feat(blocks): Add ClaudeCodeBlock for executing tasks via Claude Code in E2B sandbox (#11761)

Introduces a new ClaudeCodeBlock that enables execution of coding tasks
using Anthropic's Claude Code in an E2B sandbox. This block unlocks
powerful agentic coding capabilities - Claude Code can autonomously
create files, install packages, run commands, and build complete
applications within a secure sandboxed environment.

Changes 🏗️

- New file backend/blocks/claude_code.py:
  - ClaudeCodeBlock - Execute tasks using Claude Code in an E2B sandbox
- Dual credential support: E2B API key (sandbox) + Anthropic API key
(Claude Code)
- Session continuation support via session_id, sandbox_id, and
conversation_history
- Automatic file extraction with path, relative_path, name, and content
fields
  - Configurable timeout, setup commands, and working directory
- dispose_sandbox option to keep sandbox alive for multi-turn
conversations

Checklist 📋

For code changes:

- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Create and execute ClaudeCodeBlock with a simple prompt ("Create a
hello world HTML file")
- [x] Verify files output includes correct path, relative_path, name,
and content
- [x] Test session continuation by passing session_id and sandbox_id
back
- [x] Build "Any API → Instant App" demo agent combining Firecrawl +
ClaudeCodeBlock + GitHub blocks
- [x] Verify generated files are pushed to GitHub with correct folder
structure using relative_path

Here are two example agents i made that can be used to test this agent,
they require github, anthropic and e2b access via api keys that are set
via the user/on the platform is testing on dev

The first agent is my

Any API → Instant App
"Transform any API documentation into a fully functional web
application. Just provide a docs URL and get a complete, ready-to-deploy
app pushed to a new GitHub repository."

[Any API → Instant
App_v36.json](https://github.com/user-attachments/files/24600326/Any.API.Instant.App_v36.json)


The second agent is my
Idea to project
"Simply enter your coding project's idea and this agent will make all of
the base initial code needed for you to start working on that project
and place it on github for you!"

[Idea to
project_v11.json](https://github.com/user-attachments/files/24600346/Idea.to.project_v11.json)

If you have any questions or issues let me know.

References
https://e2b.dev/blog/python-guide-run-claude-code-in-an-e2b-sandbox

https://github.com/e2b-dev/e2b-cookbook/tree/main/examples/anthropic-claude-code-in-sandbox-python
https://code.claude.com/docs/en/cli-reference

I tried to use E2b's "anthropic-claude-code" template but it kept
complaining it was out of date, so I make it manually spin up a E2b
instance and make it install the latest claude code and it uses that

.github/workflows/claude-ci-failure-auto-fix.yml

@@ -93,5 +93,5 @@ jobs:
 
             Error logs:
             ${{ toJSON(fromJSON(steps.failure_details.outputs.result).errorLogs) }}
-          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
+          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
           claude_args: "--allowedTools 'Edit,MultiEdit,Write,Read,Glob,Grep,LS,Bash(git:*),Bash(bun:*),Bash(npm:*),Bash(npx:*),Bash(gh:*)'"

.github/workflows/claude-dependabot.yml

@@ -7,7 +7,7 @@
 # - Provide actionable recommendations for the development team
 #
 # Triggered on: Dependabot PRs (opened, synchronize)
-# Requirements: ANTHROPIC_API_KEY secret must be configured
+# Requirements: CLAUDE_CODE_OAUTH_TOKEN secret must be configured
 
 name: Claude Dependabot PR Review
 
@@ -308,7 +308,7 @@ jobs:
         id: claude_review
         uses: anthropics/claude-code-action@v1
         with:
-          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
+          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
           claude_args: |
             --allowedTools "Bash(npm:*),Bash(pnpm:*),Bash(poetry:*),Bash(git:*),Edit,Replace,NotebookEditCell,mcp__github_inline_comment__create_inline_comment,Bash(gh pr comment:*), Bash(gh pr diff:*), Bash(gh pr view:*)"
           prompt: |

.github/workflows/claude.yml

@@ -323,7 +323,7 @@ jobs:
         id: claude
         uses: anthropics/claude-code-action@v1
         with:
-          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
+          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
           claude_args: |
             --allowedTools "Bash(npm:*),Bash(pnpm:*),Bash(poetry:*),Bash(git:*),Edit,Replace,NotebookEditCell,mcp__github_inline_comment__create_inline_comment,Bash(gh pr comment:*), Bash(gh pr diff:*), Bash(gh pr view:*), Bash(gh pr edit:*)"
             --model opus

.github/workflows/docs-block-sync.yml

@@ -0,0 +1,78 @@
+name: Block Documentation Sync Check
+
+on:
+  push:
+    branches: [master, dev]
+    paths:
+      - "autogpt_platform/backend/backend/blocks/**"
+      - "docs/integrations/**"
+      - "autogpt_platform/backend/scripts/generate_block_docs.py"
+      - ".github/workflows/docs-block-sync.yml"
+  pull_request:
+    branches: [master, dev]
+    paths:
+      - "autogpt_platform/backend/backend/blocks/**"
+      - "docs/integrations/**"
+      - "autogpt_platform/backend/scripts/generate_block_docs.py"
+      - ".github/workflows/docs-block-sync.yml"
+
+jobs:
+  check-docs-sync:
+    runs-on: ubuntu-latest
+    timeout-minutes: 15
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 1
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Set up Python dependency cache
+        uses: actions/cache@v4
+        with:
+          path: ~/.cache/pypoetry
+          key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
+          restore-keys: |
+            poetry-${{ runner.os }}-
+
+      - name: Install Poetry
+        run: |
+          cd autogpt_platform/backend
+          HEAD_POETRY_VERSION=$(python3 ../../.github/workflows/scripts/get_package_version_from_lockfile.py poetry)
+          echo "Found Poetry version ${HEAD_POETRY_VERSION} in backend/poetry.lock"
+          curl -sSL https://install.python-poetry.org | POETRY_VERSION=$HEAD_POETRY_VERSION python3 -
+          echo "$HOME/.local/bin" >> $GITHUB_PATH
+
+      - name: Install dependencies
+        working-directory: autogpt_platform/backend
+        run: |
+          poetry install --only main
+          poetry run prisma generate
+
+      - name: Check block documentation is in sync
+        working-directory: autogpt_platform/backend
+        run: |
+          echo "Checking if block documentation is in sync with code..."
+          poetry run python scripts/generate_block_docs.py --check
+
+      - name: Show diff if out of sync
+        if: failure()
+        working-directory: autogpt_platform/backend
+        run: |
+          echo "::error::Block documentation is out of sync with code!"
+          echo ""
+          echo "To fix this, run the following command locally:"
+          echo "  cd autogpt_platform/backend && poetry run python scripts/generate_block_docs.py"
+          echo ""
+          echo "Then commit the updated documentation files."
+          echo ""
+          echo "Regenerating docs to show diff..."
+          poetry run python scripts/generate_block_docs.py
+          echo ""
+          echo "Changes detected:"
+          git diff ../../docs/integrations/ || true

.github/workflows/docs-claude-review.yml

@@ -0,0 +1,95 @@
+name: Claude Block Docs Review
+
+on:
+  pull_request:
+    types: [opened, synchronize]
+    paths:
+      - "docs/integrations/**"
+      - "autogpt_platform/backend/backend/blocks/**"
+
+jobs:
+  claude-review:
+    # Only run for PRs from members/collaborators
+    if: |
+      github.event.pull_request.author_association == 'OWNER' ||
+      github.event.pull_request.author_association == 'MEMBER' ||
+      github.event.pull_request.author_association == 'COLLABORATOR'
+    runs-on: ubuntu-latest
+    timeout-minutes: 15
+    permissions:
+      contents: read
+      pull-requests: write
+      id-token: write
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Set up Python dependency cache
+        uses: actions/cache@v4
+        with:
+          path: ~/.cache/pypoetry
+          key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
+          restore-keys: |
+            poetry-${{ runner.os }}-
+
+      - name: Install Poetry
+        run: |
+          cd autogpt_platform/backend
+          HEAD_POETRY_VERSION=$(python3 ../../.github/workflows/scripts/get_package_version_from_lockfile.py poetry)
+          curl -sSL https://install.python-poetry.org | POETRY_VERSION=$HEAD_POETRY_VERSION python3 -
+          echo "$HOME/.local/bin" >> $GITHUB_PATH
+
+      - name: Install dependencies
+        working-directory: autogpt_platform/backend
+        run: |
+          poetry install --only main
+          poetry run prisma generate
+
+      - name: Run Claude Code Review
+        uses: anthropics/claude-code-action@v1
+        with:
+          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
+          claude_args: |
+            --allowedTools "Read,Glob,Grep,Bash(gh pr comment:*),Bash(gh pr diff:*),Bash(gh pr view:*)"
+          prompt: |
+            You are reviewing a PR that modifies block documentation or block code for AutoGPT.
+
+            ## Your Task
+            Review the changes in this PR and provide constructive feedback. Focus on:
+
+            1. **Documentation Accuracy**: For any block code changes, verify that:
+               - Input/output tables in docs match the actual block schemas
+               - Description text accurately reflects what the block does
+               - Any new blocks have corresponding documentation
+
+            2. **Manual Content Quality**: Check manual sections (marked with `<!-- MANUAL: -->` markers):
+               - "How it works" sections should have clear technical explanations
+               - "Possible use case" sections should have practical, real-world examples
+               - Content should be helpful for users trying to understand the blocks
+
+            3. **Template Compliance**: Ensure docs follow the standard template:
+               - What it is (brief intro)
+               - What it does (description)
+               - How it works (technical explanation)
+               - Inputs table
+               - Outputs table
+               - Possible use case
+
+            4. **Cross-references**: Check that links and anchors are correct
+
+            ## Review Process
+            1. First, get the PR diff to see what changed: `gh pr diff ${{ github.event.pull_request.number }}`
+            2. Read any modified block files to understand the implementation
+            3. Read corresponding documentation files to verify accuracy
+            4. Provide your feedback as a PR comment
+
+            Be constructive and specific. If everything looks good, say so!
+            If there are issues, explain what's wrong and suggest how to fix it.

.github/workflows/docs-enhance.yml

@@ -0,0 +1,194 @@
+name: Enhance Block Documentation
+
+on:
+  workflow_dispatch:
+    inputs:
+      block_pattern:
+        description: 'Block file pattern to enhance (e.g., "google/*.md" or "*" for all blocks)'
+        required: true
+        default: '*'
+        type: string
+      dry_run:
+        description: 'Dry run mode - show proposed changes without committing'
+        type: boolean
+        default: true
+      max_blocks:
+        description: 'Maximum number of blocks to process (0 for unlimited)'
+        type: number
+        default: 10
+
+jobs:
+  enhance-docs:
+    runs-on: ubuntu-latest
+    timeout-minutes: 45
+    permissions:
+      contents: write
+      pull-requests: write
+      id-token: write
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 1
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Set up Python dependency cache
+        uses: actions/cache@v4
+        with:
+          path: ~/.cache/pypoetry
+          key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
+          restore-keys: |
+            poetry-${{ runner.os }}-
+
+      - name: Install Poetry
+        run: |
+          cd autogpt_platform/backend
+          HEAD_POETRY_VERSION=$(python3 ../../.github/workflows/scripts/get_package_version_from_lockfile.py poetry)
+          curl -sSL https://install.python-poetry.org | POETRY_VERSION=$HEAD_POETRY_VERSION python3 -
+          echo "$HOME/.local/bin" >> $GITHUB_PATH
+
+      - name: Install dependencies
+        working-directory: autogpt_platform/backend
+        run: |
+          poetry install --only main
+          poetry run prisma generate
+
+      - name: Run Claude Enhancement
+        uses: anthropics/claude-code-action@v1
+        with:
+          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
+          claude_args: |
+            --allowedTools "Read,Edit,Glob,Grep,Write,Bash(git:*),Bash(gh:*),Bash(find:*),Bash(ls:*)"
+          prompt: |
+            You are enhancing block documentation for AutoGPT. Your task is to improve the MANUAL sections
+            of block documentation files by reading the actual block implementations and writing helpful content.
+
+            ## Configuration
+            - Block pattern: ${{ inputs.block_pattern }}
+            - Dry run: ${{ inputs.dry_run }}
+            - Max blocks to process: ${{ inputs.max_blocks }}
+
+            ## Your Task
+
+            1. **Find Documentation Files**
+               Find block documentation files matching the pattern in `docs/integrations/`
+               Pattern: ${{ inputs.block_pattern }}
+
+               Use: `find docs/integrations -name "*.md" -type f`
+
+            2. **For Each Documentation File** (up to ${{ inputs.max_blocks }} files):
+
+               a. Read the documentation file
+
+               b. Identify which block(s) it documents (look for the block class name)
+
+               c. Find and read the corresponding block implementation in `autogpt_platform/backend/backend/blocks/`
+
+               d. Improve the MANUAL sections:
+
+                  **"How it works" section** (within `<!-- MANUAL: how_it_works -->` markers):
+                  - Explain the technical flow of the block
+                  - Describe what APIs or services it connects to
+                  - Note any important configuration or prerequisites
+                  - Keep it concise but informative (2-4 paragraphs)
+
+                  **"Possible use case" section** (within `<!-- MANUAL: use_case -->` markers):
+                  - Provide 2-3 practical, real-world examples
+                  - Make them specific and actionable
+                  - Show how this block could be used in an automation workflow
+
+            3. **Important Rules**
+               - ONLY modify content within `<!-- MANUAL: -->` and `<!-- END MANUAL -->` markers
+               - Do NOT modify auto-generated sections (inputs/outputs tables, descriptions)
+               - Keep content accurate based on the actual block implementation
+               - Write for users who may not be technical experts
+
+            4. **Output**
+               ${{ inputs.dry_run == true && 'DRY RUN MODE: Show proposed changes for each file but do NOT actually edit the files. Describe what you would change.' || 'LIVE MODE: Actually edit the files to improve the documentation.' }}
+
+            ## Example Improvements
+
+            **Before (How it works):**
+            ```
+            _Add technical explanation here._
+            ```
+
+            **After (How it works):**
+            ```
+            This block connects to the GitHub API to retrieve issue information. When executed,
+            it authenticates using your GitHub credentials and fetches issue details including
+            title, body, labels, and assignees.
+
+            The block requires a valid GitHub OAuth connection with repository access permissions.
+            It supports both public and private repositories you have access to.
+            ```
+
+            **Before (Possible use case):**
+            ```
+            _Add practical use case examples here._
+            ```
+
+            **After (Possible use case):**
+            ```
+            **Customer Support Automation**: Monitor a GitHub repository for new issues with
+            the "bug" label, then automatically create a ticket in your support system and
+            notify the on-call engineer via Slack.
+
+            **Release Notes Generation**: When a new release is published, gather all closed
+            issues since the last release and generate a summary for your changelog.
+            ```
+
+            Begin by finding and listing the documentation files to process.
+
+      - name: Create PR with enhanced documentation
+        if: ${{ inputs.dry_run == false }}
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          # Check if there are changes
+          if git diff --quiet docs/integrations/; then
+            echo "No changes to commit"
+            exit 0
+          fi
+
+          # Configure git
+          git config user.name "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+
+          # Create branch and commit
+          BRANCH_NAME="docs/enhance-blocks-$(date +%Y%m%d-%H%M%S)"
+          git checkout -b "$BRANCH_NAME"
+          git add docs/integrations/
+          git commit -m "docs: enhance block documentation with LLM-generated content
+
+          Pattern: ${{ inputs.block_pattern }}
+          Max blocks: ${{ inputs.max_blocks }}
+
+          🤖 Generated with [Claude Code](https://claude.com/claude-code)
+
+          Co-Authored-By: Claude <noreply@anthropic.com>"
+
+          # Push and create PR
+          git push -u origin "$BRANCH_NAME"
+          gh pr create \
+            --title "docs: LLM-enhanced block documentation" \
+            --body "## Summary
+          This PR contains LLM-enhanced documentation for block files matching pattern: \`${{ inputs.block_pattern }}\`
+
+          The following manual sections were improved:
+          - **How it works**: Technical explanations based on block implementations
+          - **Possible use case**: Practical, real-world examples
+
+          ## Review Checklist
+          - [ ] Content is accurate based on block implementations
+          - [ ] Examples are practical and helpful
+          - [ ] No auto-generated sections were modified
+
+          ---
+          🤖 Generated with [Claude Code](https://claude.com/claude-code)" \
+            --base dev

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

@@ -128,7 +128,7 @@ jobs:
           token: ${{ secrets.GITHUB_TOKEN }}
           exitOnceUploaded: true
 
-  test:
+  e2e_test:
     runs-on: big-boi
     needs: setup
     strategy:
@@ -258,3 +258,39 @@ jobs:
       - name: Print Final Docker Compose logs
         if: always()
         run: docker compose -f ../docker-compose.yml logs
+
+  integration_test:
+    runs-on: ubuntu-latest
+    needs: setup
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          submodules: recursive
+
+      - name: Set up Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: "22.18.0"
+
+      - name: Enable corepack
+        run: corepack enable
+
+      - name: Restore dependencies cache
+        uses: actions/cache@v4
+        with:
+          path: ~/.pnpm-store
+          key: ${{ needs.setup.outputs.cache-key }}
+          restore-keys: |
+            ${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml') }}
+            ${{ runner.os }}-pnpm-
+
+      - name: Install dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Generate API client
+        run: pnpm generate:api
+
+      - name: Run Integration Tests
+        run: pnpm test:unit

AGENTS.md

@@ -16,6 +16,32 @@ See `docs/content/platform/getting-started.md` for setup instructions.
 - Format Python code with `poetry run format`.
 - Format frontend code using `pnpm format`.
 
+
+## Frontend guidelines:
+
+See `/frontend/CONTRIBUTING.md` for complete patterns. Quick reference:
+
+1. **Pages**: Create in `src/app/(platform)/feature-name/page.tsx`
+   - Add `usePageName.ts` hook for logic
+   - Put sub-components in local `components/` folder
+2. **Components**: Structure as `ComponentName/ComponentName.tsx` + `useComponentName.ts` + `helpers.ts`
+   - Use design system components from `src/components/` (atoms, molecules, organisms)
+   - Never use `src/components/__legacy__/*`
+3. **Data fetching**: Use generated API hooks from `@/app/api/__generated__/endpoints/`
+   - Regenerate with `pnpm generate:api`
+   - Pattern: `use{Method}{Version}{OperationName}`
+4. **Styling**: Tailwind CSS only, use design tokens, Phosphor Icons only
+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
+- Separate render logic from business logic (component.tsx + useComponent.ts + helpers.ts)
+- Colocate state when possible and avoid creating large components, use sub-components ( local `/components` folder next to the parent component ) when sensible
+- Avoid large hooks, abstract logic into `helpers.ts` files when sensible
+- Use function declarations for components, arrow functions only for callbacks
+- No barrel files or `index.ts` re-exports
+- Do not use `useCallback` or `useMemo` unless strictly needed
+- Avoid comments at all times unless the code is very complex
+
 ## Testing
 
 - Backend: `poetry run test` (runs pytest with a docker based postgres + prisma).

autogpt_platform/CLAUDE.md

@@ -201,7 +201,7 @@ If you get any pushback or hit complex block conditions check the new_blocks gui
 3. Write tests alongside the route file
 4. Run `poetry run test` to verify
 
-**Frontend feature development:**
+### Frontend guidelines:
 
 See `/frontend/CONTRIBUTING.md` for complete patterns. Quick reference:
 
@@ -217,6 +217,14 @@ See `/frontend/CONTRIBUTING.md` for complete patterns. Quick reference:
 4. **Styling**: Tailwind CSS only, use design tokens, Phosphor Icons only
 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
+- Separate render logic from business logic (component.tsx + useComponent.ts + helpers.ts)
+- Colocate state when possible and avoid creating large components, use sub-components ( local `/components` folder next to the parent component ) when sensible
+- Avoid large hooks, abstract logic into `helpers.ts` files when sensible
+- Use function declarations for components, arrow functions only for callbacks
+- No barrel files or `index.ts` re-exports
+- Do not use `useCallback` or `useMemo` unless strictly needed
+- Avoid comments at all times unless the code is very complex
 
 ### Security Implementation
 

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

@@ -290,6 +290,11 @@ async def _cache_session(session: ChatSession) -> None:
     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 _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)

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

@@ -172,12 +172,12 @@ async def get_session(
         user_id: The optional authenticated user ID, or None for anonymous access.
 
     Returns:
-        SessionDetailResponse: Details for the requested session; raises NotFoundError if not found.
+        SessionDetailResponse: Details for the requested session, or None if not found.
 
     """
     session = await get_chat_session(session_id, user_id)
     if not session:
-        raise NotFoundError(f"Session {session_id} not found")
+        raise NotFoundError(f"Session {session_id} not found.")
 
     messages = [message.model_dump() for message in session.messages]
     logger.info(
@@ -222,6 +222,8 @@ async def stream_chat_post(
     session = await _validate_and_get_session(session_id, user_id)
 
     async def event_generator() -> AsyncGenerator[str, None]:
+        chunk_count = 0
+        first_chunk_type: str | None = None
         async for chunk in chat_service.stream_chat_completion(
             session_id,
             request.message,
@@ -230,7 +232,26 @@ async def stream_chat_post(
             session=session,  # Pass pre-fetched session to avoid double-fetch
             context=request.context,
         ):
+            if chunk_count < 3:
+                logger.info(
+                    "Chat stream chunk",
+                    extra={
+                        "session_id": session_id,
+                        "chunk_type": str(chunk.type),
+                    },
+                )
+            if not first_chunk_type:
+                first_chunk_type = str(chunk.type)
+            chunk_count += 1
             yield chunk.to_sse()
+        logger.info(
+            "Chat stream completed",
+            extra={
+                "session_id": session_id,
+                "chunk_count": chunk_count,
+                "first_chunk_type": first_chunk_type,
+            },
+        )
         # AI SDK protocol termination
         yield "data: [DONE]\n\n"
 
@@ -275,6 +296,8 @@ async def stream_chat_get(
     session = await _validate_and_get_session(session_id, user_id)
 
     async def event_generator() -> AsyncGenerator[str, None]:
+        chunk_count = 0
+        first_chunk_type: str | None = None
         async for chunk in chat_service.stream_chat_completion(
             session_id,
             message,
@@ -282,7 +305,26 @@ async def stream_chat_get(
             user_id=user_id,
             session=session,  # Pass pre-fetched session to avoid double-fetch
         ):
+            if chunk_count < 3:
+                logger.info(
+                    "Chat stream chunk",
+                    extra={
+                        "session_id": session_id,
+                        "chunk_type": str(chunk.type),
+                    },
+                )
+            if not first_chunk_type:
+                first_chunk_type = str(chunk.type)
+            chunk_count += 1
             yield chunk.to_sse()
+        logger.info(
+            "Chat stream completed",
+            extra={
+                "session_id": session_id,
+                "chunk_count": chunk_count,
+                "first_chunk_type": first_chunk_type,
+            },
+        )
         # AI SDK protocol termination
         yield "data: [DONE]\n\n"
 

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

@@ -1,15 +1,18 @@
 import asyncio
 import logging
+import time
+from asyncio import CancelledError
 from collections.abc import AsyncGenerator
 from typing import Any
 
 import orjson
-from langfuse import Langfuse
+from langfuse import get_client, propagate_attributes
+from langfuse.openai import openai  # type: ignore
 from openai import (
     APIConnectionError,
     APIError,
     APIStatusError,
-    AsyncOpenAI,
+    PermissionDeniedError,
     RateLimitError,
 )
 from openai.types.chat import ChatCompletionChunk, ChatCompletionToolParam
@@ -21,12 +24,12 @@ from backend.data.understanding import (
 from backend.util.exceptions import NotFoundError
 from backend.util.settings import Settings
 
-from . import db as chat_db
 from .config import ChatConfig
 from .model import (
     ChatMessage,
     ChatSession,
     Usage,
+    cache_chat_session,
     get_chat_session,
     update_session_title,
     upsert_chat_session,
@@ -50,10 +53,10 @@ logger = logging.getLogger(__name__)
 
 config = ChatConfig()
 settings = Settings()
-client = AsyncOpenAI(api_key=config.api_key, base_url=config.base_url)
+client = openai.AsyncOpenAI(api_key=config.api_key, base_url=config.base_url)
 
-# Langfuse client (lazy initialization)
+
-_langfuse_client: Langfuse | None = None
+langfuse = get_client()
 
 
 class LangfuseNotConfiguredError(Exception):
@@ -69,65 +72,6 @@ def _is_langfuse_configured() -> bool:
     )
 
 
-def _get_langfuse_client() -> Langfuse:
-    """Get or create the Langfuse client for prompt management and tracing."""
-    global _langfuse_client
-    if _langfuse_client is None:
-        if not _is_langfuse_configured():
-            raise LangfuseNotConfiguredError(
-                "Langfuse is not configured. The chat feature requires Langfuse for prompt management. "
-                "Please set the LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY environment variables."
-            )
-        _langfuse_client = Langfuse(
-            public_key=settings.secrets.langfuse_public_key,
-            secret_key=settings.secrets.langfuse_secret_key,
-            host=settings.secrets.langfuse_host or "https://cloud.langfuse.com",
-        )
-    return _langfuse_client
-
-
-def _get_environment() -> str:
-    """Get the current environment name for Langfuse tagging."""
-    return settings.config.app_env.value
-
-
-def _get_langfuse_prompt() -> str:
-    """Fetch the latest production prompt from Langfuse.
-
-    Returns:
-        The compiled prompt text from Langfuse.
-
-    Raises:
-        Exception: If Langfuse is unavailable or prompt fetch fails.
-    """
-    try:
-        langfuse = _get_langfuse_client()
-        # cache_ttl_seconds=0 disables SDK caching to always get the latest prompt
-        prompt = langfuse.get_prompt(config.langfuse_prompt_name, cache_ttl_seconds=0)
-        compiled = prompt.compile()
-        logger.info(
-            f"Fetched prompt '{config.langfuse_prompt_name}' from Langfuse "
-            f"(version: {prompt.version})"
-        )
-        return compiled
-    except Exception as e:
-        logger.error(f"Failed to fetch prompt from Langfuse: {e}")
-        raise
-
-
-async def _is_first_session(user_id: str) -> bool:
-    """Check if this is the user's first chat session.
-
-    Returns True if the user has 1 or fewer sessions (meaning this is their first).
-    """
-    try:
-        session_count = await chat_db.get_user_session_count(user_id)
-        return session_count <= 1
-    except Exception as e:
-        logger.warning(f"Failed to check session count for user {user_id}: {e}")
-        return False  # Default to non-onboarding if we can't check
-
-
 async def _build_system_prompt(user_id: str | None) -> tuple[str, Any]:
     """Build the full system prompt including business understanding if available.
 
@@ -139,8 +83,6 @@ async def _build_system_prompt(user_id: str | None) -> tuple[str, Any]:
         Tuple of (compiled prompt string, Langfuse prompt object for tracing)
     """
 
-    langfuse = _get_langfuse_client()
-
     # cache_ttl_seconds=0 disables SDK caching to always get the latest prompt
     prompt = langfuse.get_prompt(config.langfuse_prompt_name, cache_ttl_seconds=0)
 
@@ -158,7 +100,7 @@ async def _build_system_prompt(user_id: str | None) -> tuple[str, Any]:
         context = "This is the first time you are meeting the user. Greet them and introduce them to the platform"
 
     compiled = prompt.compile(users_information=context)
-    return compiled, prompt
+    return compiled, understanding
 
 
 async def _generate_session_title(message: str) -> str | None:
@@ -217,6 +159,7 @@ async def assign_user_to_session(
 async def stream_chat_completion(
     session_id: str,
     message: str | None = None,
+    tool_call_response: str | None = None,
     is_user_message: bool = True,
     user_id: str | None = None,
     retry_count: int = 0,
@@ -256,11 +199,6 @@ async def stream_chat_completion(
         yield StreamFinish()
         return
 
-    # Langfuse observations will be created after session is loaded (need messages for input)
-    # Initialize to None so finally block can safely check and end them
-    trace = None
-    generation = None
-
     # Only fetch from Redis if session not provided (initial call)
     if session is None:
         session = await get_chat_session(session_id, user_id)
@@ -336,297 +274,349 @@ async def stream_chat_completion(
             asyncio.create_task(_update_title())
 
     # Build system prompt with business understanding
-    system_prompt, langfuse_prompt = await _build_system_prompt(user_id)
+    system_prompt, understanding = await _build_system_prompt(user_id)
-
-    # Build input messages including system prompt for complete Langfuse logging
-    trace_input_messages = [{"role": "system", "content": system_prompt}] + [
-        m.model_dump() for m in session.messages
-    ]
 
     # Create Langfuse trace for this LLM call (each call gets its own trace, grouped by session_id)
     # Using v3 SDK: start_observation creates a root span, update_trace sets trace-level attributes
-    try:
+    input = message
-        langfuse = _get_langfuse_client()
+    if not message and tool_call_response:
-        env = _get_environment()
+        input = tool_call_response
-        trace = langfuse.start_observation(
+
-            name="chat_completion",
+    langfuse = get_client()
-            input={"messages": trace_input_messages},
+    with langfuse.start_as_current_observation(
-            metadata={
+        as_type="span",
-                "environment": env,
+        name="user-copilot-request",
-                "model": config.model,
+        input=input,
-                "message_count": len(session.messages),
+    ) as span:
-                "prompt_name": langfuse_prompt.name if langfuse_prompt else None,
+        with propagate_attributes(
-                "prompt_version": langfuse_prompt.version if langfuse_prompt else None,
-            },
-        )
-        # Set trace-level attributes (session_id, user_id, tags)
-        trace.update_trace(
             session_id=session_id,
             user_id=user_id,
-            tags=[env, "copilot"],
+            tags=["copilot"],
-        )
+            metadata={
-    except Exception as e:
+                "users_information": format_understanding_for_prompt(understanding)[
-        logger.warning(f"Failed to create Langfuse trace: {e}")
+                    :200
+                ]  # langfuse only accepts upto to 200 chars
+            },
+        ):
 
-    # Initialize variables that will be used in finally block (must be defined before try)
+            # Initialize variables that will be used in finally block (must be defined before try)
-    assistant_response = ChatMessage(
+            assistant_response = ChatMessage(
-        role="assistant",
+                role="assistant",
-        content="",
+                content="",
-    )
-    accumulated_tool_calls: list[dict[str, Any]] = []
-
-    # Wrap main logic in try/finally to ensure Langfuse observations are always ended
-    try:
-        has_yielded_end = False
-        has_yielded_error = False
-        has_done_tool_call = False
-        has_received_text = False
-        text_streaming_ended = False
-        tool_response_messages: list[ChatMessage] = []
-        should_retry = False
-
-        # Generate unique IDs for AI SDK protocol
-        import uuid as uuid_module
-
-        message_id = str(uuid_module.uuid4())
-        text_block_id = str(uuid_module.uuid4())
-
-        # Yield message start
-        yield StreamStart(messageId=message_id)
-
-        # Create Langfuse generation for each LLM call, linked to the prompt
-        # Using v3 SDK: start_observation with as_type="generation"
-        generation = (
-            trace.start_observation(
-                as_type="generation",
-                name="llm_call",
-                model=config.model,
-                input={"messages": trace_input_messages},
-                prompt=langfuse_prompt,
             )
-            if trace
+            accumulated_tool_calls: list[dict[str, Any]] = []
-            else None
+            has_saved_assistant_message = False
-        )
+            has_appended_streaming_message = False
+            last_cache_time = 0.0
+            last_cache_content_len = 0
 
-        try:
+            # Wrap main logic in try/finally to ensure Langfuse observations are always ended
-            async for chunk in _stream_chat_chunks(
+            has_yielded_end = False
-                session=session,
+            has_yielded_error = False
-                tools=tools,
+            has_done_tool_call = False
-                system_prompt=system_prompt,
+            has_received_text = False
-                text_block_id=text_block_id,
+            text_streaming_ended = False
-            ):
+            tool_response_messages: list[ChatMessage] = []
+            should_retry = False
 
-                if isinstance(chunk, StreamTextStart):
+            # Generate unique IDs for AI SDK protocol
-                    # Emit text-start before first text delta
+            import uuid as uuid_module
-                    if not has_received_text:
+
+            message_id = str(uuid_module.uuid4())
+            text_block_id = str(uuid_module.uuid4())
+
+            # Yield message start
+            yield StreamStart(messageId=message_id)
+
+            try:
+                async for chunk in _stream_chat_chunks(
+                    session=session,
+                    tools=tools,
+                    system_prompt=system_prompt,
+                    text_block_id=text_block_id,
+                ):
+
+                    if isinstance(chunk, StreamTextStart):
+                        # Emit text-start before first text delta
+                        if not has_received_text:
+                            yield chunk
+                    elif isinstance(chunk, StreamTextDelta):
+                        delta = chunk.delta or ""
+                        assert assistant_response.content is not None
+                        assistant_response.content += delta
+                        has_received_text = True
+                        if not has_appended_streaming_message:
+                            session.messages.append(assistant_response)
+                            has_appended_streaming_message = True
+                        current_time = time.monotonic()
+                        content_len = len(assistant_response.content)
+                        if (
+                            current_time - last_cache_time >= 1.0
+                            and content_len > last_cache_content_len
+                        ):
+                            try:
+                                await cache_chat_session(session)
+                            except Exception as e:
+                                logger.warning(
+                                    f"Failed to cache partial session {session.session_id}: {e}"
+                                )
+                            last_cache_time = current_time
+                            last_cache_content_len = content_len
                         yield chunk
-                elif isinstance(chunk, StreamTextDelta):
+                    elif isinstance(chunk, StreamTextEnd):
-                    delta = chunk.delta or ""
+                        # Emit text-end after text completes
-                    assert assistant_response.content is not None
+                        if has_received_text and not text_streaming_ended:
-                    assistant_response.content += delta
+                            text_streaming_ended = True
-                    has_received_text = True
+                            if assistant_response.content:
-                    yield chunk
+                                logger.warn(
-                elif isinstance(chunk, StreamTextEnd):
+                                    f"StreamTextEnd: Attempting to set output {assistant_response.content}"
-                    # Emit text-end after text completes
+                                )
-                    if has_received_text and not text_streaming_ended:
+                                span.update_trace(output=assistant_response.content)
-                        text_streaming_ended = True
+                                span.update(output=assistant_response.content)
-                        yield chunk
+                            yield chunk
-                elif isinstance(chunk, StreamToolInputStart):
+                    elif isinstance(chunk, StreamToolInputStart):
-                    # Emit text-end before first tool call, but only if we've received text
+                        # Emit text-end before first tool call, but only if we've received text
-                    if has_received_text and not text_streaming_ended:
-                        yield StreamTextEnd(id=text_block_id)
-                        text_streaming_ended = True
-                    yield chunk
-                elif isinstance(chunk, StreamToolInputAvailable):
-                    # Accumulate tool calls in OpenAI format
-                    accumulated_tool_calls.append(
-                        {
-                            "id": chunk.toolCallId,
-                            "type": "function",
-                            "function": {
-                                "name": chunk.toolName,
-                                "arguments": orjson.dumps(chunk.input).decode("utf-8"),
-                            },
-                        }
-                    )
-                elif isinstance(chunk, StreamToolOutputAvailable):
-                    result_content = (
-                        chunk.output
-                        if isinstance(chunk.output, str)
-                        else orjson.dumps(chunk.output).decode("utf-8")
-                    )
-                    tool_response_messages.append(
-                        ChatMessage(
-                            role="tool",
-                            content=result_content,
-                            tool_call_id=chunk.toolCallId,
-                        )
-                    )
-                    has_done_tool_call = True
-                    # Track if any tool execution failed
-                    if not chunk.success:
-                        logger.warning(
-                            f"Tool {chunk.toolName} (ID: {chunk.toolCallId}) execution failed"
-                        )
-                    yield chunk
-                elif isinstance(chunk, StreamFinish):
-                    if not has_done_tool_call:
-                        # Emit text-end before finish if we received text but haven't closed it
                         if has_received_text and not text_streaming_ended:
                             yield StreamTextEnd(id=text_block_id)
                             text_streaming_ended = True
-                        has_yielded_end = True
                         yield chunk
-                elif isinstance(chunk, StreamError):
+                    elif isinstance(chunk, StreamToolInputAvailable):
-                    has_yielded_error = True
+                        # Accumulate tool calls in OpenAI format
-                elif isinstance(chunk, StreamUsage):
+                        accumulated_tool_calls.append(
-                    session.usage.append(
+                            {
-                        Usage(
+                                "id": chunk.toolCallId,
-                            prompt_tokens=chunk.promptTokens,
+                                "type": "function",
-                            completion_tokens=chunk.completionTokens,
+                                "function": {
-                            total_tokens=chunk.totalTokens,
+                                    "name": chunk.toolName,
+                                    "arguments": orjson.dumps(chunk.input).decode(
+                                        "utf-8"
+                                    ),
+                                },
+                            }
                         )
-                    )
+                    elif isinstance(chunk, StreamToolOutputAvailable):
-                else:
+                        result_content = (
-                    logger.error(f"Unknown chunk type: {type(chunk)}", exc_info=True)
+                            chunk.output
-        except Exception as e:
+                            if isinstance(chunk.output, str)
-            logger.error(f"Error during stream: {e!s}", exc_info=True)
+                            else orjson.dumps(chunk.output).decode("utf-8")
+                        )
+                        tool_response_messages.append(
+                            ChatMessage(
+                                role="tool",
+                                content=result_content,
+                                tool_call_id=chunk.toolCallId,
+                            )
+                        )
+                        has_done_tool_call = True
+                        # Track if any tool execution failed
+                        if not chunk.success:
+                            logger.warning(
+                                f"Tool {chunk.toolName} (ID: {chunk.toolCallId}) execution failed"
+                            )
+                        yield chunk
+                    elif isinstance(chunk, StreamFinish):
+                        if not has_done_tool_call:
+                            # Emit text-end before finish if we received text but haven't closed it
+                            if has_received_text and not text_streaming_ended:
+                                yield StreamTextEnd(id=text_block_id)
+                                text_streaming_ended = True
 
-            # Check if this is a retryable error (JSON parsing, incomplete tool calls, etc.)
+                            # Save assistant message before yielding finish to ensure it's persisted
-            is_retryable = isinstance(e, (orjson.JSONDecodeError, KeyError, TypeError))
+                            # even if client disconnects immediately after receiving StreamFinish
+                            if not has_saved_assistant_message:
+                                messages_to_save_early: list[ChatMessage] = []
+                                if accumulated_tool_calls:
+                                    assistant_response.tool_calls = (
+                                        accumulated_tool_calls
+                                    )
+                                if not has_appended_streaming_message and (
+                                    assistant_response.content
+                                    or assistant_response.tool_calls
+                                ):
+                                    messages_to_save_early.append(assistant_response)
+                                messages_to_save_early.extend(tool_response_messages)
 
-            if is_retryable and retry_count < config.max_retries:
+                                if messages_to_save_early:
-                logger.info(
+                                    session.messages.extend(messages_to_save_early)
-                    f"Retryable error encountered. Attempt {retry_count + 1}/{config.max_retries}"
+                                    logger.info(
+                                        f"Saving assistant message before StreamFinish: "
+                                        f"content_len={len(assistant_response.content or '')}, "
+                                        f"tool_calls={len(assistant_response.tool_calls or [])}, "
+                                        f"tool_responses={len(tool_response_messages)}"
+                                    )
+                                if (
+                                    messages_to_save_early
+                                    or has_appended_streaming_message
+                                ):
+                                    await upsert_chat_session(session)
+                                    has_saved_assistant_message = True
+
+                            has_yielded_end = True
+                            yield chunk
+                    elif isinstance(chunk, StreamError):
+                        has_yielded_error = True
+                        yield chunk
+                    elif isinstance(chunk, StreamUsage):
+                        session.usage.append(
+                            Usage(
+                                prompt_tokens=chunk.promptTokens,
+                                completion_tokens=chunk.completionTokens,
+                                total_tokens=chunk.totalTokens,
+                            )
+                        )
+                    else:
+                        logger.error(
+                            f"Unknown chunk type: {type(chunk)}", exc_info=True
+                        )
+                if assistant_response.content:
+                    langfuse.update_current_trace(output=assistant_response.content)
+                    langfuse.update_current_span(output=assistant_response.content)
+                elif tool_response_messages:
+                    langfuse.update_current_trace(output=str(tool_response_messages))
+                    langfuse.update_current_span(output=str(tool_response_messages))
+
+            except CancelledError:
+                if not has_saved_assistant_message:
+                    if accumulated_tool_calls:
+                        assistant_response.tool_calls = accumulated_tool_calls
+                    if assistant_response.content:
+                        assistant_response.content = (
+                            f"{assistant_response.content}\n\n[interrupted]"
+                        )
+                    else:
+                        assistant_response.content = "[interrupted]"
+                    if not has_appended_streaming_message:
+                        session.messages.append(assistant_response)
+                    if tool_response_messages:
+                        session.messages.extend(tool_response_messages)
+                    try:
+                        await upsert_chat_session(session)
+                    except Exception as e:
+                        logger.warning(
+                            f"Failed to save interrupted session {session.session_id}: {e}"
+                        )
+                raise
+            except Exception as e:
+                logger.error(f"Error during stream: {e!s}", exc_info=True)
+
+                # Check if this is a retryable error (JSON parsing, incomplete tool calls, etc.)
+                is_retryable = isinstance(
+                    e, (orjson.JSONDecodeError, KeyError, TypeError)
                 )
-                should_retry = True
+
-            else:
+                if is_retryable and retry_count < config.max_retries:
-                # Non-retryable error or max retries exceeded
+                    logger.info(
-                # Save any partial progress before reporting error
+                        f"Retryable error encountered. Attempt {retry_count + 1}/{config.max_retries}"
+                    )
+                    should_retry = True
+                else:
+                    # Non-retryable error or max retries exceeded
+                    # Save any partial progress before reporting error
+                    messages_to_save: list[ChatMessage] = []
+
+                    # Add assistant message if it has content or tool calls
+                    if accumulated_tool_calls:
+                        assistant_response.tool_calls = accumulated_tool_calls
+                    if not has_appended_streaming_message and (
+                        assistant_response.content or assistant_response.tool_calls
+                    ):
+                        messages_to_save.append(assistant_response)
+
+                    # Add tool response messages after assistant message
+                    messages_to_save.extend(tool_response_messages)
+
+                    if not has_saved_assistant_message:
+                        if messages_to_save:
+                            session.messages.extend(messages_to_save)
+                        if messages_to_save or has_appended_streaming_message:
+                            await upsert_chat_session(session)
+
+                    if not has_yielded_error:
+                        error_message = str(e)
+                        if not is_retryable:
+                            error_message = f"Non-retryable error: {error_message}"
+                        elif retry_count >= config.max_retries:
+                            error_message = f"Max retries ({config.max_retries}) exceeded: {error_message}"
+
+                        error_response = StreamError(errorText=error_message)
+                        yield error_response
+                    if not has_yielded_end:
+                        yield StreamFinish()
+                    return
+
+            # Handle retry outside of exception handler to avoid nesting
+            if should_retry and retry_count < config.max_retries:
+                logger.info(
+                    f"Retrying stream_chat_completion for session {session_id}, attempt {retry_count + 1}"
+                )
+                async for chunk in stream_chat_completion(
+                    session_id=session.session_id,
+                    user_id=user_id,
+                    retry_count=retry_count + 1,
+                    session=session,
+                    context=context,
+                ):
+                    yield chunk
+                return  # Exit after retry to avoid double-saving in finally block
+
+            # Normal completion path - save session and handle tool call continuation
+            # Only save if we haven't already saved when StreamFinish was received
+            if not has_saved_assistant_message:
+                logger.info(
+                    f"Normal completion path: session={session.session_id}, "
+                    f"current message_count={len(session.messages)}"
+                )
+
+                # Build the messages list in the correct order
                 messages_to_save: list[ChatMessage] = []
 
-                # Add assistant message if it has content or tool calls
+                # Add assistant message with tool_calls if any
                 if accumulated_tool_calls:
                     assistant_response.tool_calls = accumulated_tool_calls
-                if assistant_response.content or assistant_response.tool_calls:
+                    logger.info(
+                        f"Added {len(accumulated_tool_calls)} tool calls to assistant message"
+                    )
+                if not has_appended_streaming_message and (
+                    assistant_response.content or assistant_response.tool_calls
+                ):
                     messages_to_save.append(assistant_response)
+                    logger.info(
+                        f"Saving assistant message with content_len={len(assistant_response.content or '')}, tool_calls={len(assistant_response.tool_calls or [])}"
+                    )
 
                 # Add tool response messages after assistant message
                 messages_to_save.extend(tool_response_messages)
-
+                logger.info(
-                session.messages.extend(messages_to_save)
+                    f"Saving {len(tool_response_messages)} tool response messages, "
-                await upsert_chat_session(session)
+                    f"total_to_save={len(messages_to_save)}"
-
-                if not has_yielded_error:
-                    error_message = str(e)
-                    if not is_retryable:
-                        error_message = f"Non-retryable error: {error_message}"
-                    elif retry_count >= config.max_retries:
-                        error_message = f"Max retries ({config.max_retries}) exceeded: {error_message}"
-
-                    error_response = StreamError(errorText=error_message)
-                    yield error_response
-                if not has_yielded_end:
-                    yield StreamFinish()
-                return
-
-        # Handle retry outside of exception handler to avoid nesting
-        if should_retry and retry_count < config.max_retries:
-            logger.info(
-                f"Retrying stream_chat_completion for session {session_id}, attempt {retry_count + 1}"
-            )
-            async for chunk in stream_chat_completion(
-                session_id=session.session_id,
-                user_id=user_id,
-                retry_count=retry_count + 1,
-                session=session,
-                context=context,
-            ):
-                yield chunk
-            return  # Exit after retry to avoid double-saving in finally block
-
-        # Normal completion path - save session and handle tool call continuation
-        logger.info(
-            f"Normal completion path: session={session.session_id}, "
-            f"current message_count={len(session.messages)}"
-        )
-
-        # Build the messages list in the correct order
-        messages_to_save: list[ChatMessage] = []
-
-        # Add assistant message with tool_calls if any
-        if accumulated_tool_calls:
-            assistant_response.tool_calls = accumulated_tool_calls
-            logger.info(
-                f"Added {len(accumulated_tool_calls)} tool calls to assistant message"
-            )
-        if assistant_response.content or assistant_response.tool_calls:
-            messages_to_save.append(assistant_response)
-            logger.info(
-                f"Saving assistant message with content_len={len(assistant_response.content or '')}, tool_calls={len(assistant_response.tool_calls or [])}"
-            )
-
-        # Add tool response messages after assistant message
-        messages_to_save.extend(tool_response_messages)
-        logger.info(
-            f"Saving {len(tool_response_messages)} tool response messages, "
-            f"total_to_save={len(messages_to_save)}"
-        )
-
-        session.messages.extend(messages_to_save)
-        logger.info(
-            f"Extended session messages, new message_count={len(session.messages)}"
-        )
-        await upsert_chat_session(session)
-
-        # If we did a tool call, stream the chat completion again to get the next response
-        if has_done_tool_call:
-            logger.info(
-                "Tool call executed, streaming chat completion again to get assistant response"
-            )
-            async for chunk in stream_chat_completion(
-                session_id=session.session_id,
-                user_id=user_id,
-                session=session,  # Pass session object to avoid Redis refetch
-                context=context,
-            ):
-                yield chunk
-
-    finally:
-        # Always end Langfuse observations to prevent resource leaks
-        # Guard against None and catch errors to avoid masking original exceptions
-        if generation is not None:
-            try:
-                latest_usage = session.usage[-1] if session.usage else None
-                generation.update(
-                    model=config.model,
-                    output={
-                        "content": assistant_response.content,
-                        "tool_calls": accumulated_tool_calls or None,
-                    },
-                    usage_details=(
-                        {
-                            "input": latest_usage.prompt_tokens,
-                            "output": latest_usage.completion_tokens,
-                            "total": latest_usage.total_tokens,
-                        }
-                        if latest_usage
-                        else None
-                    ),
                 )
-                generation.end()
-            except Exception as e:
-                logger.warning(f"Failed to end Langfuse generation: {e}")
 
-        if trace is not None:
+                if messages_to_save:
-            try:
+                    session.messages.extend(messages_to_save)
-                if accumulated_tool_calls:
+                    logger.info(
-                    trace.update_trace(output={"tool_calls": accumulated_tool_calls})
+                        f"Extended session messages, new message_count={len(session.messages)}"
-                else:
+                    )
-                    trace.update_trace(output={"response": assistant_response.content})
+                if messages_to_save or has_appended_streaming_message:
-                trace.end()
+                    await upsert_chat_session(session)
-            except Exception as e:
+            else:
-                logger.warning(f"Failed to end Langfuse trace: {e}")
+                logger.info(
+                    "Assistant message already saved when StreamFinish was received, "
+                    "skipping duplicate save"
+                )
+
+            # If we did a tool call, stream the chat completion again to get the next response
+            if has_done_tool_call:
+                logger.info(
+                    "Tool call executed, streaming chat completion again to get assistant response"
+                )
+                async for chunk in stream_chat_completion(
+                    session_id=session.session_id,
+                    user_id=user_id,
+                    session=session,  # Pass session object to avoid Redis refetch
+                    context=context,
+                    tool_call_response=str(tool_response_messages),
+                ):
+                    yield chunk
 
 
 # Retry configuration for OpenAI API calls
@@ -654,6 +644,12 @@ def _is_retryable_error(error: Exception) -> bool:
     return False
 
 
+def _is_region_blocked_error(error: Exception) -> bool:
+    if isinstance(error, PermissionDeniedError):
+        return "not available in your region" in str(error).lower()
+    return "not available in your region" in str(error).lower()
+
+
 async def _stream_chat_chunks(
     session: ChatSession,
     tools: list[ChatCompletionToolParam],
@@ -846,7 +842,18 @@ async def _stream_chat_chunks(
                         f"Error in stream (not retrying): {e!s}",
                         exc_info=True,
                     )
-                    error_response = StreamError(errorText=str(e))
+                    error_code = None
+                    error_text = str(e)
+                    if _is_region_blocked_error(e):
+                        error_code = "MODEL_NOT_AVAILABLE_REGION"
+                        error_text = (
+                            "This model is not available in your region. "
+                            "Please connect via VPN and try again."
+                        )
+                    error_response = StreamError(
+                        errorText=error_text,
+                        code=error_code,
+                    )
                     yield error_response
                     yield StreamFinish()
                     return
@@ -900,5 +907,4 @@ async def _yield_tool_call(
         session=session,
     )
 
-    logger.info(f"Yielding Tool execution response: {tool_execution_response}")
     yield tool_execution_response

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

@@ -30,7 +30,7 @@ TOOL_REGISTRY: dict[str, BaseTool] = {
     "find_library_agent": FindLibraryAgentTool(),
     "run_agent": RunAgentTool(),
     "run_block": RunBlockTool(),
-    "agent_output": AgentOutputTool(),
+    "view_agent_output": AgentOutputTool(),
     "search_docs": SearchDocsTool(),
     "get_doc_page": GetDocPageTool(),
 }

autogpt_platform/backend/backend/api/features/chat/tools/add_understanding.py

@@ -3,6 +3,8 @@
 import logging
 from typing import Any
 
+from langfuse import observe
+
 from backend.api.features.chat.model import ChatSession
 from backend.data.understanding import (
     BusinessUnderstandingInput,
@@ -59,6 +61,7 @@ and automations for the user's specific needs."""
         """Requires authentication to store user-specific data."""
         return True
 
+    @observe(as_type="tool", name="add_understanding")
     async def _execute(
         self,
         user_id: str | None,

autogpt_platform/backend/backend/api/features/chat/tools/agent_generator/core.py

@@ -218,6 +218,7 @@ async def save_agent_to_library(
     library_agents = await library_db.create_library_agent(
         graph=created_graph,
         user_id=user_id,
+        sensitive_action_safe_mode=True,
         create_library_agents_for_sub_graphs=False,
     )
 

autogpt_platform/backend/backend/api/features/chat/tools/agent_output.py

@@ -5,6 +5,7 @@ import re
 from datetime import datetime, timedelta, timezone
 from typing import Any
 
+from langfuse import observe
 from pydantic import BaseModel, field_validator
 
 from backend.api.features.chat.model import ChatSession
@@ -103,7 +104,7 @@ class AgentOutputTool(BaseTool):
 
     @property
     def name(self) -> str:
-        return "agent_output"
+        return "view_agent_output"
 
     @property
     def description(self) -> str:
@@ -328,6 +329,7 @@ class AgentOutputTool(BaseTool):
             total_executions=len(available_executions) if available_executions else 1,
         )
 
+    @observe(as_type="tool", name="view_agent_output")
     async def _execute(
         self,
         user_id: str | None,

autogpt_platform/backend/backend/api/features/chat/tools/create_agent.py

@@ -3,6 +3,8 @@
 import logging
 from typing import Any
 
+from langfuse import observe
+
 from backend.api.features.chat.model import ChatSession
 
 from .agent_generator import (
@@ -78,6 +80,7 @@ class CreateAgentTool(BaseTool):
             "required": ["description"],
         }
 
+    @observe(as_type="tool", name="create_agent")
     async def _execute(
         self,
         user_id: str | None,

autogpt_platform/backend/backend/api/features/chat/tools/edit_agent.py

@@ -3,6 +3,8 @@
 import logging
 from typing import Any
 
+from langfuse import observe
+
 from backend.api.features.chat.model import ChatSession
 
 from .agent_generator import (
@@ -85,6 +87,7 @@ class EditAgentTool(BaseTool):
             "required": ["agent_id", "changes"],
         }
 
+    @observe(as_type="tool", name="edit_agent")
     async def _execute(
         self,
         user_id: str | None,

autogpt_platform/backend/backend/api/features/chat/tools/find_agent.py

@@ -2,6 +2,8 @@
 
 from typing import Any
 
+from langfuse import observe
+
 from backend.api.features.chat.model import ChatSession
 
 from .agent_search import search_agents
@@ -35,6 +37,7 @@ class FindAgentTool(BaseTool):
             "required": ["query"],
         }
 
+    @observe(as_type="tool", name="find_agent")
     async def _execute(
         self, user_id: str | None, session: ChatSession, **kwargs
     ) -> ToolResponseBase:

autogpt_platform/backend/backend/api/features/chat/tools/find_block.py

@@ -1,6 +1,7 @@
 import logging
 from typing import Any
 
+from langfuse import observe
 from prisma.enums import ContentType
 
 from backend.api.features.chat.model import ChatSession
@@ -55,6 +56,7 @@ class FindBlockTool(BaseTool):
     def requires_auth(self) -> bool:
         return True
 
+    @observe(as_type="tool", name="find_block")
     async def _execute(
         self,
         user_id: str | None,

autogpt_platform/backend/backend/api/features/chat/tools/find_library_agent.py

@@ -2,6 +2,8 @@
 
 from typing import Any
 
+from langfuse import observe
+
 from backend.api.features.chat.model import ChatSession
 
 from .agent_search import search_agents
@@ -41,6 +43,7 @@ class FindLibraryAgentTool(BaseTool):
     def requires_auth(self) -> bool:
         return True
 
+    @observe(as_type="tool", name="find_library_agent")
     async def _execute(
         self, user_id: str | None, session: ChatSession, **kwargs
     ) -> ToolResponseBase:

autogpt_platform/backend/backend/api/features/chat/tools/get_doc_page.py

@@ -4,6 +4,8 @@ import logging
 from pathlib import Path
 from typing import Any
 
+from langfuse import observe
+
 from backend.api.features.chat.model import ChatSession
 from backend.api.features.chat.tools.base import BaseTool
 from backend.api.features.chat.tools.models import (
@@ -71,6 +73,7 @@ class GetDocPageTool(BaseTool):
         url_path = path.rsplit(".", 1)[0] if "." in path else path
         return f"{DOCS_BASE_URL}/{url_path}"
 
+    @observe(as_type="tool", name="get_doc_page")
     async def _execute(
         self,
         user_id: str | None,

autogpt_platform/backend/backend/api/features/chat/tools/run_agent.py

@@ -3,6 +3,7 @@
 import logging
 from typing import Any
 
+from langfuse import observe
 from pydantic import BaseModel, Field, field_validator
 
 from backend.api.features.chat.config import ChatConfig
@@ -32,7 +33,7 @@ from .models import (
     UserReadiness,
 )
 from .utils import (
-    check_user_has_required_credentials,
+    build_missing_credentials_from_graph,
     extract_credentials_from_schema,
     fetch_graph_from_store_slug,
     get_or_create_library_agent,
@@ -154,6 +155,7 @@ class RunAgentTool(BaseTool):
         """All operations require authentication."""
         return True
 
+    @observe(as_type="tool", name="run_agent")
     async def _execute(
         self,
         user_id: str | None,
@@ -235,15 +237,13 @@ class RunAgentTool(BaseTool):
                 # Return credentials needed response with input data info
                 # The UI handles credential setup automatically, so the message
                 # focuses on asking about input data
-                credentials = extract_credentials_from_schema(
+                requirements_creds_dict = build_missing_credentials_from_graph(
-                    graph.credentials_input_schema
+                    graph, None
                 )
-                missing_creds_check = await check_user_has_required_credentials(
+                missing_credentials_dict = build_missing_credentials_from_graph(
-                    user_id, credentials
+                    graph, graph_credentials
                 )
-                missing_credentials_dict = {
+                requirements_creds_list = list(requirements_creds_dict.values())
-                    c.id: c.model_dump() for c in missing_creds_check
-                }
 
                 return SetupRequirementsResponse(
                     message=self._build_inputs_message(graph, MSG_WHAT_VALUES_TO_USE),
@@ -257,7 +257,7 @@ class RunAgentTool(BaseTool):
                             ready_to_run=False,
                         ),
                         requirements={
-                            "credentials": [c.model_dump() for c in credentials],
+                            "credentials": requirements_creds_list,
                             "inputs": self._get_inputs_list(graph.input_schema),
                             "execution_modes": self._get_execution_modes(graph),
                         },

autogpt_platform/backend/backend/api/features/chat/tools/run_block.py

@@ -4,6 +4,8 @@ import logging
 from collections import defaultdict
 from typing import Any
 
+from langfuse import observe
+
 from backend.api.features.chat.model import ChatSession
 from backend.data.block import get_block
 from backend.data.execution import ExecutionContext
@@ -20,6 +22,7 @@ from .models import (
     ToolResponseBase,
     UserReadiness,
 )
+from .utils import build_missing_credentials_from_field_info
 
 logger = logging.getLogger(__name__)
 
@@ -127,6 +130,7 @@ class RunBlockTool(BaseTool):
 
         return matched_credentials, missing_credentials
 
+    @observe(as_type="tool", name="run_block")
     async def _execute(
         self,
         user_id: str | None,
@@ -186,7 +190,11 @@ class RunBlockTool(BaseTool):
 
         if missing_credentials:
             # Return setup requirements response with missing credentials
-            missing_creds_dict = {c.id: c.model_dump() for c in missing_credentials}
+            credentials_fields_info = block.input_schema.get_credentials_fields_info()
+            missing_creds_dict = build_missing_credentials_from_field_info(
+                credentials_fields_info, set(matched_credentials.keys())
+            )
+            missing_creds_list = list(missing_creds_dict.values())
 
             return SetupRequirementsResponse(
                 message=(
@@ -203,7 +211,7 @@ class RunBlockTool(BaseTool):
                         ready_to_run=False,
                     ),
                     requirements={
-                        "credentials": [c.model_dump() for c in missing_credentials],
+                        "credentials": missing_creds_list,
                         "inputs": self._get_inputs_list(block),
                         "execution_modes": ["immediate"],
                     },

autogpt_platform/backend/backend/api/features/chat/tools/search_docs.py

@@ -3,6 +3,7 @@
 import logging
 from typing import Any
 
+from langfuse import observe
 from prisma.enums import ContentType
 
 from backend.api.features.chat.model import ChatSession
@@ -87,6 +88,7 @@ class SearchDocsTool(BaseTool):
         url_path = path.rsplit(".", 1)[0] if "." in path else path
         return f"{DOCS_BASE_URL}/{url_path}"
 
+    @observe(as_type="tool", name="search_docs")
     async def _execute(
         self,
         user_id: str | None,

autogpt_platform/backend/backend/api/features/chat/tools/utils.py

@@ -8,7 +8,7 @@ from backend.api.features.library import model as library_model
 from backend.api.features.store import db as store_db
 from backend.data import graph as graph_db
 from backend.data.graph import GraphModel
-from backend.data.model import CredentialsMetaInput
+from backend.data.model import CredentialsFieldInfo, CredentialsMetaInput
 from backend.integrations.creds_manager import IntegrationCredentialsManager
 from backend.util.exceptions import NotFoundError
 
@@ -89,6 +89,59 @@ def extract_credentials_from_schema(
     return credentials
 
 
+def _serialize_missing_credential(
+    field_key: str, field_info: CredentialsFieldInfo
+) -> dict[str, Any]:
+    """
+    Convert credential field info into a serializable dict that preserves all supported
+    credential types (e.g., api_key + oauth2) so the UI can offer multiple options.
+    """
+    supported_types = sorted(field_info.supported_types)
+    provider = next(iter(field_info.provider), "unknown")
+    scopes = sorted(field_info.required_scopes or [])
+
+    return {
+        "id": field_key,
+        "title": field_key.replace("_", " ").title(),
+        "provider": provider,
+        "provider_name": provider.replace("_", " ").title(),
+        "type": supported_types[0] if supported_types else "api_key",
+        "types": supported_types,
+        "scopes": scopes,
+    }
+
+
+def build_missing_credentials_from_graph(
+    graph: GraphModel, matched_credentials: dict[str, CredentialsMetaInput] | None
+) -> dict[str, Any]:
+    """
+    Build a missing_credentials mapping from a graph's aggregated credentials inputs,
+    preserving all supported credential types for each field.
+    """
+    matched_keys = set(matched_credentials.keys()) if matched_credentials else set()
+    aggregated_fields = graph.aggregate_credentials_inputs()
+
+    return {
+        field_key: _serialize_missing_credential(field_key, field_info)
+        for field_key, (field_info, _node_fields) in aggregated_fields.items()
+        if field_key not in matched_keys
+    }
+
+
+def build_missing_credentials_from_field_info(
+    credential_fields: dict[str, CredentialsFieldInfo],
+    matched_keys: set[str],
+) -> dict[str, Any]:
+    """
+    Build missing_credentials mapping from a simple credentials field info dictionary.
+    """
+    return {
+        field_key: _serialize_missing_credential(field_key, field_info)
+        for field_key, field_info in credential_fields.items()
+        if field_key not in matched_keys
+    }
+
+
 def extract_credentials_as_dict(
     credentials_input_schema: dict[str, Any] | None,
 ) -> dict[str, CredentialsMetaInput]:

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

@@ -401,27 +401,11 @@ async def add_generated_agent_image(
     )
 
 
-def _initialize_graph_settings(graph: graph_db.GraphModel) -> GraphSettings:
-    """
-    Initialize GraphSettings based on graph content.
-
-    Args:
-        graph: The graph to analyze
-
-    Returns:
-        GraphSettings with appropriate human_in_the_loop_safe_mode value
-    """
-    if graph.has_human_in_the_loop:
-        # Graph has HITL blocks - set safe mode to True by default
-        return GraphSettings(human_in_the_loop_safe_mode=True)
-    else:
-        # Graph has no HITL blocks - keep None
-        return GraphSettings(human_in_the_loop_safe_mode=None)
-
-
 async def create_library_agent(
     graph: graph_db.GraphModel,
     user_id: str,
+    hitl_safe_mode: bool = True,
+    sensitive_action_safe_mode: bool = False,
     create_library_agents_for_sub_graphs: bool = True,
 ) -> list[library_model.LibraryAgent]:
     """
@@ -430,6 +414,8 @@ async def create_library_agent(
     Args:
         agent: The agent/Graph to add to the library.
         user_id: The user to whom the agent will be added.
+        hitl_safe_mode: Whether HITL blocks require manual review (default True).
+        sensitive_action_safe_mode: Whether sensitive action blocks require review.
         create_library_agents_for_sub_graphs: If True, creates LibraryAgent records for sub-graphs as well.
 
     Returns:
@@ -465,7 +451,11 @@ async def create_library_agent(
                             }
                         },
                         settings=SafeJson(
-                            _initialize_graph_settings(graph_entry).model_dump()
+                            GraphSettings.from_graph(
+                                graph_entry,
+                                hitl_safe_mode=hitl_safe_mode,
+                                sensitive_action_safe_mode=sensitive_action_safe_mode,
+                            ).model_dump()
                         ),
                     ),
                     include=library_agent_include(
@@ -627,33 +617,6 @@ async def update_library_agent(
         raise DatabaseError("Failed to update library agent") from e
 
 
-async def update_library_agent_settings(
-    user_id: str,
-    agent_id: str,
-    settings: GraphSettings,
-) -> library_model.LibraryAgent:
-    """
-    Updates the settings for a specific LibraryAgent.
-
-    Args:
-        user_id: The owner of the LibraryAgent.
-        agent_id: The ID of the LibraryAgent to update.
-        settings: New GraphSettings to apply.
-
-    Returns:
-        The updated LibraryAgent.
-
-    Raises:
-        NotFoundError: If the specified LibraryAgent does not exist.
-        DatabaseError: If there's an error in the update operation.
-    """
-    return await update_library_agent(
-        library_agent_id=agent_id,
-        user_id=user_id,
-        settings=settings,
-    )
-
-
 async def delete_library_agent(
     library_agent_id: str, user_id: str, soft_delete: bool = True
 ) -> None:
@@ -838,7 +801,7 @@ async def add_store_agent_to_library(
                 "isCreatedByUser": False,
                 "useGraphIsActiveVersion": False,
                 "settings": SafeJson(
-                    _initialize_graph_settings(graph_model).model_dump()
+                    GraphSettings.from_graph(graph_model).model_dump()
                 ),
             },
             include=library_agent_include(
@@ -1228,8 +1191,15 @@ async def fork_library_agent(
         )
         new_graph = await on_graph_activate(new_graph, user_id=user_id)
 
-        # Create a library agent for the new graph
+        # Create a library agent for the new graph, preserving safe mode settings
-        return (await create_library_agent(new_graph, user_id))[0]
+        return (
+            await create_library_agent(
+                new_graph,
+                user_id,
+                hitl_safe_mode=original_agent.settings.human_in_the_loop_safe_mode,
+                sensitive_action_safe_mode=original_agent.settings.sensitive_action_safe_mode,
+            )
+        )[0]
     except prisma.errors.PrismaError as e:
         logger.error(f"Database error cloning library agent: {e}")
         raise DatabaseError("Failed to fork library agent") from e

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

@@ -73,6 +73,12 @@ class LibraryAgent(pydantic.BaseModel):
     has_external_trigger: bool = pydantic.Field(
         description="Whether the agent has an external trigger (e.g. webhook) node"
     )
+    has_human_in_the_loop: bool = pydantic.Field(
+        description="Whether the agent has human-in-the-loop blocks"
+    )
+    has_sensitive_action: bool = pydantic.Field(
+        description="Whether the agent has sensitive action blocks"
+    )
     trigger_setup_info: Optional[GraphTriggerInfo] = None
 
     # Indicates whether there's a new output (based on recent runs)
@@ -180,6 +186,8 @@ class LibraryAgent(pydantic.BaseModel):
                 graph.credentials_input_schema if sub_graphs is not None else None
             ),
             has_external_trigger=graph.has_external_trigger,
+            has_human_in_the_loop=graph.has_human_in_the_loop,
+            has_sensitive_action=graph.has_sensitive_action,
             trigger_setup_info=graph.trigger_setup_info,
             new_output=new_output,
             can_access_graph=can_access_graph,

autogpt_platform/backend/backend/api/features/library/routes_test.py

@@ -52,6 +52,8 @@ async def test_get_library_agents_success(
                 output_schema={"type": "object", "properties": {}},
                 credentials_input_schema={"type": "object", "properties": {}},
                 has_external_trigger=False,
+                has_human_in_the_loop=False,
+                has_sensitive_action=False,
                 status=library_model.LibraryAgentStatus.COMPLETED,
                 recommended_schedule_cron=None,
                 new_output=False,
@@ -75,6 +77,8 @@ async def test_get_library_agents_success(
                 output_schema={"type": "object", "properties": {}},
                 credentials_input_schema={"type": "object", "properties": {}},
                 has_external_trigger=False,
+                has_human_in_the_loop=False,
+                has_sensitive_action=False,
                 status=library_model.LibraryAgentStatus.COMPLETED,
                 recommended_schedule_cron=None,
                 new_output=False,
@@ -150,6 +154,8 @@ async def test_get_favorite_library_agents_success(
                 output_schema={"type": "object", "properties": {}},
                 credentials_input_schema={"type": "object", "properties": {}},
                 has_external_trigger=False,
+                has_human_in_the_loop=False,
+                has_sensitive_action=False,
                 status=library_model.LibraryAgentStatus.COMPLETED,
                 recommended_schedule_cron=None,
                 new_output=False,
@@ -218,6 +224,8 @@ def test_add_agent_to_library_success(
         output_schema={"type": "object", "properties": {}},
         credentials_input_schema={"type": "object", "properties": {}},
         has_external_trigger=False,
+        has_human_in_the_loop=False,
+        has_sensitive_action=False,
         status=library_model.LibraryAgentStatus.COMPLETED,
         new_output=False,
         can_access_graph=True,

autogpt_platform/backend/backend/api/features/store/embeddings.py

@@ -154,6 +154,7 @@ async def store_content_embedding(
 
         # Upsert the embedding
         # WHERE clause in DO UPDATE prevents PostgreSQL 15 bug with NULLS NOT DISTINCT
+        # Use unqualified ::vector - pgvector is in search_path on all environments
         await execute_raw_with_schema(
             """
             INSERT INTO {schema_prefix}"UnifiedContentEmbedding" (
@@ -177,7 +178,6 @@ async def store_content_embedding(
             searchable_text,
             metadata_json,
             client=client,
-            set_public_search_path=True,
         )
 
         logger.info(f"Stored embedding for {content_type}:{content_id}")
@@ -236,7 +236,6 @@ async def get_content_embedding(
             content_type,
             content_id,
             user_id,
-            set_public_search_path=True,
         )
 
         if result and len(result) > 0:
@@ -871,31 +870,45 @@ async def semantic_search(
         # Add content type parameters and build placeholders dynamically
         content_type_start_idx = len(params) + 1
         content_type_placeholders = ", ".join(
-            f'${content_type_start_idx + i}::{{{{schema_prefix}}}}"ContentType"'
+            "$" + str(content_type_start_idx + i) + '::{schema_prefix}"ContentType"'
             for i in range(len(content_types))
         )
         params.extend([ct.value for ct in content_types])
 
-        sql = f"""
+        # Build min_similarity param index before appending
+        min_similarity_idx = len(params) + 1
+        params.append(min_similarity)
+
+        # Use unqualified ::vector and <=> operator - pgvector is in search_path on all environments
+        sql = (
+            """
             SELECT
                 "contentId" as content_id,
                 "contentType" as content_type,
                 "searchableText" as searchable_text,
                 metadata,
-                1 - (embedding <=> '{embedding_str}'::vector) as similarity
+                1 - (embedding <=> '"""
-            FROM {{{{schema_prefix}}}}"UnifiedContentEmbedding"
+            + embedding_str
-            WHERE "contentType" IN ({content_type_placeholders})
+            + """'::vector) as similarity
-            {user_filter}
+            FROM {schema_prefix}"UnifiedContentEmbedding"
-            AND 1 - (embedding <=> '{embedding_str}'::vector) >= ${len(params) + 1}
+            WHERE "contentType" IN ("""
+            + content_type_placeholders
+            + """)
+            """
+            + user_filter
+            + """
+            AND 1 - (embedding <=> '"""
+            + embedding_str
+            + """'::vector) >= $"""
+            + str(min_similarity_idx)
+            + """
             ORDER BY similarity DESC
             LIMIT $1
         """
-        params.append(min_similarity)
+        )
 
         try:
-            results = await query_raw_with_schema(
+            results = await query_raw_with_schema(sql, *params)
-                sql, *params, set_public_search_path=True
-            )
             return [
                 {
                     "content_id": row["content_id"],
@@ -922,31 +935,41 @@ async def semantic_search(
     # Add content type parameters and build placeholders dynamically
     content_type_start_idx = len(params_lexical) + 1
     content_type_placeholders_lexical = ", ".join(
-        f'${content_type_start_idx + i}::{{{{schema_prefix}}}}"ContentType"'
+        "$" + str(content_type_start_idx + i) + '::{schema_prefix}"ContentType"'
         for i in range(len(content_types))
     )
     params_lexical.extend([ct.value for ct in content_types])
 
-    sql_lexical = f"""
+    # Build query param index before appending
+    query_param_idx = len(params_lexical) + 1
+    params_lexical.append(f"%{query}%")
+
+    # Use regular string (not f-string) for template to preserve {schema_prefix} placeholders
+    sql_lexical = (
+        """
         SELECT
             "contentId" as content_id,
             "contentType" as content_type,
             "searchableText" as searchable_text,
             metadata,
             0.0 as similarity
-        FROM {{{{schema_prefix}}}}"UnifiedContentEmbedding"
+        FROM {schema_prefix}"UnifiedContentEmbedding"
-        WHERE "contentType" IN ({content_type_placeholders_lexical})
+        WHERE "contentType" IN ("""
-        {user_filter}
+        + content_type_placeholders_lexical
-        AND "searchableText" ILIKE ${len(params_lexical) + 1}
+        + """)
+        """
+        + user_filter
+        + """
+        AND "searchableText" ILIKE $"""
+        + str(query_param_idx)
+        + """
         ORDER BY "updatedAt" DESC
         LIMIT $1
     """
-    params_lexical.append(f"%{query}%")
+    )
 
     try:
-        results = await query_raw_with_schema(
+        results = await query_raw_with_schema(sql_lexical, *params_lexical)
-            sql_lexical, *params_lexical, set_public_search_path=True
-        )
         return [
             {
                 "content_id": row["content_id"],

autogpt_platform/backend/backend/api/features/store/embeddings_test.py

@@ -155,18 +155,14 @@ async def test_store_embedding_success(mocker):
     )
 
     assert result is True
-    # execute_raw is called twice: once for SET search_path, once for INSERT
+    # execute_raw is called once for INSERT (no separate SET search_path needed)
-    assert mock_client.execute_raw.call_count == 2
+    assert mock_client.execute_raw.call_count == 1
 
-    # First call: SET search_path
+    # Verify the INSERT query with the actual data
-    first_call_args = mock_client.execute_raw.call_args_list[0][0]
+    call_args = mock_client.execute_raw.call_args_list[0][0]
-    assert "SET search_path" in first_call_args[0]
+    assert "test-version-id" in call_args
-
+    assert "[0.1,0.2,0.3]" in call_args
-    # Second call: INSERT query with the actual data
+    assert None in call_args  # userId should be None for store agents
-    second_call_args = mock_client.execute_raw.call_args_list[1][0]
-    assert "test-version-id" in second_call_args
-    assert "[0.1,0.2,0.3]" in second_call_args
-    assert None in second_call_args  # userId should be None for store agents
 
 
 @pytest.mark.asyncio(loop_scope="session")

autogpt_platform/backend/backend/api/features/store/hybrid_search.py

@@ -12,7 +12,7 @@ from dataclasses import dataclass
 from typing import Any, Literal
 
 from prisma.enums import ContentType
-from rank_bm25 import BM25Okapi
+from rank_bm25 import BM25Okapi  # type: ignore[import-untyped]
 
 from backend.api.features.store.embeddings import (
     EMBEDDING_DIM,
@@ -363,9 +363,7 @@ async def unified_hybrid_search(
         LIMIT {limit_param} OFFSET {offset_param}
     """
 
-    results = await query_raw_with_schema(
+    results = await query_raw_with_schema(sql_query, *params)
-        sql_query, *params, set_public_search_path=True
-    )
 
     total = results[0]["total_count"] if results else 0
     # Apply BM25 reranking
@@ -688,9 +686,7 @@ async def hybrid_search(
         LIMIT {limit_param} OFFSET {offset_param}
     """
 
-    results = await query_raw_with_schema(
+    results = await query_raw_with_schema(sql_query, *params)
-        sql_query, *params, set_public_search_path=True
-    )
 
     total = results[0]["total_count"] if results else 0
 

autogpt_platform/backend/backend/api/features/v1.py

@@ -761,10 +761,8 @@ async def create_new_graph(
     graph.reassign_ids(user_id=user_id, reassign_graph_id=True)
     graph.validate_graph(for_run=False)
 
-    # The return value of the create graph & library function is intentionally not used here,
-    # as the graph already valid and no sub-graphs are returned back.
     await graph_db.create_graph(graph, user_id=user_id)
-    await library_db.create_library_agent(graph, user_id=user_id)
+    await library_db.create_library_agent(graph, user_id)
     activated_graph = await on_graph_activate(graph, user_id=user_id)
 
     if create_graph.source == "builder":
@@ -888,21 +886,19 @@ async def set_graph_active_version(
 async def _update_library_agent_version_and_settings(
     user_id: str, agent_graph: graph_db.GraphModel
 ) -> library_model.LibraryAgent:
-    # Keep the library agent up to date with the new active version
     library = await library_db.update_agent_version_in_library(
         user_id, agent_graph.id, agent_graph.version
     )
-    # If the graph has HITL node, initialize the setting if it's not already set.
+    updated_settings = GraphSettings.from_graph(
-    if (
+        graph=agent_graph,
-        agent_graph.has_human_in_the_loop
+        hitl_safe_mode=library.settings.human_in_the_loop_safe_mode,
-        and library.settings.human_in_the_loop_safe_mode is None
+        sensitive_action_safe_mode=library.settings.sensitive_action_safe_mode,
-    ):
+    )
-        await library_db.update_library_agent_settings(
+    if updated_settings != library.settings:
+        library = await library_db.update_library_agent(
+            library_agent_id=library.id,
             user_id=user_id,
-            agent_id=library.id,
+            settings=updated_settings,
-            settings=library.settings.model_copy(
-                update={"human_in_the_loop_safe_mode": True}
-            ),
         )
     return library
 
@@ -919,21 +915,18 @@ async def update_graph_settings(
     user_id: Annotated[str, Security(get_user_id)],
 ) -> GraphSettings:
     """Update graph settings for the user's library agent."""
-    # Get the library agent for this graph
     library_agent = await library_db.get_library_agent_by_graph_id(
         graph_id=graph_id, user_id=user_id
     )
     if not library_agent:
         raise HTTPException(404, f"Graph #{graph_id} not found in user's library")
 
-    # Update the library agent settings
+    updated_agent = await library_db.update_library_agent(
-    updated_agent = await library_db.update_library_agent_settings(
+        library_agent_id=library_agent.id,
         user_id=user_id,
-        agent_id=library_agent.id,
         settings=settings,
     )
 
-    # Return the updated settings
     return GraphSettings.model_validate(updated_agent.settings)
 
 

autogpt_platform/backend/backend/blocks/ai_shortform_video_block.py

@@ -174,7 +174,7 @@ class AIShortformVideoCreatorBlock(Block):
         )
         frame_rate: int = SchemaField(description="Frame rate of the video", default=60)
         generation_preset: GenerationPreset = SchemaField(
-            description="Generation preset for visual style - only effects AI generated visuals",
+            description="Generation preset for visual style - only affects AI-generated visuals",
             default=GenerationPreset.LEONARDO,
             placeholder=GenerationPreset.LEONARDO,
         )

autogpt_platform/backend/backend/blocks/apollo/models.py

@@ -381,7 +381,7 @@ Each range you add needs to be a string, with the upper and lower numbers of the
     organization_locations: Optional[list[str]] = SchemaField(
         description="""The location of the company headquarters. You can search across cities, US states, and countries.
 
-If a company has several office locations, results are still based on the headquarters location. For example, if you search chicago but a company's HQ location is in boston, any Boston-based companies will not appearch in your search results, even if they match other parameters.
+If a company has several office locations, results are still based on the headquarters location. For example, if you search chicago but a company's HQ location is in boston, any Boston-based companies will not appear in your search results, even if they match other parameters.
 
 To exclude companies based on location, use the organization_not_locations parameter.
 """,

autogpt_platform/backend/backend/blocks/apollo/organization.py

@@ -34,7 +34,7 @@ Each range you add needs to be a string, with the upper and lower numbers of the
         organization_locations: list[str] = SchemaField(
             description="""The location of the company headquarters. You can search across cities, US states, and countries.
 
-If a company has several office locations, results are still based on the headquarters location. For example, if you search chicago but a company's HQ location is in boston, any Boston-based companies will not appearch in your search results, even if they match other parameters.
+If a company has several office locations, results are still based on the headquarters location. For example, if you search chicago but a company's HQ location is in boston, any Boston-based companies will not appear in your search results, even if they match other parameters.
 
 To exclude companies based on location, use the organization_not_locations parameter.
 """,

autogpt_platform/backend/backend/blocks/basic.py

@@ -81,7 +81,7 @@ class StoreValueBlock(Block):
     def __init__(self):
         super().__init__(
             id="1ff065e9-88e8-4358-9d82-8dc91f622ba9",
-            description="This block forwards an input value as output, allowing reuse without change.",
+            description="A basic block that stores and forwards a value throughout workflows, allowing it to be reused without changes across multiple blocks.",
             categories={BlockCategory.BASIC},
             input_schema=StoreValueBlock.Input,
             output_schema=StoreValueBlock.Output,
@@ -111,7 +111,7 @@ class PrintToConsoleBlock(Block):
     def __init__(self):
         super().__init__(
             id="f3b1c1b2-4c4f-4f0d-8d2f-4c4f0d8d2f4c",
-            description="Print the given text to the console, this is used for a debugging purpose.",
+            description="A debugging block that outputs text to the console for monitoring and troubleshooting workflow execution.",
             categories={BlockCategory.BASIC},
             input_schema=PrintToConsoleBlock.Input,
             output_schema=PrintToConsoleBlock.Output,
@@ -137,7 +137,7 @@ class NoteBlock(Block):
     def __init__(self):
         super().__init__(
             id="cc10ff7b-7753-4ff2-9af6-9399b1a7eddc",
-            description="This block is used to display a sticky note with the given text.",
+            description="A visual annotation block that displays a sticky note in the workflow editor for documentation and organization purposes.",
             categories={BlockCategory.BASIC},
             input_schema=NoteBlock.Input,
             output_schema=NoteBlock.Output,

autogpt_platform/backend/backend/blocks/claude_code.py

@@ -0,0 +1,659 @@
+import json
+import shlex
+import uuid
+from typing import Literal, Optional
+
+from e2b import AsyncSandbox as BaseAsyncSandbox
+from pydantic import BaseModel, SecretStr
+
+from backend.data.block import (
+    Block,
+    BlockCategory,
+    BlockOutput,
+    BlockSchemaInput,
+    BlockSchemaOutput,
+)
+from backend.data.model import (
+    APIKeyCredentials,
+    CredentialsField,
+    CredentialsMetaInput,
+    SchemaField,
+)
+from backend.integrations.providers import ProviderName
+
+
+class ClaudeCodeExecutionError(Exception):
+    """Exception raised when Claude Code execution fails.
+
+    Carries the sandbox_id so it can be returned to the user for cleanup
+    when dispose_sandbox=False.
+    """
+
+    def __init__(self, message: str, sandbox_id: str = ""):
+        super().__init__(message)
+        self.sandbox_id = sandbox_id
+
+
+# Test credentials for E2B
+TEST_E2B_CREDENTIALS = APIKeyCredentials(
+    id="01234567-89ab-cdef-0123-456789abcdef",
+    provider="e2b",
+    api_key=SecretStr("mock-e2b-api-key"),
+    title="Mock E2B API key",
+    expires_at=None,
+)
+TEST_E2B_CREDENTIALS_INPUT = {
+    "provider": TEST_E2B_CREDENTIALS.provider,
+    "id": TEST_E2B_CREDENTIALS.id,
+    "type": TEST_E2B_CREDENTIALS.type,
+    "title": TEST_E2B_CREDENTIALS.title,
+}
+
+# Test credentials for Anthropic
+TEST_ANTHROPIC_CREDENTIALS = APIKeyCredentials(
+    id="2e568a2b-b2ea-475a-8564-9a676bf31c56",
+    provider="anthropic",
+    api_key=SecretStr("mock-anthropic-api-key"),
+    title="Mock Anthropic API key",
+    expires_at=None,
+)
+TEST_ANTHROPIC_CREDENTIALS_INPUT = {
+    "provider": TEST_ANTHROPIC_CREDENTIALS.provider,
+    "id": TEST_ANTHROPIC_CREDENTIALS.id,
+    "type": TEST_ANTHROPIC_CREDENTIALS.type,
+    "title": TEST_ANTHROPIC_CREDENTIALS.title,
+}
+
+
+class ClaudeCodeBlock(Block):
+    """
+    Execute tasks using Claude Code (Anthropic's AI coding assistant) in an E2B sandbox.
+
+    Claude Code can create files, install tools, run commands, and perform complex
+    coding tasks autonomously within a secure sandbox environment.
+    """
+
+    # Use base template - we'll install Claude Code ourselves for latest version
+    DEFAULT_TEMPLATE = "base"
+
+    class Input(BlockSchemaInput):
+        e2b_credentials: CredentialsMetaInput[
+            Literal[ProviderName.E2B], Literal["api_key"]
+        ] = CredentialsField(
+            description=(
+                "API key for the E2B platform to create the sandbox. "
+                "Get one on the [e2b website](https://e2b.dev/docs)"
+            ),
+        )
+
+        anthropic_credentials: CredentialsMetaInput[
+            Literal[ProviderName.ANTHROPIC], Literal["api_key"]
+        ] = CredentialsField(
+            description=(
+                "API key for Anthropic to power Claude Code. "
+                "Get one at [Anthropic's website](https://console.anthropic.com)"
+            ),
+        )
+
+        prompt: str = SchemaField(
+            description=(
+                "The task or instruction for Claude Code to execute. "
+                "Claude Code can create files, install packages, run commands, "
+                "and perform complex coding tasks."
+            ),
+            placeholder="Create a hello world index.html file",
+            default="",
+            advanced=False,
+        )
+
+        timeout: int = SchemaField(
+            description=(
+                "Sandbox timeout in seconds. Claude Code tasks can take "
+                "a while, so set this appropriately for your task complexity. "
+                "Note: This only applies when creating a new sandbox. "
+                "When reconnecting to an existing sandbox via sandbox_id, "
+                "the original timeout is retained."
+            ),
+            default=300,  # 5 minutes default
+            advanced=True,
+        )
+
+        setup_commands: list[str] = SchemaField(
+            description=(
+                "Optional shell commands to run before executing Claude Code. "
+                "Useful for installing dependencies or setting up the environment."
+            ),
+            default_factory=list,
+            advanced=True,
+        )
+
+        working_directory: str = SchemaField(
+            description="Working directory for Claude Code to operate in.",
+            default="/home/user",
+            advanced=True,
+        )
+
+        # Session/continuation support
+        session_id: str = SchemaField(
+            description=(
+                "Session ID to resume a previous conversation. "
+                "Leave empty for a new conversation. "
+                "Use the session_id from a previous run to continue that conversation."
+            ),
+            default="",
+            advanced=True,
+        )
+
+        sandbox_id: str = SchemaField(
+            description=(
+                "Sandbox ID to reconnect to an existing sandbox. "
+                "Required when resuming a session (along with session_id). "
+                "Use the sandbox_id from a previous run where dispose_sandbox was False."
+            ),
+            default="",
+            advanced=True,
+        )
+
+        conversation_history: str = SchemaField(
+            description=(
+                "Previous conversation history to continue from. "
+                "Use this to restore context on a fresh sandbox if the previous one timed out. "
+                "Pass the conversation_history output from a previous run."
+            ),
+            default="",
+            advanced=True,
+        )
+
+        dispose_sandbox: bool = SchemaField(
+            description=(
+                "Whether to dispose of the sandbox immediately after execution. "
+                "Set to False if you want to continue the conversation later "
+                "(you'll need both sandbox_id and session_id from the output)."
+            ),
+            default=True,
+            advanced=True,
+        )
+
+    class FileOutput(BaseModel):
+        """A file extracted from the sandbox."""
+
+        path: str
+        relative_path: str  # Path relative to working directory (for GitHub, etc.)
+        name: str
+        content: str
+
+    class Output(BlockSchemaOutput):
+        response: str = SchemaField(
+            description="The output/response from Claude Code execution"
+        )
+        files: list["ClaudeCodeBlock.FileOutput"] = SchemaField(
+            description=(
+                "List of text files created/modified by Claude Code during this execution. "
+                "Each file has 'path', 'relative_path', 'name', and 'content' fields."
+            )
+        )
+        conversation_history: str = SchemaField(
+            description=(
+                "Full conversation history including this turn. "
+                "Pass this to conversation_history input to continue on a fresh sandbox "
+                "if the previous sandbox timed out."
+            )
+        )
+        session_id: str = SchemaField(
+            description=(
+                "Session ID for this conversation. "
+                "Pass this back along with sandbox_id to continue the conversation."
+            )
+        )
+        sandbox_id: Optional[str] = SchemaField(
+            description=(
+                "ID of the sandbox instance. "
+                "Pass this back along with session_id to continue the conversation. "
+                "This is None if dispose_sandbox was True (sandbox was disposed)."
+            ),
+            default=None,
+        )
+        error: str = SchemaField(description="Error message if execution failed")
+
+    def __init__(self):
+        super().__init__(
+            id="4e34f4a5-9b89-4326-ba77-2dd6750b7194",
+            description=(
+                "Execute tasks using Claude Code in an E2B sandbox. "
+                "Claude Code can create files, install tools, run commands, "
+                "and perform complex coding tasks autonomously."
+            ),
+            categories={BlockCategory.DEVELOPER_TOOLS, BlockCategory.AI},
+            input_schema=ClaudeCodeBlock.Input,
+            output_schema=ClaudeCodeBlock.Output,
+            test_credentials={
+                "e2b_credentials": TEST_E2B_CREDENTIALS,
+                "anthropic_credentials": TEST_ANTHROPIC_CREDENTIALS,
+            },
+            test_input={
+                "e2b_credentials": TEST_E2B_CREDENTIALS_INPUT,
+                "anthropic_credentials": TEST_ANTHROPIC_CREDENTIALS_INPUT,
+                "prompt": "Create a hello world HTML file",
+                "timeout": 300,
+                "setup_commands": [],
+                "working_directory": "/home/user",
+                "session_id": "",
+                "sandbox_id": "",
+                "conversation_history": "",
+                "dispose_sandbox": True,
+            },
+            test_output=[
+                ("response", "Created index.html with hello world content"),
+                (
+                    "files",
+                    [
+                        {
+                            "path": "/home/user/index.html",
+                            "relative_path": "index.html",
+                            "name": "index.html",
+                            "content": "<html>Hello World</html>",
+                        }
+                    ],
+                ),
+                (
+                    "conversation_history",
+                    "User: Create a hello world HTML file\n"
+                    "Claude: Created index.html with hello world content",
+                ),
+                ("session_id", str),
+                ("sandbox_id", None),  # None because dispose_sandbox=True in test_input
+            ],
+            test_mock={
+                "execute_claude_code": lambda *args, **kwargs: (
+                    "Created index.html with hello world content",  # response
+                    [
+                        ClaudeCodeBlock.FileOutput(
+                            path="/home/user/index.html",
+                            relative_path="index.html",
+                            name="index.html",
+                            content="<html>Hello World</html>",
+                        )
+                    ],  # files
+                    "User: Create a hello world HTML file\n"
+                    "Claude: Created index.html with hello world content",  # conversation_history
+                    "test-session-id",  # session_id
+                    "sandbox_id",  # sandbox_id
+                ),
+            },
+        )
+
+    async def execute_claude_code(
+        self,
+        e2b_api_key: str,
+        anthropic_api_key: str,
+        prompt: str,
+        timeout: int,
+        setup_commands: list[str],
+        working_directory: str,
+        session_id: str,
+        existing_sandbox_id: str,
+        conversation_history: str,
+        dispose_sandbox: bool,
+    ) -> tuple[str, list["ClaudeCodeBlock.FileOutput"], str, str, str]:
+        """
+        Execute Claude Code in an E2B sandbox.
+
+        Returns:
+            Tuple of (response, files, conversation_history, session_id, sandbox_id)
+        """
+
+        # Validate that sandbox_id is provided when resuming a session
+        if session_id and not existing_sandbox_id:
+            raise ValueError(
+                "sandbox_id is required when resuming a session with session_id. "
+                "The session state is stored in the original sandbox. "
+                "If the sandbox has timed out, use conversation_history instead "
+                "to restore context on a fresh sandbox."
+            )
+
+        sandbox = None
+        sandbox_id = ""
+
+        try:
+            # Either reconnect to existing sandbox or create a new one
+            if existing_sandbox_id:
+                # Reconnect to existing sandbox for conversation continuation
+                sandbox = await BaseAsyncSandbox.connect(
+                    sandbox_id=existing_sandbox_id,
+                    api_key=e2b_api_key,
+                )
+            else:
+                # Create new sandbox
+                sandbox = await BaseAsyncSandbox.create(
+                    template=self.DEFAULT_TEMPLATE,
+                    api_key=e2b_api_key,
+                    timeout=timeout,
+                    envs={"ANTHROPIC_API_KEY": anthropic_api_key},
+                )
+
+                # Install Claude Code from npm (ensures we get the latest version)
+                install_result = await sandbox.commands.run(
+                    "npm install -g @anthropic-ai/claude-code@latest",
+                    timeout=120,  # 2 min timeout for install
+                )
+                if install_result.exit_code != 0:
+                    raise Exception(
+                        f"Failed to install Claude Code: {install_result.stderr}"
+                    )
+
+                # Run any user-provided setup commands
+                for cmd in setup_commands:
+                    setup_result = await sandbox.commands.run(cmd)
+                    if setup_result.exit_code != 0:
+                        raise Exception(
+                            f"Setup command failed: {cmd}\n"
+                            f"Exit code: {setup_result.exit_code}\n"
+                            f"Stdout: {setup_result.stdout}\n"
+                            f"Stderr: {setup_result.stderr}"
+                        )
+
+            # Capture sandbox_id immediately after creation/connection
+            # so it's available for error recovery if dispose_sandbox=False
+            sandbox_id = sandbox.sandbox_id
+
+            # Generate or use provided session ID
+            current_session_id = session_id if session_id else str(uuid.uuid4())
+
+            # Build base Claude flags
+            base_flags = "-p --dangerously-skip-permissions --output-format json"
+
+            # Add conversation history context if provided (for fresh sandbox continuation)
+            history_flag = ""
+            if conversation_history and not session_id:
+                # Inject previous conversation as context via system prompt
+                # Use consistent escaping via _escape_prompt helper
+                escaped_history = self._escape_prompt(
+                    f"Previous conversation context: {conversation_history}"
+                )
+                history_flag = f" --append-system-prompt {escaped_history}"
+
+            # Build Claude command based on whether we're resuming or starting new
+            # Use shlex.quote for working_directory and session IDs to prevent injection
+            safe_working_dir = shlex.quote(working_directory)
+            if session_id:
+                # Resuming existing session (sandbox still alive)
+                safe_session_id = shlex.quote(session_id)
+                claude_command = (
+                    f"cd {safe_working_dir} && "
+                    f"echo {self._escape_prompt(prompt)} | "
+                    f"claude --resume {safe_session_id} {base_flags}"
+                )
+            else:
+                # New session with specific ID
+                safe_current_session_id = shlex.quote(current_session_id)
+                claude_command = (
+                    f"cd {safe_working_dir} && "
+                    f"echo {self._escape_prompt(prompt)} | "
+                    f"claude --session-id {safe_current_session_id} {base_flags}{history_flag}"
+                )
+
+            # Capture timestamp before running Claude Code to filter files later
+            # Capture timestamp 1 second in the past to avoid race condition with file creation
+            timestamp_result = await sandbox.commands.run(
+                "date -u -d '1 second ago' +%Y-%m-%dT%H:%M:%S"
+            )
+            if timestamp_result.exit_code != 0:
+                raise RuntimeError(
+                    f"Failed to capture timestamp: {timestamp_result.stderr}"
+                )
+            start_timestamp = (
+                timestamp_result.stdout.strip() if timestamp_result.stdout else None
+            )
+
+            result = await sandbox.commands.run(
+                claude_command,
+                timeout=0,  # No command timeout - let sandbox timeout handle it
+            )
+
+            # Check for command failure
+            if result.exit_code != 0:
+                error_msg = result.stderr or result.stdout or "Unknown error"
+                raise Exception(
+                    f"Claude Code command failed with exit code {result.exit_code}:\n"
+                    f"{error_msg}"
+                )
+
+            raw_output = result.stdout or ""
+
+            # Parse JSON output to extract response and build conversation history
+            response = ""
+            new_conversation_history = conversation_history or ""
+
+            try:
+                # The JSON output contains the result
+                output_data = json.loads(raw_output)
+                response = output_data.get("result", raw_output)
+
+                # Build conversation history entry
+                turn_entry = f"User: {prompt}\nClaude: {response}"
+                if new_conversation_history:
+                    new_conversation_history = (
+                        f"{new_conversation_history}\n\n{turn_entry}"
+                    )
+                else:
+                    new_conversation_history = turn_entry
+
+            except json.JSONDecodeError:
+                # If not valid JSON, use raw output
+                response = raw_output
+                turn_entry = f"User: {prompt}\nClaude: {response}"
+                if new_conversation_history:
+                    new_conversation_history = (
+                        f"{new_conversation_history}\n\n{turn_entry}"
+                    )
+                else:
+                    new_conversation_history = turn_entry
+
+            # Extract files created/modified during this run
+            files = await self._extract_files(
+                sandbox, working_directory, start_timestamp
+            )
+
+            return (
+                response,
+                files,
+                new_conversation_history,
+                current_session_id,
+                sandbox_id,
+            )
+
+        except Exception as e:
+            # Wrap exception with sandbox_id so caller can access/cleanup
+            # the preserved sandbox when dispose_sandbox=False
+            raise ClaudeCodeExecutionError(str(e), sandbox_id) from e
+
+        finally:
+            if dispose_sandbox and sandbox:
+                await sandbox.kill()
+
+    async def _extract_files(
+        self,
+        sandbox: BaseAsyncSandbox,
+        working_directory: str,
+        since_timestamp: str | None = None,
+    ) -> list["ClaudeCodeBlock.FileOutput"]:
+        """
+        Extract text files created/modified during this Claude Code execution.
+
+        Args:
+            sandbox: The E2B sandbox instance
+            working_directory: Directory to search for files
+            since_timestamp: ISO timestamp - only return files modified after this time
+
+        Returns:
+            List of FileOutput objects with path, relative_path, name, and content
+        """
+        files: list[ClaudeCodeBlock.FileOutput] = []
+
+        # Text file extensions we can safely read as text
+        text_extensions = {
+            ".txt",
+            ".md",
+            ".html",
+            ".htm",
+            ".css",
+            ".js",
+            ".ts",
+            ".jsx",
+            ".tsx",
+            ".json",
+            ".xml",
+            ".yaml",
+            ".yml",
+            ".toml",
+            ".ini",
+            ".cfg",
+            ".conf",
+            ".py",
+            ".rb",
+            ".php",
+            ".java",
+            ".c",
+            ".cpp",
+            ".h",
+            ".hpp",
+            ".cs",
+            ".go",
+            ".rs",
+            ".swift",
+            ".kt",
+            ".scala",
+            ".sh",
+            ".bash",
+            ".zsh",
+            ".sql",
+            ".graphql",
+            ".env",
+            ".gitignore",
+            ".dockerfile",
+            "Dockerfile",
+            ".vue",
+            ".svelte",
+            ".astro",
+            ".mdx",
+            ".rst",
+            ".tex",
+            ".csv",
+            ".log",
+        }
+
+        try:
+            # List files recursively using find command
+            # Exclude node_modules and .git directories, but allow hidden files
+            # like .env and .gitignore (they're filtered by text_extensions later)
+            # Filter by timestamp to only get files created/modified during this run
+            safe_working_dir = shlex.quote(working_directory)
+            timestamp_filter = ""
+            if since_timestamp:
+                timestamp_filter = f"-newermt {shlex.quote(since_timestamp)} "
+            find_result = await sandbox.commands.run(
+                f"find {safe_working_dir} -type f "
+                f"{timestamp_filter}"
+                f"-not -path '*/node_modules/*' "
+                f"-not -path '*/.git/*' "
+                f"2>/dev/null"
+            )
+
+            if find_result.stdout:
+                for file_path in find_result.stdout.strip().split("\n"):
+                    if not file_path:
+                        continue
+
+                    # Check if it's a text file we can read
+                    is_text = any(
+                        file_path.endswith(ext) for ext in text_extensions
+                    ) or file_path.endswith("Dockerfile")
+
+                    if is_text:
+                        try:
+                            content = await sandbox.files.read(file_path)
+                            # Handle bytes or string
+                            if isinstance(content, bytes):
+                                content = content.decode("utf-8", errors="replace")
+
+                            # Extract filename from path
+                            file_name = file_path.split("/")[-1]
+
+                            # Calculate relative path by stripping working directory
+                            relative_path = file_path
+                            if file_path.startswith(working_directory):
+                                relative_path = file_path[len(working_directory) :]
+                                # Remove leading slash if present
+                                if relative_path.startswith("/"):
+                                    relative_path = relative_path[1:]
+
+                            files.append(
+                                ClaudeCodeBlock.FileOutput(
+                                    path=file_path,
+                                    relative_path=relative_path,
+                                    name=file_name,
+                                    content=content,
+                                )
+                            )
+                        except Exception:
+                            # Skip files that can't be read
+                            pass
+
+        except Exception:
+            # If file extraction fails, return empty results
+            pass
+
+        return files
+
+    def _escape_prompt(self, prompt: str) -> str:
+        """Escape the prompt for safe shell execution."""
+        # Use single quotes and escape any single quotes in the prompt
+        escaped = prompt.replace("'", "'\"'\"'")
+        return f"'{escaped}'"
+
+    async def run(
+        self,
+        input_data: Input,
+        *,
+        e2b_credentials: APIKeyCredentials,
+        anthropic_credentials: APIKeyCredentials,
+        **kwargs,
+    ) -> BlockOutput:
+        try:
+            (
+                response,
+                files,
+                conversation_history,
+                session_id,
+                sandbox_id,
+            ) = await self.execute_claude_code(
+                e2b_api_key=e2b_credentials.api_key.get_secret_value(),
+                anthropic_api_key=anthropic_credentials.api_key.get_secret_value(),
+                prompt=input_data.prompt,
+                timeout=input_data.timeout,
+                setup_commands=input_data.setup_commands,
+                working_directory=input_data.working_directory,
+                session_id=input_data.session_id,
+                existing_sandbox_id=input_data.sandbox_id,
+                conversation_history=input_data.conversation_history,
+                dispose_sandbox=input_data.dispose_sandbox,
+            )
+
+            yield "response", response
+            # Always yield files (empty list if none) to match Output schema
+            yield "files", [f.model_dump() for f in files]
+            # Always yield conversation_history so user can restore context on fresh sandbox
+            yield "conversation_history", conversation_history
+            # Always yield session_id so user can continue conversation
+            yield "session_id", session_id
+            # Always yield sandbox_id (None if disposed) to match Output schema
+            yield "sandbox_id", sandbox_id if not input_data.dispose_sandbox else None
+
+        except ClaudeCodeExecutionError as e:
+            yield "error", str(e)
+            # If sandbox was preserved (dispose_sandbox=False), yield sandbox_id
+            # so user can reconnect to or clean up the orphaned sandbox
+            if not input_data.dispose_sandbox and e.sandbox_id:
+                yield "sandbox_id", e.sandbox_id
+        except Exception as e:
+            yield "error", str(e)

autogpt_platform/backend/backend/blocks/data_manipulation.py

@@ -159,7 +159,7 @@ class FindInDictionaryBlock(Block):
     def __init__(self):
         super().__init__(
             id="0e50422c-6dee-4145-83d6-3a5a392f65de",
-            description="Lookup the given key in the input dictionary/object/list and return the value.",
+            description="A block that looks up a value in a dictionary, list, or object by key or index and returns the corresponding value.",
             input_schema=FindInDictionaryBlock.Input,
             output_schema=FindInDictionaryBlock.Output,
             test_input=[
@@ -680,3 +680,58 @@ class ListIsEmptyBlock(Block):
 
     async def run(self, input_data: Input, **kwargs) -> BlockOutput:
         yield "is_empty", len(input_data.list) == 0
+
+
+class ConcatenateListsBlock(Block):
+    class Input(BlockSchemaInput):
+        lists: List[List[Any]] = SchemaField(
+            description="A list of lists to concatenate together. All lists will be combined in order into a single list.",
+            placeholder="e.g., [[1, 2], [3, 4], [5, 6]]",
+        )
+
+    class Output(BlockSchemaOutput):
+        concatenated_list: List[Any] = SchemaField(
+            description="The concatenated list containing all elements from all input lists in order."
+        )
+        error: str = SchemaField(
+            description="Error message if concatenation failed due to invalid input types."
+        )
+
+    def __init__(self):
+        super().__init__(
+            id="3cf9298b-5817-4141-9d80-7c2cc5199c8e",
+            description="Concatenates multiple lists into a single list. All elements from all input lists are combined in order.",
+            categories={BlockCategory.BASIC},
+            input_schema=ConcatenateListsBlock.Input,
+            output_schema=ConcatenateListsBlock.Output,
+            test_input=[
+                {"lists": [[1, 2, 3], [4, 5, 6]]},
+                {"lists": [["a", "b"], ["c"], ["d", "e", "f"]]},
+                {"lists": [[1, 2], []]},
+                {"lists": []},
+            ],
+            test_output=[
+                ("concatenated_list", [1, 2, 3, 4, 5, 6]),
+                ("concatenated_list", ["a", "b", "c", "d", "e", "f"]),
+                ("concatenated_list", [1, 2]),
+                ("concatenated_list", []),
+            ],
+        )
+
+    async def run(self, input_data: Input, **kwargs) -> BlockOutput:
+        concatenated = []
+        for idx, lst in enumerate(input_data.lists):
+            if lst is None:
+                # Skip None values to avoid errors
+                continue
+            if not isinstance(lst, list):
+                # Type validation: each item must be a list
+                # Strings are iterable and would cause extend() to iterate character-by-character
+                # Non-iterable types would raise TypeError
+                yield "error", (
+                    f"Invalid input at index {idx}: expected a list, got {type(lst).__name__}. "
+                    f"All items in 'lists' must be lists (e.g., [[1, 2], [3, 4]])."
+                )
+                return
+            concatenated.extend(lst)
+        yield "concatenated_list", concatenated

autogpt_platform/backend/backend/blocks/github/issues.py

@@ -51,7 +51,7 @@ class GithubCommentBlock(Block):
     def __init__(self):
         super().__init__(
             id="a8db4d8d-db1c-4a25-a1b0-416a8c33602b",
-            description="This block posts a comment on a specified GitHub issue or pull request.",
+            description="A block that posts comments on GitHub issues or pull requests using the GitHub API.",
             categories={BlockCategory.DEVELOPER_TOOLS},
             input_schema=GithubCommentBlock.Input,
             output_schema=GithubCommentBlock.Output,
@@ -151,7 +151,7 @@ class GithubUpdateCommentBlock(Block):
     def __init__(self):
         super().__init__(
             id="b3f4d747-10e3-4e69-8c51-f2be1d99c9a7",
-            description="This block updates a comment on a specified GitHub issue or pull request.",
+            description="A block that updates an existing comment on a GitHub issue or pull request.",
             categories={BlockCategory.DEVELOPER_TOOLS},
             input_schema=GithubUpdateCommentBlock.Input,
             output_schema=GithubUpdateCommentBlock.Output,
@@ -249,7 +249,7 @@ class GithubListCommentsBlock(Block):
     def __init__(self):
         super().__init__(
             id="c4b5fb63-0005-4a11-b35a-0c2467bd6b59",
-            description="This block lists all comments for a specified GitHub issue or pull request.",
+            description="A block that retrieves all comments from a GitHub issue or pull request, including comment metadata and content.",
             categories={BlockCategory.DEVELOPER_TOOLS},
             input_schema=GithubListCommentsBlock.Input,
             output_schema=GithubListCommentsBlock.Output,
@@ -363,7 +363,7 @@ class GithubMakeIssueBlock(Block):
     def __init__(self):
         super().__init__(
             id="691dad47-f494-44c3-a1e8-05b7990f2dab",
-            description="This block creates a new issue on a specified GitHub repository.",
+            description="A block that creates new issues on GitHub repositories with a title and body content.",
             categories={BlockCategory.DEVELOPER_TOOLS},
             input_schema=GithubMakeIssueBlock.Input,
             output_schema=GithubMakeIssueBlock.Output,
@@ -433,7 +433,7 @@ class GithubReadIssueBlock(Block):
     def __init__(self):
         super().__init__(
             id="6443c75d-032a-4772-9c08-230c707c8acc",
-            description="This block reads the body, title, and user of a specified GitHub issue.",
+            description="A block that retrieves information about a specific GitHub issue, including its title, body content, and creator.",
             categories={BlockCategory.DEVELOPER_TOOLS},
             input_schema=GithubReadIssueBlock.Input,
             output_schema=GithubReadIssueBlock.Output,
@@ -510,7 +510,7 @@ class GithubListIssuesBlock(Block):
     def __init__(self):
         super().__init__(
             id="c215bfd7-0e57-4573-8f8c-f7d4963dcd74",
-            description="This block lists all issues for a specified GitHub repository.",
+            description="A block that retrieves a list of issues from a GitHub repository with their titles and URLs.",
             categories={BlockCategory.DEVELOPER_TOOLS},
             input_schema=GithubListIssuesBlock.Input,
             output_schema=GithubListIssuesBlock.Output,
@@ -597,7 +597,7 @@ class GithubAddLabelBlock(Block):
     def __init__(self):
         super().__init__(
             id="98bd6b77-9506-43d5-b669-6b9733c4b1f1",
-            description="This block adds a label to a specified GitHub issue or pull request.",
+            description="A block that adds a label to a GitHub issue or pull request for categorization and organization.",
             categories={BlockCategory.DEVELOPER_TOOLS},
             input_schema=GithubAddLabelBlock.Input,
             output_schema=GithubAddLabelBlock.Output,
@@ -657,7 +657,7 @@ class GithubRemoveLabelBlock(Block):
     def __init__(self):
         super().__init__(
             id="78f050c5-3e3a-48c0-9e5b-ef1ceca5589c",
-            description="This block removes a label from a specified GitHub issue or pull request.",
+            description="A block that removes a label from a GitHub issue or pull request.",
             categories={BlockCategory.DEVELOPER_TOOLS},
             input_schema=GithubRemoveLabelBlock.Input,
             output_schema=GithubRemoveLabelBlock.Output,
@@ -720,7 +720,7 @@ class GithubAssignIssueBlock(Block):
     def __init__(self):
         super().__init__(
             id="90507c72-b0ff-413a-886a-23bbbd66f542",
-            description="This block assigns a user to a specified GitHub issue.",
+            description="A block that assigns a GitHub user to an issue for task ownership and tracking.",
             categories={BlockCategory.DEVELOPER_TOOLS},
             input_schema=GithubAssignIssueBlock.Input,
             output_schema=GithubAssignIssueBlock.Output,
@@ -786,7 +786,7 @@ class GithubUnassignIssueBlock(Block):
     def __init__(self):
         super().__init__(
             id="d154002a-38f4-46c2-962d-2488f2b05ece",
-            description="This block unassigns a user from a specified GitHub issue.",
+            description="A block that removes a user's assignment from a GitHub issue.",
             categories={BlockCategory.DEVELOPER_TOOLS},
             input_schema=GithubUnassignIssueBlock.Input,
             output_schema=GithubUnassignIssueBlock.Output,

autogpt_platform/backend/backend/blocks/google/gmail.py

@@ -353,7 +353,7 @@ class GmailReadBlock(GmailBase):
     def __init__(self):
         super().__init__(
             id="25310c70-b89b-43ba-b25c-4dfa7e2a481c",
-            description="This block reads emails from Gmail.",
+            description="A block that retrieves and reads emails from a Gmail account based on search criteria, returning detailed message information including subject, sender, body, and attachments.",
             categories={BlockCategory.COMMUNICATION},
             disabled=not GOOGLE_OAUTH_IS_CONFIGURED,
             input_schema=GmailReadBlock.Input,
@@ -743,7 +743,7 @@ class GmailListLabelsBlock(GmailBase):
     def __init__(self):
         super().__init__(
             id="3e1c2c1c-c689-4520-b956-1f3bf4e02bb7",
-            description="This block lists all labels in Gmail.",
+            description="A block that retrieves all labels (categories) from a Gmail account for organizing and categorizing emails.",
             categories={BlockCategory.COMMUNICATION},
             input_schema=GmailListLabelsBlock.Input,
             output_schema=GmailListLabelsBlock.Output,
@@ -807,7 +807,7 @@ class GmailAddLabelBlock(GmailBase):
     def __init__(self):
         super().__init__(
             id="f884b2fb-04f4-4265-9658-14f433926ac9",
-            description="This block adds a label to a Gmail message.",
+            description="A block that adds a label to a specific email message in Gmail, creating the label if it doesn't exist.",
             categories={BlockCategory.COMMUNICATION},
             input_schema=GmailAddLabelBlock.Input,
             output_schema=GmailAddLabelBlock.Output,
@@ -893,7 +893,7 @@ class GmailRemoveLabelBlock(GmailBase):
     def __init__(self):
         super().__init__(
             id="0afc0526-aba1-4b2b-888e-a22b7c3f359d",
-            description="This block removes a label from a Gmail message.",
+            description="A block that removes a label from a specific email message in a Gmail account.",
             categories={BlockCategory.COMMUNICATION},
             input_schema=GmailRemoveLabelBlock.Input,
             output_schema=GmailRemoveLabelBlock.Output,
@@ -961,7 +961,7 @@ class GmailGetThreadBlock(GmailBase):
     def __init__(self):
         super().__init__(
             id="21a79166-9df7-4b5f-9f36-96f639d86112",
-            description="Get a full Gmail thread by ID",
+            description="A block that retrieves an entire Gmail thread (email conversation) by ID, returning all messages with decoded bodies for reading complete conversations.",
             categories={BlockCategory.COMMUNICATION},
             input_schema=GmailGetThreadBlock.Input,
             output_schema=GmailGetThreadBlock.Output,

autogpt_platform/backend/backend/blocks/google/sheets.py

@@ -282,7 +282,7 @@ class GoogleSheetsReadBlock(Block):
     def __init__(self):
         super().__init__(
             id="5724e902-3635-47e9-a108-aaa0263a4988",
-            description="This block reads data from a Google Sheets spreadsheet.",
+            description="A block that reads data from a Google Sheets spreadsheet using A1 notation range selection.",
             categories={BlockCategory.DATA},
             input_schema=GoogleSheetsReadBlock.Input,
             output_schema=GoogleSheetsReadBlock.Output,
@@ -409,7 +409,7 @@ class GoogleSheetsWriteBlock(Block):
     def __init__(self):
         super().__init__(
             id="d9291e87-301d-47a8-91fe-907fb55460e5",
-            description="This block writes data to a Google Sheets spreadsheet.",
+            description="A block that writes data to a Google Sheets spreadsheet at a specified A1 notation range.",
             categories={BlockCategory.DATA},
             input_schema=GoogleSheetsWriteBlock.Input,
             output_schema=GoogleSheetsWriteBlock.Output,

autogpt_platform/backend/backend/blocks/helpers/review.py

@@ -84,7 +84,7 @@ class HITLReviewHelper:
             Exception: If review creation or status update fails
         """
         # Skip review if safe mode is disabled - return auto-approved result
-        if not execution_context.safe_mode:
+        if not execution_context.human_in_the_loop_safe_mode:
             logger.info(
                 f"Block {block_name} skipping review for node {node_exec_id} - safe mode disabled"
             )

autogpt_platform/backend/backend/blocks/human_in_the_loop.py

@@ -104,7 +104,7 @@ class HumanInTheLoopBlock(Block):
         execution_context: ExecutionContext,
         **_kwargs,
     ) -> BlockOutput:
-        if not execution_context.safe_mode:
+        if not execution_context.human_in_the_loop_safe_mode:
             logger.info(
                 f"HITL block skipping review for node {node_exec_id} - safe mode disabled"
             )

autogpt_platform/backend/backend/blocks/io.py

@@ -76,7 +76,7 @@ class AgentInputBlock(Block):
         super().__init__(
             **{
                 "id": "c0a8e994-ebf1-4a9c-a4d8-89d09c86741b",
-                "description": "Base block for user inputs.",
+                "description": "A block that accepts and processes user input values within a workflow, supporting various input types and validation.",
                 "input_schema": AgentInputBlock.Input,
                 "output_schema": AgentInputBlock.Output,
                 "test_input": [
@@ -168,7 +168,7 @@ class AgentOutputBlock(Block):
     def __init__(self):
         super().__init__(
             id="363ae599-353e-4804-937e-b2ee3cef3da4",
-            description="Stores the output of the graph for users to see.",
+            description="A block that records and formats workflow results for display to users, with optional Jinja2 template formatting support.",
             input_schema=AgentOutputBlock.Input,
             output_schema=AgentOutputBlock.Output,
             test_input=[

autogpt_platform/backend/backend/blocks/llm.py

@@ -79,6 +79,10 @@ class ModelMetadata(NamedTuple):
     provider: str
     context_window: int
     max_output_tokens: int | None
+    display_name: str
+    provider_name: str
+    creator_name: str
+    price_tier: Literal[1, 2, 3]
 
 
 class LlmModelMeta(EnumMeta):
@@ -171,6 +175,26 @@ class LlmModel(str, Enum, metaclass=LlmModelMeta):
     V0_1_5_LG = "v0-1.5-lg"
     V0_1_0_MD = "v0-1.0-md"
 
+    @classmethod
+    def __get_pydantic_json_schema__(cls, schema, handler):
+        json_schema = handler(schema)
+        llm_model_metadata = {}
+        for model in cls:
+            model_name = model.value
+            metadata = model.metadata
+            llm_model_metadata[model_name] = {
+                "creator": metadata.creator_name,
+                "creator_name": metadata.creator_name,
+                "title": metadata.display_name,
+                "provider": metadata.provider,
+                "provider_name": metadata.provider_name,
+                "name": model_name,
+                "price_tier": metadata.price_tier,
+            }
+        json_schema["llm_model"] = True
+        json_schema["llm_model_metadata"] = llm_model_metadata
+        return json_schema
+
     @property
     def metadata(self) -> ModelMetadata:
         return MODEL_METADATA[self]
@@ -190,119 +214,291 @@ class LlmModel(str, Enum, metaclass=LlmModelMeta):
 
 MODEL_METADATA = {
     # https://platform.openai.com/docs/models
-    LlmModel.O3: ModelMetadata("openai", 200000, 100000),
+    LlmModel.O3: ModelMetadata("openai", 200000, 100000, "O3", "OpenAI", "OpenAI", 2),
-    LlmModel.O3_MINI: ModelMetadata("openai", 200000, 100000),  # o3-mini-2025-01-31
+    LlmModel.O3_MINI: ModelMetadata(
-    LlmModel.O1: ModelMetadata("openai", 200000, 100000),  # o1-2024-12-17
+        "openai", 200000, 100000, "O3 Mini", "OpenAI", "OpenAI", 1
-    LlmModel.O1_MINI: ModelMetadata("openai", 128000, 65536),  # o1-mini-2024-09-12
+    ),  # o3-mini-2025-01-31
+    LlmModel.O1: ModelMetadata(
+        "openai", 200000, 100000, "O1", "OpenAI", "OpenAI", 3
+    ),  # o1-2024-12-17
+    LlmModel.O1_MINI: ModelMetadata(
+        "openai", 128000, 65536, "O1 Mini", "OpenAI", "OpenAI", 2
+    ),  # o1-mini-2024-09-12
     # GPT-5 models
-    LlmModel.GPT5_2: ModelMetadata("openai", 400000, 128000),
+    LlmModel.GPT5_2: ModelMetadata(
-    LlmModel.GPT5_1: ModelMetadata("openai", 400000, 128000),
+        "openai", 400000, 128000, "GPT-5.2", "OpenAI", "OpenAI", 3
-    LlmModel.GPT5: ModelMetadata("openai", 400000, 128000),
+    ),
-    LlmModel.GPT5_MINI: ModelMetadata("openai", 400000, 128000),
+    LlmModel.GPT5_1: ModelMetadata(
-    LlmModel.GPT5_NANO: ModelMetadata("openai", 400000, 128000),
+        "openai", 400000, 128000, "GPT-5.1", "OpenAI", "OpenAI", 2
-    LlmModel.GPT5_CHAT: ModelMetadata("openai", 400000, 16384),
+    ),
-    LlmModel.GPT41: ModelMetadata("openai", 1047576, 32768),
+    LlmModel.GPT5: ModelMetadata(
-    LlmModel.GPT41_MINI: ModelMetadata("openai", 1047576, 32768),
+        "openai", 400000, 128000, "GPT-5", "OpenAI", "OpenAI", 1
+    ),
+    LlmModel.GPT5_MINI: ModelMetadata(
+        "openai", 400000, 128000, "GPT-5 Mini", "OpenAI", "OpenAI", 1
+    ),
+    LlmModel.GPT5_NANO: ModelMetadata(
+        "openai", 400000, 128000, "GPT-5 Nano", "OpenAI", "OpenAI", 1
+    ),
+    LlmModel.GPT5_CHAT: ModelMetadata(
+        "openai", 400000, 16384, "GPT-5 Chat Latest", "OpenAI", "OpenAI", 2
+    ),
+    LlmModel.GPT41: ModelMetadata(
+        "openai", 1047576, 32768, "GPT-4.1", "OpenAI", "OpenAI", 1
+    ),
+    LlmModel.GPT41_MINI: ModelMetadata(
+        "openai", 1047576, 32768, "GPT-4.1 Mini", "OpenAI", "OpenAI", 1
+    ),
     LlmModel.GPT4O_MINI: ModelMetadata(
-        "openai", 128000, 16384
+        "openai", 128000, 16384, "GPT-4o Mini", "OpenAI", "OpenAI", 1
     ),  # gpt-4o-mini-2024-07-18
-    LlmModel.GPT4O: ModelMetadata("openai", 128000, 16384),  # gpt-4o-2024-08-06
+    LlmModel.GPT4O: ModelMetadata(
+        "openai", 128000, 16384, "GPT-4o", "OpenAI", "OpenAI", 2
+    ),  # gpt-4o-2024-08-06
     LlmModel.GPT4_TURBO: ModelMetadata(
-        "openai", 128000, 4096
+        "openai", 128000, 4096, "GPT-4 Turbo", "OpenAI", "OpenAI", 3
     ),  # gpt-4-turbo-2024-04-09
-    LlmModel.GPT3_5_TURBO: ModelMetadata("openai", 16385, 4096),  # gpt-3.5-turbo-0125
+    LlmModel.GPT3_5_TURBO: ModelMetadata(
+        "openai", 16385, 4096, "GPT-3.5 Turbo", "OpenAI", "OpenAI", 1
+    ),  # gpt-3.5-turbo-0125
     # https://docs.anthropic.com/en/docs/about-claude/models
     LlmModel.CLAUDE_4_1_OPUS: ModelMetadata(
-        "anthropic", 200000, 32000
+        "anthropic", 200000, 32000, "Claude Opus 4.1", "Anthropic", "Anthropic", 3
     ),  # claude-opus-4-1-20250805
     LlmModel.CLAUDE_4_OPUS: ModelMetadata(
-        "anthropic", 200000, 32000
+        "anthropic", 200000, 32000, "Claude Opus 4", "Anthropic", "Anthropic", 3
     ),  # claude-4-opus-20250514
     LlmModel.CLAUDE_4_SONNET: ModelMetadata(
-        "anthropic", 200000, 64000
+        "anthropic", 200000, 64000, "Claude Sonnet 4", "Anthropic", "Anthropic", 2
     ),  # claude-4-sonnet-20250514
     LlmModel.CLAUDE_4_5_OPUS: ModelMetadata(
-        "anthropic", 200000, 64000
+        "anthropic", 200000, 64000, "Claude Opus 4.5", "Anthropic", "Anthropic", 3
     ),  # claude-opus-4-5-20251101
     LlmModel.CLAUDE_4_5_SONNET: ModelMetadata(
-        "anthropic", 200000, 64000
+        "anthropic", 200000, 64000, "Claude Sonnet 4.5", "Anthropic", "Anthropic", 3
     ),  # claude-sonnet-4-5-20250929
     LlmModel.CLAUDE_4_5_HAIKU: ModelMetadata(
-        "anthropic", 200000, 64000
+        "anthropic", 200000, 64000, "Claude Haiku 4.5", "Anthropic", "Anthropic", 2
     ),  # claude-haiku-4-5-20251001
     LlmModel.CLAUDE_3_7_SONNET: ModelMetadata(
-        "anthropic", 200000, 64000
+        "anthropic", 200000, 64000, "Claude 3.7 Sonnet", "Anthropic", "Anthropic", 2
     ),  # claude-3-7-sonnet-20250219
     LlmModel.CLAUDE_3_HAIKU: ModelMetadata(
-        "anthropic", 200000, 4096
+        "anthropic", 200000, 4096, "Claude 3 Haiku", "Anthropic", "Anthropic", 1
     ),  # claude-3-haiku-20240307
     # https://docs.aimlapi.com/api-overview/model-database/text-models
-    LlmModel.AIML_API_QWEN2_5_72B: ModelMetadata("aiml_api", 32000, 8000),
+    LlmModel.AIML_API_QWEN2_5_72B: ModelMetadata(
-    LlmModel.AIML_API_LLAMA3_1_70B: ModelMetadata("aiml_api", 128000, 40000),
+        "aiml_api", 32000, 8000, "Qwen 2.5 72B Instruct Turbo", "AI/ML", "Qwen", 1
-    LlmModel.AIML_API_LLAMA3_3_70B: ModelMetadata("aiml_api", 128000, None),
+    ),
-    LlmModel.AIML_API_META_LLAMA_3_1_70B: ModelMetadata("aiml_api", 131000, 2000),
+    LlmModel.AIML_API_LLAMA3_1_70B: ModelMetadata(
-    LlmModel.AIML_API_LLAMA_3_2_3B: ModelMetadata("aiml_api", 128000, None),
+        "aiml_api",
-    # https://console.groq.com/docs/models
+        128000,
-    LlmModel.LLAMA3_3_70B: ModelMetadata("groq", 128000, 32768),
+        40000,
-    LlmModel.LLAMA3_1_8B: ModelMetadata("groq", 128000, 8192),
+        "Llama 3.1 Nemotron 70B Instruct",
-    # https://ollama.com/library
+        "AI/ML",
-    LlmModel.OLLAMA_LLAMA3_3: ModelMetadata("ollama", 8192, None),
+        "Nvidia",
-    LlmModel.OLLAMA_LLAMA3_2: ModelMetadata("ollama", 8192, None),
+        1,
-    LlmModel.OLLAMA_LLAMA3_8B: ModelMetadata("ollama", 8192, None),
+    ),
-    LlmModel.OLLAMA_LLAMA3_405B: ModelMetadata("ollama", 8192, None),
+    LlmModel.AIML_API_LLAMA3_3_70B: ModelMetadata(
-    LlmModel.OLLAMA_DOLPHIN: ModelMetadata("ollama", 32768, None),
+        "aiml_api", 128000, None, "Llama 3.3 70B Instruct Turbo", "AI/ML", "Meta", 1
-    # https://openrouter.ai/models
+    ),
-    LlmModel.GEMINI_2_5_PRO: ModelMetadata("open_router", 1050000, 8192),
+    LlmModel.AIML_API_META_LLAMA_3_1_70B: ModelMetadata(
-    LlmModel.GEMINI_3_PRO_PREVIEW: ModelMetadata("open_router", 1048576, 65535),
+        "aiml_api", 131000, 2000, "Llama 3.1 70B Instruct Turbo", "AI/ML", "Meta", 1
-    LlmModel.GEMINI_2_5_FLASH: ModelMetadata("open_router", 1048576, 65535),
+    ),
-    LlmModel.GEMINI_2_0_FLASH: ModelMetadata("open_router", 1048576, 8192),
+    LlmModel.AIML_API_LLAMA_3_2_3B: ModelMetadata(
-    LlmModel.GEMINI_2_5_FLASH_LITE_PREVIEW: ModelMetadata(
+        "aiml_api", 128000, None, "Llama 3.2 3B Instruct Turbo", "AI/ML", "Meta", 1
-        "open_router", 1048576, 65535
+    ),
+    # https://console.groq.com/docs/models
+    LlmModel.LLAMA3_3_70B: ModelMetadata(
+        "groq", 128000, 32768, "Llama 3.3 70B Versatile", "Groq", "Meta", 1
+    ),
+    LlmModel.LLAMA3_1_8B: ModelMetadata(
+        "groq", 128000, 8192, "Llama 3.1 8B Instant", "Groq", "Meta", 1
+    ),
+    # https://ollama.com/library
+    LlmModel.OLLAMA_LLAMA3_3: ModelMetadata(
+        "ollama", 8192, None, "Llama 3.3", "Ollama", "Meta", 1
+    ),
+    LlmModel.OLLAMA_LLAMA3_2: ModelMetadata(
+        "ollama", 8192, None, "Llama 3.2", "Ollama", "Meta", 1
+    ),
+    LlmModel.OLLAMA_LLAMA3_8B: ModelMetadata(
+        "ollama", 8192, None, "Llama 3", "Ollama", "Meta", 1
+    ),
+    LlmModel.OLLAMA_LLAMA3_405B: ModelMetadata(
+        "ollama", 8192, None, "Llama 3.1 405B", "Ollama", "Meta", 1
+    ),
+    LlmModel.OLLAMA_DOLPHIN: ModelMetadata(
+        "ollama", 32768, None, "Dolphin Mistral Latest", "Ollama", "Mistral AI", 1
+    ),
+    # https://openrouter.ai/models
+    LlmModel.GEMINI_2_5_PRO: ModelMetadata(
+        "open_router",
+        1050000,
+        8192,
+        "Gemini 2.5 Pro Preview 03.25",
+        "OpenRouter",
+        "Google",
+        2,
+    ),
+    LlmModel.GEMINI_3_PRO_PREVIEW: ModelMetadata(
+        "open_router", 1048576, 65535, "Gemini 3 Pro Preview", "OpenRouter", "Google", 2
+    ),
+    LlmModel.GEMINI_2_5_FLASH: ModelMetadata(
+        "open_router", 1048576, 65535, "Gemini 2.5 Flash", "OpenRouter", "Google", 1
+    ),
+    LlmModel.GEMINI_2_0_FLASH: ModelMetadata(
+        "open_router", 1048576, 8192, "Gemini 2.0 Flash 001", "OpenRouter", "Google", 1
+    ),
+    LlmModel.GEMINI_2_5_FLASH_LITE_PREVIEW: ModelMetadata(
+        "open_router",
+        1048576,
+        65535,
+        "Gemini 2.5 Flash Lite Preview 06.17",
+        "OpenRouter",
+        "Google",
+        1,
+    ),
+    LlmModel.GEMINI_2_0_FLASH_LITE: ModelMetadata(
+        "open_router",
+        1048576,
+        8192,
+        "Gemini 2.0 Flash Lite 001",
+        "OpenRouter",
+        "Google",
+        1,
+    ),
+    LlmModel.MISTRAL_NEMO: ModelMetadata(
+        "open_router", 128000, 4096, "Mistral Nemo", "OpenRouter", "Mistral AI", 1
+    ),
+    LlmModel.COHERE_COMMAND_R_08_2024: ModelMetadata(
+        "open_router", 128000, 4096, "Command R 08.2024", "OpenRouter", "Cohere", 1
+    ),
+    LlmModel.COHERE_COMMAND_R_PLUS_08_2024: ModelMetadata(
+        "open_router", 128000, 4096, "Command R Plus 08.2024", "OpenRouter", "Cohere", 2
+    ),
+    LlmModel.DEEPSEEK_CHAT: ModelMetadata(
+        "open_router", 64000, 2048, "DeepSeek Chat", "OpenRouter", "DeepSeek", 1
+    ),
+    LlmModel.DEEPSEEK_R1_0528: ModelMetadata(
+        "open_router", 163840, 163840, "DeepSeek R1 0528", "OpenRouter", "DeepSeek", 1
+    ),
+    LlmModel.PERPLEXITY_SONAR: ModelMetadata(
+        "open_router", 127000, 8000, "Sonar", "OpenRouter", "Perplexity", 1
+    ),
+    LlmModel.PERPLEXITY_SONAR_PRO: ModelMetadata(
+        "open_router", 200000, 8000, "Sonar Pro", "OpenRouter", "Perplexity", 2
     ),
-    LlmModel.GEMINI_2_0_FLASH_LITE: ModelMetadata("open_router", 1048576, 8192),
-    LlmModel.MISTRAL_NEMO: ModelMetadata("open_router", 128000, 4096),
-    LlmModel.COHERE_COMMAND_R_08_2024: ModelMetadata("open_router", 128000, 4096),
-    LlmModel.COHERE_COMMAND_R_PLUS_08_2024: ModelMetadata("open_router", 128000, 4096),
-    LlmModel.DEEPSEEK_CHAT: ModelMetadata("open_router", 64000, 2048),
-    LlmModel.DEEPSEEK_R1_0528: ModelMetadata("open_router", 163840, 163840),
-    LlmModel.PERPLEXITY_SONAR: ModelMetadata("open_router", 127000, 8000),
-    LlmModel.PERPLEXITY_SONAR_PRO: ModelMetadata("open_router", 200000, 8000),
     LlmModel.PERPLEXITY_SONAR_DEEP_RESEARCH: ModelMetadata(
         "open_router",
         128000,
         16000,
+        "Sonar Deep Research",
+        "OpenRouter",
+        "Perplexity",
+        3,
     ),
     LlmModel.NOUSRESEARCH_HERMES_3_LLAMA_3_1_405B: ModelMetadata(
-        "open_router", 131000, 4096
+        "open_router",
+        131000,
+        4096,
+        "Hermes 3 Llama 3.1 405B",
+        "OpenRouter",
+        "Nous Research",
+        1,
     ),
     LlmModel.NOUSRESEARCH_HERMES_3_LLAMA_3_1_70B: ModelMetadata(
-        "open_router", 12288, 12288
+        "open_router",
+        12288,
+        12288,
+        "Hermes 3 Llama 3.1 70B",
+        "OpenRouter",
+        "Nous Research",
+        1,
+    ),
+    LlmModel.OPENAI_GPT_OSS_120B: ModelMetadata(
+        "open_router", 131072, 131072, "GPT-OSS 120B", "OpenRouter", "OpenAI", 1
+    ),
+    LlmModel.OPENAI_GPT_OSS_20B: ModelMetadata(
+        "open_router", 131072, 32768, "GPT-OSS 20B", "OpenRouter", "OpenAI", 1
+    ),
+    LlmModel.AMAZON_NOVA_LITE_V1: ModelMetadata(
+        "open_router", 300000, 5120, "Nova Lite V1", "OpenRouter", "Amazon", 1
+    ),
+    LlmModel.AMAZON_NOVA_MICRO_V1: ModelMetadata(
+        "open_router", 128000, 5120, "Nova Micro V1", "OpenRouter", "Amazon", 1
+    ),
+    LlmModel.AMAZON_NOVA_PRO_V1: ModelMetadata(
+        "open_router", 300000, 5120, "Nova Pro V1", "OpenRouter", "Amazon", 1
+    ),
+    LlmModel.MICROSOFT_WIZARDLM_2_8X22B: ModelMetadata(
+        "open_router", 65536, 4096, "WizardLM 2 8x22B", "OpenRouter", "Microsoft", 1
+    ),
+    LlmModel.GRYPHE_MYTHOMAX_L2_13B: ModelMetadata(
+        "open_router", 4096, 4096, "MythoMax L2 13B", "OpenRouter", "Gryphe", 1
+    ),
+    LlmModel.META_LLAMA_4_SCOUT: ModelMetadata(
+        "open_router", 131072, 131072, "Llama 4 Scout", "OpenRouter", "Meta", 1
+    ),
+    LlmModel.META_LLAMA_4_MAVERICK: ModelMetadata(
+        "open_router", 1048576, 1000000, "Llama 4 Maverick", "OpenRouter", "Meta", 1
+    ),
+    LlmModel.GROK_4: ModelMetadata(
+        "open_router", 256000, 256000, "Grok 4", "OpenRouter", "xAI", 3
+    ),
+    LlmModel.GROK_4_FAST: ModelMetadata(
+        "open_router", 2000000, 30000, "Grok 4 Fast", "OpenRouter", "xAI", 1
+    ),
+    LlmModel.GROK_4_1_FAST: ModelMetadata(
+        "open_router", 2000000, 30000, "Grok 4.1 Fast", "OpenRouter", "xAI", 1
+    ),
+    LlmModel.GROK_CODE_FAST_1: ModelMetadata(
+        "open_router", 256000, 10000, "Grok Code Fast 1", "OpenRouter", "xAI", 1
+    ),
+    LlmModel.KIMI_K2: ModelMetadata(
+        "open_router", 131000, 131000, "Kimi K2", "OpenRouter", "Moonshot AI", 1
+    ),
+    LlmModel.QWEN3_235B_A22B_THINKING: ModelMetadata(
+        "open_router",
+        262144,
+        262144,
+        "Qwen 3 235B A22B Thinking 2507",
+        "OpenRouter",
+        "Qwen",
+        1,
+    ),
+    LlmModel.QWEN3_CODER: ModelMetadata(
+        "open_router", 262144, 262144, "Qwen 3 Coder", "OpenRouter", "Qwen", 3
     ),
-    LlmModel.OPENAI_GPT_OSS_120B: ModelMetadata("open_router", 131072, 131072),
-    LlmModel.OPENAI_GPT_OSS_20B: ModelMetadata("open_router", 131072, 32768),
-    LlmModel.AMAZON_NOVA_LITE_V1: ModelMetadata("open_router", 300000, 5120),
-    LlmModel.AMAZON_NOVA_MICRO_V1: ModelMetadata("open_router", 128000, 5120),
-    LlmModel.AMAZON_NOVA_PRO_V1: ModelMetadata("open_router", 300000, 5120),
-    LlmModel.MICROSOFT_WIZARDLM_2_8X22B: ModelMetadata("open_router", 65536, 4096),
-    LlmModel.GRYPHE_MYTHOMAX_L2_13B: ModelMetadata("open_router", 4096, 4096),
-    LlmModel.META_LLAMA_4_SCOUT: ModelMetadata("open_router", 131072, 131072),
-    LlmModel.META_LLAMA_4_MAVERICK: ModelMetadata("open_router", 1048576, 1000000),
-    LlmModel.GROK_4: ModelMetadata("open_router", 256000, 256000),
-    LlmModel.GROK_4_FAST: ModelMetadata("open_router", 2000000, 30000),
-    LlmModel.GROK_4_1_FAST: ModelMetadata("open_router", 2000000, 30000),
-    LlmModel.GROK_CODE_FAST_1: ModelMetadata("open_router", 256000, 10000),
-    LlmModel.KIMI_K2: ModelMetadata("open_router", 131000, 131000),
-    LlmModel.QWEN3_235B_A22B_THINKING: ModelMetadata("open_router", 262144, 262144),
-    LlmModel.QWEN3_CODER: ModelMetadata("open_router", 262144, 262144),
     # Llama API models
-    LlmModel.LLAMA_API_LLAMA_4_SCOUT: ModelMetadata("llama_api", 128000, 4028),
+    LlmModel.LLAMA_API_LLAMA_4_SCOUT: ModelMetadata(
-    LlmModel.LLAMA_API_LLAMA4_MAVERICK: ModelMetadata("llama_api", 128000, 4028),
+        "llama_api",
-    LlmModel.LLAMA_API_LLAMA3_3_8B: ModelMetadata("llama_api", 128000, 4028),
+        128000,
-    LlmModel.LLAMA_API_LLAMA3_3_70B: ModelMetadata("llama_api", 128000, 4028),
+        4028,
+        "Llama 4 Scout 17B 16E Instruct FP8",
+        "Llama API",
+        "Meta",
+        1,
+    ),
+    LlmModel.LLAMA_API_LLAMA4_MAVERICK: ModelMetadata(
+        "llama_api",
+        128000,
+        4028,
+        "Llama 4 Maverick 17B 128E Instruct FP8",
+        "Llama API",
+        "Meta",
+        1,
+    ),
+    LlmModel.LLAMA_API_LLAMA3_3_8B: ModelMetadata(
+        "llama_api", 128000, 4028, "Llama 3.3 8B Instruct", "Llama API", "Meta", 1
+    ),
+    LlmModel.LLAMA_API_LLAMA3_3_70B: ModelMetadata(
+        "llama_api", 128000, 4028, "Llama 3.3 70B Instruct", "Llama API", "Meta", 1
+    ),
     # v0 by Vercel models
-    LlmModel.V0_1_5_MD: ModelMetadata("v0", 128000, 64000),
+    LlmModel.V0_1_5_MD: ModelMetadata("v0", 128000, 64000, "v0 1.5 MD", "V0", "V0", 1),
-    LlmModel.V0_1_5_LG: ModelMetadata("v0", 512000, 64000),
+    LlmModel.V0_1_5_LG: ModelMetadata("v0", 512000, 64000, "v0 1.5 LG", "V0", "V0", 1),
-    LlmModel.V0_1_0_MD: ModelMetadata("v0", 128000, 64000),
+    LlmModel.V0_1_0_MD: ModelMetadata("v0", 128000, 64000, "v0 1.0 MD", "V0", "V0", 1),
 }
 
 DEFAULT_LLM_MODEL = LlmModel.GPT5_2
@@ -854,7 +1050,7 @@ class AIStructuredResponseGeneratorBlock(AIBlockBase):
     def __init__(self):
         super().__init__(
             id="ed55ac19-356e-4243-a6cb-bc599e9b716f",
-            description="Call a Large Language Model (LLM) to generate formatted object based on the given prompt.",
+            description="A block that generates structured JSON responses using a Large Language Model (LLM), with schema validation and format enforcement.",
             categories={BlockCategory.AI},
             input_schema=AIStructuredResponseGeneratorBlock.Input,
             output_schema=AIStructuredResponseGeneratorBlock.Output,
@@ -1265,7 +1461,7 @@ class AITextGeneratorBlock(AIBlockBase):
     def __init__(self):
         super().__init__(
             id="1f292d4a-41a4-4977-9684-7c8d560b9f91",
-            description="Call a Large Language Model (LLM) to generate a string based on the given prompt.",
+            description="A block that produces text responses using a Large Language Model (LLM) based on customizable prompts and system instructions.",
             categories={BlockCategory.AI},
             input_schema=AITextGeneratorBlock.Input,
             output_schema=AITextGeneratorBlock.Output,
@@ -1361,7 +1557,7 @@ class AITextSummarizerBlock(AIBlockBase):
     def __init__(self):
         super().__init__(
             id="a0a69be1-4528-491c-a85a-a4ab6873e3f0",
-            description="Utilize a Large Language Model (LLM) to summarize a long text.",
+            description="A block that summarizes long texts using a Large Language Model (LLM), with configurable focus topics and summary styles.",
             categories={BlockCategory.AI, BlockCategory.TEXT},
             input_schema=AITextSummarizerBlock.Input,
             output_schema=AITextSummarizerBlock.Output,
@@ -1562,7 +1758,7 @@ class AIConversationBlock(AIBlockBase):
     def __init__(self):
         super().__init__(
             id="32a87eab-381e-4dd4-bdb8-4c47151be35a",
-            description="Advanced LLM call that takes a list of messages and sends them to the language model.",
+            description="A block that facilitates multi-turn conversations with a Large Language Model (LLM), maintaining context across message exchanges.",
             categories={BlockCategory.AI},
             input_schema=AIConversationBlock.Input,
             output_schema=AIConversationBlock.Output,
@@ -1682,7 +1878,7 @@ class AIListGeneratorBlock(AIBlockBase):
     def __init__(self):
         super().__init__(
             id="9c0b0450-d199-458b-a731-072189dd6593",
-            description="Generate a list of values based on the given prompt using a Large Language Model (LLM).",
+            description="A block that creates lists of items based on prompts using a Large Language Model (LLM), with optional source data for context.",
             categories={BlockCategory.AI, BlockCategory.TEXT},
             input_schema=AIListGeneratorBlock.Input,
             output_schema=AIListGeneratorBlock.Output,

autogpt_platform/backend/backend/blocks/medium.py

@@ -46,7 +46,7 @@ class PublishToMediumBlock(Block):
     class Input(BlockSchemaInput):
         author_id: BlockSecret = SecretField(
             key="medium_author_id",
-            description="""The Medium AuthorID of the user. You can get this by calling the /me endpoint of the Medium API.\n\ncurl -H "Authorization: Bearer YOUR_ACCESS_TOKEN" https://api.medium.com/v1/me" the response will contain the authorId field.""",
+            description="""The Medium AuthorID of the user. You can get this by calling the /me endpoint of the Medium API.\n\ncurl -H "Authorization: Bearer YOUR_ACCESS_TOKEN" https://api.medium.com/v1/me\n\nThe response will contain the authorId field.""",
             placeholder="Enter the author's Medium AuthorID",
         )
         title: str = SchemaField(

autogpt_platform/backend/backend/blocks/talking_head.py

@@ -50,7 +50,7 @@ class CreateTalkingAvatarVideoBlock(Block):
             description="The voice provider to use", default="microsoft"
         )
         voice_id: str = SchemaField(
-            description="The voice ID to use, get list of voices [here](https://docs.agpt.co/server/d_id)",
+            description="The voice ID to use, see [available voice IDs](https://agpt.co/docs/platform/using-ai-services/d_id)",
             default="en-US-JennyNeural",
         )
         presenter_id: str = SchemaField(

autogpt_platform/backend/backend/blocks/test/test_smart_decision_maker.py

@@ -242,7 +242,7 @@ async def test_smart_decision_maker_tracks_llm_stats():
         outputs = {}
         # Create execution context
 
-        mock_execution_context = ExecutionContext(safe_mode=False)
+        mock_execution_context = ExecutionContext(human_in_the_loop_safe_mode=False)
 
         # Create a mock execution processor for tests
 
@@ -343,7 +343,7 @@ async def test_smart_decision_maker_parameter_validation():
 
         # Create execution context
 
-        mock_execution_context = ExecutionContext(safe_mode=False)
+        mock_execution_context = ExecutionContext(human_in_the_loop_safe_mode=False)
 
         # Create a mock execution processor for tests
 
@@ -409,7 +409,7 @@ async def test_smart_decision_maker_parameter_validation():
 
         # Create execution context
 
-        mock_execution_context = ExecutionContext(safe_mode=False)
+        mock_execution_context = ExecutionContext(human_in_the_loop_safe_mode=False)
 
         # Create a mock execution processor for tests
 
@@ -471,7 +471,7 @@ async def test_smart_decision_maker_parameter_validation():
         outputs = {}
         # Create execution context
 
-        mock_execution_context = ExecutionContext(safe_mode=False)
+        mock_execution_context = ExecutionContext(human_in_the_loop_safe_mode=False)
 
         # Create a mock execution processor for tests
 
@@ -535,7 +535,7 @@ async def test_smart_decision_maker_parameter_validation():
         outputs = {}
         # Create execution context
 
-        mock_execution_context = ExecutionContext(safe_mode=False)
+        mock_execution_context = ExecutionContext(human_in_the_loop_safe_mode=False)
 
         # Create a mock execution processor for tests
 
@@ -658,7 +658,7 @@ async def test_smart_decision_maker_raw_response_conversion():
         outputs = {}
         # Create execution context
 
-        mock_execution_context = ExecutionContext(safe_mode=False)
+        mock_execution_context = ExecutionContext(human_in_the_loop_safe_mode=False)
 
         # Create a mock execution processor for tests
 
@@ -730,7 +730,7 @@ async def test_smart_decision_maker_raw_response_conversion():
         outputs = {}
         # Create execution context
 
-        mock_execution_context = ExecutionContext(safe_mode=False)
+        mock_execution_context = ExecutionContext(human_in_the_loop_safe_mode=False)
 
         # Create a mock execution processor for tests
 
@@ -786,7 +786,7 @@ async def test_smart_decision_maker_raw_response_conversion():
         outputs = {}
         # Create execution context
 
-        mock_execution_context = ExecutionContext(safe_mode=False)
+        mock_execution_context = ExecutionContext(human_in_the_loop_safe_mode=False)
 
         # Create a mock execution processor for tests
 
@@ -905,7 +905,7 @@ async def test_smart_decision_maker_agent_mode():
         # Create a mock execution context
 
         mock_execution_context = ExecutionContext(
-            safe_mode=False,
+            human_in_the_loop_safe_mode=False,
         )
 
         # Create a mock execution processor for agent mode tests
@@ -1027,7 +1027,7 @@ async def test_smart_decision_maker_traditional_mode_default():
 
         # Create execution context
 
-        mock_execution_context = ExecutionContext(safe_mode=False)
+        mock_execution_context = ExecutionContext(human_in_the_loop_safe_mode=False)
 
         # Create a mock execution processor for tests
 

autogpt_platform/backend/backend/blocks/test/test_smart_decision_maker_dynamic_fields.py

@@ -386,7 +386,7 @@ async def test_output_yielding_with_dynamic_fields():
             outputs = {}
             from backend.data.execution import ExecutionContext
 
-            mock_execution_context = ExecutionContext(safe_mode=False)
+            mock_execution_context = ExecutionContext(human_in_the_loop_safe_mode=False)
             mock_execution_processor = MagicMock()
 
             async for output_name, output_value in block.run(
@@ -609,7 +609,9 @@ async def test_validation_errors_dont_pollute_conversation():
                 outputs = {}
                 from backend.data.execution import ExecutionContext
 
-                mock_execution_context = ExecutionContext(safe_mode=False)
+                mock_execution_context = ExecutionContext(
+                    human_in_the_loop_safe_mode=False
+                )
 
                 # Create a proper mock execution processor for agent mode
                 from collections import defaultdict

autogpt_platform/backend/backend/data/block.py

@@ -474,7 +474,7 @@ class Block(ABC, Generic[BlockSchemaInputType, BlockSchemaOutputType]):
         self.block_type = block_type
         self.webhook_config = webhook_config
         self.execution_stats: NodeExecutionStats = NodeExecutionStats()
-        self.requires_human_review: bool = False
+        self.is_sensitive_action: bool = False
 
         if self.webhook_config:
             if isinstance(self.webhook_config, BlockWebhookConfig):
@@ -637,8 +637,9 @@ class Block(ABC, Generic[BlockSchemaInputType, BlockSchemaOutputType]):
             - should_pause: True if execution should be paused for review
             - input_data_to_use: The input data to use (may be modified by reviewer)
         """
-        # Skip review if not required or safe mode is disabled
+        if not (
-        if not self.requires_human_review or not execution_context.safe_mode:
+            self.is_sensitive_action and execution_context.sensitive_action_safe_mode
+        ):
             return False, input_data
 
         from backend.blocks.helpers.review import HITLReviewHelper

autogpt_platform/backend/backend/data/block_cost_config.py

@@ -99,10 +99,15 @@ MODEL_COST: dict[LlmModel, int] = {
     LlmModel.OPENAI_GPT_OSS_20B: 1,
     LlmModel.GEMINI_2_5_PRO: 4,
     LlmModel.GEMINI_3_PRO_PREVIEW: 5,
+    LlmModel.GEMINI_2_5_FLASH: 1,
+    LlmModel.GEMINI_2_0_FLASH: 1,
+    LlmModel.GEMINI_2_5_FLASH_LITE_PREVIEW: 1,
+    LlmModel.GEMINI_2_0_FLASH_LITE: 1,
     LlmModel.MISTRAL_NEMO: 1,
     LlmModel.COHERE_COMMAND_R_08_2024: 1,
     LlmModel.COHERE_COMMAND_R_PLUS_08_2024: 3,
     LlmModel.DEEPSEEK_CHAT: 2,
+    LlmModel.DEEPSEEK_R1_0528: 1,
     LlmModel.PERPLEXITY_SONAR: 1,
     LlmModel.PERPLEXITY_SONAR_PRO: 5,
     LlmModel.PERPLEXITY_SONAR_DEEP_RESEARCH: 10,
@@ -126,11 +131,6 @@ MODEL_COST: dict[LlmModel, int] = {
     LlmModel.KIMI_K2: 1,
     LlmModel.QWEN3_235B_A22B_THINKING: 1,
     LlmModel.QWEN3_CODER: 9,
-    LlmModel.GEMINI_2_5_FLASH: 1,
-    LlmModel.GEMINI_2_0_FLASH: 1,
-    LlmModel.GEMINI_2_5_FLASH_LITE_PREVIEW: 1,
-    LlmModel.GEMINI_2_0_FLASH_LITE: 1,
-    LlmModel.DEEPSEEK_R1_0528: 1,
     # v0 by Vercel models
     LlmModel.V0_1_5_MD: 1,
     LlmModel.V0_1_5_LG: 2,

autogpt_platform/backend/backend/data/db.py

@@ -38,20 +38,6 @@ POOL_TIMEOUT = os.getenv("DB_POOL_TIMEOUT")
 if POOL_TIMEOUT:
     DATABASE_URL = add_param(DATABASE_URL, "pool_timeout", POOL_TIMEOUT)
 
-# Add public schema to search_path for pgvector type access
-# The vector extension is in public schema, but search_path is determined by schema parameter
-# Extract the schema from DATABASE_URL or default to 'public' (matching get_database_schema())
-parsed_url = urlparse(DATABASE_URL)
-url_params = dict(parse_qsl(parsed_url.query))
-db_schema = url_params.get("schema", "public")
-# Build search_path, avoiding duplicates if db_schema is already 'public'
-search_path_schemas = list(
-    dict.fromkeys([db_schema, "public"])
-)  # Preserves order, removes duplicates
-search_path = ",".join(search_path_schemas)
-# This allows using ::vector without schema qualification
-DATABASE_URL = add_param(DATABASE_URL, "options", f"-c search_path={search_path}")
-
 HTTP_TIMEOUT = int(POOL_TIMEOUT) if POOL_TIMEOUT else None
 
 prisma = Prisma(
@@ -127,38 +113,48 @@ async def _raw_with_schema(
     *args,
     execute: bool = False,
     client: Prisma | None = None,
-    set_public_search_path: bool = False,
 ) -> list[dict] | int:
     """Internal: Execute raw SQL with proper schema handling.
 
     Use query_raw_with_schema() or execute_raw_with_schema() instead.
 
+    Supports placeholders:
+        - {schema_prefix}: Table/type prefix (e.g., "platform".)
+        - {schema}: Raw schema name for application tables (e.g., platform)
+
+    Note on pgvector types:
+        Use unqualified ::vector and <=> operator in queries. PostgreSQL resolves
+        these via search_path, which includes the schema where pgvector is installed
+        on all environments (local, CI, dev).
+
     Args:
-        query_template: SQL query with {schema_prefix} placeholder
+        query_template: SQL query with {schema_prefix} and/or {schema} placeholders
         *args: Query parameters
         execute: If False, executes SELECT query. If True, executes INSERT/UPDATE/DELETE.
         client: Optional Prisma client for transactions (only used when execute=True).
-        set_public_search_path: If True, sets search_path to include public schema.
-                                Needed for pgvector types and other public schema objects.
 
     Returns:
         - list[dict] if execute=False (query results)
         - int if execute=True (number of affected rows)
+
+    Example with vector type:
+        await execute_raw_with_schema(
+            'INSERT INTO {schema_prefix}"Embedding" (vec) VALUES ($1::vector)',
+            embedding_data
+        )
     """
     schema = get_database_schema()
     schema_prefix = f'"{schema}".' if schema != "public" else ""
-    formatted_query = query_template.format(schema_prefix=schema_prefix)
+
+    formatted_query = query_template.format(
+        schema_prefix=schema_prefix,
+        schema=schema,
+    )
 
     import prisma as prisma_module
 
     db_client = client if client else prisma_module.get_client()
 
-    # Set search_path to include public schema if requested
-    # Prisma doesn't support the 'options' connection parameter, so we set it per-session
-    # This is idempotent and safe to call multiple times
-    if set_public_search_path:
-        await db_client.execute_raw(f"SET search_path = {schema}, public")  # type: ignore
-
     if execute:
         result = await db_client.execute_raw(formatted_query, *args)  # type: ignore
     else:
@@ -167,16 +163,12 @@ async def _raw_with_schema(
     return result
 
 
-async def query_raw_with_schema(
+async def query_raw_with_schema(query_template: str, *args) -> list[dict]:
-    query_template: str, *args, set_public_search_path: bool = False
-) -> list[dict]:
     """Execute raw SQL SELECT query with proper schema handling.
 
     Args:
-        query_template: SQL query with {schema_prefix} placeholder
+        query_template: SQL query with {schema_prefix} and/or {schema} placeholders
         *args: Query parameters
-        set_public_search_path: If True, sets search_path to include public schema.
-                                Needed for pgvector types and other public schema objects.
 
     Returns:
         List of result rows as dictionaries
@@ -187,23 +179,20 @@ async def query_raw_with_schema(
             user_id
         )
     """
-    return await _raw_with_schema(query_template, *args, execute=False, set_public_search_path=set_public_search_path)  # type: ignore
+    return await _raw_with_schema(query_template, *args, execute=False)  # type: ignore
 
 
 async def execute_raw_with_schema(
     query_template: str,
     *args,
     client: Prisma | None = None,
-    set_public_search_path: bool = False,
 ) -> int:
     """Execute raw SQL command (INSERT/UPDATE/DELETE) with proper schema handling.
 
     Args:
-        query_template: SQL query with {schema_prefix} placeholder
+        query_template: SQL query with {schema_prefix} and/or {schema} placeholders
         *args: Query parameters
         client: Optional Prisma client for transactions
-        set_public_search_path: If True, sets search_path to include public schema.
-                                Needed for pgvector types and other public schema objects.
 
     Returns:
         Number of affected rows
@@ -215,7 +204,7 @@ async def execute_raw_with_schema(
             client=tx  # Optional transaction client
         )
     """
-    return await _raw_with_schema(query_template, *args, execute=True, client=client, set_public_search_path=set_public_search_path)  # type: ignore
+    return await _raw_with_schema(query_template, *args, execute=True, client=client)  # type: ignore
 
 
 class BaseDbModel(BaseModel):

autogpt_platform/backend/backend/data/event_bus.py

@@ -103,8 +103,18 @@ class RedisEventBus(BaseRedisEventBus[M], ABC):
         return redis.get_redis()
 
     def publish_event(self, event: M, channel_key: str):
-        message, full_channel_name = self._serialize_message(event, channel_key)
+        """
-        self.connection.publish(full_channel_name, message)
+        Publish an event to Redis. Gracefully handles connection failures
+        by logging the error instead of raising exceptions.
+        """
+        try:
+            message, full_channel_name = self._serialize_message(event, channel_key)
+            self.connection.publish(full_channel_name, message)
+        except Exception:
+            logger.exception(
+                f"Failed to publish event to Redis channel {channel_key}. "
+                "Event bus operation will continue without Redis connectivity."
+            )
 
     def listen_events(self, channel_key: str) -> Generator[M, None, None]:
         pubsub, full_channel_name = self._get_pubsub_channel(
@@ -128,9 +138,19 @@ class AsyncRedisEventBus(BaseRedisEventBus[M], ABC):
         return await redis.get_redis_async()
 
     async def publish_event(self, event: M, channel_key: str):
-        message, full_channel_name = self._serialize_message(event, channel_key)
+        """
-        connection = await self.connection
+        Publish an event to Redis. Gracefully handles connection failures
-        await connection.publish(full_channel_name, message)
+        by logging the error instead of raising exceptions.
+        """
+        try:
+            message, full_channel_name = self._serialize_message(event, channel_key)
+            connection = await self.connection
+            await connection.publish(full_channel_name, message)
+        except Exception:
+            logger.exception(
+                f"Failed to publish event to Redis channel {channel_key}. "
+                "Event bus operation will continue without Redis connectivity."
+            )
 
     async def listen_events(self, channel_key: str) -> AsyncGenerator[M, None]:
         pubsub, full_channel_name = self._get_pubsub_channel(

autogpt_platform/backend/backend/data/event_bus_test.py

@@ -0,0 +1,56 @@
+"""
+Tests for event_bus graceful degradation when Redis is unavailable.
+"""
+
+from unittest.mock import AsyncMock, patch
+
+import pytest
+from pydantic import BaseModel
+
+from backend.data.event_bus import AsyncRedisEventBus
+
+
+class TestEvent(BaseModel):
+    """Test event model."""
+
+    message: str
+
+
+class TestNotificationBus(AsyncRedisEventBus[TestEvent]):
+    """Test implementation of AsyncRedisEventBus."""
+
+    Model = TestEvent
+
+    @property
+    def event_bus_name(self) -> str:
+        return "test_event_bus"
+
+
+@pytest.mark.asyncio
+async def test_publish_event_handles_connection_failure_gracefully():
+    """Test that publish_event logs exception instead of raising when Redis is unavailable."""
+    bus = TestNotificationBus()
+    event = TestEvent(message="test message")
+
+    # Mock get_redis_async to raise connection error
+    with patch(
+        "backend.data.event_bus.redis.get_redis_async",
+        side_effect=ConnectionError("Authentication required."),
+    ):
+        # Should not raise exception
+        await bus.publish_event(event, "test_channel")
+
+
+@pytest.mark.asyncio
+async def test_publish_event_works_with_redis_available():
+    """Test that publish_event works normally when Redis is available."""
+    bus = TestNotificationBus()
+    event = TestEvent(message="test message")
+
+    # Mock successful Redis connection
+    mock_redis = AsyncMock()
+    mock_redis.publish = AsyncMock()
+
+    with patch("backend.data.event_bus.redis.get_redis_async", return_value=mock_redis):
+        await bus.publish_event(event, "test_channel")
+        mock_redis.publish.assert_called_once()

autogpt_platform/backend/backend/data/execution.py

@@ -81,7 +81,10 @@ class ExecutionContext(BaseModel):
     This includes information needed by blocks, sub-graphs, and execution management.
     """
 
-    safe_mode: bool = True
+    model_config = {"extra": "ignore"}
+
+    human_in_the_loop_safe_mode: bool = True
+    sensitive_action_safe_mode: bool = False
     user_timezone: str = "UTC"
     root_execution_id: Optional[str] = None
     parent_execution_id: Optional[str] = None

autogpt_platform/backend/backend/data/graph.py

@@ -3,7 +3,7 @@ import logging
 import uuid
 from collections import defaultdict
 from datetime import datetime, timezone
-from typing import TYPE_CHECKING, Any, Literal, Optional, cast
+from typing import TYPE_CHECKING, Annotated, Any, Literal, Optional, cast
 
 from prisma.enums import SubmissionStatus
 from prisma.models import (
@@ -20,7 +20,7 @@ from prisma.types import (
     AgentNodeLinkCreateInput,
     StoreListingVersionWhereInput,
 )
-from pydantic import BaseModel, Field, create_model
+from pydantic import BaseModel, BeforeValidator, Field, create_model
 from pydantic.fields import computed_field
 
 from backend.blocks.agent import AgentExecutorBlock
@@ -62,7 +62,31 @@ logger = logging.getLogger(__name__)
 
 
 class GraphSettings(BaseModel):
-    human_in_the_loop_safe_mode: bool | None = None
+    # Use Annotated with BeforeValidator to coerce None to default values.
+    # This handles cases where the database has null values for these fields.
+    model_config = {"extra": "ignore"}
+
+    human_in_the_loop_safe_mode: Annotated[
+        bool, BeforeValidator(lambda v: v if v is not None else True)
+    ] = True
+    sensitive_action_safe_mode: Annotated[
+        bool, BeforeValidator(lambda v: v if v is not None else False)
+    ] = False
+
+    @classmethod
+    def from_graph(
+        cls,
+        graph: "GraphModel",
+        hitl_safe_mode: bool | None = None,
+        sensitive_action_safe_mode: bool = False,
+    ) -> "GraphSettings":
+        # Default to True if not explicitly set
+        if hitl_safe_mode is None:
+            hitl_safe_mode = True
+        return cls(
+            human_in_the_loop_safe_mode=hitl_safe_mode,
+            sensitive_action_safe_mode=sensitive_action_safe_mode,
+        )
 
 
 class Link(BaseDbModel):
@@ -244,10 +268,14 @@ class BaseGraph(BaseDbModel):
         return any(
             node.block_id
             for node in self.nodes
-            if (
+            if node.block.block_type == BlockType.HUMAN_IN_THE_LOOP
-                node.block.block_type == BlockType.HUMAN_IN_THE_LOOP
+        )
-                or node.block.requires_human_review
+
-            )
+    @computed_field
+    @property
+    def has_sensitive_action(self) -> bool:
+        return any(
+            node.block_id for node in self.nodes if node.block.is_sensitive_action
         )
 
     @property

autogpt_platform/backend/backend/data/understanding.py

@@ -328,6 +328,8 @@ async def clear_business_understanding(user_id: str) -> bool:
 
 def format_understanding_for_prompt(understanding: BusinessUnderstanding) -> str:
     """Format business understanding as text for system prompt injection."""
+    if not understanding:
+        return ""
     sections = []
 
     # User info section

autogpt_platform/backend/backend/executor/scheduler.py

@@ -309,7 +309,7 @@ def ensure_embeddings_coverage():
 
         # Process in batches until no more missing embeddings
         while True:
-            result = db_client.backfill_missing_embeddings(batch_size=10)
+            result = db_client.backfill_missing_embeddings(batch_size=100)
 
             total_processed += result["processed"]
             total_success += result["success"]

autogpt_platform/backend/backend/executor/utils.py

@@ -873,11 +873,8 @@ async def add_graph_execution(
         settings = await gdb.get_graph_settings(user_id=user_id, graph_id=graph_id)
 
         execution_context = ExecutionContext(
-            safe_mode=(
+            human_in_the_loop_safe_mode=settings.human_in_the_loop_safe_mode,
-                settings.human_in_the_loop_safe_mode
+            sensitive_action_safe_mode=settings.sensitive_action_safe_mode,
-                if settings.human_in_the_loop_safe_mode is not None
-                else True
-            ),
             user_timezone=(
                 user.timezone if user.timezone != USER_TIMEZONE_NOT_SET else "UTC"
             ),

autogpt_platform/backend/backend/executor/utils_test.py

@@ -386,6 +386,7 @@ async def test_add_graph_execution_is_repeatable(mocker: MockerFixture):
     mock_user.timezone = "UTC"
     mock_settings = mocker.MagicMock()
     mock_settings.human_in_the_loop_safe_mode = True
+    mock_settings.sensitive_action_safe_mode = False
 
     mock_udb.get_user_by_id = mocker.AsyncMock(return_value=mock_user)
     mock_gdb.get_graph_settings = mocker.AsyncMock(return_value=mock_settings)
@@ -651,6 +652,7 @@ async def test_add_graph_execution_with_nodes_to_skip(mocker: MockerFixture):
     mock_user.timezone = "UTC"
     mock_settings = mocker.MagicMock()
     mock_settings.human_in_the_loop_safe_mode = True
+    mock_settings.sensitive_action_safe_mode = False
 
     mock_udb.get_user_by_id = mocker.AsyncMock(return_value=mock_user)
     mock_gdb.get_graph_settings = mocker.AsyncMock(return_value=mock_settings)

autogpt_platform/backend/migrations/20260109181714_add_docs_embedding/migration.sql

@@ -1,11 +1,37 @@
 -- CreateExtension
 -- Supabase: pgvector must be enabled via Dashboard → Database → Extensions first
--- Create in public schema so vector type is available across all schemas
+-- Ensures vector extension is in the current schema (from DATABASE_URL ?schema= param)
+-- If it exists in a different schema (e.g., public), we drop and recreate it in the current schema
+-- This ensures vector type is in the same schema as tables, making ::vector work without explicit qualification
 DO $$
+DECLARE
+    current_schema_name text;
+    vector_schema text;
 BEGIN
-    CREATE EXTENSION IF NOT EXISTS "vector" WITH SCHEMA "public";
+    -- Get the current schema from search_path
-EXCEPTION WHEN OTHERS THEN
+    SELECT current_schema() INTO current_schema_name;
-    RAISE NOTICE 'vector extension not available or already exists, skipping';
+
+    -- Check if vector extension exists and which schema it's in
+    SELECT n.nspname INTO vector_schema
+    FROM pg_extension e
+    JOIN pg_namespace n ON e.extnamespace = n.oid
+    WHERE e.extname = 'vector';
+
+    -- Handle removal if in wrong schema
+    IF vector_schema IS NOT NULL AND vector_schema != current_schema_name THEN
+        BEGIN
+            -- Vector exists in a different schema, drop it first
+            RAISE WARNING 'pgvector found in schema "%" but need it in "%". Dropping and reinstalling...',
+                vector_schema, current_schema_name;
+            EXECUTE 'DROP EXTENSION IF EXISTS vector CASCADE';
+        EXCEPTION WHEN OTHERS THEN
+            RAISE EXCEPTION 'Failed to drop pgvector from schema "%": %. You may need to drop it manually.',
+                vector_schema, SQLERRM;
+        END;
+    END IF;
+
+    -- Create extension in current schema (let it fail naturally if not available)
+    EXECUTE format('CREATE EXTENSION IF NOT EXISTS vector SCHEMA %I', current_schema_name);
 END $$;
 
 -- CreateEnum
@@ -19,7 +45,7 @@ CREATE TABLE "UnifiedContentEmbedding" (
     "contentType" "ContentType" NOT NULL,
     "contentId" TEXT NOT NULL,
     "userId" TEXT,
-    "embedding" public.vector(1536) NOT NULL,
+    "embedding" vector(1536) NOT NULL,
     "searchableText" TEXT NOT NULL,
     "metadata" JSONB NOT NULL DEFAULT '{}',
 
@@ -45,4 +71,4 @@ CREATE UNIQUE INDEX "UnifiedContentEmbedding_contentType_contentId_userId_key" O
 -- Uses cosine distance operator (<=>), which matches the query in hybrid_search.py
 -- Note: Drop first in case Prisma created a btree index (Prisma doesn't support HNSW)
 DROP INDEX IF EXISTS "UnifiedContentEmbedding_embedding_idx";
-CREATE INDEX "UnifiedContentEmbedding_embedding_idx" ON "UnifiedContentEmbedding" USING hnsw ("embedding" public.vector_cosine_ops);
+CREATE INDEX "UnifiedContentEmbedding_embedding_idx" ON "UnifiedContentEmbedding" USING hnsw ("embedding" vector_cosine_ops);

autogpt_platform/backend/migrations/20260112173500_add_supabase_extensions_to_platform_schema/migration.sql

@@ -1,71 +0,0 @@
--- Acknowledge Supabase-managed extensions to prevent drift warnings
--- These extensions are pre-installed by Supabase in specific schemas
--- This migration ensures they exist where available (Supabase) or skips gracefully (CI)
-
--- Create schemas (safe in both CI and Supabase)
-CREATE SCHEMA IF NOT EXISTS "extensions";
-
--- Extensions that exist in both CI and Supabase
-DO $$
-BEGIN
-    CREATE EXTENSION IF NOT EXISTS "pgcrypto" WITH SCHEMA "extensions";
-EXCEPTION WHEN OTHERS THEN
-    RAISE NOTICE 'pgcrypto extension not available, skipping';
-END $$;
-
-DO $$
-BEGIN
-    CREATE EXTENSION IF NOT EXISTS "uuid-ossp" WITH SCHEMA "extensions";
-EXCEPTION WHEN OTHERS THEN
-    RAISE NOTICE 'uuid-ossp extension not available, skipping';
-END $$;
-
--- Supabase-specific extensions (skip gracefully in CI)
-DO $$
-BEGIN
-    CREATE EXTENSION IF NOT EXISTS "pg_stat_statements" WITH SCHEMA "extensions";
-EXCEPTION WHEN OTHERS THEN
-    RAISE NOTICE 'pg_stat_statements extension not available, skipping';
-END $$;
-
-DO $$
-BEGIN
-    CREATE EXTENSION IF NOT EXISTS "pg_net" WITH SCHEMA "extensions";
-EXCEPTION WHEN OTHERS THEN
-    RAISE NOTICE 'pg_net extension not available, skipping';
-END $$;
-
-DO $$
-BEGIN
-    CREATE EXTENSION IF NOT EXISTS "pgjwt" WITH SCHEMA "extensions";
-EXCEPTION WHEN OTHERS THEN
-    RAISE NOTICE 'pgjwt extension not available, skipping';
-END $$;
-
-DO $$
-BEGIN
-    CREATE SCHEMA IF NOT EXISTS "graphql";
-    CREATE EXTENSION IF NOT EXISTS "pg_graphql" WITH SCHEMA "graphql";
-EXCEPTION WHEN OTHERS THEN
-    RAISE NOTICE 'pg_graphql extension not available, skipping';
-END $$;
-
-DO $$
-BEGIN
-    CREATE SCHEMA IF NOT EXISTS "pgsodium";
-    CREATE EXTENSION IF NOT EXISTS "pgsodium" WITH SCHEMA "pgsodium";
-EXCEPTION WHEN OTHERS THEN
-    RAISE NOTICE 'pgsodium extension not available, skipping';
-END $$;
-
-DO $$
-BEGIN
-    CREATE SCHEMA IF NOT EXISTS "vault";
-    CREATE EXTENSION IF NOT EXISTS "supabase_vault" WITH SCHEMA "vault";
-EXCEPTION WHEN OTHERS THEN
-    RAISE NOTICE 'supabase_vault extension not available, skipping';
-END $$;
-
-
--- Return to platform
-CREATE SCHEMA IF NOT EXISTS "platform";

autogpt_platform/backend/scripts/generate_block_docs.py

@@ -0,0 +1,998 @@
+#!/usr/bin/env python3
+"""
+Block Documentation Generator
+
+Generates markdown documentation for all blocks from code introspection.
+Preserves manually-written content between marker comments.
+
+Usage:
+    # Generate all docs
+    poetry run python scripts/generate_block_docs.py
+
+    # Check mode for CI (exits 1 if stale)
+    poetry run python scripts/generate_block_docs.py --check
+
+    # Verbose output
+    poetry run python scripts/generate_block_docs.py -v
+"""
+
+import argparse
+import inspect
+import logging
+import re
+import sys
+from collections import defaultdict
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any
+
+# Add backend to path for imports
+backend_dir = Path(__file__).parent.parent
+sys.path.insert(0, str(backend_dir))
+
+logger = logging.getLogger(__name__)
+
+# Default output directory relative to repo root
+DEFAULT_OUTPUT_DIR = (
+    Path(__file__).parent.parent.parent.parent
+    / "docs"
+    / "integrations"
+    / "block-integrations"
+)
+
+
+@dataclass
+class FieldDoc:
+    """Documentation for a single input/output field."""
+
+    name: str
+    description: str
+    type_str: str
+    required: bool
+    default: Any = None
+    advanced: bool = False
+    hidden: bool = False
+    placeholder: str | None = None
+
+
+@dataclass
+class BlockDoc:
+    """Documentation data extracted from a block."""
+
+    id: str
+    name: str
+    class_name: str
+    description: str
+    categories: list[str]
+    category_descriptions: dict[str, str]
+    inputs: list[FieldDoc]
+    outputs: list[FieldDoc]
+    block_type: str
+    source_file: str
+    contributors: list[str] = field(default_factory=list)
+
+
+# Category to human-readable name mapping
+CATEGORY_DISPLAY_NAMES = {
+    "AI": "AI and Language Models",
+    "BASIC": "Basic Operations",
+    "TEXT": "Text Processing",
+    "SEARCH": "Search and Information Retrieval",
+    "SOCIAL": "Social Media and Content",
+    "DEVELOPER_TOOLS": "Developer Tools",
+    "DATA": "Data Processing",
+    "LOGIC": "Logic and Control Flow",
+    "COMMUNICATION": "Communication",
+    "INPUT": "Input/Output",
+    "OUTPUT": "Input/Output",
+    "MULTIMEDIA": "Media Generation",
+    "PRODUCTIVITY": "Productivity",
+    "HARDWARE": "Hardware",
+    "AGENT": "Agent Integration",
+    "CRM": "CRM Services",
+    "SAFETY": "AI Safety",
+    "ISSUE_TRACKING": "Issue Tracking",
+    "MARKETING": "Marketing",
+}
+
+# Category to doc file mapping (for grouping related blocks)
+CATEGORY_FILE_MAP = {
+    "BASIC": "basic",
+    "TEXT": "text",
+    "AI": "llm",
+    "SEARCH": "search",
+    "DATA": "data",
+    "LOGIC": "logic",
+    "COMMUNICATION": "communication",
+    "MULTIMEDIA": "multimedia",
+    "PRODUCTIVITY": "productivity",
+}
+
+
+def class_name_to_display_name(class_name: str) -> str:
+    """Convert BlockClassName to 'Block Class Name'."""
+    # Remove 'Block' suffix (only at the end, not all occurrences)
+    name = class_name.removesuffix("Block")
+    # Insert space before capitals
+    name = re.sub(r"([a-z])([A-Z])", r"\1 \2", name)
+    # Handle consecutive capitals (e.g., 'HTTPRequest' -> 'HTTP Request')
+    name = re.sub(r"([A-Z]+)([A-Z][a-z])", r"\1 \2", name)
+    return name.strip()
+
+
+def type_to_readable(type_schema: dict[str, Any] | Any) -> str:
+    """Convert JSON schema type to human-readable string."""
+    if not isinstance(type_schema, dict):
+        return str(type_schema) if type_schema else "Any"
+
+    if "anyOf" in type_schema:
+        # Union type - show options
+        any_of = type_schema["anyOf"]
+        if not isinstance(any_of, list):
+            return "Any"
+        options = []
+        for opt in any_of:
+            if isinstance(opt, dict) and opt.get("type") == "null":
+                continue
+            options.append(type_to_readable(opt))
+        if not options:
+            return "None"
+        if len(options) == 1:
+            return options[0]
+        return " | ".join(options)
+
+    if "allOf" in type_schema:
+        all_of = type_schema["allOf"]
+        if not isinstance(all_of, list) or not all_of:
+            return "Any"
+        return type_to_readable(all_of[0])
+
+    schema_type = type_schema.get("type")
+
+    if schema_type == "array":
+        items = type_schema.get("items", {})
+        item_type = type_to_readable(items)
+        return f"List[{item_type}]"
+
+    if schema_type == "object":
+        if "additionalProperties" in type_schema:
+            additional_props = type_schema["additionalProperties"]
+            # additionalProperties: true means any value type is allowed
+            if additional_props is True:
+                return "Dict[str, Any]"
+            value_type = type_to_readable(additional_props)
+            return f"Dict[str, {value_type}]"
+        # Check if it's a specific model
+        title = type_schema.get("title", "Object")
+        return title
+
+    if schema_type == "string":
+        if "enum" in type_schema:
+            return " | ".join(f'"{v}"' for v in type_schema["enum"])
+        if "format" in type_schema:
+            return f"str ({type_schema['format']})"
+        return "str"
+
+    if schema_type == "integer":
+        return "int"
+
+    if schema_type == "number":
+        return "float"
+
+    if schema_type == "boolean":
+        return "bool"
+
+    if schema_type == "null":
+        return "None"
+
+    # Fallback
+    return type_schema.get("title", schema_type or "Any")
+
+
+def safe_get(d: Any, key: str, default: Any = None) -> Any:
+    """Safely get a value from a dict-like object."""
+    if isinstance(d, dict):
+        return d.get(key, default)
+    return default
+
+
+def file_path_to_title(file_path: str) -> str:
+    """Convert file path to a readable title.
+
+    Examples:
+        "github/issues.md" -> "GitHub Issues"
+        "basic.md" -> "Basic"
+        "llm.md" -> "LLM"
+        "google/sheets.md" -> "Google Sheets"
+    """
+    # Special case replacements (applied after title casing)
+    TITLE_FIXES = {
+        "Llm": "LLM",
+        "Github": "GitHub",
+        "Api": "API",
+        "Ai": "AI",
+        "Oauth": "OAuth",
+        "Url": "URL",
+        "Ci": "CI",
+        "Pr": "PR",
+        "Gmb": "GMB",  # Google My Business
+        "Hubspot": "HubSpot",
+        "Linkedin": "LinkedIn",
+        "Tiktok": "TikTok",
+        "Youtube": "YouTube",
+    }
+
+    def apply_fixes(text: str) -> str:
+        # Split into words, fix each word, rejoin
+        words = text.split()
+        fixed_words = [TITLE_FIXES.get(word, word) for word in words]
+        return " ".join(fixed_words)
+
+    path = Path(file_path)
+    name = path.stem  # e.g., "issues" or "sheets"
+
+    # Get parent dir if exists
+    parent = path.parent.name if path.parent.name != "." else None
+
+    # Title case and apply fixes
+    if parent:
+        parent_title = apply_fixes(parent.replace("_", " ").title())
+        name_title = apply_fixes(name.replace("_", " ").title())
+        return f"{parent_title} {name_title}"
+    return apply_fixes(name.replace("_", " ").title())
+
+
+def extract_block_doc(block_cls: type) -> BlockDoc:
+    """Extract documentation data from a block class."""
+    block = block_cls.create()
+
+    # Get source file
+    try:
+        source_file = inspect.getfile(block_cls)
+        # Make relative to blocks directory
+        blocks_dir = Path(source_file).parent
+        while blocks_dir.name != "blocks" and blocks_dir.parent != blocks_dir:
+            blocks_dir = blocks_dir.parent
+        source_file = str(Path(source_file).relative_to(blocks_dir.parent))
+    except (TypeError, ValueError):
+        source_file = "unknown"
+
+    # Extract input fields
+    input_schema = block.input_schema.jsonschema()
+    input_properties = safe_get(input_schema, "properties", {})
+    if not isinstance(input_properties, dict):
+        input_properties = {}
+    required_raw = safe_get(input_schema, "required", [])
+    # Handle edge cases where required might not be a list
+    if isinstance(required_raw, (list, set, tuple)):
+        required_inputs = set(required_raw)
+    else:
+        required_inputs = set()
+
+    inputs = []
+    for field_name, field_schema in input_properties.items():
+        if not isinstance(field_schema, dict):
+            continue
+        # Skip credentials fields in docs (they're auto-handled)
+        if "credentials" in field_name.lower():
+            continue
+
+        inputs.append(
+            FieldDoc(
+                name=field_name,
+                description=safe_get(field_schema, "description", ""),
+                type_str=type_to_readable(field_schema),
+                required=field_name in required_inputs,
+                default=safe_get(field_schema, "default"),
+                advanced=safe_get(field_schema, "advanced", False) or False,
+                hidden=safe_get(field_schema, "hidden", False) or False,
+                placeholder=safe_get(field_schema, "placeholder"),
+            )
+        )
+
+    # Extract output fields
+    output_schema = block.output_schema.jsonschema()
+    output_properties = safe_get(output_schema, "properties", {})
+    if not isinstance(output_properties, dict):
+        output_properties = {}
+
+    outputs = []
+    for field_name, field_schema in output_properties.items():
+        if not isinstance(field_schema, dict):
+            continue
+        outputs.append(
+            FieldDoc(
+                name=field_name,
+                description=safe_get(field_schema, "description", ""),
+                type_str=type_to_readable(field_schema),
+                required=True,  # Outputs are always produced
+                hidden=safe_get(field_schema, "hidden", False) or False,
+            )
+        )
+
+    # Get category info (sort for deterministic ordering since it's a set)
+    categories = []
+    category_descriptions = {}
+    for cat in sorted(block.categories, key=lambda c: c.name):
+        categories.append(cat.name)
+        category_descriptions[cat.name] = cat.value
+
+    # Get contributors
+    contributors = []
+    for contrib in block.contributors:
+        contributors.append(contrib.name if hasattr(contrib, "name") else str(contrib))
+
+    return BlockDoc(
+        id=block.id,
+        name=class_name_to_display_name(block.name),
+        class_name=block.name,
+        description=block.description,
+        categories=categories,
+        category_descriptions=category_descriptions,
+        inputs=inputs,
+        outputs=outputs,
+        block_type=block.block_type.value,
+        source_file=source_file,
+        contributors=contributors,
+    )
+
+
+def generate_anchor(name: str) -> str:
+    """Generate markdown anchor from block name."""
+    return name.lower().replace(" ", "-").replace("(", "").replace(")", "")
+
+
+def extract_manual_content(existing_content: str) -> dict[str, str]:
+    """Extract content between MANUAL markers from existing file."""
+    manual_sections = {}
+
+    # Pattern: <!-- MANUAL: section_name -->content<!-- END MANUAL -->
+    pattern = r"<!-- MANUAL: (\w+) -->\s*(.*?)\s*<!-- END MANUAL -->"
+    matches = re.findall(pattern, existing_content, re.DOTALL)
+
+    for section_name, content in matches:
+        manual_sections[section_name] = content.strip()
+
+    return manual_sections
+
+
+def generate_block_markdown(
+    block: BlockDoc,
+    manual_content: dict[str, str] | None = None,
+) -> str:
+    """Generate markdown documentation for a single block."""
+    manual_content = manual_content or {}
+    lines = []
+
+    # All blocks use ## heading, sections use ### (consistent siblings)
+    lines.append(f"## {block.name}")
+    lines.append("")
+
+    # What it is (full description)
+    lines.append("### What it is")
+    lines.append(block.description or "No description available.")
+    lines.append("")
+
+    # How it works (manual section)
+    lines.append("### How it works")
+    how_it_works = manual_content.get(
+        "how_it_works", "_Add technical explanation here._"
+    )
+    lines.append("<!-- MANUAL: how_it_works -->")
+    lines.append(how_it_works)
+    lines.append("<!-- END MANUAL -->")
+    lines.append("")
+
+    # Inputs table (auto-generated)
+    visible_inputs = [f for f in block.inputs if not f.hidden]
+    if visible_inputs:
+        lines.append("### Inputs")
+        lines.append("")
+        lines.append("| Input | Description | Type | Required |")
+        lines.append("|-------|-------------|------|----------|")
+        for inp in visible_inputs:
+            required = "Yes" if inp.required else "No"
+            desc = inp.description or "-"
+            type_str = inp.type_str or "-"
+            # Normalize newlines and escape pipes for valid table syntax
+            desc = desc.replace("\n", " ").replace("|", "\\|")
+            type_str = type_str.replace("|", "\\|")
+            lines.append(f"| {inp.name} | {desc} | {type_str} | {required} |")
+        lines.append("")
+
+    # Outputs table (auto-generated)
+    visible_outputs = [f for f in block.outputs if not f.hidden]
+    if visible_outputs:
+        lines.append("### Outputs")
+        lines.append("")
+        lines.append("| Output | Description | Type |")
+        lines.append("|--------|-------------|------|")
+        for out in visible_outputs:
+            desc = out.description or "-"
+            type_str = out.type_str or "-"
+            # Normalize newlines and escape pipes for valid table syntax
+            desc = desc.replace("\n", " ").replace("|", "\\|")
+            type_str = type_str.replace("|", "\\|")
+            lines.append(f"| {out.name} | {desc} | {type_str} |")
+        lines.append("")
+
+    # Possible use case (manual section)
+    lines.append("### Possible use case")
+    use_case = manual_content.get("use_case", "_Add practical use case examples here._")
+    lines.append("<!-- MANUAL: use_case -->")
+    lines.append(use_case)
+    lines.append("<!-- END MANUAL -->")
+    lines.append("")
+
+    # Optional per-block extras (only include if has content)
+    extras = manual_content.get("extras", "")
+    if extras:
+        lines.append("<!-- MANUAL: extras -->")
+        lines.append(extras)
+        lines.append("<!-- END MANUAL -->")
+        lines.append("")
+
+    lines.append("---")
+    lines.append("")
+
+    return "\n".join(lines)
+
+
+def get_block_file_mapping(blocks: list[BlockDoc]) -> dict[str, list[BlockDoc]]:
+    """
+    Map blocks to their documentation files.
+
+    Returns dict of {relative_file_path: [blocks]}
+    """
+    file_mapping = defaultdict(list)
+
+    for block in blocks:
+        # Determine file path based on source file or category
+        source_path = Path(block.source_file)
+
+        # If source is in a subdirectory (e.g., google/gmail.py), use that structure
+        if len(source_path.parts) > 2:  # blocks/subdir/file.py
+            subdir = source_path.parts[1]  # e.g., "google"
+            # Use the Python filename as the md filename
+            md_file = source_path.stem + ".md"  # e.g., "gmail.md"
+            file_path = f"{subdir}/{md_file}"
+        else:
+            # Use category-based grouping for top-level blocks
+            primary_category = block.categories[0] if block.categories else "BASIC"
+            file_name = CATEGORY_FILE_MAP.get(primary_category, "misc")
+            file_path = f"{file_name}.md"
+
+        file_mapping[file_path].append(block)
+
+    return dict(file_mapping)
+
+
+def generate_overview_table(blocks: list[BlockDoc], block_dir_prefix: str = "") -> str:
+    """Generate the overview table markdown (blocks.md).
+
+    Args:
+        blocks: List of block documentation objects
+        block_dir_prefix: Prefix for block file links (e.g., "block-integrations/")
+    """
+    lines = []
+
+    # GitBook YAML frontmatter
+    lines.append("---")
+    lines.append("layout:")
+    lines.append("  width: default")
+    lines.append("  title:")
+    lines.append("    visible: true")
+    lines.append("  description:")
+    lines.append("    visible: true")
+    lines.append("  tableOfContents:")
+    lines.append("    visible: false")
+    lines.append("  outline:")
+    lines.append("    visible: true")
+    lines.append("  pagination:")
+    lines.append("    visible: true")
+    lines.append("  metadata:")
+    lines.append("    visible: true")
+    lines.append("---")
+    lines.append("")
+
+    lines.append("# AutoGPT Blocks Overview")
+    lines.append("")
+    lines.append(
+        'AutoGPT uses a modular approach with various "blocks" to handle different tasks. These blocks are the building blocks of AutoGPT workflows, allowing users to create complex automations by combining simple, specialized components.'
+    )
+    lines.append("")
+    lines.append('{% hint style="info" %}')
+    lines.append("**Creating Your Own Blocks**")
+    lines.append("")
+    lines.append("Want to create your own custom blocks? Check out our guides:")
+    lines.append("")
+    lines.append(
+        "* [Build your own Blocks](https://docs.agpt.co/platform/new_blocks/) - Step-by-step tutorial with examples"
+    )
+    lines.append(
+        "* [Block SDK Guide](https://docs.agpt.co/platform/block-sdk-guide/) - Advanced SDK patterns with OAuth, webhooks, and provider configuration"
+    )
+    lines.append("{% endhint %}")
+    lines.append("")
+    lines.append(
+        "Below is a comprehensive list of all available blocks, categorized by their primary function. Click on any block name to view its detailed documentation."
+    )
+    lines.append("")
+
+    # Group blocks by category
+    by_category = defaultdict(list)
+    for block in blocks:
+        primary_cat = block.categories[0] if block.categories else "BASIC"
+        by_category[primary_cat].append(block)
+
+    # Sort categories
+    category_order = [
+        "BASIC",
+        "DATA",
+        "TEXT",
+        "AI",
+        "SEARCH",
+        "SOCIAL",
+        "COMMUNICATION",
+        "DEVELOPER_TOOLS",
+        "MULTIMEDIA",
+        "PRODUCTIVITY",
+        "LOGIC",
+        "INPUT",
+        "OUTPUT",
+        "AGENT",
+        "CRM",
+        "SAFETY",
+        "ISSUE_TRACKING",
+        "HARDWARE",
+        "MARKETING",
+    ]
+
+    # Track emitted display names to avoid duplicate headers
+    # (e.g., INPUT and OUTPUT both map to "Input/Output")
+    emitted_display_names: set[str] = set()
+
+    for category in category_order:
+        if category not in by_category:
+            continue
+
+        display_name = CATEGORY_DISPLAY_NAMES.get(category, category)
+
+        # Collect all blocks for this display name (may span multiple categories)
+        if display_name in emitted_display_names:
+            # Already emitted header, just add rows to existing table
+            # Find the position before the last empty line and insert rows
+            cat_blocks = sorted(by_category[category], key=lambda b: b.name)
+            # Remove the trailing empty line, add rows, then re-add empty line
+            lines.pop()
+            for block in cat_blocks:
+                file_mapping = get_block_file_mapping([block])
+                file_path = list(file_mapping.keys())[0]
+                anchor = generate_anchor(block.name)
+                short_desc = (
+                    block.description.split(".")[0]
+                    if block.description
+                    else "No description"
+                )
+                short_desc = short_desc.replace("\n", " ").replace("|", "\\|")
+                link_path = f"{block_dir_prefix}{file_path}"
+                lines.append(f"| [{block.name}]({link_path}#{anchor}) | {short_desc} |")
+            lines.append("")
+            continue
+
+        emitted_display_names.add(display_name)
+        cat_blocks = sorted(by_category[category], key=lambda b: b.name)
+
+        lines.append(f"## {display_name}")
+        lines.append("")
+        lines.append("| Block Name | Description |")
+        lines.append("|------------|-------------|")
+
+        for block in cat_blocks:
+            # Determine link path
+            file_mapping = get_block_file_mapping([block])
+            file_path = list(file_mapping.keys())[0]
+            anchor = generate_anchor(block.name)
+
+            # Short description (first sentence)
+            short_desc = (
+                block.description.split(".")[0]
+                if block.description
+                else "No description"
+            )
+            short_desc = short_desc.replace("\n", " ").replace("|", "\\|")
+
+            link_path = f"{block_dir_prefix}{file_path}"
+            lines.append(f"| [{block.name}]({link_path}#{anchor}) | {short_desc} |")
+
+        lines.append("")
+
+    return "\n".join(lines)
+
+
+def generate_summary_md(
+    blocks: list[BlockDoc], root_dir: Path, block_dir_prefix: str = ""
+) -> str:
+    """Generate SUMMARY.md for GitBook navigation.
+
+    Args:
+        blocks: List of block documentation objects
+        root_dir: The root docs directory (e.g., docs/integrations/)
+        block_dir_prefix: Prefix for block file links (e.g., "block-integrations/")
+    """
+    lines = []
+    lines.append("# Table of contents")
+    lines.append("")
+    lines.append("* [AutoGPT Blocks Overview](README.md)")
+    lines.append("")
+
+    # Check for guides/ directory at the root level (docs/integrations/guides/)
+    guides_dir = root_dir / "guides"
+    if guides_dir.exists():
+        lines.append("## Guides")
+        lines.append("")
+        for guide_file in sorted(guides_dir.glob("*.md")):
+            # Use just the file name for title (replace hyphens/underscores with spaces)
+            title = file_path_to_title(guide_file.stem.replace("-", "_") + ".md")
+            lines.append(f"* [{title}](guides/{guide_file.name})")
+        lines.append("")
+
+    lines.append("## Block Integrations")
+    lines.append("")
+
+    file_mapping = get_block_file_mapping(blocks)
+    for file_path in sorted(file_mapping.keys()):
+        title = file_path_to_title(file_path)
+        link_path = f"{block_dir_prefix}{file_path}"
+        lines.append(f"* [{title}]({link_path})")
+
+    lines.append("")
+
+    return "\n".join(lines)
+
+
+def load_all_blocks_for_docs() -> list[BlockDoc]:
+    """Load all blocks and extract documentation."""
+    from backend.blocks import load_all_blocks
+
+    block_classes = load_all_blocks()
+    blocks = []
+
+    for _block_id, block_cls in block_classes.items():
+        try:
+            block_doc = extract_block_doc(block_cls)
+            blocks.append(block_doc)
+        except Exception as e:
+            logger.warning(f"Failed to extract docs for {block_cls.__name__}: {e}")
+
+    return blocks
+
+
+def write_block_docs(
+    output_dir: Path,
+    blocks: list[BlockDoc],
+    verbose: bool = False,
+) -> dict[str, str]:
+    """
+    Write block documentation files.
+
+    Returns dict of {file_path: content} for all generated files.
+    """
+    output_dir = Path(output_dir)
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    file_mapping = get_block_file_mapping(blocks)
+    generated_files = {}
+
+    for file_path, file_blocks in file_mapping.items():
+        full_path = output_dir / file_path
+
+        # Create subdirectories if needed
+        full_path.parent.mkdir(parents=True, exist_ok=True)
+
+        # Load existing content for manual section preservation
+        existing_content = ""
+        if full_path.exists():
+            existing_content = full_path.read_text()
+
+        # Always generate title from file path (with fixes applied)
+        file_title = file_path_to_title(file_path)
+
+        # Extract existing file description if present (preserve manual content)
+        file_header_pattern = (
+            r"^# .+?\n<!-- MANUAL: file_description -->\n(.*?)\n<!-- END MANUAL -->"
+        )
+        file_header_match = re.search(file_header_pattern, existing_content, re.DOTALL)
+
+        if file_header_match:
+            file_description = file_header_match.group(1)
+        else:
+            file_description = "_Add a description of this category of blocks._"
+
+        # Generate file header
+        file_header = f"# {file_title}\n"
+        file_header += "<!-- MANUAL: file_description -->\n"
+        file_header += f"{file_description}\n"
+        file_header += "<!-- END MANUAL -->\n"
+
+        # Generate content for each block
+        content_parts = []
+        for block in sorted(file_blocks, key=lambda b: b.name):
+            # Extract manual content specific to this block
+            # Match block heading (h2) and capture until --- separator
+            block_pattern = rf"(?:^|\n)## {re.escape(block.name)}\s*\n(.*?)(?=\n---|\Z)"
+            block_match = re.search(block_pattern, existing_content, re.DOTALL)
+            if block_match:
+                manual_content = extract_manual_content(block_match.group(1))
+            else:
+                manual_content = {}
+
+            content_parts.append(
+                generate_block_markdown(
+                    block,
+                    manual_content,
+                )
+            )
+
+        # Add file-level additional_content section if present
+        file_additional = extract_manual_content(existing_content).get(
+            "additional_content", ""
+        )
+        if file_additional:
+            content_parts.append("<!-- MANUAL: additional_content -->")
+            content_parts.append(file_additional)
+            content_parts.append("<!-- END MANUAL -->")
+            content_parts.append("")
+
+        full_content = file_header + "\n" + "\n".join(content_parts)
+        generated_files[str(file_path)] = full_content
+
+        if verbose:
+            print(f"  Writing {file_path} ({len(file_blocks)} blocks)")
+
+        full_path.write_text(full_content)
+
+    # Generate overview file at the parent directory (docs/integrations/)
+    # with links prefixed to point into block-integrations/
+    root_dir = output_dir.parent
+    block_dir_name = output_dir.name  # "block-integrations"
+    block_dir_prefix = f"{block_dir_name}/"
+
+    overview_content = generate_overview_table(blocks, block_dir_prefix)
+    overview_path = root_dir / "README.md"
+    generated_files["README.md"] = overview_content
+    overview_path.write_text(overview_content)
+
+    if verbose:
+        print("  Writing README.md (overview) to parent directory")
+
+    # Generate SUMMARY.md for GitBook navigation at the parent directory
+    summary_content = generate_summary_md(blocks, root_dir, block_dir_prefix)
+    summary_path = root_dir / "SUMMARY.md"
+    generated_files["SUMMARY.md"] = summary_content
+    summary_path.write_text(summary_content)
+
+    if verbose:
+        print("  Writing SUMMARY.md (navigation) to parent directory")
+
+    return generated_files
+
+
+def check_docs_in_sync(output_dir: Path, blocks: list[BlockDoc]) -> bool:
+    """
+    Check if generated docs match existing docs.
+
+    Returns True if in sync, False otherwise.
+    """
+    output_dir = Path(output_dir)
+    file_mapping = get_block_file_mapping(blocks)
+
+    all_match = True
+    out_of_sync_details: list[tuple[str, list[str]]] = []
+
+    for file_path, file_blocks in file_mapping.items():
+        full_path = output_dir / file_path
+
+        if not full_path.exists():
+            block_names = [b.name for b in sorted(file_blocks, key=lambda b: b.name)]
+            print(f"MISSING: {file_path}")
+            print(f"  Blocks: {', '.join(block_names)}")
+            out_of_sync_details.append((file_path, block_names))
+            all_match = False
+            continue
+
+        existing_content = full_path.read_text()
+
+        # Always generate title from file path (with fixes applied)
+        file_title = file_path_to_title(file_path)
+
+        # Extract existing file description if present (preserve manual content)
+        file_header_pattern = (
+            r"^# .+?\n<!-- MANUAL: file_description -->\n(.*?)\n<!-- END MANUAL -->"
+        )
+        file_header_match = re.search(file_header_pattern, existing_content, re.DOTALL)
+
+        if file_header_match:
+            file_description = file_header_match.group(1)
+        else:
+            file_description = "_Add a description of this category of blocks._"
+
+        # Generate expected file header
+        file_header = f"# {file_title}\n"
+        file_header += "<!-- MANUAL: file_description -->\n"
+        file_header += f"{file_description}\n"
+        file_header += "<!-- END MANUAL -->\n"
+
+        # Extract manual content from existing file
+        manual_sections_by_block = {}
+        for block in file_blocks:
+            block_pattern = rf"(?:^|\n)## {re.escape(block.name)}\s*\n(.*?)(?=\n---|\Z)"
+            block_match = re.search(block_pattern, existing_content, re.DOTALL)
+            if block_match:
+                manual_sections_by_block[block.name] = extract_manual_content(
+                    block_match.group(1)
+                )
+
+        # Generate expected content and check each block individually
+        content_parts = []
+        mismatched_blocks = []
+        for block in sorted(file_blocks, key=lambda b: b.name):
+            manual_content = manual_sections_by_block.get(block.name, {})
+            expected_block_content = generate_block_markdown(
+                block,
+                manual_content,
+            )
+            content_parts.append(expected_block_content)
+
+            # Check if this specific block's section exists and matches
+            # Include the --- separator to match generate_block_markdown output
+            block_pattern = rf"(?:^|\n)(## {re.escape(block.name)}\s*\n.*?\n---\n)"
+            block_match = re.search(block_pattern, existing_content, re.DOTALL)
+            if not block_match:
+                mismatched_blocks.append(f"{block.name} (missing)")
+            elif block_match.group(1).strip() != expected_block_content.strip():
+                mismatched_blocks.append(block.name)
+
+        # Add file-level additional_content to expected content (matches write_block_docs)
+        file_additional = extract_manual_content(existing_content).get(
+            "additional_content", ""
+        )
+        if file_additional:
+            content_parts.append("<!-- MANUAL: additional_content -->")
+            content_parts.append(file_additional)
+            content_parts.append("<!-- END MANUAL -->")
+            content_parts.append("")
+
+        expected_content = file_header + "\n" + "\n".join(content_parts)
+
+        if existing_content.strip() != expected_content.strip():
+            print(f"OUT OF SYNC: {file_path}")
+            if mismatched_blocks:
+                print(f"  Affected blocks: {', '.join(mismatched_blocks)}")
+            out_of_sync_details.append((file_path, mismatched_blocks))
+            all_match = False
+
+    # Check overview at the parent directory (docs/integrations/)
+    root_dir = output_dir.parent
+    block_dir_name = output_dir.name  # "block-integrations"
+    block_dir_prefix = f"{block_dir_name}/"
+
+    overview_path = root_dir / "README.md"
+    if overview_path.exists():
+        existing_overview = overview_path.read_text()
+        expected_overview = generate_overview_table(blocks, block_dir_prefix)
+        if existing_overview.strip() != expected_overview.strip():
+            print("OUT OF SYNC: README.md (overview)")
+            print("  The blocks overview table needs regeneration")
+            out_of_sync_details.append(("README.md", ["overview table"]))
+            all_match = False
+    else:
+        print("MISSING: README.md (overview)")
+        out_of_sync_details.append(("README.md", ["overview table"]))
+        all_match = False
+
+    # Check SUMMARY.md at the parent directory
+    summary_path = root_dir / "SUMMARY.md"
+    if summary_path.exists():
+        existing_summary = summary_path.read_text()
+        expected_summary = generate_summary_md(blocks, root_dir, block_dir_prefix)
+        if existing_summary.strip() != expected_summary.strip():
+            print("OUT OF SYNC: SUMMARY.md (navigation)")
+            print("  The GitBook navigation needs regeneration")
+            out_of_sync_details.append(("SUMMARY.md", ["navigation"]))
+            all_match = False
+    else:
+        print("MISSING: SUMMARY.md (navigation)")
+        out_of_sync_details.append(("SUMMARY.md", ["navigation"]))
+        all_match = False
+
+    # Check for unfilled manual sections
+    unfilled_patterns = [
+        "_Add a description of this category of blocks._",
+        "_Add technical explanation here._",
+        "_Add practical use case examples here._",
+    ]
+    files_with_unfilled = []
+    for file_path in file_mapping.keys():
+        full_path = output_dir / file_path
+        if full_path.exists():
+            content = full_path.read_text()
+            unfilled_count = sum(1 for p in unfilled_patterns if p in content)
+            if unfilled_count > 0:
+                files_with_unfilled.append((file_path, unfilled_count))
+
+    if files_with_unfilled:
+        print("\nWARNING: Files with unfilled manual sections:")
+        for file_path, count in sorted(files_with_unfilled):
+            print(f"  {file_path}: {count} unfilled section(s)")
+        print(
+            f"\nTotal: {len(files_with_unfilled)} files with unfilled manual sections"
+        )
+
+    return all_match
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Generate block documentation from code introspection"
+    )
+    parser.add_argument(
+        "--output-dir",
+        type=Path,
+        default=DEFAULT_OUTPUT_DIR,
+        help="Output directory for generated docs",
+    )
+    parser.add_argument(
+        "--check",
+        action="store_true",
+        help="Check if docs are in sync (for CI), exit 1 if not",
+    )
+    parser.add_argument(
+        "-v",
+        "--verbose",
+        action="store_true",
+        help="Verbose output",
+    )
+
+    args = parser.parse_args()
+
+    logging.basicConfig(
+        level=logging.DEBUG if args.verbose else logging.INFO,
+        format="%(levelname)s: %(message)s",
+    )
+
+    print("Loading blocks...")
+    blocks = load_all_blocks_for_docs()
+    print(f"Found {len(blocks)} blocks")
+
+    if args.check:
+        print(f"Checking docs in {args.output_dir}...")
+        in_sync = check_docs_in_sync(args.output_dir, blocks)
+        if in_sync:
+            print("All documentation is in sync!")
+            sys.exit(0)
+        else:
+            print("\n" + "=" * 60)
+            print("Documentation is out of sync!")
+            print("=" * 60)
+            print("\nTo fix this, run one of the following:")
+            print("\n  Option 1 - Run locally:")
+            print(
+                "    cd autogpt_platform/backend && poetry run python scripts/generate_block_docs.py"
+            )
+            print("\n  Option 2 - Ask Claude Code to run it:")
+            print('    "Run the block docs generator script to sync documentation"')
+            print("\n" + "=" * 60)
+            sys.exit(1)
+    else:
+        print(f"Generating docs to {args.output_dir}...")
+        write_block_docs(
+            args.output_dir,
+            blocks,
+            verbose=args.verbose,
+        )
+        print("Done!")
+
+
+if __name__ == "__main__":
+    main()

autogpt_platform/backend/scripts/test_generate_block_docs.py

@@ -0,0 +1,208 @@
+#!/usr/bin/env python3
+"""Tests for the block documentation generator."""
+import pytest
+
+from scripts.generate_block_docs import (
+    class_name_to_display_name,
+    extract_manual_content,
+    generate_anchor,
+    type_to_readable,
+)
+
+
+class TestClassNameToDisplayName:
+    """Tests for class_name_to_display_name function."""
+
+    def test_simple_block_name(self):
+        assert class_name_to_display_name("PrintBlock") == "Print"
+
+    def test_multi_word_block_name(self):
+        assert class_name_to_display_name("GetWeatherBlock") == "Get Weather"
+
+    def test_consecutive_capitals(self):
+        assert class_name_to_display_name("HTTPRequestBlock") == "HTTP Request"
+
+    def test_ai_prefix(self):
+        assert class_name_to_display_name("AIConditionBlock") == "AI Condition"
+
+    def test_no_block_suffix(self):
+        assert class_name_to_display_name("SomeClass") == "Some Class"
+
+
+class TestTypeToReadable:
+    """Tests for type_to_readable function."""
+
+    def test_string_type(self):
+        assert type_to_readable({"type": "string"}) == "str"
+
+    def test_integer_type(self):
+        assert type_to_readable({"type": "integer"}) == "int"
+
+    def test_number_type(self):
+        assert type_to_readable({"type": "number"}) == "float"
+
+    def test_boolean_type(self):
+        assert type_to_readable({"type": "boolean"}) == "bool"
+
+    def test_array_type(self):
+        result = type_to_readable({"type": "array", "items": {"type": "string"}})
+        assert result == "List[str]"
+
+    def test_object_type(self):
+        result = type_to_readable({"type": "object", "title": "MyModel"})
+        assert result == "MyModel"
+
+    def test_anyof_with_null(self):
+        result = type_to_readable({"anyOf": [{"type": "string"}, {"type": "null"}]})
+        assert result == "str"
+
+    def test_anyof_multiple_types(self):
+        result = type_to_readable({"anyOf": [{"type": "string"}, {"type": "integer"}]})
+        assert result == "str | int"
+
+    def test_enum_type(self):
+        result = type_to_readable(
+            {"type": "string", "enum": ["option1", "option2", "option3"]}
+        )
+        assert result == '"option1" | "option2" | "option3"'
+
+    def test_none_input(self):
+        assert type_to_readable(None) == "Any"
+
+    def test_non_dict_input(self):
+        assert type_to_readable("string") == "string"
+
+
+class TestExtractManualContent:
+    """Tests for extract_manual_content function."""
+
+    def test_extract_how_it_works(self):
+        content = """
+### How it works
+<!-- MANUAL: how_it_works -->
+This is how it works.
+<!-- END MANUAL -->
+"""
+        result = extract_manual_content(content)
+        assert result == {"how_it_works": "This is how it works."}
+
+    def test_extract_use_case(self):
+        content = """
+### Possible use case
+<!-- MANUAL: use_case -->
+Example use case here.
+<!-- END MANUAL -->
+"""
+        result = extract_manual_content(content)
+        assert result == {"use_case": "Example use case here."}
+
+    def test_extract_multiple_sections(self):
+        content = """
+<!-- MANUAL: how_it_works -->
+How it works content.
+<!-- END MANUAL -->
+
+<!-- MANUAL: use_case -->
+Use case content.
+<!-- END MANUAL -->
+"""
+        result = extract_manual_content(content)
+        assert result == {
+            "how_it_works": "How it works content.",
+            "use_case": "Use case content.",
+        }
+
+    def test_empty_content(self):
+        result = extract_manual_content("")
+        assert result == {}
+
+    def test_no_markers(self):
+        result = extract_manual_content("Some content without markers")
+        assert result == {}
+
+
+class TestGenerateAnchor:
+    """Tests for generate_anchor function."""
+
+    def test_simple_name(self):
+        assert generate_anchor("Print") == "print"
+
+    def test_multi_word_name(self):
+        assert generate_anchor("Get Weather") == "get-weather"
+
+    def test_name_with_parentheses(self):
+        assert generate_anchor("Something (Optional)") == "something-optional"
+
+    def test_already_lowercase(self):
+        assert generate_anchor("already lowercase") == "already-lowercase"
+
+
+class TestIntegration:
+    """Integration tests that require block loading."""
+
+    def test_load_blocks(self):
+        """Test that blocks can be loaded successfully."""
+        import logging
+        import sys
+        from pathlib import Path
+
+        logging.disable(logging.CRITICAL)
+        sys.path.insert(0, str(Path(__file__).parent.parent))
+
+        from scripts.generate_block_docs import load_all_blocks_for_docs
+
+        blocks = load_all_blocks_for_docs()
+        assert len(blocks) > 0, "Should load at least one block"
+
+    def test_block_doc_has_required_fields(self):
+        """Test that extracted block docs have required fields."""
+        import logging
+        import sys
+        from pathlib import Path
+
+        logging.disable(logging.CRITICAL)
+        sys.path.insert(0, str(Path(__file__).parent.parent))
+
+        from scripts.generate_block_docs import load_all_blocks_for_docs
+
+        blocks = load_all_blocks_for_docs()
+        block = blocks[0]
+
+        assert hasattr(block, "id")
+        assert hasattr(block, "name")
+        assert hasattr(block, "description")
+        assert hasattr(block, "categories")
+        assert hasattr(block, "inputs")
+        assert hasattr(block, "outputs")
+
+    def test_file_mapping_is_deterministic(self):
+        """Test that file mapping produces consistent results."""
+        import logging
+        import sys
+        from pathlib import Path
+
+        logging.disable(logging.CRITICAL)
+        sys.path.insert(0, str(Path(__file__).parent.parent))
+
+        from scripts.generate_block_docs import (
+            get_block_file_mapping,
+            load_all_blocks_for_docs,
+        )
+
+        # Load blocks twice and compare mappings
+        blocks1 = load_all_blocks_for_docs()
+        blocks2 = load_all_blocks_for_docs()
+
+        mapping1 = get_block_file_mapping(blocks1)
+        mapping2 = get_block_file_mapping(blocks2)
+
+        # Check same files are generated
+        assert set(mapping1.keys()) == set(mapping2.keys())
+
+        # Check same block counts per file
+        for file_path in mapping1:
+            assert len(mapping1[file_path]) == len(mapping2[file_path])
+
+
+if __name__ == "__main__":
+    pytest.main([__file__, "-v"])

autogpt_platform/backend/snapshots/grph_single

@@ -11,6 +11,7 @@
   "forked_from_version": null,
   "has_external_trigger": false,
   "has_human_in_the_loop": false,
+  "has_sensitive_action": false,
   "id": "graph-123",
   "input_schema": {
     "properties": {},

autogpt_platform/backend/snapshots/grphs_all

@@ -11,6 +11,7 @@
     "forked_from_version": null,
     "has_external_trigger": false,
     "has_human_in_the_loop": false,
+    "has_sensitive_action": false,
     "id": "graph-123",
     "input_schema": {
       "properties": {},

autogpt_platform/backend/snapshots/lib_agts_search

@@ -27,6 +27,8 @@
         "properties": {}
       },
       "has_external_trigger": false,
+      "has_human_in_the_loop": false,
+      "has_sensitive_action": false,
       "trigger_setup_info": null,
       "new_output": false,
       "can_access_graph": true,
@@ -34,7 +36,8 @@
       "is_favorite": false,
       "recommended_schedule_cron": null,
       "settings": {
-        "human_in_the_loop_safe_mode": null
+        "human_in_the_loop_safe_mode": true,
+        "sensitive_action_safe_mode": false
       },
       "marketplace_listing": null
     },
@@ -65,6 +68,8 @@
         "properties": {}
       },
       "has_external_trigger": false,
+      "has_human_in_the_loop": false,
+      "has_sensitive_action": false,
       "trigger_setup_info": null,
       "new_output": false,
       "can_access_graph": false,
@@ -72,7 +77,8 @@
       "is_favorite": false,
       "recommended_schedule_cron": null,
       "settings": {
-        "human_in_the_loop_safe_mode": null
+        "human_in_the_loop_safe_mode": true,
+        "sensitive_action_safe_mode": false
       },
       "marketplace_listing": null
     }

autogpt_platform/frontend/.env.default

@@ -29,4 +29,4 @@ NEXT_PUBLIC_CLOUDFLARE_TURNSTILE_SITE_KEY=
 NEXT_PUBLIC_TURNSTILE=disabled
 
 # PR previews
-NEXT_PUBLIC_PREVIEW_STEALING_DEV=
+NEXT_PUBLIC_PREVIEW_STEALING_DEV=

autogpt_platform/frontend/CONTRIBUTING.md

@@ -175,6 +175,8 @@ While server components and actions are cool and cutting-edge, they introduce a
 
 - Prefer [React Query](https://tanstack.com/query/latest/docs/framework/react/overview) for server state, colocated near consumers (see [state colocation](https://kentcdodds.com/blog/state-colocation-will-make-your-react-app-faster))
 - Co-locate UI state inside components/hooks; keep global state minimal
+- Avoid `useMemo` and `useCallback` unless you have a measured performance issue
+- Do not abuse `useEffect`; prefer state colocation and derive values directly when possible
 
 ### Styling and components
 
@@ -549,9 +551,48 @@ Files:
 Types:
 
 - Prefer `interface` for object shapes
-- Component props should be `interface Props { ... }`
+- Component props should be `interface Props { ... }` (not exported)
+- Only use specific exported names (e.g., `export interface MyComponentProps`) when the interface needs to be used outside the component
+- Keep type definitions inline with the component - do not create separate `types.ts` files unless types are shared across multiple files
 - Use precise types; avoid `any` and unsafe casts
 
+**Props naming examples:**
+
+```tsx
+// ✅ Good - internal props, not exported
+interface Props {
+  title: string;
+  onClose: () => void;
+}
+
+export function Modal({ title, onClose }: Props) {
+  // ...
+}
+
+// ✅ Good - exported when needed externally
+export interface ModalProps {
+  title: string;
+  onClose: () => void;
+}
+
+export function Modal({ title, onClose }: ModalProps) {
+  // ...
+}
+
+// ❌ Bad - unnecessarily specific name for internal use
+interface ModalComponentProps {
+  title: string;
+  onClose: () => void;
+}
+
+// ❌ Bad - separate types.ts file for single component
+// types.ts
+export interface ModalProps { ... }
+
+// Modal.tsx
+import type { ModalProps } from './types';
+```
+
 Parameters:
 
 - If more than one parameter is needed, pass a single `Args` object for clarity

autogpt_platform/frontend/orval.config.ts

@@ -16,6 +16,12 @@ export default defineConfig({
       client: "react-query",
       httpClient: "fetch",
       indexFiles: false,
+      mock: {
+        type: "msw",
+        baseUrl: "http://localhost:3000/api/proxy",
+        generateEachHttpStatus: true,
+        delay: 0,
+      },
       override: {
         mutator: {
           path: "./mutators/custom-mutator.ts",

autogpt_platform/frontend/package.json

@@ -15,6 +15,8 @@
     "types": "tsc --noEmit",
     "test": "NEXT_PUBLIC_PW_TEST=true next build --turbo && playwright test",
     "test-ui": "NEXT_PUBLIC_PW_TEST=true next build --turbo && playwright test --ui",
+    "test:unit": "vitest run",
+    "test:unit:watch": "vitest",
     "test:no-build": "playwright test",
     "gentests": "playwright codegen http://localhost:3000",
     "storybook": "storybook dev -p 6006",
@@ -118,6 +120,7 @@
   },
   "devDependencies": {
     "@chromatic-com/storybook": "4.1.2",
+    "happy-dom": "20.3.4",
     "@opentelemetry/instrumentation": "0.209.0",
     "@playwright/test": "1.56.1",
     "@storybook/addon-a11y": "9.1.5",
@@ -127,6 +130,8 @@
     "@storybook/nextjs": "9.1.5",
     "@tanstack/eslint-plugin-query": "5.91.2",
     "@tanstack/react-query-devtools": "5.90.2",
+    "@testing-library/dom": "10.4.1",
+    "@testing-library/react": "16.3.2",
     "@types/canvas-confetti": "1.9.0",
     "@types/lodash": "4.17.20",
     "@types/negotiator": "0.6.4",
@@ -135,6 +140,7 @@
     "@types/react-dom": "18.3.5",
     "@types/react-modal": "3.16.3",
     "@types/react-window": "1.8.8",
+    "@vitejs/plugin-react": "5.1.2",
     "axe-playwright": "2.2.2",
     "chromatic": "13.3.3",
     "concurrently": "9.2.1",
@@ -153,7 +159,9 @@
     "require-in-the-middle": "8.0.1",
     "storybook": "9.1.5",
     "tailwindcss": "3.4.17",
-    "typescript": "5.9.3"
+    "typescript": "5.9.3",
+    "vite-tsconfig-paths": "6.0.4",
+    "vitest": "4.0.17"
   },
   "msw": {
     "workerDirectory": [

autogpt_platform/frontend/src/app/(no-navbar)/logout/page.tsx

@@ -0,0 +1,58 @@
+"use client";
+
+import { LoadingSpinner } from "@/components/atoms/LoadingSpinner/LoadingSpinner";
+import { Text } from "@/components/atoms/Text/Text";
+import { useToast } from "@/components/molecules/Toast/use-toast";
+import { useSupabase } from "@/lib/supabase/hooks/useSupabase";
+import { useRouter } from "next/navigation";
+import { useEffect, useRef } from "react";
+
+const LOGOUT_REDIRECT_DELAY_MS = 400;
+
+function wait(ms: number): Promise<void> {
+  return new Promise(function resolveAfterDelay(resolve) {
+    setTimeout(resolve, ms);
+  });
+}
+
+export default function LogoutPage() {
+  const { logOut } = useSupabase();
+  const { toast } = useToast();
+  const router = useRouter();
+  const hasStartedRef = useRef(false);
+
+  useEffect(
+    function handleLogoutEffect() {
+      if (hasStartedRef.current) return;
+      hasStartedRef.current = true;
+
+      async function runLogout() {
+        try {
+          await logOut();
+        } catch {
+          toast({
+            title: "Failed to log out. Redirecting to login.",
+            variant: "destructive",
+          });
+        } finally {
+          await wait(LOGOUT_REDIRECT_DELAY_MS);
+          router.replace("/login");
+        }
+      }
+
+      void runLogout();
+    },
+    [logOut, router, toast],
+  );
+
+  return (
+    <div className="flex min-h-screen items-center justify-center px-4">
+      <div className="flex flex-col items-center justify-center gap-4 py-8">
+        <LoadingSpinner size="large" />
+        <Text variant="body" className="text-center">
+          Logging you out...
+        </Text>
+      </div>
+    </div>
+  );
+}

autogpt_platform/frontend/src/app/(platform)/auth/callback/route.ts

@@ -9,7 +9,7 @@ export async function GET(request: Request) {
   const { searchParams, origin } = new URL(request.url);
   const code = searchParams.get("code");
 
-  let next = "/marketplace";
+  let next = "/";
 
   if (code) {
     const supabase = await getServerSupabase();

autogpt_platform/frontend/src/app/(platform)/build/components/BuilderActions/components/RunGraph/RunGraph.tsx

@@ -5,10 +5,11 @@ import {
   TooltipContent,
   TooltipTrigger,
 } from "@/components/atoms/Tooltip/BaseTooltip";
-import { PlayIcon, StopIcon } from "@phosphor-icons/react";
+import { CircleNotchIcon, PlayIcon, StopIcon } from "@phosphor-icons/react";
 import { useShallow } from "zustand/react/shallow";
 import { RunInputDialog } from "../RunInputDialog/RunInputDialog";
 import { useRunGraph } from "./useRunGraph";
+import { cn } from "@/lib/utils";
 
 export const RunGraph = ({ flowID }: { flowID: string | null }) => {
   const {
@@ -24,6 +25,31 @@ export const RunGraph = ({ flowID }: { flowID: string | null }) => {
     useShallow((state) => state.isGraphRunning),
   );
 
+  const isLoading = isExecutingGraph || isTerminatingGraph || isSaving;
+
+  // Determine which icon to show with proper animation
+  const renderIcon = () => {
+    const iconClass = cn(
+      "size-4 transition-transform duration-200 ease-out",
+      !isLoading && "group-hover:scale-110",
+    );
+
+    if (isLoading) {
+      return (
+        <CircleNotchIcon
+          className={cn(iconClass, "animate-spin")}
+          weight="bold"
+        />
+      );
+    }
+
+    if (isGraphRunning) {
+      return <StopIcon className={iconClass} weight="fill" />;
+    }
+
+    return <PlayIcon className={iconClass} weight="fill" />;
+  };
+
   return (
     <>
       <Tooltip>
@@ -33,18 +59,18 @@ export const RunGraph = ({ flowID }: { flowID: string | null }) => {
             variant={isGraphRunning ? "destructive" : "primary"}
             data-id={isGraphRunning ? "stop-graph-button" : "run-graph-button"}
             onClick={isGraphRunning ? handleStopGraph : handleRunGraph}
-            disabled={!flowID || isExecutingGraph || isTerminatingGraph}
+            disabled={!flowID || isLoading}
-            loading={isExecutingGraph || isTerminatingGraph || isSaving}
+            className="group"
           >
-            {!isGraphRunning ? (
+            {renderIcon()}
-              <PlayIcon className="size-4" />
-            ) : (
-              <StopIcon className="size-4" />
-            )}
           </Button>
         </TooltipTrigger>
         <TooltipContent>
-          {isGraphRunning ? "Stop agent" : "Run agent"}
+          {isLoading
+            ? "Processing..."
+            : isGraphRunning
+              ? "Stop agent"
+              : "Run agent"}
         </TooltipContent>
       </Tooltip>
       <RunInputDialog

autogpt_platform/frontend/src/app/(platform)/build/components/BuilderActions/components/RunInputDialog/RunInputDialog.tsx

@@ -10,6 +10,7 @@ import { useRunInputDialog } from "./useRunInputDialog";
 import { CronSchedulerDialog } from "../CronSchedulerDialog/CronSchedulerDialog";
 import { useTutorialStore } from "@/app/(platform)/build/stores/tutorialStore";
 import { useEffect } from "react";
+import { CredentialsGroupedView } from "@/components/contextual/CredentialsInput/components/CredentialsGroupedView/CredentialsGroupedView";
 
 export const RunInputDialog = ({
   isOpen,
@@ -23,19 +24,17 @@ export const RunInputDialog = ({
   const hasInputs = useGraphStore((state) => state.hasInputs);
   const hasCredentials = useGraphStore((state) => state.hasCredentials);
   const inputSchema = useGraphStore((state) => state.inputSchema);
-  const credentialsSchema = useGraphStore(
-    (state) => state.credentialsInputSchema,
-  );
 
   const {
-    credentialsUiSchema,
+    credentialFields,
+    requiredCredentials,
     handleManualRun,
     handleInputChange,
     openCronSchedulerDialog,
     setOpenCronSchedulerDialog,
     inputValues,
     credentialValues,
-    handleCredentialChange,
+    handleCredentialFieldChange,
     isExecutingGraph,
   } = useRunInputDialog({ setIsOpen });
 
@@ -62,67 +61,67 @@ export const RunInputDialog = ({
           isOpen,
           set: setIsOpen,
         }}
-        styling={{ maxWidth: "600px", minWidth: "600px" }}
+        styling={{ maxWidth: "700px", minWidth: "700px" }}
       >
         <Dialog.Content>
-          <div className="space-y-6 p-1" data-id="run-input-dialog-content">
+          <div
-            {/* Credentials Section */}
+            className="grid grid-cols-[1fr_auto] gap-10 p-1"
-            {hasCredentials() && (
+            data-id="run-input-dialog-content"
-              <div data-id="run-input-credentials-section">
+          >
-                <div className="mb-4">
+            <div className="space-y-6">
-                  <Text variant="h4" className="text-gray-900">
+              {/* Credentials Section */}
-                    Credentials
+              {hasCredentials() && credentialFields.length > 0 && (
-                  </Text>
+                <div data-id="run-input-credentials-section">
+                  <div className="mb-4">
+                    <Text variant="h4" className="text-gray-900">
+                      Credentials
+                    </Text>
+                  </div>
+                  <div className="px-2" data-id="run-input-credentials-form">
+                    <CredentialsGroupedView
+                      credentialFields={credentialFields}
+                      requiredCredentials={requiredCredentials}
+                      inputCredentials={credentialValues}
+                      inputValues={inputValues}
+                      onCredentialChange={handleCredentialFieldChange}
+                    />
+                  </div>
                 </div>
-                <div className="px-2" data-id="run-input-credentials-form">
+              )}
-                  <FormRenderer
-                    jsonSchema={credentialsSchema as RJSFSchema}
-                    handleChange={(v) => handleCredentialChange(v.formData)}
-                    uiSchema={credentialsUiSchema}
-                    initialValues={{}}
-                    formContext={{
-                      showHandles: false,
-                      size: "large",
-                      showOptionalToggle: false,
-                    }}
-                  />
-                </div>
-              </div>
-            )}
 
-            {/* Inputs Section */}
+              {/* Inputs Section */}
-            {hasInputs() && (
+              {hasInputs() && (
-              <div data-id="run-input-inputs-section">
+                <div data-id="run-input-inputs-section">
-                <div className="mb-4">
+                  <div className="mb-4">
-                  <Text variant="h4" className="text-gray-900">
+                    <Text variant="h4" className="text-gray-900">
-                    Inputs
+                      Inputs
-                  </Text>
+                    </Text>
+                  </div>
+                  <div data-id="run-input-inputs-form">
+                    <FormRenderer
+                      jsonSchema={inputSchema as RJSFSchema}
+                      handleChange={(v) => handleInputChange(v.formData)}
+                      uiSchema={uiSchema}
+                      initialValues={{}}
+                      formContext={{
+                        showHandles: false,
+                        size: "large",
+                      }}
+                    />
+                  </div>
                 </div>
-                <div data-id="run-input-inputs-form">
+              )}
-                  <FormRenderer
+            </div>
-                    jsonSchema={inputSchema as RJSFSchema}
-                    handleChange={(v) => handleInputChange(v.formData)}
-                    uiSchema={uiSchema}
-                    initialValues={{}}
-                    formContext={{
-                      showHandles: false,
-                      size: "large",
-                    }}
-                  />
-                </div>
-              </div>
-            )}
 
-            {/* Action Button */}
             <div
-              className="flex justify-end pt-2"
+              className="flex flex-col items-end justify-start"
               data-id="run-input-actions-section"
             >
               {purpose === "run" && (
                 <Button
                   variant="primary"
                   size="large"
-                  className="group h-fit min-w-0 gap-2"
+                  className="group h-fit min-w-0 gap-2 px-10"
                   onClick={handleManualRun}
                   loading={isExecutingGraph}
                   data-id="run-input-manual-run-button"
@@ -137,7 +136,7 @@ export const RunInputDialog = ({
                 <Button
                   variant="primary"
                   size="large"
-                  className="group h-fit min-w-0 gap-2"
+                  className="group h-fit min-w-0 gap-2 px-10"
                   onClick={() => setOpenCronSchedulerDialog(true)}
                   data-id="run-input-schedule-button"
                 >

autogpt_platform/frontend/src/app/(platform)/build/components/BuilderActions/components/RunInputDialog/useRunInputDialog.ts

@@ -7,12 +7,11 @@ import {
   GraphExecutionMeta,
 } from "@/lib/autogpt-server-api";
 import { parseAsInteger, parseAsString, useQueryStates } from "nuqs";
-import { useMemo, useState } from "react";
+import { useCallback, useMemo, useState } from "react";
-import { uiSchema } from "../../../FlowEditor/nodes/uiSchema";
-import { isCredentialFieldSchema } from "@/components/renderers/InputRenderer/custom/CredentialField/helpers";
 import { useNodeStore } from "@/app/(platform)/build/stores/nodeStore";
 import { useToast } from "@/components/molecules/Toast/use-toast";
 import { useReactFlow } from "@xyflow/react";
+import type { CredentialField } from "@/components/contextual/CredentialsInput/components/CredentialsGroupedView/helpers";
 
 export const useRunInputDialog = ({
   setIsOpen,
@@ -120,27 +119,32 @@ export const useRunInputDialog = ({
       },
     });
 
-  // We are rendering the credentials field differently compared to other fields.
+  // Convert credentials schema to credential fields array for CredentialsGroupedView
-  // In the node, we have the field name as "credential" - so our library catches it and renders it differently.
+  const credentialFields: CredentialField[] = useMemo(() => {
-  // But here we have a different name, something like `Firecrawl credentials`, so here we are telling the library that this field is a credential field type.
+    if (!credentialsSchema?.properties) return [];
+    return Object.entries(credentialsSchema.properties);
+  }, [credentialsSchema]);
 
-  const credentialsUiSchema = useMemo(() => {
+  // Get required credentials as a Set
-    const dynamicUiSchema: any = { ...uiSchema };
+  const requiredCredentials = useMemo(() => {
+    return new Set<string>(credentialsSchema?.required || []);
+  }, [credentialsSchema]);
 
-    if (credentialsSchema?.properties) {
+  // Handler for individual credential changes
-      Object.keys(credentialsSchema.properties).forEach((fieldName) => {
+  const handleCredentialFieldChange = useCallback(
-        const fieldSchema = credentialsSchema.properties[fieldName];
+    (key: string, value?: CredentialsMetaInput) => {
-        if (isCredentialFieldSchema(fieldSchema)) {
+      setCredentialValues((prev) => {
-          dynamicUiSchema[fieldName] = {
+        if (value) {
-            ...dynamicUiSchema[fieldName],
+          return { ...prev, [key]: value };
-            "ui:field": "custom/credential_field",
+        } else {
-          };
+          const next = { ...prev };
+          delete next[key];
+          return next;
         }
       });
-    }
+    },
-
+    [],
-    return dynamicUiSchema;
+  );
-  }, [credentialsSchema]);
 
   const handleManualRun = async () => {
     // Filter out incomplete credentials (those without a valid id)
@@ -173,12 +177,14 @@ export const useRunInputDialog = ({
   };
 
   return {
-    credentialsUiSchema,
+    credentialFields,
+    requiredCredentials,
     inputValues,
     credentialValues,
     isExecutingGraph,
     handleInputChange,
     handleCredentialChange,
+    handleCredentialFieldChange,
     handleManualRun,
     openCronSchedulerDialog,
     setOpenCronSchedulerDialog,

autogpt_platform/frontend/src/app/(platform)/build/components/FloatingSafeModeToogle.tsx

@@ -18,69 +18,118 @@ interface Props {
   fullWidth?: boolean;
 }
 
+interface SafeModeButtonProps {
+  isEnabled: boolean;
+  label: string;
+  tooltipEnabled: string;
+  tooltipDisabled: string;
+  onToggle: () => void;
+  isPending: boolean;
+  fullWidth?: boolean;
+}
+
+function SafeModeButton({
+  isEnabled,
+  label,
+  tooltipEnabled,
+  tooltipDisabled,
+  onToggle,
+  isPending,
+  fullWidth = false,
+}: SafeModeButtonProps) {
+  return (
+    <Tooltip delayDuration={100}>
+      <TooltipTrigger asChild>
+        <Button
+          variant={isEnabled ? "primary" : "outline"}
+          size="small"
+          onClick={onToggle}
+          disabled={isPending}
+          className={cn("justify-start", fullWidth ? "w-full" : "")}
+        >
+          {isEnabled ? (
+            <>
+              <ShieldCheckIcon weight="bold" size={16} />
+              <Text variant="body" className="text-zinc-200">
+                {label}: ON
+              </Text>
+            </>
+          ) : (
+            <>
+              <ShieldIcon weight="bold" size={16} />
+              <Text variant="body" className="text-zinc-600">
+                {label}: OFF
+              </Text>
+            </>
+          )}
+        </Button>
+      </TooltipTrigger>
+      <TooltipContent>
+        <div className="text-center">
+          <div className="font-medium">
+            {label}: {isEnabled ? "ON" : "OFF"}
+          </div>
+          <div className="mt-1 text-xs text-muted-foreground">
+            {isEnabled ? tooltipEnabled : tooltipDisabled}
+          </div>
+        </div>
+      </TooltipContent>
+    </Tooltip>
+  );
+}
+
 export function FloatingSafeModeToggle({
   graph,
   className,
   fullWidth = false,
 }: Props) {
   const {
-    currentSafeMode,
+    currentHITLSafeMode,
+    showHITLToggle,
+    isHITLStateUndetermined,
+    handleHITLToggle,
+    currentSensitiveActionSafeMode,
+    showSensitiveActionToggle,
+    handleSensitiveActionToggle,
     isPending,
     shouldShowToggle,
-    isStateUndetermined,
-    handleToggle,
   } = useAgentSafeMode(graph);
 
-  if (!shouldShowToggle || isStateUndetermined || isPending) {
+  if (!shouldShowToggle || isPending) {
+    return null;
+  }
+
+  const showHITL = showHITLToggle && !isHITLStateUndetermined;
+  const showSensitive = showSensitiveActionToggle;
+
+  if (!showHITL && !showSensitive) {
     return null;
   }
 
   return (
-    <div className={cn("fixed z-50", className)}>
+    <div className={cn("fixed z-50 flex flex-col gap-2", className)}>
-      <Tooltip delayDuration={100}>
+      {showHITL && (
-        <TooltipTrigger asChild>
+        <SafeModeButton
-          <Button
+          isEnabled={currentHITLSafeMode}
-            variant={currentSafeMode! ? "primary" : "outline"}
+          label="Human in the loop block approval"
-            key={graph.id}
+          tooltipEnabled="The agent will pause at human-in-the-loop blocks and wait for your approval"
-            size="small"
+          tooltipDisabled="Human in the loop blocks will proceed automatically"
-            title={
+          onToggle={handleHITLToggle}
-              currentSafeMode!
+          isPending={isPending}
-                ? "Safe Mode: ON. Human in the loop blocks require manual review"
+          fullWidth={fullWidth}
-                : "Safe Mode: OFF. Human in the loop blocks proceed automatically"
+        />
-            }
+      )}
-            onClick={handleToggle}
+      {showSensitive && (
-            className={cn(fullWidth ? "w-full" : "")}
+        <SafeModeButton
-          >
+          isEnabled={currentSensitiveActionSafeMode}
-            {currentSafeMode! ? (
+          label="Sensitive actions blocks approval"
-              <>
+          tooltipEnabled="The agent will pause at sensitive action blocks and wait for your approval"
-                <ShieldCheckIcon weight="bold" size={16} />
+          tooltipDisabled="Sensitive action blocks will proceed automatically"
-                <Text variant="body" className="text-zinc-200">
+          onToggle={handleSensitiveActionToggle}
-                  Safe Mode: ON
+          isPending={isPending}
-                </Text>
+          fullWidth={fullWidth}
-              </>
+        />
-            ) : (
+      )}
-              <>
-                <ShieldIcon weight="bold" size={16} />
-                <Text variant="body" className="text-zinc-600">
-                  Safe Mode: OFF
-                </Text>
-              </>
-            )}
-          </Button>
-        </TooltipTrigger>
-        <TooltipContent>
-          <div className="text-center">
-            <div className="font-medium">
-              Safe Mode: {currentSafeMode! ? "ON" : "OFF"}
-            </div>
-            <div className="mt-1 text-xs text-muted-foreground">
-              {currentSafeMode!
-                ? "Human in the loop blocks require manual review"
-                : "Human in the loop blocks proceed automatically"}
-            </div>
-          </div>
-        </TooltipContent>
-      </Tooltip>
     </div>
   );
 }

autogpt_platform/frontend/src/app/(platform)/build/components/FlowEditor/Flow/components/CustomControl.tsx

@@ -53,14 +53,14 @@ export const CustomControls = memo(
     const controls = [
       {
         id: "zoom-in-button",
-        icon: <PlusIcon className="size-4" />,
+        icon: <PlusIcon className="size-3.5 text-zinc-600" />,
         label: "Zoom In",
         onClick: () => zoomIn(),
         className: "h-10 w-10 border-none",
       },
       {
         id: "zoom-out-button",
-        icon: <MinusIcon className="size-4" />,
+        icon: <MinusIcon className="size-3.5 text-zinc-600" />,
         label: "Zoom Out",
         onClick: () => zoomOut(),
         className: "h-10 w-10 border-none",
@@ -68,9 +68,9 @@ export const CustomControls = memo(
       {
         id: "tutorial-button",
         icon: isTutorialLoading ? (
-          <CircleNotchIcon className="size-4 animate-spin" />
+          <CircleNotchIcon className="size-3.5 animate-spin text-zinc-600" />
         ) : (
-          <ChalkboardIcon className="size-4" />
+          <ChalkboardIcon className="size-3.5 text-zinc-600" />
         ),
         label: isTutorialLoading ? "Loading Tutorial..." : "Start Tutorial",
         onClick: handleTutorialClick,
@@ -79,7 +79,7 @@ export const CustomControls = memo(
       },
       {
         id: "fit-view-button",
-        icon: <FrameCornersIcon className="size-4" />,
+        icon: <FrameCornersIcon className="size-3.5 text-zinc-600" />,
         label: "Fit View",
         onClick: () => fitView({ padding: 0.2, duration: 800, maxZoom: 1 }),
         className: "h-10 w-10 border-none",
@@ -87,9 +87,9 @@ export const CustomControls = memo(
       {
         id: "lock-button",
         icon: !isLocked ? (
-          <LockOpenIcon className="size-4" />
+          <LockOpenIcon className="size-3.5 text-zinc-600" />
         ) : (
-          <LockIcon className="size-4" />
+          <LockIcon className="size-3.5 text-zinc-600" />
         ),
         label: "Toggle Lock",
         onClick: () => setIsLocked(!isLocked),

autogpt_platform/frontend/src/app/(platform)/build/components/FlowEditor/Flow/useFlow.ts

@@ -139,14 +139,6 @@ export const useFlow = () => {
       useNodeStore.getState().setNodes([]);
       useNodeStore.getState().clearResolutionState();
       addNodes(customNodes);
-
-      // Sync hardcoded values with handle IDs.
-      // If a key–value field has a key without a value, the backend omits it from hardcoded values.
-      // But if a handleId exists for that key, it causes inconsistency.
-      // This ensures hardcoded values stay in sync with handle IDs.
-      customNodes.forEach((node) => {
-        useNodeStore.getState().syncHardcodedValuesWithHandleIds(node.id);
-      });
     }
   }, [customNodes, addNodes]);
 
@@ -158,6 +150,14 @@ export const useFlow = () => {
     }
   }, [graph?.links, addLinks]);
 
+  useEffect(() => {
+    if (customNodes.length > 0 && graph?.links) {
+      customNodes.forEach((node) => {
+        useNodeStore.getState().syncHardcodedValuesWithHandleIds(node.id);
+      });
+    }
+  }, [customNodes, graph?.links]);
+
   // update node execution status in nodes
   useEffect(() => {
     if (

autogpt_platform/frontend/src/app/(platform)/build/components/FlowEditor/edges/CustomEdge.tsx

@@ -19,6 +19,8 @@ export type CustomEdgeData = {
   beadUp?: number;
   beadDown?: number;
   beadData?: Map<string, NodeExecutionResult["status"]>;
+  edgeColorClass?: string;
+  edgeHexColor?: string;
 };
 
 export type CustomEdge = XYEdge<CustomEdgeData, "custom">;
@@ -36,7 +38,6 @@ const CustomEdge = ({
   selected,
 }: EdgeProps<CustomEdge>) => {
   const removeConnection = useEdgeStore((state) => state.removeEdge);
-  // Subscribe to the brokenEdgeIDs map and check if this edge is broken across any node
   const isBroken = useNodeStore((state) => state.isEdgeBroken(id));
   const [isHovered, setIsHovered] = useState(false);
 
@@ -52,6 +53,7 @@ const CustomEdge = ({
   const isStatic = data?.isStatic ?? false;
   const beadUp = data?.beadUp ?? 0;
   const beadDown = data?.beadDown ?? 0;
+  const edgeColorClass = data?.edgeColorClass;
 
   const handleRemoveEdge = () => {
     removeConnection(id);
@@ -70,7 +72,9 @@ const CustomEdge = ({
             ? "!stroke-red-500 !stroke-[2px] [stroke-dasharray:4]"
             : selected
               ? "stroke-zinc-800"
-              : "stroke-zinc-500/50 hover:stroke-zinc-500",
+              : edgeColorClass
+                ? cn(edgeColorClass, "opacity-70 hover:opacity-100")
+                : "stroke-zinc-500/50 hover:stroke-zinc-500",
         )}
       />
       <JSBeads

autogpt_platform/frontend/src/app/(platform)/build/components/FlowEditor/edges/useCustomEdge.ts

@@ -8,6 +8,7 @@ import { useCallback } from "react";
 import { useNodeStore } from "../../../stores/nodeStore";
 import { useHistoryStore } from "../../../stores/historyStore";
 import { CustomEdge } from "./CustomEdge";
+import { getEdgeColorFromOutputType } from "../nodes/helpers";
 
 export const useCustomEdge = () => {
   const edges = useEdgeStore((s) => s.edges);
@@ -34,8 +35,13 @@ export const useCustomEdge = () => {
       if (exists) return;
 
       const nodes = useNodeStore.getState().nodes;
-      const isStatic = nodes.find((n) => n.id === conn.source)?.data
+      const sourceNode = nodes.find((n) => n.id === conn.source);
-        ?.staticOutput;
+      const isStatic = sourceNode?.data?.staticOutput;
+
+      const { colorClass, hexColor } = getEdgeColorFromOutputType(
+        sourceNode?.data?.outputSchema,
+        conn.sourceHandle,
+      );
 
       addEdge({
         source: conn.source,
@@ -44,6 +50,8 @@ export const useCustomEdge = () => {
         targetHandle: conn.targetHandle,
         data: {
           isStatic,
+          edgeColorClass: colorClass,
+          edgeHexColor: hexColor,
         },
       });
     },

autogpt_platform/frontend/src/app/(platform)/build/components/FlowEditor/nodes/CustomNode/components/NodeOutput/NodeOutput.tsx

@@ -1,22 +1,21 @@
 import { Button } from "@/components/atoms/Button/Button";
 import { Text } from "@/components/atoms/Text/Text";
+import {
+  Accordion,
+  AccordionContent,
+  AccordionItem,
+  AccordionTrigger,
+} from "@/components/molecules/Accordion/Accordion";
 import { beautifyString, cn } from "@/lib/utils";
-import { CaretDownIcon, CopyIcon, CheckIcon } from "@phosphor-icons/react";
+import { CopyIcon, CheckIcon } from "@phosphor-icons/react";
 import { NodeDataViewer } from "./components/NodeDataViewer/NodeDataViewer";
 import { ContentRenderer } from "./components/ContentRenderer";
 import { useNodeOutput } from "./useNodeOutput";
 import { ViewMoreData } from "./components/ViewMoreData";
 
 export const NodeDataRenderer = ({ nodeId }: { nodeId: string }) => {
-  const {
+  const { outputData, copiedKey, handleCopy, executionResultId, inputData } =
-    outputData,
+    useNodeOutput(nodeId);
-    isExpanded,
-    setIsExpanded,
-    copiedKey,
-    handleCopy,
-    executionResultId,
-    inputData,
-  } = useNodeOutput(nodeId);
 
   if (Object.keys(outputData).length === 0) {
     return null;
@@ -25,122 +24,117 @@ export const NodeDataRenderer = ({ nodeId }: { nodeId: string }) => {
   return (
     <div
       data-tutorial-id={`node-output`}
-      className="flex flex-col gap-3 rounded-b-xl border-t border-zinc-200 px-4 py-4"
+      className="rounded-b-xl border-t border-zinc-200 px-4 py-2"
     >
-      <div className="flex items-center justify-between">
+      <Accordion type="single" collapsible defaultValue="node-output">
-        <Text variant="body-medium" className="!font-semibold text-slate-700">
+        <AccordionItem value="node-output" className="border-none">
-          Node Output
+          <AccordionTrigger className="py-2 hover:no-underline">
-        </Text>
+            <Text
-        <Button
+              variant="body-medium"
-          variant="ghost"
+              className="!font-semibold text-slate-700"
-          size="small"
+            >
-          onClick={() => setIsExpanded(!isExpanded)}
+              Node Output
-          className="h-fit min-w-0 p-1 text-slate-600 hover:text-slate-900"
+            </Text>
-        >
+          </AccordionTrigger>
-          <CaretDownIcon
+          <AccordionContent className="pt-2">
-            size={16}
+            <div className="flex max-w-[350px] flex-col gap-4">
-            weight="bold"
+              <div className="space-y-2">
-            className={`transition-transform ${isExpanded ? "rotate-180" : ""}`}
+                <Text variant="small-medium">Input</Text>
-          />
-        </Button>
-      </div>
 
-      {isExpanded && (
+                <ContentRenderer value={inputData} shortContent={false} />
-        <>
-          <div className="flex max-w-[350px] flex-col gap-4">
-            <div className="space-y-2">
-              <Text variant="small-medium">Input</Text>
 
-              <ContentRenderer value={inputData} shortContent={false} />
+                <div className="mt-1 flex justify-end gap-1">
-
+                  <NodeDataViewer
-              <div className="mt-1 flex justify-end gap-1">
+                    data={inputData}
-                <NodeDataViewer
+                    pinName="Input"
-                  data={inputData}
+                    execId={executionResultId}
-                  pinName="Input"
+                  />
-                  execId={executionResultId}
+                  <Button
-                />
+                    variant="secondary"
-                <Button
+                    size="small"
-                  variant="secondary"
+                    onClick={() => handleCopy("input", inputData)}
-                  size="small"
+                    className={cn(
-                  onClick={() => handleCopy("input", inputData)}
+                      "h-fit min-w-0 gap-1.5 border border-zinc-200 p-2 text-black hover:text-slate-900",
-                  className={cn(
+                      copiedKey === "input" &&
-                    "h-fit min-w-0 gap-1.5 border border-zinc-200 p-2 text-black hover:text-slate-900",
+                        "border-green-400 bg-green-100 hover:border-green-400 hover:bg-green-200",
-                    copiedKey === "input" &&
+                    )}
-                      "border-green-400 bg-green-100 hover:border-green-400 hover:bg-green-200",
+                  >
-                  )}
+                    {copiedKey === "input" ? (
-                >
+                      <CheckIcon size={12} className="text-green-600" />
-                  {copiedKey === "input" ? (
+                    ) : (
-                    <CheckIcon size={12} className="text-green-600" />
+                      <CopyIcon size={12} />
-                  ) : (
+                    )}
-                    <CopyIcon size={12} />
+                  </Button>
-                  )}
+                </div>
-                </Button>
               </div>
-            </div>
 
-            {Object.entries(outputData)
+              {Object.entries(outputData)
-              .slice(0, 2)
+                .slice(0, 2)
-              .map(([key, value]) => (
+                .map(([key, value]) => (
-                <div key={key} className="flex flex-col gap-2">
+                  <div key={key} className="flex flex-col gap-2">
-                  <div className="flex items-center gap-2">
+                    <div className="flex items-center gap-2">
-                    <Text
+                      <Text
-                      variant="small-medium"
+                        variant="small-medium"
-                      className="!font-semibold text-slate-600"
+                        className="!font-semibold text-slate-600"
-                    >
+                      >
-                      Pin:
+                        Pin:
-                    </Text>
+                      </Text>
-                    <Text variant="small" className="text-slate-700">
+                      <Text variant="small" className="text-slate-700">
-                      {beautifyString(key)}
+                        {beautifyString(key)}
-                    </Text>
+                      </Text>
-                  </div>
+                    </div>
-                  <div className="w-full space-y-2">
+                    <div className="w-full space-y-2">
-                    <Text
+                      <Text
-                      variant="small"
+                        variant="small"
-                      className="!font-semibold text-slate-600"
+                        className="!font-semibold text-slate-600"
-                    >
+                      >
-                      Data:
+                        Data:
-                    </Text>
+                      </Text>
-                    <div className="relative space-y-2">
+                      <div className="relative space-y-2">
-                      {value.map((item, index) => (
+                        {value.map((item, index) => (
-                        <div key={index}>
+                          <div key={index}>
-                          <ContentRenderer value={item} shortContent={true} />
+                            <ContentRenderer value={item} shortContent={true} />
+                          </div>
+                        ))}
+
+                        <div className="mt-1 flex justify-end gap-1">
+                          <NodeDataViewer
+                            data={value}
+                            pinName={key}
+                            execId={executionResultId}
+                          />
+                          <Button
+                            variant="secondary"
+                            size="small"
+                            onClick={() => handleCopy(key, value)}
+                            className={cn(
+                              "h-fit min-w-0 gap-1.5 border border-zinc-200 p-2 text-black hover:text-slate-900",
+                              copiedKey === key &&
+                                "border-green-400 bg-green-100 hover:border-green-400 hover:bg-green-200",
+                            )}
+                          >
+                            {copiedKey === key ? (
+                              <CheckIcon size={12} className="text-green-600" />
+                            ) : (
+                              <CopyIcon size={12} />
+                            )}
+                          </Button>
                         </div>
-                      ))}
-
-                      <div className="mt-1 flex justify-end gap-1">
-                        <NodeDataViewer
-                          data={value}
-                          pinName={key}
-                          execId={executionResultId}
-                        />
-                        <Button
-                          variant="secondary"
-                          size="small"
-                          onClick={() => handleCopy(key, value)}
-                          className={cn(
-                            "h-fit min-w-0 gap-1.5 border border-zinc-200 p-2 text-black hover:text-slate-900",
-                            copiedKey === key &&
-                              "border-green-400 bg-green-100 hover:border-green-400 hover:bg-green-200",
-                          )}
-                        >
-                          {copiedKey === key ? (
-                            <CheckIcon size={12} className="text-green-600" />
-                          ) : (
-                            <CopyIcon size={12} />
-                          )}
-                        </Button>
                       </div>
                     </div>
                   </div>
-                </div>
+                ))}
-              ))}
+            </div>
-          </div>
 
-          {Object.keys(outputData).length > 2 && (
+            {Object.keys(outputData).length > 2 && (
-            <ViewMoreData outputData={outputData} execId={executionResultId} />
+              <ViewMoreData
-          )}
+                outputData={outputData}
-        </>
+                execId={executionResultId}
-      )}
+              />
+            )}
+          </AccordionContent>
+        </AccordionItem>
+      </Accordion>
     </div>
   );
 };

autogpt_platform/frontend/src/app/(platform)/build/components/FlowEditor/nodes/CustomNode/components/NodeOutput/useNodeOutput.tsx

@@ -4,7 +4,6 @@ import { useShallow } from "zustand/react/shallow";
 import { useState } from "react";
 
 export const useNodeOutput = (nodeId: string) => {
-  const [isExpanded, setIsExpanded] = useState(true);
   const [copiedKey, setCopiedKey] = useState<string | null>(null);
   const { toast } = useToast();
 
@@ -37,13 +36,10 @@ export const useNodeOutput = (nodeId: string) => {
     }
   };
   return {
-    outputData: outputData,
+    outputData,
-    inputData: inputData,
+    inputData,
-    isExpanded: isExpanded,
+    copiedKey,
-    setIsExpanded: setIsExpanded,
+    handleCopy,
-    copiedKey: copiedKey,
-    setCopiedKey: setCopiedKey,
-    handleCopy: handleCopy,
     executionResultId: nodeExecutionResult?.node_exec_id,
   };
 };

autogpt_platform/frontend/src/app/(platform)/build/components/FlowEditor/nodes/helpers.ts

@@ -187,3 +187,38 @@ export const getTypeDisplayInfo = (schema: any) => {
     hexColor,
   };
 };
+
+export function getEdgeColorFromOutputType(
+  outputSchema: RJSFSchema | undefined,
+  sourceHandle: string,
+): { colorClass: string; hexColor: string } {
+  const defaultColor = {
+    colorClass: "stroke-zinc-500/50",
+    hexColor: "#6b7280",
+  };
+
+  if (!outputSchema?.properties) return defaultColor;
+
+  const properties = outputSchema.properties as Record<string, unknown>;
+  const handleParts = sourceHandle.split("_#_");
+  let currentSchema: Record<string, unknown> = properties;
+
+  for (let i = 0; i < handleParts.length; i++) {
+    const part = handleParts[i];
+    const fieldSchema = currentSchema[part] as Record<string, unknown>;
+    if (!fieldSchema) return defaultColor;
+
+    if (i === handleParts.length - 1) {
+      const { hexColor, colorClass } = getTypeDisplayInfo(fieldSchema);
+      return { colorClass: colorClass.replace("!text-", "stroke-"), hexColor };
+    }
+
+    if (fieldSchema.properties) {
+      currentSchema = fieldSchema.properties as Record<string, unknown>;
+    } else {
+      return defaultColor;
+    }
+  }
+
+  return defaultColor;
+}

autogpt_platform/frontend/src/app/(platform)/build/components/FlowEditor/tutorial/icons.ts

@@ -1,7 +1,32 @@
-// These are SVG Phosphor icons
+type IconOptions = {
+  size?: number;
+  color?: string;
+};
+
+const DEFAULT_SIZE = 16;
+const DEFAULT_COLOR = "#52525b"; // zinc-600
+
+const iconPaths = {
+  ClickIcon: `M88,24V16a8,8,0,0,1,16,0v8a8,8,0,0,1-16,0ZM16,104h8a8,8,0,0,0,0-16H16a8,8,0,0,0,0,16ZM124.42,39.16a8,8,0,0,0,10.74-3.58l8-16a8,8,0,0,0-14.31-7.16l-8,16A8,8,0,0,0,124.42,39.16Zm-96,81.69-16,8a8,8,0,0,0,7.16,14.31l16-8a8,8,0,1,0-7.16-14.31ZM219.31,184a16,16,0,0,1,0,22.63l-12.68,12.68a16,16,0,0,1-22.63,0L132.7,168,115,214.09c0,.1-.08.21-.13.32a15.83,15.83,0,0,1-14.6,9.59l-.79,0a15.83,15.83,0,0,1-14.41-11L32.8,52.92A16,16,0,0,1,52.92,32.8L213,85.07a16,16,0,0,1,1.41,29.8l-.32.13L168,132.69ZM208,195.31,156.69,144h0a16,16,0,0,1,4.93-26l.32-.14,45.95-17.64L48,48l52.2,159.86,17.65-46c0-.11.08-.22.13-.33a16,16,0,0,1,11.69-9.34,16.72,16.72,0,0,1,3-.28,16,16,0,0,1,11.3,4.69L195.31,208Z`,
+  Keyboard: `M224,48H32A16,16,0,0,0,16,64V192a16,16,0,0,0,16,16H224a16,16,0,0,0,16-16V64A16,16,0,0,0,224,48Zm0,144H32V64H224V192Zm-16-64a8,8,0,0,1-8,8H56a8,8,0,0,1,0-16H200A8,8,0,0,1,208,128Zm0-32a8,8,0,0,1-8,8H56a8,8,0,0,1,0-16H200A8,8,0,0,1,208,96ZM72,160a8,8,0,0,1-8,8H56a8,8,0,0,1,0-16h8A8,8,0,0,1,72,160Zm96,0a8,8,0,0,1-8,8H96a8,8,0,0,1,0-16h64A8,8,0,0,1,168,160Zm40,0a8,8,0,0,1-8,8h-8a8,8,0,0,1,0-16h8A8,8,0,0,1,208,160Z`,
+  Drag: `M188,80a27.79,27.79,0,0,0-13.36,3.4,28,28,0,0,0-46.64-11A28,28,0,0,0,80,92v20H68a28,28,0,0,0-28,28v12a88,88,0,0,0,176,0V108A28,28,0,0,0,188,80Zm12,72a72,72,0,0,1-144,0V140a12,12,0,0,1,12-12H80v24a8,8,0,0,0,16,0V92a12,12,0,0,1,24,0v28a8,8,0,0,0,16,0V92a12,12,0,0,1,24,0v28a8,8,0,0,0,16,0V108a12,12,0,0,1,24,0Z`,
+};
+
+function createIcon(path: string, options: IconOptions = {}): string {
+  const size = options.size ?? DEFAULT_SIZE;
+  const color = options.color ?? DEFAULT_COLOR;
+  return `<svg xmlns="http://www.w3.org/2000/svg" width="${size}" height="${size}" fill="${color}" viewBox="0 0 256 256"><path d="${path}"></path></svg>`;
+}
 
 export const ICONS = {
-  ClickIcon: `<svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" fill="#000000" viewBox="0 0 256 256"><path d="M88,24V16a8,8,0,0,1,16,0v8a8,8,0,0,1-16,0ZM16,104h8a8,8,0,0,0,0-16H16a8,8,0,0,0,0,16ZM124.42,39.16a8,8,0,0,0,10.74-3.58l8-16a8,8,0,0,0-14.31-7.16l-8,16A8,8,0,0,0,124.42,39.16Zm-96,81.69-16,8a8,8,0,0,0,7.16,14.31l16-8a8,8,0,1,0-7.16-14.31ZM219.31,184a16,16,0,0,1,0,22.63l-12.68,12.68a16,16,0,0,1-22.63,0L132.7,168,115,214.09c0,.1-.08.21-.13.32a15.83,15.83,0,0,1-14.6,9.59l-.79,0a15.83,15.83,0,0,1-14.41-11L32.8,52.92A16,16,0,0,1,52.92,32.8L213,85.07a16,16,0,0,1,1.41,29.8l-.32.13L168,132.69ZM208,195.31,156.69,144h0a16,16,0,0,1,4.93-26l.32-.14,45.95-17.64L48,48l52.2,159.86,17.65-46c0-.11.08-.22.13-.33a16,16,0,0,1,11.69-9.34,16.72,16.72,0,0,1,3-.28,16,16,0,0,1,11.3,4.69L195.31,208Z"></path></svg>`,
+  ClickIcon: createIcon(iconPaths.ClickIcon),
-  Keyboard: `<svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" fill="#000000" viewBox="0 0 256 256"><path d="M224,48H32A16,16,0,0,0,16,64V192a16,16,0,0,0,16,16H224a16,16,0,0,0,16-16V64A16,16,0,0,0,224,48Zm0,144H32V64H224V192Zm-16-64a8,8,0,0,1-8,8H56a8,8,0,0,1,0-16H200A8,8,0,0,1,208,128Zm0-32a8,8,0,0,1-8,8H56a8,8,0,0,1,0-16H200A8,8,0,0,1,208,96ZM72,160a8,8,0,0,1-8,8H56a8,8,0,0,1,0-16h8A8,8,0,0,1,72,160Zm96,0a8,8,0,0,1-8,8H96a8,8,0,0,1,0-16h64A8,8,0,0,1,168,160Zm40,0a8,8,0,0,1-8,8h-8a8,8,0,0,1,0-16h8A8,8,0,0,1,208,160Z"></path></svg>`,
+  Keyboard: createIcon(iconPaths.Keyboard),
-  Drag: `<svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" fill="#000000" viewBox="0 0 256 256"><path d="M188,80a27.79,27.79,0,0,0-13.36,3.4,28,28,0,0,0-46.64-11A28,28,0,0,0,80,92v20H68a28,28,0,0,0-28,28v12a88,88,0,0,0,176,0V108A28,28,0,0,0,188,80Zm12,72a72,72,0,0,1-144,0V140a12,12,0,0,1,12-12H80v24a8,8,0,0,0,16,0V92a12,12,0,0,1,24,0v28a8,8,0,0,0,16,0V92a12,12,0,0,1,24,0v28a8,8,0,0,0,16,0V108a12,12,0,0,1,24,0Z"></path></svg>`,
+  Drag: createIcon(iconPaths.Drag),
 };
+
+export function getIcon(
+  name: keyof typeof iconPaths,
+  options?: IconOptions,
+): string {
+  return createIcon(iconPaths[name], options);
+}