actions/upload-artifact@v4
-        with:
-          name: playwright-report
-          path: autogpt_platform/frontend/playwright-report
-          if-no-files-found: ignore
-          retention-days: 3
-
-      - name: Upload Playwright test results
-        if: always()
-        uses: actions/upload-artifact@v4
-        with:
-          name: playwright-test-results
-          path: autogpt_platform/frontend/test-results
-          if-no-files-found: ignore
-          retention-days: 3
-
-      - name: Print Final Docker Compose logs
-        if: always()
-        run: docker compose -f ../docker-compose.resolved.yml logs

.gitignore

@@ -3,7 +3,6 @@
 classic/original_autogpt/keys.py
 classic/original_autogpt/*.json
 auto_gpt_workspace/*
-.autogpt/
 *.mpeg
 .env
 # Root .env files
@@ -17,7 +16,6 @@ log-ingestion.txt
 /logs
 *.log
 *.mp3
-!autogpt_platform/frontend/public/notification.mp3
 mem.sqlite3
 venvAutoGPT
 
@@ -161,10 +159,6 @@ CURRENT_BULLETIN.md
 
 # AgBenchmark
 classic/benchmark/agbenchmark/reports/
-classic/reports/
-classic/direct_benchmark/reports/
-classic/.benchmark_workspaces/
-classic/direct_benchmark/.benchmark_workspaces/
 
 # Nodejs
 package-lock.json
@@ -183,16 +177,7 @@ autogpt_platform/backend/settings.py
 
 *.ign.*
 .test-contents
-**/.claude/settings.local.json
 .claude/settings.local.json
 CLAUDE.local.md
 /autogpt_platform/backend/logs
-/autogpt_platform/backend/poetry.toml
-
-# Test database
-test.db
-.next
-# Implementation plans (generated by AI agents)
-plans/
-.claude/worktrees/
-test-results/
+.next

.gitleaks.toml

@@ -1,36 +0,0 @@
-title = "AutoGPT Gitleaks Config"
-
-[extend]
-useDefault = true
-
-[allowlist]
-description = "Global allowlist"
-paths = [
-    # Template/example env files (no real secrets)
-    '''\.env\.(default|example|template)$''',
-    # Lock files
-    '''pnpm-lock\.yaml$''',
-    '''poetry\.lock$''',
-    # Secrets baseline
-    '''\.secrets\.baseline$''',
-    # Build artifacts and caches (should not be committed)
-    '''__pycache__/''',
-    '''classic/frontend/build/''',
-    # Docker dev setup (local dev JWTs/keys only)
-    '''autogpt_platform/db/docker/''',
-    # Load test configs (dev JWTs)
-    '''load-tests/configs/''',
-    # Test files with fake/fixture keys (_test.py, test_*.py, conftest.py)
-    '''(_test|test_.*|conftest)\.py$''',
-    # Documentation (only contains placeholder keys in curl/API examples)
-    '''docs/.*\.md$''',
-    # Firebase config (public API keys by design)
-    '''google-services\.json$''',
-    '''classic/frontend/(lib|web)/''',
-]
-# CI test-only encryption key (marked DO NOT USE IN PRODUCTION)
-regexes = [
-    '''dvziYgz0KSK8FENhju0ZYi8''',
-    # LLM model name enum values falsely flagged as API keys
-    '''Llama-\d.*Instruct''',
-]

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "classic/forge/tests/vcr_cassettes"]
+	path = classic/forge/tests/vcr_cassettes
+	url = https://github.com/Significant-Gravitas/Auto-GPT-test-cassettes

.nvmrc

@@ -1 +0,0 @@
-22

.pre-commit-config.yaml

@@ -1,10 +1,3 @@
-default_install_hook_types:
-  - pre-commit
-  - pre-push
-  - post-checkout
-
-default_stages: [pre-commit]
-
 repos:
   - repo: https://github.com/pre-commit/pre-commit-hooks
     rev: v4.4.0
@@ -23,15 +16,8 @@ repos:
       - id: detect-secrets
         name: Detect secrets
         description: Detects high entropy strings that are likely to be passwords.
-        args: ["--baseline", ".secrets.baseline"]
         files: ^autogpt_platform/
-        exclude: (pnpm-lock\.yaml|\.env\.(default|example|template))$
-
-  - repo: https://github.com/gitleaks/gitleaks
-    rev: v8.24.3
-    hooks:
-      - id: gitleaks
-        name: Detect secrets (gitleaks)
+        stages: [pre-push]
 
   - repo: local
     # For proper type checking, all dependencies need to be up-to-date.
@@ -40,71 +26,49 @@ repos:
       - id: poetry-install
         name: Check & Install dependencies - AutoGPT Platform - Backend
         alias: poetry-install-platform-backend
+        entry: poetry -C autogpt_platform/backend install
         # include autogpt_libs source (since it's a path dependency)
-        entry: >
-          bash -c '
-          if [ -n "$PRE_COMMIT_FROM_REF" ]; then
-            git diff --name-only "$PRE_COMMIT_FROM_REF" "$PRE_COMMIT_TO_REF"
-          else
-            git diff --cached --name-only
-          fi | grep -qE "^autogpt_platform/(backend|autogpt_libs)/poetry\.lock$" || exit 0;
-          poetry -C autogpt_platform/backend install
-          '
-        always_run: true
+        files: ^autogpt_platform/(backend|autogpt_libs)/poetry\.lock$
+        types: [file]
         language: system
         pass_filenames: false
-        stages: [pre-commit, post-checkout]
 
       - id: poetry-install
         name: Check & Install dependencies - AutoGPT Platform - Libs
         alias: poetry-install-platform-libs
-        entry: >
-          bash -c '
-          if [ -n "$PRE_COMMIT_FROM_REF" ]; then
-            git diff --name-only "$PRE_COMMIT_FROM_REF" "$PRE_COMMIT_TO_REF"
-          else
-            git diff --cached --name-only
-          fi | grep -qE "^autogpt_platform/autogpt_libs/poetry\.lock$" || exit 0;
-          poetry -C autogpt_platform/autogpt_libs install
-          '
-        always_run: true
+        entry: poetry -C autogpt_platform/autogpt_libs install
+        files: ^autogpt_platform/autogpt_libs/poetry\.lock$
+        types: [file]
         language: system
         pass_filenames: false
-        stages: [pre-commit, post-checkout]
-
-      - id: pnpm-install
-        name: Check & Install dependencies - AutoGPT Platform - Frontend
-        alias: pnpm-install-platform-frontend
-        entry: >
-          bash -c '
-          if [ -n "$PRE_COMMIT_FROM_REF" ]; then
-            git diff --name-only "$PRE_COMMIT_FROM_REF" "$PRE_COMMIT_TO_REF"
-          else
-            git diff --cached --name-only
-          fi | grep -qE "^autogpt_platform/frontend/pnpm-lock\.yaml$" || exit 0;
-          pnpm --prefix autogpt_platform/frontend install
-          '
-        always_run: true
-        language: system
-        pass_filenames: false
-        stages: [pre-commit, post-checkout]
 
       - id: poetry-install
-        name: Check & Install dependencies - Classic
-        alias: poetry-install-classic
-        entry: >
-          bash -c '
-          if [ -n "$PRE_COMMIT_FROM_REF" ]; then
-            git diff --name-only "$PRE_COMMIT_FROM_REF" "$PRE_COMMIT_TO_REF"
-          else
-            git diff --cached --name-only
-          fi | grep -qE "^classic/poetry\.lock$" || exit 0;
-          poetry -C classic install
-          '
-        always_run: true
+        name: Check & Install dependencies - Classic - AutoGPT
+        alias: poetry-install-classic-autogpt
+        entry: poetry -C classic/original_autogpt install
+        # include forge source (since it's a path dependency)
+        files: ^classic/(original_autogpt|forge)/poetry\.lock$
+        types: [file]
+        language: system
+        pass_filenames: false
+
+      - id: poetry-install
+        name: Check & Install dependencies - Classic - Forge
+        alias: poetry-install-classic-forge
+        entry: poetry -C classic/forge install
+        files: ^classic/forge/poetry\.lock$
+        types: [file]
+        language: system
+        pass_filenames: false
+
+      - id: poetry-install
+        name: Check & Install dependencies - Classic - Benchmark
+        alias: poetry-install-classic-benchmark
+        entry: poetry -C classic/benchmark install
+        files: ^classic/benchmark/poetry\.lock$
+        types: [file]
         language: system
         pass_filenames: false
-        stages: [pre-commit, post-checkout]
 
   - repo: local
     # For proper type checking, Prisma client must be up-to-date.
@@ -112,54 +76,12 @@ repos:
       - id: prisma-generate
         name: Prisma Generate - AutoGPT Platform - Backend
         alias: prisma-generate-platform-backend
-        entry: >
-          bash -c '
-          if [ -n "$PRE_COMMIT_FROM_REF" ]; then
-            git diff --name-only "$PRE_COMMIT_FROM_REF" "$PRE_COMMIT_TO_REF"
-          else
-            git diff --cached --name-only
-          fi | grep -qE "^autogpt_platform/((backend|autogpt_libs)/poetry\.lock|backend/schema\.prisma)$" || exit 0;
-          cd autogpt_platform/backend
-          && poetry run prisma generate
-          && poetry run gen-prisma-stub
-          '
+        entry: bash -c 'cd autogpt_platform/backend && poetry run prisma generate'
         # include everything that triggers poetry install + the prisma schema
-        always_run: true
+        files: ^autogpt_platform/((backend|autogpt_libs)/poetry\.lock|backend/schema.prisma)$
+        types: [file]
         language: system
         pass_filenames: false
-        stages: [pre-commit, post-checkout]
-
-      - id: export-api-schema
-        name: Export API schema - AutoGPT Platform - Backend -> Frontend
-        alias: export-api-schema-platform
-        entry: >
-          bash -c '
-          cd autogpt_platform/backend
-          && poetry run export-api-schema --output ../frontend/src/app/api/openapi.json
-          && cd ../frontend
-          && pnpm prettier --write ./src/app/api/openapi.json
-          '
-        files: ^autogpt_platform/backend/
-        language: system
-        pass_filenames: false
-
-      - id: generate-api-client
-        name: Generate API client - AutoGPT Platform - Frontend
-        alias: generate-api-client-platform-frontend
-        entry: >
-          bash -c '
-          SCHEMA=autogpt_platform/frontend/src/app/api/openapi.json;
-          if [ -n "$PRE_COMMIT_FROM_REF" ]; then
-            git diff --quiet "$PRE_COMMIT_FROM_REF" "$PRE_COMMIT_TO_REF" -- "$SCHEMA" && exit 0
-          else
-            git diff --quiet HEAD -- "$SCHEMA" && exit 0
-          fi;
-          cd autogpt_platform/frontend && pnpm generate:api
-          '
-        always_run: true
-        language: system
-        pass_filenames: false
-        stages: [pre-commit, post-checkout]
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
     rev: v0.7.2
@@ -194,10 +116,26 @@ repos:
         language: system
 
       - id: isort
-        name: Lint (isort) - Classic
-        alias: isort-classic
-        entry: bash -c 'cd classic && poetry run isort $(echo "$@" | sed "s|classic/||g")' --
-        files: ^classic/(original_autogpt|forge|direct_benchmark)/
+        name: Lint (isort) - Classic - AutoGPT
+        alias: isort-classic-autogpt
+        entry: poetry -P classic/original_autogpt run isort -p autogpt
+        files: ^classic/original_autogpt/
+        types: [file, python]
+        language: system
+
+      - id: isort
+        name: Lint (isort) - Classic - Forge
+        alias: isort-classic-forge
+        entry: poetry -P classic/forge run isort -p forge
+        files: ^classic/forge/
+        types: [file, python]
+        language: system
+
+      - id: isort
+        name: Lint (isort) - Classic - Benchmark
+        alias: isort-classic-benchmark
+        entry: poetry -P classic/benchmark run isort -p agbenchmark
+        files: ^classic/benchmark/
         types: [file, python]
         language: system
 
@@ -211,13 +149,26 @@ repos:
 
   - repo: https://github.com/PyCQA/flake8
     rev: 7.0.0
-    # Use consolidated flake8 config at classic/.flake8
+    # To have flake8 load the config of the individual subprojects, we have to call
+    # them separately.
     hooks:
       - id: flake8
-        name: Lint (Flake8) - Classic
-        alias: flake8-classic
-        files: ^classic/(original_autogpt|forge|direct_benchmark)/
-        args: [--config=classic/.flake8]
+        name: Lint (Flake8) - Classic - AutoGPT
+        alias: flake8-classic-autogpt
+        files: ^classic/original_autogpt/(autogpt|scripts|tests)/
+        args: [--config=classic/original_autogpt/.flake8]
+
+      - id: flake8
+        name: Lint (Flake8) - Classic - Forge
+        alias: flake8-classic-forge
+        files: ^classic/forge/(forge|tests)/
+        args: [--config=classic/forge/.flake8]
+
+      - id: flake8
+        name: Lint (Flake8) - Classic - Benchmark
+        alias: flake8-classic-benchmark
+        files: ^classic/benchmark/(agbenchmark|tests)/((?!reports).)*[/.]
+        args: [--config=classic/benchmark/.flake8]
 
   - repo: local
     hooks:
@@ -253,10 +204,29 @@ repos:
         pass_filenames: false
 
       - id: pyright
-        name: Typecheck - Classic
-        alias: pyright-classic
-        entry: poetry -C classic run pyright
-        files: ^classic/(original_autogpt|forge|direct_benchmark)/.*\.py$|^classic/poetry\.lock$
+        name: Typecheck - Classic - AutoGPT
+        alias: pyright-classic-autogpt
+        entry: poetry -C classic/original_autogpt run pyright
+        # include forge source (since it's a path dependency) but exclude *_test.py files:
+        files: ^(classic/original_autogpt/((autogpt|scripts|tests)/|poetry\.lock$)|classic/forge/(forge/.*(?<!_test)\.py|poetry\.lock)$)
+        types: [file]
+        language: system
+        pass_filenames: false
+
+      - id: pyright
+        name: Typecheck - Classic - Forge
+        alias: pyright-classic-forge
+        entry: poetry -C classic/forge run pyright
+        files: ^classic/forge/(forge/|poetry\.lock$)
+        types: [file]
+        language: system
+        pass_filenames: false
+
+      - id: pyright
+        name: Typecheck - Classic - Benchmark
+        alias: pyright-classic-benchmark
+        entry: poetry -C classic/benchmark run pyright
+        files: ^classic/benchmark/(agbenchmark/|tests/|poetry\.lock$)
         types: [file]
         language: system
         pass_filenames: false
@@ -283,9 +253,26 @@ repos:
   #       pass_filenames: false
 
   #     - id: pytest
-  #       name: Run tests - Classic (excl. slow tests)
-  #       alias: pytest-classic
-  #       entry: bash -c 'cd classic && poetry run pytest -m "not slow"'
-  #       files: ^classic/(original_autogpt|forge|direct_benchmark)/
+  #       name: Run tests - Classic - AutoGPT (excl. slow tests)
+  #       alias: pytest-classic-autogpt
+  #       entry: bash -c 'cd classic/original_autogpt && poetry run pytest --cov=autogpt -m "not slow" tests/unit tests/integration'
+  #       # include forge source (since it's a path dependency) but exclude *_test.py files:
+  #       files: ^(classic/original_autogpt/((autogpt|tests)/|poetry\.lock$)|classic/forge/(forge/.*(?<!_test)\.py|poetry\.lock)$)
+  #       language: system
+  #       pass_filenames: false
+
+  #     - id: pytest
+  #       name: Run tests - Classic - Forge (excl. slow tests)
+  #       alias: pytest-classic-forge
+  #       entry: bash -c 'cd classic/forge && poetry run pytest --cov=forge -m "not slow"'
+  #       files: ^classic/forge/(forge/|tests/|poetry\.lock$)
+  #       language: system
+  #       pass_filenames: false
+
+  #     - id: pytest
+  #       name: Run tests - Classic - Benchmark
+  #       alias: pytest-classic-benchmark
+  #       entry: bash -c 'cd classic/benchmark && poetry run pytest --cov=benchmark'
+  #       files: ^classic/benchmark/(agbenchmark/|tests/|poetry\.lock$)
   #       language: system
   #       pass_filenames: false

.secrets.baseline

@@ -1,471 +0,0 @@
-{
-  "version": "1.5.0",
-  "plugins_used": [
-    {
-      "name": "ArtifactoryDetector"
-    },
-    {
-      "name": "AWSKeyDetector"
-    },
-    {
-      "name": "AzureStorageKeyDetector"
-    },
-    {
-      "name": "Base64HighEntropyString",
-      "limit": 4.5
-    },
-    {
-      "name": "BasicAuthDetector"
-    },
-    {
-      "name": "CloudantDetector"
-    },
-    {
-      "name": "DiscordBotTokenDetector"
-    },
-    {
-      "name": "GitHubTokenDetector"
-    },
-    {
-      "name": "GitLabTokenDetector"
-    },
-    {
-      "name": "HexHighEntropyString",
-      "limit": 3.0
-    },
-    {
-      "name": "IbmCloudIamDetector"
-    },
-    {
-      "name": "IbmCosHmacDetector"
-    },
-    {
-      "name": "IPPublicDetector"
-    },
-    {
-      "name": "JwtTokenDetector"
-    },
-    {
-      "name": "KeywordDetector",
-      "keyword_exclude": ""
-    },
-    {
-      "name": "MailchimpDetector"
-    },
-    {
-      "name": "NpmDetector"
-    },
-    {
-      "name": "OpenAIDetector"
-    },
-    {
-      "name": "PrivateKeyDetector"
-    },
-    {
-      "name": "PypiTokenDetector"
-    },
-    {
-      "name": "SendGridDetector"
-    },
-    {
-      "name": "SlackDetector"
-    },
-    {
-      "name": "SoftlayerDetector"
-    },
-    {
-      "name": "SquareOAuthDetector"
-    },
-    {
-      "name": "StripeDetector"
-    },
-    {
-      "name": "TelegramBotTokenDetector"
-    },
-    {
-      "name": "TwilioKeyDetector"
-    }
-  ],
-  "filters_used": [
-    {
-      "path": "detect_secrets.filters.allowlist.is_line_allowlisted"
-    },
-    {
-      "path": "detect_secrets.filters.common.is_baseline_file",
-      "filename": ".secrets.baseline"
-    },
-    {
-      "path": "detect_secrets.filters.common.is_ignored_due_to_verification_policies",
-      "min_level": 2
-    },
-    {
-      "path": "detect_secrets.filters.heuristic.is_indirect_reference"
-    },
-    {
-      "path": "detect_secrets.filters.heuristic.is_likely_id_string"
-    },
-    {
-      "path": "detect_secrets.filters.heuristic.is_lock_file"
-    },
-    {
-      "path": "detect_secrets.filters.heuristic.is_not_alphanumeric_string"
-    },
-    {
-      "path": "detect_secrets.filters.heuristic.is_potential_uuid"
-    },
-    {
-      "path": "detect_secrets.filters.heuristic.is_prefixed_with_dollar_sign"
-    },
-    {
-      "path": "detect_secrets.filters.heuristic.is_sequential_string"
-    },
-    {
-      "path": "detect_secrets.filters.heuristic.is_swagger_file"
-    },
-    {
-      "path": "detect_secrets.filters.heuristic.is_templated_secret"
-    },
-    {
-      "path": "detect_secrets.filters.regex.should_exclude_file",
-      "pattern": [
-        "\\.env$",
-        "pnpm-lock\\.yaml$",
-        "\\.env\\.(default|example|template)$",
-        "__pycache__",
-        "_test\\.py$",
-        "test_.*\\.py$",
-        "conftest\\.py$",
-        "poetry\\.lock$",
-        "node_modules"
-      ]
-    }
-  ],
-  "results": {
-    "autogpt_platform/backend/backend/api/external/v1/integrations.py": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/backend/backend/api/external/v1/integrations.py",
-        "hashed_secret": "665b1e3851eefefa3fb878654292f16597d25155",
-        "is_verified": false,
-        "line_number": 289
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/airtable/_config.py": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/backend/backend/blocks/airtable/_config.py",
-        "hashed_secret": "57e168b03afb7c1ee3cdc4ee3db2fe1cc6e0df26",
-        "is_verified": false,
-        "line_number": 29
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/dataforseo/_config.py": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/backend/backend/blocks/dataforseo/_config.py",
-        "hashed_secret": "32ce93887331fa5d192f2876ea15ec000c7d58b8",
-        "is_verified": false,
-        "line_number": 12
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/github/checks.py": [
-      {
-        "type": "Hex High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/github/checks.py",
-        "hashed_secret": "8ac6f92737d8586790519c5d7bfb4d2eb172c238",
-        "is_verified": false,
-        "line_number": 108
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/github/ci.py": [
-      {
-        "type": "Hex High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/github/ci.py",
-        "hashed_secret": "90bd1b48e958257948487b90bee080ba5ed00caa",
-        "is_verified": false,
-        "line_number": 123
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/github/example_payloads/pull_request.synchronize.json": [
-      {
-        "type": "Hex High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/github/example_payloads/pull_request.synchronize.json",
-        "hashed_secret": "f96896dafced7387dcd22343b8ea29d3d2c65663",
-        "is_verified": false,
-        "line_number": 42
-      },
-      {
-        "type": "Hex High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/github/example_payloads/pull_request.synchronize.json",
-        "hashed_secret": "b80a94d5e70bedf4f5f89d2f5a5255cc9492d12e",
-        "is_verified": false,
-        "line_number": 193
-      },
-      {
-        "type": "Hex High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/github/example_payloads/pull_request.synchronize.json",
-        "hashed_secret": "75b17e517fe1b3136394f6bec80c4f892da75e42",
-        "is_verified": false,
-        "line_number": 344
-      },
-      {
-        "type": "Hex High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/github/example_payloads/pull_request.synchronize.json",
-        "hashed_secret": "b0bfb5e4e2394e7f8906e5ed1dffd88b2bc89dd5",
-        "is_verified": false,
-        "line_number": 534
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/github/statuses.py": [
-      {
-        "type": "Hex High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/github/statuses.py",
-        "hashed_secret": "8ac6f92737d8586790519c5d7bfb4d2eb172c238",
-        "is_verified": false,
-        "line_number": 85
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/google/docs.py": [
-      {
-        "type": "Hex High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/google/docs.py",
-        "hashed_secret": "c95da0c6696342c867ef0c8258d2f74d20fd94d4",
-        "is_verified": false,
-        "line_number": 203
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/google/sheets.py": [
-      {
-        "type": "Base64 High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/google/sheets.py",
-        "hashed_secret": "bd5a04fa3667e693edc13239b6d310c5c7a8564b",
-        "is_verified": false,
-        "line_number": 57
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/linear/_config.py": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/backend/backend/blocks/linear/_config.py",
-        "hashed_secret": "b37f020f42d6d613b6ce30103e4d408c4499b3bb",
-        "is_verified": false,
-        "line_number": 53
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/medium.py": [
-      {
-        "type": "Hex High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/medium.py",
-        "hashed_secret": "ff998abc1ce6d8f01a675fa197368e44c8916e9c",
-        "is_verified": false,
-        "line_number": 131
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/replicate/replicate_block.py": [
-      {
-        "type": "Hex High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/replicate/replicate_block.py",
-        "hashed_secret": "8bbdd6f26368f58ea4011d13d7f763cb662e66f0",
-        "is_verified": false,
-        "line_number": 55
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/slant3d/webhook.py": [
-      {
-        "type": "Hex High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/slant3d/webhook.py",
-        "hashed_secret": "36263c76947443b2f6e6b78153967ac4a7da99f9",
-        "is_verified": false,
-        "line_number": 100
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/talking_head.py": [
-      {
-        "type": "Base64 High Entropy String",
-        "filename": "autogpt_platform/backend/backend/blocks/talking_head.py",
-        "hashed_secret": "44ce2d66222529eea4a32932823466fc0601c799",
-        "is_verified": false,
-        "line_number": 113
-      }
-    ],
-    "autogpt_platform/backend/backend/blocks/wordpress/_config.py": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/backend/backend/blocks/wordpress/_config.py",
-        "hashed_secret": "e62679512436161b78e8a8d68c8829c2a1031ccb",
-        "is_verified": false,
-        "line_number": 17
-      }
-    ],
-    "autogpt_platform/backend/backend/util/cache.py": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/backend/backend/util/cache.py",
-        "hashed_secret": "37f0c918c3fa47ca4a70e42037f9f123fdfbc75b",
-        "is_verified": false,
-        "line_number": 449
-      }
-    ],
-    "autogpt_platform/frontend/src/app/(platform)/build/components/FlowEditor/nodes/helpers.ts": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/app/(platform)/build/components/FlowEditor/nodes/helpers.ts",
-        "hashed_secret": "5baa61e4c9b93f3f0682250b6cf8331b7ee68fd8",
-        "is_verified": false,
-        "line_number": 6
-      }
-    ],
-    "autogpt_platform/frontend/src/app/(platform)/dictionaries/en.json": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/app/(platform)/dictionaries/en.json",
-        "hashed_secret": "8be3c943b1609fffbfc51aad666d0a04adf83c9d",
-        "is_verified": false,
-        "line_number": 5
-      }
-    ],
-    "autogpt_platform/frontend/src/app/(platform)/dictionaries/es.json": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/app/(platform)/dictionaries/es.json",
-        "hashed_secret": "5a6d1c612954979ea99ee33dbb2d231b00f6ac0a",
-        "is_verified": false,
-        "line_number": 5
-      }
-    ],
-    "autogpt_platform/frontend/src/app/(platform)/library/agents/[id]/components/NewAgentLibraryView/components/modals/AgentInputsReadOnly/helpers.ts": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/app/(platform)/library/agents/[id]/components/NewAgentLibraryView/components/modals/AgentInputsReadOnly/helpers.ts",
-        "hashed_secret": "cf678cab87dc1f7d1b95b964f15375e088461679",
-        "is_verified": false,
-        "line_number": 6
-      },
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/app/(platform)/library/agents/[id]/components/NewAgentLibraryView/components/modals/AgentInputsReadOnly/helpers.ts",
-        "hashed_secret": "f72cbb45464d487064610c5411c576ca4019d380",
-        "is_verified": false,
-        "line_number": 8
-      }
-    ],
-    "autogpt_platform/frontend/src/app/(platform)/library/agents/[id]/components/NewAgentLibraryView/components/modals/RunAgentModal/components/ModalRunSection/helpers.ts": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/app/(platform)/library/agents/[id]/components/NewAgentLibraryView/components/modals/RunAgentModal/components/ModalRunSection/helpers.ts",
-        "hashed_secret": "cf678cab87dc1f7d1b95b964f15375e088461679",
-        "is_verified": false,
-        "line_number": 5
-      },
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/app/(platform)/library/agents/[id]/components/NewAgentLibraryView/components/modals/RunAgentModal/components/ModalRunSection/helpers.ts",
-        "hashed_secret": "f72cbb45464d487064610c5411c576ca4019d380",
-        "is_verified": false,
-        "line_number": 7
-      }
-    ],
-    "autogpt_platform/frontend/src/app/(platform)/profile/(user)/integrations/page.tsx": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/app/(platform)/profile/(user)/integrations/page.tsx",
-        "hashed_secret": "cf678cab87dc1f7d1b95b964f15375e088461679",
-        "is_verified": false,
-        "line_number": 192
-      },
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/app/(platform)/profile/(user)/integrations/page.tsx",
-        "hashed_secret": "86275db852204937bbdbdebe5fabe8536e030ab6",
-        "is_verified": false,
-        "line_number": 193
-      }
-    ],
-    "autogpt_platform/frontend/src/components/contextual/CredentialsInput/helpers.ts": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/components/contextual/CredentialsInput/helpers.ts",
-        "hashed_secret": "47acd2028cf81b5da88ddeedb2aea4eca4b71fbd",
-        "is_verified": false,
-        "line_number": 102
-      },
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/components/contextual/CredentialsInput/helpers.ts",
-        "hashed_secret": "8be3c943b1609fffbfc51aad666d0a04adf83c9d",
-        "is_verified": false,
-        "line_number": 103
-      }
-    ],
-    "autogpt_platform/frontend/src/lib/autogpt-server-api/utils.ts": [
-      {
-        "type": "Base64 High Entropy String",
-        "filename": "autogpt_platform/frontend/src/lib/autogpt-server-api/utils.ts",
-        "hashed_secret": "9c486c92f1a7420e1045c7ad963fbb7ba3621025",
-        "is_verified": false,
-        "line_number": 73
-      },
-      {
-        "type": "Base64 High Entropy String",
-        "filename": "autogpt_platform/frontend/src/lib/autogpt-server-api/utils.ts",
-        "hashed_secret": "9277508c7a6effc8fb59163efbfada189e35425c",
-        "is_verified": false,
-        "line_number": 75
-      },
-      {
-        "type": "Base64 High Entropy String",
-        "filename": "autogpt_platform/frontend/src/lib/autogpt-server-api/utils.ts",
-        "hashed_secret": "8dc7e2cb1d0935897d541bf5facab389b8a50340",
-        "is_verified": false,
-        "line_number": 77
-      },
-      {
-        "type": "Base64 High Entropy String",
-        "filename": "autogpt_platform/frontend/src/lib/autogpt-server-api/utils.ts",
-        "hashed_secret": "79a26ad48775944299be6aaf9fb1d5302c1ed75b",
-        "is_verified": false,
-        "line_number": 79
-      },
-      {
-        "type": "Base64 High Entropy String",
-        "filename": "autogpt_platform/frontend/src/lib/autogpt-server-api/utils.ts",
-        "hashed_secret": "a3b62b44500a1612e48d4cab8294df81561b3b1a",
-        "is_verified": false,
-        "line_number": 81
-      },
-      {
-        "type": "Base64 High Entropy String",
-        "filename": "autogpt_platform/frontend/src/lib/autogpt-server-api/utils.ts",
-        "hashed_secret": "a58979bd0b21ef4f50417d001008e60dd7a85c64",
-        "is_verified": false,
-        "line_number": 83
-      },
-      {
-        "type": "Base64 High Entropy String",
-        "filename": "autogpt_platform/frontend/src/lib/autogpt-server-api/utils.ts",
-        "hashed_secret": "6cb6e075f8e8c7c850f9d128d6608e5dbe209a79",
-        "is_verified": false,
-        "line_number": 85
-      }
-    ],
-    "autogpt_platform/frontend/src/lib/constants.ts": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/lib/constants.ts",
-        "hashed_secret": "27b924db06a28cc755fb07c54f0fddc30659fe4d",
-        "is_verified": false,
-        "line_number": 13
-      }
-    ],
-    "autogpt_platform/frontend/src/tests/credentials/index.ts": [
-      {
-        "type": "Secret Keyword",
-        "filename": "autogpt_platform/frontend/src/tests/credentials/index.ts",
-        "hashed_secret": "c18006fc138809314751cd1991f1e0b820fabd37",
-        "is_verified": false,
-        "line_number": 4
-      }
-    ]
-  },
-  "generated_at": "2026-04-09T14:20:23Z"
-}

AGENTS.md

@@ -1,6 +1,6 @@
 # AutoGPT Platform Contribution Guide
 
-This guide provides context for coding agents when updating the **autogpt_platform** folder.
+This guide provides context for Codex when updating the **autogpt_platform** folder.
 
 ## Directory overview
 
@@ -30,7 +30,7 @@ See `/frontend/CONTRIBUTING.md` for complete patterns. Quick reference:
    - Regenerate with `pnpm generate:api`
    - Pattern: `use{Method}{Version}{OperationName}`
 4. **Styling**: Tailwind CSS only, use design tokens, Phosphor Icons only
-5. **Testing**: Integration tests (Vitest + RTL + MSW) are the default (~90%, page-level). Playwright for E2E critical flows. Storybook for design system components. See `autogpt_platform/frontend/TESTING.md`
+5. **Testing**: Add Storybook stories for new components, Playwright for E2E
 6. **Code conventions**: Function declarations (not arrow functions) for components/handlers
 
 - Component props should be `interface Props { ... }` (not exported) unless the interface needs to be used outside the component
@@ -47,9 +47,7 @@ See `/frontend/CONTRIBUTING.md` for complete patterns. Quick reference:
 ## Testing
 
 - Backend: `poetry run test` (runs pytest with a docker based postgres + prisma).
-- Frontend integration tests: `pnpm test:unit` (Vitest + RTL + MSW, primary testing approach).
-- Frontend E2E tests: `pnpm test` or `pnpm test-ui` for Playwright tests.
-- See `autogpt_platform/frontend/TESTING.md` for the full testing strategy.
+- Frontend: `pnpm test` or `pnpm test-ui` for Playwright tests. See `docs/content/platform/contributing/tests.md` for tips.
 
 Always run the relevant linters and tests before committing.
 Use conventional commit messages for all commits (e.g. `feat(backend): add API`).

CLAUDE.md

@@ -1 +0,0 @@
-@AGENTS.md

README.md

@@ -83,13 +83,13 @@ The AutoGPT frontend is where users interact with our powerful AI automation pla
 
    **Agent Builder:** For those who want to customize, our intuitive, low-code interface allows you to design and configure your own AI agents. 
    
-   **Workflow Management:** Build, modify, and optimize your automation workflows with ease. You build your agent by connecting blocks, where each block performs a single action.
+   **Workflow Management:** Build, modify, and optimize your automation workflows with ease. You build your agent by connecting blocks, where each block     performs a single action.
    
    **Deployment Controls:** Manage the lifecycle of your agents, from testing to production.
    
    **Ready-to-Use Agents:** Don't want to build? Simply select from our library of pre-configured agents and put them to work immediately.
    
-   **Agent Interaction:** Whether you've built your own or are using pre-configured agents, easily run and interact with them through our user-friendly interface.
+   **Agent Interaction:** Whether you've built your own or are using pre-configured agents, easily run and interact with them through our user-friendly      interface.
 
    **Monitoring and Analytics:** Keep track of your agents' performance and gain insights to continually improve your automation processes.
 

autogpt_platform/.gitignore

@@ -1,6 +1,2 @@
 *.ignore.*
-*.ign.*
-.application.logs
-
-# Claude Code local settings only — the rest of .claude/ is shared (skills etc.)
-.claude/settings.local.json
+*.ign.*

autogpt_platform/AGENTS.md

@@ -1,120 +0,0 @@
-# AutoGPT Platform
-
-This file provides guidance to coding agents when working with code in this repository.
-
-## Repository Overview
-
-AutoGPT Platform is a monorepo containing:
-
-- **Backend** (`backend`): Python FastAPI server with async support
-- **Frontend** (`frontend`): Next.js React application
-- **Shared Libraries** (`autogpt_libs`): Common Python utilities
-
-## Component Documentation
-
-- **Backend**: See @backend/AGENTS.md for backend-specific commands, architecture, and development tasks
-- **Frontend**: See @frontend/AGENTS.md for frontend-specific commands, architecture, and development patterns
-
-## Key Concepts
-
-1. **Agent Graphs**: Workflow definitions stored as JSON, executed by the backend
-2. **Blocks**: Reusable components in `backend/backend/blocks/` that perform specific tasks
-3. **Integrations**: OAuth and API connections stored per user
-4. **Store**: Marketplace for sharing agent templates
-5. **Virus Scanning**: ClamAV integration for file upload security
-
-### Environment Configuration
-
-#### Configuration Files
-
-- **Backend**: `backend/.env.default` (defaults) → `backend/.env` (user overrides)
-- **Frontend**: `frontend/.env.default` (defaults) → `frontend/.env` (user overrides)
-- **Platform**: `.env.default` (Supabase/shared defaults) → `.env` (user overrides)
-
-#### Docker Environment Loading Order
-
-1. `.env.default` files provide base configuration (tracked in git)
-2. `.env` files provide user-specific overrides (gitignored)
-3. Docker Compose `environment:` sections provide service-specific overrides
-4. Shell environment variables have highest precedence
-
-#### Key Points
-
-- All services use hardcoded defaults in docker-compose files (no `${VARIABLE}` substitutions)
-- The `env_file` directive loads variables INTO containers at runtime
-- Backend/Frontend services use YAML anchors for consistent configuration
-- Supabase services (`db/docker/docker-compose.yml`) follow the same pattern
-
-### Branching Strategy
-
-- **`dev`** is the main development branch. All PRs should target `dev`.
-- **`master`** is the production branch. Only used for production releases.
-
-### Creating Pull Requests
-
-- Create the PR against the `dev` branch of the repository.
-- **Split PRs by concern** — each PR should have a single clear purpose. For example, "usage tracking" and "credit charging" should be separate PRs even if related. Combining multiple concerns makes it harder for reviewers to understand what belongs to what.
-- Ensure the branch name is descriptive (e.g., `feature/add-new-block`)
-- Use conventional commit messages (see below)
-- **Structure the PR description with Why / What / How** — Why: the motivation (what problem it solves, what's broken/missing without it); What: high-level summary of changes; How: approach, key implementation details, or architecture decisions. Reviewers need all three to judge whether the approach fits the problem.
-- Fill out the .github/PULL_REQUEST_TEMPLATE.md template as the PR description
-- Always use `--body-file` to pass PR body — avoids shell interpretation of backticks and special characters:
-  ```bash
-  PR_BODY=$(mktemp)
-  cat > "$PR_BODY" << 'PREOF'
-  ## Summary
-  - use `backticks` freely here
-  PREOF
-  gh pr create --title "..." --body-file "$PR_BODY" --base dev
-  rm "$PR_BODY"
-  ```
-- Run the github pre-commit hooks to ensure code quality.
-
-### Test-Driven Development (TDD)
-
-When fixing a bug or adding a feature, follow a test-first approach:
-
-1. **Write a failing test first** — create a test that reproduces the bug or validates the new behavior, marked with `@pytest.mark.xfail` (backend) or `.fixme` (Playwright). Run it to confirm it fails for the right reason.
-2. **Implement the fix/feature** — write the minimal code to make the test pass.
-3. **Remove the xfail marker** — once the test passes, remove the `xfail`/`.fixme` annotation and run the full test suite to confirm nothing else broke.
-
-This ensures every change is covered by a test and that the test actually validates the intended behavior.
-
-### Reviewing/Revising Pull Requests
-
-Use `/pr-review` to review a PR or `/pr-address` to address comments.
-
-When fetching comments manually:
-- `gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/reviews --paginate` — top-level reviews
-- `gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/comments --paginate` — inline review comments (always paginate to avoid missing comments beyond page 1)
-- `gh api repos/Significant-Gravitas/AutoGPT/issues/{N}/comments` — PR conversation comments
-
-### Conventional Commits
-
-Use this format for commit messages and Pull Request titles:
-
-**Conventional Commit Types:**
-
-- `feat`: Introduces a new feature to the codebase
-- `fix`: Patches a bug in the codebase
-- `refactor`: Code change that neither fixes a bug nor adds a feature; also applies to removing features
-- `ci`: Changes to CI configuration
-- `docs`: Documentation-only changes
-- `dx`: Improvements to the developer experience
-
-**Recommended Base Scopes:**
-
-- `platform`: Changes affecting both frontend and backend
-- `frontend`
-- `backend`
-- `infra`
-- `blocks`: Modifications/additions of individual blocks
-
-**Subscope Examples:**
-
-- `backend/executor`
-- `backend/db`
-- `frontend/builder` (includes changes to the block UI component)
-- `infra/prod`
-
-Use these scopes and subscopes for clarity and consistency in commit messages.

autogpt_platform/CLAUDE.md

@@ -1 +1,95 @@
-@AGENTS.md
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Repository Overview
+
+AutoGPT Platform is a monorepo containing:
+
+- **Backend** (`backend`): Python FastAPI server with async support
+- **Frontend** (`frontend`): Next.js React application
+- **Shared Libraries** (`autogpt_libs`): Common Python utilities
+
+## Component Documentation
+
+- **Backend**: See @backend/CLAUDE.md for backend-specific commands, architecture, and development tasks
+- **Frontend**: See @frontend/CLAUDE.md for frontend-specific commands, architecture, and development patterns
+
+## Key Concepts
+
+1. **Agent Graphs**: Workflow definitions stored as JSON, executed by the backend
+2. **Blocks**: Reusable components in `backend/backend/blocks/` that perform specific tasks
+3. **Integrations**: OAuth and API connections stored per user
+4. **Store**: Marketplace for sharing agent templates
+5. **Virus Scanning**: ClamAV integration for file upload security
+
+### Environment Configuration
+
+#### Configuration Files
+
+- **Backend**: `backend/.env.default` (defaults) → `backend/.env` (user overrides)
+- **Frontend**: `frontend/.env.default` (defaults) → `frontend/.env` (user overrides)
+- **Platform**: `.env.default` (Supabase/shared defaults) → `.env` (user overrides)
+
+#### Docker Environment Loading Order
+
+1. `.env.default` files provide base configuration (tracked in git)
+2. `.env` files provide user-specific overrides (gitignored)
+3. Docker Compose `environment:` sections provide service-specific overrides
+4. Shell environment variables have highest precedence
+
+#### Key Points
+
+- All services use hardcoded defaults in docker-compose files (no `${VARIABLE}` substitutions)
+- The `env_file` directive loads variables INTO containers at runtime
+- Backend/Frontend services use YAML anchors for consistent configuration
+- Supabase services (`db/docker/docker-compose.yml`) follow the same pattern
+
+### Branching Strategy
+
+- **`dev`** is the main development branch. All PRs should target `dev`.
+- **`master`** is the production branch. Only used for production releases.
+
+### Creating Pull Requests
+
+- Create the PR against the `dev` branch of the repository.
+- Ensure the branch name is descriptive (e.g., `feature/add-new-block`)
+- Use conventional commit messages (see below)
+- Fill out the .github/PULL_REQUEST_TEMPLATE.md template as the PR description
+- Run the github pre-commit hooks to ensure code quality.
+
+### Reviewing/Revising Pull Requests
+
+- When the user runs /pr-comments or tries to fetch them, also run gh api /repos/Significant-Gravitas/AutoGPT/pulls/[issuenum]/reviews to get the reviews
+- Use gh api /repos/Significant-Gravitas/AutoGPT/pulls/[issuenum]/reviews/[review_id]/comments to get the review contents
+- Use gh api /repos/Significant-Gravitas/AutoGPT/issues/9924/comments to get the pr specific comments
+
+### Conventional Commits
+
+Use this format for commit messages and Pull Request titles:
+
+**Conventional Commit Types:**
+
+- `feat`: Introduces a new feature to the codebase
+- `fix`: Patches a bug in the codebase
+- `refactor`: Code change that neither fixes a bug nor adds a feature; also applies to removing features
+- `ci`: Changes to CI configuration
+- `docs`: Documentation-only changes
+- `dx`: Improvements to the developer experience
+
+**Recommended Base Scopes:**
+
+- `platform`: Changes affecting both frontend and backend
+- `frontend`
+- `backend`
+- `infra`
+- `blocks`: Modifications/additions of individual blocks
+
+**Subscope Examples:**
+
+- `backend/executor`
+- `backend/db`
+- `frontend/builder` (includes changes to the block UI component)
+- `infra/prod`
+
+Use these scopes and subscopes for clarity and consistency in commit messages.

autogpt_platform/analytics/queries/auth_activities.sql

@@ -1,40 +0,0 @@
--- =============================================================
--- View: analytics.auth_activities
--- Looker source alias: ds49  |  Charts: 1
--- =============================================================
--- DESCRIPTION
---   Tracks authentication events (login, logout, SSO, password
---   reset, etc.) from Supabase's internal audit log.
---   Useful for monitoring sign-in patterns and detecting anomalies.
---
--- SOURCE TABLES
---   auth.audit_log_entries  — Supabase internal auth event log
---
--- OUTPUT COLUMNS
---   created_at      TIMESTAMPTZ  When the auth event occurred
---   actor_id        TEXT         User ID who triggered the event
---   actor_via_sso   TEXT         Whether the action was via SSO ('true'/'false')
---   action          TEXT         Event type (e.g. 'login', 'logout', 'token_refreshed')
---
--- WINDOW
---   Rolling 90 days from current date
---
--- EXAMPLE QUERIES
---   -- Daily login counts
---   SELECT DATE_TRUNC('day', created_at) AS day, COUNT(*) AS logins
---   FROM analytics.auth_activities
---   WHERE action = 'login'
---   GROUP BY 1 ORDER BY 1;
---
---   -- SSO vs password login breakdown
---   SELECT actor_via_sso, COUNT(*) FROM analytics.auth_activities
---   WHERE action = 'login' GROUP BY 1;
--- =============================================================
-
-SELECT
-    created_at,
-    payload->>'actor_id'      AS actor_id,
-    payload->>'actor_via_sso' AS actor_via_sso,
-    payload->>'action'        AS action
-FROM auth.audit_log_entries
-WHERE created_at >= NOW() - INTERVAL '90 days'

autogpt_platform/analytics/queries/graph_execution.sql

@@ -1,105 +0,0 @@
--- =============================================================
--- View: analytics.graph_execution
--- Looker source alias: ds16  |  Charts: 21
--- =============================================================
--- DESCRIPTION
---   One row per agent graph execution (last 90 days).
---   Unpacks the JSONB stats column into individual numeric columns
---   and normalises the executionStatus — runs that failed due to
---   insufficient credits are reclassified as 'NO_CREDITS' for
---   easier filtering.  Error messages are scrubbed of IDs and URLs
---   to allow safe grouping.
---
--- SOURCE TABLES
---   platform.AgentGraphExecution  — Execution records
---   platform.AgentGraph           — Agent graph metadata (for name)
---   platform.LibraryAgent         — To flag possibly-AI (safe-mode) agents
---
--- OUTPUT COLUMNS
---   id                TEXT         Execution UUID
---   agentGraphId      TEXT         Agent graph UUID
---   agentGraphVersion INT          Graph version number
---   executionStatus   TEXT         COMPLETED | FAILED | NO_CREDITS | RUNNING | QUEUED | TERMINATED
---   createdAt         TIMESTAMPTZ  When the execution was queued
---   updatedAt         TIMESTAMPTZ  Last status update time
---   userId            TEXT         Owner user UUID
---   agentGraphName    TEXT         Human-readable agent name
---   cputime           DECIMAL      Total CPU seconds consumed
---   walltime          DECIMAL      Total wall-clock seconds
---   node_count        DECIMAL      Number of nodes in the graph
---   nodes_cputime     DECIMAL      CPU time across all nodes
---   nodes_walltime    DECIMAL      Wall time across all nodes
---   execution_cost    DECIMAL      Credit cost of this execution
---   correctness_score FLOAT        AI correctness score (if available)
---   possibly_ai       BOOLEAN      True if agent has sensitive_action_safe_mode enabled
---   groupedErrorMessage TEXT       Scrubbed error string (IDs/URLs replaced with wildcards)
---
--- WINDOW
---   Rolling 90 days (createdAt > CURRENT_DATE - 90 days)
---
--- EXAMPLE QUERIES
---   -- Daily execution counts by status
---   SELECT DATE_TRUNC('day', "createdAt") AS day, "executionStatus", COUNT(*)
---   FROM analytics.graph_execution
---   GROUP BY 1, 2 ORDER BY 1;
---
---   -- Average cost per execution by agent
---   SELECT "agentGraphName", AVG("execution_cost") AS avg_cost, COUNT(*) AS runs
---   FROM analytics.graph_execution
---   WHERE "executionStatus" = 'COMPLETED'
---   GROUP BY 1 ORDER BY avg_cost DESC;
---
---   -- Top error messages
---   SELECT "groupedErrorMessage", COUNT(*) AS occurrences
---   FROM analytics.graph_execution
---   WHERE "executionStatus" = 'FAILED'
---   GROUP BY 1 ORDER BY 2 DESC LIMIT 20;
--- =============================================================
-
-SELECT
-    ge."id"                                                        AS id,
-    ge."agentGraphId"                                              AS agentGraphId,
-    ge."agentGraphVersion"                                         AS agentGraphVersion,
-    CASE
-        WHEN jsonb_exists(ge."stats"::jsonb, 'error')
-         AND (
-               (ge."stats"::jsonb->>'error') ILIKE '%insufficient balance%'
-            OR (ge."stats"::jsonb->>'error') ILIKE '%you have no credits left%'
-             )
-        THEN 'NO_CREDITS'
-        ELSE CAST(ge."executionStatus" AS TEXT)
-    END                                                            AS executionStatus,
-    ge."createdAt"                                                 AS createdAt,
-    ge."updatedAt"                                                 AS updatedAt,
-    ge."userId"                                                    AS userId,
-    g."name"                                                       AS agentGraphName,
-    (ge."stats"::jsonb->>'cputime')::decimal                       AS cputime,
-    (ge."stats"::jsonb->>'walltime')::decimal                      AS walltime,
-    (ge."stats"::jsonb->>'node_count')::decimal                    AS node_count,
-    (ge."stats"::jsonb->>'nodes_cputime')::decimal                 AS nodes_cputime,
-    (ge."stats"::jsonb->>'nodes_walltime')::decimal                AS nodes_walltime,
-    (ge."stats"::jsonb->>'cost')::decimal                          AS execution_cost,
-    (ge."stats"::jsonb->>'correctness_score')::float               AS correctness_score,
-    COALESCE(la.possibly_ai, FALSE)                                AS possibly_ai,
-    REGEXP_REPLACE(
-        REGEXP_REPLACE(
-            TRIM(BOTH '"' FROM ge."stats"::jsonb->>'error'),
-            '(https?://)([A-Za-z0-9.-]+)(:[0-9]+)?(/[^\s]*)?',
-            '\1\2/...', 'gi'
-        ),
-        '[a-zA-Z0-9_:-]*\d[a-zA-Z0-9_:-]*', '*', 'g'
-    )                                                              AS groupedErrorMessage
-FROM platform."AgentGraphExecution" ge
-LEFT JOIN platform."AgentGraph" g
-       ON ge."agentGraphId" = g."id"
-      AND ge."agentGraphVersion" = g."version"
-LEFT JOIN (
-    SELECT DISTINCT ON ("userId", "agentGraphId")
-           "userId", "agentGraphId",
-           ("settings"::jsonb->>'sensitive_action_safe_mode')::boolean AS possibly_ai
-    FROM platform."LibraryAgent"
-    WHERE "isDeleted"  = FALSE
-      AND "isArchived" = FALSE
-    ORDER BY "userId", "agentGraphId", "agentGraphVersion" DESC
-) la ON la."userId" = ge."userId" AND la."agentGraphId" = ge."agentGraphId"
-WHERE ge."createdAt" > CURRENT_DATE - INTERVAL '90 days'

autogpt_platform/analytics/queries/node_block_execution.sql

@@ -1,101 +0,0 @@
--- =============================================================
--- View: analytics.node_block_execution
--- Looker source alias: ds14  |  Charts: 11
--- =============================================================
--- DESCRIPTION
---   One row per node (block) execution (last 90 days).
---   Unpacks stats JSONB and joins to identify which block type
---   was run.  For failed nodes, joins the error output and
---   scrubs it for safe grouping.
---
--- SOURCE TABLES
---   platform.AgentNodeExecution              — Node execution records
---   platform.AgentNode                       — Node → block mapping
---   platform.AgentBlock                      — Block name/ID
---   platform.AgentNodeExecutionInputOutput   — Error output values
---
--- OUTPUT COLUMNS
---   id                    TEXT         Node execution UUID
---   agentGraphExecutionId TEXT         Parent graph execution UUID
---   agentNodeId           TEXT         Node UUID within the graph
---   executionStatus       TEXT         COMPLETED | FAILED | QUEUED | RUNNING | TERMINATED
---   addedTime             TIMESTAMPTZ  When the node was queued
---   queuedTime            TIMESTAMPTZ  When it entered the queue
---   startedTime           TIMESTAMPTZ  When execution started
---   endedTime             TIMESTAMPTZ  When execution finished
---   inputSize             BIGINT       Input payload size in bytes
---   outputSize            BIGINT       Output payload size in bytes
---   walltime              NUMERIC      Wall-clock seconds for this node
---   cputime               NUMERIC      CPU seconds for this node
---   llmRetryCount         INT          Number of LLM retries
---   llmCallCount          INT          Number of LLM API calls made
---   inputTokenCount       BIGINT       LLM input tokens consumed
---   outputTokenCount      BIGINT       LLM output tokens produced
---   blockName             TEXT         Human-readable block name (e.g. 'OpenAIBlock')
---   blockId               TEXT         Block UUID
---   groupedErrorMessage   TEXT         Scrubbed error (IDs/URLs wildcarded)
---   errorMessage          TEXT         Raw error output (only set when FAILED)
---
--- WINDOW
---   Rolling 90 days (addedTime > CURRENT_DATE - 90 days)
---
--- EXAMPLE QUERIES
---   -- Most-used blocks by execution count
---   SELECT "blockName", COUNT(*) AS executions,
---          COUNT(*) FILTER (WHERE "executionStatus"='FAILED') AS failures
---   FROM analytics.node_block_execution
---   GROUP BY 1 ORDER BY executions DESC LIMIT 20;
---
---   -- Average LLM token usage per block
---   SELECT "blockName",
---          AVG("inputTokenCount") AS avg_input_tokens,
---          AVG("outputTokenCount") AS avg_output_tokens
---   FROM analytics.node_block_execution
---   WHERE "llmCallCount" > 0
---   GROUP BY 1 ORDER BY avg_input_tokens DESC;
---
---   -- Top failure reasons
---   SELECT "blockName", "groupedErrorMessage", COUNT(*) AS count
---   FROM analytics.node_block_execution
---   WHERE "executionStatus" = 'FAILED'
---   GROUP BY 1, 2 ORDER BY count DESC LIMIT 20;
--- =============================================================
-
-SELECT
-    ne."id"                                                            AS id,
-    ne."agentGraphExecutionId"                                         AS agentGraphExecutionId,
-    ne."agentNodeId"                                                   AS agentNodeId,
-    CAST(ne."executionStatus" AS TEXT)                                 AS executionStatus,
-    ne."addedTime"                                                     AS addedTime,
-    ne."queuedTime"                                                    AS queuedTime,
-    ne."startedTime"                                                   AS startedTime,
-    ne."endedTime"                                                     AS endedTime,
-    (ne."stats"::jsonb->>'input_size')::bigint                         AS inputSize,
-    (ne."stats"::jsonb->>'output_size')::bigint                        AS outputSize,
-    (ne."stats"::jsonb->>'walltime')::numeric                          AS walltime,
-    (ne."stats"::jsonb->>'cputime')::numeric                           AS cputime,
-    (ne."stats"::jsonb->>'llm_retry_count')::int                       AS llmRetryCount,
-    (ne."stats"::jsonb->>'llm_call_count')::int                        AS llmCallCount,
-    (ne."stats"::jsonb->>'input_token_count')::bigint                  AS inputTokenCount,
-    (ne."stats"::jsonb->>'output_token_count')::bigint                 AS outputTokenCount,
-    b."name"                                                           AS blockName,
-    b."id"                                                             AS blockId,
-    REGEXP_REPLACE(
-        REGEXP_REPLACE(
-            TRIM(BOTH '"' FROM eio."data"::text),
-            '(https?://)([A-Za-z0-9.-]+)(:[0-9]+)?(/[^\s]*)?',
-            '\1\2/...', 'gi'
-        ),
-        '[a-zA-Z0-9_:-]*\d[a-zA-Z0-9_:-]*', '*', 'g'
-    )                                                                  AS groupedErrorMessage,
-    eio."data"                                                         AS errorMessage
-FROM platform."AgentNodeExecution" ne
-LEFT JOIN platform."AgentNode" nd
-       ON ne."agentNodeId" = nd."id"
-LEFT JOIN platform."AgentBlock" b
-       ON nd."agentBlockId" = b."id"
-LEFT JOIN platform."AgentNodeExecutionInputOutput" eio
-       ON eio."referencedByOutputExecId" = ne."id"
-      AND eio."name" = 'error'
-      AND ne."executionStatus" = 'FAILED'
-WHERE ne."addedTime" > CURRENT_DATE - INTERVAL '90 days'

autogpt_platform/analytics/queries/platform_cost_log.sql

@@ -1,100 +0,0 @@
--- =============================================================
--- View: analytics.platform_cost_log
--- Looker source alias: ds115  |  Charts: 0
--- =============================================================
--- DESCRIPTION
---   One row per platform cost log entry (last 90 days).
---   Tracks real API spend at the call level: provider, model,
---   token counts (including Anthropic cache tokens), cost in
---   microdollars, and the block/execution that incurred the cost.
---   Joins the User table to provide email for per-user breakdowns.
---
--- SOURCE TABLES
---   platform.PlatformCostLog  — Per-call cost records
---   platform.User             — User email
---
--- OUTPUT COLUMNS
---   id                      TEXT         Log entry UUID
---   createdAt               TIMESTAMPTZ  When the cost was recorded
---   userId                  TEXT         User who incurred the cost (nullable)
---   email                   TEXT         User email (nullable)
---   graphExecId             TEXT         Graph execution UUID (nullable)
---   nodeExecId              TEXT         Node execution UUID (nullable)
---   blockName               TEXT         Block that made the API call (nullable)
---   provider                TEXT         API provider, lowercase (e.g. 'openai', 'anthropic')
---   model                   TEXT         Model name (nullable)
---   trackingType            TEXT         Cost unit: 'tokens' | 'cost_usd' | 'characters' | etc.
---   costMicrodollars        BIGINT       Cost in microdollars (divide by 1,000,000 for USD)
---   costUsd                 FLOAT        Cost in USD (costMicrodollars / 1,000,000)
---   inputTokens             INT          Prompt/input tokens (nullable)
---   outputTokens            INT          Completion/output tokens (nullable)
---   cacheReadTokens         INT          Anthropic cache-read tokens billed at 10% (nullable)
---   cacheCreationTokens     INT          Anthropic cache-write tokens billed at 125% (nullable)
---   totalTokens             INT          inputTokens + outputTokens (nullable if either is null)
---   duration                FLOAT        API call duration in seconds (nullable)
---
--- WINDOW
---   Rolling 90 days (createdAt > CURRENT_DATE - 90 days)
---
--- EXAMPLE QUERIES
---   -- Total spend by provider (last 90 days)
---   SELECT provider, SUM("costUsd") AS total_usd, COUNT(*) AS calls
---   FROM analytics.platform_cost_log
---   GROUP BY 1 ORDER BY total_usd DESC;
---
---   -- Spend by model
---   SELECT provider, model, SUM("costUsd") AS total_usd,
---          SUM("inputTokens") AS input_tokens,
---          SUM("outputTokens") AS output_tokens
---   FROM analytics.platform_cost_log
---   WHERE model IS NOT NULL
---   GROUP BY 1, 2 ORDER BY total_usd DESC;
---
---   -- Top 20 users by spend
---   SELECT "userId", email, SUM("costUsd") AS total_usd, COUNT(*) AS calls
---   FROM analytics.platform_cost_log
---   WHERE "userId" IS NOT NULL
---   GROUP BY 1, 2 ORDER BY total_usd DESC LIMIT 20;
---
---   -- Daily spend trend
---   SELECT DATE_TRUNC('day', "createdAt") AS day,
---          SUM("costUsd") AS daily_usd,
---          COUNT(*) AS calls
---   FROM analytics.platform_cost_log
---   GROUP BY 1 ORDER BY 1;
---
---   -- Cache hit rate for Anthropic (cache reads vs total reads)
---   SELECT DATE_TRUNC('day', "createdAt") AS day,
---          SUM("cacheReadTokens")::float /
---            NULLIF(SUM("inputTokens" + COALESCE("cacheReadTokens", 0)), 0) AS cache_hit_rate
---   FROM analytics.platform_cost_log
---   WHERE provider = 'anthropic'
---   GROUP BY 1 ORDER BY 1;
--- =============================================================
-
-SELECT
-    p."id"                                                        AS id,
-    p."createdAt"                                                 AS createdAt,
-    p."userId"                                                    AS userId,
-    u."email"                                                     AS email,
-    p."graphExecId"                                               AS graphExecId,
-    p."nodeExecId"                                                AS nodeExecId,
-    p."blockName"                                                 AS blockName,
-    p."provider"                                                  AS provider,
-    p."model"                                                     AS model,
-    p."trackingType"                                              AS trackingType,
-    p."costMicrodollars"                                          AS costMicrodollars,
-    p."costMicrodollars"::float / 1000000.0                       AS costUsd,
-    p."inputTokens"                                               AS inputTokens,
-    p."outputTokens"                                              AS outputTokens,
-    p."cacheReadTokens"                                           AS cacheReadTokens,
-    p."cacheCreationTokens"                                       AS cacheCreationTokens,
-    CASE
-        WHEN p."inputTokens" IS NOT NULL AND p."outputTokens" IS NOT NULL
-        THEN p."inputTokens" + p."outputTokens"
-        ELSE NULL
-    END                                                           AS totalTokens,
-    p."duration"                                                  AS duration
-FROM platform."PlatformCostLog" p
-LEFT JOIN platform."User" u ON u."id" = p."userId"
-WHERE p."createdAt" > CURRENT_DATE - INTERVAL '90 days'

autogpt_platform/analytics/queries/retention_agent.sql

@@ -1,97 +0,0 @@
--- =============================================================
--- View: analytics.retention_agent
--- Looker source alias: ds35  |  Charts: 2
--- =============================================================
--- DESCRIPTION
---   Weekly cohort retention broken down per individual agent.
---   Cohort = week of a user's first use of THAT specific agent.
---   Tells you which agents keep users coming back vs. one-shot
---   use. Only includes cohorts from the last 180 days.
---
--- SOURCE TABLES
---   platform.AgentGraphExecution  — Execution records (user × agent × time)
---   platform.AgentGraph           — Agent names
---
--- OUTPUT COLUMNS
---   agent_id            TEXT   Agent graph UUID
---   agent_label         TEXT   'AgentName [first8chars]'
---   agent_label_n       TEXT   'AgentName [first8chars] (n=total_users)'
---   cohort_week_start   DATE   Week users first ran this agent
---   cohort_label        TEXT   ISO week label
---   cohort_label_n      TEXT   ISO week label with cohort size
---   user_lifetime_week  INT    Weeks since first use of this agent
---   cohort_users        BIGINT Users in this cohort for this agent
---   active_users        BIGINT Users who ran the agent again in week k
---   retention_rate      FLOAT  active_users / cohort_users
---   cohort_users_w0     BIGINT cohort_users only at week 0 (safe to SUM)
---   agent_total_users   BIGINT Total users across all cohorts for this agent
---
--- EXAMPLE QUERIES
---   -- Best-retained agents at week 2
---   SELECT agent_label, AVG(retention_rate) AS w2_retention
---   FROM analytics.retention_agent
---   WHERE user_lifetime_week = 2 AND cohort_users >= 10
---   GROUP BY 1 ORDER BY w2_retention DESC LIMIT 10;
---
---   -- Agents with most unique users
---   SELECT DISTINCT agent_label, agent_total_users
---   FROM analytics.retention_agent
---   ORDER BY agent_total_users DESC LIMIT 20;
--- =============================================================
-
-WITH params AS (SELECT 12::int AS max_weeks, (CURRENT_DATE - INTERVAL '180 days') AS cohort_start),
-events AS (
-  SELECT e."userId"::text AS user_id, e."agentGraphId" AS agent_id,
-         e."createdAt"::timestamptz AS created_at,
-         DATE_TRUNC('week', e."createdAt")::date AS week_start
-  FROM platform."AgentGraphExecution" e
-),
-first_use AS (
-  SELECT user_id, agent_id, MIN(created_at) AS first_use_at,
-         DATE_TRUNC('week', MIN(created_at))::date AS cohort_week_start
-  FROM events GROUP BY 1,2
-  HAVING MIN(created_at) >= (SELECT cohort_start FROM params)
-),
-activity_weeks AS (SELECT DISTINCT user_id, agent_id, week_start FROM events),
-user_week_age AS (
-  SELECT aw.user_id, aw.agent_id, fu.cohort_week_start,
-         ((aw.week_start - DATE_TRUNC('week',fu.first_use_at)::date)/7)::int AS user_lifetime_week
-  FROM activity_weeks aw JOIN first_use fu USING (user_id, agent_id)
-  WHERE aw.week_start >= DATE_TRUNC('week',fu.first_use_at)::date
-),
-active_counts AS (
-  SELECT agent_id, cohort_week_start, user_lifetime_week, COUNT(DISTINCT user_id) AS active_users
-  FROM user_week_age WHERE user_lifetime_week >= 0 GROUP BY 1,2,3
-),
-cohort_sizes AS (
-  SELECT agent_id, cohort_week_start, COUNT(DISTINCT user_id) AS cohort_users FROM first_use GROUP BY 1,2
-),
-cohort_caps AS (
-  SELECT cs.agent_id, cs.cohort_week_start, cs.cohort_users,
-         LEAST((SELECT max_weeks FROM params),
-               GREATEST(0,((DATE_TRUNC('week',CURRENT_DATE)::date-cs.cohort_week_start)/7)::int)) AS cap_weeks
-  FROM cohort_sizes cs
-),
-grid AS (
-  SELECT cc.agent_id, cc.cohort_week_start, gs AS user_lifetime_week, cc.cohort_users
-  FROM cohort_caps cc CROSS JOIN LATERAL generate_series(0, cc.cap_weeks) gs
-),
-agent_names AS (SELECT DISTINCT ON (g."id") g."id" AS agent_id, g."name" AS agent_name FROM platform."AgentGraph" g ORDER BY g."id", g."version" DESC),
-agent_total_users AS (SELECT agent_id, SUM(cohort_users) AS agent_total_users FROM cohort_sizes GROUP BY 1)
-SELECT
-  g.agent_id,
-  COALESCE(an.agent_name,'(unnamed)')||' ['||LEFT(g.agent_id::text,8)||']'  AS agent_label,
-  COALESCE(an.agent_name,'(unnamed)')||' ['||LEFT(g.agent_id::text,8)||'] (n='||COALESCE(atu.agent_total_users,0)||')' AS agent_label_n,
-  g.cohort_week_start,
-  TO_CHAR(g.cohort_week_start,'IYYY-"W"IW')                               AS cohort_label,
-  TO_CHAR(g.cohort_week_start,'IYYY-"W"IW')||' (n='||g.cohort_users||')'  AS cohort_label_n,
-  g.user_lifetime_week, g.cohort_users,
-  COALESCE(ac.active_users,0)                                              AS active_users,
-  COALESCE(ac.active_users,0)::float / NULLIF(g.cohort_users,0)           AS retention_rate,
-  CASE WHEN g.user_lifetime_week=0 THEN g.cohort_users ELSE 0 END         AS cohort_users_w0,
-  COALESCE(atu.agent_total_users,0)                                        AS agent_total_users
-FROM grid g
-LEFT JOIN active_counts     ac  ON ac.agent_id=g.agent_id AND ac.cohort_week_start=g.cohort_week_start AND ac.user_lifetime_week=g.user_lifetime_week
-LEFT JOIN agent_names       an  ON an.agent_id=g.agent_id
-LEFT JOIN agent_total_users atu ON atu.agent_id=g.agent_id
-ORDER BY agent_label, g.cohort_week_start, g.user_lifetime_week;

autogpt_platform/analytics/queries/retention_execution_daily.sql

@@ -1,81 +0,0 @@
--- =============================================================
--- View: analytics.retention_execution_daily
--- Looker source alias: ds111  |  Charts: 1
--- =============================================================
--- DESCRIPTION
---   Daily cohort retention based on agent executions.
---   Cohort anchor = day of user's FIRST ever execution.
---   Only includes cohorts from the last 90 days, up to day 30.
---   Great for early engagement analysis (did users run another
---   agent the next day?).
---
--- SOURCE TABLES
---   platform.AgentGraphExecution  — Execution records
---
--- OUTPUT COLUMNS
---   Same pattern as retention_login_daily.
---   cohort_day_start = day of first execution (not first login)
---
--- EXAMPLE QUERIES
---   -- Day-3 execution retention
---   SELECT cohort_label, retention_rate_bounded AS d3_retention
---   FROM analytics.retention_execution_daily
---   WHERE user_lifetime_day = 3 ORDER BY cohort_day_start;
--- =============================================================
-
-WITH params AS (SELECT 30::int AS max_days, (CURRENT_DATE - INTERVAL '90 days') AS cohort_start),
-events AS (
-  SELECT e."userId"::text AS user_id, e."createdAt"::timestamptz AS created_at,
-         DATE_TRUNC('day', e."createdAt")::date AS day_start
-  FROM platform."AgentGraphExecution" e WHERE e."userId" IS NOT NULL
-),
-first_exec AS (
-  SELECT user_id, MIN(created_at) AS first_exec_at,
-         DATE_TRUNC('day', MIN(created_at))::date AS cohort_day_start
-  FROM events GROUP BY 1
-  HAVING MIN(created_at) >= (SELECT cohort_start FROM params)
-),
-activity_days AS (SELECT DISTINCT user_id, day_start FROM events),
-user_day_age AS (
-  SELECT ad.user_id, fe.cohort_day_start,
-         (ad.day_start - DATE_TRUNC('day',fe.first_exec_at)::date)::int AS user_lifetime_day
-  FROM activity_days ad JOIN first_exec fe USING (user_id)
-  WHERE ad.day_start >= DATE_TRUNC('day',fe.first_exec_at)::date
-),
-bounded_counts AS (
-  SELECT cohort_day_start, user_lifetime_day, COUNT(DISTINCT user_id) AS active_users_bounded
-  FROM user_day_age WHERE user_lifetime_day >= 0 GROUP BY 1,2
-),
-last_active AS (
-  SELECT cohort_day_start, user_id, MAX(user_lifetime_day) AS last_active_day FROM user_day_age GROUP BY 1,2
-),
-unbounded_counts AS (
-  SELECT la.cohort_day_start, gs AS user_lifetime_day, COUNT(*) AS retained_users_unbounded
-  FROM last_active la
-  CROSS JOIN LATERAL generate_series(0, LEAST(la.last_active_day,(SELECT max_days FROM params))) gs
-  GROUP BY 1,2
-),
-cohort_sizes AS (SELECT cohort_day_start, COUNT(DISTINCT user_id) AS cohort_users FROM first_exec GROUP BY 1),
-cohort_caps AS (
-  SELECT cs.cohort_day_start, cs.cohort_users,
-         LEAST((SELECT max_days FROM params), GREATEST(0,(CURRENT_DATE-cs.cohort_day_start)::int)) AS cap_days
-  FROM cohort_sizes cs
-),
-grid AS (
-  SELECT cc.cohort_day_start, gs AS user_lifetime_day, cc.cohort_users
-  FROM cohort_caps cc CROSS JOIN LATERAL generate_series(0, cc.cap_days) gs
-)
-SELECT
-  g.cohort_day_start,
-  TO_CHAR(g.cohort_day_start,'YYYY-MM-DD')                                AS cohort_label,
-  TO_CHAR(g.cohort_day_start,'YYYY-MM-DD')||' (n='||g.cohort_users||')'   AS cohort_label_n,
-  g.user_lifetime_day, g.cohort_users,
-  COALESCE(b.active_users_bounded,0)     AS active_users_bounded,
-  COALESCE(u.retained_users_unbounded,0) AS retained_users_unbounded,
-  CASE WHEN g.cohort_users>0 THEN COALESCE(b.active_users_bounded,0)::float/g.cohort_users END    AS retention_rate_bounded,
-  CASE WHEN g.cohort_users>0 THEN COALESCE(u.retained_users_unbounded,0)::float/g.cohort_users END AS retention_rate_unbounded,
-  CASE WHEN g.user_lifetime_day=0 THEN g.cohort_users ELSE 0 END          AS cohort_users_d0
-FROM grid g
-LEFT JOIN bounded_counts   b ON b.cohort_day_start=g.cohort_day_start AND b.user_lifetime_day=g.user_lifetime_day
-LEFT JOIN unbounded_counts u ON u.cohort_day_start=g.cohort_day_start AND u.user_lifetime_day=g.user_lifetime_day
-ORDER BY g.cohort_day_start, g.user_lifetime_day;

autogpt_platform/analytics/queries/retention_execution_weekly.sql

@@ -1,81 +0,0 @@
--- =============================================================
--- View: analytics.retention_execution_weekly
--- Looker source alias: ds92  |  Charts: 2
--- =============================================================
--- DESCRIPTION
---   Weekly cohort retention based on agent executions.
---   Cohort anchor = week of user's FIRST ever agent execution
---   (not first login). Only includes cohorts from the last 180 days.
---   Useful when you care about product engagement, not just visits.
---
--- SOURCE TABLES
---   platform.AgentGraphExecution  — Execution records
---
--- OUTPUT COLUMNS
---   Same pattern as retention_login_weekly.
---   cohort_week_start = week of first execution (not first login)
---
--- EXAMPLE QUERIES
---   -- Week-2 execution retention
---   SELECT cohort_label, retention_rate_bounded
---   FROM analytics.retention_execution_weekly
---   WHERE user_lifetime_week = 2 ORDER BY cohort_week_start;
--- =============================================================
-
-WITH params AS (SELECT 12::int AS max_weeks, (CURRENT_DATE - INTERVAL '180 days') AS cohort_start),
-events AS (
-  SELECT e."userId"::text AS user_id, e."createdAt"::timestamptz AS created_at,
-         DATE_TRUNC('week', e."createdAt")::date AS week_start
-  FROM platform."AgentGraphExecution" e WHERE e."userId" IS NOT NULL
-),
-first_exec AS (
-  SELECT user_id, MIN(created_at) AS first_exec_at,
-         DATE_TRUNC('week', MIN(created_at))::date AS cohort_week_start
-  FROM events GROUP BY 1
-  HAVING MIN(created_at) >= (SELECT cohort_start FROM params)
-),
-activity_weeks AS (SELECT DISTINCT user_id, week_start FROM events),
-user_week_age AS (
-  SELECT aw.user_id, fe.cohort_week_start,
-         ((aw.week_start - DATE_TRUNC('week',fe.first_exec_at)::date)/7)::int AS user_lifetime_week
-  FROM activity_weeks aw JOIN first_exec fe USING (user_id)
-  WHERE aw.week_start >= DATE_TRUNC('week',fe.first_exec_at)::date
-),
-bounded_counts AS (
-  SELECT cohort_week_start, user_lifetime_week, COUNT(DISTINCT user_id) AS active_users_bounded
-  FROM user_week_age WHERE user_lifetime_week >= 0 GROUP BY 1,2
-),
-last_active AS (
-  SELECT cohort_week_start, user_id, MAX(user_lifetime_week) AS last_active_week FROM user_week_age GROUP BY 1,2
-),
-unbounded_counts AS (
-  SELECT la.cohort_week_start, gs AS user_lifetime_week, COUNT(*) AS retained_users_unbounded
-  FROM last_active la
-  CROSS JOIN LATERAL generate_series(0, LEAST(la.last_active_week,(SELECT max_weeks FROM params))) gs
-  GROUP BY 1,2
-),
-cohort_sizes AS (SELECT cohort_week_start, COUNT(DISTINCT user_id) AS cohort_users FROM first_exec GROUP BY 1),
-cohort_caps AS (
-  SELECT cs.cohort_week_start, cs.cohort_users,
-         LEAST((SELECT max_weeks FROM params),
-               GREATEST(0,((DATE_TRUNC('week',CURRENT_DATE)::date-cs.cohort_week_start)/7)::int)) AS cap_weeks
-  FROM cohort_sizes cs
-),
-grid AS (
-  SELECT cc.cohort_week_start, gs AS user_lifetime_week, cc.cohort_users
-  FROM cohort_caps cc CROSS JOIN LATERAL generate_series(0, cc.cap_weeks) gs
-)
-SELECT
-  g.cohort_week_start,
-  TO_CHAR(g.cohort_week_start,'IYYY-"W"IW')                               AS cohort_label,
-  TO_CHAR(g.cohort_week_start,'IYYY-"W"IW')||' (n='||g.cohort_users||')'  AS cohort_label_n,
-  g.user_lifetime_week, g.cohort_users,
-  COALESCE(b.active_users_bounded,0)     AS active_users_bounded,
-  COALESCE(u.retained_users_unbounded,0) AS retained_users_unbounded,
-  CASE WHEN g.cohort_users>0 THEN COALESCE(b.active_users_bounded,0)::float/g.cohort_users END    AS retention_rate_bounded,
-  CASE WHEN g.cohort_users>0 THEN COALESCE(u.retained_users_unbounded,0)::float/g.cohort_users END AS retention_rate_unbounded,
-  CASE WHEN g.user_lifetime_week=0 THEN g.cohort_users ELSE 0 END         AS cohort_users_w0
-FROM grid g
-LEFT JOIN bounded_counts   b ON b.cohort_week_start=g.cohort_week_start AND b.user_lifetime_week=g.user_lifetime_week
-LEFT JOIN unbounded_counts u ON u.cohort_week_start=g.cohort_week_start AND u.user_lifetime_week=g.user_lifetime_week
-ORDER BY g.cohort_week_start, g.user_lifetime_week;

autogpt_platform/analytics/queries/retention_login_daily.sql

@@ -1,94 +0,0 @@
--- =============================================================
--- View: analytics.retention_login_daily
--- Looker source alias: ds112  |  Charts: 1
--- =============================================================
--- DESCRIPTION
---   Daily cohort retention based on login sessions.
---   Same logic as retention_login_weekly but at day granularity,
---   showing up to day 30 for cohorts from the last 90 days.
---   Useful for analysing early activation (days 1-7) in detail.
---
--- SOURCE TABLES
---   auth.sessions  — Login session records
---
--- OUTPUT COLUMNS (same pattern as retention_login_weekly)
---   cohort_day_start          DATE     First day the cohort logged in
---   cohort_label              TEXT     Date string (e.g. '2025-03-01')
---   cohort_label_n            TEXT     Date + cohort size (e.g. '2025-03-01 (n=12)')
---   user_lifetime_day         INT      Days since first login (0 = signup day)
---   cohort_users              BIGINT   Total users in cohort
---   active_users_bounded      BIGINT   Users active on exactly day k
---   retained_users_unbounded  BIGINT   Users active any time on/after day k
---   retention_rate_bounded    FLOAT    bounded / cohort_users
---   retention_rate_unbounded  FLOAT    unbounded / cohort_users
---   cohort_users_d0           BIGINT   cohort_users only at day 0, else 0 (safe to SUM)
---
--- EXAMPLE QUERIES
---   -- Day-1 retention rate (came back next day)
---   SELECT cohort_label, retention_rate_bounded AS d1_retention
---   FROM analytics.retention_login_daily
---   WHERE user_lifetime_day = 1 ORDER BY cohort_day_start;
---
---   -- Average retention curve across all cohorts
---   SELECT user_lifetime_day,
---          SUM(active_users_bounded)::float / NULLIF(SUM(cohort_users_d0), 0) AS avg_retention
---   FROM analytics.retention_login_daily
---   GROUP BY 1 ORDER BY 1;
--- =============================================================
-
-WITH params AS (SELECT 30::int AS max_days, (CURRENT_DATE - INTERVAL '90 days')::date AS cohort_start),
-events AS (
-  SELECT s.user_id::text AS user_id, s.created_at::timestamptz AS created_at,
-         DATE_TRUNC('day', s.created_at)::date AS day_start
-  FROM auth.sessions s WHERE s.user_id IS NOT NULL
-),
-first_login AS (
-  SELECT user_id, MIN(created_at) AS first_login_time,
-         DATE_TRUNC('day', MIN(created_at))::date AS cohort_day_start
-  FROM events GROUP BY 1
-  HAVING MIN(created_at) >= (SELECT cohort_start FROM params)
-),
-activity_days AS (SELECT DISTINCT user_id, day_start FROM events),
-user_day_age AS (
-  SELECT ad.user_id, fl.cohort_day_start,
-         (ad.day_start - DATE_TRUNC('day', fl.first_login_time)::date)::int AS user_lifetime_day
-  FROM activity_days ad JOIN first_login fl USING (user_id)
-  WHERE ad.day_start >= DATE_TRUNC('day', fl.first_login_time)::date
-),
-bounded_counts AS (
-  SELECT cohort_day_start, user_lifetime_day, COUNT(DISTINCT user_id) AS active_users_bounded
-  FROM user_day_age WHERE user_lifetime_day >= 0 GROUP BY 1,2
-),
-last_active AS (
-  SELECT cohort_day_start, user_id, MAX(user_lifetime_day) AS last_active_day FROM user_day_age GROUP BY 1,2
-),
-unbounded_counts AS (
-  SELECT la.cohort_day_start, gs AS user_lifetime_day, COUNT(*) AS retained_users_unbounded
-  FROM last_active la
-  CROSS JOIN LATERAL generate_series(0, LEAST(la.last_active_day,(SELECT max_days FROM params))) gs
-  GROUP BY 1,2
-),
-cohort_sizes AS (SELECT cohort_day_start, COUNT(DISTINCT user_id) AS cohort_users FROM first_login GROUP BY 1),
-cohort_caps AS (
-  SELECT cs.cohort_day_start, cs.cohort_users,
-         LEAST((SELECT max_days FROM params), GREATEST(0,(CURRENT_DATE-cs.cohort_day_start)::int)) AS cap_days
-  FROM cohort_sizes cs
-),
-grid AS (
-  SELECT cc.cohort_day_start, gs AS user_lifetime_day, cc.cohort_users
-  FROM cohort_caps cc CROSS JOIN LATERAL generate_series(0, cc.cap_days) gs
-)
-SELECT
-  g.cohort_day_start,
-  TO_CHAR(g.cohort_day_start,'YYYY-MM-DD')                                  AS cohort_label,
-  TO_CHAR(g.cohort_day_start,'YYYY-MM-DD')||' (n='||g.cohort_users||')'     AS cohort_label_n,
-  g.user_lifetime_day, g.cohort_users,
-  COALESCE(b.active_users_bounded,0)     AS active_users_bounded,
-  COALESCE(u.retained_users_unbounded,0) AS retained_users_unbounded,
-  CASE WHEN g.cohort_users>0 THEN COALESCE(b.active_users_bounded,0)::float/g.cohort_users END    AS retention_rate_bounded,
-  CASE WHEN g.cohort_users>0 THEN COALESCE(u.retained_users_unbounded,0)::float/g.cohort_users END AS retention_rate_unbounded,
-  CASE WHEN g.user_lifetime_day=0 THEN g.cohort_users ELSE 0 END            AS cohort_users_d0
-FROM grid g
-LEFT JOIN bounded_counts   b ON b.cohort_day_start=g.cohort_day_start AND b.user_lifetime_day=g.user_lifetime_day
-LEFT JOIN unbounded_counts u ON u.cohort_day_start=g.cohort_day_start AND u.user_lifetime_day=g.user_lifetime_day
-ORDER BY g.cohort_day_start, g.user_lifetime_day;

autogpt_platform/analytics/queries/retention_login_onboarded_weekly.sql

@@ -1,96 +0,0 @@
--- =============================================================
--- View: analytics.retention_login_onboarded_weekly
--- Looker source alias: ds101  |  Charts: 2
--- =============================================================
--- DESCRIPTION
---   Weekly cohort retention from login sessions, restricted to
---   users who "onboarded" — defined as running at least one
---   agent within 365 days of their first login.
---   Filters out users who signed up but never activated,
---   giving a cleaner view of engaged-user retention.
---
--- SOURCE TABLES
---   auth.sessions                  — Login session records
---   platform.AgentGraphExecution   — Used to identify onboarders
---
--- OUTPUT COLUMNS
---   Same as retention_login_weekly (cohort_week_start, user_lifetime_week,
---   retention_rate_bounded, retention_rate_unbounded, etc.)
---   Only difference: cohort is filtered to onboarded users only.
---
--- EXAMPLE QUERIES
---   -- Compare week-4 retention: all users vs onboarded only
---   SELECT 'all_users' AS segment, AVG(retention_rate_bounded) AS w4_retention
---   FROM analytics.retention_login_weekly WHERE user_lifetime_week = 4
---   UNION ALL
---   SELECT 'onboarded', AVG(retention_rate_bounded)
---   FROM analytics.retention_login_onboarded_weekly WHERE user_lifetime_week = 4;
--- =============================================================
-
-WITH params AS (SELECT 12::int AS max_weeks, 365::int AS onboarding_window_days),
-events AS (
-  SELECT s.user_id::text AS user_id, s.created_at::timestamptz AS created_at,
-         DATE_TRUNC('week', s.created_at)::date AS week_start
-  FROM auth.sessions s WHERE s.user_id IS NOT NULL
-),
-first_login_all AS (
-  SELECT user_id, MIN(created_at) AS first_login_time,
-         DATE_TRUNC('week', MIN(created_at))::date AS cohort_week_start
-  FROM events GROUP BY 1
-),
-onboarders AS (
-  SELECT fl.user_id FROM first_login_all fl
-  WHERE EXISTS (
-    SELECT 1 FROM platform."AgentGraphExecution" e
-    WHERE e."userId"::text = fl.user_id
-      AND e."createdAt" >= fl.first_login_time
-      AND e."createdAt" < fl.first_login_time
-          + make_interval(days => (SELECT onboarding_window_days FROM params))
-  )
-),
-first_login AS (SELECT * FROM first_login_all WHERE user_id IN (SELECT user_id FROM onboarders)),
-activity_weeks AS (SELECT DISTINCT user_id, week_start FROM events),
-user_week_age AS (
-  SELECT aw.user_id, fl.cohort_week_start,
-         ((aw.week_start - DATE_TRUNC('week',fl.first_login_time)::date)/7)::int AS user_lifetime_week
-  FROM activity_weeks aw JOIN first_login fl USING (user_id)
-  WHERE aw.week_start >= DATE_TRUNC('week',fl.first_login_time)::date
-),
-bounded_counts AS (
-  SELECT cohort_week_start, user_lifetime_week, COUNT(DISTINCT user_id) AS active_users_bounded
-  FROM user_week_age WHERE user_lifetime_week >= 0 GROUP BY 1,2
-),
-last_active AS (
-  SELECT cohort_week_start, user_id, MAX(user_lifetime_week) AS last_active_week FROM user_week_age GROUP BY 1,2
-),
-unbounded_counts AS (
-  SELECT la.cohort_week_start, gs AS user_lifetime_week, COUNT(*) AS retained_users_unbounded
-  FROM last_active la
-  CROSS JOIN LATERAL generate_series(0, LEAST(la.last_active_week,(SELECT max_weeks FROM params))) gs
-  GROUP BY 1,2
-),
-cohort_sizes AS (SELECT cohort_week_start, COUNT(DISTINCT user_id) AS cohort_users FROM first_login GROUP BY 1),
-cohort_caps AS (
-  SELECT cs.cohort_week_start, cs.cohort_users,
-         LEAST((SELECT max_weeks FROM params),
-               GREATEST(0,((DATE_TRUNC('week',CURRENT_DATE)::date-cs.cohort_week_start)/7)::int)) AS cap_weeks
-  FROM cohort_sizes cs
-),
-grid AS (
-  SELECT cc.cohort_week_start, gs AS user_lifetime_week, cc.cohort_users
-  FROM cohort_caps cc CROSS JOIN LATERAL generate_series(0, cc.cap_weeks) gs
-)
-SELECT
-  g.cohort_week_start,
-  TO_CHAR(g.cohort_week_start,'IYYY-"W"IW')                               AS cohort_label,
-  TO_CHAR(g.cohort_week_start,'IYYY-"W"IW')||' (n='||g.cohort_users||')'  AS cohort_label_n,
-  g.user_lifetime_week, g.cohort_users,
-  COALESCE(b.active_users_bounded,0)     AS active_users_bounded,
-  COALESCE(u.retained_users_unbounded,0) AS retained_users_unbounded,
-  CASE WHEN g.cohort_users>0 THEN COALESCE(b.active_users_bounded,0)::float/g.cohort_users END    AS retention_rate_bounded,
-  CASE WHEN g.cohort_users>0 THEN COALESCE(u.retained_users_unbounded,0)::float/g.cohort_users END AS retention_rate_unbounded,
-  CASE WHEN g.user_lifetime_week=0 THEN g.cohort_users ELSE 0 END         AS cohort_users_w0
-FROM grid g
-LEFT JOIN bounded_counts   b ON b.cohort_week_start=g.cohort_week_start AND b.user_lifetime_week=g.user_lifetime_week
-LEFT JOIN unbounded_counts u ON u.cohort_week_start=g.cohort_week_start AND u.user_lifetime_week=g.user_lifetime_week
-ORDER BY g.cohort_week_start, g.user_lifetime_week;

autogpt_platform/analytics/queries/retention_login_weekly.sql

@@ -1,103 +0,0 @@
--- =============================================================
--- View: analytics.retention_login_weekly
--- Looker source alias: ds83  |  Charts: 2
--- =============================================================
--- DESCRIPTION
---   Weekly cohort retention based on login sessions.
---   Users are grouped by the ISO week of their first ever login.
---   For each cohort × lifetime-week combination, outputs both:
---     - bounded rate: % active in exactly that week
---     - unbounded rate: % who were ever active on or after that week
---   Weeks are capped to the cohort's actual age (no future data points).
---
--- SOURCE TABLES
---   auth.sessions  — Login session records
---
--- HOW TO READ THE OUTPUT
---   cohort_week_start   The Monday of the week users first logged in
---   user_lifetime_week  0 = signup week, 1 = one week later, etc.
---   retention_rate_bounded   = active_users_bounded / cohort_users
---   retention_rate_unbounded = retained_users_unbounded / cohort_users
---
--- OUTPUT COLUMNS
---   cohort_week_start         DATE     First day of the cohort's signup week
---   cohort_label              TEXT     ISO week label (e.g. '2025-W01')
---   cohort_label_n            TEXT     ISO week label with cohort size (e.g. '2025-W01 (n=42)')
---   user_lifetime_week        INT      Weeks since first login (0 = signup week)
---   cohort_users              BIGINT   Total users in this cohort (denominator)
---   active_users_bounded      BIGINT   Users active in exactly week k
---   retained_users_unbounded  BIGINT   Users active any time on/after week k
---   retention_rate_bounded    FLOAT    bounded active / cohort_users
---   retention_rate_unbounded  FLOAT    unbounded retained / cohort_users
---   cohort_users_w0           BIGINT   cohort_users only at week 0, else 0 (safe to SUM in pivot tables)
---
--- EXAMPLE QUERIES
---   -- Week-1 retention rate per cohort
---   SELECT cohort_label, retention_rate_bounded AS w1_retention
---   FROM analytics.retention_login_weekly
---   WHERE user_lifetime_week = 1
---   ORDER BY cohort_week_start;
---
---   -- Overall average retention curve (all cohorts combined)
---   SELECT user_lifetime_week,
---          SUM(active_users_bounded)::float / NULLIF(SUM(cohort_users_w0), 0) AS avg_retention
---   FROM analytics.retention_login_weekly
---   GROUP BY 1 ORDER BY 1;
--- =============================================================
-
-WITH params AS (SELECT 12::int AS max_weeks),
-events AS (
-  SELECT s.user_id::text AS user_id, s.created_at::timestamptz AS created_at,
-         DATE_TRUNC('week', s.created_at)::date AS week_start
-  FROM auth.sessions s WHERE s.user_id IS NOT NULL
-),
-first_login AS (
-  SELECT user_id, MIN(created_at) AS first_login_time,
-         DATE_TRUNC('week', MIN(created_at))::date AS cohort_week_start
-  FROM events GROUP BY 1
-),
-activity_weeks AS (SELECT DISTINCT user_id, week_start FROM events),
-user_week_age AS (
-  SELECT aw.user_id, fl.cohort_week_start,
-         ((aw.week_start - DATE_TRUNC('week', fl.first_login_time)::date) / 7)::int AS user_lifetime_week
-  FROM activity_weeks aw JOIN first_login fl USING (user_id)
-  WHERE aw.week_start >= DATE_TRUNC('week', fl.first_login_time)::date
-),
-bounded_counts AS (
-  SELECT cohort_week_start, user_lifetime_week, COUNT(DISTINCT user_id) AS active_users_bounded
-  FROM user_week_age WHERE user_lifetime_week >= 0 GROUP BY 1,2
-),
-last_active AS (
-  SELECT cohort_week_start, user_id, MAX(user_lifetime_week) AS last_active_week FROM user_week_age GROUP BY 1,2
-),
-unbounded_counts AS (
-  SELECT la.cohort_week_start, gs AS user_lifetime_week, COUNT(*) AS retained_users_unbounded
-  FROM last_active la
-  CROSS JOIN LATERAL generate_series(0, LEAST(la.last_active_week,(SELECT max_weeks FROM params))) gs
-  GROUP BY 1,2
-),
-cohort_sizes AS (SELECT cohort_week_start, COUNT(DISTINCT user_id) AS cohort_users FROM first_login GROUP BY 1),
-cohort_caps AS (
-  SELECT cs.cohort_week_start, cs.cohort_users,
-         LEAST((SELECT max_weeks FROM params),
-               GREATEST(0,((DATE_TRUNC('week',CURRENT_DATE)::date - cs.cohort_week_start)/7)::int)) AS cap_weeks
-  FROM cohort_sizes cs
-),
-grid AS (
-  SELECT cc.cohort_week_start, gs AS user_lifetime_week, cc.cohort_users
-  FROM cohort_caps cc CROSS JOIN LATERAL generate_series(0, cc.cap_weeks) gs
-)
-SELECT
-  g.cohort_week_start,
-  TO_CHAR(g.cohort_week_start,'IYYY-"W"IW')                                    AS cohort_label,
-  TO_CHAR(g.cohort_week_start,'IYYY-"W"IW')||' (n='||g.cohort_users||')'       AS cohort_label_n,
-  g.user_lifetime_week, g.cohort_users,
-  COALESCE(b.active_users_bounded,0)     AS active_users_bounded,
-  COALESCE(u.retained_users_unbounded,0) AS retained_users_unbounded,
-  CASE WHEN g.cohort_users>0 THEN COALESCE(b.active_users_bounded,0)::float/g.cohort_users END    AS retention_rate_bounded,
-  CASE WHEN g.cohort_users>0 THEN COALESCE(u.retained_users_unbounded,0)::float/g.cohort_users END AS retention_rate_unbounded,
-  CASE WHEN g.user_lifetime_week=0 THEN g.cohort_users ELSE 0 END               AS cohort_users_w0
-FROM grid g
-LEFT JOIN bounded_counts   b ON b.cohort_week_start=g.cohort_week_start AND b.user_lifetime_week=g.user_lifetime_week
-LEFT JOIN unbounded_counts u ON u.cohort_week_start=g.cohort_week_start AND u.user_lifetime_week=g.user_lifetime_week
-ORDER BY g.cohort_week_start, g.user_lifetime_week

autogpt_platform/analytics/queries/user_block_spending.sql

@@ -1,71 +0,0 @@
--- =============================================================
--- View: analytics.user_block_spending
--- Looker source alias: ds6  |  Charts: 5
--- =============================================================
--- DESCRIPTION
---   One row per credit transaction (last 90 days).
---   Shows how users spend credits broken down by block type,
---   LLM provider and model.  Joins node execution stats for
---   token-level detail.
---
--- SOURCE TABLES
---   platform.CreditTransaction   — Credit debit/credit records
---   platform.AgentNodeExecution  — Node execution stats (for token counts)
---
--- OUTPUT COLUMNS
---   transactionKey        TEXT         Unique transaction identifier
---   userId                TEXT         User who was charged
---   amount                DECIMAL      Credit amount (positive = credit, negative = debit)
---   negativeAmount        DECIMAL      amount * -1 (convenience for spend charts)
---   transactionType       TEXT         Transaction type (e.g. 'USAGE', 'REFUND', 'TOP_UP')
---   transactionTime       TIMESTAMPTZ  When the transaction was recorded
---   blockId               TEXT         Block UUID that triggered the spend
---   blockName             TEXT         Human-readable block name
---   llm_provider          TEXT         LLM provider (e.g. 'openai', 'anthropic')
---   llm_model             TEXT         Model name (e.g. 'gpt-4o', 'claude-3-5-sonnet')
---   node_exec_id          TEXT         Linked node execution UUID
---   llm_call_count        INT          LLM API calls made in that execution
---   llm_retry_count       INT          LLM retries in that execution
---   llm_input_token_count INT          Input tokens consumed
---   llm_output_token_count INT         Output tokens produced
---
--- WINDOW
---   Rolling 90 days (createdAt > CURRENT_DATE - 90 days)
---
--- EXAMPLE QUERIES
---   -- Total spend per user (last 90 days)
---   SELECT "userId", SUM("negativeAmount") AS total_spent
---   FROM analytics.user_block_spending
---   WHERE "transactionType" = 'USAGE'
---   GROUP BY 1 ORDER BY total_spent DESC;
---
---   -- Spend by LLM provider + model
---   SELECT "llm_provider", "llm_model",
---          SUM("negativeAmount") AS total_cost,
---          SUM("llm_input_token_count") AS input_tokens,
---          SUM("llm_output_token_count") AS output_tokens
---   FROM analytics.user_block_spending
---   WHERE "llm_provider" IS NOT NULL
---   GROUP BY 1, 2 ORDER BY total_cost DESC;
--- =============================================================
-
-SELECT
-    c."transactionKey"                                        AS transactionKey,
-    c."userId"                                                AS userId,
-    c."amount"                                                AS amount,
-    c."amount" * -1                                           AS negativeAmount,
-    c."type"                                                  AS transactionType,
-    c."createdAt"                                             AS transactionTime,
-    c.metadata->>'block_id'                                   AS blockId,
-    c.metadata->>'block'                                      AS blockName,
-    c.metadata->'input'->'credentials'->>'provider'           AS llm_provider,
-    c.metadata->'input'->>'model'                             AS llm_model,
-    c.metadata->>'node_exec_id'                               AS node_exec_id,
-    (ne."stats"->>'llm_call_count')::int                       AS llm_call_count,
-    (ne."stats"->>'llm_retry_count')::int                      AS llm_retry_count,
-    (ne."stats"->>'input_token_count')::int                    AS llm_input_token_count,
-    (ne."stats"->>'output_token_count')::int                   AS llm_output_token_count
-FROM platform."CreditTransaction" c
-LEFT JOIN platform."AgentNodeExecution" ne
-       ON (c.metadata->>'node_exec_id') = ne."id"::text
-WHERE c."createdAt" > CURRENT_DATE - INTERVAL '90 days'

autogpt_platform/analytics/queries/user_onboarding.sql

@@ -1,45 +0,0 @@
--- =============================================================
--- View: analytics.user_onboarding
--- Looker source alias: ds68  |  Charts: 3
--- =============================================================
--- DESCRIPTION
---   One row per user onboarding record.  Contains the user's
---   stated usage reason, selected integrations, completed
---   onboarding steps and optional first agent selection.
---   Full history (no date filter) since onboarding happens
---   once per user.
---
--- SOURCE TABLES
---   platform.UserOnboarding  — Onboarding state per user
---
--- OUTPUT COLUMNS
---   id                            TEXT         Onboarding record UUID
---   createdAt                     TIMESTAMPTZ  When onboarding started
---   updatedAt                     TIMESTAMPTZ  Last update to onboarding state
---   usageReason                   TEXT         Why user signed up (e.g. 'work', 'personal')
---   integrations                  TEXT[]       Array of integration names the user selected
---   userId                        TEXT         User UUID
---   completedSteps                TEXT[]       Array of onboarding step enums completed
---   selectedStoreListingVersionId TEXT         First marketplace agent the user chose (if any)
---
--- EXAMPLE QUERIES
---   -- Usage reason breakdown
---   SELECT "usageReason", COUNT(*) FROM analytics.user_onboarding GROUP BY 1;
---
---   -- Completion rate per step
---   SELECT step, COUNT(*) AS users_completed
---   FROM analytics.user_onboarding
---   CROSS JOIN LATERAL UNNEST("completedSteps") AS step
---   GROUP BY 1 ORDER BY users_completed DESC;
--- =============================================================
-
-SELECT
-    id,
-    "createdAt",
-    "updatedAt",
-    "usageReason",
-    integrations,
-    "userId",
-    "completedSteps",
-    "selectedStoreListingVersionId"
-FROM platform."UserOnboarding"

autogpt_platform/analytics/queries/user_onboarding_funnel.sql

@@ -1,100 +0,0 @@
--- =============================================================
--- View: analytics.user_onboarding_funnel
--- Looker source alias: ds74  |  Charts: 1
--- =============================================================
--- DESCRIPTION
---   Pre-aggregated onboarding funnel showing how many users
---   completed each step and the drop-off percentage from the
---   previous step.  One row per onboarding step (all 22 steps
---   always present, even with 0 completions — prevents sparse
---   gaps from making LAG compare the wrong predecessors).
---
--- SOURCE TABLES
---   platform.UserOnboarding  — Onboarding records with completedSteps array
---
--- OUTPUT COLUMNS
---   step             TEXT     Onboarding step enum name (e.g. 'WELCOME', 'CONGRATS')
---   step_order       INT      Numeric position in the funnel (1=first, 22=last)
---   users_completed  BIGINT   Distinct users who completed this step
---   pct_from_prev    NUMERIC  % of users from the previous step who reached this one
---
--- STEP ORDER
---   1  WELCOME               9  MARKETPLACE_VISIT     17  SCHEDULE_AGENT
---   2  USAGE_REASON         10  MARKETPLACE_ADD_AGENT  18  RUN_AGENTS
---   3  INTEGRATIONS         11  MARKETPLACE_RUN_AGENT  19  RUN_3_DAYS
---   4  AGENT_CHOICE         12  BUILDER_OPEN           20  TRIGGER_WEBHOOK
---   5  AGENT_NEW_RUN        13  BUILDER_SAVE_AGENT     21  RUN_14_DAYS
---   6  AGENT_INPUT          14  BUILDER_RUN_AGENT      22  RUN_AGENTS_100
---   7  CONGRATS             15  VISIT_COPILOT
---   8  GET_RESULTS          16  RE_RUN_AGENT
---
--- WINDOW
---   Users who started onboarding in the last 90 days
---
--- EXAMPLE QUERIES
---   -- Full funnel
---   SELECT * FROM analytics.user_onboarding_funnel ORDER BY step_order;
---
---   -- Biggest drop-off point
---   SELECT step, pct_from_prev FROM analytics.user_onboarding_funnel
---   ORDER BY pct_from_prev ASC LIMIT 3;
--- =============================================================
-
-WITH all_steps AS (
-  -- Complete ordered grid of all 22 steps so zero-completion steps
-  -- are always present, keeping LAG comparisons correct.
-  SELECT step_name, step_order
-  FROM (VALUES
-    ('WELCOME',               1),
-    ('USAGE_REASON',          2),
-    ('INTEGRATIONS',          3),
-    ('AGENT_CHOICE',          4),
-    ('AGENT_NEW_RUN',         5),
-    ('AGENT_INPUT',           6),
-    ('CONGRATS',              7),
-    ('GET_RESULTS',           8),
-    ('MARKETPLACE_VISIT',     9),
-    ('MARKETPLACE_ADD_AGENT', 10),
-    ('MARKETPLACE_RUN_AGENT', 11),
-    ('BUILDER_OPEN',          12),
-    ('BUILDER_SAVE_AGENT',    13),
-    ('BUILDER_RUN_AGENT',     14),
-    ('VISIT_COPILOT',         15),
-    ('RE_RUN_AGENT',          16),
-    ('SCHEDULE_AGENT',        17),
-    ('RUN_AGENTS',            18),
-    ('RUN_3_DAYS',            19),
-    ('TRIGGER_WEBHOOK',       20),
-    ('RUN_14_DAYS',           21),
-    ('RUN_AGENTS_100',        22)
-  ) AS t(step_name, step_order)
-),
-raw AS (
-  SELECT
-      u."userId",
-      step_txt::text AS step
-  FROM platform."UserOnboarding" u
-  CROSS JOIN LATERAL UNNEST(u."completedSteps") AS step_txt
-  WHERE u."createdAt" >= CURRENT_DATE - INTERVAL '90 days'
-),
-step_counts AS (
-  SELECT step, COUNT(DISTINCT "userId") AS users_completed
-  FROM raw GROUP BY step
-),
-funnel AS (
-  SELECT
-      a.step_name                          AS step,
-      a.step_order,
-      COALESCE(sc.users_completed, 0)      AS users_completed,
-      ROUND(
-        100.0 * COALESCE(sc.users_completed, 0)
-        / NULLIF(
-            LAG(COALESCE(sc.users_completed, 0)) OVER (ORDER BY a.step_order),
-            0
-          ),
-        2
-      )                                    AS pct_from_prev
-  FROM all_steps a
-  LEFT JOIN step_counts sc ON sc.step = a.step_name
-)
-SELECT * FROM funnel ORDER BY step_order

autogpt_platform/analytics/queries/user_onboarding_integration.sql

@@ -1,41 +0,0 @@
--- =============================================================
--- View: analytics.user_onboarding_integration
--- Looker source alias: ds75  |  Charts: 1
--- =============================================================
--- DESCRIPTION
---   Pre-aggregated count of users who selected each integration
---   during onboarding.  One row per integration type, sorted
---   by popularity.
---
--- SOURCE TABLES
---   platform.UserOnboarding  — integrations array column
---
--- OUTPUT COLUMNS
---   integration            TEXT    Integration name (e.g. 'github', 'slack', 'notion')
---   users_with_integration BIGINT  Distinct users who selected this integration
---
--- WINDOW
---   Users who started onboarding in the last 90 days
---
--- EXAMPLE QUERIES
---   -- Full integration popularity ranking
---   SELECT * FROM analytics.user_onboarding_integration;
---
---   -- Top 5 integrations
---   SELECT * FROM analytics.user_onboarding_integration LIMIT 5;
--- =============================================================
-
-WITH exploded AS (
-  SELECT
-      u."userId" AS user_id,
-      UNNEST(u."integrations") AS integration
-  FROM platform."UserOnboarding" u
-  WHERE u."createdAt" >= CURRENT_DATE - INTERVAL '90 days'
-)
-SELECT
-    integration,
-    COUNT(DISTINCT user_id) AS users_with_integration
-FROM exploded
-WHERE integration IS NOT NULL AND integration <> ''
-GROUP BY integration
-ORDER BY users_with_integration DESC

autogpt_platform/analytics/queries/users_activities.sql

@@ -1,145 +0,0 @@
--- =============================================================
--- View: analytics.users_activities
--- Looker source alias: ds56  |  Charts: 5
--- =============================================================
--- DESCRIPTION
---   One row per user with lifetime activity summary.
---   Joins login sessions with agent graphs, executions and
---   node-level runs to give a full picture of how engaged
---   each user is.  Includes a convenience flag for 7-day
---   activation (did the user return at least 7 days after
---   their first login?).
---
--- SOURCE TABLES
---   auth.sessions                    — Login/session records
---   platform.AgentGraph              — Graphs (agents) built by the user
---   platform.AgentGraphExecution     — Agent run history
---   platform.AgentNodeExecution      — Individual block execution history
---
--- PERFORMANCE NOTE
---   Each CTE aggregates its own table independently by userId.
---   This avoids the fan-out that occurs when driving every join
---   from user_logins across the two largest tables
---   (AgentGraphExecution and AgentNodeExecution).
---
--- OUTPUT COLUMNS
---   user_id                   TEXT         Supabase user UUID
---   first_login_time          TIMESTAMPTZ  First ever session created_at
---   last_login_time           TIMESTAMPTZ  Most recent session created_at
---   last_visit_time           TIMESTAMPTZ  Max of last refresh or login
---   last_agent_save_time      TIMESTAMPTZ  Last time user saved an agent graph
---   agent_count               BIGINT       Number of distinct active graphs built (0 if none)
---   first_agent_run_time      TIMESTAMPTZ  First ever graph execution
---   last_agent_run_time       TIMESTAMPTZ  Most recent graph execution
---   unique_agent_runs         BIGINT       Distinct agent graphs ever run (0 if none)
---   agent_runs                BIGINT       Total graph execution count (0 if none)
---   node_execution_count      BIGINT       Total node executions across all runs
---   node_execution_failed     BIGINT       Node executions with FAILED status
---   node_execution_completed  BIGINT       Node executions with COMPLETED status
---   node_execution_terminated BIGINT       Node executions with TERMINATED status
---   node_execution_queued     BIGINT       Node executions with QUEUED status
---   node_execution_running    BIGINT       Node executions with RUNNING status
---   is_active_after_7d        INT          1=returned after day 7, 0=did not, NULL=too early to tell
---   node_execution_incomplete BIGINT       Node executions with INCOMPLETE status
---   node_execution_review     BIGINT       Node executions with REVIEW status
---
--- EXAMPLE QUERIES
---   -- Users who ran at least one agent and returned after 7 days
---   SELECT COUNT(*) FROM analytics.users_activities
---   WHERE agent_runs > 0 AND is_active_after_7d = 1;
---
---   -- Top 10 most active users by agent runs
---   SELECT user_id, agent_runs, node_execution_count
---   FROM analytics.users_activities
---   ORDER BY agent_runs DESC LIMIT 10;
---
---   -- 7-day activation rate
---   SELECT
---     SUM(CASE WHEN is_active_after_7d = 1 THEN 1 ELSE 0 END)::float
---     / NULLIF(COUNT(CASE WHEN is_active_after_7d IS NOT NULL THEN 1 END), 0)
---     AS activation_rate
---   FROM analytics.users_activities;
--- =============================================================
-
-WITH user_logins AS (
-  SELECT
-    user_id::text                                    AS user_id,
-    MIN(created_at)                                  AS first_login_time,
-    MAX(created_at)                                  AS last_login_time,
-    GREATEST(
-      MAX(refreshed_at)::timestamptz,
-      MAX(created_at)::timestamptz
-    )                                                AS last_visit_time
-  FROM auth.sessions
-  GROUP BY user_id
-),
-user_agents AS (
-  -- Aggregate AgentGraph directly by userId (no fan-out from user_logins)
-  SELECT
-    "userId"::text                AS user_id,
-    MAX("updatedAt")              AS last_agent_save_time,
-    COUNT(DISTINCT "id")          AS agent_count
-  FROM platform."AgentGraph"
-  WHERE "isActive"
-  GROUP BY "userId"
-),
-user_graph_runs AS (
-  -- Aggregate AgentGraphExecution directly by userId
-  SELECT
-    "userId"::text                        AS user_id,
-    MIN("createdAt")                      AS first_agent_run_time,
-    MAX("createdAt")                      AS last_agent_run_time,
-    COUNT(DISTINCT "agentGraphId")        AS unique_agent_runs,
-    COUNT("id")                           AS agent_runs
-  FROM platform."AgentGraphExecution"
-  GROUP BY "userId"
-),
-user_node_runs AS (
-  -- Aggregate AgentNodeExecution directly; resolve userId via a
-  -- single join to AgentGraphExecution instead of fanning out from
-  -- user_logins through both large tables.
-  SELECT
-    g."userId"::text                                                   AS user_id,
-    COUNT(*)                                                           AS node_execution_count,
-    COUNT(*) FILTER (WHERE n."executionStatus" = 'FAILED')             AS node_execution_failed,
-    COUNT(*) FILTER (WHERE n."executionStatus" = 'COMPLETED')          AS node_execution_completed,
-    COUNT(*) FILTER (WHERE n."executionStatus" = 'TERMINATED')         AS node_execution_terminated,
-    COUNT(*) FILTER (WHERE n."executionStatus" = 'QUEUED')             AS node_execution_queued,
-    COUNT(*) FILTER (WHERE n."executionStatus" = 'RUNNING')            AS node_execution_running,
-    COUNT(*) FILTER (WHERE n."executionStatus" = 'INCOMPLETE')         AS node_execution_incomplete,
-    COUNT(*) FILTER (WHERE n."executionStatus" = 'REVIEW')             AS node_execution_review
-  FROM platform."AgentNodeExecution" n
-  JOIN platform."AgentGraphExecution" g
-    ON g."id" = n."agentGraphExecutionId"
-  GROUP BY g."userId"
-)
-SELECT
-  ul.user_id,
-  ul.first_login_time,
-  ul.last_login_time,
-  ul.last_visit_time,
-  ua.last_agent_save_time,
-  COALESCE(ua.agent_count, 0)             AS agent_count,
-  gr.first_agent_run_time,
-  gr.last_agent_run_time,
-  COALESCE(gr.unique_agent_runs, 0)       AS unique_agent_runs,
-  COALESCE(gr.agent_runs, 0)              AS agent_runs,
-  COALESCE(nr.node_execution_count, 0)      AS node_execution_count,
-  COALESCE(nr.node_execution_failed, 0)     AS node_execution_failed,
-  COALESCE(nr.node_execution_completed, 0)  AS node_execution_completed,
-  COALESCE(nr.node_execution_terminated, 0) AS node_execution_terminated,
-  COALESCE(nr.node_execution_queued, 0)     AS node_execution_queued,
-  COALESCE(nr.node_execution_running, 0)    AS node_execution_running,
-  CASE
-    WHEN ul.first_login_time < NOW() - INTERVAL '7 days'
-     AND ul.last_visit_time  >= ul.first_login_time + INTERVAL '7 days' THEN 1
-    WHEN ul.first_login_time < NOW() - INTERVAL '7 days'
-     AND ul.last_visit_time  <  ul.first_login_time + INTERVAL '7 days' THEN 0
-    ELSE NULL
-  END AS is_active_after_7d,
-  COALESCE(nr.node_execution_incomplete, 0) AS node_execution_incomplete,
-  COALESCE(nr.node_execution_review, 0)     AS node_execution_review
-FROM user_logins ul
-LEFT JOIN user_agents     ua ON ul.user_id = ua.user_id
-LEFT JOIN user_graph_runs gr ON ul.user_id = gr.user_id
-LEFT JOIN user_node_runs  nr ON ul.user_id = nr.user_id

autogpt_platform/autogpt_libs/poetry.lock

@@ -1,4 +1,4 @@
-# This file is automatically @generated by Poetry 2.2.1 and should not be changed by hand.
+# This file is automatically @generated by Poetry 2.1.1 and should not be changed by hand.
 
 [[package]]
 name = "annotated-doc"
@@ -67,7 +67,7 @@ description = "Backport of asyncio.Runner, a context manager that controls event
 optional = false
 python-versions = "<3.11,>=3.8"
 groups = ["dev"]
-markers = "python_version == \"3.10\""
+markers = "python_version < \"3.11\""
 files = [
     {file = "backports_asyncio_runner-1.2.0-py3-none-any.whl", hash = "sha256:0da0a936a8aeb554eccb426dc55af3ba63bcdc69fa1a600b5bb305413a4477b5"},
     {file = "backports_asyncio_runner-1.2.0.tar.gz", hash = "sha256:a5aa7b2b7d8f8bfcaa2b57313f70792df84e32a2a746f585213373f900b42162"},
@@ -541,7 +541,7 @@ description = "Backport of PEP 654 (exception groups)"
 optional = false
 python-versions = ">=3.7"
 groups = ["main", "dev"]
-markers = "python_version == \"3.10\""
+markers = "python_version < \"3.11\""
 files = [
     {file = "exceptiongroup-1.3.0-py3-none-any.whl", hash = "sha256:4d111e6e0c13d0644cad6ddaa7ed0261a0b36971f6d23e7ec9b4b9097da78a10"},
     {file = "exceptiongroup-1.3.0.tar.gz", hash = "sha256:b241f5885f560bc56a59ee63ca4c6a8bfa46ae4ad651af316d4e81817bb9fd88"},
@@ -2181,14 +2181,14 @@ testing = ["coverage (>=6.2)", "hypothesis (>=5.7.1)"]
 
 [[package]]
 name = "pytest-cov"
-version = "7.1.0"
+version = "7.0.0"
 description = "Pytest plugin for measuring coverage."
 optional = false
 python-versions = ">=3.9"
 groups = ["dev"]
 files = [
-    {file = "pytest_cov-7.1.0-py3-none-any.whl", hash = "sha256:a0461110b7865f9a271aa1b51e516c9a95de9d696734a2f71e3e78f46e1d4678"},
-    {file = "pytest_cov-7.1.0.tar.gz", hash = "sha256:30674f2b5f6351aa09702a9c8c364f6a01c27aae0c1366ae8016160d1efc56b2"},
+    {file = "pytest_cov-7.0.0-py3-none-any.whl", hash = "sha256:3b8e9558b16cc1479da72058bdecf8073661c7f57f7d3c5f22a1c23507f2d861"},
+    {file = "pytest_cov-7.0.0.tar.gz", hash = "sha256:33c97eda2e049a0c5298e91f519302a1334c26ac65c1a483d6206fd458361af1"},
 ]
 
 [package.dependencies]
@@ -2342,30 +2342,30 @@ pyasn1 = ">=0.1.3"
 
 [[package]]
 name = "ruff"
-version = "0.15.7"
+version = "0.15.0"
 description = "An extremely fast Python linter and code formatter, written in Rust."
 optional = false
 python-versions = ">=3.7"
 groups = ["dev"]
 files = [
-    {file = "ruff-0.15.7-py3-none-linux_armv6l.whl", hash = "sha256:a81cc5b6910fb7dfc7c32d20652e50fa05963f6e13ead3c5915c41ac5d16668e"},
-    {file = "ruff-0.15.7-py3-none-macosx_10_12_x86_64.whl", hash = "sha256:722d165bd52403f3bdabc0ce9e41fc47070ac56d7a91b4e0d097b516a53a3477"},
-    {file = "ruff-0.15.7-py3-none-macosx_11_0_arm64.whl", hash = "sha256:7fbc2448094262552146cbe1b9643a92f66559d3761f1ad0656d4991491af49e"},
-    {file = "ruff-0.15.7-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:6b39329b60eba44156d138275323cc726bbfbddcec3063da57caa8a8b1d50adf"},
-    {file = "ruff-0.15.7-py3-none-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:87768c151808505f2bfc93ae44e5f9e7c8518943e5074f76ac21558ef5627c85"},
-    {file = "ruff-0.15.7-py3-none-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:fb0511670002c6c529ec66c0e30641c976c8963de26a113f3a30456b702468b0"},
-    {file = "ruff-0.15.7-py3-none-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:e0d19644f801849229db8345180a71bee5407b429dd217f853ec515e968a6912"},
-    {file = "ruff-0.15.7-py3-none-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:4806d8e09ef5e84eb19ba833d0442f7e300b23fe3f0981cae159a248a10f0036"},
-    {file = "ruff-0.15.7-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:dce0896488562f09a27b9c91b1f58a097457143931f3c4d519690dea54e624c5"},
-    {file = "ruff-0.15.7-py3-none-manylinux_2_31_riscv64.whl", hash = "sha256:1852ce241d2bc89e5dc823e03cff4ce73d816b5c6cdadd27dbfe7b03217d2a12"},
-    {file = "ruff-0.15.7-py3-none-musllinux_1_2_aarch64.whl", hash = "sha256:5f3e4b221fb4bd293f79912fc5e93a9063ebd6d0dcbd528f91b89172a9b8436c"},
-    {file = "ruff-0.15.7-py3-none-musllinux_1_2_armv7l.whl", hash = "sha256:b15e48602c9c1d9bdc504b472e90b90c97dc7d46c7028011ae67f3861ceba7b4"},
-    {file = "ruff-0.15.7-py3-none-musllinux_1_2_i686.whl", hash = "sha256:1b4705e0e85cedc74b0a23cf6a179dbb3df184cb227761979cc76c0440b5ab0d"},
-    {file = "ruff-0.15.7-py3-none-musllinux_1_2_x86_64.whl", hash = "sha256:112c1fa316a558bb34319282c1200a8bf0495f1b735aeb78bfcb2991e6087580"},
-    {file = "ruff-0.15.7-py3-none-win32.whl", hash = "sha256:6d39e2d3505b082323352f733599f28169d12e891f7dd407f2d4f54b4c2886de"},
-    {file = "ruff-0.15.7-py3-none-win_amd64.whl", hash = "sha256:4d53d712ddebcd7dace1bc395367aec12c057aacfe9adbb6d832302575f4d3a1"},
-    {file = "ruff-0.15.7-py3-none-win_arm64.whl", hash = "sha256:18e8d73f1c3fdf27931497972250340f92e8c861722161a9caeb89a58ead6ed2"},
-    {file = "ruff-0.15.7.tar.gz", hash = "sha256:04f1ae61fc20fe0b148617c324d9d009b5f63412c0b16474f3d5f1a1a665f7ac"},
+    {file = "ruff-0.15.0-py3-none-linux_armv6l.whl", hash = "sha256:aac4ebaa612a82b23d45964586f24ae9bc23ca101919f5590bdb368d74ad5455"},
+    {file = "ruff-0.15.0-py3-none-macosx_10_12_x86_64.whl", hash = "sha256:dcd4be7cc75cfbbca24a98d04d0b9b36a270d0833241f776b788d59f4142b14d"},
+    {file = "ruff-0.15.0-py3-none-macosx_11_0_arm64.whl", hash = "sha256:d747e3319b2bce179c7c1eaad3d884dc0a199b5f4d5187620530adf9105268ce"},
+    {file = "ruff-0.15.0-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:650bd9c56ae03102c51a5e4b554d74d825ff3abe4db22b90fd32d816c2e90621"},
+    {file = "ruff-0.15.0-py3-none-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:a6664b7eac559e3048223a2da77769c2f92b43a6dfd4720cef42654299a599c9"},
+    {file = "ruff-0.15.0-py3-none-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:6f811f97b0f092b35320d1556f3353bf238763420ade5d9e62ebd2b73f2ff179"},
+    {file = "ruff-0.15.0-py3-none-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:761ec0a66680fab6454236635a39abaf14198818c8cdf691e036f4bc0f406b2d"},
+    {file = "ruff-0.15.0-py3-none-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:940f11c2604d317e797b289f4f9f3fa5555ffe4fb574b55ed006c3d9b6f0eb78"},
+    {file = "ruff-0.15.0-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:bcbca3d40558789126da91d7ef9a7c87772ee107033db7191edefa34e2c7f1b4"},
+    {file = "ruff-0.15.0-py3-none-manylinux_2_31_riscv64.whl", hash = "sha256:9a121a96db1d75fa3eb39c4539e607f628920dd72ff1f7c5ee4f1b768ac62d6e"},
+    {file = "ruff-0.15.0-py3-none-musllinux_1_2_aarch64.whl", hash = "sha256:5298d518e493061f2eabd4abd067c7e4fb89e2f63291c94332e35631c07c3662"},
+    {file = "ruff-0.15.0-py3-none-musllinux_1_2_armv7l.whl", hash = "sha256:afb6e603d6375ff0d6b0cee563fa21ab570fd15e65c852cb24922cef25050cf1"},
+    {file = "ruff-0.15.0-py3-none-musllinux_1_2_i686.whl", hash = "sha256:77e515f6b15f828b94dc17d2b4ace334c9ddb7d9468c54b2f9ed2b9c1593ef16"},
+    {file = "ruff-0.15.0-py3-none-musllinux_1_2_x86_64.whl", hash = "sha256:6f6e80850a01eb13b3e42ee0ebdf6e4497151b48c35051aab51c101266d187a3"},
+    {file = "ruff-0.15.0-py3-none-win32.whl", hash = "sha256:238a717ef803e501b6d51e0bdd0d2c6e8513fe9eec14002445134d3907cd46c3"},
+    {file = "ruff-0.15.0-py3-none-win_amd64.whl", hash = "sha256:dd5e4d3301dc01de614da3cdffc33d4b1b96fb89e45721f1598e5532ccf78b18"},
+    {file = "ruff-0.15.0-py3-none-win_arm64.whl", hash = "sha256:c480d632cc0ca3f0727acac8b7d053542d9e114a462a145d0b00e7cd658c515a"},
+    {file = "ruff-0.15.0.tar.gz", hash = "sha256:6bdea47cdbea30d40f8f8d7d69c0854ba7c15420ec75a26f463290949d7f7e9a"},
 ]
 
 [[package]]
@@ -2564,7 +2564,7 @@ description = "A lil' TOML parser"
 optional = false
 python-versions = ">=3.8"
 groups = ["dev"]
-markers = "python_version == \"3.10\""
+markers = "python_version < \"3.11\""
 files = [
     {file = "tomli-2.2.1-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:678e4fa69e4575eb77d103de3df8a895e1591b48e740211bd1067378c69e8249"},
     {file = "tomli-2.2.1-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:023aa114dd824ade0100497eb2318602af309e5a55595f76b626d6d9f3b7b0a6"},
@@ -2912,4 +2912,4 @@ type = ["pytest-mypy"]
 [metadata]
 lock-version = "2.1"
 python-versions = ">=3.10,<4.0"
-content-hash = "e0936a065565550afed18f6298b7e04e814b44100def7049f1a0d68662624a39"
+content-hash = "9619cae908ad38fa2c48016a58bcf4241f6f5793aa0e6cc140276e91c433cbbb"

autogpt_platform/autogpt_libs/pyproject.toml

@@ -26,8 +26,8 @@ pyright = "^1.1.408"
 pytest = "^8.4.1"
 pytest-asyncio = "^1.3.0"
 pytest-mock = "^3.15.1"
-pytest-cov = "^7.1.0"
-ruff = "^0.15.7"
+pytest-cov = "^7.0.0"
+ruff = "^0.15.0"
 
 [build-system]
 requires = ["poetry-core"]

autogpt_platform/backend/.env.default

@@ -58,17 +58,6 @@ V0_API_KEY=
 OPEN_ROUTER_API_KEY=
 NVIDIA_API_KEY=
 
-# Graphiti Temporal Knowledge Graph Memory
-# Rollout controlled by LaunchDarkly flag "graphiti-memory"
-# LLM key falls back to CHAT_API_KEY (AutoPilot), then OPEN_ROUTER_API_KEY.
-# Embedder key falls back to CHAT_OPENAI_API_KEY (AutoPilot), then OPENAI_API_KEY.
-GRAPHITI_FALKORDB_HOST=localhost
-GRAPHITI_FALKORDB_PORT=6380
-GRAPHITI_FALKORDB_PASSWORD=
-GRAPHITI_LLM_MODEL=gpt-4.1-mini
-GRAPHITI_EMBEDDER_MODEL=text-embedding-3-small
-GRAPHITI_SEMAPHORE_LIMIT=5
-
 # Langfuse Prompt Management
 # Used for managing the CoPilot system prompt externally
 # Get credentials from https://cloud.langfuse.com or your self-hosted instance
@@ -179,9 +168,6 @@ MEM0_API_KEY=
 OPENWEATHERMAP_API_KEY=
 GOOGLE_MAPS_API_KEY=
 
-# Platform Bot Linking
-PLATFORM_LINK_BASE_URL=http://localhost:3000/link
-
 # Communication Services
 DISCORD_BOT_TOKEN=
 MEDIUM_API_KEY=
@@ -192,7 +178,6 @@ SMTP_USERNAME=
 SMTP_PASSWORD=
 
 # Business & Marketing Tools
-AGENTMAIL_API_KEY=
 APOLLO_API_KEY=
 ENRICHLAYER_API_KEY=
 AYRSHARE_API_KEY=
@@ -205,8 +190,5 @@ ZEROBOUNCE_API_KEY=
 POSTHOG_API_KEY=
 POSTHOG_HOST=https://eu.i.posthog.com
 
-# Tally Form Integration (pre-populate business understanding on signup)
-TALLY_API_KEY=
-
 # Other Services
 AUTOMOD_API_KEY=

autogpt_platform/backend/AGENTS.md

@@ -1,227 +0,0 @@
-# Backend
-
-This file provides guidance to coding agents when working with the backend.
-
-## Essential Commands
-
-To run something with Python package dependencies you MUST use `poetry run ...`.
-
-```bash
-# Install dependencies
-poetry install
-
-# Run database migrations
-poetry run prisma migrate dev
-
-# Start all services (database, redis, rabbitmq, clamav)
-docker compose up -d
-
-# Run the backend as a whole
-poetry run app
-
-# Run tests
-poetry run test
-
-# Run specific test
-poetry run pytest path/to/test_file.py::test_function_name
-
-# Run block tests (tests that validate all blocks work correctly)
-poetry run pytest backend/blocks/test/test_block.py -xvs
-
-# Run tests for a specific block (e.g., GetCurrentTimeBlock)
-poetry run pytest 'backend/blocks/test/test_block.py::test_available_blocks[GetCurrentTimeBlock]' -xvs
-
-# Lint and format
-# prefer format if you want to just "fix" it and only get the errors that can't be autofixed
-poetry run format  # Black + isort
-poetry run lint    # ruff
-```
-
-More details can be found in @TESTING.md
-
-### Creating/Updating Snapshots
-
-When you first write a test or when the expected output changes:
-
-```bash
-poetry run pytest path/to/test.py --snapshot-update
-```
-
-⚠️ **Important**: Always review snapshot changes before committing! Use `git diff` to verify the changes are expected.
-
-## Architecture
-
-- **API Layer**: FastAPI with REST and WebSocket endpoints
-- **Database**: PostgreSQL with Prisma ORM, includes pgvector for embeddings
-- **Queue System**: RabbitMQ for async task processing
-- **Execution Engine**: Separate executor service processes agent workflows
-- **Authentication**: JWT-based with Supabase integration
-- **Security**: Cache protection middleware prevents sensitive data caching in browsers/proxies
-
-## Code Style
-
-- **Top-level imports only** — no local/inner imports (lazy imports only for heavy optional deps like `openpyxl`)
-- **Absolute imports** — use `from backend.module import ...` for cross-package imports. Single-dot relative (`from .sibling import ...`) is acceptable for sibling modules within the same package (e.g., blocks). Avoid double-dot relative imports (`from ..parent import ...`) — use the absolute path instead
-- **No duck typing** — no `hasattr`/`getattr`/`isinstance` for type dispatch; use typed interfaces/unions/protocols
-- **Pydantic models** over dataclass/namedtuple/dict for structured data
-- **No linter suppressors** — no `# type: ignore`, `# noqa`, `# pyright: ignore`; fix the type/code
-- **List comprehensions** over manual loop-and-append
-- **Early return** — guard clauses first, avoid deep nesting
-- **f-strings vs printf syntax in log statements** — Use `%s` for deferred interpolation in `debug` statements, f-strings elsewhere for readability: `logger.debug("Processing %s items", count)`, `logger.info(f"Processing {count} items")`
-- **Sanitize error paths** — `os.path.basename()` in error messages to avoid leaking directory structure
-- **TOCTOU awareness** — avoid check-then-act patterns for file access and credit charging
-- **`Security()` vs `Depends()`** — use `Security()` for auth deps to get proper OpenAPI security spec
-- **Redis pipelines** — `transaction=True` for atomicity on multi-step operations
-- **`max(0, value)` guards** — for computed values that should never be negative
-- **SSE protocol** — `data:` lines for frontend-parsed events (must match Zod schema), `: comment` lines for heartbeats/status
-- **File length** — keep files under ~300 lines; if a file grows beyond this, split by responsibility (e.g. extract helpers, models, or a sub-module into a new file). Never keep appending to a long file.
-- **Function length** — keep functions under ~40 lines; extract named helpers when a function grows longer. Long functions are a sign of mixed concerns, not complexity.
-- **Top-down ordering** — define the main/public function or class first, then the helpers it uses below. A reader should encounter high-level logic before implementation details.
-
-## Testing Approach
-
-- Uses pytest with snapshot testing for API responses
-- Test files are colocated with source files (`*_test.py`)
-- Mock at boundaries — mock where the symbol is **used**, not where it's **defined**
-- After refactoring, update mock targets to match new module paths
-- Use `AsyncMock` for async functions (`from unittest.mock import AsyncMock`)
-
-### Test-Driven Development (TDD)
-
-When fixing a bug or adding a feature, write the test **before** the implementation:
-
-```python
-# 1. Write a failing test marked xfail
-@pytest.mark.xfail(reason="Bug #1234: widget crashes on empty input")
-def test_widget_handles_empty_input():
-    result = widget.process("")
-    assert result == Widget.EMPTY_RESULT
-
-# 2. Run it — confirm it fails (XFAIL)
-# poetry run pytest path/to/test.py::test_widget_handles_empty_input -xvs
-
-# 3. Implement the fix
-
-# 4. Remove xfail, run again — confirm it passes
-def test_widget_handles_empty_input():
-    result = widget.process("")
-    assert result == Widget.EMPTY_RESULT
-```
-
-This catches regressions and proves the fix actually works. **Every bug fix should include a test that would have caught it.**
-
-## Database Schema
-
-Key models (defined in `schema.prisma`):
-
-- `User`: Authentication and profile data
-- `AgentGraph`: Workflow definitions with version control
-- `AgentGraphExecution`: Execution history and results
-- `AgentNode`: Individual nodes in a workflow
-- `StoreListing`: Marketplace listings for sharing agents
-
-## Environment Configuration
-
-- **Backend**: `.env.default` (defaults) → `.env` (user overrides)
-
-## Common Development Tasks
-
-### Adding a new block
-
-Follow the comprehensive [Block SDK Guide](@../../docs/platform/block-sdk-guide.md) which covers:
-
-- Provider configuration with `ProviderBuilder`
-- Block schema definition
-- Authentication (API keys, OAuth, webhooks)
-- Testing and validation
-- File organization
-
-Quick steps:
-
-1. Create new file in `backend/blocks/`
-2. Configure provider using `ProviderBuilder` in `_config.py`
-3. Inherit from `Block` base class
-4. Define input/output schemas using `BlockSchema`
-5. Implement async `run` method
-6. Generate unique block ID using `uuid.uuid4()`
-7. Test with `poetry run pytest backend/blocks/test/test_block.py`
-
-Note: when making many new blocks analyze the interfaces for each of these blocks and picture if they would go well together in a graph-based editor or would they struggle to connect productively?
-ex: do the inputs and outputs tie well together?
-
-If you get any pushback or hit complex block conditions check the new_blocks guide in the docs.
-
-#### Handling files in blocks with `store_media_file()`
-
-When blocks need to work with files (images, videos, documents), use `store_media_file()` from `backend.util.file`. The `return_format` parameter determines what you get back:
-
-| Format | Use When | Returns |
-|--------|----------|---------|
-| `"for_local_processing"` | Processing with local tools (ffmpeg, MoviePy, PIL) | Local file path (e.g., `"image.png"`) |
-| `"for_external_api"` | Sending content to external APIs (Replicate, OpenAI) | Data URI (e.g., `"data:image/png;base64,..."`) |
-| `"for_block_output"` | Returning output from your block | Smart: `workspace://` in CoPilot, data URI in graphs |
-
-**Examples:**
-
-```python
-# INPUT: Need to process file locally with ffmpeg
-local_path = await store_media_file(
-    file=input_data.video,
-    execution_context=execution_context,
-    return_format="for_local_processing",
-)
-# local_path = "video.mp4" - use with Path/ffmpeg/etc
-
-# INPUT: Need to send to external API like Replicate
-image_b64 = await store_media_file(
-    file=input_data.image,
-    execution_context=execution_context,
-    return_format="for_external_api",
-)
-# image_b64 = "data:image/png;base64,iVBORw0..." - send to API
-
-# OUTPUT: Returning result from block
-result_url = await store_media_file(
-    file=generated_image_url,
-    execution_context=execution_context,
-    return_format="for_block_output",
-)
-yield "image_url", result_url
-# In CoPilot: result_url = "workspace://abc123"
-# In graphs:  result_url = "data:image/png;base64,..."
-```
-
-**Key points:**
-
-- `for_block_output` is the ONLY format that auto-adapts to execution context
-- Always use `for_block_output` for block outputs unless you have a specific reason not to
-- Never hardcode workspace checks - let `for_block_output` handle it
-
-### Modifying the API
-
-1. Update route in `backend/api/features/`
-2. Add/update Pydantic models in same directory
-3. Write tests alongside the route file
-4. Run `poetry run test` to verify
-
-## Workspace & Media Files
-
-**Read [Workspace & Media Architecture](../../docs/platform/workspace-media-architecture.md) when:**
-- Working on CoPilot file upload/download features
-- Building blocks that handle `MediaFileType` inputs/outputs
-- Modifying `WorkspaceManager` or `store_media_file()`
-- Debugging file persistence or virus scanning issues
-
-Covers: `WorkspaceManager` (persistent storage with session scoping), `store_media_file()` (media normalization pipeline), and responsibility boundaries for virus scanning and persistence.
-
-## Security Implementation
-
-### Cache Protection Middleware
-
-- Located in `backend/api/middleware/security.py`
-- Default behavior: Disables caching for ALL endpoints with `Cache-Control: no-store, no-cache, must-revalidate, private`
-- Uses an allow list approach - only explicitly permitted paths can be cached
-- Cacheable paths include: static assets (`static/*`, `_next/static/*`), health checks, public store pages, documentation
-- Prevents sensitive data (auth tokens, API keys, user data) from being cached by browsers/proxies
-- To allow caching for a new endpoint, add it to `CACHEABLE_PATHS` in the middleware
-- Applied to both main API server and external API applications

autogpt_platform/backend/CLAUDE.md

@@ -1 +1,170 @@
-@AGENTS.md
+# CLAUDE.md - Backend
+
+This file provides guidance to Claude Code when working with the backend.
+
+## Essential Commands
+
+To run something with Python package dependencies you MUST use `poetry run ...`.
+
+```bash
+# Install dependencies
+poetry install
+
+# Run database migrations
+poetry run prisma migrate dev
+
+# Start all services (database, redis, rabbitmq, clamav)
+docker compose up -d
+
+# Run the backend as a whole
+poetry run app
+
+# Run tests
+poetry run test
+
+# Run specific test
+poetry run pytest path/to/test_file.py::test_function_name
+
+# Run block tests (tests that validate all blocks work correctly)
+poetry run pytest backend/blocks/test/test_block.py -xvs
+
+# Run tests for a specific block (e.g., GetCurrentTimeBlock)
+poetry run pytest 'backend/blocks/test/test_block.py::test_available_blocks[GetCurrentTimeBlock]' -xvs
+
+# Lint and format
+# prefer format if you want to just "fix" it and only get the errors that can't be autofixed
+poetry run format  # Black + isort
+poetry run lint    # ruff
+```
+
+More details can be found in @TESTING.md
+
+### Creating/Updating Snapshots
+
+When you first write a test or when the expected output changes:
+
+```bash
+poetry run pytest path/to/test.py --snapshot-update
+```
+
+⚠️ **Important**: Always review snapshot changes before committing! Use `git diff` to verify the changes are expected.
+
+## Architecture
+
+- **API Layer**: FastAPI with REST and WebSocket endpoints
+- **Database**: PostgreSQL with Prisma ORM, includes pgvector for embeddings
+- **Queue System**: RabbitMQ for async task processing
+- **Execution Engine**: Separate executor service processes agent workflows
+- **Authentication**: JWT-based with Supabase integration
+- **Security**: Cache protection middleware prevents sensitive data caching in browsers/proxies
+
+## Testing Approach
+
+- Uses pytest with snapshot testing for API responses
+- Test files are colocated with source files (`*_test.py`)
+
+## Database Schema
+
+Key models (defined in `schema.prisma`):
+
+- `User`: Authentication and profile data
+- `AgentGraph`: Workflow definitions with version control
+- `AgentGraphExecution`: Execution history and results
+- `AgentNode`: Individual nodes in a workflow
+- `StoreListing`: Marketplace listings for sharing agents
+
+## Environment Configuration
+
+- **Backend**: `.env.default` (defaults) → `.env` (user overrides)
+
+## Common Development Tasks
+
+### Adding a new block
+
+Follow the comprehensive [Block SDK Guide](@../../docs/content/platform/block-sdk-guide.md) which covers:
+
+- Provider configuration with `ProviderBuilder`
+- Block schema definition
+- Authentication (API keys, OAuth, webhooks)
+- Testing and validation
+- File organization
+
+Quick steps:
+
+1. Create new file in `backend/blocks/`
+2. Configure provider using `ProviderBuilder` in `_config.py`
+3. Inherit from `Block` base class
+4. Define input/output schemas using `BlockSchema`
+5. Implement async `run` method
+6. Generate unique block ID using `uuid.uuid4()`
+7. Test with `poetry run pytest backend/blocks/test/test_block.py`
+
+Note: when making many new blocks analyze the interfaces for each of these blocks and picture if they would go well together in a graph-based editor or would they struggle to connect productively?
+ex: do the inputs and outputs tie well together?
+
+If you get any pushback or hit complex block conditions check the new_blocks guide in the docs.
+
+#### Handling files in blocks with `store_media_file()`
+
+When blocks need to work with files (images, videos, documents), use `store_media_file()` from `backend.util.file`. The `return_format` parameter determines what you get back:
+
+| Format | Use When | Returns |
+|--------|----------|---------|
+| `"for_local_processing"` | Processing with local tools (ffmpeg, MoviePy, PIL) | Local file path (e.g., `"image.png"`) |
+| `"for_external_api"` | Sending content to external APIs (Replicate, OpenAI) | Data URI (e.g., `"data:image/png;base64,..."`) |
+| `"for_block_output"` | Returning output from your block | Smart: `workspace://` in CoPilot, data URI in graphs |
+
+**Examples:**
+
+```python
+# INPUT: Need to process file locally with ffmpeg
+local_path = await store_media_file(
+    file=input_data.video,
+    execution_context=execution_context,
+    return_format="for_local_processing",
+)
+# local_path = "video.mp4" - use with Path/ffmpeg/etc
+
+# INPUT: Need to send to external API like Replicate
+image_b64 = await store_media_file(
+    file=input_data.image,
+    execution_context=execution_context,
+    return_format="for_external_api",
+)
+# image_b64 = "data:image/png;base64,iVBORw0..." - send to API
+
+# OUTPUT: Returning result from block
+result_url = await store_media_file(
+    file=generated_image_url,
+    execution_context=execution_context,
+    return_format="for_block_output",
+)
+yield "image_url", result_url
+# In CoPilot: result_url = "workspace://abc123"
+# In graphs:  result_url = "data:image/png;base64,..."
+```
+
+**Key points:**
+
+- `for_block_output` is the ONLY format that auto-adapts to execution context
+- Always use `for_block_output` for block outputs unless you have a specific reason not to
+- Never hardcode workspace checks - let `for_block_output` handle it
+
+### Modifying the API
+
+1. Update route in `backend/api/features/`
+2. Add/update Pydantic models in same directory
+3. Write tests alongside the route file
+4. Run `poetry run test` to verify
+
+## Security Implementation
+
+### Cache Protection Middleware
+
+- Located in `backend/api/middleware/security.py`
+- Default behavior: Disables caching for ALL endpoints with `Cache-Control: no-store, no-cache, must-revalidate, private`
+- Uses an allow list approach - only explicitly permitted paths can be cached
+- Cacheable paths include: static assets (`static/*`, `_next/static/*`), health checks, public store pages, documentation
+- Prevents sensitive data (auth tokens, API keys, user data) from being cached by browsers/proxies
+- To allow caching for a new endpoint, add it to `CACHEABLE_PATHS` in the middleware
+- Applied to both main API server and external API applications

autogpt_platform/backend/Dockerfile

@@ -50,7 +50,7 @@ RUN poetry install --no-ansi --no-root
 # Generate Prisma client
 COPY autogpt_platform/backend/schema.prisma ./
 COPY autogpt_platform/backend/backend/data/partial_types.py ./backend/data/partial_types.py
-COPY autogpt_platform/backend/scripts/gen_prisma_types_stub.py ./scripts/
+COPY autogpt_platform/backend/gen_prisma_types_stub.py ./
 RUN poetry run prisma generate && poetry run gen-prisma-stub
 
 # =============================== DB MIGRATOR =============================== #
@@ -82,7 +82,7 @@ RUN pip3 install prisma>=0.15.0 --break-system-packages
 
 COPY autogpt_platform/backend/schema.prisma ./
 COPY autogpt_platform/backend/backend/data/partial_types.py ./backend/data/partial_types.py
-COPY autogpt_platform/backend/scripts/gen_prisma_types_stub.py ./scripts/
+COPY autogpt_platform/backend/gen_prisma_types_stub.py ./
 COPY autogpt_platform/backend/migrations ./migrations
 
 # ============================== BACKEND SERVER ============================== #
@@ -95,7 +95,7 @@ ENV DEBIAN_FRONTEND=noninteractive
 
 # Install Python, FFmpeg, ImageMagick, and CLI tools for agent use.
 # bubblewrap provides OS-level sandbox (whitelist-only FS + no network)
-# for the bash_exec MCP tool (fallback when E2B is not configured).
+# for the bash_exec MCP tool.
 # Using --no-install-recommends saves ~650MB by skipping unnecessary deps like llvm, mesa, etc.
 RUN apt-get update && apt-get install -y --no-install-recommends \
     python3.13 \
@@ -111,31 +111,13 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
 # Copy poetry (build-time only, for `poetry install --only-root` to create entry points)
 COPY --from=builder /usr/local/lib/python3* /usr/local/lib/python3*
 COPY --from=builder /usr/local/bin/poetry /usr/local/bin/poetry
-# Copy Node.js installation for Prisma and agent-browser.
-# npm/npx are symlinks in the builder (-> ../lib/node_modules/npm/bin/*-cli.js);
-# COPY resolves them to regular files, breaking require() paths.  Recreate as
-# proper symlinks so npm/npx can find their modules.
+# Copy Node.js installation for Prisma
 COPY --from=builder /usr/bin/node /usr/bin/node
 COPY --from=builder /usr/lib/node_modules /usr/lib/node_modules
-RUN ln -s ../lib/node_modules/npm/bin/npm-cli.js /usr/bin/npm \
-    && ln -s ../lib/node_modules/npm/bin/npx-cli.js /usr/bin/npx
+COPY --from=builder /usr/bin/npm /usr/bin/npm
+COPY --from=builder /usr/bin/npx /usr/bin/npx
 COPY --from=builder /root/.cache/prisma-python/binaries /root/.cache/prisma-python/binaries
 
-# Install agent-browser (Copilot browser tool) using the system chromium package.
-# Chrome for Testing (the binary agent-browser downloads via `agent-browser install`)
-# has no ARM64 builds, so we use the distro-packaged chromium instead — verified to
-# work with agent-browser via Docker tests on arm64; amd64 is validated in CI.
-# Note: system chromium tracks the Debian package schedule rather than a pinned
-# Chrome for Testing release. If agent-browser requires a specific Chrome version,
-# verify compatibility against the chromium package version in the base image.
-RUN apt-get update \
-    && apt-get install -y --no-install-recommends chromium fonts-liberation \
-    && rm -rf /var/lib/apt/lists/* \
-    && npm install -g agent-browser \
-    && rm -rf /tmp/* /root/.npm
-
-ENV AGENT_BROWSER_EXECUTABLE_PATH=/usr/bin/chromium
-
 WORKDIR /app/autogpt_platform/backend
 
 # Copy only the .venv from builder (not the entire /app directory)

autogpt_platform/backend/agents/calculator-agent.json

@@ -1,166 +0,0 @@
-{
-  "id": "858e2226-e047-4d19-a832-3be4a134d155",
-  "version": 2,
-  "is_active": true,
-  "name": "Calculator agent",
-  "description": "",
-  "instructions": null,
-  "recommended_schedule_cron": null,
-  "forked_from_id": null,
-  "forked_from_version": null,
-  "user_id": "",
-  "created_at": "2026-04-13T03:45:11.241Z",
-  "nodes": [
-    {
-      "id": "6762da5d-6915-4836-a431-6dcd7d36a54a",
-      "block_id": "c0a8e994-ebf1-4a9c-a4d8-89d09c86741b",
-      "input_default": {
-        "name": "Input",
-        "secret": false,
-        "advanced": false
-      },
-      "metadata": {
-        "position": {
-          "x": -188.2244873046875,
-          "y": 95
-        }
-      },
-      "input_links": [],
-      "output_links": [
-        {
-          "id": "432c7caa-49b9-4b70-bd21-2fa33a569601",
-          "source_id": "6762da5d-6915-4836-a431-6dcd7d36a54a",
-          "sink_id": "bf4a15ff-b0c4-4032-a21b-5880224af690",
-          "source_name": "result",
-          "sink_name": "a",
-          "is_static": true
-        }
-      ],
-      "graph_id": "858e2226-e047-4d19-a832-3be4a134d155",
-      "graph_version": 2,
-      "webhook_id": null
-    },
-    {
-      "id": "65429c9e-a0c6-4032-a421-6899c394fa74",
-      "block_id": "363ae599-353e-4804-937e-b2ee3cef3da4",
-      "input_default": {
-        "name": "Output",
-        "secret": false,
-        "advanced": false,
-        "escape_html": false
-      },
-      "metadata": {
-        "position": {
-          "x": 825.198974609375,
-          "y": 123.75
-        }
-      },
-      "input_links": [
-        {
-          "id": "8cdb2f33-5b10-4cc2-8839-f8ccb70083a3",
-          "source_id": "bf4a15ff-b0c4-4032-a21b-5880224af690",
-          "sink_id": "65429c9e-a0c6-4032-a421-6899c394fa74",
-          "source_name": "result",
-          "sink_name": "value",
-          "is_static": false
-        }
-      ],
-      "output_links": [],
-      "graph_id": "858e2226-e047-4d19-a832-3be4a134d155",
-      "graph_version": 2,
-      "webhook_id": null
-    },
-    {
-      "id": "bf4a15ff-b0c4-4032-a21b-5880224af690",
-      "block_id": "b1ab9b19-67a6-406d-abf5-2dba76d00c79",
-      "input_default": {
-        "b": 34,
-        "operation": "Add",
-        "round_result": false
-      },
-      "metadata": {
-        "position": {
-          "x": 323.0255126953125,
-          "y": 121.25
-        }
-      },
-      "input_links": [
-        {
-          "id": "432c7caa-49b9-4b70-bd21-2fa33a569601",
-          "source_id": "6762da5d-6915-4836-a431-6dcd7d36a54a",
-          "sink_id": "bf4a15ff-b0c4-4032-a21b-5880224af690",
-          "source_name": "result",
-          "sink_name": "a",
-          "is_static": true
-        }
-      ],
-      "output_links": [
-        {
-          "id": "8cdb2f33-5b10-4cc2-8839-f8ccb70083a3",
-          "source_id": "bf4a15ff-b0c4-4032-a21b-5880224af690",
-          "sink_id": "65429c9e-a0c6-4032-a421-6899c394fa74",
-          "source_name": "result",
-          "sink_name": "value",
-          "is_static": false
-        }
-      ],
-      "graph_id": "858e2226-e047-4d19-a832-3be4a134d155",
-      "graph_version": 2,
-      "webhook_id": null
-    }
-  ],
-  "links": [
-    {
-      "id": "8cdb2f33-5b10-4cc2-8839-f8ccb70083a3",
-      "source_id": "bf4a15ff-b0c4-4032-a21b-5880224af690",
-      "sink_id": "65429c9e-a0c6-4032-a421-6899c394fa74",
-      "source_name": "result",
-      "sink_name": "value",
-      "is_static": false
-    },
-    {
-      "id": "432c7caa-49b9-4b70-bd21-2fa33a569601",
-      "source_id": "6762da5d-6915-4836-a431-6dcd7d36a54a",
-      "sink_id": "bf4a15ff-b0c4-4032-a21b-5880224af690",
-      "source_name": "result",
-      "sink_name": "a",
-      "is_static": true
-    }
-  ],
-  "sub_graphs": [],
-  "input_schema": {
-    "type": "object",
-    "properties": {
-      "Input": {
-        "advanced": false,
-        "secret": false,
-        "title": "Input"
-      }
-    },
-    "required": [
-      "Input"
-    ]
-  },
-  "output_schema": {
-    "type": "object",
-    "properties": {
-      "Output": {
-        "advanced": false,
-        "secret": false,
-        "title": "Output"
-      }
-    },
-    "required": [
-      "Output"
-    ]
-  },
-  "has_external_trigger": false,
-  "has_human_in_the_loop": false,
-  "has_sensitive_action": false,
-  "trigger_setup_info": null,
-  "credentials_input_schema": {
-    "type": "object",
-    "properties": {},
-    "required": []
-  }
-}

autogpt_platform/backend/backend/api/conftest.py

@@ -1,9 +1,4 @@
-"""Common test fixtures for server tests.
-
-Note: Common fixtures like test_user_id, admin_user_id, target_user_id,
-setup_test_user, and setup_admin_user are defined in the parent conftest.py
-(backend/conftest.py) and are available here automatically.
-"""
+"""Common test fixtures for server tests."""
 
 import pytest
 from pytest_snapshot.plugin import Snapshot
@@ -16,6 +11,54 @@ def configured_snapshot(snapshot: Snapshot) -> Snapshot:
     return snapshot
 
 
+@pytest.fixture
+def test_user_id() -> str:
+    """Test user ID fixture."""
+    return "3e53486c-cf57-477e-ba2a-cb02dc828e1a"
+
+
+@pytest.fixture
+def admin_user_id() -> str:
+    """Admin user ID fixture."""
+    return "4e53486c-cf57-477e-ba2a-cb02dc828e1b"
+
+
+@pytest.fixture
+def target_user_id() -> str:
+    """Target user ID fixture."""
+    return "5e53486c-cf57-477e-ba2a-cb02dc828e1c"
+
+
+@pytest.fixture
+async def setup_test_user(test_user_id):
+    """Create test user in database before tests."""
+    from backend.data.user import get_or_create_user
+
+    # Create the test user in the database using JWT token format
+    user_data = {
+        "sub": test_user_id,
+        "email": "test@example.com",
+        "user_metadata": {"name": "Test User"},
+    }
+    await get_or_create_user(user_data)
+    return test_user_id
+
+
+@pytest.fixture
+async def setup_admin_user(admin_user_id):
+    """Create admin user in database before tests."""
+    from backend.data.user import get_or_create_user
+
+    # Create the admin user in the database using JWT token format
+    user_data = {
+        "sub": admin_user_id,
+        "email": "test-admin@example.com",
+        "user_metadata": {"name": "Test Admin"},
+    }
+    await get_or_create_user(user_data)
+    return admin_user_id
+
+
 @pytest.fixture
 def mock_jwt_user(test_user_id):
     """Provide mock JWT payload for regular user testing."""

autogpt_platform/backend/backend/api/external/middleware.py

@@ -88,23 +88,20 @@ async def require_auth(
     )
 
 
-def require_permission(*permissions: APIKeyPermission):
+def require_permission(permission: APIKeyPermission):
     """
-    Dependency function for checking required permissions.
-    All listed permissions must be present.
+    Dependency function for checking specific permissions
     (works with API keys and OAuth tokens)
     """
 
-    async def check_permissions(
+    async def check_permission(
         auth: APIAuthorizationInfo = Security(require_auth),
     ) -> APIAuthorizationInfo:
-        missing = [p for p in permissions if p not in auth.scopes]
-        if missing:
+        if permission not in auth.scopes:
             raise HTTPException(
                 status_code=status.HTTP_403_FORBIDDEN,
-                detail=f"Missing required permission(s): "
-                f"{', '.join(p.value for p in missing)}",
+                detail=f"Missing required permission: {permission.value}",
             )
         return auth
 
-    return check_permissions
+    return check_permission

autogpt_platform/backend/backend/api/external/v1/integrations.py

@@ -18,22 +18,14 @@ from pydantic import BaseModel, Field, SecretStr
 
 from backend.api.external.middleware import require_permission
 from backend.api.features.integrations.models import get_all_provider_names
-from backend.api.features.integrations.router import (
-    CredentialsMetaResponse,
-    to_meta_response,
-)
 from backend.data.auth.base import APIAuthorizationInfo
 from backend.data.model import (
     APIKeyCredentials,
     Credentials,
     CredentialsType,
     HostScopedCredentials,
+    OAuth2Credentials,
     UserPasswordCredentials,
-    is_sdk_default,
-)
-from backend.integrations.credentials_store import (
-    is_system_credential,
-    provider_matches,
 )
 from backend.integrations.creds_manager import IntegrationCredentialsManager
 from backend.integrations.oauth import CREDENTIALS_BY_PROVIDER, HANDLERS_BY_NAME
@@ -99,6 +91,18 @@ class OAuthCompleteResponse(BaseModel):
     )
 
 
+class CredentialSummary(BaseModel):
+    """Summary of a credential without sensitive data."""
+
+    id: str
+    provider: str
+    type: CredentialsType
+    title: Optional[str] = None
+    scopes: Optional[list[str]] = None
+    username: Optional[str] = None
+    host: Optional[str] = None
+
+
 class ProviderInfo(BaseModel):
     """Information about an integration provider."""
 
@@ -469,12 +473,12 @@ async def complete_oauth(
     )
 
 
-@integrations_router.get("/credentials", response_model=list[CredentialsMetaResponse])
+@integrations_router.get("/credentials", response_model=list[CredentialSummary])
 async def list_credentials(
     auth: APIAuthorizationInfo = Security(
         require_permission(APIKeyPermission.READ_INTEGRATIONS)
     ),
-) -> list[CredentialsMetaResponse]:
+) -> list[CredentialSummary]:
     """
     List all credentials for the authenticated user.
 
@@ -482,19 +486,28 @@ async def list_credentials(
     """
     credentials = await creds_manager.store.get_all_creds(auth.user_id)
     return [
-        to_meta_response(cred) for cred in credentials if not is_sdk_default(cred.id)
+        CredentialSummary(
+            id=cred.id,
+            provider=cred.provider,
+            type=cred.type,
+            title=cred.title,
+            scopes=cred.scopes if isinstance(cred, OAuth2Credentials) else None,
+            username=cred.username if isinstance(cred, OAuth2Credentials) else None,
+            host=cred.host if isinstance(cred, HostScopedCredentials) else None,
+        )
+        for cred in credentials
     ]
 
 
 @integrations_router.get(
-    "/{provider}/credentials", response_model=list[CredentialsMetaResponse]
+    "/{provider}/credentials", response_model=list[CredentialSummary]
 )
 async def list_credentials_by_provider(
     provider: Annotated[str, Path(title="The provider to list credentials for")],
     auth: APIAuthorizationInfo = Security(
         require_permission(APIKeyPermission.READ_INTEGRATIONS)
     ),
-) -> list[CredentialsMetaResponse]:
+) -> list[CredentialSummary]:
     """
     List credentials for a specific provider.
     """
@@ -502,7 +515,16 @@ async def list_credentials_by_provider(
         auth.user_id, provider
     )
     return [
-        to_meta_response(cred) for cred in credentials if not is_sdk_default(cred.id)
+        CredentialSummary(
+            id=cred.id,
+            provider=cred.provider,
+            type=cred.type,
+            title=cred.title,
+            scopes=cred.scopes if isinstance(cred, OAuth2Credentials) else None,
+            username=cred.username if isinstance(cred, OAuth2Credentials) else None,
+            host=cred.host if isinstance(cred, HostScopedCredentials) else None,
+        )
+        for cred in credentials
     ]
 
 
@@ -575,11 +597,11 @@ async def create_credential(
     # Store credentials
     try:
         await creds_manager.create(auth.user_id, credentials)
-    except Exception:
-        logger.exception("Failed to store credentials")
+    except Exception as e:
+        logger.error(f"Failed to store credentials: {e}")
         raise HTTPException(
             status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
-            detail="Failed to store credentials",
+            detail=f"Failed to store credentials: {str(e)}",
         )
 
     logger.info(f"Created {request.type} credentials for provider {provider}")
@@ -617,23 +639,15 @@ async def delete_credential(
     use the main API's delete endpoint which handles webhook cleanup and
     token revocation.
     """
-    if is_sdk_default(cred_id):
-        raise HTTPException(
-            status_code=status.HTTP_404_NOT_FOUND, detail="Credentials not found"
-        )
-    if is_system_credential(cred_id):
-        raise HTTPException(
-            status_code=status.HTTP_403_FORBIDDEN,
-            detail="System-managed credentials cannot be deleted",
-        )
     creds = await creds_manager.store.get_creds_by_id(auth.user_id, cred_id)
     if not creds:
         raise HTTPException(
             status_code=status.HTTP_404_NOT_FOUND, detail="Credentials not found"
         )
-    if not provider_matches(creds.provider, provider):
+    if creds.provider != provider:
         raise HTTPException(
-            status_code=status.HTTP_404_NOT_FOUND, detail="Credentials not found"
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail="Credentials do not match the specified provider",
         )
 
     await creds_manager.delete(auth.user_id, cred_id)

autogpt_platform/backend/backend/api/external/v1/routes.py

@@ -1,7 +1,7 @@
 import logging
 import urllib.parse
 from collections import defaultdict
-from typing import Annotated, Any, Optional, Sequence
+from typing import Annotated, Any, Literal, Optional, Sequence
 
 from fastapi import APIRouter, Body, HTTPException, Security
 from prisma.enums import AgentExecutionStatus, APIKeyPermission
@@ -9,17 +9,15 @@ from pydantic import BaseModel, Field
 from typing_extensions import TypedDict
 
 import backend.api.features.store.cache as store_cache
-import backend.api.features.store.db as store_db
 import backend.api.features.store.model as store_model
 import backend.blocks
-from backend.api.external.middleware import require_auth, require_permission
+from backend.api.external.middleware import require_permission
 from backend.data import execution as execution_db
 from backend.data import graph as graph_db
 from backend.data import user as user_db
 from backend.data.auth.base import APIAuthorizationInfo
 from backend.data.block import BlockInput, CompletedBlockOutput
 from backend.executor.utils import add_graph_execution
-from backend.integrations.webhooks.graph_lifecycle_hooks import on_graph_activate
 from backend.util.settings import Settings
 
 from .integrations import integrations_router
@@ -97,43 +95,6 @@ async def execute_graph_block(
     return output
 
 
-@v1_router.post(
-    path="/graphs",
-    tags=["graphs"],
-    status_code=201,
-    dependencies=[
-        Security(
-            require_permission(
-                APIKeyPermission.WRITE_GRAPH, APIKeyPermission.WRITE_LIBRARY
-            )
-        )
-    ],
-)
-async def create_graph(
-    graph: graph_db.Graph,
-    auth: APIAuthorizationInfo = Security(
-        require_permission(APIKeyPermission.WRITE_GRAPH, APIKeyPermission.WRITE_LIBRARY)
-    ),
-) -> graph_db.GraphModel:
-    """
-    Create a new agent graph.
-
-    The graph will be validated and assigned a new ID.
-    It is automatically added to the user's library.
-    """
-    from backend.api.features.library import db as library_db
-
-    graph_model = graph_db.make_graph_model(graph, auth.user_id)
-    graph_model.reassign_ids(user_id=auth.user_id, reassign_graph_id=True)
-    graph_model.validate_graph(for_run=False)
-
-    await graph_db.create_graph(graph_model, user_id=auth.user_id)
-    await library_db.create_library_agent(graph_model, auth.user_id)
-    activated_graph = await on_graph_activate(graph_model, user_id=auth.user_id)
-
-    return activated_graph
-
-
 @v1_router.post(
     path="/graphs/{graph_id}/execute/{graph_version}",
     tags=["graphs"],
@@ -231,13 +192,13 @@ async def get_graph_execution_results(
 @v1_router.get(
     path="/store/agents",
     tags=["store"],
-    dependencies=[Security(require_auth)],  # data is public; auth required as anti-DDoS
+    dependencies=[Security(require_permission(APIKeyPermission.READ_STORE))],
     response_model=store_model.StoreAgentsResponse,
 )
 async def get_store_agents(
     featured: bool = False,
     creator: str | None = None,
-    sorted_by: store_db.StoreAgentsSortOptions | None = None,
+    sorted_by: Literal["rating", "runs", "name", "updated_at"] | None = None,
     search_query: str | None = None,
     category: str | None = None,
     page: int = 1,
@@ -279,7 +240,7 @@ async def get_store_agents(
 @v1_router.get(
     path="/store/agents/{username}/{agent_name}",
     tags=["store"],
-    dependencies=[Security(require_auth)],  # data is public; auth required as anti-DDoS
+    dependencies=[Security(require_permission(APIKeyPermission.READ_STORE))],
     response_model=store_model.StoreAgentDetails,
 )
 async def get_store_agent(
@@ -307,13 +268,13 @@ async def get_store_agent(
 @v1_router.get(
     path="/store/creators",
     tags=["store"],
-    dependencies=[Security(require_auth)],  # data is public; auth required as anti-DDoS
+    dependencies=[Security(require_permission(APIKeyPermission.READ_STORE))],
     response_model=store_model.CreatorsResponse,
 )
 async def get_store_creators(
     featured: bool = False,
     search_query: str | None = None,
-    sorted_by: store_db.StoreCreatorsSortOptions | None = None,
+    sorted_by: Literal["agent_rating", "agent_runs", "num_agents"] | None = None,
     page: int = 1,
     page_size: int = 20,
 ) -> store_model.CreatorsResponse:
@@ -349,7 +310,7 @@ async def get_store_creators(
 @v1_router.get(
     path="/store/creators/{username}",
     tags=["store"],
-    dependencies=[Security(require_auth)],  # data is public; auth required as anti-DDoS
+    dependencies=[Security(require_permission(APIKeyPermission.READ_STORE))],
     response_model=store_model.CreatorDetails,
 )
 async def get_store_creator(

autogpt_platform/backend/backend/api/external/v1/tools.py

@@ -15,9 +15,9 @@ from prisma.enums import APIKeyPermission
 from pydantic import BaseModel, Field
 
 from backend.api.external.middleware import require_permission
-from backend.copilot.model import ChatSession
-from backend.copilot.tools import find_agent_tool, run_agent_tool
-from backend.copilot.tools.models import ToolResponseBase
+from backend.api.features.chat.model import ChatSession
+from backend.api.features.chat.tools import find_agent_tool, run_agent_tool
+from backend.api.features.chat.tools.models import ToolResponseBase
 from backend.data.auth.base import APIAuthorizationInfo
 
 logger = logging.getLogger(__name__)
@@ -72,7 +72,7 @@ class RunAgentRequest(BaseModel):
 
 def _create_ephemeral_session(user_id: str) -> ChatSession:
     """Create an ephemeral session for stateless API requests."""
-    return ChatSession.new(user_id, dry_run=False)
+    return ChatSession.new(user_id)
 
 
 @tools_router.post(

autogpt_platform/backend/backend/api/features/admin/diagnostics_admin_routes.py

@@ -1,932 +0,0 @@
-import asyncio
-import logging
-from typing import List
-
-from autogpt_libs.auth import requires_admin_user
-from autogpt_libs.auth.models import User as AuthUser
-from fastapi import APIRouter, HTTPException, Security
-from prisma.enums import AgentExecutionStatus
-from pydantic import BaseModel
-
-from backend.api.features.admin.model import (
-    AgentDiagnosticsResponse,
-    ExecutionDiagnosticsResponse,
-)
-from backend.data.diagnostics import (
-    FailedExecutionDetail,
-    OrphanedScheduleDetail,
-    RunningExecutionDetail,
-    ScheduleDetail,
-    ScheduleHealthMetrics,
-    cleanup_all_stuck_queued_executions,
-    cleanup_orphaned_executions_bulk,
-    cleanup_orphaned_schedules_bulk,
-    get_agent_diagnostics,
-    get_all_orphaned_execution_ids,
-    get_all_schedules_details,
-    get_all_stuck_queued_execution_ids,
-    get_execution_diagnostics,
-    get_failed_executions_count,
-    get_failed_executions_details,
-    get_invalid_executions_details,
-    get_long_running_executions_details,
-    get_orphaned_executions_details,
-    get_orphaned_schedules_details,
-    get_running_executions_details,
-    get_schedule_health_metrics,
-    get_stuck_queued_executions_details,
-    stop_all_long_running_executions,
-)
-from backend.data.execution import get_graph_executions
-from backend.executor.utils import add_graph_execution, stop_graph_execution
-
-logger = logging.getLogger(__name__)
-
-router = APIRouter(
-    prefix="/admin",
-    tags=["diagnostics", "admin"],
-    dependencies=[Security(requires_admin_user)],
-)
-
-
-class RunningExecutionsListResponse(BaseModel):
-    """Response model for list of running executions"""
-
-    executions: List[RunningExecutionDetail]
-    total: int
-
-
-class FailedExecutionsListResponse(BaseModel):
-    """Response model for list of failed executions"""
-
-    executions: List[FailedExecutionDetail]
-    total: int
-
-
-class StopExecutionRequest(BaseModel):
-    """Request model for stopping a single execution"""
-
-    execution_id: str
-
-
-class StopExecutionsRequest(BaseModel):
-    """Request model for stopping multiple executions"""
-
-    execution_ids: List[str]
-
-
-class StopExecutionResponse(BaseModel):
-    """Response model for stop execution operations"""
-
-    success: bool
-    stopped_count: int = 0
-    message: str
-
-
-class RequeueExecutionResponse(BaseModel):
-    """Response model for requeue execution operations"""
-
-    success: bool
-    requeued_count: int = 0
-    message: str
-
-
-@router.get(
-    "/diagnostics/executions",
-    response_model=ExecutionDiagnosticsResponse,
-    summary="Get Execution Diagnostics",
-)
-async def get_execution_diagnostics_endpoint():
-    """
-    Get comprehensive diagnostic information about execution status.
-
-    Returns all execution metrics including:
-    - Current state (running, queued)
-    - Orphaned executions (>24h old, likely not in executor)
-    - Failure metrics (1h, 24h, rate)
-    - Long-running detection (stuck >1h, >24h)
-    - Stuck queued detection
-    - Throughput metrics (completions/hour)
-    - RabbitMQ queue depths
-    """
-    logger.info("Getting execution diagnostics")
-
-    diagnostics = await get_execution_diagnostics()
-
-    response = ExecutionDiagnosticsResponse(
-        running_executions=diagnostics.running_count,
-        queued_executions_db=diagnostics.queued_db_count,
-        queued_executions_rabbitmq=diagnostics.rabbitmq_queue_depth,
-        cancel_queue_depth=diagnostics.cancel_queue_depth,
-        orphaned_running=diagnostics.orphaned_running,
-        orphaned_queued=diagnostics.orphaned_queued,
-        failed_count_1h=diagnostics.failed_count_1h,
-        failed_count_24h=diagnostics.failed_count_24h,
-        failure_rate_24h=diagnostics.failure_rate_24h,
-        stuck_running_24h=diagnostics.stuck_running_24h,
-        stuck_running_1h=diagnostics.stuck_running_1h,
-        oldest_running_hours=diagnostics.oldest_running_hours,
-        stuck_queued_1h=diagnostics.stuck_queued_1h,
-        queued_never_started=diagnostics.queued_never_started,
-        invalid_queued_with_start=diagnostics.invalid_queued_with_start,
-        invalid_running_without_start=diagnostics.invalid_running_without_start,
-        completed_1h=diagnostics.completed_1h,
-        completed_24h=diagnostics.completed_24h,
-        throughput_per_hour=diagnostics.throughput_per_hour,
-        timestamp=diagnostics.timestamp,
-    )
-
-    logger.info(
-        f"Execution diagnostics: running={diagnostics.running_count}, "
-        f"queued_db={diagnostics.queued_db_count}, "
-        f"orphaned={diagnostics.orphaned_running + diagnostics.orphaned_queued}, "
-        f"failed_24h={diagnostics.failed_count_24h}"
-    )
-
-    return response
-
-
-@router.get(
-    "/diagnostics/agents",
-    response_model=AgentDiagnosticsResponse,
-    summary="Get Agent Diagnostics",
-)
-async def get_agent_diagnostics_endpoint():
-    """
-    Get diagnostic information about agents.
-
-    Returns:
-        - agents_with_active_executions: Number of unique agents with running/queued executions
-        - timestamp: Current timestamp
-    """
-    logger.info("Getting agent diagnostics")
-
-    diagnostics = await get_agent_diagnostics()
-
-    response = AgentDiagnosticsResponse(
-        agents_with_active_executions=diagnostics.agents_with_active_executions,
-        timestamp=diagnostics.timestamp,
-    )
-
-    logger.info(
-        f"Agent diagnostics: with_active_executions={diagnostics.agents_with_active_executions}"
-    )
-
-    return response
-
-
-@router.get(
-    "/diagnostics/executions/running",
-    response_model=RunningExecutionsListResponse,
-    summary="List Running Executions",
-)
-async def list_running_executions(
-    limit: int = 100,
-    offset: int = 0,
-):
-    """
-    Get detailed list of running and queued executions (recent, likely active).
-
-    Args:
-        limit: Maximum number of executions to return (default 100)
-        offset: Number of executions to skip (default 0)
-
-    Returns:
-        List of running executions with details
-    """
-    logger.info(f"Listing running executions (limit={limit}, offset={offset})")
-
-    executions = await get_running_executions_details(limit=limit, offset=offset)
-
-    # Get total count for pagination
-    diagnostics = await get_execution_diagnostics()
-    total = diagnostics.running_count + diagnostics.queued_db_count
-
-    return RunningExecutionsListResponse(executions=executions, total=total)
-
-
-@router.get(
-    "/diagnostics/executions/orphaned",
-    response_model=RunningExecutionsListResponse,
-    summary="List Orphaned Executions",
-)
-async def list_orphaned_executions(
-    limit: int = 100,
-    offset: int = 0,
-):
-    """
-    Get detailed list of orphaned executions (>24h old, likely not in executor).
-
-    Args:
-        limit: Maximum number of executions to return (default 100)
-        offset: Number of executions to skip (default 0)
-
-    Returns:
-        List of orphaned executions with details
-    """
-    logger.info(f"Listing orphaned executions (limit={limit}, offset={offset})")
-
-    executions = await get_orphaned_executions_details(limit=limit, offset=offset)
-
-    # Get total count for pagination
-    diagnostics = await get_execution_diagnostics()
-    total = diagnostics.orphaned_running + diagnostics.orphaned_queued
-
-    return RunningExecutionsListResponse(executions=executions, total=total)
-
-
-@router.get(
-    "/diagnostics/executions/failed",
-    response_model=FailedExecutionsListResponse,
-    summary="List Failed Executions",
-)
-async def list_failed_executions(
-    limit: int = 100,
-    offset: int = 0,
-    hours: int = 24,
-):
-    """
-    Get detailed list of failed executions.
-
-    Args:
-        limit: Maximum number of executions to return (default 100)
-        offset: Number of executions to skip (default 0)
-        hours: Number of hours to look back (default 24)
-
-    Returns:
-        List of failed executions with error details
-    """
-    logger.info(
-        f"Listing failed executions (limit={limit}, offset={offset}, hours={hours})"
-    )
-
-    executions = await get_failed_executions_details(
-        limit=limit, offset=offset, hours=hours
-    )
-
-    # Get total count for pagination
-    # Always count actual total for given hours parameter
-    total = await get_failed_executions_count(hours=hours)
-
-    return FailedExecutionsListResponse(executions=executions, total=total)
-
-
-@router.get(
-    "/diagnostics/executions/long-running",
-    response_model=RunningExecutionsListResponse,
-    summary="List Long-Running Executions",
-)
-async def list_long_running_executions(
-    limit: int = 100,
-    offset: int = 0,
-):
-    """
-    Get detailed list of long-running executions (RUNNING status >24h).
-
-    Args:
-        limit: Maximum number of executions to return (default 100)
-        offset: Number of executions to skip (default 0)
-
-    Returns:
-        List of long-running executions with details
-    """
-    logger.info(f"Listing long-running executions (limit={limit}, offset={offset})")
-
-    executions = await get_long_running_executions_details(limit=limit, offset=offset)
-
-    # Get total count for pagination
-    diagnostics = await get_execution_diagnostics()
-    total = diagnostics.stuck_running_24h
-
-    return RunningExecutionsListResponse(executions=executions, total=total)
-
-
-@router.get(
-    "/diagnostics/executions/stuck-queued",
-    response_model=RunningExecutionsListResponse,
-    summary="List Stuck Queued Executions",
-)
-async def list_stuck_queued_executions(
-    limit: int = 100,
-    offset: int = 0,
-):
-    """
-    Get detailed list of stuck queued executions (QUEUED >1h, never started).
-
-    Args:
-        limit: Maximum number of executions to return (default 100)
-        offset: Number of executions to skip (default 0)
-
-    Returns:
-        List of stuck queued executions with details
-    """
-    logger.info(f"Listing stuck queued executions (limit={limit}, offset={offset})")
-
-    executions = await get_stuck_queued_executions_details(limit=limit, offset=offset)
-
-    # Get total count for pagination
-    diagnostics = await get_execution_diagnostics()
-    total = diagnostics.stuck_queued_1h
-
-    return RunningExecutionsListResponse(executions=executions, total=total)
-
-
-@router.get(
-    "/diagnostics/executions/invalid",
-    response_model=RunningExecutionsListResponse,
-    summary="List Invalid Executions",
-)
-async def list_invalid_executions(
-    limit: int = 100,
-    offset: int = 0,
-):
-    """
-    Get detailed list of executions in invalid states (READ-ONLY).
-
-    Invalid states indicate data corruption and require manual investigation:
-    - QUEUED but has startedAt (impossible - can't start while queued)
-    - RUNNING but no startedAt (impossible - can't run without starting)
-
-    ⚠️ NO BULK ACTIONS PROVIDED - These need case-by-case investigation.
-
-    Each invalid execution likely has a different root cause (crashes, race conditions,
-    DB corruption). Investigate the execution history and logs to determine appropriate
-    action (manual cleanup, status fix, or leave as-is if system recovered).
-
-    Args:
-        limit: Maximum number of executions to return (default 100)
-        offset: Number of executions to skip (default 0)
-
-    Returns:
-        List of invalid state executions with details
-    """
-    logger.info(f"Listing invalid state executions (limit={limit}, offset={offset})")
-
-    executions = await get_invalid_executions_details(limit=limit, offset=offset)
-
-    # Get total count for pagination
-    diagnostics = await get_execution_diagnostics()
-    total = (
-        diagnostics.invalid_queued_with_start
-        + diagnostics.invalid_running_without_start
-    )
-
-    return RunningExecutionsListResponse(executions=executions, total=total)
-
-
-@router.post(
-    "/diagnostics/executions/requeue",
-    response_model=RequeueExecutionResponse,
-    summary="Requeue Stuck Execution",
-)
-async def requeue_single_execution(
-    request: StopExecutionRequest,  # Reuse same request model (has execution_id)
-    user: AuthUser = Security(requires_admin_user),
-):
-    """
-    Requeue a stuck QUEUED execution (admin only).
-
-    Uses add_graph_execution with existing graph_exec_id to requeue.
-
-    ⚠️ WARNING: Only use for stuck executions. This will re-execute and may cost credits.
-
-    Args:
-        request: Contains execution_id to requeue
-
-    Returns:
-        Success status and message
-    """
-    logger.info(f"Admin {user.user_id} requeueing execution {request.execution_id}")
-
-    # Get the execution (validation - must be QUEUED)
-    executions = await get_graph_executions(
-        graph_exec_id=request.execution_id,
-        statuses=[AgentExecutionStatus.QUEUED],
-    )
-
-    if not executions:
-        raise HTTPException(
-            status_code=404,
-            detail="Execution not found or not in QUEUED status",
-        )
-
-    execution = executions[0]
-
-    # Use add_graph_execution in requeue mode
-    await add_graph_execution(
-        graph_id=execution.graph_id,
-        user_id=execution.user_id,
-        graph_version=execution.graph_version,
-        graph_exec_id=request.execution_id,  # Requeue existing execution
-    )
-
-    return RequeueExecutionResponse(
-        success=True,
-        requeued_count=1,
-        message="Execution requeued successfully",
-    )
-
-
-@router.post(
-    "/diagnostics/executions/requeue-bulk",
-    response_model=RequeueExecutionResponse,
-    summary="Requeue Multiple Stuck Executions",
-)
-async def requeue_multiple_executions(
-    request: StopExecutionsRequest,  # Reuse same request model (has execution_ids)
-    user: AuthUser = Security(requires_admin_user),
-):
-    """
-    Requeue multiple stuck QUEUED executions (admin only).
-
-    Uses add_graph_execution with existing graph_exec_id to requeue.
-
-    ⚠️ WARNING: Only use for stuck executions. This will re-execute and may cost credits.
-
-    Args:
-        request: Contains list of execution_ids to requeue
-
-    Returns:
-        Number of executions requeued and success message
-    """
-    logger.info(
-        f"Admin {user.user_id} requeueing {len(request.execution_ids)} executions"
-    )
-
-    # Get executions by ID list (must be QUEUED)
-    executions = await get_graph_executions(
-        execution_ids=request.execution_ids,
-        statuses=[AgentExecutionStatus.QUEUED],
-    )
-
-    if not executions:
-        return RequeueExecutionResponse(
-            success=False,
-            requeued_count=0,
-            message="No QUEUED executions found to requeue",
-        )
-
-    # Requeue all executions in parallel using add_graph_execution
-    async def requeue_one(exec) -> bool:
-        try:
-            await add_graph_execution(
-                graph_id=exec.graph_id,
-                user_id=exec.user_id,
-                graph_version=exec.graph_version,
-                graph_exec_id=exec.id,  # Requeue existing
-            )
-            return True
-        except Exception as e:
-            logger.error(f"Failed to requeue {exec.id}: {e}")
-            return False
-
-    results = await asyncio.gather(
-        *[requeue_one(exec) for exec in executions], return_exceptions=False
-    )
-
-    requeued_count = sum(1 for success in results if success)
-
-    return RequeueExecutionResponse(
-        success=requeued_count > 0,
-        requeued_count=requeued_count,
-        message=f"Requeued {requeued_count} of {len(request.execution_ids)} executions",
-    )
-
-
-@router.post(
-    "/diagnostics/executions/stop",
-    response_model=StopExecutionResponse,
-    summary="Stop Single Execution",
-)
-async def stop_single_execution(
-    request: StopExecutionRequest,
-    user: AuthUser = Security(requires_admin_user),
-):
-    """
-    Stop a single execution (admin only).
-
-    Uses robust stop_graph_execution which cascades to children and waits for termination.
-
-    Args:
-        request: Contains execution_id to stop
-
-    Returns:
-        Success status and message
-    """
-    logger.info(f"Admin {user.user_id} stopping execution {request.execution_id}")
-
-    # Get the execution to find its owner user_id (required by stop_graph_execution)
-    executions = await get_graph_executions(
-        graph_exec_id=request.execution_id,
-    )
-
-    if not executions:
-        raise HTTPException(status_code=404, detail="Execution not found")
-
-    execution = executions[0]
-
-    # Use robust stop_graph_execution (cascades to children, waits for termination)
-    await stop_graph_execution(
-        user_id=execution.user_id,
-        graph_exec_id=request.execution_id,
-        wait_timeout=15.0,
-        cascade=True,
-    )
-
-    return StopExecutionResponse(
-        success=True,
-        stopped_count=1,
-        message="Execution stopped successfully",
-    )
-
-
-@router.post(
-    "/diagnostics/executions/stop-bulk",
-    response_model=StopExecutionResponse,
-    summary="Stop Multiple Executions",
-)
-async def stop_multiple_executions(
-    request: StopExecutionsRequest,
-    user: AuthUser = Security(requires_admin_user),
-):
-    """
-    Stop multiple active executions (admin only).
-
-    Uses robust stop_graph_execution which cascades to children and waits for termination.
-
-    Args:
-        request: Contains list of execution_ids to stop
-
-    Returns:
-        Number of executions stopped and success message
-    """
-
-    logger.info(
-        f"Admin {user.user_id} stopping {len(request.execution_ids)} executions"
-    )
-
-    # Get executions by ID list
-    executions = await get_graph_executions(
-        execution_ids=request.execution_ids,
-    )
-
-    if not executions:
-        return StopExecutionResponse(
-            success=False,
-            stopped_count=0,
-            message="No executions found",
-        )
-
-    # Stop all executions in parallel using robust stop_graph_execution
-    async def stop_one(exec) -> bool:
-        try:
-            await stop_graph_execution(
-                user_id=exec.user_id,
-                graph_exec_id=exec.id,
-                wait_timeout=15.0,
-                cascade=True,
-            )
-            return True
-        except Exception as e:
-            logger.error(f"Failed to stop execution {exec.id}: {e}")
-            return False
-
-    results = await asyncio.gather(
-        *[stop_one(exec) for exec in executions], return_exceptions=False
-    )
-
-    stopped_count = sum(1 for success in results if success)
-
-    return StopExecutionResponse(
-        success=stopped_count > 0,
-        stopped_count=stopped_count,
-        message=f"Stopped {stopped_count} of {len(request.execution_ids)} executions",
-    )
-
-
-@router.post(
-    "/diagnostics/executions/cleanup-orphaned",
-    response_model=StopExecutionResponse,
-    summary="Cleanup Orphaned Executions",
-)
-async def cleanup_orphaned_executions(
-    request: StopExecutionsRequest,
-    user: AuthUser = Security(requires_admin_user),
-):
-    """
-    Cleanup orphaned executions by directly updating DB status (admin only).
-    For executions in DB but not actually running in executor (old/stale records).
-
-    Args:
-        request: Contains list of execution_ids to cleanup
-
-    Returns:
-        Number of executions cleaned up and success message
-    """
-    logger.info(
-        f"Admin {user.user_id} cleaning up {len(request.execution_ids)} orphaned executions"
-    )
-
-    cleaned_count = await cleanup_orphaned_executions_bulk(
-        request.execution_ids, user.user_id
-    )
-
-    return StopExecutionResponse(
-        success=cleaned_count > 0,
-        stopped_count=cleaned_count,
-        message=f"Cleaned up {cleaned_count} of {len(request.execution_ids)} orphaned executions",
-    )
-
-
-# ============================================================================
-# SCHEDULE DIAGNOSTICS ENDPOINTS
-# ============================================================================
-
-
-class SchedulesListResponse(BaseModel):
-    """Response model for list of schedules"""
-
-    schedules: List[ScheduleDetail]
-    total: int
-
-
-class OrphanedSchedulesListResponse(BaseModel):
-    """Response model for list of orphaned schedules"""
-
-    schedules: List[OrphanedScheduleDetail]
-    total: int
-
-
-class ScheduleCleanupRequest(BaseModel):
-    """Request model for cleaning up schedules"""
-
-    schedule_ids: List[str]
-
-
-class ScheduleCleanupResponse(BaseModel):
-    """Response model for schedule cleanup operations"""
-
-    success: bool
-    deleted_count: int = 0
-    message: str
-
-
-@router.get(
-    "/diagnostics/schedules",
-    response_model=ScheduleHealthMetrics,
-    summary="Get Schedule Diagnostics",
-)
-async def get_schedule_diagnostics_endpoint():
-    """
-    Get comprehensive diagnostic information about schedule health.
-
-    Returns schedule metrics including:
-    - Total schedules (user vs system)
-    - Orphaned schedules by category
-    - Upcoming executions
-    """
-    logger.info("Getting schedule diagnostics")
-
-    diagnostics = await get_schedule_health_metrics()
-
-    logger.info(
-        f"Schedule diagnostics: total={diagnostics.total_schedules}, "
-        f"user={diagnostics.user_schedules}, "
-        f"orphaned={diagnostics.total_orphaned}"
-    )
-
-    return diagnostics
-
-
-@router.get(
-    "/diagnostics/schedules/all",
-    response_model=SchedulesListResponse,
-    summary="List All User Schedules",
-)
-async def list_all_schedules(
-    limit: int = 100,
-    offset: int = 0,
-):
-    """
-    Get detailed list of all user schedules (excludes system monitoring jobs).
-
-    Args:
-        limit: Maximum number of schedules to return (default 100)
-        offset: Number of schedules to skip (default 0)
-
-    Returns:
-        List of schedules with details
-    """
-    logger.info(f"Listing all schedules (limit={limit}, offset={offset})")
-
-    schedules = await get_all_schedules_details(limit=limit, offset=offset)
-
-    # Get total count
-    diagnostics = await get_schedule_health_metrics()
-    total = diagnostics.user_schedules
-
-    return SchedulesListResponse(schedules=schedules, total=total)
-
-
-@router.get(
-    "/diagnostics/schedules/orphaned",
-    response_model=OrphanedSchedulesListResponse,
-    summary="List Orphaned Schedules",
-)
-async def list_orphaned_schedules():
-    """
-    Get detailed list of orphaned schedules with orphan reasons.
-
-    Returns:
-        List of orphaned schedules categorized by orphan type
-    """
-    logger.info("Listing orphaned schedules")
-
-    schedules = await get_orphaned_schedules_details()
-
-    return OrphanedSchedulesListResponse(schedules=schedules, total=len(schedules))
-
-
-@router.post(
-    "/diagnostics/schedules/cleanup-orphaned",
-    response_model=ScheduleCleanupResponse,
-    summary="Cleanup Orphaned Schedules",
-)
-async def cleanup_orphaned_schedules(
-    request: ScheduleCleanupRequest,
-    user: AuthUser = Security(requires_admin_user),
-):
-    """
-    Cleanup orphaned schedules by deleting from scheduler (admin only).
-
-    Args:
-        request: Contains list of schedule_ids to delete
-
-    Returns:
-        Number of schedules deleted and success message
-    """
-    logger.info(
-        f"Admin {user.user_id} cleaning up {len(request.schedule_ids)} orphaned schedules"
-    )
-
-    deleted_count = await cleanup_orphaned_schedules_bulk(
-        request.schedule_ids, user.user_id
-    )
-
-    return ScheduleCleanupResponse(
-        success=deleted_count > 0,
-        deleted_count=deleted_count,
-        message=f"Deleted {deleted_count} of {len(request.schedule_ids)} orphaned schedules",
-    )
-
-
-@router.post(
-    "/diagnostics/executions/stop-all-long-running",
-    response_model=StopExecutionResponse,
-    summary="Stop ALL Long-Running Executions",
-)
-async def stop_all_long_running_executions_endpoint(
-    user: AuthUser = Security(requires_admin_user),
-):
-    """
-    Stop ALL long-running executions (RUNNING >24h) by sending cancel signals (admin only).
-    Operates on entire dataset, not limited to pagination.
-
-    Returns:
-        Number of executions stopped and success message
-    """
-    logger.info(f"Admin {user.user_id} stopping ALL long-running executions")
-
-    stopped_count = await stop_all_long_running_executions(user.user_id)
-
-    return StopExecutionResponse(
-        success=stopped_count > 0,
-        stopped_count=stopped_count,
-        message=f"Stopped {stopped_count} long-running executions",
-    )
-
-
-@router.post(
-    "/diagnostics/executions/cleanup-all-orphaned",
-    response_model=StopExecutionResponse,
-    summary="Cleanup ALL Orphaned Executions",
-)
-async def cleanup_all_orphaned_executions(
-    user: AuthUser = Security(requires_admin_user),
-):
-    """
-    Cleanup ALL orphaned executions (>24h old) by directly updating DB status.
-    Operates on all executions, not just paginated results.
-
-    Returns:
-        Number of executions cleaned up and success message
-    """
-    logger.info(f"Admin {user.user_id} cleaning up ALL orphaned executions")
-
-    # Fetch all orphaned execution IDs
-    execution_ids = await get_all_orphaned_execution_ids()
-
-    if not execution_ids:
-        return StopExecutionResponse(
-            success=True,
-            stopped_count=0,
-            message="No orphaned executions to cleanup",
-        )
-
-    cleaned_count = await cleanup_orphaned_executions_bulk(execution_ids, user.user_id)
-
-    return StopExecutionResponse(
-        success=cleaned_count > 0,
-        stopped_count=cleaned_count,
-        message=f"Cleaned up {cleaned_count} orphaned executions",
-    )
-
-
-@router.post(
-    "/diagnostics/executions/cleanup-all-stuck-queued",
-    response_model=StopExecutionResponse,
-    summary="Cleanup ALL Stuck Queued Executions",
-)
-async def cleanup_all_stuck_queued_executions_endpoint(
-    user: AuthUser = Security(requires_admin_user),
-):
-    """
-    Cleanup ALL stuck queued executions (QUEUED >1h) by updating DB status (admin only).
-    Operates on entire dataset, not limited to pagination.
-
-    Returns:
-        Number of executions cleaned up and success message
-    """
-    logger.info(f"Admin {user.user_id} cleaning up ALL stuck queued executions")
-
-    cleaned_count = await cleanup_all_stuck_queued_executions(user.user_id)
-
-    return StopExecutionResponse(
-        success=cleaned_count > 0,
-        stopped_count=cleaned_count,
-        message=f"Cleaned up {cleaned_count} stuck queued executions",
-    )
-
-
-@router.post(
-    "/diagnostics/executions/requeue-all-stuck",
-    response_model=RequeueExecutionResponse,
-    summary="Requeue ALL Stuck Queued Executions",
-)
-async def requeue_all_stuck_executions(
-    user: AuthUser = Security(requires_admin_user),
-):
-    """
-    Requeue ALL stuck queued executions (QUEUED >1h) by publishing to RabbitMQ.
-    Operates on all executions, not just paginated results.
-
-    Uses add_graph_execution with existing graph_exec_id to requeue.
-
-    ⚠️ WARNING: This will re-execute ALL stuck executions and may cost significant credits.
-
-    Returns:
-        Number of executions requeued and success message
-    """
-    logger.info(f"Admin {user.user_id} requeueing ALL stuck queued executions")
-
-    # Fetch all stuck queued execution IDs
-    execution_ids = await get_all_stuck_queued_execution_ids()
-
-    if not execution_ids:
-        return RequeueExecutionResponse(
-            success=True,
-            requeued_count=0,
-            message="No stuck queued executions to requeue",
-        )
-
-    # Get stuck executions by ID list (must be QUEUED)
-    executions = await get_graph_executions(
-        execution_ids=execution_ids,
-        statuses=[AgentExecutionStatus.QUEUED],
-    )
-
-    # Requeue all in parallel using add_graph_execution
-    async def requeue_one(exec) -> bool:
-        try:
-            await add_graph_execution(
-                graph_id=exec.graph_id,
-                user_id=exec.user_id,
-                graph_version=exec.graph_version,
-                graph_exec_id=exec.id,  # Requeue existing
-            )
-            return True
-        except Exception as e:
-            logger.error(f"Failed to requeue {exec.id}: {e}")
-            return False
-
-    results = await asyncio.gather(
-        *[requeue_one(exec) for exec in executions], return_exceptions=False
-    )
-
-    requeued_count = sum(1 for success in results if success)
-
-    return RequeueExecutionResponse(
-        success=requeued_count > 0,
-        requeued_count=requeued_count,
-        message=f"Requeued {requeued_count} stuck executions",
-    )

autogpt_platform/backend/backend/api/features/admin/diagnostics_admin_routes_test.py

@@ -1,889 +0,0 @@
-from datetime import datetime, timezone
-from unittest.mock import AsyncMock
-
-import fastapi
-import fastapi.testclient
-import pytest
-import pytest_mock
-from autogpt_libs.auth.jwt_utils import get_jwt_payload
-from prisma.enums import AgentExecutionStatus
-
-import backend.api.features.admin.diagnostics_admin_routes as diagnostics_admin_routes
-from backend.data.diagnostics import (
-    AgentDiagnosticsSummary,
-    ExecutionDiagnosticsSummary,
-    FailedExecutionDetail,
-    OrphanedScheduleDetail,
-    RunningExecutionDetail,
-    ScheduleDetail,
-    ScheduleHealthMetrics,
-)
-from backend.data.execution import GraphExecutionMeta
-
-app = fastapi.FastAPI()
-app.include_router(diagnostics_admin_routes.router)
-
-client = fastapi.testclient.TestClient(app)
-
-
-@pytest.fixture(autouse=True)
-def setup_app_admin_auth(mock_jwt_admin):
-    """Setup admin auth overrides for all tests in this module"""
-    app.dependency_overrides[get_jwt_payload] = mock_jwt_admin["get_jwt_payload"]
-    yield
-    app.dependency_overrides.clear()
-
-
-def test_get_execution_diagnostics_success(
-    mocker: pytest_mock.MockFixture,
-):
-    """Test fetching execution diagnostics with invalid state detection"""
-    mock_diagnostics = ExecutionDiagnosticsSummary(
-        running_count=10,
-        queued_db_count=5,
-        rabbitmq_queue_depth=3,
-        cancel_queue_depth=0,
-        orphaned_running=2,
-        orphaned_queued=1,
-        failed_count_1h=5,
-        failed_count_24h=20,
-        failure_rate_24h=0.83,
-        stuck_running_24h=1,
-        stuck_running_1h=3,
-        oldest_running_hours=26.5,
-        stuck_queued_1h=2,
-        queued_never_started=1,
-        invalid_queued_with_start=1,  # New invalid state
-        invalid_running_without_start=1,  # New invalid state
-        completed_1h=50,
-        completed_24h=1200,
-        throughput_per_hour=50.0,
-        timestamp=datetime.now(timezone.utc).isoformat(),
-    )
-
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_execution_diagnostics",
-        return_value=mock_diagnostics,
-    )
-
-    response = client.get("/admin/diagnostics/executions")
-
-    assert response.status_code == 200
-    data = response.json()
-
-    # Verify new invalid state fields are included
-    assert data["invalid_queued_with_start"] == 1
-    assert data["invalid_running_without_start"] == 1
-    # Verify all expected fields present
-    assert "running_executions" in data
-    assert "orphaned_running" in data
-    assert "failed_count_24h" in data
-
-
-def test_list_invalid_executions(
-    mocker: pytest_mock.MockFixture,
-):
-    """Test listing executions in invalid states (read-only endpoint)"""
-    mock_invalid_executions = [
-        RunningExecutionDetail(
-            execution_id="exec-invalid-1",
-            graph_id="graph-123",
-            graph_name="Test Graph",
-            graph_version=1,
-            user_id="user-123",
-            user_email="test@example.com",
-            status="QUEUED",
-            created_at=datetime.now(timezone.utc),
-            started_at=datetime.now(
-                timezone.utc
-            ),  # QUEUED but has startedAt - INVALID!
-            queue_status=None,
-        ),
-        RunningExecutionDetail(
-            execution_id="exec-invalid-2",
-            graph_id="graph-456",
-            graph_name="Another Graph",
-            graph_version=2,
-            user_id="user-456",
-            user_email="user@example.com",
-            status="RUNNING",
-            created_at=datetime.now(timezone.utc),
-            started_at=None,  # RUNNING but no startedAt - INVALID!
-            queue_status=None,
-        ),
-    ]
-
-    mock_diagnostics = ExecutionDiagnosticsSummary(
-        running_count=10,
-        queued_db_count=5,
-        rabbitmq_queue_depth=3,
-        cancel_queue_depth=0,
-        orphaned_running=0,
-        orphaned_queued=0,
-        failed_count_1h=0,
-        failed_count_24h=0,
-        failure_rate_24h=0.0,
-        stuck_running_24h=0,
-        stuck_running_1h=0,
-        oldest_running_hours=None,
-        stuck_queued_1h=0,
-        queued_never_started=0,
-        invalid_queued_with_start=1,
-        invalid_running_without_start=1,
-        completed_1h=0,
-        completed_24h=0,
-        throughput_per_hour=0.0,
-        timestamp=datetime.now(timezone.utc).isoformat(),
-    )
-
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_invalid_executions_details",
-        return_value=mock_invalid_executions,
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_execution_diagnostics",
-        return_value=mock_diagnostics,
-    )
-
-    response = client.get("/admin/diagnostics/executions/invalid?limit=100&offset=0")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["total"] == 2  # Sum of both invalid state types
-    assert len(data["executions"]) == 2
-    # Verify both types of invalid states are returned
-    assert data["executions"][0]["execution_id"] in [
-        "exec-invalid-1",
-        "exec-invalid-2",
-    ]
-    assert data["executions"][1]["execution_id"] in [
-        "exec-invalid-1",
-        "exec-invalid-2",
-    ]
-
-
-def test_requeue_single_execution_with_add_graph_execution(
-    mocker: pytest_mock.MockFixture,
-    admin_user_id: str,
-):
-    """Test requeueing uses add_graph_execution in requeue mode"""
-    mock_exec_meta = GraphExecutionMeta(
-        id="exec-stuck-123",
-        user_id="user-123",
-        graph_id="graph-456",
-        graph_version=1,
-        inputs=None,
-        credential_inputs=None,
-        nodes_input_masks=None,
-        preset_id=None,
-        status=AgentExecutionStatus.QUEUED,
-        started_at=datetime.now(timezone.utc),
-        ended_at=datetime.now(timezone.utc),
-        stats=None,
-    )
-
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_graph_executions",
-        return_value=[mock_exec_meta],
-    )
-
-    mock_add_graph_execution = mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.add_graph_execution",
-        return_value=AsyncMock(),
-    )
-
-    response = client.post(
-        "/admin/diagnostics/executions/requeue",
-        json={"execution_id": "exec-stuck-123"},
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is True
-    assert data["requeued_count"] == 1
-
-    # Verify it used add_graph_execution in requeue mode
-    mock_add_graph_execution.assert_called_once()
-    call_kwargs = mock_add_graph_execution.call_args.kwargs
-    assert call_kwargs["graph_exec_id"] == "exec-stuck-123"  # Requeue mode!
-    assert call_kwargs["graph_id"] == "graph-456"
-    assert call_kwargs["user_id"] == "user-123"
-
-
-def test_stop_single_execution_with_stop_graph_execution(
-    mocker: pytest_mock.MockFixture,
-    admin_user_id: str,
-):
-    """Test stopping uses robust stop_graph_execution"""
-    mock_exec_meta = GraphExecutionMeta(
-        id="exec-running-123",
-        user_id="user-789",
-        graph_id="graph-999",
-        graph_version=2,
-        inputs=None,
-        credential_inputs=None,
-        nodes_input_masks=None,
-        preset_id=None,
-        status=AgentExecutionStatus.RUNNING,
-        started_at=datetime.now(timezone.utc),
-        ended_at=datetime.now(timezone.utc),
-        stats=None,
-    )
-
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_graph_executions",
-        return_value=[mock_exec_meta],
-    )
-
-    mock_stop_graph_execution = mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.stop_graph_execution",
-        return_value=AsyncMock(),
-    )
-
-    response = client.post(
-        "/admin/diagnostics/executions/stop",
-        json={"execution_id": "exec-running-123"},
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is True
-    assert data["stopped_count"] == 1
-
-    # Verify it used stop_graph_execution with cascade
-    mock_stop_graph_execution.assert_called_once()
-    call_kwargs = mock_stop_graph_execution.call_args.kwargs
-    assert call_kwargs["graph_exec_id"] == "exec-running-123"
-    assert call_kwargs["user_id"] == "user-789"
-    assert call_kwargs["cascade"] is True  # Stops children too!
-    assert call_kwargs["wait_timeout"] == 15.0
-
-
-def test_requeue_not_queued_execution_fails(
-    mocker: pytest_mock.MockFixture,
-):
-    """Test that requeue fails if execution is not in QUEUED status"""
-    # Mock an execution that's RUNNING (not QUEUED)
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_graph_executions",
-        return_value=[],  # No QUEUED executions found
-    )
-
-    response = client.post(
-        "/admin/diagnostics/executions/requeue",
-        json={"execution_id": "exec-running-123"},
-    )
-
-    assert response.status_code == 404
-    assert "not found or not in QUEUED status" in response.json()["detail"]
-
-
-def test_list_invalid_executions_no_bulk_actions(
-    mocker: pytest_mock.MockFixture,
-):
-    """Verify invalid executions endpoint is read-only (no bulk actions)"""
-    # This is a documentation test - the endpoint exists but should not
-    # have corresponding cleanup/stop/requeue endpoints
-
-    # These endpoints should NOT exist for invalid states:
-    invalid_bulk_endpoints = [
-        "/admin/diagnostics/executions/cleanup-invalid",
-        "/admin/diagnostics/executions/stop-invalid",
-        "/admin/diagnostics/executions/requeue-invalid",
-    ]
-
-    for endpoint in invalid_bulk_endpoints:
-        response = client.post(endpoint, json={"execution_ids": ["test"]})
-        assert response.status_code == 404, f"{endpoint} should not exist (read-only)"
-
-
-def test_execution_ids_filter_efficiency(
-    mocker: pytest_mock.MockFixture,
-):
-    """Test that bulk operations use efficient execution_ids filter"""
-    mock_exec_metas = [
-        GraphExecutionMeta(
-            id=f"exec-{i}",
-            user_id=f"user-{i}",
-            graph_id="graph-123",
-            graph_version=1,
-            inputs=None,
-            credential_inputs=None,
-            nodes_input_masks=None,
-            preset_id=None,
-            status=AgentExecutionStatus.QUEUED,
-            started_at=datetime.now(timezone.utc),
-            ended_at=datetime.now(timezone.utc),
-            stats=None,
-        )
-        for i in range(3)
-    ]
-
-    mock_get_graph_executions = mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_graph_executions",
-        return_value=mock_exec_metas,
-    )
-
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.add_graph_execution",
-        return_value=AsyncMock(),
-    )
-
-    response = client.post(
-        "/admin/diagnostics/executions/requeue-bulk",
-        json={"execution_ids": ["exec-0", "exec-1", "exec-2"]},
-    )
-
-    assert response.status_code == 200
-
-    # Verify it used execution_ids filter (not fetching all queued)
-    mock_get_graph_executions.assert_called_once()
-    call_kwargs = mock_get_graph_executions.call_args.kwargs
-    assert "execution_ids" in call_kwargs
-    assert call_kwargs["execution_ids"] == ["exec-0", "exec-1", "exec-2"]
-    assert call_kwargs["statuses"] == [AgentExecutionStatus.QUEUED]
-
-
-# ---------------------------------------------------------------------------
-# Helper: reusable mock diagnostics summary
-# ---------------------------------------------------------------------------
-
-
-def _make_mock_diagnostics(**overrides) -> ExecutionDiagnosticsSummary:
-    defaults = dict(
-        running_count=10,
-        queued_db_count=5,
-        rabbitmq_queue_depth=3,
-        cancel_queue_depth=0,
-        orphaned_running=2,
-        orphaned_queued=1,
-        failed_count_1h=5,
-        failed_count_24h=20,
-        failure_rate_24h=0.83,
-        stuck_running_24h=3,
-        stuck_running_1h=5,
-        oldest_running_hours=26.5,
-        stuck_queued_1h=2,
-        queued_never_started=1,
-        invalid_queued_with_start=1,
-        invalid_running_without_start=1,
-        completed_1h=50,
-        completed_24h=1200,
-        throughput_per_hour=50.0,
-        timestamp=datetime.now(timezone.utc).isoformat(),
-    )
-    defaults.update(overrides)
-    return ExecutionDiagnosticsSummary(**defaults)
-
-
-_SENTINEL = object()
-
-
-def _make_mock_execution(
-    exec_id: str = "exec-1",
-    status: str = "RUNNING",
-    started_at: datetime | None | object = _SENTINEL,
-) -> RunningExecutionDetail:
-    return RunningExecutionDetail(
-        execution_id=exec_id,
-        graph_id="graph-123",
-        graph_name="Test Graph",
-        graph_version=1,
-        user_id="user-123",
-        user_email="test@example.com",
-        status=status,
-        created_at=datetime.now(timezone.utc),
-        started_at=(
-            datetime.now(timezone.utc) if started_at is _SENTINEL else started_at
-        ),
-        queue_status=None,
-    )
-
-
-def _make_mock_failed_execution(
-    exec_id: str = "exec-fail-1",
-) -> FailedExecutionDetail:
-    return FailedExecutionDetail(
-        execution_id=exec_id,
-        graph_id="graph-123",
-        graph_name="Test Graph",
-        graph_version=1,
-        user_id="user-123",
-        user_email="test@example.com",
-        status="FAILED",
-        created_at=datetime.now(timezone.utc),
-        started_at=datetime.now(timezone.utc),
-        failed_at=datetime.now(timezone.utc),
-        error_message="Something went wrong",
-    )
-
-
-def _make_mock_schedule_health(**overrides) -> ScheduleHealthMetrics:
-    defaults = dict(
-        total_schedules=15,
-        user_schedules=10,
-        system_schedules=5,
-        orphaned_deleted_graph=2,
-        orphaned_no_library_access=1,
-        orphaned_invalid_credentials=0,
-        orphaned_validation_failed=0,
-        total_orphaned=3,
-        schedules_next_hour=4,
-        schedules_next_24h=8,
-        total_runs_next_hour=12,
-        total_runs_next_24h=48,
-        timestamp=datetime.now(timezone.utc).isoformat(),
-    )
-    defaults.update(overrides)
-    return ScheduleHealthMetrics(**defaults)
-
-
-# ---------------------------------------------------------------------------
-# GET endpoints: execution list variants
-# ---------------------------------------------------------------------------
-
-
-def test_list_running_executions(mocker: pytest_mock.MockFixture):
-    mock_execs = [
-        _make_mock_execution("exec-run-1"),
-        _make_mock_execution("exec-run-2"),
-    ]
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_running_executions_details",
-        return_value=mock_execs,
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_execution_diagnostics",
-        return_value=_make_mock_diagnostics(),
-    )
-
-    response = client.get("/admin/diagnostics/executions/running?limit=50&offset=0")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["total"] == 15  # running_count(10) + queued_db_count(5)
-    assert len(data["executions"]) == 2
-    assert data["executions"][0]["execution_id"] == "exec-run-1"
-
-
-def test_list_orphaned_executions(mocker: pytest_mock.MockFixture):
-    mock_execs = [_make_mock_execution("exec-orphan-1", status="RUNNING")]
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_orphaned_executions_details",
-        return_value=mock_execs,
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_execution_diagnostics",
-        return_value=_make_mock_diagnostics(),
-    )
-
-    response = client.get("/admin/diagnostics/executions/orphaned?limit=50&offset=0")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["total"] == 3  # orphaned_running(2) + orphaned_queued(1)
-    assert len(data["executions"]) == 1
-
-
-def test_list_failed_executions(mocker: pytest_mock.MockFixture):
-    mock_execs = [_make_mock_failed_execution("exec-fail-1")]
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_failed_executions_details",
-        return_value=mock_execs,
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_failed_executions_count",
-        return_value=42,
-    )
-
-    response = client.get(
-        "/admin/diagnostics/executions/failed?limit=50&offset=0&hours=24"
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["total"] == 42
-    assert len(data["executions"]) == 1
-    assert data["executions"][0]["error_message"] == "Something went wrong"
-
-
-def test_list_long_running_executions(mocker: pytest_mock.MockFixture):
-    mock_execs = [_make_mock_execution("exec-long-1")]
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_long_running_executions_details",
-        return_value=mock_execs,
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_execution_diagnostics",
-        return_value=_make_mock_diagnostics(),
-    )
-
-    response = client.get(
-        "/admin/diagnostics/executions/long-running?limit=50&offset=0"
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["total"] == 3  # stuck_running_24h
-    assert len(data["executions"]) == 1
-
-
-def test_list_stuck_queued_executions(mocker: pytest_mock.MockFixture):
-    mock_execs = [
-        _make_mock_execution("exec-stuck-1", status="QUEUED", started_at=None)
-    ]
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_stuck_queued_executions_details",
-        return_value=mock_execs,
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_execution_diagnostics",
-        return_value=_make_mock_diagnostics(),
-    )
-
-    response = client.get(
-        "/admin/diagnostics/executions/stuck-queued?limit=50&offset=0"
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["total"] == 2  # stuck_queued_1h
-    assert len(data["executions"]) == 1
-
-
-# ---------------------------------------------------------------------------
-# GET endpoints: agent + schedule diagnostics
-# ---------------------------------------------------------------------------
-
-
-def test_get_agent_diagnostics(mocker: pytest_mock.MockFixture):
-    mock_diag = AgentDiagnosticsSummary(
-        agents_with_active_executions=7,
-        timestamp=datetime.now(timezone.utc).isoformat(),
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_agent_diagnostics",
-        return_value=mock_diag,
-    )
-
-    response = client.get("/admin/diagnostics/agents")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["agents_with_active_executions"] == 7
-
-
-def test_get_schedule_diagnostics(mocker: pytest_mock.MockFixture):
-    mock_metrics = _make_mock_schedule_health()
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_schedule_health_metrics",
-        return_value=mock_metrics,
-    )
-
-    response = client.get("/admin/diagnostics/schedules")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["user_schedules"] == 10
-    assert data["total_orphaned"] == 3
-    assert data["total_runs_next_hour"] == 12
-
-
-def test_list_all_schedules(mocker: pytest_mock.MockFixture):
-    mock_schedules = [
-        ScheduleDetail(
-            schedule_id="sched-1",
-            schedule_name="Daily Run",
-            graph_id="graph-1",
-            graph_name="My Agent",
-            graph_version=1,
-            user_id="user-1",
-            user_email="alice@example.com",
-            cron="0 9 * * *",
-            timezone="UTC",
-            next_run_time=datetime.now(timezone.utc).isoformat(),
-        ),
-    ]
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_all_schedules_details",
-        return_value=mock_schedules,
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_schedule_health_metrics",
-        return_value=_make_mock_schedule_health(),
-    )
-
-    response = client.get("/admin/diagnostics/schedules/all?limit=50&offset=0")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["total"] == 10
-    assert len(data["schedules"]) == 1
-    assert data["schedules"][0]["schedule_name"] == "Daily Run"
-
-
-def test_list_orphaned_schedules(mocker: pytest_mock.MockFixture):
-    mock_orphans = [
-        OrphanedScheduleDetail(
-            schedule_id="sched-orphan-1",
-            schedule_name="Ghost Schedule",
-            graph_id="graph-deleted",
-            graph_version=1,
-            user_id="user-1",
-            orphan_reason="deleted_graph",
-            error_detail=None,
-            next_run_time=datetime.now(timezone.utc).isoformat(),
-        ),
-    ]
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_orphaned_schedules_details",
-        return_value=mock_orphans,
-    )
-
-    response = client.get("/admin/diagnostics/schedules/orphaned")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["total"] == 1
-    assert data["schedules"][0]["orphan_reason"] == "deleted_graph"
-
-
-# ---------------------------------------------------------------------------
-# POST endpoints: bulk stop, cleanup, requeue
-# ---------------------------------------------------------------------------
-
-
-def test_stop_multiple_executions(mocker: pytest_mock.MockFixture):
-    mock_exec_metas = [
-        GraphExecutionMeta(
-            id=f"exec-{i}",
-            user_id=f"user-{i}",
-            graph_id="graph-123",
-            graph_version=1,
-            inputs=None,
-            credential_inputs=None,
-            nodes_input_masks=None,
-            preset_id=None,
-            status=AgentExecutionStatus.RUNNING,
-            started_at=datetime.now(timezone.utc),
-            ended_at=None,
-            stats=None,
-        )
-        for i in range(2)
-    ]
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_graph_executions",
-        return_value=mock_exec_metas,
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.stop_graph_execution",
-        return_value=AsyncMock(),
-    )
-
-    response = client.post(
-        "/admin/diagnostics/executions/stop-bulk",
-        json={"execution_ids": ["exec-0", "exec-1"]},
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is True
-    assert data["stopped_count"] == 2
-
-
-def test_stop_multiple_executions_none_found(mocker: pytest_mock.MockFixture):
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_graph_executions",
-        return_value=[],
-    )
-
-    response = client.post(
-        "/admin/diagnostics/executions/stop-bulk",
-        json={"execution_ids": ["nonexistent"]},
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is False
-    assert data["stopped_count"] == 0
-
-
-def test_cleanup_orphaned_executions(mocker: pytest_mock.MockFixture):
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.cleanup_orphaned_executions_bulk",
-        return_value=3,
-    )
-
-    response = client.post(
-        "/admin/diagnostics/executions/cleanup-orphaned",
-        json={"execution_ids": ["exec-1", "exec-2", "exec-3"]},
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is True
-    assert data["stopped_count"] == 3
-
-
-def test_cleanup_orphaned_schedules(mocker: pytest_mock.MockFixture):
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.cleanup_orphaned_schedules_bulk",
-        return_value=2,
-    )
-
-    response = client.post(
-        "/admin/diagnostics/schedules/cleanup-orphaned",
-        json={"schedule_ids": ["sched-1", "sched-2"]},
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is True
-    assert data["deleted_count"] == 2
-
-
-def test_stop_all_long_running_executions(mocker: pytest_mock.MockFixture):
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.stop_all_long_running_executions",
-        return_value=5,
-    )
-
-    response = client.post("/admin/diagnostics/executions/stop-all-long-running")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is True
-    assert data["stopped_count"] == 5
-
-
-def test_cleanup_all_orphaned_executions(mocker: pytest_mock.MockFixture):
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_all_orphaned_execution_ids",
-        return_value=["exec-1", "exec-2"],
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.cleanup_orphaned_executions_bulk",
-        return_value=2,
-    )
-
-    response = client.post("/admin/diagnostics/executions/cleanup-all-orphaned")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is True
-    assert data["stopped_count"] == 2
-
-
-def test_cleanup_all_orphaned_executions_none(mocker: pytest_mock.MockFixture):
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_all_orphaned_execution_ids",
-        return_value=[],
-    )
-
-    response = client.post("/admin/diagnostics/executions/cleanup-all-orphaned")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is True
-    assert data["stopped_count"] == 0
-    assert "No orphaned" in data["message"]
-
-
-def test_cleanup_all_stuck_queued_executions(mocker: pytest_mock.MockFixture):
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.cleanup_all_stuck_queued_executions",
-        return_value=4,
-    )
-
-    response = client.post("/admin/diagnostics/executions/cleanup-all-stuck-queued")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is True
-    assert data["stopped_count"] == 4
-
-
-def test_requeue_all_stuck_executions(mocker: pytest_mock.MockFixture):
-    mock_exec_metas = [
-        GraphExecutionMeta(
-            id=f"exec-stuck-{i}",
-            user_id=f"user-{i}",
-            graph_id="graph-123",
-            graph_version=1,
-            inputs=None,
-            credential_inputs=None,
-            nodes_input_masks=None,
-            preset_id=None,
-            status=AgentExecutionStatus.QUEUED,
-            started_at=None,
-            ended_at=None,
-            stats=None,
-        )
-        for i in range(3)
-    ]
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_all_stuck_queued_execution_ids",
-        return_value=["exec-stuck-0", "exec-stuck-1", "exec-stuck-2"],
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_graph_executions",
-        return_value=mock_exec_metas,
-    )
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.add_graph_execution",
-        return_value=AsyncMock(),
-    )
-
-    response = client.post("/admin/diagnostics/executions/requeue-all-stuck")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is True
-    assert data["requeued_count"] == 3
-
-
-def test_requeue_all_stuck_executions_none(mocker: pytest_mock.MockFixture):
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_all_stuck_queued_execution_ids",
-        return_value=[],
-    )
-
-    response = client.post("/admin/diagnostics/executions/requeue-all-stuck")
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is True
-    assert data["requeued_count"] == 0
-    assert "No stuck" in data["message"]
-
-
-def test_requeue_bulk_none_found(mocker: pytest_mock.MockFixture):
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_graph_executions",
-        return_value=[],
-    )
-
-    response = client.post(
-        "/admin/diagnostics/executions/requeue-bulk",
-        json={"execution_ids": ["nonexistent"]},
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["success"] is False
-    assert data["requeued_count"] == 0
-
-
-def test_stop_single_execution_not_found(mocker: pytest_mock.MockFixture):
-    mocker.patch(
-        "backend.api.features.admin.diagnostics_admin_routes.get_graph_executions",
-        return_value=[],
-    )
-
-    response = client.post(
-        "/admin/diagnostics/executions/stop",
-        json={"execution_id": "nonexistent"},
-    )
-
-    assert response.status_code == 404
-    assert "not found" in response.json()["detail"]

autogpt_platform/backend/backend/api/features/admin/model.py

@@ -14,70 +14,3 @@ class UserHistoryResponse(BaseModel):
 class AddUserCreditsResponse(BaseModel):
     new_balance: int
     transaction_key: str
-
-
-class ExecutionDiagnosticsResponse(BaseModel):
-    """Response model for execution diagnostics"""
-
-    # Current execution state
-    running_executions: int
-    queued_executions_db: int
-    queued_executions_rabbitmq: int
-    cancel_queue_depth: int
-
-    # Orphaned execution detection
-    orphaned_running: int
-    orphaned_queued: int
-
-    # Failure metrics
-    failed_count_1h: int
-    failed_count_24h: int
-    failure_rate_24h: float
-
-    # Long-running detection
-    stuck_running_24h: int
-    stuck_running_1h: int
-    oldest_running_hours: float | None
-
-    # Stuck queued detection
-    stuck_queued_1h: int
-    queued_never_started: int
-
-    # Invalid state detection (data corruption - no auto-actions)
-    invalid_queued_with_start: int
-    invalid_running_without_start: int
-
-    # Throughput metrics
-    completed_1h: int
-    completed_24h: int
-    throughput_per_hour: float
-
-    timestamp: str
-
-
-class AgentDiagnosticsResponse(BaseModel):
-    """Response model for agent diagnostics"""
-
-    agents_with_active_executions: int
-    timestamp: str
-
-
-class ScheduleHealthMetrics(BaseModel):
-    """Response model for schedule diagnostics"""
-
-    total_schedules: int
-    user_schedules: int
-    system_schedules: int
-
-    # Orphan detection
-    orphaned_deleted_graph: int
-    orphaned_no_library_access: int
-    orphaned_invalid_credentials: int
-    orphaned_validation_failed: int
-    total_orphaned: int
-
-    # Upcoming
-    schedules_next_hour: int
-    schedules_next_24h: int
-
-    timestamp: str

autogpt_platform/backend/backend/api/features/admin/platform_cost_routes.py

@@ -1,141 +0,0 @@
-import logging
-from datetime import datetime
-
-from autogpt_libs.auth import get_user_id, requires_admin_user
-from fastapi import APIRouter, Query, Security
-from pydantic import BaseModel
-
-from backend.data.platform_cost import (
-    CostLogRow,
-    PlatformCostDashboard,
-    get_platform_cost_dashboard,
-    get_platform_cost_logs,
-    get_platform_cost_logs_for_export,
-)
-from backend.util.models import Pagination
-
-logger = logging.getLogger(__name__)
-
-
-router = APIRouter(
-    prefix="/platform-costs",
-    tags=["platform-cost", "admin"],
-    dependencies=[Security(requires_admin_user)],
-)
-
-
-class PlatformCostLogsResponse(BaseModel):
-    logs: list[CostLogRow]
-    pagination: Pagination
-
-
-@router.get(
-    "/dashboard",
-    response_model=PlatformCostDashboard,
-    summary="Get Platform Cost Dashboard",
-)
-async def get_cost_dashboard(
-    admin_user_id: str = Security(get_user_id),
-    start: datetime | None = Query(None),
-    end: datetime | None = Query(None),
-    provider: str | None = Query(None),
-    user_id: str | None = Query(None),
-    model: str | None = Query(None),
-    block_name: str | None = Query(None),
-    tracking_type: str | None = Query(None),
-    graph_exec_id: str | None = Query(None),
-):
-    logger.info("Admin %s fetching platform cost dashboard", admin_user_id)
-    return await get_platform_cost_dashboard(
-        start=start,
-        end=end,
-        provider=provider,
-        user_id=user_id,
-        model=model,
-        block_name=block_name,
-        tracking_type=tracking_type,
-        graph_exec_id=graph_exec_id,
-    )
-
-
-@router.get(
-    "/logs",
-    response_model=PlatformCostLogsResponse,
-    summary="Get Platform Cost Logs",
-)
-async def get_cost_logs(
-    admin_user_id: str = Security(get_user_id),
-    start: datetime | None = Query(None),
-    end: datetime | None = Query(None),
-    provider: str | None = Query(None),
-    user_id: str | None = Query(None),
-    page: int = Query(1, ge=1),
-    page_size: int = Query(50, ge=1, le=200),
-    model: str | None = Query(None),
-    block_name: str | None = Query(None),
-    tracking_type: str | None = Query(None),
-    graph_exec_id: str | None = Query(None),
-):
-    logger.info("Admin %s fetching platform cost logs", admin_user_id)
-    logs, total = await get_platform_cost_logs(
-        start=start,
-        end=end,
-        provider=provider,
-        user_id=user_id,
-        page=page,
-        page_size=page_size,
-        model=model,
-        block_name=block_name,
-        tracking_type=tracking_type,
-        graph_exec_id=graph_exec_id,
-    )
-    total_pages = (total + page_size - 1) // page_size
-    return PlatformCostLogsResponse(
-        logs=logs,
-        pagination=Pagination(
-            total_items=total,
-            total_pages=total_pages,
-            current_page=page,
-            page_size=page_size,
-        ),
-    )
-
-
-class PlatformCostExportResponse(BaseModel):
-    logs: list[CostLogRow]
-    total_rows: int
-    truncated: bool
-
-
-@router.get(
-    "/logs/export",
-    response_model=PlatformCostExportResponse,
-    summary="Export Platform Cost Logs",
-)
-async def export_cost_logs(
-    admin_user_id: str = Security(get_user_id),
-    start: datetime | None = Query(None),
-    end: datetime | None = Query(None),
-    provider: str | None = Query(None),
-    user_id: str | None = Query(None),
-    model: str | None = Query(None),
-    block_name: str | None = Query(None),
-    tracking_type: str | None = Query(None),
-    graph_exec_id: str | None = Query(None),
-):
-    logger.info("Admin %s exporting platform cost logs", admin_user_id)
-    logs, truncated = await get_platform_cost_logs_for_export(
-        start=start,
-        end=end,
-        provider=provider,
-        user_id=user_id,
-        model=model,
-        block_name=block_name,
-        tracking_type=tracking_type,
-        graph_exec_id=graph_exec_id,
-    )
-    return PlatformCostExportResponse(
-        logs=logs,
-        total_rows=len(logs),
-        truncated=truncated,
-    )

autogpt_platform/backend/backend/api/features/admin/platform_cost_routes_test.py

@@ -1,291 +0,0 @@
-from datetime import datetime, timezone
-from unittest.mock import AsyncMock
-
-import fastapi
-import fastapi.testclient
-import pytest
-import pytest_mock
-from autogpt_libs.auth.jwt_utils import get_jwt_payload
-
-from backend.data.platform_cost import CostLogRow, PlatformCostDashboard
-
-from .platform_cost_routes import router as platform_cost_router
-
-app = fastapi.FastAPI()
-app.include_router(platform_cost_router)
-
-client = fastapi.testclient.TestClient(app)
-
-
-@pytest.fixture(autouse=True)
-def setup_app_admin_auth(mock_jwt_admin):
-    """Setup admin auth overrides for all tests in this module"""
-    app.dependency_overrides[get_jwt_payload] = mock_jwt_admin["get_jwt_payload"]
-    yield
-    app.dependency_overrides.clear()
-
-
-def test_get_dashboard_success(
-    mocker: pytest_mock.MockerFixture,
-) -> None:
-    real_dashboard = PlatformCostDashboard(
-        by_provider=[],
-        by_user=[],
-        total_cost_microdollars=0,
-        total_requests=0,
-        total_users=0,
-    )
-    mocker.patch(
-        "backend.api.features.admin.platform_cost_routes.get_platform_cost_dashboard",
-        AsyncMock(return_value=real_dashboard),
-    )
-
-    response = client.get("/platform-costs/dashboard")
-    assert response.status_code == 200
-    data = response.json()
-    assert "by_provider" in data
-    assert "by_user" in data
-    assert data["total_cost_microdollars"] == 0
-
-
-def test_get_logs_success(
-    mocker: pytest_mock.MockerFixture,
-) -> None:
-    mocker.patch(
-        "backend.api.features.admin.platform_cost_routes.get_platform_cost_logs",
-        AsyncMock(return_value=([], 0)),
-    )
-
-    response = client.get("/platform-costs/logs")
-    assert response.status_code == 200
-    data = response.json()
-    assert data["logs"] == []
-    assert data["pagination"]["total_items"] == 0
-
-
-def test_get_dashboard_with_filters(
-    mocker: pytest_mock.MockerFixture,
-) -> None:
-    real_dashboard = PlatformCostDashboard(
-        by_provider=[],
-        by_user=[],
-        total_cost_microdollars=0,
-        total_requests=0,
-        total_users=0,
-    )
-    mock_dashboard = AsyncMock(return_value=real_dashboard)
-    mocker.patch(
-        "backend.api.features.admin.platform_cost_routes.get_platform_cost_dashboard",
-        mock_dashboard,
-    )
-
-    response = client.get(
-        "/platform-costs/dashboard",
-        params={
-            "start": "2026-01-01T00:00:00",
-            "end": "2026-04-01T00:00:00",
-            "provider": "openai",
-            "user_id": "test-user-123",
-        },
-    )
-    assert response.status_code == 200
-    mock_dashboard.assert_called_once()
-    call_kwargs = mock_dashboard.call_args.kwargs
-    assert call_kwargs["provider"] == "openai"
-    assert call_kwargs["user_id"] == "test-user-123"
-    assert call_kwargs["start"] is not None
-    assert call_kwargs["end"] is not None
-
-
-def test_get_logs_with_pagination(
-    mocker: pytest_mock.MockerFixture,
-) -> None:
-    mocker.patch(
-        "backend.api.features.admin.platform_cost_routes.get_platform_cost_logs",
-        AsyncMock(return_value=([], 0)),
-    )
-
-    response = client.get(
-        "/platform-costs/logs",
-        params={"page": 2, "page_size": 25, "provider": "anthropic"},
-    )
-    assert response.status_code == 200
-    data = response.json()
-    assert data["pagination"]["current_page"] == 2
-    assert data["pagination"]["page_size"] == 25
-
-
-def test_get_dashboard_requires_admin() -> None:
-    import fastapi
-    from fastapi import HTTPException
-
-    def reject_jwt(request: fastapi.Request):
-        raise HTTPException(status_code=401, detail="Not authenticated")
-
-    app.dependency_overrides[get_jwt_payload] = reject_jwt
-    try:
-        response = client.get("/platform-costs/dashboard")
-        assert response.status_code == 401
-        response = client.get("/platform-costs/logs")
-        assert response.status_code == 401
-    finally:
-        app.dependency_overrides.clear()
-
-
-def test_get_dashboard_rejects_non_admin(mock_jwt_user, mock_jwt_admin) -> None:
-    """Non-admin JWT must be rejected with 403 by requires_admin_user."""
-    app.dependency_overrides[get_jwt_payload] = mock_jwt_user["get_jwt_payload"]
-    try:
-        response = client.get("/platform-costs/dashboard")
-        assert response.status_code == 403
-        response = client.get("/platform-costs/logs")
-        assert response.status_code == 403
-    finally:
-        app.dependency_overrides[get_jwt_payload] = mock_jwt_admin["get_jwt_payload"]
-
-
-def test_get_logs_invalid_page_size_too_large() -> None:
-    """page_size > 200 must be rejected with 422."""
-    response = client.get("/platform-costs/logs", params={"page_size": 201})
-    assert response.status_code == 422
-
-
-def test_get_logs_invalid_page_size_zero() -> None:
-    """page_size = 0 (below ge=1) must be rejected with 422."""
-    response = client.get("/platform-costs/logs", params={"page_size": 0})
-    assert response.status_code == 422
-
-
-def test_get_logs_invalid_page_negative() -> None:
-    """page < 1 must be rejected with 422."""
-    response = client.get("/platform-costs/logs", params={"page": 0})
-    assert response.status_code == 422
-
-
-def test_get_dashboard_invalid_date_format() -> None:
-    """Malformed start date must be rejected with 422."""
-    response = client.get("/platform-costs/dashboard", params={"start": "not-a-date"})
-    assert response.status_code == 422
-
-
-def test_get_dashboard_repeated_requests(
-    mocker: pytest_mock.MockerFixture,
-) -> None:
-    """Repeated requests to the dashboard route both return 200."""
-    real_dashboard = PlatformCostDashboard(
-        by_provider=[],
-        by_user=[],
-        total_cost_microdollars=42,
-        total_requests=1,
-        total_users=1,
-    )
-    mocker.patch(
-        "backend.api.features.admin.platform_cost_routes.get_platform_cost_dashboard",
-        AsyncMock(return_value=real_dashboard),
-    )
-
-    r1 = client.get("/platform-costs/dashboard")
-    r2 = client.get("/platform-costs/dashboard")
-
-    assert r1.status_code == 200
-    assert r2.status_code == 200
-    assert r1.json()["total_cost_microdollars"] == 42
-    assert r2.json()["total_cost_microdollars"] == 42
-
-
-def _make_cost_log_row() -> CostLogRow:
-    return CostLogRow(
-        id="log-1",
-        created_at=datetime(2026, 1, 1, tzinfo=timezone.utc),
-        user_id="user-1",
-        email="u***@example.com",
-        graph_exec_id="graph-1",
-        node_exec_id="node-1",
-        block_name="LlmCallBlock",
-        provider="anthropic",
-        tracking_type="token",
-        cost_microdollars=500,
-        input_tokens=100,
-        output_tokens=50,
-        cache_read_tokens=10,
-        cache_creation_tokens=5,
-        duration=1.5,
-        model="claude-3-5-sonnet-20241022",
-    )
-
-
-def test_export_logs_success(
-    mocker: pytest_mock.MockerFixture,
-) -> None:
-    row = _make_cost_log_row()
-    mocker.patch(
-        "backend.api.features.admin.platform_cost_routes.get_platform_cost_logs_for_export",
-        AsyncMock(return_value=([row], False)),
-    )
-
-    response = client.get("/platform-costs/logs/export")
-    assert response.status_code == 200
-    data = response.json()
-    assert data["total_rows"] == 1
-    assert data["truncated"] is False
-    assert len(data["logs"]) == 1
-    assert data["logs"][0]["cache_read_tokens"] == 10
-    assert data["logs"][0]["cache_creation_tokens"] == 5
-
-
-def test_export_logs_truncated(
-    mocker: pytest_mock.MockerFixture,
-) -> None:
-    rows = [_make_cost_log_row() for _ in range(3)]
-    mocker.patch(
-        "backend.api.features.admin.platform_cost_routes.get_platform_cost_logs_for_export",
-        AsyncMock(return_value=(rows, True)),
-    )
-
-    response = client.get("/platform-costs/logs/export")
-    assert response.status_code == 200
-    data = response.json()
-    assert data["total_rows"] == 3
-    assert data["truncated"] is True
-
-
-def test_export_logs_with_filters(
-    mocker: pytest_mock.MockerFixture,
-) -> None:
-    mock_export = AsyncMock(return_value=([], False))
-    mocker.patch(
-        "backend.api.features.admin.platform_cost_routes.get_platform_cost_logs_for_export",
-        mock_export,
-    )
-
-    response = client.get(
-        "/platform-costs/logs/export",
-        params={
-            "provider": "anthropic",
-            "model": "claude-3-5-sonnet-20241022",
-            "block_name": "LlmCallBlock",
-            "tracking_type": "token",
-        },
-    )
-    assert response.status_code == 200
-    mock_export.assert_called_once()
-    call_kwargs = mock_export.call_args.kwargs
-    assert call_kwargs["provider"] == "anthropic"
-    assert call_kwargs["model"] == "claude-3-5-sonnet-20241022"
-    assert call_kwargs["block_name"] == "LlmCallBlock"
-    assert call_kwargs["tracking_type"] == "token"
-
-
-def test_export_logs_requires_admin() -> None:
-    import fastapi
-    from fastapi import HTTPException
-
-    def reject_jwt(request: fastapi.Request):
-        raise HTTPException(status_code=401, detail="Not authenticated")
-
-    app.dependency_overrides[get_jwt_payload] = reject_jwt
-    try:
-        response = client.get("/platform-costs/logs/export")
-        assert response.status_code == 401
-    finally:
-        app.dependency_overrides.clear()

autogpt_platform/backend/backend/api/features/admin/rate_limit_admin_routes.py

@@ -1,263 +0,0 @@
-"""Admin endpoints for checking and resetting user CoPilot rate limit usage."""
-
-import logging
-from typing import Optional
-
-from autogpt_libs.auth import get_user_id, requires_admin_user
-from fastapi import APIRouter, Body, HTTPException, Security
-from pydantic import BaseModel
-
-from backend.copilot.config import ChatConfig
-from backend.copilot.rate_limit import (
-    SubscriptionTier,
-    get_global_rate_limits,
-    get_usage_status,
-    get_user_tier,
-    reset_user_usage,
-    set_user_tier,
-)
-from backend.data.user import get_user_by_email, get_user_email_by_id, search_users
-
-logger = logging.getLogger(__name__)
-
-config = ChatConfig()
-
-router = APIRouter(
-    prefix="/admin",
-    tags=["copilot", "admin"],
-    dependencies=[Security(requires_admin_user)],
-)
-
-
-class UserRateLimitResponse(BaseModel):
-    user_id: str
-    user_email: Optional[str] = None
-    daily_cost_limit_microdollars: int
-    weekly_cost_limit_microdollars: int
-    daily_cost_used_microdollars: int
-    weekly_cost_used_microdollars: int
-    tier: SubscriptionTier
-
-
-class UserTierResponse(BaseModel):
-    user_id: str
-    tier: SubscriptionTier
-
-
-class SetUserTierRequest(BaseModel):
-    user_id: str
-    tier: SubscriptionTier
-
-
-async def _resolve_user_id(
-    user_id: Optional[str], email: Optional[str]
-) -> tuple[str, Optional[str]]:
-    """Resolve a user_id and email from the provided parameters.
-
-    Returns (user_id, email). Accepts either user_id or email; at least one
-    must be provided.  When both are provided, ``email`` takes precedence.
-    """
-    if email:
-        user = await get_user_by_email(email)
-        if not user:
-            raise HTTPException(
-                status_code=404, detail="No user found with the provided email."
-            )
-        return user.id, email
-
-    if not user_id:
-        raise HTTPException(
-            status_code=400,
-            detail="Either user_id or email query parameter is required.",
-        )
-
-    # We have a user_id; try to look up their email for display purposes.
-    # This is non-critical -- a failure should not block the response.
-    try:
-        resolved_email = await get_user_email_by_id(user_id)
-    except Exception:
-        logger.warning("Failed to resolve email for user %s", user_id, exc_info=True)
-        resolved_email = None
-    return user_id, resolved_email
-
-
-@router.get(
-    "/rate_limit",
-    response_model=UserRateLimitResponse,
-    summary="Get User Rate Limit",
-)
-async def get_user_rate_limit(
-    user_id: Optional[str] = None,
-    email: Optional[str] = None,
-    admin_user_id: str = Security(get_user_id),
-) -> UserRateLimitResponse:
-    """Get a user's current usage and effective rate limits. Admin-only.
-
-    Accepts either ``user_id`` or ``email`` as a query parameter.
-    When ``email`` is provided the user is looked up by email first.
-    """
-    resolved_id, resolved_email = await _resolve_user_id(user_id, email)
-
-    logger.info("Admin %s checking rate limit for user %s", admin_user_id, resolved_id)
-
-    daily_limit, weekly_limit, tier = await get_global_rate_limits(
-        resolved_id,
-        config.daily_cost_limit_microdollars,
-        config.weekly_cost_limit_microdollars,
-    )
-    usage = await get_usage_status(resolved_id, daily_limit, weekly_limit, tier=tier)
-
-    return UserRateLimitResponse(
-        user_id=resolved_id,
-        user_email=resolved_email,
-        daily_cost_limit_microdollars=daily_limit,
-        weekly_cost_limit_microdollars=weekly_limit,
-        daily_cost_used_microdollars=usage.daily.used,
-        weekly_cost_used_microdollars=usage.weekly.used,
-        tier=tier,
-    )
-
-
-@router.post(
-    "/rate_limit/reset",
-    response_model=UserRateLimitResponse,
-    summary="Reset User Rate Limit Usage",
-)
-async def reset_user_rate_limit(
-    user_id: str = Body(embed=True),
-    reset_weekly: bool = Body(False, embed=True),
-    admin_user_id: str = Security(get_user_id),
-) -> UserRateLimitResponse:
-    """Reset a user's daily usage counter (and optionally weekly). Admin-only."""
-    logger.info(
-        "Admin %s resetting rate limit for user %s (reset_weekly=%s)",
-        admin_user_id,
-        user_id,
-        reset_weekly,
-    )
-
-    try:
-        await reset_user_usage(user_id, reset_weekly=reset_weekly)
-    except Exception as e:
-        logger.exception("Failed to reset user usage")
-        raise HTTPException(status_code=500, detail="Failed to reset usage") from e
-
-    daily_limit, weekly_limit, tier = await get_global_rate_limits(
-        user_id,
-        config.daily_cost_limit_microdollars,
-        config.weekly_cost_limit_microdollars,
-    )
-    usage = await get_usage_status(user_id, daily_limit, weekly_limit, tier=tier)
-
-    try:
-        resolved_email = await get_user_email_by_id(user_id)
-    except Exception:
-        logger.warning("Failed to resolve email for user %s", user_id, exc_info=True)
-        resolved_email = None
-
-    return UserRateLimitResponse(
-        user_id=user_id,
-        user_email=resolved_email,
-        daily_cost_limit_microdollars=daily_limit,
-        weekly_cost_limit_microdollars=weekly_limit,
-        daily_cost_used_microdollars=usage.daily.used,
-        weekly_cost_used_microdollars=usage.weekly.used,
-        tier=tier,
-    )
-
-
-@router.get(
-    "/rate_limit/tier",
-    response_model=UserTierResponse,
-    summary="Get User Rate Limit Tier",
-)
-async def get_user_rate_limit_tier(
-    user_id: str,
-    admin_user_id: str = Security(get_user_id),
-) -> UserTierResponse:
-    """Get a user's current rate-limit tier. Admin-only.
-
-    Returns 404 if the user does not exist in the database.
-    """
-    logger.info("Admin %s checking tier for user %s", admin_user_id, user_id)
-
-    resolved_email = await get_user_email_by_id(user_id)
-    if resolved_email is None:
-        raise HTTPException(status_code=404, detail=f"User {user_id} not found")
-
-    tier = await get_user_tier(user_id)
-    return UserTierResponse(user_id=user_id, tier=tier)
-
-
-@router.post(
-    "/rate_limit/tier",
-    response_model=UserTierResponse,
-    summary="Set User Rate Limit Tier",
-)
-async def set_user_rate_limit_tier(
-    request: SetUserTierRequest,
-    admin_user_id: str = Security(get_user_id),
-) -> UserTierResponse:
-    """Set a user's rate-limit tier. Admin-only.
-
-    Returns 404 if the user does not exist in the database.
-    """
-    try:
-        resolved_email = await get_user_email_by_id(request.user_id)
-    except Exception:
-        logger.warning(
-            "Failed to resolve email for user %s",
-            request.user_id,
-            exc_info=True,
-        )
-        resolved_email = None
-
-    if resolved_email is None:
-        raise HTTPException(status_code=404, detail=f"User {request.user_id} not found")
-
-    old_tier = await get_user_tier(request.user_id)
-    logger.info(
-        "Admin %s changing tier for user %s (%s): %s -> %s",
-        admin_user_id,
-        request.user_id,
-        resolved_email,
-        old_tier.value,
-        request.tier.value,
-    )
-    try:
-        await set_user_tier(request.user_id, request.tier)
-    except Exception as e:
-        logger.exception("Failed to set user tier")
-        raise HTTPException(status_code=500, detail="Failed to set tier") from e
-
-    return UserTierResponse(user_id=request.user_id, tier=request.tier)
-
-
-class UserSearchResult(BaseModel):
-    user_id: str
-    user_email: Optional[str] = None
-
-
-@router.get(
-    "/rate_limit/search_users",
-    response_model=list[UserSearchResult],
-    summary="Search Users by Name or Email",
-)
-async def admin_search_users(
-    query: str,
-    limit: int = 20,
-    admin_user_id: str = Security(get_user_id),
-) -> list[UserSearchResult]:
-    """Search users by partial email or name. Admin-only.
-
-    Queries the User table directly — returns results even for users
-    without credit transaction history.
-    """
-    if len(query.strip()) < 3:
-        raise HTTPException(
-            status_code=400,
-            detail="Search query must be at least 3 characters.",
-        )
-    logger.info("Admin %s searching users with query=%r", admin_user_id, query)
-    results = await search_users(query, limit=max(1, min(limit, 50)))
-    return [UserSearchResult(user_id=uid, user_email=email) for uid, email in results]

autogpt_platform/backend/backend/api/features/admin/rate_limit_admin_routes_test.py

@@ -1,566 +0,0 @@
-import json
-from types import SimpleNamespace
-from unittest.mock import AsyncMock
-
-import fastapi
-import fastapi.testclient
-import pytest
-import pytest_mock
-from autogpt_libs.auth.jwt_utils import get_jwt_payload
-from pytest_snapshot.plugin import Snapshot
-
-from backend.copilot.rate_limit import CoPilotUsageStatus, SubscriptionTier, UsageWindow
-
-from .rate_limit_admin_routes import router as rate_limit_admin_router
-
-app = fastapi.FastAPI()
-app.include_router(rate_limit_admin_router)
-
-client = fastapi.testclient.TestClient(app)
-
-_MOCK_MODULE = "backend.api.features.admin.rate_limit_admin_routes"
-
-_TARGET_EMAIL = "target@example.com"
-
-
-@pytest.fixture(autouse=True)
-def setup_app_admin_auth(mock_jwt_admin):
-    """Setup admin auth overrides for all tests in this module"""
-    app.dependency_overrides[get_jwt_payload] = mock_jwt_admin["get_jwt_payload"]
-    yield
-    app.dependency_overrides.clear()
-
-
-def _mock_usage_status(
-    daily_used: int = 500_000, weekly_used: int = 3_000_000
-) -> CoPilotUsageStatus:
-    from datetime import UTC, datetime, timedelta
-
-    now = datetime.now(UTC)
-    return CoPilotUsageStatus(
-        daily=UsageWindow(
-            used=daily_used, limit=2_500_000, resets_at=now + timedelta(hours=6)
-        ),
-        weekly=UsageWindow(
-            used=weekly_used, limit=12_500_000, resets_at=now + timedelta(days=3)
-        ),
-    )
-
-
-def _patch_rate_limit_deps(
-    mocker: pytest_mock.MockerFixture,
-    target_user_id: str,
-    daily_used: int = 500_000,
-    weekly_used: int = 3_000_000,
-):
-    """Patch the common rate-limit + user-lookup dependencies."""
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_global_rate_limits",
-        new_callable=AsyncMock,
-        return_value=(2_500_000, 12_500_000, SubscriptionTier.FREE),
-    )
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_usage_status",
-        new_callable=AsyncMock,
-        return_value=_mock_usage_status(daily_used=daily_used, weekly_used=weekly_used),
-    )
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_email_by_id",
-        new_callable=AsyncMock,
-        return_value=_TARGET_EMAIL,
-    )
-
-
-def test_get_rate_limit(
-    mocker: pytest_mock.MockerFixture,
-    configured_snapshot: Snapshot,
-    target_user_id: str,
-) -> None:
-    """Test getting rate limit and usage for a user."""
-    _patch_rate_limit_deps(mocker, target_user_id)
-
-    response = client.get("/admin/rate_limit", params={"user_id": target_user_id})
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["user_id"] == target_user_id
-    assert data["user_email"] == _TARGET_EMAIL
-    assert data["daily_cost_limit_microdollars"] == 2_500_000
-    assert data["weekly_cost_limit_microdollars"] == 12_500_000
-    assert data["daily_cost_used_microdollars"] == 500_000
-    assert data["weekly_cost_used_microdollars"] == 3_000_000
-    assert data["tier"] == "FREE"
-
-    configured_snapshot.assert_match(
-        json.dumps(data, indent=2, sort_keys=True) + "\n",
-        "get_rate_limit",
-    )
-
-
-def test_get_rate_limit_by_email(
-    mocker: pytest_mock.MockerFixture,
-    target_user_id: str,
-) -> None:
-    """Test looking up rate limits via email instead of user_id."""
-    _patch_rate_limit_deps(mocker, target_user_id)
-
-    mock_user = SimpleNamespace(id=target_user_id, email=_TARGET_EMAIL)
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_by_email",
-        new_callable=AsyncMock,
-        return_value=mock_user,
-    )
-
-    response = client.get("/admin/rate_limit", params={"email": _TARGET_EMAIL})
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["user_id"] == target_user_id
-    assert data["user_email"] == _TARGET_EMAIL
-    assert data["daily_cost_limit_microdollars"] == 2_500_000
-
-
-def test_get_rate_limit_by_email_not_found(
-    mocker: pytest_mock.MockerFixture,
-) -> None:
-    """Test that looking up a non-existent email returns 404."""
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_by_email",
-        new_callable=AsyncMock,
-        return_value=None,
-    )
-
-    response = client.get("/admin/rate_limit", params={"email": "nobody@example.com"})
-
-    assert response.status_code == 404
-
-
-def test_get_rate_limit_no_params() -> None:
-    """Test that omitting both user_id and email returns 400."""
-    response = client.get("/admin/rate_limit")
-    assert response.status_code == 400
-
-
-def test_reset_user_usage_daily_only(
-    mocker: pytest_mock.MockerFixture,
-    configured_snapshot: Snapshot,
-    target_user_id: str,
-) -> None:
-    """Test resetting only daily usage (default behaviour)."""
-    mock_reset = mocker.patch(
-        f"{_MOCK_MODULE}.reset_user_usage",
-        new_callable=AsyncMock,
-    )
-    _patch_rate_limit_deps(mocker, target_user_id, daily_used=0, weekly_used=3_000_000)
-
-    response = client.post(
-        "/admin/rate_limit/reset",
-        json={"user_id": target_user_id},
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["daily_cost_used_microdollars"] == 0
-    # Weekly is untouched
-    assert data["weekly_cost_used_microdollars"] == 3_000_000
-    assert data["tier"] == "FREE"
-
-    mock_reset.assert_awaited_once_with(target_user_id, reset_weekly=False)
-
-    configured_snapshot.assert_match(
-        json.dumps(data, indent=2, sort_keys=True) + "\n",
-        "reset_user_usage_daily_only",
-    )
-
-
-def test_reset_user_usage_daily_and_weekly(
-    mocker: pytest_mock.MockerFixture,
-    configured_snapshot: Snapshot,
-    target_user_id: str,
-) -> None:
-    """Test resetting both daily and weekly usage."""
-    mock_reset = mocker.patch(
-        f"{_MOCK_MODULE}.reset_user_usage",
-        new_callable=AsyncMock,
-    )
-    _patch_rate_limit_deps(mocker, target_user_id, daily_used=0, weekly_used=0)
-
-    response = client.post(
-        "/admin/rate_limit/reset",
-        json={"user_id": target_user_id, "reset_weekly": True},
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["daily_cost_used_microdollars"] == 0
-    assert data["weekly_cost_used_microdollars"] == 0
-    assert data["tier"] == "FREE"
-
-    mock_reset.assert_awaited_once_with(target_user_id, reset_weekly=True)
-
-    configured_snapshot.assert_match(
-        json.dumps(data, indent=2, sort_keys=True) + "\n",
-        "reset_user_usage_daily_and_weekly",
-    )
-
-
-def test_reset_user_usage_redis_failure(
-    mocker: pytest_mock.MockerFixture,
-    target_user_id: str,
-) -> None:
-    """Test that Redis failure on reset returns 500."""
-    mocker.patch(
-        f"{_MOCK_MODULE}.reset_user_usage",
-        new_callable=AsyncMock,
-        side_effect=Exception("Redis connection refused"),
-    )
-
-    response = client.post(
-        "/admin/rate_limit/reset",
-        json={"user_id": target_user_id},
-    )
-
-    assert response.status_code == 500
-
-
-def test_get_rate_limit_email_lookup_failure(
-    mocker: pytest_mock.MockerFixture,
-    target_user_id: str,
-) -> None:
-    """Test that failing to resolve a user email degrades gracefully."""
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_global_rate_limits",
-        new_callable=AsyncMock,
-        return_value=(2_500_000, 12_500_000, SubscriptionTier.FREE),
-    )
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_usage_status",
-        new_callable=AsyncMock,
-        return_value=_mock_usage_status(),
-    )
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_email_by_id",
-        new_callable=AsyncMock,
-        side_effect=Exception("DB connection lost"),
-    )
-
-    response = client.get("/admin/rate_limit", params={"user_id": target_user_id})
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["user_id"] == target_user_id
-    assert data["user_email"] is None
-
-
-def test_admin_endpoints_require_admin_role(mock_jwt_user) -> None:
-    """Test that rate limit admin endpoints require admin role."""
-    app.dependency_overrides[get_jwt_payload] = mock_jwt_user["get_jwt_payload"]
-
-    response = client.get("/admin/rate_limit", params={"user_id": "test"})
-    assert response.status_code == 403
-
-    response = client.post(
-        "/admin/rate_limit/reset",
-        json={"user_id": "test"},
-    )
-    assert response.status_code == 403
-
-
-# ---------------------------------------------------------------------------
-# Tier management endpoints
-# ---------------------------------------------------------------------------
-
-
-def test_get_user_tier(
-    mocker: pytest_mock.MockerFixture,
-    target_user_id: str,
-) -> None:
-    """Test getting a user's rate-limit tier."""
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_email_by_id",
-        new_callable=AsyncMock,
-        return_value=_TARGET_EMAIL,
-    )
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_tier",
-        new_callable=AsyncMock,
-        return_value=SubscriptionTier.PRO,
-    )
-
-    response = client.get("/admin/rate_limit/tier", params={"user_id": target_user_id})
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["user_id"] == target_user_id
-    assert data["tier"] == "PRO"
-
-
-def test_get_user_tier_user_not_found(
-    mocker: pytest_mock.MockerFixture,
-    target_user_id: str,
-) -> None:
-    """Test that getting tier for a non-existent user returns 404."""
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_email_by_id",
-        new_callable=AsyncMock,
-        return_value=None,
-    )
-
-    response = client.get("/admin/rate_limit/tier", params={"user_id": target_user_id})
-
-    assert response.status_code == 404
-
-
-def test_set_user_tier(
-    mocker: pytest_mock.MockerFixture,
-    target_user_id: str,
-) -> None:
-    """Test setting a user's rate-limit tier (upgrade)."""
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_email_by_id",
-        new_callable=AsyncMock,
-        return_value=_TARGET_EMAIL,
-    )
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_tier",
-        new_callable=AsyncMock,
-        return_value=SubscriptionTier.FREE,
-    )
-    mock_set = mocker.patch(
-        f"{_MOCK_MODULE}.set_user_tier",
-        new_callable=AsyncMock,
-    )
-
-    response = client.post(
-        "/admin/rate_limit/tier",
-        json={"user_id": target_user_id, "tier": "ENTERPRISE"},
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["user_id"] == target_user_id
-    assert data["tier"] == "ENTERPRISE"
-    mock_set.assert_awaited_once_with(target_user_id, SubscriptionTier.ENTERPRISE)
-
-
-def test_set_user_tier_downgrade(
-    mocker: pytest_mock.MockerFixture,
-    target_user_id: str,
-) -> None:
-    """Test downgrading a user's tier from PRO to FREE."""
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_email_by_id",
-        new_callable=AsyncMock,
-        return_value=_TARGET_EMAIL,
-    )
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_tier",
-        new_callable=AsyncMock,
-        return_value=SubscriptionTier.PRO,
-    )
-    mock_set = mocker.patch(
-        f"{_MOCK_MODULE}.set_user_tier",
-        new_callable=AsyncMock,
-    )
-
-    response = client.post(
-        "/admin/rate_limit/tier",
-        json={"user_id": target_user_id, "tier": "FREE"},
-    )
-
-    assert response.status_code == 200
-    data = response.json()
-    assert data["user_id"] == target_user_id
-    assert data["tier"] == "FREE"
-    mock_set.assert_awaited_once_with(target_user_id, SubscriptionTier.FREE)
-
-
-def test_set_user_tier_invalid_tier(
-    target_user_id: str,
-) -> None:
-    """Test that setting an invalid tier returns 422."""
-    response = client.post(
-        "/admin/rate_limit/tier",
-        json={"user_id": target_user_id, "tier": "invalid"},
-    )
-
-    assert response.status_code == 422
-
-
-def test_set_user_tier_invalid_tier_uppercase(
-    target_user_id: str,
-) -> None:
-    """Test that setting an unrecognised uppercase tier (e.g. 'INVALID') returns 422.
-
-    Regression: ensures Pydantic enum validation rejects values that are not
-    members of SubscriptionTier, even when they look like valid enum names.
-    """
-    response = client.post(
-        "/admin/rate_limit/tier",
-        json={"user_id": target_user_id, "tier": "INVALID"},
-    )
-
-    assert response.status_code == 422
-    body = response.json()
-    assert "detail" in body
-
-
-def test_set_user_tier_email_lookup_failure_returns_404(
-    mocker: pytest_mock.MockerFixture,
-    target_user_id: str,
-) -> None:
-    """Test that email lookup failure returns 404 (user unverifiable)."""
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_email_by_id",
-        new_callable=AsyncMock,
-        side_effect=Exception("DB connection failed"),
-    )
-
-    response = client.post(
-        "/admin/rate_limit/tier",
-        json={"user_id": target_user_id, "tier": "PRO"},
-    )
-
-    assert response.status_code == 404
-
-
-def test_set_user_tier_user_not_found(
-    mocker: pytest_mock.MockerFixture,
-    target_user_id: str,
-) -> None:
-    """Test that setting tier for a non-existent user returns 404."""
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_email_by_id",
-        new_callable=AsyncMock,
-        return_value=None,
-    )
-
-    response = client.post(
-        "/admin/rate_limit/tier",
-        json={"user_id": target_user_id, "tier": "PRO"},
-    )
-
-    assert response.status_code == 404
-
-
-def test_set_user_tier_db_failure(
-    mocker: pytest_mock.MockerFixture,
-    target_user_id: str,
-) -> None:
-    """Test that DB failure on set tier returns 500."""
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_email_by_id",
-        new_callable=AsyncMock,
-        return_value=_TARGET_EMAIL,
-    )
-    mocker.patch(
-        f"{_MOCK_MODULE}.get_user_tier",
-        new_callable=AsyncMock,
-        return_value=SubscriptionTier.FREE,
-    )
-    mocker.patch(
-        f"{_MOCK_MODULE}.set_user_tier",
-        new_callable=AsyncMock,
-        side_effect=Exception("DB connection refused"),
-    )
-
-    response = client.post(
-        "/admin/rate_limit/tier",
-        json={"user_id": target_user_id, "tier": "PRO"},
-    )
-
-    assert response.status_code == 500
-
-
-def test_tier_endpoints_require_admin_role(mock_jwt_user) -> None:
-    """Test that tier admin endpoints require admin role."""
-    app.dependency_overrides[get_jwt_payload] = mock_jwt_user["get_jwt_payload"]
-
-    response = client.get("/admin/rate_limit/tier", params={"user_id": "test"})
-    assert response.status_code == 403
-
-    response = client.post(
-        "/admin/rate_limit/tier",
-        json={"user_id": "test", "tier": "PRO"},
-    )
-    assert response.status_code == 403
-
-
-# ─── search_users endpoint ──────────────────────────────────────────
-
-
-def test_search_users_returns_matching_users(
-    mocker: pytest_mock.MockerFixture,
-    admin_user_id: str,
-) -> None:
-    """Partial search should return all matching users from the User table."""
-    mocker.patch(
-        _MOCK_MODULE + ".search_users",
-        new_callable=AsyncMock,
-        return_value=[
-            ("user-1", "zamil.majdy@gmail.com"),
-            ("user-2", "zamil.majdy@agpt.co"),
-        ],
-    )
-
-    response = client.get("/admin/rate_limit/search_users", params={"query": "zamil"})
-
-    assert response.status_code == 200
-    results = response.json()
-    assert len(results) == 2
-    assert results[0]["user_email"] == "zamil.majdy@gmail.com"
-    assert results[1]["user_email"] == "zamil.majdy@agpt.co"
-
-
-def test_search_users_empty_results(
-    mocker: pytest_mock.MockerFixture,
-    admin_user_id: str,
-) -> None:
-    """Search with no matches returns empty list."""
-    mocker.patch(
-        _MOCK_MODULE + ".search_users",
-        new_callable=AsyncMock,
-        return_value=[],
-    )
-
-    response = client.get(
-        "/admin/rate_limit/search_users", params={"query": "nonexistent"}
-    )
-
-    assert response.status_code == 200
-    assert response.json() == []
-
-
-def test_search_users_short_query_rejected(
-    admin_user_id: str,
-) -> None:
-    """Query shorter than 3 characters should return 400."""
-    response = client.get("/admin/rate_limit/search_users", params={"query": "ab"})
-    assert response.status_code == 400
-
-
-def test_search_users_negative_limit_clamped(
-    mocker: pytest_mock.MockerFixture,
-    admin_user_id: str,
-) -> None:
-    """Negative limit should be clamped to 1, not passed through."""
-    mock_search = mocker.patch(
-        _MOCK_MODULE + ".search_users",
-        new_callable=AsyncMock,
-        return_value=[],
-    )
-
-    response = client.get(
-        "/admin/rate_limit/search_users", params={"query": "test", "limit": -1}
-    )
-
-    assert response.status_code == 200
-    mock_search.assert_awaited_once_with("test", limit=1)
-
-
-def test_search_users_requires_admin_role(mock_jwt_user) -> None:
-    """Test that the search_users endpoint requires admin role."""
-    app.dependency_overrides[get_jwt_payload] = mock_jwt_user["get_jwt_payload"]
-
-    response = client.get("/admin/rate_limit/search_users", params={"query": "test"})
-    assert response.status_code == 403

autogpt_platform/backend/backend/api/features/admin/store_admin_routes.py

@@ -7,8 +7,6 @@ import fastapi
 import fastapi.responses
 import prisma.enums
 
-import backend.api.features.library.db as library_db
-import backend.api.features.library.model as library_model
 import backend.api.features.store.cache as store_cache
 import backend.api.features.store.db as store_db
 import backend.api.features.store.model as store_model
@@ -26,13 +24,14 @@ router = fastapi.APIRouter(
 @router.get(
     "/listings",
     summary="Get Admin Listings History",
+    response_model=store_model.StoreListingsWithVersionsResponse,
 )
 async def get_admin_listings_with_versions(
     status: typing.Optional[prisma.enums.SubmissionStatus] = None,
     search: typing.Optional[str] = None,
     page: int = 1,
     page_size: int = 20,
-) -> store_model.StoreListingsWithVersionsAdminViewResponse:
+):
     """
     Get store listings with their version history for admins.
 
@@ -46,26 +45,36 @@ async def get_admin_listings_with_versions(
         page_size: Number of items per page
 
     Returns:
-        Paginated listings with their versions
+        StoreListingsWithVersionsResponse with listings and their versions
     """
-    listings = await store_db.get_admin_listings_with_versions(
-        status=status,
-        search_query=search,
-        page=page,
-        page_size=page_size,
-    )
-    return listings
+    try:
+        listings = await store_db.get_admin_listings_with_versions(
+            status=status,
+            search_query=search,
+            page=page,
+            page_size=page_size,
+        )
+        return listings
+    except Exception as e:
+        logger.exception("Error getting admin listings with versions: %s", e)
+        return fastapi.responses.JSONResponse(
+            status_code=500,
+            content={
+                "detail": "An error occurred while retrieving listings with versions"
+            },
+        )
 
 
 @router.post(
     "/submissions/{store_listing_version_id}/review",
     summary="Review Store Submission",
+    response_model=store_model.StoreSubmission,
 )
 async def review_submission(
     store_listing_version_id: str,
     request: store_model.ReviewSubmissionRequest,
     user_id: str = fastapi.Security(autogpt_libs.auth.get_user_id),
-) -> store_model.StoreSubmissionAdminView:
+):
     """
     Review a store listing submission.
 
@@ -75,24 +84,31 @@ async def review_submission(
         user_id: Authenticated admin user performing the review
 
     Returns:
-        StoreSubmissionAdminView with updated review information
+        StoreSubmission with updated review information
     """
-    already_approved = await store_db.check_submission_already_approved(
-        store_listing_version_id=store_listing_version_id,
-    )
-    submission = await store_db.review_store_submission(
-        store_listing_version_id=store_listing_version_id,
-        is_approved=request.is_approved,
-        external_comments=request.comments,
-        internal_comments=request.internal_comments or "",
-        reviewer_id=user_id,
-    )
+    try:
+        already_approved = await store_db.check_submission_already_approved(
+            store_listing_version_id=store_listing_version_id,
+        )
+        submission = await store_db.review_store_submission(
+            store_listing_version_id=store_listing_version_id,
+            is_approved=request.is_approved,
+            external_comments=request.comments,
+            internal_comments=request.internal_comments or "",
+            reviewer_id=user_id,
+        )
 
-    state_changed = already_approved != request.is_approved
-    # Clear caches whenever approval state changes, since store visibility can change
-    if state_changed:
-        store_cache.clear_all_caches()
-    return submission
+        state_changed = already_approved != request.is_approved
+        # Clear caches when the request is approved as it updates what is shown on the store
+        if state_changed:
+            store_cache.clear_all_caches()
+        return submission
+    except Exception as e:
+        logger.exception("Error reviewing submission: %s", e)
+        return fastapi.responses.JSONResponse(
+            status_code=500,
+            content={"detail": "An error occurred while reviewing the submission"},
+        )
 
 
 @router.get(
@@ -134,40 +150,3 @@ async def admin_download_agent_file(
         return fastapi.responses.FileResponse(
             tmp_file.name, filename=file_name, media_type="application/json"
         )
-
-
-@router.get(
-    "/submissions/{store_listing_version_id}/preview",
-    summary="Admin Preview Submission Listing",
-)
-async def admin_preview_submission(
-    store_listing_version_id: str,
-) -> store_model.StoreAgentDetails:
-    """
-    Preview a marketplace submission as it would appear on the listing page.
-    Bypasses the APPROVED-only StoreAgent view so admins can preview pending
-    submissions before approving.
-    """
-    return await store_db.get_store_agent_details_as_admin(store_listing_version_id)
-
-
-@router.post(
-    "/submissions/{store_listing_version_id}/add-to-library",
-    summary="Admin Add Pending Agent to Library",
-    status_code=201,
-)
-async def admin_add_agent_to_library(
-    store_listing_version_id: str,
-    user_id: str = fastapi.Security(autogpt_libs.auth.get_user_id),
-) -> library_model.LibraryAgent:
-    """
-    Add a pending marketplace agent to the admin's library for review.
-    Uses admin-level access to bypass marketplace APPROVED-only checks.
-
-    The builder can load the graph because get_graph() checks library
-    membership as a fallback: "you added it, you keep it."
-    """
-    return await library_db.add_store_agent_to_library_as_admin(
-        store_listing_version_id=store_listing_version_id,
-        user_id=user_id,
-    )

autogpt_platform/backend/backend/api/features/admin/store_admin_routes_test.py

@@ -1,335 +0,0 @@
-"""Tests for admin store routes and the bypass logic they depend on.
-
-Tests are organized by what they protect:
-- SECRT-2162: get_graph_as_admin bypasses ownership/marketplace checks
-- SECRT-2167 security: admin endpoints reject non-admin users
-- SECRT-2167 bypass: preview queries StoreListingVersion (not StoreAgent view),
-  and add-to-library uses get_graph_as_admin (not get_graph)
-"""
-
-from datetime import datetime, timezone
-from unittest.mock import AsyncMock, MagicMock, patch
-
-import fastapi
-import fastapi.responses
-import fastapi.testclient
-import pytest
-import pytest_mock
-from autogpt_libs.auth.jwt_utils import get_jwt_payload
-
-from backend.data.graph import get_graph_as_admin
-from backend.util.exceptions import NotFoundError
-
-from .store_admin_routes import router as store_admin_router
-
-# Shared constants
-ADMIN_USER_ID = "admin-user-id"
-CREATOR_USER_ID = "other-creator-id"
-GRAPH_ID = "test-graph-id"
-GRAPH_VERSION = 3
-SLV_ID = "test-store-listing-version-id"
-
-
-def _make_mock_graph(user_id: str = CREATOR_USER_ID) -> MagicMock:
-    graph = MagicMock()
-    graph.userId = user_id
-    graph.id = GRAPH_ID
-    graph.version = GRAPH_VERSION
-    graph.Nodes = []
-    return graph
-
-
-# ---- SECRT-2162: get_graph_as_admin bypasses ownership checks ---- #
-
-
-@pytest.mark.asyncio
-async def test_admin_can_access_pending_agent_not_owned() -> None:
-    """get_graph_as_admin must return a graph even when the admin doesn't own
-    it and it's not APPROVED in the marketplace."""
-    mock_graph = _make_mock_graph()
-    mock_graph_model = MagicMock(name="GraphModel")
-
-    with (
-        patch("backend.data.graph.AgentGraph.prisma") as mock_prisma,
-        patch(
-            "backend.data.graph.GraphModel.from_db",
-            return_value=mock_graph_model,
-        ),
-    ):
-        mock_prisma.return_value.find_first = AsyncMock(return_value=mock_graph)
-
-        result = await get_graph_as_admin(
-            graph_id=GRAPH_ID,
-            version=GRAPH_VERSION,
-            user_id=ADMIN_USER_ID,
-            for_export=False,
-        )
-
-    assert result is mock_graph_model
-
-
-@pytest.mark.asyncio
-async def test_admin_download_pending_agent_with_subagents() -> None:
-    """get_graph_as_admin with for_export=True must call get_sub_graphs
-    and pass sub_graphs to GraphModel.from_db."""
-    mock_graph = _make_mock_graph()
-    mock_sub_graph = MagicMock(name="SubGraph")
-    mock_graph_model = MagicMock(name="GraphModel")
-
-    with (
-        patch("backend.data.graph.AgentGraph.prisma") as mock_prisma,
-        patch(
-            "backend.data.graph.get_sub_graphs",
-            new_callable=AsyncMock,
-            return_value=[mock_sub_graph],
-        ) as mock_get_sub,
-        patch(
-            "backend.data.graph.GraphModel.from_db",
-            return_value=mock_graph_model,
-        ) as mock_from_db,
-    ):
-        mock_prisma.return_value.find_first = AsyncMock(return_value=mock_graph)
-
-        result = await get_graph_as_admin(
-            graph_id=GRAPH_ID,
-            version=GRAPH_VERSION,
-            user_id=ADMIN_USER_ID,
-            for_export=True,
-        )
-
-    assert result is mock_graph_model
-    mock_get_sub.assert_awaited_once_with(mock_graph)
-    mock_from_db.assert_called_once_with(
-        graph=mock_graph,
-        sub_graphs=[mock_sub_graph],
-        for_export=True,
-    )
-
-
-# ---- SECRT-2167 security: admin endpoints reject non-admin users ---- #
-
-app = fastapi.FastAPI()
-app.include_router(store_admin_router)
-
-
-@app.exception_handler(NotFoundError)
-async def _not_found_handler(
-    request: fastapi.Request, exc: NotFoundError
-) -> fastapi.responses.JSONResponse:
-    return fastapi.responses.JSONResponse(status_code=404, content={"detail": str(exc)})
-
-
-client = fastapi.testclient.TestClient(app)
-
-
-@pytest.fixture(autouse=True)
-def setup_app_admin_auth(mock_jwt_admin):
-    """Setup admin auth overrides for all route tests in this module."""
-    app.dependency_overrides[get_jwt_payload] = mock_jwt_admin["get_jwt_payload"]
-    yield
-    app.dependency_overrides.clear()
-
-
-def test_preview_requires_admin(mock_jwt_user) -> None:
-    """Non-admin users must get 403 on the preview endpoint."""
-    app.dependency_overrides[get_jwt_payload] = mock_jwt_user["get_jwt_payload"]
-    response = client.get(f"/admin/submissions/{SLV_ID}/preview")
-    assert response.status_code == 403
-
-
-def test_add_to_library_requires_admin(mock_jwt_user) -> None:
-    """Non-admin users must get 403 on the add-to-library endpoint."""
-    app.dependency_overrides[get_jwt_payload] = mock_jwt_user["get_jwt_payload"]
-    response = client.post(f"/admin/submissions/{SLV_ID}/add-to-library")
-    assert response.status_code == 403
-
-
-def test_preview_nonexistent_submission(
-    mocker: pytest_mock.MockerFixture,
-) -> None:
-    """Preview of a nonexistent submission returns 404."""
-    mocker.patch(
-        "backend.api.features.admin.store_admin_routes.store_db"
-        ".get_store_agent_details_as_admin",
-        side_effect=NotFoundError("not found"),
-    )
-    response = client.get(f"/admin/submissions/{SLV_ID}/preview")
-    assert response.status_code == 404
-
-
-# ---- SECRT-2167 bypass: verify the right data sources are used ---- #
-
-
-@pytest.mark.asyncio
-async def test_preview_queries_store_listing_version_not_store_agent() -> None:
-    """get_store_agent_details_as_admin must query StoreListingVersion
-    directly (not the APPROVED-only StoreAgent view). This is THE test that
-    prevents the bypass from being accidentally reverted."""
-    from backend.api.features.store.db import get_store_agent_details_as_admin
-
-    mock_slv = MagicMock()
-    mock_slv.id = SLV_ID
-    mock_slv.name = "Test Agent"
-    mock_slv.subHeading = "Short desc"
-    mock_slv.description = "Long desc"
-    mock_slv.videoUrl = None
-    mock_slv.agentOutputDemoUrl = None
-    mock_slv.imageUrls = ["https://example.com/img.png"]
-    mock_slv.instructions = None
-    mock_slv.categories = ["productivity"]
-    mock_slv.version = 1
-    mock_slv.agentGraphId = GRAPH_ID
-    mock_slv.agentGraphVersion = GRAPH_VERSION
-    mock_slv.updatedAt = datetime(2026, 3, 24, tzinfo=timezone.utc)
-    mock_slv.recommendedScheduleCron = "0 9 * * *"
-
-    mock_listing = MagicMock()
-    mock_listing.id = "listing-id"
-    mock_listing.slug = "test-agent"
-    mock_listing.activeVersionId = SLV_ID
-    mock_listing.hasApprovedVersion = False
-    mock_listing.CreatorProfile = MagicMock(username="creator", avatarUrl="")
-    mock_slv.StoreListing = mock_listing
-
-    with (
-        patch(
-            "backend.api.features.store.db.prisma.models" ".StoreListingVersion.prisma",
-        ) as mock_slv_prisma,
-        patch(
-            "backend.api.features.store.db.prisma.models.StoreAgent.prisma",
-        ) as mock_store_agent_prisma,
-    ):
-        mock_slv_prisma.return_value.find_unique = AsyncMock(return_value=mock_slv)
-
-        result = await get_store_agent_details_as_admin(SLV_ID)
-
-    # Verify it queried StoreListingVersion (not the APPROVED-only StoreAgent)
-    mock_slv_prisma.return_value.find_unique.assert_awaited_once()
-    await_args = mock_slv_prisma.return_value.find_unique.await_args
-    assert await_args is not None
-    assert await_args.kwargs["where"] == {"id": SLV_ID}
-
-    # Verify the APPROVED-only StoreAgent view was NOT touched
-    mock_store_agent_prisma.assert_not_called()
-
-    # Verify the result has the right data
-    assert result.agent_name == "Test Agent"
-    assert result.agent_image == ["https://example.com/img.png"]
-    assert result.has_approved_version is False
-    assert result.runs == 0
-    assert result.rating == 0.0
-
-
-@pytest.mark.asyncio
-async def test_resolve_graph_admin_uses_get_graph_as_admin() -> None:
-    """resolve_graph_for_library(admin=True) must call get_graph_as_admin,
-    not get_graph. This is THE test that prevents the add-to-library bypass
-    from being accidentally reverted."""
-    from backend.api.features.library._add_to_library import resolve_graph_for_library
-
-    mock_slv = MagicMock()
-    mock_slv.AgentGraph = MagicMock(id=GRAPH_ID, version=GRAPH_VERSION)
-    mock_graph_model = MagicMock(name="GraphModel")
-
-    with (
-        patch(
-            "backend.api.features.library._add_to_library.prisma.models"
-            ".StoreListingVersion.prisma",
-        ) as mock_prisma,
-        patch(
-            "backend.api.features.library._add_to_library.graph_db"
-            ".get_graph_as_admin",
-            new_callable=AsyncMock,
-            return_value=mock_graph_model,
-        ) as mock_admin,
-        patch(
-            "backend.api.features.library._add_to_library.graph_db.get_graph",
-            new_callable=AsyncMock,
-        ) as mock_regular,
-    ):
-        mock_prisma.return_value.find_unique = AsyncMock(return_value=mock_slv)
-
-        result = await resolve_graph_for_library(SLV_ID, ADMIN_USER_ID, admin=True)
-
-    assert result is mock_graph_model
-    mock_admin.assert_awaited_once_with(
-        graph_id=GRAPH_ID, version=GRAPH_VERSION, user_id=ADMIN_USER_ID
-    )
-    mock_regular.assert_not_awaited()
-
-
-@pytest.mark.asyncio
-async def test_resolve_graph_regular_uses_get_graph() -> None:
-    """resolve_graph_for_library(admin=False) must call get_graph,
-    not get_graph_as_admin. Ensures the non-admin path is preserved."""
-    from backend.api.features.library._add_to_library import resolve_graph_for_library
-
-    mock_slv = MagicMock()
-    mock_slv.AgentGraph = MagicMock(id=GRAPH_ID, version=GRAPH_VERSION)
-    mock_graph_model = MagicMock(name="GraphModel")
-
-    with (
-        patch(
-            "backend.api.features.library._add_to_library.prisma.models"
-            ".StoreListingVersion.prisma",
-        ) as mock_prisma,
-        patch(
-            "backend.api.features.library._add_to_library.graph_db"
-            ".get_graph_as_admin",
-            new_callable=AsyncMock,
-        ) as mock_admin,
-        patch(
-            "backend.api.features.library._add_to_library.graph_db.get_graph",
-            new_callable=AsyncMock,
-            return_value=mock_graph_model,
-        ) as mock_regular,
-    ):
-        mock_prisma.return_value.find_unique = AsyncMock(return_value=mock_slv)
-
-        result = await resolve_graph_for_library(SLV_ID, "regular-user-id", admin=False)
-
-    assert result is mock_graph_model
-    mock_regular.assert_awaited_once_with(
-        graph_id=GRAPH_ID, version=GRAPH_VERSION, user_id="regular-user-id"
-    )
-    mock_admin.assert_not_awaited()
-
-
-# ---- Library membership grants graph access (product decision) ---- #
-
-
-@pytest.mark.asyncio
-async def test_library_member_can_view_pending_agent_in_builder() -> None:
-    """After adding a pending agent to their library, the user should be
-    able to load the graph in the builder via get_graph()."""
-    mock_graph = _make_mock_graph()
-    mock_graph_model = MagicMock(name="GraphModel")
-    mock_library_agent = MagicMock()
-    mock_library_agent.AgentGraph = mock_graph
-
-    with (
-        patch("backend.data.graph.AgentGraph.prisma") as mock_ag_prisma,
-        patch(
-            "backend.data.graph.StoreListingVersion.prisma",
-        ) as mock_slv_prisma,
-        patch("backend.data.graph.LibraryAgent.prisma") as mock_lib_prisma,
-        patch(
-            "backend.data.graph.GraphModel.from_db",
-            return_value=mock_graph_model,
-        ),
-    ):
-        mock_ag_prisma.return_value.find_first = AsyncMock(return_value=None)
-        mock_slv_prisma.return_value.find_first = AsyncMock(return_value=None)
-        mock_lib_prisma.return_value.find_first = AsyncMock(
-            return_value=mock_library_agent
-        )
-
-        from backend.data.graph import get_graph
-
-        result = await get_graph(
-            graph_id=GRAPH_ID,
-            version=GRAPH_VERSION,
-            user_id=ADMIN_USER_ID,
-        )
-
-    assert result is mock_graph_model, "Library membership should grant graph access"

autogpt_platform/backend/backend/api/features/builder/db.py

@@ -1,10 +1,10 @@
 import logging
 from dataclasses import dataclass
+from datetime import datetime, timedelta, timezone
 from difflib import SequenceMatcher
-from typing import Any, Sequence, get_args, get_origin
+from typing import Sequence
 
 import prisma
-from prisma.models import mv_suggested_blocks
 
 import backend.api.features.library.db as library_db
 import backend.api.features.library.model as library_model
@@ -19,10 +19,10 @@ from backend.blocks._base import (
     BlockType,
 )
 from backend.blocks.llm import LlmModel
+from backend.data.db import query_raw_with_schema
 from backend.integrations.providers import ProviderName
 from backend.util.cache import cached
 from backend.util.models import Pagination
-from backend.util.text import split_camelcase
 
 from .model import (
     BlockCategoryResponse,
@@ -42,16 +42,6 @@ MAX_LIBRARY_AGENT_RESULTS = 100
 MAX_MARKETPLACE_AGENT_RESULTS = 100
 MIN_SCORE_FOR_FILTERED_RESULTS = 10.0
 
-# Boost blocks over marketplace agents in search results
-BLOCK_SCORE_BOOST = 50.0
-
-# Block IDs to exclude from search results
-EXCLUDED_BLOCK_IDS = frozenset(
-    {
-        "e189baac-8c20-45a1-94a7-55177ea42565",  # AgentExecutorBlock
-    }
-)
-
 SearchResultItem = BlockInfo | library_model.LibraryAgent | store_model.StoreAgent
 
 
@@ -74,8 +64,8 @@ def get_block_categories(category_blocks: int = 3) -> list[BlockCategoryResponse
 
     for block_type in load_all_blocks().values():
         block: AnyBlockSchema = block_type()
-        # Skip disabled and excluded blocks
-        if block.disabled or block.id in EXCLUDED_BLOCK_IDS:
+        # Skip disabled blocks
+        if block.disabled:
             continue
         # Skip blocks that don't have categories (all should have at least one)
         if not block.categories:
@@ -126,9 +116,6 @@ def get_blocks(
         # Skip disabled blocks
         if block.disabled:
             continue
-        # Skip excluded blocks
-        if block.id in EXCLUDED_BLOCK_IDS:
-            continue
         # Skip blocks that don't match the category
         if category and category not in {c.name.lower() for c in block.categories}:
             continue
@@ -268,25 +255,14 @@ async def _build_cached_search_results(
         "my_agents": 0,
     }
 
-    # Use hybrid search when query is present, otherwise list all blocks
-    if (include_blocks or include_integrations) and normalized_query:
-        block_results, block_total, integration_total = await _text_search_blocks(
-            query=search_query,
-            include_blocks=include_blocks,
-            include_integrations=include_integrations,
-        )
-        scored_items.extend(block_results)
-        total_items["blocks"] = block_total
-        total_items["integrations"] = integration_total
-    elif include_blocks or include_integrations:
-        # No query - list all blocks using in-memory approach
-        block_results, block_total, integration_total = _collect_block_results(
-            include_blocks=include_blocks,
-            include_integrations=include_integrations,
-        )
-        scored_items.extend(block_results)
-        total_items["blocks"] = block_total
-        total_items["integrations"] = integration_total
+    block_results, block_total, integration_total = _collect_block_results(
+        normalized_query=normalized_query,
+        include_blocks=include_blocks,
+        include_integrations=include_integrations,
+    )
+    scored_items.extend(block_results)
+    total_items["blocks"] = block_total
+    total_items["integrations"] = integration_total
 
     if include_library_agents:
         library_response = await library_db.list_library_agents(
@@ -331,14 +307,10 @@ async def _build_cached_search_results(
 
 def _collect_block_results(
     *,
+    normalized_query: str,
     include_blocks: bool,
     include_integrations: bool,
 ) -> tuple[list[_ScoredItem], int, int]:
-    """
-    Collect all blocks for listing (no search query).
-
-    All blocks get BLOCK_SCORE_BOOST to prioritize them over marketplace agents.
-    """
     results: list[_ScoredItem] = []
     block_count = 0
     integration_count = 0
@@ -351,10 +323,6 @@ def _collect_block_results(
         if block.disabled:
             continue
 
-        # Skip excluded blocks
-        if block.id in EXCLUDED_BLOCK_IDS:
-            continue
-
         block_info = block.get_info()
         credentials = list(block.input_schema.get_credentials_fields().values())
         is_integration = len(credentials) > 0
@@ -364,6 +332,10 @@ def _collect_block_results(
         if not is_integration and not include_blocks:
             continue
 
+        score = _score_block(block, block_info, normalized_query)
+        if not _should_include_item(score, normalized_query):
+            continue
+
         filter_type: FilterType = "integrations" if is_integration else "blocks"
         if is_integration:
             integration_count += 1
@@ -374,86 +346,14 @@ def _collect_block_results(
             _ScoredItem(
                 item=block_info,
                 filter_type=filter_type,
-                score=BLOCK_SCORE_BOOST,
-                sort_key=block_info.name.lower(),
+                score=score,
+                sort_key=_get_item_name(block_info),
             )
         )
 
     return results, block_count, integration_count
 
 
-async def _text_search_blocks(
-    *,
-    query: str,
-    include_blocks: bool,
-    include_integrations: bool,
-) -> tuple[list[_ScoredItem], int, int]:
-    """
-    Search blocks using in-memory text matching over the block registry.
-
-    All blocks are already loaded in memory, so this is fast and reliable
-    regardless of whether OpenAI embeddings are available.
-
-    Scoring:
-        - Base: text relevance via _score_primary_fields, plus BLOCK_SCORE_BOOST
-          to prioritize blocks over marketplace agents in combined results
-        - +20 if the block has an LlmModel field and the query matches an LLM model name
-    """
-    results: list[_ScoredItem] = []
-
-    if not include_blocks and not include_integrations:
-        return results, 0, 0
-
-    normalized_query = query.strip().lower()
-
-    all_results, _, _ = _collect_block_results(
-        include_blocks=include_blocks,
-        include_integrations=include_integrations,
-    )
-
-    all_blocks = load_all_blocks()
-
-    for item in all_results:
-        block_info = item.item
-        assert isinstance(block_info, BlockInfo)
-        name = split_camelcase(block_info.name).lower()
-
-        # Build rich description including input field descriptions,
-        # matching the searchable text that the embedding pipeline uses
-        desc_parts = [block_info.description or ""]
-        block_cls = all_blocks.get(block_info.id)
-        if block_cls is not None:
-            block: AnyBlockSchema = block_cls()
-            desc_parts += [
-                f"{f}: {info.description}"
-                for f, info in block.input_schema.model_fields.items()
-                if info.description
-            ]
-        description = " ".join(desc_parts).lower()
-
-        score = _score_primary_fields(name, description, normalized_query)
-
-        # Add LLM model match bonus
-        if block_cls is not None and _matches_llm_model(
-            block_cls().input_schema, normalized_query
-        ):
-            score += 20
-
-        if score >= MIN_SCORE_FOR_FILTERED_RESULTS:
-            results.append(
-                _ScoredItem(
-                    item=block_info,
-                    filter_type=item.filter_type,
-                    score=score + BLOCK_SCORE_BOOST,
-                    sort_key=name,
-                )
-            )
-
-    block_count = sum(1 for r in results if r.filter_type == "blocks")
-    integration_count = sum(1 for r in results if r.filter_type == "integrations")
-    return results, block_count, integration_count
-
-
 def _build_library_items(
     *,
     agents: list[library_model.LibraryAgent],
@@ -572,8 +472,6 @@ async def _get_static_counts():
         block: AnyBlockSchema = block_type()
         if block.disabled:
             continue
-        if block.id in EXCLUDED_BLOCK_IDS:
-            continue
 
         all_blocks += 1
 
@@ -600,25 +498,47 @@ async def _get_static_counts():
     }
 
 
-def _contains_type(annotation: Any, target: type) -> bool:
-    """Check if an annotation is or contains the target type (handles Optional/Union/Annotated)."""
-    if annotation is target:
-        return True
-    origin = get_origin(annotation)
-    if origin is None:
-        return False
-    return any(_contains_type(arg, target) for arg in get_args(annotation))
-
-
 def _matches_llm_model(schema_cls: type[BlockSchema], query: str) -> bool:
     for field in schema_cls.model_fields.values():
-        if _contains_type(field.annotation, LlmModel):
+        if field.annotation == LlmModel:
             # Check if query matches any value in llm_models
             if any(query in name for name in llm_models):
                 return True
     return False
 
 
+def _score_block(
+    block: AnyBlockSchema,
+    block_info: BlockInfo,
+    normalized_query: str,
+) -> float:
+    if not normalized_query:
+        return 0.0
+
+    name = block_info.name.lower()
+    description = block_info.description.lower()
+    score = _score_primary_fields(name, description, normalized_query)
+
+    category_text = " ".join(
+        category.get("category", "").lower() for category in block_info.categories
+    )
+    score += _score_additional_field(category_text, normalized_query, 12, 6)
+
+    credentials_info = block.input_schema.get_credentials_fields_info().values()
+    provider_names = [
+        provider.value.lower()
+        for info in credentials_info
+        for provider in info.provider
+    ]
+    provider_text = " ".join(provider_names)
+    score += _score_additional_field(provider_text, normalized_query, 15, 6)
+
+    if _matches_llm_model(block.input_schema, normalized_query):
+        score += 20
+
+    return score
+
+
 def _score_library_agent(
     agent: library_model.LibraryAgent,
     normalized_query: str,
@@ -725,20 +645,31 @@ def _get_all_providers() -> dict[ProviderName, Provider]:
     return providers
 
 
-@cached(ttl_seconds=3600, shared_cache=True)
+@cached(ttl_seconds=3600)
 async def get_suggested_blocks(count: int = 5) -> list[BlockInfo]:
-    """Return the most-executed blocks from the last 14 days.
+    suggested_blocks = []
+    # Sum the number of executions for each block type
+    # Prisma cannot group by nested relations, so we do a raw query
+    # Calculate the cutoff timestamp
+    timestamp_threshold = datetime.now(timezone.utc) - timedelta(days=30)
 
-    Queries the mv_suggested_blocks materialized view (refreshed hourly via pg_cron)
-    and returns the top `count` blocks sorted by execution count, excluding
-    Input/Output/Agent block types and blocks in EXCLUDED_BLOCK_IDS.
-    """
-    results = await mv_suggested_blocks.prisma().find_many()
+    results = await query_raw_with_schema(
+        """
+        SELECT
+            agent_node."agentBlockId" AS block_id,
+            COUNT(execution.id) AS execution_count
+        FROM {schema_prefix}"AgentNodeExecution" execution
+        JOIN {schema_prefix}"AgentNode" agent_node ON execution."agentNodeId" = agent_node.id
+        WHERE execution."endedTime" >= $1::timestamp
+        GROUP BY agent_node."agentBlockId"
+        ORDER BY execution_count DESC;
+        """,
+        timestamp_threshold,
+    )
 
     # Get the top blocks based on execution count
-    # But ignore Input, Output, Agent, and excluded blocks
+    # But ignore Input and Output blocks
     blocks: list[tuple[BlockInfo, int]] = []
-    execution_counts = {row.block_id: row.execution_count for row in results}
 
     for block_type in load_all_blocks().values():
         block: AnyBlockSchema = block_type()
@@ -748,9 +679,11 @@ async def get_suggested_blocks(count: int = 5) -> list[BlockInfo]:
             BlockType.AGENT,
         ):
             continue
-        if block.id in EXCLUDED_BLOCK_IDS:
-            continue
-        execution_count = execution_counts.get(block.id, 0)
+        # Find the execution count for this block
+        execution_count = next(
+            (row["execution_count"] for row in results if row["block_id"] == block.id),
+            0,
+        )
         blocks.append((block.get_info(), execution_count))
     # Sort blocks by execution count
     blocks.sort(key=lambda x: x[1], reverse=True)

autogpt_platform/backend/backend/api/features/builder/model.py

@@ -27,6 +27,7 @@ class SearchEntry(BaseModel):
 
 # Suggestions
 class SuggestionsResponse(BaseModel):
+    otto_suggestions: list[str]
     recent_searches: list[SearchEntry]
     providers: list[ProviderName]
     top_blocks: list[BlockInfo]

autogpt_platform/backend/backend/api/features/builder/routes.py

@@ -1,5 +1,5 @@
 import logging
-from typing import Annotated, Sequence, cast, get_args
+from typing import Annotated, Sequence
 
 import fastapi
 from autogpt_libs.auth.dependencies import get_user_id, requires_user
@@ -10,8 +10,6 @@ from backend.util.models import Pagination
 from . import db as builder_db
 from . import model as builder_model
 
-VALID_FILTER_VALUES = get_args(builder_model.FilterType)
-
 logger = logging.getLogger(__name__)
 
 router = fastapi.APIRouter(
@@ -51,6 +49,11 @@ async def get_suggestions(
     Get all suggestions for the Blocks Menu.
     """
     return builder_model.SuggestionsResponse(
+        otto_suggestions=[
+            "What blocks do I need to get started?",
+            "Help me create a list",
+            "Help me feed my data to Google Maps",
+        ],
         recent_searches=await builder_db.get_recent_searches(user_id),
         providers=[
             ProviderName.TWITTER,
@@ -148,7 +151,7 @@ async def get_providers(
 async def search(
     user_id: Annotated[str, fastapi.Security(get_user_id)],
     search_query: Annotated[str | None, fastapi.Query()] = None,
-    filter: Annotated[str | None, fastapi.Query()] = None,
+    filter: Annotated[list[builder_model.FilterType] | None, fastapi.Query()] = None,
     search_id: Annotated[str | None, fastapi.Query()] = None,
     by_creator: Annotated[list[str] | None, fastapi.Query()] = None,
     page: Annotated[int, fastapi.Query()] = 1,
@@ -157,20 +160,9 @@ async def search(
     """
     Search for blocks (including integrations), marketplace agents, and user library agents.
     """
-    # Parse and validate filter parameter
-    filters: list[builder_model.FilterType]
-    if filter:
-        filter_values = [f.strip() for f in filter.split(",")]
-        invalid_filters = [f for f in filter_values if f not in VALID_FILTER_VALUES]
-        if invalid_filters:
-            raise fastapi.HTTPException(
-                status_code=400,
-                detail=f"Invalid filter value(s): {', '.join(invalid_filters)}. "
-                f"Valid values are: {', '.join(VALID_FILTER_VALUES)}",
-            )
-        filters = cast(list[builder_model.FilterType], filter_values)
-    else:
-        filters = [
+    # If no filters are provided, then we will return all types
+    if not filter:
+        filter = [
             "blocks",
             "integrations",
             "marketplace_agents",
@@ -182,7 +174,7 @@ async def search(
     cached_results = await builder_db.get_sorted_search_results(
         user_id=user_id,
         search_query=search_query,
-        filters=filters,
+        filters=filter,
         by_creator=by_creator,
     )
 
@@ -204,7 +196,7 @@ async def search(
         user_id,
         builder_model.SearchEntry(
             search_query=search_query,
-            filter=filters,
+            filter=filter,
             by_creator=by_creator,
             search_id=search_id,
         ),

autogpt_platform/backend/backend/api/features/chat/completion_consumer.py

@@ -0,0 +1,368 @@
+"""Redis Streams consumer for operation completion messages.
+
+This module provides a consumer (ChatCompletionConsumer) that listens for
+completion notifications (OperationCompleteMessage) from external services
+(like Agent Generator) and triggers the appropriate stream registry and
+chat service updates via process_operation_success/process_operation_failure.
+
+Why Redis Streams instead of RabbitMQ?
+--------------------------------------
+While the project typically uses RabbitMQ for async task queues (e.g., execution
+queue), Redis Streams was chosen for chat completion notifications because:
+
+1. **Unified Infrastructure**: The SSE reconnection feature already uses Redis
+   Streams (via stream_registry) for message persistence and replay. Using Redis
+   Streams for completion notifications keeps all chat streaming infrastructure
+   in one system, simplifying operations and reducing cross-system coordination.
+
+2. **Message Replay**: Redis Streams support XREAD with arbitrary message IDs,
+   allowing consumers to replay missed messages after reconnection. This aligns
+   with the SSE reconnection pattern where clients can resume from last_message_id.
+
+3. **Consumer Groups with XAUTOCLAIM**: Redis consumer groups provide automatic
+   load balancing across pods with explicit message claiming (XAUTOCLAIM) for
+   recovering from dead consumers - ideal for the completion callback pattern.
+
+4. **Lower Latency**: For real-time SSE updates, Redis (already in-memory for
+   stream_registry) provides lower latency than an additional RabbitMQ hop.
+
+5. **Atomicity with Task State**: Completion processing often needs to update
+   task metadata stored in Redis. Keeping both in Redis enables simpler
+   transactional semantics without distributed coordination.
+
+The consumer uses Redis Streams with consumer groups for reliable message
+processing across multiple platform pods, with XAUTOCLAIM for reclaiming
+stale pending messages from dead consumers.
+"""
+
+import asyncio
+import logging
+import os
+import uuid
+from typing import Any
+
+import orjson
+from prisma import Prisma
+from pydantic import BaseModel
+from redis.exceptions import ResponseError
+
+from backend.data.redis_client import get_redis_async
+
+from . import stream_registry
+from .completion_handler import process_operation_failure, process_operation_success
+from .config import ChatConfig
+
+logger = logging.getLogger(__name__)
+config = ChatConfig()
+
+
+class OperationCompleteMessage(BaseModel):
+    """Message format for operation completion notifications."""
+
+    operation_id: str
+    task_id: str
+    success: bool
+    result: dict | str | None = None
+    error: str | None = None
+
+
+class ChatCompletionConsumer:
+    """Consumer for chat operation completion messages from Redis Streams.
+
+    This consumer initializes its own Prisma client in start() to ensure
+    database operations work correctly within this async context.
+
+    Uses Redis consumer groups to allow multiple platform pods to consume
+    messages reliably with automatic redelivery on failure.
+    """
+
+    def __init__(self):
+        self._consumer_task: asyncio.Task | None = None
+        self._running = False
+        self._prisma: Prisma | None = None
+        self._consumer_name = f"consumer-{uuid.uuid4().hex[:8]}"
+
+    async def start(self) -> None:
+        """Start the completion consumer."""
+        if self._running:
+            logger.warning("Completion consumer already running")
+            return
+
+        # Create consumer group if it doesn't exist
+        try:
+            redis = await get_redis_async()
+            await redis.xgroup_create(
+                config.stream_completion_name,
+                config.stream_consumer_group,
+                id="0",
+                mkstream=True,
+            )
+            logger.info(
+                f"Created consumer group '{config.stream_consumer_group}' "
+                f"on stream '{config.stream_completion_name}'"
+            )
+        except ResponseError as e:
+            if "BUSYGROUP" in str(e):
+                logger.debug(
+                    f"Consumer group '{config.stream_consumer_group}' already exists"
+                )
+            else:
+                raise
+
+        self._running = True
+        self._consumer_task = asyncio.create_task(self._consume_messages())
+        logger.info(
+            f"Chat completion consumer started (consumer: {self._consumer_name})"
+        )
+
+    async def _ensure_prisma(self) -> Prisma:
+        """Lazily initialize Prisma client on first use."""
+        if self._prisma is None:
+            database_url = os.getenv("DATABASE_URL", "postgresql://localhost:5432")
+            self._prisma = Prisma(datasource={"url": database_url})
+            await self._prisma.connect()
+            logger.info("[COMPLETION] Consumer Prisma client connected (lazy init)")
+        return self._prisma
+
+    async def stop(self) -> None:
+        """Stop the completion consumer."""
+        self._running = False
+
+        if self._consumer_task:
+            self._consumer_task.cancel()
+            try:
+                await self._consumer_task
+            except asyncio.CancelledError:
+                pass
+            self._consumer_task = None
+
+        if self._prisma:
+            await self._prisma.disconnect()
+            self._prisma = None
+            logger.info("[COMPLETION] Consumer Prisma client disconnected")
+
+        logger.info("Chat completion consumer stopped")
+
+    async def _consume_messages(self) -> None:
+        """Main message consumption loop with retry logic."""
+        max_retries = 10
+        retry_delay = 5  # seconds
+        retry_count = 0
+        block_timeout = 5000  # milliseconds
+
+        while self._running and retry_count < max_retries:
+            try:
+                redis = await get_redis_async()
+
+                # Reset retry count on successful connection
+                retry_count = 0
+
+                while self._running:
+                    # First, claim any stale pending messages from dead consumers
+                    # Redis does NOT auto-redeliver pending messages; we must explicitly
+                    # claim them using XAUTOCLAIM
+                    try:
+                        claimed_result = await redis.xautoclaim(
+                            name=config.stream_completion_name,
+                            groupname=config.stream_consumer_group,
+                            consumername=self._consumer_name,
+                            min_idle_time=config.stream_claim_min_idle_ms,
+                            start_id="0-0",
+                            count=10,
+                        )
+                        # xautoclaim returns: (next_start_id, [(id, data), ...], [deleted_ids])
+                        if claimed_result and len(claimed_result) >= 2:
+                            claimed_entries = claimed_result[1]
+                            if claimed_entries:
+                                logger.info(
+                                    f"Claimed {len(claimed_entries)} stale pending messages"
+                                )
+                                for entry_id, data in claimed_entries:
+                                    if not self._running:
+                                        return
+                                    await self._process_entry(redis, entry_id, data)
+                    except Exception as e:
+                        logger.warning(f"XAUTOCLAIM failed (non-fatal): {e}")
+
+                    # Read new messages from the stream
+                    messages = await redis.xreadgroup(
+                        groupname=config.stream_consumer_group,
+                        consumername=self._consumer_name,
+                        streams={config.stream_completion_name: ">"},
+                        block=block_timeout,
+                        count=10,
+                    )
+
+                    if not messages:
+                        continue
+
+                    for stream_name, entries in messages:
+                        for entry_id, data in entries:
+                            if not self._running:
+                                return
+                            await self._process_entry(redis, entry_id, data)
+
+            except asyncio.CancelledError:
+                logger.info("Consumer cancelled")
+                return
+            except Exception as e:
+                retry_count += 1
+                logger.error(
+                    f"Consumer error (retry {retry_count}/{max_retries}): {e}",
+                    exc_info=True,
+                )
+                if self._running and retry_count < max_retries:
+                    await asyncio.sleep(retry_delay)
+                else:
+                    logger.error("Max retries reached, stopping consumer")
+                    return
+
+    async def _process_entry(
+        self, redis: Any, entry_id: str, data: dict[str, Any]
+    ) -> None:
+        """Process a single stream entry and acknowledge it on success.
+
+        Args:
+            redis: Redis client connection
+            entry_id: The stream entry ID
+            data: The entry data dict
+        """
+        try:
+            # Handle the message
+            message_data = data.get("data")
+            if message_data:
+                await self._handle_message(
+                    message_data.encode()
+                    if isinstance(message_data, str)
+                    else message_data
+                )
+
+            # Acknowledge the message after successful processing
+            await redis.xack(
+                config.stream_completion_name,
+                config.stream_consumer_group,
+                entry_id,
+            )
+        except Exception as e:
+            logger.error(
+                f"Error processing completion message {entry_id}: {e}",
+                exc_info=True,
+            )
+            # Message remains in pending state and will be claimed by
+            # XAUTOCLAIM after min_idle_time expires
+
+    async def _handle_message(self, body: bytes) -> None:
+        """Handle a completion message using our own Prisma client."""
+        try:
+            data = orjson.loads(body)
+            message = OperationCompleteMessage(**data)
+        except Exception as e:
+            logger.error(f"Failed to parse completion message: {e}")
+            return
+
+        logger.info(
+            f"[COMPLETION] Received completion for operation {message.operation_id} "
+            f"(task_id={message.task_id}, success={message.success})"
+        )
+
+        # Find task in registry
+        task = await stream_registry.find_task_by_operation_id(message.operation_id)
+        if task is None:
+            task = await stream_registry.get_task(message.task_id)
+
+        if task is None:
+            logger.warning(
+                f"[COMPLETION] Task not found for operation {message.operation_id} "
+                f"(task_id={message.task_id})"
+            )
+            return
+
+        logger.info(
+            f"[COMPLETION] Found task: task_id={task.task_id}, "
+            f"session_id={task.session_id}, tool_call_id={task.tool_call_id}"
+        )
+
+        # Guard against empty task fields
+        if not task.task_id or not task.session_id or not task.tool_call_id:
+            logger.error(
+                f"[COMPLETION] Task has empty critical fields! "
+                f"task_id={task.task_id!r}, session_id={task.session_id!r}, "
+                f"tool_call_id={task.tool_call_id!r}"
+            )
+            return
+
+        if message.success:
+            await self._handle_success(task, message)
+        else:
+            await self._handle_failure(task, message)
+
+    async def _handle_success(
+        self,
+        task: stream_registry.ActiveTask,
+        message: OperationCompleteMessage,
+    ) -> None:
+        """Handle successful operation completion."""
+        prisma = await self._ensure_prisma()
+        await process_operation_success(task, message.result, prisma)
+
+    async def _handle_failure(
+        self,
+        task: stream_registry.ActiveTask,
+        message: OperationCompleteMessage,
+    ) -> None:
+        """Handle failed operation completion."""
+        prisma = await self._ensure_prisma()
+        await process_operation_failure(task, message.error, prisma)
+
+
+# Module-level consumer instance
+_consumer: ChatCompletionConsumer | None = None
+
+
+async def start_completion_consumer() -> None:
+    """Start the global completion consumer."""
+    global _consumer
+    if _consumer is None:
+        _consumer = ChatCompletionConsumer()
+    await _consumer.start()
+
+
+async def stop_completion_consumer() -> None:
+    """Stop the global completion consumer."""
+    global _consumer
+    if _consumer:
+        await _consumer.stop()
+        _consumer = None
+
+
+async def publish_operation_complete(
+    operation_id: str,
+    task_id: str,
+    success: bool,
+    result: dict | str | None = None,
+    error: str | None = None,
+) -> None:
+    """Publish an operation completion message to Redis Streams.
+
+    Args:
+        operation_id: The operation ID that completed.
+        task_id: The task ID associated with the operation.
+        success: Whether the operation succeeded.
+        result: The result data (for success).
+        error: The error message (for failure).
+    """
+    message = OperationCompleteMessage(
+        operation_id=operation_id,
+        task_id=task_id,
+        success=success,
+        result=result,
+        error=error,
+    )
+
+    redis = await get_redis_async()
+    await redis.xadd(
+        config.stream_completion_name,
+        {"data": message.model_dump_json()},
+        maxlen=config.stream_max_length,
+    )
+    logger.info(f"Published completion for operation {operation_id}")

autogpt_platform/backend/backend/api/features/chat/completion_handler.py

@@ -0,0 +1,344 @@
+"""Shared completion handling for operation success and failure.
+
+This module provides common logic for handling operation completion from both:
+- The Redis Streams consumer (completion_consumer.py)
+- The HTTP webhook endpoint (routes.py)
+"""
+
+import logging
+from typing import Any
+
+import orjson
+from prisma import Prisma
+
+from . import service as chat_service
+from . import stream_registry
+from .response_model import StreamError, StreamToolOutputAvailable
+from .tools.models import ErrorResponse
+
+logger = logging.getLogger(__name__)
+
+# Tools that produce agent_json that needs to be saved to library
+AGENT_GENERATION_TOOLS = {"create_agent", "edit_agent"}
+
+# Keys that should be stripped from agent_json when returning in error responses
+SENSITIVE_KEYS = frozenset(
+    {
+        "api_key",
+        "apikey",
+        "api_secret",
+        "password",
+        "secret",
+        "credentials",
+        "credential",
+        "token",
+        "access_token",
+        "refresh_token",
+        "private_key",
+        "privatekey",
+        "auth",
+        "authorization",
+    }
+)
+
+
+def _sanitize_agent_json(obj: Any) -> Any:
+    """Recursively sanitize agent_json by removing sensitive keys.
+
+    Args:
+        obj: The object to sanitize (dict, list, or primitive)
+
+    Returns:
+        Sanitized copy with sensitive keys removed/redacted
+    """
+    if isinstance(obj, dict):
+        return {
+            k: "[REDACTED]" if k.lower() in SENSITIVE_KEYS else _sanitize_agent_json(v)
+            for k, v in obj.items()
+        }
+    elif isinstance(obj, list):
+        return [_sanitize_agent_json(item) for item in obj]
+    else:
+        return obj
+
+
+class ToolMessageUpdateError(Exception):
+    """Raised when updating a tool message in the database fails."""
+
+    pass
+
+
+async def _update_tool_message(
+    session_id: str,
+    tool_call_id: str,
+    content: str,
+    prisma_client: Prisma | None,
+) -> None:
+    """Update tool message in database.
+
+    Args:
+        session_id: The session ID
+        tool_call_id: The tool call ID to update
+        content: The new content for the message
+        prisma_client: Optional Prisma client. If None, uses chat_service.
+
+    Raises:
+        ToolMessageUpdateError: If the database update fails. The caller should
+            handle this to avoid marking the task as completed with inconsistent state.
+    """
+    try:
+        if prisma_client:
+            # Use provided Prisma client (for consumer with its own connection)
+            updated_count = await prisma_client.chatmessage.update_many(
+                where={
+                    "sessionId": session_id,
+                    "toolCallId": tool_call_id,
+                },
+                data={"content": content},
+            )
+            # Check if any rows were updated - 0 means message not found
+            if updated_count == 0:
+                raise ToolMessageUpdateError(
+                    f"No message found with tool_call_id={tool_call_id} in session {session_id}"
+                )
+        else:
+            # Use service function (for webhook endpoint)
+            await chat_service._update_pending_operation(
+                session_id=session_id,
+                tool_call_id=tool_call_id,
+                result=content,
+            )
+    except ToolMessageUpdateError:
+        raise
+    except Exception as e:
+        logger.error(f"[COMPLETION] Failed to update tool message: {e}", exc_info=True)
+        raise ToolMessageUpdateError(
+            f"Failed to update tool message for tool_call_id={tool_call_id}: {e}"
+        ) from e
+
+
+def serialize_result(result: dict | list | str | int | float | bool | None) -> str:
+    """Serialize result to JSON string with sensible defaults.
+
+    Args:
+        result: The result to serialize. Can be a dict, list, string,
+            number, boolean, or None.
+
+    Returns:
+        JSON string representation of the result. Returns '{"status": "completed"}'
+        only when result is explicitly None.
+    """
+    if isinstance(result, str):
+        return result
+    if result is None:
+        return '{"status": "completed"}'
+    return orjson.dumps(result).decode("utf-8")
+
+
+async def _save_agent_from_result(
+    result: dict[str, Any],
+    user_id: str | None,
+    tool_name: str,
+) -> dict[str, Any]:
+    """Save agent to library if result contains agent_json.
+
+    Args:
+        result: The result dict that may contain agent_json
+        user_id: The user ID to save the agent for
+        tool_name: The tool name (create_agent or edit_agent)
+
+    Returns:
+        Updated result dict with saved agent details, or original result if no agent_json
+    """
+    if not user_id:
+        logger.warning("[COMPLETION] Cannot save agent: no user_id in task")
+        return result
+
+    agent_json = result.get("agent_json")
+    if not agent_json:
+        logger.warning(
+            f"[COMPLETION] {tool_name} completed but no agent_json in result"
+        )
+        return result
+
+    try:
+        from .tools.agent_generator import save_agent_to_library
+
+        is_update = tool_name == "edit_agent"
+        created_graph, library_agent = await save_agent_to_library(
+            agent_json, user_id, is_update=is_update
+        )
+
+        logger.info(
+            f"[COMPLETION] Saved agent '{created_graph.name}' to library "
+            f"(graph_id={created_graph.id}, library_agent_id={library_agent.id})"
+        )
+
+        # Return a response similar to AgentSavedResponse
+        return {
+            "type": "agent_saved",
+            "message": f"Agent '{created_graph.name}' has been saved to your library!",
+            "agent_id": created_graph.id,
+            "agent_name": created_graph.name,
+            "library_agent_id": library_agent.id,
+            "library_agent_link": f"/library/agents/{library_agent.id}",
+            "agent_page_link": f"/build?flowID={created_graph.id}",
+        }
+    except Exception as e:
+        logger.error(
+            f"[COMPLETION] Failed to save agent to library: {e}",
+            exc_info=True,
+        )
+        # Return error but don't fail the whole operation
+        # Sanitize agent_json to remove sensitive keys before returning
+        return {
+            "type": "error",
+            "message": f"Agent was generated but failed to save: {str(e)}",
+            "error": str(e),
+            "agent_json": _sanitize_agent_json(agent_json),
+        }
+
+
+async def process_operation_success(
+    task: stream_registry.ActiveTask,
+    result: dict | str | None,
+    prisma_client: Prisma | None = None,
+) -> None:
+    """Handle successful operation completion.
+
+    Publishes the result to the stream registry, updates the database,
+    generates LLM continuation, and marks the task as completed.
+
+    Args:
+        task: The active task that completed
+        result: The result data from the operation
+        prisma_client: Optional Prisma client for database operations.
+            If None, uses chat_service._update_pending_operation instead.
+
+    Raises:
+        ToolMessageUpdateError: If the database update fails. The task will be
+            marked as failed instead of completed to avoid inconsistent state.
+    """
+    # For agent generation tools, save the agent to library
+    if task.tool_name in AGENT_GENERATION_TOOLS and isinstance(result, dict):
+        result = await _save_agent_from_result(result, task.user_id, task.tool_name)
+
+    # Serialize result for output (only substitute default when result is exactly None)
+    result_output = result if result is not None else {"status": "completed"}
+    output_str = (
+        result_output
+        if isinstance(result_output, str)
+        else orjson.dumps(result_output).decode("utf-8")
+    )
+
+    # Publish result to stream registry
+    await stream_registry.publish_chunk(
+        task.task_id,
+        StreamToolOutputAvailable(
+            toolCallId=task.tool_call_id,
+            toolName=task.tool_name,
+            output=output_str,
+            success=True,
+        ),
+    )
+
+    # Update pending operation in database
+    # If this fails, we must not continue to mark the task as completed
+    result_str = serialize_result(result)
+    try:
+        await _update_tool_message(
+            session_id=task.session_id,
+            tool_call_id=task.tool_call_id,
+            content=result_str,
+            prisma_client=prisma_client,
+        )
+    except ToolMessageUpdateError:
+        # DB update failed - mark task as failed to avoid inconsistent state
+        logger.error(
+            f"[COMPLETION] DB update failed for task {task.task_id}, "
+            "marking as failed instead of completed"
+        )
+        await stream_registry.publish_chunk(
+            task.task_id,
+            StreamError(errorText="Failed to save operation result to database"),
+        )
+        await stream_registry.mark_task_completed(task.task_id, status="failed")
+        raise
+
+    # Generate LLM continuation with streaming
+    try:
+        await chat_service._generate_llm_continuation_with_streaming(
+            session_id=task.session_id,
+            user_id=task.user_id,
+            task_id=task.task_id,
+        )
+    except Exception as e:
+        logger.error(
+            f"[COMPLETION] Failed to generate LLM continuation: {e}",
+            exc_info=True,
+        )
+
+    # Mark task as completed and release Redis lock
+    await stream_registry.mark_task_completed(task.task_id, status="completed")
+    try:
+        await chat_service._mark_operation_completed(task.tool_call_id)
+    except Exception as e:
+        logger.error(f"[COMPLETION] Failed to mark operation completed: {e}")
+
+    logger.info(
+        f"[COMPLETION] Successfully processed completion for task {task.task_id}"
+    )
+
+
+async def process_operation_failure(
+    task: stream_registry.ActiveTask,
+    error: str | None,
+    prisma_client: Prisma | None = None,
+) -> None:
+    """Handle failed operation completion.
+
+    Publishes the error to the stream registry, updates the database with
+    the error response, and marks the task as failed.
+
+    Args:
+        task: The active task that failed
+        error: The error message from the operation
+        prisma_client: Optional Prisma client for database operations.
+            If None, uses chat_service._update_pending_operation instead.
+    """
+    error_msg = error or "Operation failed"
+
+    # Publish error to stream registry
+    await stream_registry.publish_chunk(
+        task.task_id,
+        StreamError(errorText=error_msg),
+    )
+
+    # Update pending operation with error
+    # If this fails, we still continue to mark the task as failed
+    error_response = ErrorResponse(
+        message=error_msg,
+        error=error,
+    )
+    try:
+        await _update_tool_message(
+            session_id=task.session_id,
+            tool_call_id=task.tool_call_id,
+            content=error_response.model_dump_json(),
+            prisma_client=prisma_client,
+        )
+    except ToolMessageUpdateError:
+        # DB update failed - log but continue with cleanup
+        logger.error(
+            f"[COMPLETION] DB update failed while processing failure for task {task.task_id}, "
+            "continuing with cleanup"
+        )
+
+    # Mark task as failed and release Redis lock
+    await stream_registry.mark_task_completed(task.task_id, status="failed")
+    try:
+        await chat_service._mark_operation_completed(task.tool_call_id)
+    except Exception as e:
+        logger.error(f"[COMPLETION] Failed to mark operation completed: {e}")
+
+    logger.info(f"[COMPLETION] Processed failure for task {task.task_id}: {error_msg}")

autogpt_platform/backend/backend/api/features/chat/config.py

@@ -0,0 +1,187 @@
+"""Configuration management for chat system."""
+
+import os
+
+from pydantic import Field, field_validator
+from pydantic_settings import BaseSettings
+
+
+class ChatConfig(BaseSettings):
+    """Configuration for the chat system."""
+
+    # OpenAI API Configuration
+    model: str = Field(
+        default="anthropic/claude-opus-4.6", description="Default model to use"
+    )
+    title_model: str = Field(
+        default="openai/gpt-4o-mini",
+        description="Model to use for generating session titles (should be fast/cheap)",
+    )
+    api_key: str | None = Field(default=None, description="OpenAI API key")
+    base_url: str | None = Field(
+        default="https://openrouter.ai/api/v1",
+        description="Base URL for API (e.g., for OpenRouter)",
+    )
+
+    # Session TTL Configuration - 12 hours
+    session_ttl: int = Field(default=43200, description="Session TTL in seconds")
+
+    # Streaming Configuration
+    stream_timeout: int = Field(default=300, description="Stream timeout in seconds")
+    max_retries: int = Field(
+        default=3,
+        description="Max retries for fallback path (SDK handles retries internally)",
+    )
+    max_agent_runs: int = Field(default=30, description="Maximum number of agent runs")
+    max_agent_schedules: int = Field(
+        default=30, description="Maximum number of agent schedules"
+    )
+
+    # Long-running operation configuration
+    long_running_operation_ttl: int = Field(
+        default=600,
+        description="TTL in seconds for long-running operation tracking in Redis (safety net if pod dies)",
+    )
+
+    # Stream registry configuration for SSE reconnection
+    stream_ttl: int = Field(
+        default=3600,
+        description="TTL in seconds for stream data in Redis (1 hour)",
+    )
+    stream_max_length: int = Field(
+        default=10000,
+        description="Maximum number of messages to store per stream",
+    )
+
+    # Redis Streams configuration for completion consumer
+    stream_completion_name: str = Field(
+        default="chat:completions",
+        description="Redis Stream name for operation completions",
+    )
+    stream_consumer_group: str = Field(
+        default="chat_consumers",
+        description="Consumer group name for completion stream",
+    )
+    stream_claim_min_idle_ms: int = Field(
+        default=60000,
+        description="Minimum idle time in milliseconds before claiming pending messages from dead consumers",
+    )
+
+    # Redis key prefixes for stream registry
+    task_meta_prefix: str = Field(
+        default="chat:task:meta:",
+        description="Prefix for task metadata hash keys",
+    )
+    task_stream_prefix: str = Field(
+        default="chat:stream:",
+        description="Prefix for task message stream keys",
+    )
+    task_op_prefix: str = Field(
+        default="chat:task:op:",
+        description="Prefix for operation ID to task ID mapping keys",
+    )
+    internal_api_key: str | None = Field(
+        default=None,
+        description="API key for internal webhook callbacks (env: CHAT_INTERNAL_API_KEY)",
+    )
+
+    # Langfuse Prompt Management Configuration
+    # Note: Langfuse credentials are in Settings().secrets (settings.py)
+    langfuse_prompt_name: str = Field(
+        default="CoPilot Prompt",
+        description="Name of the prompt in Langfuse to fetch",
+    )
+
+    # Claude Agent SDK Configuration
+    use_claude_agent_sdk: bool = Field(
+        default=True,
+        description="Use Claude Agent SDK for chat completions",
+    )
+    claude_agent_model: str | None = Field(
+        default=None,
+        description="Model for the Claude Agent SDK path. If None, derives from "
+        "the `model` field by stripping the OpenRouter provider prefix.",
+    )
+    claude_agent_max_buffer_size: int = Field(
+        default=10 * 1024 * 1024,  # 10MB (default SDK is 1MB)
+        description="Max buffer size in bytes for Claude Agent SDK JSON message parsing. "
+        "Increase if tool outputs exceed the limit.",
+    )
+    claude_agent_max_subtasks: int = Field(
+        default=10,
+        description="Max number of sub-agent Tasks the SDK can spawn per session.",
+    )
+    claude_agent_use_resume: bool = Field(
+        default=True,
+        description="Use --resume for multi-turn conversations instead of "
+        "history compression. Falls back to compression when unavailable.",
+    )
+
+    # Extended thinking configuration for Claude models
+    thinking_enabled: bool = Field(
+        default=True,
+        description="Enable adaptive thinking for Claude models via OpenRouter",
+    )
+
+    @field_validator("api_key", mode="before")
+    @classmethod
+    def get_api_key(cls, v):
+        """Get API key from environment if not provided."""
+        if v is None:
+            # Try to get from environment variables
+            # First check for CHAT_API_KEY (Pydantic prefix)
+            v = os.getenv("CHAT_API_KEY")
+            if not v:
+                # Fall back to OPEN_ROUTER_API_KEY
+                v = os.getenv("OPEN_ROUTER_API_KEY")
+            if not v:
+                # Fall back to OPENAI_API_KEY
+                v = os.getenv("OPENAI_API_KEY")
+        return v
+
+    @field_validator("base_url", mode="before")
+    @classmethod
+    def get_base_url(cls, v):
+        """Get base URL from environment if not provided."""
+        if v is None:
+            # Check for OpenRouter or custom base URL
+            v = os.getenv("CHAT_BASE_URL")
+            if not v:
+                v = os.getenv("OPENROUTER_BASE_URL")
+            if not v:
+                v = os.getenv("OPENAI_BASE_URL")
+            if not v:
+                v = "https://openrouter.ai/api/v1"
+        return v
+
+    @field_validator("internal_api_key", mode="before")
+    @classmethod
+    def get_internal_api_key(cls, v):
+        """Get internal API key from environment if not provided."""
+        if v is None:
+            v = os.getenv("CHAT_INTERNAL_API_KEY")
+        return v
+
+    @field_validator("use_claude_agent_sdk", mode="before")
+    @classmethod
+    def get_use_claude_agent_sdk(cls, v):
+        """Get use_claude_agent_sdk from environment if not provided."""
+        # Check environment variable - default to True if not set
+        env_val = os.getenv("CHAT_USE_CLAUDE_AGENT_SDK", "").lower()
+        if env_val:
+            return env_val in ("true", "1", "yes", "on")
+        # Default to True (SDK enabled by default)
+        return True if v is None else v
+
+    # Prompt paths for different contexts
+    PROMPT_PATHS: dict[str, str] = {
+        "default": "prompts/chat_system.md",
+        "onboarding": "prompts/onboarding_system.md",
+    }
+
+    class Config:
+        """Pydantic config."""
+
+        env_file = ".env"
+        env_file_encoding = "utf-8"
+        extra = "ignore"  # Ignore extra environment variables

autogpt_platform/backend/backend/api/features/chat/db.py

@@ -0,0 +1,288 @@
+"""Database operations for chat sessions."""
+
+import asyncio
+import logging
+from datetime import UTC, datetime
+from typing import Any, cast
+
+from prisma.models import ChatMessage as PrismaChatMessage
+from prisma.models import ChatSession as PrismaChatSession
+from prisma.types import (
+    ChatMessageCreateInput,
+    ChatSessionCreateInput,
+    ChatSessionUpdateInput,
+    ChatSessionWhereInput,
+)
+
+from backend.data.db import transaction
+from backend.util.json import SafeJson
+
+logger = logging.getLogger(__name__)
+
+
+async def get_chat_session(session_id: str) -> PrismaChatSession | None:
+    """Get a chat session by ID from the database."""
+    session = await PrismaChatSession.prisma().find_unique(
+        where={"id": session_id},
+        include={"Messages": True},
+    )
+    if session and session.Messages:
+        # Sort messages by sequence in Python - Prisma Python client doesn't support
+        # order_by in include clauses (unlike Prisma JS), so we sort after fetching
+        session.Messages.sort(key=lambda m: m.sequence)
+    return session
+
+
+async def create_chat_session(
+    session_id: str,
+    user_id: str,
+) -> PrismaChatSession:
+    """Create a new chat session in the database."""
+    data = ChatSessionCreateInput(
+        id=session_id,
+        userId=user_id,
+        credentials=SafeJson({}),
+        successfulAgentRuns=SafeJson({}),
+        successfulAgentSchedules=SafeJson({}),
+    )
+    return await PrismaChatSession.prisma().create(data=data)
+
+
+async def update_chat_session(
+    session_id: str,
+    credentials: dict[str, Any] | None = None,
+    successful_agent_runs: dict[str, Any] | None = None,
+    successful_agent_schedules: dict[str, Any] | None = None,
+    total_prompt_tokens: int | None = None,
+    total_completion_tokens: int | None = None,
+    title: str | None = None,
+) -> PrismaChatSession | None:
+    """Update a chat session's metadata."""
+    data: ChatSessionUpdateInput = {"updatedAt": datetime.now(UTC)}
+
+    if credentials is not None:
+        data["credentials"] = SafeJson(credentials)
+    if successful_agent_runs is not None:
+        data["successfulAgentRuns"] = SafeJson(successful_agent_runs)
+    if successful_agent_schedules is not None:
+        data["successfulAgentSchedules"] = SafeJson(successful_agent_schedules)
+    if total_prompt_tokens is not None:
+        data["totalPromptTokens"] = total_prompt_tokens
+    if total_completion_tokens is not None:
+        data["totalCompletionTokens"] = total_completion_tokens
+    if title is not None:
+        data["title"] = title
+
+    session = await PrismaChatSession.prisma().update(
+        where={"id": session_id},
+        data=data,
+        include={"Messages": True},
+    )
+    if session and session.Messages:
+        # Sort in Python - Prisma Python doesn't support order_by in include clauses
+        session.Messages.sort(key=lambda m: m.sequence)
+    return session
+
+
+async def add_chat_message(
+    session_id: str,
+    role: str,
+    sequence: int,
+    content: str | None = None,
+    name: str | None = None,
+    tool_call_id: str | None = None,
+    refusal: str | None = None,
+    tool_calls: list[dict[str, Any]] | None = None,
+    function_call: dict[str, Any] | None = None,
+) -> PrismaChatMessage:
+    """Add a message to a chat session."""
+    # Build input dict dynamically rather than using ChatMessageCreateInput directly
+    # because Prisma's TypedDict validation rejects optional fields set to None.
+    # We only include fields that have values, then cast at the end.
+    data: dict[str, Any] = {
+        "Session": {"connect": {"id": session_id}},
+        "role": role,
+        "sequence": sequence,
+    }
+
+    # Add optional string fields
+    if content is not None:
+        data["content"] = content
+    if name is not None:
+        data["name"] = name
+    if tool_call_id is not None:
+        data["toolCallId"] = tool_call_id
+    if refusal is not None:
+        data["refusal"] = refusal
+
+    # Add optional JSON fields only when they have values
+    if tool_calls is not None:
+        data["toolCalls"] = SafeJson(tool_calls)
+    if function_call is not None:
+        data["functionCall"] = SafeJson(function_call)
+
+    # Run message create and session timestamp update in parallel for lower latency
+    _, message = await asyncio.gather(
+        PrismaChatSession.prisma().update(
+            where={"id": session_id},
+            data={"updatedAt": datetime.now(UTC)},
+        ),
+        PrismaChatMessage.prisma().create(data=cast(ChatMessageCreateInput, data)),
+    )
+    return message
+
+
+async def add_chat_messages_batch(
+    session_id: str,
+    messages: list[dict[str, Any]],
+    start_sequence: int,
+) -> list[PrismaChatMessage]:
+    """Add multiple messages to a chat session in a batch.
+
+    Uses a transaction for atomicity - if any message creation fails,
+    the entire batch is rolled back.
+    """
+    if not messages:
+        return []
+
+    created_messages = []
+
+    async with transaction() as tx:
+        for i, msg in enumerate(messages):
+            # Build input dict dynamically rather than using ChatMessageCreateInput
+            # directly because Prisma's TypedDict validation rejects optional fields
+            # set to None. We only include fields that have values, then cast.
+            data: dict[str, Any] = {
+                "Session": {"connect": {"id": session_id}},
+                "role": msg["role"],
+                "sequence": start_sequence + i,
+            }
+
+            # Add optional string fields
+            if msg.get("content") is not None:
+                data["content"] = msg["content"]
+            if msg.get("name") is not None:
+                data["name"] = msg["name"]
+            if msg.get("tool_call_id") is not None:
+                data["toolCallId"] = msg["tool_call_id"]
+            if msg.get("refusal") is not None:
+                data["refusal"] = msg["refusal"]
+
+            # Add optional JSON fields only when they have values
+            if msg.get("tool_calls") is not None:
+                data["toolCalls"] = SafeJson(msg["tool_calls"])
+            if msg.get("function_call") is not None:
+                data["functionCall"] = SafeJson(msg["function_call"])
+
+            created = await PrismaChatMessage.prisma(tx).create(
+                data=cast(ChatMessageCreateInput, data)
+            )
+            created_messages.append(created)
+
+        # Update session's updatedAt timestamp within the same transaction.
+        # Note: Token usage (total_prompt_tokens, total_completion_tokens) is updated
+        # separately via update_chat_session() after streaming completes.
+        await PrismaChatSession.prisma(tx).update(
+            where={"id": session_id},
+            data={"updatedAt": datetime.now(UTC)},
+        )
+
+    return created_messages
+
+
+async def get_user_chat_sessions(
+    user_id: str,
+    limit: int = 50,
+    offset: int = 0,
+) -> list[PrismaChatSession]:
+    """Get chat sessions for a user, ordered by most recent."""
+    return await PrismaChatSession.prisma().find_many(
+        where={"userId": user_id},
+        order={"updatedAt": "desc"},
+        take=limit,
+        skip=offset,
+    )
+
+
+async def get_user_session_count(user_id: str) -> int:
+    """Get the total number of chat sessions for a user."""
+    return await PrismaChatSession.prisma().count(where={"userId": user_id})
+
+
+async def delete_chat_session(session_id: str, user_id: str | None = None) -> bool:
+    """Delete a chat session and all its messages.
+
+    Args:
+        session_id: The session ID to delete.
+        user_id: If provided, validates that the session belongs to this user
+            before deletion. This prevents unauthorized deletion of other
+            users' sessions.
+
+    Returns:
+        True if deleted successfully, False otherwise.
+    """
+    try:
+        # Build typed where clause with optional user_id validation
+        where_clause: ChatSessionWhereInput = {"id": session_id}
+        if user_id is not None:
+            where_clause["userId"] = user_id
+
+        result = await PrismaChatSession.prisma().delete_many(where=where_clause)
+        if result == 0:
+            logger.warning(
+                f"No session deleted for {session_id} "
+                f"(user_id validation: {user_id is not None})"
+            )
+            return False
+        return True
+    except Exception as e:
+        logger.error(f"Failed to delete chat session {session_id}: {e}")
+        return False
+
+
+async def get_chat_session_message_count(session_id: str) -> int:
+    """Get the number of messages in a chat session."""
+    count = await PrismaChatMessage.prisma().count(where={"sessionId": session_id})
+    return count
+
+
+async def update_tool_message_content(
+    session_id: str,
+    tool_call_id: str,
+    new_content: str,
+) -> bool:
+    """Update the content of a tool message in chat history.
+
+    Used by background tasks to update pending operation messages with final results.
+
+    Args:
+        session_id: The chat session ID.
+        tool_call_id: The tool call ID to find the message.
+        new_content: The new content to set.
+
+    Returns:
+        True if a message was updated, False otherwise.
+    """
+    try:
+        result = await PrismaChatMessage.prisma().update_many(
+            where={
+                "sessionId": session_id,
+                "toolCallId": tool_call_id,
+            },
+            data={
+                "content": new_content,
+            },
+        )
+        if result == 0:
+            logger.warning(
+                f"No message found to update for session {session_id}, "
+                f"tool_call_id {tool_call_id}"
+            )
+            return False
+        return True
+    except Exception as e:
+        logger.error(
+            f"Failed to update tool message for session {session_id}, "
+            f"tool_call_id {tool_call_id}: {e}"
+        )
+        return False

autogpt_platform/backend/backend/api/features/chat/model.py

@@ -0,0 +1,713 @@
+import asyncio
+import logging
+import uuid
+from datetime import UTC, datetime
+from typing import Any, cast
+from weakref import WeakValueDictionary
+
+from openai.types.chat import (
+    ChatCompletionAssistantMessageParam,
+    ChatCompletionDeveloperMessageParam,
+    ChatCompletionFunctionMessageParam,
+    ChatCompletionMessageParam,
+    ChatCompletionSystemMessageParam,
+    ChatCompletionToolMessageParam,
+    ChatCompletionUserMessageParam,
+)
+from openai.types.chat.chat_completion_assistant_message_param import FunctionCall
+from openai.types.chat.chat_completion_message_tool_call_param import (
+    ChatCompletionMessageToolCallParam,
+    Function,
+)
+from prisma.models import ChatMessage as PrismaChatMessage
+from prisma.models import ChatSession as PrismaChatSession
+from pydantic import BaseModel
+
+from backend.data.redis_client import get_redis_async
+from backend.util import json
+from backend.util.exceptions import DatabaseError, RedisError
+
+from . import db as chat_db
+from .config import ChatConfig
+
+logger = logging.getLogger(__name__)
+config = ChatConfig()
+
+
+def _parse_json_field(value: str | dict | list | None, default: Any = None) -> Any:
+    """Parse a JSON field that may be stored as string or already parsed."""
+    if value is None:
+        return default
+    if isinstance(value, str):
+        return json.loads(value)
+    return value
+
+
+# Redis cache key prefix for chat sessions
+CHAT_SESSION_CACHE_PREFIX = "chat:session:"
+
+
+def _get_session_cache_key(session_id: str) -> str:
+    """Get the Redis cache key for a chat session."""
+    return f"{CHAT_SESSION_CACHE_PREFIX}{session_id}"
+
+
+# Session-level locks to prevent race conditions during concurrent upserts.
+# Uses WeakValueDictionary to automatically garbage collect locks when no longer referenced,
+# preventing unbounded memory growth while maintaining lock semantics for active sessions.
+# Invalidation: Locks are auto-removed by GC when no coroutine holds a reference (after
+# async with lock: completes). Explicit cleanup also occurs in delete_chat_session().
+_session_locks: WeakValueDictionary[str, asyncio.Lock] = WeakValueDictionary()
+_session_locks_mutex = asyncio.Lock()
+
+
+async def _get_session_lock(session_id: str) -> asyncio.Lock:
+    """Get or create a lock for a specific session to prevent concurrent upserts.
+
+    Uses WeakValueDictionary for automatic cleanup: locks are garbage collected
+    when no coroutine holds a reference to them, preventing memory leaks from
+    unbounded growth of session locks.
+    """
+    async with _session_locks_mutex:
+        lock = _session_locks.get(session_id)
+        if lock is None:
+            lock = asyncio.Lock()
+            _session_locks[session_id] = lock
+        return lock
+
+
+class ChatMessage(BaseModel):
+    role: str
+    content: str | None = None
+    name: str | None = None
+    tool_call_id: str | None = None
+    refusal: str | None = None
+    tool_calls: list[dict] | None = None
+    function_call: dict | None = None
+
+
+class Usage(BaseModel):
+    prompt_tokens: int
+    completion_tokens: int
+    total_tokens: int
+
+
+class ChatSession(BaseModel):
+    session_id: str
+    user_id: str
+    title: str | None = None
+    messages: list[ChatMessage]
+    usage: list[Usage]
+    credentials: dict[str, dict] = {}  # Map of provider -> credential metadata
+    started_at: datetime
+    updated_at: datetime
+    successful_agent_runs: dict[str, int] = {}
+    successful_agent_schedules: dict[str, int] = {}
+
+    def add_tool_call_to_current_turn(self, tool_call: dict) -> None:
+        """Attach a tool_call to the current turn's assistant message.
+
+        Searches backwards for the most recent assistant message (stopping at
+        any user message boundary). If found, appends the tool_call to it.
+        Otherwise creates a new assistant message with the tool_call.
+        """
+        for msg in reversed(self.messages):
+            if msg.role == "user":
+                break
+            if msg.role == "assistant":
+                if not msg.tool_calls:
+                    msg.tool_calls = []
+                msg.tool_calls.append(tool_call)
+                return
+
+        self.messages.append(
+            ChatMessage(role="assistant", content="", tool_calls=[tool_call])
+        )
+
+    @staticmethod
+    def new(user_id: str) -> "ChatSession":
+        return ChatSession(
+            session_id=str(uuid.uuid4()),
+            user_id=user_id,
+            title=None,
+            messages=[],
+            usage=[],
+            credentials={},
+            started_at=datetime.now(UTC),
+            updated_at=datetime.now(UTC),
+        )
+
+    @staticmethod
+    def from_db(
+        prisma_session: PrismaChatSession,
+        prisma_messages: list[PrismaChatMessage] | None = None,
+    ) -> "ChatSession":
+        """Convert Prisma models to Pydantic ChatSession."""
+        messages = []
+        if prisma_messages:
+            for msg in prisma_messages:
+                messages.append(
+                    ChatMessage(
+                        role=msg.role,
+                        content=msg.content,
+                        name=msg.name,
+                        tool_call_id=msg.toolCallId,
+                        refusal=msg.refusal,
+                        tool_calls=_parse_json_field(msg.toolCalls),
+                        function_call=_parse_json_field(msg.functionCall),
+                    )
+                )
+
+        # Parse JSON fields from Prisma
+        credentials = _parse_json_field(prisma_session.credentials, default={})
+        successful_agent_runs = _parse_json_field(
+            prisma_session.successfulAgentRuns, default={}
+        )
+        successful_agent_schedules = _parse_json_field(
+            prisma_session.successfulAgentSchedules, default={}
+        )
+
+        # Calculate usage from token counts
+        usage = []
+        if prisma_session.totalPromptTokens or prisma_session.totalCompletionTokens:
+            usage.append(
+                Usage(
+                    prompt_tokens=prisma_session.totalPromptTokens or 0,
+                    completion_tokens=prisma_session.totalCompletionTokens or 0,
+                    total_tokens=(prisma_session.totalPromptTokens or 0)
+                    + (prisma_session.totalCompletionTokens or 0),
+                )
+            )
+
+        return ChatSession(
+            session_id=prisma_session.id,
+            user_id=prisma_session.userId,
+            title=prisma_session.title,
+            messages=messages,
+            usage=usage,
+            credentials=credentials,
+            started_at=prisma_session.createdAt,
+            updated_at=prisma_session.updatedAt,
+            successful_agent_runs=successful_agent_runs,
+            successful_agent_schedules=successful_agent_schedules,
+        )
+
+    @staticmethod
+    def _merge_consecutive_assistant_messages(
+        messages: list[ChatCompletionMessageParam],
+    ) -> list[ChatCompletionMessageParam]:
+        """Merge consecutive assistant messages into single messages.
+
+        Long-running tool flows can create split assistant messages: one with
+        text content and another with tool_calls. Anthropic's API requires
+        tool_result blocks to reference a tool_use in the immediately preceding
+        assistant message, so these splits cause 400 errors via OpenRouter.
+        """
+        if len(messages) < 2:
+            return messages
+
+        result: list[ChatCompletionMessageParam] = [messages[0]]
+        for msg in messages[1:]:
+            prev = result[-1]
+            if prev.get("role") != "assistant" or msg.get("role") != "assistant":
+                result.append(msg)
+                continue
+
+            prev = cast(ChatCompletionAssistantMessageParam, prev)
+            curr = cast(ChatCompletionAssistantMessageParam, msg)
+
+            curr_content = curr.get("content") or ""
+            if curr_content:
+                prev_content = prev.get("content") or ""
+                prev["content"] = (
+                    f"{prev_content}\n{curr_content}" if prev_content else curr_content
+                )
+
+            curr_tool_calls = curr.get("tool_calls")
+            if curr_tool_calls:
+                prev_tool_calls = prev.get("tool_calls")
+                prev["tool_calls"] = (
+                    list(prev_tool_calls) + list(curr_tool_calls)
+                    if prev_tool_calls
+                    else list(curr_tool_calls)
+                )
+        return result
+
+    def to_openai_messages(self) -> list[ChatCompletionMessageParam]:
+        messages = []
+        for message in self.messages:
+            if message.role == "developer":
+                m = ChatCompletionDeveloperMessageParam(
+                    role="developer",
+                    content=message.content or "",
+                )
+                if message.name:
+                    m["name"] = message.name
+                messages.append(m)
+            elif message.role == "system":
+                m = ChatCompletionSystemMessageParam(
+                    role="system",
+                    content=message.content or "",
+                )
+                if message.name:
+                    m["name"] = message.name
+                messages.append(m)
+            elif message.role == "user":
+                m = ChatCompletionUserMessageParam(
+                    role="user",
+                    content=message.content or "",
+                )
+                if message.name:
+                    m["name"] = message.name
+                messages.append(m)
+            elif message.role == "assistant":
+                m = ChatCompletionAssistantMessageParam(
+                    role="assistant",
+                    content=message.content or "",
+                )
+                if message.function_call:
+                    m["function_call"] = FunctionCall(
+                        arguments=message.function_call["arguments"],
+                        name=message.function_call["name"],
+                    )
+                if message.refusal:
+                    m["refusal"] = message.refusal
+                if message.tool_calls:
+                    t: list[ChatCompletionMessageToolCallParam] = []
+                    for tool_call in message.tool_calls:
+                        # Tool calls are stored with nested structure: {id, type, function: {name, arguments}}
+                        function_data = tool_call.get("function", {})
+
+                        # Skip tool calls that are missing required fields
+                        if "id" not in tool_call or "name" not in function_data:
+                            logger.warning(
+                                f"Skipping invalid tool call: missing required fields. "
+                                f"Got: {tool_call.keys()}, function keys: {function_data.keys()}"
+                            )
+                            continue
+
+                        # Arguments are stored as a JSON string
+                        arguments_str = function_data.get("arguments", "{}")
+
+                        t.append(
+                            ChatCompletionMessageToolCallParam(
+                                id=tool_call["id"],
+                                type="function",
+                                function=Function(
+                                    arguments=arguments_str,
+                                    name=function_data["name"],
+                                ),
+                            )
+                        )
+                    m["tool_calls"] = t
+                if message.name:
+                    m["name"] = message.name
+                messages.append(m)
+            elif message.role == "tool":
+                messages.append(
+                    ChatCompletionToolMessageParam(
+                        role="tool",
+                        content=message.content or "",
+                        tool_call_id=message.tool_call_id or "",
+                    )
+                )
+            elif message.role == "function":
+                messages.append(
+                    ChatCompletionFunctionMessageParam(
+                        role="function",
+                        content=message.content,
+                        name=message.name or "",
+                    )
+                )
+        return self._merge_consecutive_assistant_messages(messages)
+
+
+async def _get_session_from_cache(session_id: str) -> ChatSession | None:
+    """Get a chat session from Redis cache."""
+    redis_key = _get_session_cache_key(session_id)
+    async_redis = await get_redis_async()
+    raw_session: bytes | None = await async_redis.get(redis_key)
+
+    if raw_session is None:
+        return None
+
+    try:
+        session = ChatSession.model_validate_json(raw_session)
+        logger.info(
+            f"[CACHE] Loaded session {session_id}: {len(session.messages)} messages, "
+            f"last_roles={[m.role for m in session.messages[-3:]]}"  # Last 3 roles
+        )
+        return session
+    except Exception as e:
+        logger.error(f"Failed to deserialize session {session_id}: {e}", exc_info=True)
+        raise RedisError(f"Corrupted session data for {session_id}") from e
+
+
+async def _cache_session(session: ChatSession) -> None:
+    """Cache a chat session in Redis."""
+    redis_key = _get_session_cache_key(session.session_id)
+    async_redis = await get_redis_async()
+    await async_redis.setex(redis_key, config.session_ttl, session.model_dump_json())
+
+
+async def cache_chat_session(session: ChatSession) -> None:
+    """Cache a chat session without persisting to the database."""
+    await _cache_session(session)
+
+
+async def invalidate_session_cache(session_id: str) -> None:
+    """Invalidate a chat session from Redis cache.
+
+    Used by background tasks to ensure fresh data is loaded on next access.
+    This is best-effort - Redis failures are logged but don't fail the operation.
+    """
+    try:
+        redis_key = _get_session_cache_key(session_id)
+        async_redis = await get_redis_async()
+        await async_redis.delete(redis_key)
+    except Exception as e:
+        # Best-effort: log but don't fail - cache will expire naturally
+        logger.warning(f"Failed to invalidate session cache for {session_id}: {e}")
+
+
+async def _get_session_from_db(session_id: str) -> ChatSession | None:
+    """Get a chat session from the database."""
+    prisma_session = await chat_db.get_chat_session(session_id)
+    if not prisma_session:
+        return None
+
+    messages = prisma_session.Messages
+    logger.debug(
+        f"[DB] Loaded session {session_id}: {len(messages) if messages else 0} messages, "
+        f"roles={[m.role for m in messages[-3:]] if messages else []}"  # Last 3 roles
+    )
+
+    return ChatSession.from_db(prisma_session, messages)
+
+
+async def _save_session_to_db(
+    session: ChatSession, existing_message_count: int
+) -> None:
+    """Save or update a chat session in the database."""
+    # Check if session exists in DB
+    existing = await chat_db.get_chat_session(session.session_id)
+
+    if not existing:
+        # Create new session
+        await chat_db.create_chat_session(
+            session_id=session.session_id,
+            user_id=session.user_id,
+        )
+        existing_message_count = 0
+
+    # Calculate total tokens from usage
+    total_prompt = sum(u.prompt_tokens for u in session.usage)
+    total_completion = sum(u.completion_tokens for u in session.usage)
+
+    # Update session metadata
+    await chat_db.update_chat_session(
+        session_id=session.session_id,
+        credentials=session.credentials,
+        successful_agent_runs=session.successful_agent_runs,
+        successful_agent_schedules=session.successful_agent_schedules,
+        total_prompt_tokens=total_prompt,
+        total_completion_tokens=total_completion,
+    )
+
+    # Add new messages (only those after existing count)
+    new_messages = session.messages[existing_message_count:]
+    if new_messages:
+        messages_data = []
+        for msg in new_messages:
+            messages_data.append(
+                {
+                    "role": msg.role,
+                    "content": msg.content,
+                    "name": msg.name,
+                    "tool_call_id": msg.tool_call_id,
+                    "refusal": msg.refusal,
+                    "tool_calls": msg.tool_calls,
+                    "function_call": msg.function_call,
+                }
+            )
+        logger.debug(
+            f"[DB] Saving {len(new_messages)} messages to session {session.session_id}, "
+            f"roles={[m['role'] for m in messages_data]}"
+        )
+        await chat_db.add_chat_messages_batch(
+            session_id=session.session_id,
+            messages=messages_data,
+            start_sequence=existing_message_count,
+        )
+
+
+async def get_chat_session(
+    session_id: str,
+    user_id: str | None = None,
+) -> ChatSession | None:
+    """Get a chat session by ID.
+
+    Checks Redis cache first, falls back to database if not found.
+    Caches database results back to Redis.
+
+    Args:
+        session_id: The session ID to fetch.
+        user_id: If provided, validates that the session belongs to this user.
+            If None, ownership is not validated (admin/system access).
+    """
+    # Try cache first
+    try:
+        session = await _get_session_from_cache(session_id)
+        if session:
+            # Verify user ownership if user_id was provided for validation
+            if user_id is not None and session.user_id != user_id:
+                logger.warning(
+                    f"Session {session_id} user id mismatch: {session.user_id} != {user_id}"
+                )
+                return None
+            return session
+    except RedisError:
+        logger.warning(f"Cache error for session {session_id}, trying database")
+    except Exception as e:
+        logger.warning(f"Unexpected cache error for session {session_id}: {e}")
+
+    # Fall back to database
+    logger.debug(f"Session {session_id} not in cache, checking database")
+    session = await _get_session_from_db(session_id)
+
+    if session is None:
+        logger.warning(f"Session {session_id} not found in cache or database")
+        return None
+
+    # Verify user ownership if user_id was provided for validation
+    if user_id is not None and session.user_id != user_id:
+        logger.warning(
+            f"Session {session_id} user id mismatch: {session.user_id} != {user_id}"
+        )
+        return None
+
+    # Cache the session from DB
+    try:
+        await _cache_session(session)
+    except Exception as e:
+        logger.warning(f"Failed to cache session {session_id}: {e}")
+
+    return session
+
+
+async def upsert_chat_session(
+    session: ChatSession,
+) -> ChatSession:
+    """Update a chat session in both cache and database.
+
+    Uses session-level locking to prevent race conditions when concurrent
+    operations (e.g., background title update and main stream handler)
+    attempt to upsert the same session simultaneously.
+
+    Raises:
+        DatabaseError: If the database write fails. The cache is still updated
+            as a best-effort optimization, but the error is propagated to ensure
+            callers are aware of the persistence failure.
+        RedisError: If the cache write fails (after successful DB write).
+    """
+    # Acquire session-specific lock to prevent concurrent upserts
+    lock = await _get_session_lock(session.session_id)
+
+    async with lock:
+        # Get existing message count from DB for incremental saves
+        existing_message_count = await chat_db.get_chat_session_message_count(
+            session.session_id
+        )
+
+        db_error: Exception | None = None
+
+        # Save to database (primary storage)
+        try:
+            await _save_session_to_db(session, existing_message_count)
+        except Exception as e:
+            logger.error(
+                f"Failed to save session {session.session_id} to database: {e}"
+            )
+            db_error = e
+
+        # Save to cache (best-effort, even if DB failed)
+        try:
+            await _cache_session(session)
+        except Exception as e:
+            # If DB succeeded but cache failed, raise cache error
+            if db_error is None:
+                raise RedisError(
+                    f"Failed to persist chat session {session.session_id} to Redis: {e}"
+                ) from e
+            # If both failed, log cache error but raise DB error (more critical)
+            logger.warning(
+                f"Cache write also failed for session {session.session_id}: {e}"
+            )
+
+        # Propagate DB error after attempting cache (prevents data loss)
+        if db_error is not None:
+            raise DatabaseError(
+                f"Failed to persist chat session {session.session_id} to database"
+            ) from db_error
+
+        return session
+
+
+async def append_and_save_message(session_id: str, message: ChatMessage) -> ChatSession:
+    """Atomically append a message to a session and persist it.
+
+    Acquires the session lock, re-fetches the latest session state,
+    appends the message, and saves — preventing message loss when
+    concurrent requests modify the same session.
+    """
+    lock = await _get_session_lock(session_id)
+
+    async with lock:
+        session = await get_chat_session(session_id)
+        if session is None:
+            raise ValueError(f"Session {session_id} not found")
+
+        session.messages.append(message)
+        existing_message_count = await chat_db.get_chat_session_message_count(
+            session_id
+        )
+
+        try:
+            await _save_session_to_db(session, existing_message_count)
+        except Exception as e:
+            raise DatabaseError(
+                f"Failed to persist message to session {session_id}"
+            ) from e
+
+        try:
+            await _cache_session(session)
+        except Exception as e:
+            logger.warning(f"Cache write failed for session {session_id}: {e}")
+
+        return session
+
+
+async def create_chat_session(user_id: str) -> ChatSession:
+    """Create a new chat session and persist it.
+
+    Raises:
+        DatabaseError: If the database write fails. We fail fast to ensure
+            callers never receive a non-persisted session that only exists
+            in cache (which would be lost when the cache expires).
+    """
+    session = ChatSession.new(user_id)
+
+    # Create in database first - fail fast if this fails
+    try:
+        await chat_db.create_chat_session(
+            session_id=session.session_id,
+            user_id=user_id,
+        )
+    except Exception as e:
+        logger.error(f"Failed to create session {session.session_id} in database: {e}")
+        raise DatabaseError(
+            f"Failed to create chat session {session.session_id} in database"
+        ) from e
+
+    # Cache the session (best-effort optimization, DB is source of truth)
+    try:
+        await _cache_session(session)
+    except Exception as e:
+        logger.warning(f"Failed to cache new session {session.session_id}: {e}")
+
+    return session
+
+
+async def get_user_sessions(
+    user_id: str,
+    limit: int = 50,
+    offset: int = 0,
+) -> tuple[list[ChatSession], int]:
+    """Get chat sessions for a user from the database with total count.
+
+    Returns:
+        A tuple of (sessions, total_count) where total_count is the overall
+        number of sessions for the user (not just the current page).
+    """
+    prisma_sessions = await chat_db.get_user_chat_sessions(user_id, limit, offset)
+    total_count = await chat_db.get_user_session_count(user_id)
+
+    sessions = []
+    for prisma_session in prisma_sessions:
+        # Convert without messages for listing (lighter weight)
+        sessions.append(ChatSession.from_db(prisma_session, None))
+
+    return sessions, total_count
+
+
+async def delete_chat_session(session_id: str, user_id: str | None = None) -> bool:
+    """Delete a chat session from both cache and database.
+
+    Args:
+        session_id: The session ID to delete.
+        user_id: If provided, validates that the session belongs to this user
+            before deletion. This prevents unauthorized deletion.
+
+    Returns:
+        True if deleted successfully, False otherwise.
+    """
+    # Delete from database first (with optional user_id validation)
+    # This confirms ownership before invalidating cache
+    deleted = await chat_db.delete_chat_session(session_id, user_id)
+
+    if not deleted:
+        return False
+
+    # Only invalidate cache and clean up lock after DB confirms deletion
+    try:
+        redis_key = _get_session_cache_key(session_id)
+        async_redis = await get_redis_async()
+        await async_redis.delete(redis_key)
+    except Exception as e:
+        logger.warning(f"Failed to delete session {session_id} from cache: {e}")
+
+    # Clean up session lock (belt-and-suspenders with WeakValueDictionary)
+    async with _session_locks_mutex:
+        _session_locks.pop(session_id, None)
+
+    return True
+
+
+async def update_session_title(session_id: str, title: str) -> bool:
+    """Update only the title of a chat session.
+
+    This is a lightweight operation that doesn't touch messages, avoiding
+    race conditions with concurrent message updates. Use this for background
+    title generation instead of upsert_chat_session.
+
+    Args:
+        session_id: The session ID to update.
+        title: The new title to set.
+
+    Returns:
+        True if updated successfully, False otherwise.
+    """
+    try:
+        result = await chat_db.update_chat_session(session_id=session_id, title=title)
+        if result is None:
+            logger.warning(f"Session {session_id} not found for title update")
+            return False
+
+        # Update title in cache if it exists (instead of invalidating).
+        # This prevents race conditions where cache invalidation causes
+        # the frontend to see stale DB data while streaming is still in progress.
+        try:
+            cached = await _get_session_from_cache(session_id)
+            if cached:
+                cached.title = title
+                await _cache_session(cached)
+        except Exception as e:
+            # Not critical - title will be correct on next full cache refresh
+            logger.warning(
+                f"Failed to update title in cache for session {session_id}: {e}"
+            )
+
+        return True
+    except Exception as e:
+        logger.error(f"Failed to update title for session {session_id}: {e}")
+        return False

autogpt_platform/backend/backend/api/features/chat/model_test.py

@@ -0,0 +1,333 @@
+from typing import cast
+
+import pytest
+from openai.types.chat import (
+    ChatCompletionAssistantMessageParam,
+    ChatCompletionMessageParam,
+    ChatCompletionToolMessageParam,
+    ChatCompletionUserMessageParam,
+)
+from openai.types.chat.chat_completion_message_tool_call_param import (
+    ChatCompletionMessageToolCallParam,
+    Function,
+)
+
+from .model import (
+    ChatMessage,
+    ChatSession,
+    Usage,
+    get_chat_session,
+    upsert_chat_session,
+)
+
+messages = [
+    ChatMessage(content="Hello, how are you?", role="user"),
+    ChatMessage(
+        content="I'm fine, thank you!",
+        role="assistant",
+        tool_calls=[
+            {
+                "id": "t123",
+                "type": "function",
+                "function": {
+                    "name": "get_weather",
+                    "arguments": '{"city": "New York"}',
+                },
+            }
+        ],
+    ),
+    ChatMessage(
+        content="I'm using the tool to get the weather",
+        role="tool",
+        tool_call_id="t123",
+    ),
+]
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_chatsession_serialization_deserialization():
+    s = ChatSession.new(user_id="abc123")
+    s.messages = messages
+    s.usage = [Usage(prompt_tokens=100, completion_tokens=200, total_tokens=300)]
+    serialized = s.model_dump_json()
+    s2 = ChatSession.model_validate_json(serialized)
+    assert s2.model_dump() == s.model_dump()
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_chatsession_redis_storage(setup_test_user, test_user_id):
+
+    s = ChatSession.new(user_id=test_user_id)
+    s.messages = messages
+
+    s = await upsert_chat_session(s)
+
+    s2 = await get_chat_session(
+        session_id=s.session_id,
+        user_id=s.user_id,
+    )
+
+    assert s2 == s
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_chatsession_redis_storage_user_id_mismatch(
+    setup_test_user, test_user_id
+):
+
+    s = ChatSession.new(user_id=test_user_id)
+    s.messages = messages
+    s = await upsert_chat_session(s)
+
+    s2 = await get_chat_session(s.session_id, "different_user_id")
+
+    assert s2 is None
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_chatsession_db_storage(setup_test_user, test_user_id):
+    """Test that messages are correctly saved to and loaded from DB (not cache)."""
+    from backend.data.redis_client import get_redis_async
+
+    # Create session with messages including assistant message
+    s = ChatSession.new(user_id=test_user_id)
+    s.messages = messages  # Contains user, assistant, and tool messages
+    assert s.session_id is not None, "Session id is not set"
+    # Upsert to save to both cache and DB
+    s = await upsert_chat_session(s)
+
+    # Clear the Redis cache to force DB load
+    redis_key = f"chat:session:{s.session_id}"
+    async_redis = await get_redis_async()
+    await async_redis.delete(redis_key)
+
+    # Load from DB (cache was cleared)
+    s2 = await get_chat_session(
+        session_id=s.session_id,
+        user_id=s.user_id,
+    )
+
+    assert s2 is not None, "Session not found after loading from DB"
+    assert len(s2.messages) == len(
+        s.messages
+    ), f"Message count mismatch: expected {len(s.messages)}, got {len(s2.messages)}"
+
+    # Verify all roles are present
+    roles = [m.role for m in s2.messages]
+    assert "user" in roles, f"User message missing. Roles found: {roles}"
+    assert "assistant" in roles, f"Assistant message missing. Roles found: {roles}"
+    assert "tool" in roles, f"Tool message missing. Roles found: {roles}"
+
+    # Verify message content
+    for orig, loaded in zip(s.messages, s2.messages):
+        assert orig.role == loaded.role, f"Role mismatch: {orig.role} != {loaded.role}"
+        assert (
+            orig.content == loaded.content
+        ), f"Content mismatch for {orig.role}: {orig.content} != {loaded.content}"
+        if orig.tool_calls:
+            assert (
+                loaded.tool_calls is not None
+            ), f"Tool calls missing for {orig.role} message"
+            assert len(orig.tool_calls) == len(loaded.tool_calls)
+
+
+# --------------------------------------------------------------------------- #
+#  _merge_consecutive_assistant_messages                                       #
+# --------------------------------------------------------------------------- #
+
+_tc = ChatCompletionMessageToolCallParam(
+    id="tc1", type="function", function=Function(name="do_stuff", arguments="{}")
+)
+_tc2 = ChatCompletionMessageToolCallParam(
+    id="tc2", type="function", function=Function(name="other", arguments="{}")
+)
+
+
+def test_merge_noop_when_no_consecutive_assistants():
+    """Messages without consecutive assistants are returned unchanged."""
+    msgs = [
+        ChatCompletionUserMessageParam(role="user", content="hi"),
+        ChatCompletionAssistantMessageParam(role="assistant", content="hello"),
+        ChatCompletionUserMessageParam(role="user", content="bye"),
+    ]
+    merged = ChatSession._merge_consecutive_assistant_messages(msgs)
+    assert len(merged) == 3
+    assert [m["role"] for m in merged] == ["user", "assistant", "user"]
+
+
+def test_merge_splits_text_and_tool_calls():
+    """The exact bug scenario: text-only assistant followed by tool_calls-only assistant."""
+    msgs = [
+        ChatCompletionUserMessageParam(role="user", content="build agent"),
+        ChatCompletionAssistantMessageParam(
+            role="assistant", content="Let me build that"
+        ),
+        ChatCompletionAssistantMessageParam(
+            role="assistant", content="", tool_calls=[_tc]
+        ),
+        ChatCompletionToolMessageParam(role="tool", content="ok", tool_call_id="tc1"),
+    ]
+    merged = ChatSession._merge_consecutive_assistant_messages(msgs)
+
+    assert len(merged) == 3
+    assert merged[0]["role"] == "user"
+    assert merged[2]["role"] == "tool"
+    a = cast(ChatCompletionAssistantMessageParam, merged[1])
+    assert a["role"] == "assistant"
+    assert a.get("content") == "Let me build that"
+    assert a.get("tool_calls") == [_tc]
+
+
+def test_merge_combines_tool_calls_from_both():
+    """Both consecutive assistants have tool_calls — they get merged."""
+    msgs: list[ChatCompletionAssistantMessageParam] = [
+        ChatCompletionAssistantMessageParam(
+            role="assistant", content="text", tool_calls=[_tc]
+        ),
+        ChatCompletionAssistantMessageParam(
+            role="assistant", content="", tool_calls=[_tc2]
+        ),
+    ]
+    merged = ChatSession._merge_consecutive_assistant_messages(msgs)  # type: ignore[arg-type]
+
+    assert len(merged) == 1
+    a = cast(ChatCompletionAssistantMessageParam, merged[0])
+    assert a.get("tool_calls") == [_tc, _tc2]
+    assert a.get("content") == "text"
+
+
+def test_merge_three_consecutive_assistants():
+    """Three consecutive assistants collapse into one."""
+    msgs: list[ChatCompletionAssistantMessageParam] = [
+        ChatCompletionAssistantMessageParam(role="assistant", content="a"),
+        ChatCompletionAssistantMessageParam(role="assistant", content="b"),
+        ChatCompletionAssistantMessageParam(
+            role="assistant", content="", tool_calls=[_tc]
+        ),
+    ]
+    merged = ChatSession._merge_consecutive_assistant_messages(msgs)  # type: ignore[arg-type]
+
+    assert len(merged) == 1
+    a = cast(ChatCompletionAssistantMessageParam, merged[0])
+    assert a.get("content") == "a\nb"
+    assert a.get("tool_calls") == [_tc]
+
+
+def test_merge_empty_and_single_message():
+    """Edge cases: empty list and single message."""
+    assert ChatSession._merge_consecutive_assistant_messages([]) == []
+
+    single: list[ChatCompletionMessageParam] = [
+        ChatCompletionUserMessageParam(role="user", content="hi")
+    ]
+    assert ChatSession._merge_consecutive_assistant_messages(single) == single
+
+
+# --------------------------------------------------------------------------- #
+#  add_tool_call_to_current_turn                                               #
+# --------------------------------------------------------------------------- #
+
+_raw_tc = {
+    "id": "tc1",
+    "type": "function",
+    "function": {"name": "f", "arguments": "{}"},
+}
+_raw_tc2 = {
+    "id": "tc2",
+    "type": "function",
+    "function": {"name": "g", "arguments": "{}"},
+}
+
+
+def test_add_tool_call_appends_to_existing_assistant():
+    """When the last assistant is from the current turn, tool_call is added to it."""
+    session = ChatSession.new(user_id="u")
+    session.messages = [
+        ChatMessage(role="user", content="hi"),
+        ChatMessage(role="assistant", content="working on it"),
+    ]
+    session.add_tool_call_to_current_turn(_raw_tc)
+
+    assert len(session.messages) == 2  # no new message created
+    assert session.messages[1].tool_calls == [_raw_tc]
+
+
+def test_add_tool_call_creates_assistant_when_none_exists():
+    """When there's no current-turn assistant, a new one is created."""
+    session = ChatSession.new(user_id="u")
+    session.messages = [
+        ChatMessage(role="user", content="hi"),
+    ]
+    session.add_tool_call_to_current_turn(_raw_tc)
+
+    assert len(session.messages) == 2
+    assert session.messages[1].role == "assistant"
+    assert session.messages[1].tool_calls == [_raw_tc]
+
+
+def test_add_tool_call_does_not_cross_user_boundary():
+    """A user message acts as a boundary — previous assistant is not modified."""
+    session = ChatSession.new(user_id="u")
+    session.messages = [
+        ChatMessage(role="assistant", content="old turn"),
+        ChatMessage(role="user", content="new message"),
+    ]
+    session.add_tool_call_to_current_turn(_raw_tc)
+
+    assert len(session.messages) == 3  # new assistant was created
+    assert session.messages[0].tool_calls is None  # old assistant untouched
+    assert session.messages[2].role == "assistant"
+    assert session.messages[2].tool_calls == [_raw_tc]
+
+
+def test_add_tool_call_multiple_times():
+    """Multiple long-running tool calls accumulate on the same assistant."""
+    session = ChatSession.new(user_id="u")
+    session.messages = [
+        ChatMessage(role="user", content="hi"),
+        ChatMessage(role="assistant", content="doing stuff"),
+    ]
+    session.add_tool_call_to_current_turn(_raw_tc)
+    # Simulate a pending tool result in between (like _yield_tool_call does)
+    session.messages.append(
+        ChatMessage(role="tool", content="pending", tool_call_id="tc1")
+    )
+    session.add_tool_call_to_current_turn(_raw_tc2)
+
+    assert len(session.messages) == 3  # user, assistant, tool — no extra assistant
+    assert session.messages[1].tool_calls == [_raw_tc, _raw_tc2]
+
+
+def test_to_openai_messages_merges_split_assistants():
+    """End-to-end: session with split assistants produces valid OpenAI messages."""
+    session = ChatSession.new(user_id="u")
+    session.messages = [
+        ChatMessage(role="user", content="build agent"),
+        ChatMessage(role="assistant", content="Let me build that"),
+        ChatMessage(
+            role="assistant",
+            content="",
+            tool_calls=[
+                {
+                    "id": "tc1",
+                    "type": "function",
+                    "function": {"name": "create_agent", "arguments": "{}"},
+                }
+            ],
+        ),
+        ChatMessage(role="tool", content="done", tool_call_id="tc1"),
+        ChatMessage(role="assistant", content="Saved!"),
+        ChatMessage(role="user", content="show me an example run"),
+    ]
+    openai_msgs = session.to_openai_messages()
+
+    # The two consecutive assistants at index 1,2 should be merged
+    roles = [m["role"] for m in openai_msgs]
+    assert roles == ["user", "assistant", "tool", "assistant", "user"]
+
+    # The merged assistant should have both content and tool_calls
+    merged = cast(ChatCompletionAssistantMessageParam, openai_msgs[1])
+    assert merged.get("content") == "Let me build that"
+    tc_list = merged.get("tool_calls")
+    assert tc_list is not None and len(list(tc_list)) == 1
+    assert list(tc_list)[0]["id"] == "tc1"

autogpt_platform/backend/backend/api/features/chat/response_model.py

@@ -0,0 +1,225 @@
+"""
+Response models for Vercel AI SDK UI Stream Protocol.
+
+This module implements the AI SDK UI Stream Protocol (v1) for streaming chat responses.
+See: https://ai-sdk.dev/docs/ai-sdk-ui/stream-protocol
+"""
+
+from enum import Enum
+from typing import Any
+
+from pydantic import BaseModel, Field
+
+from backend.util.json import dumps as json_dumps
+
+
+class ResponseType(str, Enum):
+    """Types of streaming responses following AI SDK protocol."""
+
+    # Message lifecycle
+    START = "start"
+    FINISH = "finish"
+
+    # Step lifecycle (one LLM API call within a message)
+    START_STEP = "start-step"
+    FINISH_STEP = "finish-step"
+
+    # Text streaming
+    TEXT_START = "text-start"
+    TEXT_DELTA = "text-delta"
+    TEXT_END = "text-end"
+
+    # Tool interaction
+    TOOL_INPUT_START = "tool-input-start"
+    TOOL_INPUT_AVAILABLE = "tool-input-available"
+    TOOL_OUTPUT_AVAILABLE = "tool-output-available"
+
+    # Other
+    ERROR = "error"
+    USAGE = "usage"
+    HEARTBEAT = "heartbeat"
+
+
+class StreamBaseResponse(BaseModel):
+    """Base response model for all streaming responses."""
+
+    type: ResponseType
+
+    def to_sse(self) -> str:
+        """Convert to SSE format."""
+        return f"data: {self.model_dump_json()}\n\n"
+
+
+# ========== Message Lifecycle ==========
+
+
+class StreamStart(StreamBaseResponse):
+    """Start of a new message."""
+
+    type: ResponseType = ResponseType.START
+    messageId: str = Field(..., description="Unique message ID")
+    taskId: str | None = Field(
+        default=None,
+        description="Task ID for SSE reconnection. Clients can reconnect using GET /tasks/{taskId}/stream",
+    )
+
+    def to_sse(self) -> str:
+        """Convert to SSE format, excluding non-protocol fields like taskId."""
+        import json
+
+        data: dict[str, Any] = {
+            "type": self.type.value,
+            "messageId": self.messageId,
+        }
+        return f"data: {json.dumps(data)}\n\n"
+
+
+class StreamFinish(StreamBaseResponse):
+    """End of message/stream."""
+
+    type: ResponseType = ResponseType.FINISH
+
+
+class StreamStartStep(StreamBaseResponse):
+    """Start of a step (one LLM API call within a message).
+
+    The AI SDK uses this to add a step-start boundary to message.parts,
+    enabling visual separation between multiple LLM calls in a single message.
+    """
+
+    type: ResponseType = ResponseType.START_STEP
+
+
+class StreamFinishStep(StreamBaseResponse):
+    """End of a step (one LLM API call within a message).
+
+    The AI SDK uses this to reset activeTextParts and activeReasoningParts,
+    so the next LLM call in a tool-call continuation starts with clean state.
+    """
+
+    type: ResponseType = ResponseType.FINISH_STEP
+
+
+# ========== Text Streaming ==========
+
+
+class StreamTextStart(StreamBaseResponse):
+    """Start of a text block."""
+
+    type: ResponseType = ResponseType.TEXT_START
+    id: str = Field(..., description="Text block ID")
+
+
+class StreamTextDelta(StreamBaseResponse):
+    """Streaming text content delta."""
+
+    type: ResponseType = ResponseType.TEXT_DELTA
+    id: str = Field(..., description="Text block ID")
+    delta: str = Field(..., description="Text content delta")
+
+
+class StreamTextEnd(StreamBaseResponse):
+    """End of a text block."""
+
+    type: ResponseType = ResponseType.TEXT_END
+    id: str = Field(..., description="Text block ID")
+
+
+# ========== Tool Interaction ==========
+
+
+class StreamToolInputStart(StreamBaseResponse):
+    """Tool call started notification."""
+
+    type: ResponseType = ResponseType.TOOL_INPUT_START
+    toolCallId: str = Field(..., description="Unique tool call ID")
+    toolName: str = Field(..., description="Name of the tool being called")
+
+
+class StreamToolInputAvailable(StreamBaseResponse):
+    """Tool input is ready for execution."""
+
+    type: ResponseType = ResponseType.TOOL_INPUT_AVAILABLE
+    toolCallId: str = Field(..., description="Unique tool call ID")
+    toolName: str = Field(..., description="Name of the tool being called")
+    input: dict[str, Any] = Field(
+        default_factory=dict, description="Tool input arguments"
+    )
+
+
+class StreamToolOutputAvailable(StreamBaseResponse):
+    """Tool execution result."""
+
+    type: ResponseType = ResponseType.TOOL_OUTPUT_AVAILABLE
+    toolCallId: str = Field(..., description="Tool call ID this responds to")
+    output: str | dict[str, Any] = Field(..., description="Tool execution output")
+    # Keep these for internal backend use
+    toolName: str | None = Field(
+        default=None, description="Name of the tool that was executed"
+    )
+    success: bool = Field(
+        default=True, description="Whether the tool execution succeeded"
+    )
+
+    def to_sse(self) -> str:
+        """Convert to SSE format, excluding non-spec fields."""
+        import json
+
+        data = {
+            "type": self.type.value,
+            "toolCallId": self.toolCallId,
+            "output": self.output,
+        }
+        return f"data: {json.dumps(data)}\n\n"
+
+
+# ========== Other ==========
+
+
+class StreamUsage(StreamBaseResponse):
+    """Token usage statistics."""
+
+    type: ResponseType = ResponseType.USAGE
+    promptTokens: int = Field(..., description="Number of prompt tokens")
+    completionTokens: int = Field(..., description="Number of completion tokens")
+    totalTokens: int = Field(..., description="Total number of tokens")
+
+
+class StreamError(StreamBaseResponse):
+    """Error response."""
+
+    type: ResponseType = ResponseType.ERROR
+    errorText: str = Field(..., description="Error message text")
+    code: str | None = Field(default=None, description="Error code")
+    details: dict[str, Any] | None = Field(
+        default=None, description="Additional error details"
+    )
+
+    def to_sse(self) -> str:
+        """Convert to SSE format, only emitting fields required by AI SDK protocol.
+
+        The AI SDK uses z.strictObject({type, errorText}) which rejects
+        any extra fields like `code` or `details`.
+        """
+        data = {
+            "type": self.type.value,
+            "errorText": self.errorText,
+        }
+        return f"data: {json_dumps(data)}\n\n"
+
+
+class StreamHeartbeat(StreamBaseResponse):
+    """Heartbeat to keep SSE connection alive during long-running operations.
+
+    Uses SSE comment format (: comment) which is ignored by clients but keeps
+    the connection alive through proxies and load balancers.
+    """
+
+    type: ResponseType = ResponseType.HEARTBEAT
+    toolCallId: str | None = Field(
+        default=None, description="Tool call ID if heartbeat is for a specific tool"
+    )
+
+    def to_sse(self) -> str:
+        """Convert to SSE comment format to keep connection alive."""
+        return ": heartbeat\n\n"

autogpt_platform/backend/backend/api/features/chat/sdk/__init__.py

@@ -0,0 +1,14 @@
+"""Claude Agent SDK integration for CoPilot.
+
+This module provides the integration layer between the Claude Agent SDK
+and the existing CoPilot tool system, enabling drop-in replacement of
+the current LLM orchestration with the battle-tested Claude Agent SDK.
+"""
+
+from .service import stream_chat_completion_sdk
+from .tool_adapter import create_copilot_mcp_server
+
+__all__ = [
+    "stream_chat_completion_sdk",
+    "create_copilot_mcp_server",
+]

autogpt_platform/backend/backend/api/features/chat/sdk/response_adapter.py

@@ -0,0 +1,203 @@
+"""Response adapter for converting Claude Agent SDK messages to Vercel AI SDK format.
+
+This module provides the adapter layer that converts streaming messages from
+the Claude Agent SDK into the Vercel AI SDK UI Stream Protocol format that
+the frontend expects.
+"""
+
+import json
+import logging
+import uuid
+
+from claude_agent_sdk import (
+    AssistantMessage,
+    Message,
+    ResultMessage,
+    SystemMessage,
+    TextBlock,
+    ToolResultBlock,
+    ToolUseBlock,
+    UserMessage,
+)
+
+from backend.api.features.chat.response_model import (
+    StreamBaseResponse,
+    StreamError,
+    StreamFinish,
+    StreamFinishStep,
+    StreamStart,
+    StreamStartStep,
+    StreamTextDelta,
+    StreamTextEnd,
+    StreamTextStart,
+    StreamToolInputAvailable,
+    StreamToolInputStart,
+    StreamToolOutputAvailable,
+)
+from backend.api.features.chat.sdk.tool_adapter import (
+    MCP_TOOL_PREFIX,
+    pop_pending_tool_output,
+)
+
+logger = logging.getLogger(__name__)
+
+
+class SDKResponseAdapter:
+    """Adapter for converting Claude Agent SDK messages to Vercel AI SDK format.
+
+    This class maintains state during a streaming session to properly track
+    text blocks, tool calls, and message lifecycle.
+    """
+
+    def __init__(self, message_id: str | None = None):
+        self.message_id = message_id or str(uuid.uuid4())
+        self.text_block_id = str(uuid.uuid4())
+        self.has_started_text = False
+        self.has_ended_text = False
+        self.current_tool_calls: dict[str, dict[str, str]] = {}
+        self.task_id: str | None = None
+        self.step_open = False
+
+    def set_task_id(self, task_id: str) -> None:
+        """Set the task ID for reconnection support."""
+        self.task_id = task_id
+
+    def convert_message(self, sdk_message: Message) -> list[StreamBaseResponse]:
+        """Convert a single SDK message to Vercel AI SDK format."""
+        responses: list[StreamBaseResponse] = []
+
+        if isinstance(sdk_message, SystemMessage):
+            if sdk_message.subtype == "init":
+                responses.append(
+                    StreamStart(messageId=self.message_id, taskId=self.task_id)
+                )
+                # Open the first step (matches non-SDK: StreamStart then StreamStartStep)
+                responses.append(StreamStartStep())
+                self.step_open = True
+
+        elif isinstance(sdk_message, AssistantMessage):
+            # After tool results, the SDK sends a new AssistantMessage for the
+            # next LLM turn. Open a new step if the previous one was closed.
+            if not self.step_open:
+                responses.append(StreamStartStep())
+                self.step_open = True
+
+            for block in sdk_message.content:
+                if isinstance(block, TextBlock):
+                    if block.text:
+                        self._ensure_text_started(responses)
+                        responses.append(
+                            StreamTextDelta(id=self.text_block_id, delta=block.text)
+                        )
+
+                elif isinstance(block, ToolUseBlock):
+                    self._end_text_if_open(responses)
+
+                    # Strip MCP prefix so frontend sees "find_block"
+                    # instead of "mcp__copilot__find_block".
+                    tool_name = block.name.removeprefix(MCP_TOOL_PREFIX)
+
+                    responses.append(
+                        StreamToolInputStart(toolCallId=block.id, toolName=tool_name)
+                    )
+                    responses.append(
+                        StreamToolInputAvailable(
+                            toolCallId=block.id,
+                            toolName=tool_name,
+                            input=block.input,
+                        )
+                    )
+                    self.current_tool_calls[block.id] = {"name": tool_name}
+
+        elif isinstance(sdk_message, UserMessage):
+            # UserMessage carries tool results back from tool execution.
+            content = sdk_message.content
+            blocks = content if isinstance(content, list) else []
+            for block in blocks:
+                if isinstance(block, ToolResultBlock) and block.tool_use_id:
+                    tool_info = self.current_tool_calls.get(block.tool_use_id, {})
+                    tool_name = tool_info.get("name", "unknown")
+
+                    # Prefer the stashed full output over the SDK's
+                    # (potentially truncated) ToolResultBlock content.
+                    # The SDK truncates large results, writing them to disk,
+                    # which breaks frontend widget parsing.
+                    output = pop_pending_tool_output(tool_name) or (
+                        _extract_tool_output(block.content)
+                    )
+
+                    responses.append(
+                        StreamToolOutputAvailable(
+                            toolCallId=block.tool_use_id,
+                            toolName=tool_name,
+                            output=output,
+                            success=not (block.is_error or False),
+                        )
+                    )
+
+            # Close the current step after tool results — the next
+            # AssistantMessage will open a new step for the continuation.
+            if self.step_open:
+                responses.append(StreamFinishStep())
+                self.step_open = False
+
+        elif isinstance(sdk_message, ResultMessage):
+            self._end_text_if_open(responses)
+            # Close the step before finishing.
+            if self.step_open:
+                responses.append(StreamFinishStep())
+                self.step_open = False
+
+            if sdk_message.subtype == "success":
+                responses.append(StreamFinish())
+            elif sdk_message.subtype in ("error", "error_during_execution"):
+                error_msg = getattr(sdk_message, "result", None) or "Unknown error"
+                responses.append(
+                    StreamError(errorText=str(error_msg), code="sdk_error")
+                )
+                responses.append(StreamFinish())
+            else:
+                logger.warning(
+                    f"Unexpected ResultMessage subtype: {sdk_message.subtype}"
+                )
+                responses.append(StreamFinish())
+
+        else:
+            logger.debug(f"Unhandled SDK message type: {type(sdk_message).__name__}")
+
+        return responses
+
+    def _ensure_text_started(self, responses: list[StreamBaseResponse]) -> None:
+        """Start (or restart) a text block if needed."""
+        if not self.has_started_text or self.has_ended_text:
+            if self.has_ended_text:
+                self.text_block_id = str(uuid.uuid4())
+                self.has_ended_text = False
+            responses.append(StreamTextStart(id=self.text_block_id))
+            self.has_started_text = True
+
+    def _end_text_if_open(self, responses: list[StreamBaseResponse]) -> None:
+        """End the current text block if one is open."""
+        if self.has_started_text and not self.has_ended_text:
+            responses.append(StreamTextEnd(id=self.text_block_id))
+            self.has_ended_text = True
+
+
+def _extract_tool_output(content: str | list[dict[str, str]] | None) -> str:
+    """Extract a string output from a ToolResultBlock's content field."""
+    if isinstance(content, str):
+        return content
+    if isinstance(content, list):
+        parts = [item.get("text", "") for item in content if item.get("type") == "text"]
+        if parts:
+            return "".join(parts)
+        try:
+            return json.dumps(content)
+        except (TypeError, ValueError):
+            return str(content)
+    if content is None:
+        return ""
+    try:
+        return json.dumps(content)
+    except (TypeError, ValueError):
+        return str(content)

autogpt_platform/backend/backend/api/features/chat/sdk/response_adapter_test.py

@@ -0,0 +1,366 @@
+"""Unit tests for the SDK response adapter."""
+
+from claude_agent_sdk import (
+    AssistantMessage,
+    ResultMessage,
+    SystemMessage,
+    TextBlock,
+    ToolResultBlock,
+    ToolUseBlock,
+    UserMessage,
+)
+
+from backend.api.features.chat.response_model import (
+    StreamBaseResponse,
+    StreamError,
+    StreamFinish,
+    StreamFinishStep,
+    StreamStart,
+    StreamStartStep,
+    StreamTextDelta,
+    StreamTextEnd,
+    StreamTextStart,
+    StreamToolInputAvailable,
+    StreamToolInputStart,
+    StreamToolOutputAvailable,
+)
+
+from .response_adapter import SDKResponseAdapter
+from .tool_adapter import MCP_TOOL_PREFIX
+
+
+def _adapter() -> SDKResponseAdapter:
+    a = SDKResponseAdapter(message_id="msg-1")
+    a.set_task_id("task-1")
+    return a
+
+
+# -- SystemMessage -----------------------------------------------------------
+
+
+def test_system_init_emits_start_and_step():
+    adapter = _adapter()
+    results = adapter.convert_message(SystemMessage(subtype="init", data={}))
+    assert len(results) == 2
+    assert isinstance(results[0], StreamStart)
+    assert results[0].messageId == "msg-1"
+    assert results[0].taskId == "task-1"
+    assert isinstance(results[1], StreamStartStep)
+
+
+def test_system_non_init_emits_nothing():
+    adapter = _adapter()
+    results = adapter.convert_message(SystemMessage(subtype="other", data={}))
+    assert results == []
+
+
+# -- AssistantMessage with TextBlock -----------------------------------------
+
+
+def test_text_block_emits_step_start_and_delta():
+    adapter = _adapter()
+    msg = AssistantMessage(content=[TextBlock(text="hello")], model="test")
+    results = adapter.convert_message(msg)
+    assert len(results) == 3
+    assert isinstance(results[0], StreamStartStep)
+    assert isinstance(results[1], StreamTextStart)
+    assert isinstance(results[2], StreamTextDelta)
+    assert results[2].delta == "hello"
+
+
+def test_empty_text_block_emits_only_step():
+    adapter = _adapter()
+    msg = AssistantMessage(content=[TextBlock(text="")], model="test")
+    results = adapter.convert_message(msg)
+    # Empty text skipped, but step still opens
+    assert len(results) == 1
+    assert isinstance(results[0], StreamStartStep)
+
+
+def test_multiple_text_deltas_reuse_block_id():
+    adapter = _adapter()
+    msg1 = AssistantMessage(content=[TextBlock(text="a")], model="test")
+    msg2 = AssistantMessage(content=[TextBlock(text="b")], model="test")
+    r1 = adapter.convert_message(msg1)
+    r2 = adapter.convert_message(msg2)
+    # First gets step+start+delta, second only delta (block & step already started)
+    assert len(r1) == 3
+    assert isinstance(r1[0], StreamStartStep)
+    assert isinstance(r1[1], StreamTextStart)
+    assert len(r2) == 1
+    assert isinstance(r2[0], StreamTextDelta)
+    assert r1[1].id == r2[0].id  # same block ID
+
+
+# -- AssistantMessage with ToolUseBlock --------------------------------------
+
+
+def test_tool_use_emits_input_start_and_available():
+    """Tool names arrive with MCP prefix and should be stripped for the frontend."""
+    adapter = _adapter()
+    msg = AssistantMessage(
+        content=[
+            ToolUseBlock(
+                id="tool-1",
+                name=f"{MCP_TOOL_PREFIX}find_agent",
+                input={"q": "x"},
+            )
+        ],
+        model="test",
+    )
+    results = adapter.convert_message(msg)
+    assert len(results) == 3
+    assert isinstance(results[0], StreamStartStep)
+    assert isinstance(results[1], StreamToolInputStart)
+    assert results[1].toolCallId == "tool-1"
+    assert results[1].toolName == "find_agent"  # prefix stripped
+    assert isinstance(results[2], StreamToolInputAvailable)
+    assert results[2].toolName == "find_agent"  # prefix stripped
+    assert results[2].input == {"q": "x"}
+
+
+def test_text_then_tool_ends_text_block():
+    adapter = _adapter()
+    text_msg = AssistantMessage(content=[TextBlock(text="thinking...")], model="test")
+    tool_msg = AssistantMessage(
+        content=[ToolUseBlock(id="t1", name=f"{MCP_TOOL_PREFIX}tool", input={})],
+        model="test",
+    )
+    adapter.convert_message(text_msg)  # opens step + text
+    results = adapter.convert_message(tool_msg)
+    # Step already open, so: TextEnd, ToolInputStart, ToolInputAvailable
+    assert len(results) == 3
+    assert isinstance(results[0], StreamTextEnd)
+    assert isinstance(results[1], StreamToolInputStart)
+
+
+# -- UserMessage with ToolResultBlock ----------------------------------------
+
+
+def test_tool_result_emits_output_and_finish_step():
+    adapter = _adapter()
+    # First register the tool call (opens step) — SDK sends prefixed name
+    tool_msg = AssistantMessage(
+        content=[ToolUseBlock(id="t1", name=f"{MCP_TOOL_PREFIX}find_agent", input={})],
+        model="test",
+    )
+    adapter.convert_message(tool_msg)
+
+    # Now send tool result
+    result_msg = UserMessage(
+        content=[ToolResultBlock(tool_use_id="t1", content="found 3 agents")]
+    )
+    results = adapter.convert_message(result_msg)
+    assert len(results) == 2
+    assert isinstance(results[0], StreamToolOutputAvailable)
+    assert results[0].toolCallId == "t1"
+    assert results[0].toolName == "find_agent"  # prefix stripped
+    assert results[0].output == "found 3 agents"
+    assert results[0].success is True
+    assert isinstance(results[1], StreamFinishStep)
+
+
+def test_tool_result_error():
+    adapter = _adapter()
+    adapter.convert_message(
+        AssistantMessage(
+            content=[
+                ToolUseBlock(id="t1", name=f"{MCP_TOOL_PREFIX}run_agent", input={})
+            ],
+            model="test",
+        )
+    )
+    result_msg = UserMessage(
+        content=[ToolResultBlock(tool_use_id="t1", content="timeout", is_error=True)]
+    )
+    results = adapter.convert_message(result_msg)
+    assert isinstance(results[0], StreamToolOutputAvailable)
+    assert results[0].success is False
+    assert isinstance(results[1], StreamFinishStep)
+
+
+def test_tool_result_list_content():
+    adapter = _adapter()
+    adapter.convert_message(
+        AssistantMessage(
+            content=[ToolUseBlock(id="t1", name=f"{MCP_TOOL_PREFIX}tool", input={})],
+            model="test",
+        )
+    )
+    result_msg = UserMessage(
+        content=[
+            ToolResultBlock(
+                tool_use_id="t1",
+                content=[
+                    {"type": "text", "text": "line1"},
+                    {"type": "text", "text": "line2"},
+                ],
+            )
+        ]
+    )
+    results = adapter.convert_message(result_msg)
+    assert isinstance(results[0], StreamToolOutputAvailable)
+    assert results[0].output == "line1line2"
+    assert isinstance(results[1], StreamFinishStep)
+
+
+def test_string_user_message_ignored():
+    """A plain string UserMessage (not tool results) produces no output."""
+    adapter = _adapter()
+    results = adapter.convert_message(UserMessage(content="hello"))
+    assert results == []
+
+
+# -- ResultMessage -----------------------------------------------------------
+
+
+def test_result_success_emits_finish_step_and_finish():
+    adapter = _adapter()
+    # Start some text first (opens step)
+    adapter.convert_message(
+        AssistantMessage(content=[TextBlock(text="done")], model="test")
+    )
+    msg = ResultMessage(
+        subtype="success",
+        duration_ms=100,
+        duration_api_ms=50,
+        is_error=False,
+        num_turns=1,
+        session_id="s1",
+    )
+    results = adapter.convert_message(msg)
+    # TextEnd + FinishStep + StreamFinish
+    assert len(results) == 3
+    assert isinstance(results[0], StreamTextEnd)
+    assert isinstance(results[1], StreamFinishStep)
+    assert isinstance(results[2], StreamFinish)
+
+
+def test_result_error_emits_error_and_finish():
+    adapter = _adapter()
+    msg = ResultMessage(
+        subtype="error",
+        duration_ms=100,
+        duration_api_ms=50,
+        is_error=True,
+        num_turns=0,
+        session_id="s1",
+        result="API rate limited",
+    )
+    results = adapter.convert_message(msg)
+    # No step was open, so no FinishStep — just Error + Finish
+    assert len(results) == 2
+    assert isinstance(results[0], StreamError)
+    assert "API rate limited" in results[0].errorText
+    assert isinstance(results[1], StreamFinish)
+
+
+# -- Text after tools (new block ID) ----------------------------------------
+
+
+def test_text_after_tool_gets_new_block_id():
+    adapter = _adapter()
+    # Text -> Tool -> ToolResult -> Text should get a new text block ID and step
+    adapter.convert_message(
+        AssistantMessage(content=[TextBlock(text="before")], model="test")
+    )
+    adapter.convert_message(
+        AssistantMessage(
+            content=[ToolUseBlock(id="t1", name=f"{MCP_TOOL_PREFIX}tool", input={})],
+            model="test",
+        )
+    )
+    # Send tool result (closes step)
+    adapter.convert_message(
+        UserMessage(content=[ToolResultBlock(tool_use_id="t1", content="ok")])
+    )
+    results = adapter.convert_message(
+        AssistantMessage(content=[TextBlock(text="after")], model="test")
+    )
+    # Should get StreamStartStep (new step) + StreamTextStart (new block) + StreamTextDelta
+    assert len(results) == 3
+    assert isinstance(results[0], StreamStartStep)
+    assert isinstance(results[1], StreamTextStart)
+    assert isinstance(results[2], StreamTextDelta)
+    assert results[2].delta == "after"
+
+
+# -- Full conversation flow --------------------------------------------------
+
+
+def test_full_conversation_flow():
+    """Simulate a complete conversation: init -> text -> tool -> result -> text -> finish."""
+    adapter = _adapter()
+    all_responses: list[StreamBaseResponse] = []
+
+    # 1. Init
+    all_responses.extend(
+        adapter.convert_message(SystemMessage(subtype="init", data={}))
+    )
+    # 2. Assistant text
+    all_responses.extend(
+        adapter.convert_message(
+            AssistantMessage(content=[TextBlock(text="Let me search")], model="test")
+        )
+    )
+    # 3. Tool use
+    all_responses.extend(
+        adapter.convert_message(
+            AssistantMessage(
+                content=[
+                    ToolUseBlock(
+                        id="t1",
+                        name=f"{MCP_TOOL_PREFIX}find_agent",
+                        input={"query": "email"},
+                    )
+                ],
+                model="test",
+            )
+        )
+    )
+    # 4. Tool result
+    all_responses.extend(
+        adapter.convert_message(
+            UserMessage(
+                content=[ToolResultBlock(tool_use_id="t1", content="Found 2 agents")]
+            )
+        )
+    )
+    # 5. More text
+    all_responses.extend(
+        adapter.convert_message(
+            AssistantMessage(content=[TextBlock(text="I found 2")], model="test")
+        )
+    )
+    # 6. Result
+    all_responses.extend(
+        adapter.convert_message(
+            ResultMessage(
+                subtype="success",
+                duration_ms=500,
+                duration_api_ms=400,
+                is_error=False,
+                num_turns=2,
+                session_id="s1",
+            )
+        )
+    )
+
+    types = [type(r).__name__ for r in all_responses]
+    assert types == [
+        "StreamStart",
+        "StreamStartStep",  # step 1: text + tool call
+        "StreamTextStart",
+        "StreamTextDelta",  # "Let me search"
+        "StreamTextEnd",  # closed before tool
+        "StreamToolInputStart",
+        "StreamToolInputAvailable",
+        "StreamToolOutputAvailable",  # tool result
+        "StreamFinishStep",  # step 1 closed after tool result
+        "StreamStartStep",  # step 2: continuation text
+        "StreamTextStart",  # new block after tool
+        "StreamTextDelta",  # "I found 2"
+        "StreamTextEnd",  # closed by result
+        "StreamFinishStep",  # step 2 closed
+        "StreamFinish",
+    ]

autogpt_platform/backend/backend/api/features/chat/sdk/security_hooks.py

@@ -0,0 +1,305 @@
+"""Security hooks for Claude Agent SDK integration.
+
+This module provides security hooks that validate tool calls before execution,
+ensuring multi-user isolation and preventing unauthorized operations.
+"""
+
+import json
+import logging
+import os
+import re
+from collections.abc import Callable
+from typing import Any, cast
+
+from backend.api.features.chat.sdk.tool_adapter import (
+    BLOCKED_TOOLS,
+    DANGEROUS_PATTERNS,
+    MCP_TOOL_PREFIX,
+    WORKSPACE_SCOPED_TOOLS,
+)
+
+logger = logging.getLogger(__name__)
+
+
+def _deny(reason: str) -> dict[str, Any]:
+    """Return a hook denial response."""
+    return {
+        "hookSpecificOutput": {
+            "hookEventName": "PreToolUse",
+            "permissionDecision": "deny",
+            "permissionDecisionReason": reason,
+        }
+    }
+
+
+def _validate_workspace_path(
+    tool_name: str, tool_input: dict[str, Any], sdk_cwd: str | None
+) -> dict[str, Any]:
+    """Validate that a workspace-scoped tool only accesses allowed paths.
+
+    Allowed directories:
+    - The SDK working directory (``/tmp/copilot-<session>/``)
+    - The SDK tool-results directory (``~/.claude/projects/…/tool-results/``)
+    """
+    path = tool_input.get("file_path") or tool_input.get("path") or ""
+    if not path:
+        # Glob/Grep without a path default to cwd which is already sandboxed
+        return {}
+
+    # Resolve relative paths against sdk_cwd (the SDK sets cwd so the LLM
+    # naturally uses relative paths like "test.txt" instead of absolute ones).
+    # Tilde paths (~/) are home-dir references, not relative — expand first.
+    if path.startswith("~"):
+        resolved = os.path.realpath(os.path.expanduser(path))
+    elif not os.path.isabs(path) and sdk_cwd:
+        resolved = os.path.realpath(os.path.join(sdk_cwd, path))
+    else:
+        resolved = os.path.realpath(path)
+
+    # Allow access within the SDK working directory
+    if sdk_cwd:
+        norm_cwd = os.path.realpath(sdk_cwd)
+        if resolved.startswith(norm_cwd + os.sep) or resolved == norm_cwd:
+            return {}
+
+    # Allow access to ~/.claude/projects/*/tool-results/ (big tool results)
+    claude_dir = os.path.realpath(os.path.expanduser("~/.claude/projects"))
+    tool_results_seg = os.sep + "tool-results" + os.sep
+    if resolved.startswith(claude_dir + os.sep) and tool_results_seg in resolved:
+        return {}
+
+    logger.warning(
+        f"Blocked {tool_name} outside workspace: {path} (resolved={resolved})"
+    )
+    workspace_hint = f" Allowed workspace: {sdk_cwd}" if sdk_cwd else ""
+    return _deny(
+        f"[SECURITY] Tool '{tool_name}' can only access files within the workspace "
+        f"directory.{workspace_hint} "
+        "This is enforced by the platform and cannot be bypassed."
+    )
+
+
+def _validate_tool_access(
+    tool_name: str, tool_input: dict[str, Any], sdk_cwd: str | None = None
+) -> dict[str, Any]:
+    """Validate that a tool call is allowed.
+
+    Returns:
+        Empty dict to allow, or dict with hookSpecificOutput to deny
+    """
+    # Block forbidden tools
+    if tool_name in BLOCKED_TOOLS:
+        logger.warning(f"Blocked tool access attempt: {tool_name}")
+        return _deny(
+            f"[SECURITY] Tool '{tool_name}' is blocked for security. "
+            "This is enforced by the platform and cannot be bypassed. "
+            "Use the CoPilot-specific MCP tools instead."
+        )
+
+    # Workspace-scoped tools: allowed only within the SDK workspace directory
+    if tool_name in WORKSPACE_SCOPED_TOOLS:
+        return _validate_workspace_path(tool_name, tool_input, sdk_cwd)
+
+    # Check for dangerous patterns in tool input
+    # Use json.dumps for predictable format (str() produces Python repr)
+    input_str = json.dumps(tool_input) if tool_input else ""
+
+    for pattern in DANGEROUS_PATTERNS:
+        if re.search(pattern, input_str, re.IGNORECASE):
+            logger.warning(
+                f"Blocked dangerous pattern in tool input: {pattern} in {tool_name}"
+            )
+            return _deny(
+                "[SECURITY] Input contains a blocked pattern. "
+                "This is enforced by the platform and cannot be bypassed."
+            )
+
+    return {}
+
+
+def _validate_user_isolation(
+    tool_name: str, tool_input: dict[str, Any], user_id: str | None
+) -> dict[str, Any]:
+    """Validate that tool calls respect user isolation."""
+    # For workspace file tools, ensure path doesn't escape
+    if "workspace" in tool_name.lower():
+        path = tool_input.get("path", "") or tool_input.get("file_path", "")
+        if path:
+            # Check for path traversal
+            if ".." in path or path.startswith("/"):
+                logger.warning(
+                    f"Blocked path traversal attempt: {path} by user {user_id}"
+                )
+                return {
+                    "hookSpecificOutput": {
+                        "hookEventName": "PreToolUse",
+                        "permissionDecision": "deny",
+                        "permissionDecisionReason": "Path traversal not allowed",
+                    }
+                }
+
+    return {}
+
+
+def create_security_hooks(
+    user_id: str | None,
+    sdk_cwd: str | None = None,
+    max_subtasks: int = 3,
+    on_stop: Callable[[str, str], None] | None = None,
+) -> dict[str, Any]:
+    """Create the security hooks configuration for Claude Agent SDK.
+
+    Includes security validation and observability hooks:
+    - PreToolUse: Security validation before tool execution
+    - PostToolUse: Log successful tool executions
+    - PostToolUseFailure: Log and handle failed tool executions
+    - PreCompact: Log context compaction events (SDK handles compaction automatically)
+    - Stop: Capture transcript path for stateless resume (when *on_stop* is provided)
+
+    Args:
+        user_id: Current user ID for isolation validation
+        sdk_cwd: SDK working directory for workspace-scoped tool validation
+        max_subtasks: Maximum Task (sub-agent) spawns allowed per session
+        on_stop: Callback ``(transcript_path, sdk_session_id)`` invoked when
+            the SDK finishes processing — used to read the JSONL transcript
+            before the CLI process exits.
+
+    Returns:
+        Hooks configuration dict for ClaudeAgentOptions
+    """
+    try:
+        from claude_agent_sdk import HookMatcher
+        from claude_agent_sdk.types import HookContext, HookInput, SyncHookJSONOutput
+
+        # Per-session counter for Task sub-agent spawns
+        task_spawn_count = 0
+
+        async def pre_tool_use_hook(
+            input_data: HookInput,
+            tool_use_id: str | None,
+            context: HookContext,
+        ) -> SyncHookJSONOutput:
+            """Combined pre-tool-use validation hook."""
+            nonlocal task_spawn_count
+            _ = context  # unused but required by signature
+            tool_name = cast(str, input_data.get("tool_name", ""))
+            tool_input = cast(dict[str, Any], input_data.get("tool_input", {}))
+
+            # Rate-limit Task (sub-agent) spawns per session
+            if tool_name == "Task":
+                task_spawn_count += 1
+                if task_spawn_count > max_subtasks:
+                    logger.warning(
+                        f"[SDK] Task limit reached ({max_subtasks}), user={user_id}"
+                    )
+                    return cast(
+                        SyncHookJSONOutput,
+                        _deny(
+                            f"Maximum {max_subtasks} sub-tasks per session. "
+                            "Please continue in the main conversation."
+                        ),
+                    )
+
+            # Strip MCP prefix for consistent validation
+            is_copilot_tool = tool_name.startswith(MCP_TOOL_PREFIX)
+            clean_name = tool_name.removeprefix(MCP_TOOL_PREFIX)
+
+            # Only block non-CoPilot tools; our MCP-registered tools
+            # (including Read for oversized results) are already sandboxed.
+            if not is_copilot_tool:
+                result = _validate_tool_access(clean_name, tool_input, sdk_cwd)
+                if result:
+                    return cast(SyncHookJSONOutput, result)
+
+            # Validate user isolation
+            result = _validate_user_isolation(clean_name, tool_input, user_id)
+            if result:
+                return cast(SyncHookJSONOutput, result)
+
+            logger.debug(f"[SDK] Tool start: {tool_name}, user={user_id}")
+            return cast(SyncHookJSONOutput, {})
+
+        async def post_tool_use_hook(
+            input_data: HookInput,
+            tool_use_id: str | None,
+            context: HookContext,
+        ) -> SyncHookJSONOutput:
+            """Log successful tool executions for observability."""
+            _ = context
+            tool_name = cast(str, input_data.get("tool_name", ""))
+            logger.debug(f"[SDK] Tool success: {tool_name}, tool_use_id={tool_use_id}")
+            return cast(SyncHookJSONOutput, {})
+
+        async def post_tool_failure_hook(
+            input_data: HookInput,
+            tool_use_id: str | None,
+            context: HookContext,
+        ) -> SyncHookJSONOutput:
+            """Log failed tool executions for debugging."""
+            _ = context
+            tool_name = cast(str, input_data.get("tool_name", ""))
+            error = input_data.get("error", "Unknown error")
+            logger.warning(
+                f"[SDK] Tool failed: {tool_name}, error={error}, "
+                f"user={user_id}, tool_use_id={tool_use_id}"
+            )
+            return cast(SyncHookJSONOutput, {})
+
+        async def pre_compact_hook(
+            input_data: HookInput,
+            tool_use_id: str | None,
+            context: HookContext,
+        ) -> SyncHookJSONOutput:
+            """Log when SDK triggers context compaction.
+
+            The SDK automatically compacts conversation history when it grows too large.
+            This hook provides visibility into when compaction happens.
+            """
+            _ = context, tool_use_id
+            trigger = input_data.get("trigger", "auto")
+            logger.info(
+                f"[SDK] Context compaction triggered: {trigger}, user={user_id}"
+            )
+            return cast(SyncHookJSONOutput, {})
+
+        # --- Stop hook: capture transcript path for stateless resume ---
+        async def stop_hook(
+            input_data: HookInput,
+            tool_use_id: str | None,
+            context: HookContext,
+        ) -> SyncHookJSONOutput:
+            """Capture transcript path when SDK finishes processing.
+
+            The Stop hook fires while the CLI process is still alive, giving us
+            a reliable window to read the JSONL transcript before SIGTERM.
+            """
+            _ = context, tool_use_id
+            transcript_path = cast(str, input_data.get("transcript_path", ""))
+            sdk_session_id = cast(str, input_data.get("session_id", ""))
+
+            if transcript_path and on_stop:
+                logger.info(
+                    f"[SDK] Stop hook: transcript_path={transcript_path}, "
+                    f"sdk_session_id={sdk_session_id[:12]}..."
+                )
+                on_stop(transcript_path, sdk_session_id)
+
+            return cast(SyncHookJSONOutput, {})
+
+        hooks: dict[str, Any] = {
+            "PreToolUse": [HookMatcher(matcher="*", hooks=[pre_tool_use_hook])],
+            "PostToolUse": [HookMatcher(matcher="*", hooks=[post_tool_use_hook])],
+            "PostToolUseFailure": [
+                HookMatcher(matcher="*", hooks=[post_tool_failure_hook])
+            ],
+            "PreCompact": [HookMatcher(matcher="*", hooks=[pre_compact_hook])],
+        }
+
+        if on_stop is not None:
+            hooks["Stop"] = [HookMatcher(matcher=None, hooks=[stop_hook])]
+
+        return hooks
+    except ImportError:
+        # Fallback for when SDK isn't available - return empty hooks
+        logger.warning("claude-agent-sdk not available, security hooks disabled")
+        return {}

autogpt_platform/backend/backend/api/features/chat/sdk/security_hooks_test.py

@@ -0,0 +1,165 @@
+"""Unit tests for SDK security hooks."""
+
+import os
+
+from .security_hooks import _validate_tool_access, _validate_user_isolation
+
+SDK_CWD = "/tmp/copilot-abc123"
+
+
+def _is_denied(result: dict) -> bool:
+    hook = result.get("hookSpecificOutput", {})
+    return hook.get("permissionDecision") == "deny"
+
+
+# -- Blocked tools -----------------------------------------------------------
+
+
+def test_blocked_tools_denied():
+    for tool in ("bash", "shell", "exec", "terminal", "command"):
+        result = _validate_tool_access(tool, {})
+        assert _is_denied(result), f"{tool} should be blocked"
+
+
+def test_unknown_tool_allowed():
+    result = _validate_tool_access("SomeCustomTool", {})
+    assert result == {}
+
+
+# -- Workspace-scoped tools --------------------------------------------------
+
+
+def test_read_within_workspace_allowed():
+    result = _validate_tool_access(
+        "Read", {"file_path": f"{SDK_CWD}/file.txt"}, sdk_cwd=SDK_CWD
+    )
+    assert result == {}
+
+
+def test_write_within_workspace_allowed():
+    result = _validate_tool_access(
+        "Write", {"file_path": f"{SDK_CWD}/output.json"}, sdk_cwd=SDK_CWD
+    )
+    assert result == {}
+
+
+def test_edit_within_workspace_allowed():
+    result = _validate_tool_access(
+        "Edit", {"file_path": f"{SDK_CWD}/src/main.py"}, sdk_cwd=SDK_CWD
+    )
+    assert result == {}
+
+
+def test_glob_within_workspace_allowed():
+    result = _validate_tool_access("Glob", {"path": f"{SDK_CWD}/src"}, sdk_cwd=SDK_CWD)
+    assert result == {}
+
+
+def test_grep_within_workspace_allowed():
+    result = _validate_tool_access("Grep", {"path": f"{SDK_CWD}/src"}, sdk_cwd=SDK_CWD)
+    assert result == {}
+
+
+def test_read_outside_workspace_denied():
+    result = _validate_tool_access(
+        "Read", {"file_path": "/etc/passwd"}, sdk_cwd=SDK_CWD
+    )
+    assert _is_denied(result)
+
+
+def test_write_outside_workspace_denied():
+    result = _validate_tool_access(
+        "Write", {"file_path": "/home/user/secrets.txt"}, sdk_cwd=SDK_CWD
+    )
+    assert _is_denied(result)
+
+
+def test_traversal_attack_denied():
+    result = _validate_tool_access(
+        "Read",
+        {"file_path": f"{SDK_CWD}/../../etc/passwd"},
+        sdk_cwd=SDK_CWD,
+    )
+    assert _is_denied(result)
+
+
+def test_no_path_allowed():
+    """Glob/Grep without a path argument defaults to cwd — should pass."""
+    result = _validate_tool_access("Glob", {}, sdk_cwd=SDK_CWD)
+    assert result == {}
+
+
+def test_read_no_cwd_denies_absolute():
+    """If no sdk_cwd is set, absolute paths are denied."""
+    result = _validate_tool_access("Read", {"file_path": "/tmp/anything"})
+    assert _is_denied(result)
+
+
+# -- Tool-results directory --------------------------------------------------
+
+
+def test_read_tool_results_allowed():
+    home = os.path.expanduser("~")
+    path = f"{home}/.claude/projects/-tmp-copilot-abc123/tool-results/12345.txt"
+    result = _validate_tool_access("Read", {"file_path": path}, sdk_cwd=SDK_CWD)
+    assert result == {}
+
+
+def test_read_claude_projects_without_tool_results_denied():
+    home = os.path.expanduser("~")
+    path = f"{home}/.claude/projects/-tmp-copilot-abc123/settings.json"
+    result = _validate_tool_access("Read", {"file_path": path}, sdk_cwd=SDK_CWD)
+    assert _is_denied(result)
+
+
+# -- Built-in Bash is blocked (use bash_exec MCP tool instead) ---------------
+
+
+def test_bash_builtin_always_blocked():
+    """SDK built-in Bash is blocked — bash_exec MCP tool with bubblewrap is used instead."""
+    result = _validate_tool_access("Bash", {"command": "echo hello"}, sdk_cwd=SDK_CWD)
+    assert _is_denied(result)
+
+
+# -- Dangerous patterns ------------------------------------------------------
+
+
+def test_dangerous_pattern_blocked():
+    result = _validate_tool_access("SomeTool", {"cmd": "sudo rm -rf /"})
+    assert _is_denied(result)
+
+
+def test_subprocess_pattern_blocked():
+    result = _validate_tool_access("SomeTool", {"code": "subprocess.run(...)"})
+    assert _is_denied(result)
+
+
+# -- User isolation ----------------------------------------------------------
+
+
+def test_workspace_path_traversal_blocked():
+    result = _validate_user_isolation(
+        "workspace_read", {"path": "../../../etc/shadow"}, user_id="user-1"
+    )
+    assert _is_denied(result)
+
+
+def test_workspace_absolute_path_blocked():
+    result = _validate_user_isolation(
+        "workspace_read", {"path": "/etc/passwd"}, user_id="user-1"
+    )
+    assert _is_denied(result)
+
+
+def test_workspace_normal_path_allowed():
+    result = _validate_user_isolation(
+        "workspace_read", {"path": "src/main.py"}, user_id="user-1"
+    )
+    assert result == {}
+
+
+def test_non_workspace_tool_passes_isolation():
+    result = _validate_user_isolation(
+        "find_agent", {"query": "email"}, user_id="user-1"
+    )
+    assert result == {}

autogpt_platform/backend/backend/api/features/chat/sdk/service.py

@@ -0,0 +1,752 @@
+"""Claude Agent SDK service layer for CoPilot chat completions."""
+
+import asyncio
+import json
+import logging
+import os
+import uuid
+from collections.abc import AsyncGenerator
+from dataclasses import dataclass
+from typing import Any
+
+from backend.util.exceptions import NotFoundError
+
+from .. import stream_registry
+from ..config import ChatConfig
+from ..model import (
+    ChatMessage,
+    ChatSession,
+    get_chat_session,
+    update_session_title,
+    upsert_chat_session,
+)
+from ..response_model import (
+    StreamBaseResponse,
+    StreamError,
+    StreamFinish,
+    StreamStart,
+    StreamTextDelta,
+    StreamToolInputAvailable,
+    StreamToolOutputAvailable,
+)
+from ..service import (
+    _build_system_prompt,
+    _execute_long_running_tool_with_streaming,
+    _generate_session_title,
+)
+from ..tools.models import OperationPendingResponse, OperationStartedResponse
+from ..tools.sandbox import WORKSPACE_PREFIX, make_session_path
+from ..tracking import track_user_message
+from .response_adapter import SDKResponseAdapter
+from .security_hooks import create_security_hooks
+from .tool_adapter import (
+    COPILOT_TOOL_NAMES,
+    SDK_DISALLOWED_TOOLS,
+    LongRunningCallback,
+    create_copilot_mcp_server,
+    set_execution_context,
+)
+from .transcript import (
+    download_transcript,
+    read_transcript_file,
+    upload_transcript,
+    validate_transcript,
+    write_transcript_to_tempfile,
+)
+
+logger = logging.getLogger(__name__)
+config = ChatConfig()
+
+# Set to hold background tasks to prevent garbage collection
+_background_tasks: set[asyncio.Task[Any]] = set()
+
+
+@dataclass
+class CapturedTranscript:
+    """Info captured by the SDK Stop hook for stateless --resume."""
+
+    path: str = ""
+    sdk_session_id: str = ""
+
+    @property
+    def available(self) -> bool:
+        return bool(self.path)
+
+
+_SDK_CWD_PREFIX = WORKSPACE_PREFIX
+
+# Appended to the system prompt to inform the agent about available tools.
+# The SDK built-in Bash is NOT available — use mcp__copilot__bash_exec instead,
+# which has kernel-level network isolation (unshare --net).
+_SDK_TOOL_SUPPLEMENT = """
+
+## Tool notes
+
+- The SDK built-in Bash tool is NOT available.  Use the `bash_exec` MCP tool
+  for shell commands — it runs in a network-isolated sandbox.
+- **Shared workspace**: The SDK Read/Write tools and `bash_exec` share the
+  same working directory. Files created by one are readable by the other.
+  These files are **ephemeral** — they exist only for the current session.
+- **Persistent storage**: Use `write_workspace_file` / `read_workspace_file`
+  for files that should persist across sessions (stored in cloud storage).
+- Long-running tools (create_agent, edit_agent, etc.) are handled
+  asynchronously.  You will receive an immediate response; the actual result
+  is delivered to the user via a background stream.
+"""
+
+
+def _build_long_running_callback(user_id: str | None) -> LongRunningCallback:
+    """Build a callback that delegates long-running tools to the non-SDK infrastructure.
+
+    Long-running tools (create_agent, edit_agent, etc.) are delegated to the
+    existing background infrastructure: stream_registry (Redis Streams),
+    database persistence, and SSE reconnection.  This means results survive
+    page refreshes / pod restarts, and the frontend shows the proper loading
+    widget with progress updates.
+
+    The returned callback matches the ``LongRunningCallback`` signature:
+    ``(tool_name, args, session) -> MCP response dict``.
+    """
+
+    async def _callback(
+        tool_name: str, args: dict[str, Any], session: ChatSession
+    ) -> dict[str, Any]:
+        operation_id = str(uuid.uuid4())
+        task_id = str(uuid.uuid4())
+        tool_call_id = f"sdk-{uuid.uuid4().hex[:12]}"
+        session_id = session.session_id
+
+        # --- Build user-friendly messages (matches non-SDK service) ---
+        if tool_name == "create_agent":
+            desc = args.get("description", "")
+            desc_preview = (desc[:100] + "...") if len(desc) > 100 else desc
+            pending_msg = (
+                f"Creating your agent: {desc_preview}"
+                if desc_preview
+                else "Creating agent... This may take a few minutes."
+            )
+            started_msg = (
+                "Agent creation started. You can close this tab - "
+                "check your library in a few minutes."
+            )
+        elif tool_name == "edit_agent":
+            changes = args.get("changes", "")
+            changes_preview = (changes[:100] + "...") if len(changes) > 100 else changes
+            pending_msg = (
+                f"Editing agent: {changes_preview}"
+                if changes_preview
+                else "Editing agent... This may take a few minutes."
+            )
+            started_msg = (
+                "Agent edit started. You can close this tab - "
+                "check your library in a few minutes."
+            )
+        else:
+            pending_msg = f"Running {tool_name}... This may take a few minutes."
+            started_msg = (
+                f"{tool_name} started. You can close this tab - "
+                "check back in a few minutes."
+            )
+
+        # --- Register task in Redis for SSE reconnection ---
+        await stream_registry.create_task(
+            task_id=task_id,
+            session_id=session_id,
+            user_id=user_id,
+            tool_call_id=tool_call_id,
+            tool_name=tool_name,
+            operation_id=operation_id,
+        )
+
+        # --- Save OperationPendingResponse to chat history ---
+        pending_message = ChatMessage(
+            role="tool",
+            content=OperationPendingResponse(
+                message=pending_msg,
+                operation_id=operation_id,
+                tool_name=tool_name,
+            ).model_dump_json(),
+            tool_call_id=tool_call_id,
+        )
+        session.messages.append(pending_message)
+        await upsert_chat_session(session)
+
+        # --- Spawn background task (reuses non-SDK infrastructure) ---
+        bg_task = asyncio.create_task(
+            _execute_long_running_tool_with_streaming(
+                tool_name=tool_name,
+                parameters=args,
+                tool_call_id=tool_call_id,
+                operation_id=operation_id,
+                task_id=task_id,
+                session_id=session_id,
+                user_id=user_id,
+            )
+        )
+        _background_tasks.add(bg_task)
+        bg_task.add_done_callback(_background_tasks.discard)
+        await stream_registry.set_task_asyncio_task(task_id, bg_task)
+
+        logger.info(
+            f"[SDK] Long-running tool {tool_name} delegated to background "
+            f"(operation_id={operation_id}, task_id={task_id})"
+        )
+
+        # --- Return OperationStartedResponse as MCP tool result ---
+        # This flows through SDK → response adapter → frontend, triggering
+        # the loading widget with SSE reconnection support.
+        started_json = OperationStartedResponse(
+            message=started_msg,
+            operation_id=operation_id,
+            tool_name=tool_name,
+            task_id=task_id,
+        ).model_dump_json()
+
+        return {
+            "content": [{"type": "text", "text": started_json}],
+            "isError": False,
+        }
+
+    return _callback
+
+
+def _resolve_sdk_model() -> str | None:
+    """Resolve the model name for the Claude Agent SDK CLI.
+
+    Uses ``config.claude_agent_model`` if set, otherwise derives from
+    ``config.model`` by stripping the OpenRouter provider prefix (e.g.,
+    ``"anthropic/claude-opus-4.6"`` → ``"claude-opus-4.6"``).
+    """
+    if config.claude_agent_model:
+        return config.claude_agent_model
+    model = config.model
+    if "/" in model:
+        return model.split("/", 1)[1]
+    return model
+
+
+def _build_sdk_env() -> dict[str, str]:
+    """Build env vars for the SDK CLI process.
+
+    Routes API calls through OpenRouter (or a custom base_url) using
+    the same ``config.api_key`` / ``config.base_url`` as the non-SDK path.
+    This gives per-call token and cost tracking on the OpenRouter dashboard.
+
+    Only overrides ``ANTHROPIC_API_KEY`` when a valid proxy URL and auth
+    token are both present — otherwise returns an empty dict so the SDK
+    falls back to its default credentials.
+    """
+    env: dict[str, str] = {}
+    if config.api_key and config.base_url:
+        # Strip /v1 suffix — SDK expects the base URL without a version path
+        base = config.base_url.rstrip("/")
+        if base.endswith("/v1"):
+            base = base[:-3]
+        if not base or not base.startswith("http"):
+            # Invalid base_url — don't override SDK defaults
+            return env
+        env["ANTHROPIC_BASE_URL"] = base
+        env["ANTHROPIC_AUTH_TOKEN"] = config.api_key
+        # Must be explicitly empty so the CLI uses AUTH_TOKEN instead
+        env["ANTHROPIC_API_KEY"] = ""
+    return env
+
+
+def _make_sdk_cwd(session_id: str) -> str:
+    """Create a safe, session-specific working directory path.
+
+    Delegates to :func:`~backend.api.features.chat.tools.sandbox.make_session_path`
+    (single source of truth for path sanitization) and adds a defence-in-depth
+    assertion.
+    """
+    cwd = make_session_path(session_id)
+    # Defence-in-depth: normpath + startswith is a CodeQL-recognised sanitizer
+    cwd = os.path.normpath(cwd)
+    if not cwd.startswith(_SDK_CWD_PREFIX):
+        raise ValueError(f"SDK cwd escaped prefix: {cwd}")
+    return cwd
+
+
+def _cleanup_sdk_tool_results(cwd: str) -> None:
+    """Remove SDK tool-result files for a specific session working directory.
+
+    The SDK creates tool-result files under ~/.claude/projects/<encoded-cwd>/tool-results/.
+    We clean only the specific cwd's results to avoid race conditions between
+    concurrent sessions.
+
+    Security: cwd MUST be created by _make_sdk_cwd() which sanitizes session_id.
+    """
+    import shutil
+
+    # Validate cwd is under the expected prefix
+    normalized = os.path.normpath(cwd)
+    if not normalized.startswith(_SDK_CWD_PREFIX):
+        logger.warning(f"[SDK] Rejecting cleanup for path outside workspace: {cwd}")
+        return
+
+    # SDK encodes the cwd path by replacing '/' with '-'
+    encoded_cwd = normalized.replace("/", "-")
+
+    # Construct the project directory path (known-safe home expansion)
+    claude_projects = os.path.expanduser("~/.claude/projects")
+    project_dir = os.path.join(claude_projects, encoded_cwd)
+
+    # Security check 3: Validate project_dir is under ~/.claude/projects
+    project_dir = os.path.normpath(project_dir)
+    if not project_dir.startswith(claude_projects):
+        logger.warning(
+            f"[SDK] Rejecting cleanup for escaped project path: {project_dir}"
+        )
+        return
+
+    results_dir = os.path.join(project_dir, "tool-results")
+    if os.path.isdir(results_dir):
+        for filename in os.listdir(results_dir):
+            file_path = os.path.join(results_dir, filename)
+            try:
+                if os.path.isfile(file_path):
+                    os.remove(file_path)
+            except OSError:
+                pass
+
+    # Also clean up the temp cwd directory itself
+    try:
+        shutil.rmtree(normalized, ignore_errors=True)
+    except OSError:
+        pass
+
+
+async def _compress_conversation_history(
+    session: ChatSession,
+) -> list[ChatMessage]:
+    """Compress prior conversation messages if they exceed the token threshold.
+
+    Uses the shared compress_context() from prompt.py which supports:
+    - LLM summarization of old messages (keeps recent ones intact)
+    - Progressive content truncation as fallback
+    - Middle-out deletion as last resort
+
+    Returns the compressed prior messages (everything except the current message).
+    """
+    prior = session.messages[:-1]
+    if len(prior) < 2:
+        return prior
+
+    from backend.util.prompt import compress_context
+
+    # Convert ChatMessages to dicts for compress_context
+    messages_dict = []
+    for msg in prior:
+        msg_dict: dict[str, Any] = {"role": msg.role}
+        if msg.content:
+            msg_dict["content"] = msg.content
+        if msg.tool_calls:
+            msg_dict["tool_calls"] = msg.tool_calls
+        if msg.tool_call_id:
+            msg_dict["tool_call_id"] = msg.tool_call_id
+        messages_dict.append(msg_dict)
+
+    try:
+        import openai
+
+        async with openai.AsyncOpenAI(
+            api_key=config.api_key, base_url=config.base_url, timeout=30.0
+        ) as client:
+            result = await compress_context(
+                messages=messages_dict,
+                model=config.model,
+                client=client,
+            )
+    except Exception as e:
+        logger.warning(f"[SDK] Context compression with LLM failed: {e}")
+        # Fall back to truncation-only (no LLM summarization)
+        result = await compress_context(
+            messages=messages_dict,
+            model=config.model,
+            client=None,
+        )
+
+    if result.was_compacted:
+        logger.info(
+            f"[SDK] Context compacted: {result.original_token_count} -> "
+            f"{result.token_count} tokens "
+            f"({result.messages_summarized} summarized, "
+            f"{result.messages_dropped} dropped)"
+        )
+        # Convert compressed dicts back to ChatMessages
+        return [
+            ChatMessage(
+                role=m["role"],
+                content=m.get("content"),
+                tool_calls=m.get("tool_calls"),
+                tool_call_id=m.get("tool_call_id"),
+            )
+            for m in result.messages
+        ]
+
+    return prior
+
+
+def _format_conversation_context(messages: list[ChatMessage]) -> str | None:
+    """Format conversation messages into a context prefix for the user message.
+
+    Returns a string like:
+        <conversation_history>
+        User: hello
+        You responded: Hi! How can I help?
+        </conversation_history>
+
+    Returns None if there are no messages to format.
+    """
+    if not messages:
+        return None
+
+    lines: list[str] = []
+    for msg in messages:
+        if not msg.content:
+            continue
+        if msg.role == "user":
+            lines.append(f"User: {msg.content}")
+        elif msg.role == "assistant":
+            lines.append(f"You responded: {msg.content}")
+        # Skip tool messages — they're internal details
+
+    if not lines:
+        return None
+
+    return "<conversation_history>\n" + "\n".join(lines) + "\n</conversation_history>"
+
+
+async def stream_chat_completion_sdk(
+    session_id: str,
+    message: str | None = None,
+    tool_call_response: str | None = None,  # noqa: ARG001
+    is_user_message: bool = True,
+    user_id: str | None = None,
+    retry_count: int = 0,  # noqa: ARG001
+    session: ChatSession | None = None,
+    context: dict[str, str] | None = None,  # noqa: ARG001
+) -> AsyncGenerator[StreamBaseResponse, None]:
+    """Stream chat completion using Claude Agent SDK.
+
+    Drop-in replacement for stream_chat_completion with improved reliability.
+    """
+
+    if session is None:
+        session = await get_chat_session(session_id, user_id)
+
+    if not session:
+        raise NotFoundError(
+            f"Session {session_id} not found. Please create a new session first."
+        )
+
+    if message:
+        session.messages.append(
+            ChatMessage(
+                role="user" if is_user_message else "assistant", content=message
+            )
+        )
+        if is_user_message:
+            track_user_message(
+                user_id=user_id, session_id=session_id, message_length=len(message)
+            )
+
+    session = await upsert_chat_session(session)
+
+    # Generate title for new sessions (first user message)
+    if is_user_message and not session.title:
+        user_messages = [m for m in session.messages if m.role == "user"]
+        if len(user_messages) == 1:
+            first_message = user_messages[0].content or message or ""
+            if first_message:
+                task = asyncio.create_task(
+                    _update_title_async(session_id, first_message, user_id)
+                )
+                _background_tasks.add(task)
+                task.add_done_callback(_background_tasks.discard)
+
+    # Build system prompt (reuses non-SDK path with Langfuse support)
+    has_history = len(session.messages) > 1
+    system_prompt, _ = await _build_system_prompt(
+        user_id, has_conversation_history=has_history
+    )
+    system_prompt += _SDK_TOOL_SUPPLEMENT
+    message_id = str(uuid.uuid4())
+    task_id = str(uuid.uuid4())
+
+    yield StreamStart(messageId=message_id, taskId=task_id)
+
+    stream_completed = False
+    # Initialise sdk_cwd before the try so the finally can reference it
+    # even if _make_sdk_cwd raises (in that case it stays as "").
+    sdk_cwd = ""
+    use_resume = False
+
+    try:
+        # Use a session-specific temp dir to avoid cleanup race conditions
+        # between concurrent sessions.
+        sdk_cwd = _make_sdk_cwd(session_id)
+        os.makedirs(sdk_cwd, exist_ok=True)
+
+        set_execution_context(
+            user_id,
+            session,
+            long_running_callback=_build_long_running_callback(user_id),
+        )
+        try:
+            from claude_agent_sdk import ClaudeAgentOptions, ClaudeSDKClient
+
+            # Fail fast when no API credentials are available at all
+            sdk_env = _build_sdk_env()
+            if not sdk_env and not os.environ.get("ANTHROPIC_API_KEY"):
+                raise RuntimeError(
+                    "No API key configured. Set OPEN_ROUTER_API_KEY "
+                    "(or CHAT_API_KEY) for OpenRouter routing, "
+                    "or ANTHROPIC_API_KEY for direct Anthropic access."
+                )
+
+            mcp_server = create_copilot_mcp_server()
+
+            sdk_model = _resolve_sdk_model()
+
+            # --- Transcript capture via Stop hook ---
+            captured_transcript = CapturedTranscript()
+
+            def _on_stop(transcript_path: str, sdk_session_id: str) -> None:
+                captured_transcript.path = transcript_path
+                captured_transcript.sdk_session_id = sdk_session_id
+
+            security_hooks = create_security_hooks(
+                user_id,
+                sdk_cwd=sdk_cwd,
+                max_subtasks=config.claude_agent_max_subtasks,
+                on_stop=_on_stop if config.claude_agent_use_resume else None,
+            )
+
+            # --- Resume strategy: download transcript from bucket ---
+            resume_file: str | None = None
+            use_resume = False
+
+            if config.claude_agent_use_resume and user_id and len(session.messages) > 1:
+                transcript_content = await download_transcript(user_id, session_id)
+                if transcript_content and validate_transcript(transcript_content):
+                    resume_file = write_transcript_to_tempfile(
+                        transcript_content, session_id, sdk_cwd
+                    )
+                    if resume_file:
+                        use_resume = True
+                        logger.info(
+                            f"[SDK] Using --resume with transcript "
+                            f"({len(transcript_content)} bytes)"
+                        )
+
+            sdk_options_kwargs: dict[str, Any] = {
+                "system_prompt": system_prompt,
+                "mcp_servers": {"copilot": mcp_server},
+                "allowed_tools": COPILOT_TOOL_NAMES,
+                "disallowed_tools": SDK_DISALLOWED_TOOLS,
+                "hooks": security_hooks,
+                "cwd": sdk_cwd,
+                "max_buffer_size": config.claude_agent_max_buffer_size,
+            }
+            if sdk_env:
+                sdk_options_kwargs["model"] = sdk_model
+                sdk_options_kwargs["env"] = sdk_env
+            if use_resume and resume_file:
+                sdk_options_kwargs["resume"] = resume_file
+
+            options = ClaudeAgentOptions(**sdk_options_kwargs)  # type: ignore[arg-type]
+
+            adapter = SDKResponseAdapter(message_id=message_id)
+            adapter.set_task_id(task_id)
+
+            async with ClaudeSDKClient(options=options) as client:
+                current_message = message or ""
+                if not current_message and session.messages:
+                    last_user = [m for m in session.messages if m.role == "user"]
+                    if last_user:
+                        current_message = last_user[-1].content or ""
+
+                if not current_message.strip():
+                    yield StreamError(
+                        errorText="Message cannot be empty.",
+                        code="empty_prompt",
+                    )
+                    yield StreamFinish()
+                    return
+
+                # Build query: with --resume the CLI already has full
+                # context, so we only send the new message.  Without
+                # resume, compress history into a context prefix.
+                query_message = current_message
+                if not use_resume and len(session.messages) > 1:
+                    logger.warning(
+                        f"[SDK] Using compression fallback for session "
+                        f"{session_id} ({len(session.messages)} messages) — "
+                        f"no transcript available for --resume"
+                    )
+                    compressed = await _compress_conversation_history(session)
+                    history_context = _format_conversation_context(compressed)
+                    if history_context:
+                        query_message = (
+                            f"{history_context}\n\n"
+                            f"Now, the user says:\n{current_message}"
+                        )
+
+                logger.info(
+                    f"[SDK] Sending query ({len(session.messages)} msgs in session)"
+                )
+                logger.debug(f"[SDK] Query preview: {current_message[:80]!r}")
+                await client.query(query_message, session_id=session_id)
+
+                assistant_response = ChatMessage(role="assistant", content="")
+                accumulated_tool_calls: list[dict[str, Any]] = []
+                has_appended_assistant = False
+                has_tool_results = False
+
+                async for sdk_msg in client.receive_messages():
+                    logger.debug(
+                        f"[SDK] Received: {type(sdk_msg).__name__} "
+                        f"{getattr(sdk_msg, 'subtype', '')}"
+                    )
+                    for response in adapter.convert_message(sdk_msg):
+                        if isinstance(response, StreamStart):
+                            continue
+
+                        yield response
+
+                        if isinstance(response, StreamTextDelta):
+                            delta = response.delta or ""
+                            # After tool results, start a new assistant
+                            # message for the post-tool text.
+                            if has_tool_results and has_appended_assistant:
+                                assistant_response = ChatMessage(
+                                    role="assistant", content=delta
+                                )
+                                accumulated_tool_calls = []
+                                has_appended_assistant = False
+                                has_tool_results = False
+                                session.messages.append(assistant_response)
+                                has_appended_assistant = True
+                            else:
+                                assistant_response.content = (
+                                    assistant_response.content or ""
+                                ) + delta
+                                if not has_appended_assistant:
+                                    session.messages.append(assistant_response)
+                                    has_appended_assistant = True
+
+                        elif isinstance(response, StreamToolInputAvailable):
+                            accumulated_tool_calls.append(
+                                {
+                                    "id": response.toolCallId,
+                                    "type": "function",
+                                    "function": {
+                                        "name": response.toolName,
+                                        "arguments": json.dumps(response.input or {}),
+                                    },
+                                }
+                            )
+                            assistant_response.tool_calls = accumulated_tool_calls
+                            if not has_appended_assistant:
+                                session.messages.append(assistant_response)
+                                has_appended_assistant = True
+
+                        elif isinstance(response, StreamToolOutputAvailable):
+                            session.messages.append(
+                                ChatMessage(
+                                    role="tool",
+                                    content=(
+                                        response.output
+                                        if isinstance(response.output, str)
+                                        else str(response.output)
+                                    ),
+                                    tool_call_id=response.toolCallId,
+                                )
+                            )
+                            has_tool_results = True
+
+                        elif isinstance(response, StreamFinish):
+                            stream_completed = True
+
+                    if stream_completed:
+                        break
+
+                if (
+                    assistant_response.content or assistant_response.tool_calls
+                ) and not has_appended_assistant:
+                    session.messages.append(assistant_response)
+
+                # --- Capture transcript while CLI is still alive ---
+                # Must happen INSIDE async with: close() sends SIGTERM
+                # which kills the CLI before it can flush the JSONL.
+                if (
+                    config.claude_agent_use_resume
+                    and user_id
+                    and captured_transcript.available
+                ):
+                    # Give CLI time to flush JSONL writes before we read
+                    await asyncio.sleep(0.5)
+                    raw_transcript = read_transcript_file(captured_transcript.path)
+                    if raw_transcript:
+                        task = asyncio.create_task(
+                            _upload_transcript_bg(user_id, session_id, raw_transcript)
+                        )
+                        _background_tasks.add(task)
+                        task.add_done_callback(_background_tasks.discard)
+                    else:
+                        logger.debug("[SDK] Stop hook fired but transcript not usable")
+
+        except ImportError:
+            raise RuntimeError(
+                "claude-agent-sdk is not installed. "
+                "Disable SDK mode (CHAT_USE_CLAUDE_AGENT_SDK=false) "
+                "to use the OpenAI-compatible fallback."
+            )
+
+        await upsert_chat_session(session)
+        logger.debug(
+            f"[SDK] Session {session_id} saved with {len(session.messages)} messages"
+        )
+        if not stream_completed:
+            yield StreamFinish()
+
+    except Exception as e:
+        logger.error(f"[SDK] Error: {e}", exc_info=True)
+        try:
+            await upsert_chat_session(session)
+        except Exception as save_err:
+            logger.error(f"[SDK] Failed to save session on error: {save_err}")
+        yield StreamError(
+            errorText="An error occurred. Please try again.",
+            code="sdk_error",
+        )
+        yield StreamFinish()
+    finally:
+        if sdk_cwd:
+            _cleanup_sdk_tool_results(sdk_cwd)
+
+
+async def _upload_transcript_bg(
+    user_id: str, session_id: str, raw_content: str
+) -> None:
+    """Background task to strip progress entries and upload transcript."""
+    try:
+        await upload_transcript(user_id, session_id, raw_content)
+    except Exception as e:
+        logger.error(f"[SDK] Failed to upload transcript for {session_id}: {e}")
+
+
+async def _update_title_async(
+    session_id: str, message: str, user_id: str | None = None
+) -> None:
+    """Background task to update session title."""
+    try:
+        title = await _generate_session_title(
+            message, user_id=user_id, session_id=session_id
+        )
+        if title:
+            await update_session_title(session_id, title)
+            logger.debug(f"[SDK] Generated title for {session_id}: {title}")
+    except Exception as e:
+        logger.warning(f"[SDK] Failed to update session title: {e}")

autogpt_platform/backend/backend/api/features/chat/sdk/tool_adapter.py

@@ -0,0 +1,363 @@
+"""Tool adapter for wrapping existing CoPilot tools as Claude Agent SDK MCP tools.
+
+This module provides the adapter layer that converts existing BaseTool implementations
+into in-process MCP tools that can be used with the Claude Agent SDK.
+
+Long-running tools (``is_long_running=True``) are delegated to the non-SDK
+background infrastructure (stream_registry, Redis persistence, SSE reconnection)
+via a callback provided by the service layer.  This avoids wasteful SDK polling
+and makes results survive page refreshes.
+"""
+
+import itertools
+import json
+import logging
+import os
+import uuid
+from collections.abc import Awaitable, Callable
+from contextvars import ContextVar
+from typing import Any
+
+from backend.api.features.chat.model import ChatSession
+from backend.api.features.chat.tools import TOOL_REGISTRY
+from backend.api.features.chat.tools.base import BaseTool
+
+logger = logging.getLogger(__name__)
+
+# Allowed base directory for the Read tool (SDK saves oversized tool results here).
+# Restricted to ~/.claude/projects/ and further validated to require "tool-results"
+# in the path — prevents reading settings, credentials, or other sensitive files.
+_SDK_PROJECTS_DIR = os.path.expanduser("~/.claude/projects/")
+
+# MCP server naming - the SDK prefixes tool names as "mcp__{server_name}__{tool}"
+MCP_SERVER_NAME = "copilot"
+MCP_TOOL_PREFIX = f"mcp__{MCP_SERVER_NAME}__"
+
+# Context variables to pass user/session info to tool execution
+_current_user_id: ContextVar[str | None] = ContextVar("current_user_id", default=None)
+_current_session: ContextVar[ChatSession | None] = ContextVar(
+    "current_session", default=None
+)
+# Stash for MCP tool outputs before the SDK potentially truncates them.
+# Keyed by tool_name → full output string. Consumed (popped) by the
+# response adapter when it builds StreamToolOutputAvailable.
+_pending_tool_outputs: ContextVar[dict[str, str]] = ContextVar(
+    "pending_tool_outputs", default=None  # type: ignore[arg-type]
+)
+
+# Callback type for delegating long-running tools to the non-SDK infrastructure.
+# Args: (tool_name, arguments, session) → MCP-formatted response dict.
+LongRunningCallback = Callable[
+    [str, dict[str, Any], ChatSession], Awaitable[dict[str, Any]]
+]
+
+# ContextVar so the service layer can inject the callback per-request.
+_long_running_callback: ContextVar[LongRunningCallback | None] = ContextVar(
+    "long_running_callback", default=None
+)
+
+
+def set_execution_context(
+    user_id: str | None,
+    session: ChatSession,
+    long_running_callback: LongRunningCallback | None = None,
+) -> None:
+    """Set the execution context for tool calls.
+
+    This must be called before streaming begins to ensure tools have access
+    to user_id and session information.
+
+    Args:
+        user_id: Current user's ID.
+        session: Current chat session.
+        long_running_callback: Optional callback to delegate long-running tools
+            to the non-SDK background infrastructure (stream_registry + Redis).
+    """
+    _current_user_id.set(user_id)
+    _current_session.set(session)
+    _pending_tool_outputs.set({})
+    _long_running_callback.set(long_running_callback)
+
+
+def get_execution_context() -> tuple[str | None, ChatSession | None]:
+    """Get the current execution context."""
+    return (
+        _current_user_id.get(),
+        _current_session.get(),
+    )
+
+
+def pop_pending_tool_output(tool_name: str) -> str | None:
+    """Pop and return the stashed full output for *tool_name*.
+
+    The SDK CLI may truncate large tool results (writing them to disk and
+    replacing the content with a file reference). This stash keeps the
+    original MCP output so the response adapter can forward it to the
+    frontend for proper widget rendering.
+
+    Returns ``None`` if nothing was stashed for *tool_name*.
+    """
+    pending = _pending_tool_outputs.get(None)
+    if pending is None:
+        return None
+    return pending.pop(tool_name, None)
+
+
+async def _execute_tool_sync(
+    base_tool: BaseTool,
+    user_id: str | None,
+    session: ChatSession,
+    args: dict[str, Any],
+) -> dict[str, Any]:
+    """Execute a tool synchronously and return MCP-formatted response."""
+    effective_id = f"sdk-{uuid.uuid4().hex[:12]}"
+    result = await base_tool.execute(
+        user_id=user_id,
+        session=session,
+        tool_call_id=effective_id,
+        **args,
+    )
+
+    text = (
+        result.output if isinstance(result.output, str) else json.dumps(result.output)
+    )
+
+    # Stash the full output before the SDK potentially truncates it.
+    pending = _pending_tool_outputs.get(None)
+    if pending is not None:
+        pending[base_tool.name] = text
+
+    return {
+        "content": [{"type": "text", "text": text}],
+        "isError": not result.success,
+    }
+
+
+def _mcp_error(message: str) -> dict[str, Any]:
+    return {
+        "content": [
+            {"type": "text", "text": json.dumps({"error": message, "type": "error"})}
+        ],
+        "isError": True,
+    }
+
+
+def create_tool_handler(base_tool: BaseTool):
+    """Create an async handler function for a BaseTool.
+
+    This wraps the existing BaseTool._execute method to be compatible
+    with the Claude Agent SDK MCP tool format.
+
+    Long-running tools (``is_long_running=True``) are delegated to the
+    non-SDK background infrastructure via a callback set in the execution
+    context.  The callback persists the operation in Redis (stream_registry)
+    so results survive page refreshes and pod restarts.
+    """
+
+    async def tool_handler(args: dict[str, Any]) -> dict[str, Any]:
+        """Execute the wrapped tool and return MCP-formatted response."""
+        user_id, session = get_execution_context()
+
+        if session is None:
+            return _mcp_error("No session context available")
+
+        # --- Long-running: delegate to non-SDK background infrastructure ---
+        if base_tool.is_long_running:
+            callback = _long_running_callback.get(None)
+            if callback:
+                try:
+                    return await callback(base_tool.name, args, session)
+                except Exception as e:
+                    logger.error(
+                        f"Long-running callback failed for {base_tool.name}: {e}",
+                        exc_info=True,
+                    )
+                    return _mcp_error(f"Failed to start {base_tool.name}: {e}")
+            # No callback — fall through to synchronous execution
+            logger.warning(
+                f"[SDK] No long-running callback for {base_tool.name}, "
+                f"executing synchronously (may block)"
+            )
+
+        # --- Normal (fast) tool: execute synchronously ---
+        try:
+            return await _execute_tool_sync(base_tool, user_id, session, args)
+        except Exception as e:
+            logger.error(f"Error executing tool {base_tool.name}: {e}", exc_info=True)
+            return _mcp_error(f"Failed to execute {base_tool.name}: {e}")
+
+    return tool_handler
+
+
+def _build_input_schema(base_tool: BaseTool) -> dict[str, Any]:
+    """Build a JSON Schema input schema for a tool."""
+    return {
+        "type": "object",
+        "properties": base_tool.parameters.get("properties", {}),
+        "required": base_tool.parameters.get("required", []),
+    }
+
+
+async def _read_file_handler(args: dict[str, Any]) -> dict[str, Any]:
+    """Read a file with optional offset/limit. Restricted to SDK working directory.
+
+    After reading, the file is deleted to prevent accumulation in long-running pods.
+    """
+    file_path = args.get("file_path", "")
+    offset = args.get("offset", 0)
+    limit = args.get("limit", 2000)
+
+    # Security: only allow reads under ~/.claude/projects/**/tool-results/
+    real_path = os.path.realpath(file_path)
+    if not real_path.startswith(_SDK_PROJECTS_DIR) or "tool-results" not in real_path:
+        return {
+            "content": [{"type": "text", "text": f"Access denied: {file_path}"}],
+            "isError": True,
+        }
+
+    try:
+        with open(real_path) as f:
+            selected = list(itertools.islice(f, offset, offset + limit))
+        content = "".join(selected)
+        # Cleanup happens in _cleanup_sdk_tool_results after session ends;
+        # don't delete here — the SDK may read in multiple chunks.
+        return {"content": [{"type": "text", "text": content}], "isError": False}
+    except FileNotFoundError:
+        return {
+            "content": [{"type": "text", "text": f"File not found: {file_path}"}],
+            "isError": True,
+        }
+    except Exception as e:
+        return {
+            "content": [{"type": "text", "text": f"Error reading file: {e}"}],
+            "isError": True,
+        }
+
+
+_READ_TOOL_NAME = "Read"
+_READ_TOOL_DESCRIPTION = (
+    "Read a file from the local filesystem. "
+    "Use offset and limit to read specific line ranges for large files."
+)
+_READ_TOOL_SCHEMA = {
+    "type": "object",
+    "properties": {
+        "file_path": {
+            "type": "string",
+            "description": "The absolute path to the file to read",
+        },
+        "offset": {
+            "type": "integer",
+            "description": "Line number to start reading from (0-indexed). Default: 0",
+        },
+        "limit": {
+            "type": "integer",
+            "description": "Number of lines to read. Default: 2000",
+        },
+    },
+    "required": ["file_path"],
+}
+
+
+# Create the MCP server configuration
+def create_copilot_mcp_server():
+    """Create an in-process MCP server configuration for CoPilot tools.
+
+    This can be passed to ClaudeAgentOptions.mcp_servers.
+
+    Note: The actual SDK MCP server creation depends on the claude-agent-sdk
+    package being available. This function returns the configuration that
+    can be used with the SDK.
+    """
+    try:
+        from claude_agent_sdk import create_sdk_mcp_server, tool
+
+        # Create decorated tool functions
+        sdk_tools = []
+
+        for tool_name, base_tool in TOOL_REGISTRY.items():
+            handler = create_tool_handler(base_tool)
+            decorated = tool(
+                tool_name,
+                base_tool.description,
+                _build_input_schema(base_tool),
+            )(handler)
+            sdk_tools.append(decorated)
+
+        # Add the Read tool so the SDK can read back oversized tool results
+        read_tool = tool(
+            _READ_TOOL_NAME,
+            _READ_TOOL_DESCRIPTION,
+            _READ_TOOL_SCHEMA,
+        )(_read_file_handler)
+        sdk_tools.append(read_tool)
+
+        server = create_sdk_mcp_server(
+            name=MCP_SERVER_NAME,
+            version="1.0.0",
+            tools=sdk_tools,
+        )
+
+        return server
+
+    except ImportError:
+        # Let ImportError propagate so service.py handles the fallback
+        raise
+
+
+# SDK built-in tools allowed within the workspace directory.
+# Security hooks validate that file paths stay within sdk_cwd.
+# Bash is NOT included — use the sandboxed MCP bash_exec tool instead,
+# which provides kernel-level network isolation via unshare --net.
+# Task allows spawning sub-agents (rate-limited by security hooks).
+# WebSearch uses Brave Search via Anthropic's API — safe, no SSRF risk.
+_SDK_BUILTIN_TOOLS = ["Read", "Write", "Edit", "Glob", "Grep", "Task", "WebSearch"]
+
+# SDK built-in tools that must be explicitly blocked.
+# Bash: dangerous — agent uses mcp__copilot__bash_exec with kernel-level
+#   network isolation (unshare --net) instead.
+# WebFetch: SSRF risk — can reach internal network (localhost, 10.x, etc.).
+#   Agent uses the SSRF-protected mcp__copilot__web_fetch tool instead.
+SDK_DISALLOWED_TOOLS = ["Bash", "WebFetch"]
+
+# Tools that are blocked entirely in security hooks (defence-in-depth).
+# Includes SDK_DISALLOWED_TOOLS plus common aliases/synonyms.
+BLOCKED_TOOLS = {
+    *SDK_DISALLOWED_TOOLS,
+    "bash",
+    "shell",
+    "exec",
+    "terminal",
+    "command",
+}
+
+# Tools allowed only when their path argument stays within the SDK workspace.
+# The SDK uses these to handle oversized tool results (writes to tool-results/
+# files, then reads them back) and for workspace file operations.
+WORKSPACE_SCOPED_TOOLS = {"Read", "Write", "Edit", "Glob", "Grep"}
+
+# Dangerous patterns in tool inputs
+DANGEROUS_PATTERNS = [
+    r"sudo",
+    r"rm\s+-rf",
+    r"dd\s+if=",
+    r"/etc/passwd",
+    r"/etc/shadow",
+    r"chmod\s+777",
+    r"curl\s+.*\|.*sh",
+    r"wget\s+.*\|.*sh",
+    r"eval\s*\(",
+    r"exec\s*\(",
+    r"__import__",
+    r"os\.system",
+    r"subprocess",
+]
+
+# List of tool names for allowed_tools configuration
+# Include MCP tools, the MCP Read tool for oversized results,
+# and SDK built-in file tools for workspace operations.
+COPILOT_TOOL_NAMES = [
+    *[f"{MCP_TOOL_PREFIX}{name}" for name in TOOL_REGISTRY.keys()],
+    f"{MCP_TOOL_PREFIX}{_READ_TOOL_NAME}",
+    *_SDK_BUILTIN_TOOLS,
+]

autogpt_platform/backend/backend/api/features/chat/sdk/transcript.py

@@ -0,0 +1,356 @@
+"""JSONL transcript management for stateless multi-turn resume.
+
+The Claude Code CLI persists conversations as JSONL files (one JSON object per
+line).  When the SDK's ``Stop`` hook fires we read this file, strip bloat
+(progress entries, metadata), and upload the result to bucket storage.  On the
+next turn we download the transcript, write it to a temp file, and pass
+``--resume`` so the CLI can reconstruct the full conversation.
+
+Storage is handled via ``WorkspaceStorageBackend`` (GCS in prod, local
+filesystem for self-hosted) — no DB column needed.
+"""
+
+import json
+import logging
+import os
+import re
+
+logger = logging.getLogger(__name__)
+
+# UUIDs are hex + hyphens; strip everything else to prevent path injection.
+_SAFE_ID_RE = re.compile(r"[^0-9a-fA-F-]")
+
+# Entry types that can be safely removed from the transcript without breaking
+# the parentUuid conversation tree that ``--resume`` relies on.
+# - progress: UI progress ticks, no message content (avg 97KB for agent_progress)
+# - file-history-snapshot: undo tracking metadata
+# - queue-operation: internal queue bookkeeping
+# - summary: session summaries
+# - pr-link: PR link metadata
+STRIPPABLE_TYPES = frozenset(
+    {"progress", "file-history-snapshot", "queue-operation", "summary", "pr-link"}
+)
+
+# Workspace storage constants — deterministic path from session_id.
+TRANSCRIPT_STORAGE_PREFIX = "chat-transcripts"
+
+
+# ---------------------------------------------------------------------------
+# Progress stripping
+# ---------------------------------------------------------------------------
+
+
+def strip_progress_entries(content: str) -> str:
+    """Remove progress/metadata entries from a JSONL transcript.
+
+    Removes entries whose ``type`` is in ``STRIPPABLE_TYPES`` and reparents
+    any remaining child entries so the ``parentUuid`` chain stays intact.
+    Typically reduces transcript size by ~30%.
+    """
+    lines = content.strip().split("\n")
+
+    entries: list[dict] = []
+    for line in lines:
+        try:
+            entries.append(json.loads(line))
+        except json.JSONDecodeError:
+            # Keep unparseable lines as-is (safety)
+            entries.append({"_raw": line})
+
+    stripped_uuids: set[str] = set()
+    uuid_to_parent: dict[str, str] = {}
+    kept: list[dict] = []
+
+    for entry in entries:
+        if "_raw" in entry:
+            kept.append(entry)
+            continue
+        uid = entry.get("uuid", "")
+        parent = entry.get("parentUuid", "")
+        entry_type = entry.get("type", "")
+
+        if uid:
+            uuid_to_parent[uid] = parent
+
+        if entry_type in STRIPPABLE_TYPES:
+            if uid:
+                stripped_uuids.add(uid)
+        else:
+            kept.append(entry)
+
+    # Reparent: walk up chain through stripped entries to find surviving ancestor
+    for entry in kept:
+        if "_raw" in entry:
+            continue
+        parent = entry.get("parentUuid", "")
+        original_parent = parent
+        while parent in stripped_uuids:
+            parent = uuid_to_parent.get(parent, "")
+        if parent != original_parent:
+            entry["parentUuid"] = parent
+
+    result_lines: list[str] = []
+    for entry in kept:
+        if "_raw" in entry:
+            result_lines.append(entry["_raw"])
+        else:
+            result_lines.append(json.dumps(entry, separators=(",", ":")))
+
+    return "\n".join(result_lines) + "\n"
+
+
+# ---------------------------------------------------------------------------
+# Local file I/O (read from CLI's JSONL, write temp file for --resume)
+# ---------------------------------------------------------------------------
+
+
+def read_transcript_file(transcript_path: str) -> str | None:
+    """Read a JSONL transcript file from disk.
+
+    Returns the raw JSONL content, or ``None`` if the file is missing, empty,
+    or only contains metadata (≤2 lines with no conversation messages).
+    """
+    if not transcript_path or not os.path.isfile(transcript_path):
+        logger.debug(f"[Transcript] File not found: {transcript_path}")
+        return None
+
+    try:
+        with open(transcript_path) as f:
+            content = f.read()
+
+        if not content.strip():
+            logger.debug(f"[Transcript] Empty file: {transcript_path}")
+            return None
+
+        lines = content.strip().split("\n")
+        if len(lines) < 3:
+            # Raw files with ≤2 lines are metadata-only
+            # (queue-operation + file-history-snapshot, no conversation).
+            logger.debug(
+                f"[Transcript] Too few lines ({len(lines)}): {transcript_path}"
+            )
+            return None
+
+        # Quick structural validation — parse first and last lines.
+        json.loads(lines[0])
+        json.loads(lines[-1])
+
+        logger.info(
+            f"[Transcript] Read {len(lines)} lines, "
+            f"{len(content)} bytes from {transcript_path}"
+        )
+        return content
+
+    except (json.JSONDecodeError, OSError) as e:
+        logger.warning(f"[Transcript] Failed to read {transcript_path}: {e}")
+        return None
+
+
+def _sanitize_id(raw_id: str, max_len: int = 36) -> str:
+    """Sanitize an ID for safe use in file paths.
+
+    Session/user IDs are expected to be UUIDs (hex + hyphens).  Strip
+    everything else and truncate to *max_len* so the result cannot introduce
+    path separators or other special characters.
+    """
+    cleaned = _SAFE_ID_RE.sub("", raw_id or "")[:max_len]
+    return cleaned or "unknown"
+
+
+_SAFE_CWD_PREFIX = os.path.realpath("/tmp/copilot-")
+
+
+def write_transcript_to_tempfile(
+    transcript_content: str,
+    session_id: str,
+    cwd: str,
+) -> str | None:
+    """Write JSONL transcript to a temp file inside *cwd* for ``--resume``.
+
+    The file lives in the session working directory so it is cleaned up
+    automatically when the session ends.
+
+    Returns the absolute path to the file, or ``None`` on failure.
+    """
+    # Validate cwd is under the expected sandbox prefix (CodeQL sanitizer).
+    real_cwd = os.path.realpath(cwd)
+    if not real_cwd.startswith(_SAFE_CWD_PREFIX):
+        logger.warning(f"[Transcript] cwd outside sandbox: {cwd}")
+        return None
+
+    try:
+        os.makedirs(real_cwd, exist_ok=True)
+        safe_id = _sanitize_id(session_id, max_len=8)
+        jsonl_path = os.path.realpath(
+            os.path.join(real_cwd, f"transcript-{safe_id}.jsonl")
+        )
+        if not jsonl_path.startswith(real_cwd):
+            logger.warning(f"[Transcript] Path escaped cwd: {jsonl_path}")
+            return None
+
+        with open(jsonl_path, "w") as f:
+            f.write(transcript_content)
+
+        logger.info(f"[Transcript] Wrote resume file: {jsonl_path}")
+        return jsonl_path
+
+    except OSError as e:
+        logger.warning(f"[Transcript] Failed to write resume file: {e}")
+        return None
+
+
+def validate_transcript(content: str | None) -> bool:
+    """Check that a transcript has actual conversation messages.
+
+    A valid transcript for resume needs at least one user message and one
+    assistant message (not just queue-operation / file-history-snapshot
+    metadata).
+    """
+    if not content or not content.strip():
+        return False
+
+    lines = content.strip().split("\n")
+    if len(lines) < 2:
+        return False
+
+    has_user = False
+    has_assistant = False
+
+    for line in lines:
+        try:
+            entry = json.loads(line)
+            msg_type = entry.get("type")
+            if msg_type == "user":
+                has_user = True
+            elif msg_type == "assistant":
+                has_assistant = True
+        except json.JSONDecodeError:
+            return False
+
+    return has_user and has_assistant
+
+
+# ---------------------------------------------------------------------------
+# Bucket storage (GCS / local via WorkspaceStorageBackend)
+# ---------------------------------------------------------------------------
+
+
+def _storage_path_parts(user_id: str, session_id: str) -> tuple[str, str, str]:
+    """Return (workspace_id, file_id, filename) for a session's transcript.
+
+    Path structure: ``chat-transcripts/{user_id}/{session_id}.jsonl``
+    IDs are sanitized to hex+hyphen to prevent path traversal.
+    """
+    return (
+        TRANSCRIPT_STORAGE_PREFIX,
+        _sanitize_id(user_id),
+        f"{_sanitize_id(session_id)}.jsonl",
+    )
+
+
+def _build_storage_path(user_id: str, session_id: str, backend: object) -> str:
+    """Build the full storage path string that ``retrieve()`` expects.
+
+    ``store()`` returns a path like ``gcs://bucket/workspaces/...`` or
+    ``local://workspace_id/file_id/filename``.  Since we use deterministic
+    arguments we can reconstruct the same path for download/delete without
+    having stored the return value.
+    """
+    from backend.util.workspace_storage import GCSWorkspaceStorage
+
+    wid, fid, fname = _storage_path_parts(user_id, session_id)
+
+    if isinstance(backend, GCSWorkspaceStorage):
+        blob = f"workspaces/{wid}/{fid}/{fname}"
+        return f"gcs://{backend.bucket_name}/{blob}"
+    else:
+        # LocalWorkspaceStorage returns local://{relative_path}
+        return f"local://{wid}/{fid}/{fname}"
+
+
+async def upload_transcript(user_id: str, session_id: str, content: str) -> None:
+    """Strip progress entries and upload transcript to bucket storage.
+
+    Safety: only overwrites when the new (stripped) transcript is larger than
+    what is already stored.  Since JSONL is append-only, the latest transcript
+    is always the longest.  This prevents a slow/stale background task from
+    clobbering a newer upload from a concurrent turn.
+    """
+    from backend.util.workspace_storage import get_workspace_storage
+
+    stripped = strip_progress_entries(content)
+    if not validate_transcript(stripped):
+        logger.warning(
+            f"[Transcript] Skipping upload — stripped content is not a valid "
+            f"transcript for session {session_id}"
+        )
+        return
+
+    storage = await get_workspace_storage()
+    wid, fid, fname = _storage_path_parts(user_id, session_id)
+    encoded = stripped.encode("utf-8")
+    new_size = len(encoded)
+
+    # Check existing transcript size to avoid overwriting newer with older
+    path = _build_storage_path(user_id, session_id, storage)
+    try:
+        existing = await storage.retrieve(path)
+        if len(existing) >= new_size:
+            logger.info(
+                f"[Transcript] Skipping upload — existing transcript "
+                f"({len(existing)}B) >= new ({new_size}B) for session "
+                f"{session_id}"
+            )
+            return
+    except (FileNotFoundError, Exception):
+        pass  # No existing transcript or retrieval error — proceed with upload
+
+    await storage.store(
+        workspace_id=wid,
+        file_id=fid,
+        filename=fname,
+        content=encoded,
+    )
+    logger.info(
+        f"[Transcript] Uploaded {new_size} bytes "
+        f"(stripped from {len(content)}) for session {session_id}"
+    )
+
+
+async def download_transcript(user_id: str, session_id: str) -> str | None:
+    """Download transcript from bucket storage.
+
+    Returns the JSONL content string, or ``None`` if not found.
+    """
+    from backend.util.workspace_storage import get_workspace_storage
+
+    storage = await get_workspace_storage()
+    path = _build_storage_path(user_id, session_id, storage)
+
+    try:
+        data = await storage.retrieve(path)
+        content = data.decode("utf-8")
+        logger.info(
+            f"[Transcript] Downloaded {len(content)} bytes for session {session_id}"
+        )
+        return content
+    except FileNotFoundError:
+        logger.debug(f"[Transcript] No transcript in storage for {session_id}")
+        return None
+    except Exception as e:
+        logger.warning(f"[Transcript] Failed to download transcript: {e}")
+        return None
+
+
+async def delete_transcript(user_id: str, session_id: str) -> None:
+    """Delete transcript from bucket storage (e.g. after resume failure)."""
+    from backend.util.workspace_storage import get_workspace_storage
+
+    storage = await get_workspace_storage()
+    path = _build_storage_path(user_id, session_id, storage)
+
+    try:
+        await storage.delete(path)
+        logger.info(f"[Transcript] Deleted transcript for session {session_id}")
+    except Exception as e:
+        logger.warning(f"[Transcript] Failed to delete transcript: {e}")