Merge main into PR 13306

Resolve settings-related merge conflicts while preserving split condenser and verification pages.

Co-authored-by: openhands <openhands@all-hands.dev>

.agents/skills/cross-repo-testing/SKILL.md

@@ -0,0 +1,202 @@
+---
+name: cross-repo-testing
+description: This skill should be used when the user asks to "test a cross-repo feature", "deploy a feature branch to staging", "test SDK against OH Cloud", "e2e test a cloud workspace feature", "test provider tokens", "test secrets inheritance", or when changes span the SDK and OpenHands server repos and need end-to-end validation against a staging deployment.
+triggers:
+- cross-repo
+- staging deployment
+- feature branch deploy
+- test against cloud
+- e2e cloud
+---
+
+# Cross-Repo Testing: SDK ↔ OpenHands Cloud
+
+How to end-to-end test features that span `OpenHands/software-agent-sdk` and `OpenHands/OpenHands` (the Cloud backend).
+
+## Repository Map
+
+| Repo | Role | What lives here |
+|------|------|-----------------|
+| [`software-agent-sdk`](https://github.com/OpenHands/software-agent-sdk) | Agent core | `openhands-sdk`, `openhands-workspace`, `openhands-tools` packages. `OpenHandsCloudWorkspace` lives here. |
+| [`OpenHands`](https://github.com/OpenHands/OpenHands) | Cloud backend | FastAPI server (`openhands/app_server/`), sandbox management, auth, enterprise integrations. Deployed as OH Cloud. |
+| [`deploy`](https://github.com/OpenHands/deploy) | Infrastructure | Helm charts + GitHub Actions that build the enterprise Docker image and deploy to staging/production. |
+
+**Data flow:** SDK client → OH Cloud API (`/api/v1/...`) → sandbox agent-server (inside runtime container)
+
+## When You Need This
+
+There are **two flows** depending on which direction the dependency goes:
+
+| Flow | When | Example |
+|------|------|---------|
+| **A — SDK client → new Cloud API** | The SDK calls an API that doesn't exist yet on production | `workspace.get_llm()` calling `GET /api/v1/users/me?expose_secrets=true` |
+| **B — OH server → new SDK code** | The Cloud server needs unreleased SDK packages or a new agent-server image | Server consumes a new tool, agent behavior, or workspace method from the SDK |
+
+Flow A only requires deploying the server PR. Flow B requires pinning the SDK to an unreleased commit in the server PR **and** using the SDK PR's agent-server image. Both flows may apply simultaneously.
+
+---
+
+## Flow A: SDK Client Tests Against New Cloud API
+
+Use this when the SDK calls an endpoint that only exists on the server PR branch.
+
+### A1. Write and test the server-side changes
+
+In the `OpenHands` repo, implement the new API endpoint(s). Run unit tests:
+
+```bash
+cd OpenHands
+poetry run pytest tests/unit/app_server/test_<relevant>.py -v
+```
+
+Push a PR. Wait for the **"Push Enterprise Image" (Docker) CI job** to succeed — this builds `ghcr.io/openhands/enterprise-server:sha-<COMMIT>`.
+
+### A2. Write the SDK-side changes
+
+In `software-agent-sdk`, implement the client code (e.g., new methods on `OpenHandsCloudWorkspace`). Run SDK unit tests:
+
+```bash
+cd software-agent-sdk
+pip install -e openhands-sdk -e openhands-workspace
+pytest tests/ -v
+```
+
+Push a PR. SDK CI is independent — it doesn't need the server changes to pass unit tests.
+
+### A3. Deploy the server PR to staging
+
+See [Deploying to a Staging Feature Environment](#deploying-to-a-staging-feature-environment) below.
+
+### A4. Run the SDK e2e test against staging
+
+See [Running E2E Tests Against Staging](#running-e2e-tests-against-staging) below.
+
+---
+
+## Flow B: OH Server Needs Unreleased SDK Code
+
+Use this when the Cloud server depends on SDK changes that haven't been released to PyPI yet. The server's runtime containers run the `agent-server` image built from the SDK repo, so the server PR must be configured to use the SDK PR's image and packages.
+
+### B1. Get the SDK PR merged (or identify the commit)
+
+The SDK PR must have CI pass so its agent-server Docker image is built. The image is tagged with the **merge-commit SHA** from GitHub Actions — NOT the head-commit SHA shown in the PR.
+
+Find the correct image tag:
+- Check the SDK PR description for an `AGENT_SERVER_IMAGES` section
+- Or check the "Consolidate Build Information" CI job for `"short_sha": "<tag>"`
+
+### B2. Pin SDK packages to the commit in the OpenHands PR
+
+In the `OpenHands` repo PR, pin all 3 SDK packages (`openhands-sdk`, `openhands-agent-server`, `openhands-tools`) to the unreleased commit and update the agent-server image tag. This involves editing 3 files and regenerating 3 lock files.
+
+Follow the **`update-sdk` skill** → "Development: Pin SDK to an Unreleased Commit" section for the full procedure and file-by-file instructions.
+
+### B3. Wait for the OpenHands enterprise image to build
+
+Push the pinned changes. The OpenHands CI will build a new enterprise Docker image (`ghcr.io/openhands/enterprise-server:sha-<OH_COMMIT>`) that bundles the unreleased SDK. Wait for the "Push Enterprise Image" job to succeed.
+
+### B4. Deploy and test
+
+Follow [Deploying to a Staging Feature Environment](#deploying-to-a-staging-feature-environment) using the new OpenHands commit SHA.
+
+### B5. Before merging: remove the pin
+
+**CI guard:** `check-package-versions.yml` blocks merge to `main` if `[tool.poetry.dependencies]` contains `rev` fields. Before the OpenHands PR can merge, the SDK PR must be merged and released to PyPI, then the pin must be replaced with the released version number.
+
+---
+
+## Deploying to a Staging Feature Environment
+
+The `deploy` repo creates preview environments from OpenHands PRs.
+
+**Option A — GitHub Actions UI (preferred):**
+Go to `OpenHands/deploy` → Actions → "Create OpenHands preview PR" → enter the OpenHands PR number. This creates a branch `ohpr-<PR>-<random>` and opens a deploy PR.
+
+**Option B — Update an existing feature branch:**
+```bash
+cd deploy
+git checkout ohpr-<PR>-<random>
+# In .github/workflows/deploy.yaml, update BOTH:
+#   OPENHANDS_SHA: "<full-40-char-commit>"
+#   OPENHANDS_RUNTIME_IMAGE_TAG: "<same-commit>-nikolaik"
+git commit -am "Update OPENHANDS_SHA to <commit>" && git push
+```
+
+**Before updating the SHA**, verify the enterprise Docker image exists:
+```bash
+gh api repos/OpenHands/OpenHands/actions/runs \
+  --jq '.workflow_runs[] | select(.head_sha=="<COMMIT>") | "\(.name): \(.conclusion)"' \
+  | grep Docker
+# Must show: "Docker: success"
+```
+
+The deploy CI auto-triggers and creates the environment at:
+```
+https://ohpr-<PR>-<random>.staging.all-hands.dev
+```
+
+**Wait for it to be live:**
+```bash
+curl -s -o /dev/null -w "%{http_code}" https://ohpr-<PR>-<random>.staging.all-hands.dev/api/v1/health
+# 401 = server is up (auth required). DNS may take 1-2 min on first deploy.
+```
+
+## Running E2E Tests Against Staging
+
+**Critical: Feature deployments have their own Keycloak instance.** API keys from `app.all-hands.dev` or `$OPENHANDS_API_KEY` will NOT work. You need a test API key issued by the specific feature deployment's Keycloak.
+
+**You (the agent) cannot obtain this key yourself** — the feature environment requires interactive browser login with credentials you do not have. You must **ask the user** to:
+1. Log in to the feature deployment at `https://ohpr-<PR>-<random>.staging.all-hands.dev` in their browser
+2. Generate a test API key from the UI
+3. Provide the key to you so you can proceed with e2e testing
+
+Do **not** attempt to log in via the browser or guess credentials. Wait for the user to supply the key before running any e2e tests.
+
+```python
+from openhands.workspace import OpenHandsCloudWorkspace
+
+STAGING = "https://ohpr-<PR>-<random>.staging.all-hands.dev"
+
+with OpenHandsCloudWorkspace(
+    cloud_api_url=STAGING,
+    cloud_api_key="<test-api-key-for-this-deployment>",
+) as workspace:
+    # Test the new feature
+    llm = workspace.get_llm()
+    secrets = workspace.get_secrets()
+    print(f"LLM: {llm.model}, secrets: {list(secrets.keys())}")
+```
+
+Or run an example script:
+```bash
+OPENHANDS_CLOUD_API_KEY="<key>" \
+OPENHANDS_CLOUD_API_URL="https://ohpr-<PR>-<random>.staging.all-hands.dev" \
+python examples/02_remote_agent_server/10_cloud_workspace_saas_credentials.py
+```
+
+### Recording results
+
+Both repos support a `.pr/` directory for temporary PR artifacts (design docs, test logs, scripts). These files are automatically removed when the PR is approved — see `.github/workflows/pr-artifacts.yml` and the "PR-Specific Artifacts" section in each repo's `AGENTS.md`.
+
+Push test output to the `.pr/logs/` directory of whichever repo you're working in:
+```bash
+mkdir -p .pr/logs
+python test_script.py 2>&1 | tee .pr/logs/<test_name>.log
+git add -f .pr/logs/
+git commit -m "docs: add e2e test results" && git push
+```
+
+Comment on **both PRs** with pass/fail summary and link to logs.
+
+## Key Gotchas
+
+| Gotcha | Details |
+|--------|---------|
+| **Feature env auth is isolated** | Each `ohpr-*` deployment has its own Keycloak. Production API keys don't work. Agents cannot log in — you must ask the user to provide a test API key from the feature deployment's UI. |
+| **Two SHAs in deploy.yaml** | `OPENHANDS_SHA` and `OPENHANDS_RUNTIME_IMAGE_TAG` must both be updated. The runtime tag is `<sha>-nikolaik`. |
+| **Enterprise image must exist** | The Docker CI job on the OpenHands PR must succeed before you can deploy. If it hasn't run, push an empty commit to trigger it. |
+| **DNS propagation** | First deployment of a new branch takes 1-2 min for DNS. Subsequent deploys are instant. |
+| **Merge-commit SHA ≠ head SHA** | SDK CI tags Docker images with GitHub Actions' merge-commit SHA, not the PR head SHA. Check the SDK PR description or CI logs for the correct tag. |
+| **SDK pin blocks merge** | `check-package-versions.yml` prevents merging an OpenHands PR that has `rev` fields in `[tool.poetry.dependencies]`. The SDK must be released to PyPI first. |
+| **Flow A: stock agent-server is fine** | When only the Cloud API changes, `OpenHandsCloudWorkspace` talks to the Cloud server, not the agent-server. No custom image needed. |
+| **Flow B: agent-server image is required** | When the server needs new SDK code inside runtime containers, you must pin to the SDK PR's agent-server image. |

.github/dependabot.yml

@@ -4,7 +4,7 @@ updates:
     directory: "/"
     schedule:
       interval: "daily"
-    open-pull-requests-limit: 1
+    open-pull-requests-limit: 5
     groups:
       # put packages in their own group if they have a history of breaking the build or needing to be reverted
       pre-commit:
@@ -29,7 +29,7 @@ updates:
     directory: "/frontend"
     schedule:
       interval: "daily"
-    open-pull-requests-limit: 1
+    open-pull-requests-limit: 5
     groups:
       docusaurus:
         patterns:
@@ -51,7 +51,7 @@ updates:
     schedule:
       interval: "weekly"
       day: "wednesday"
-    open-pull-requests-limit: 1
+    open-pull-requests-limit: 5
     groups:
       docusaurus:
         patterns:
@@ -72,9 +72,11 @@ updates:
     directory: "/"
     schedule:
       interval: "weekly"
+    open-pull-requests-limit: 5
 
   - package-ecosystem: "docker"
     directories:
       - "containers/*"
     schedule:
       interval: "weekly"
+    open-pull-requests-limit: 5

.github/workflows/ghcr-build.yml

@@ -40,11 +40,11 @@ jobs:
         run: |
           if [[ "$GITHUB_EVENT_NAME" == "pull_request" ]]; then
             json=$(jq -n -c '[
-                { image: "nikolaik/python-nodejs:python3.12-nodejs22", tag: "nikolaik" }
+                { image: "nikolaik/python-nodejs:python3.12-nodejs22-slim", tag: "nikolaik" }
               ]')
           else
             json=$(jq -n -c '[
-                { image: "nikolaik/python-nodejs:python3.12-nodejs22", tag: "nikolaik" },
+                { image: "nikolaik/python-nodejs:python3.12-nodejs22-slim", tag: "nikolaik" },
                 { image: "ubuntu:24.04", tag: "ubuntu" }
               ]')
           fi
@@ -219,11 +219,9 @@ jobs:
       - name: Determine app image tag
         shell: bash
         run: |
-          # Duplicated with build.sh
-          sanitized_ref_name=$(echo "$GITHUB_REF_NAME" | sed 's/[^a-zA-Z0-9.-]\+/-/g')
-          OPENHANDS_BUILD_VERSION=$sanitized_ref_name
-          sanitized_ref_name=$(echo "$sanitized_ref_name" | tr '[:upper:]' '[:lower:]') # lower case is required in tagging
-          echo "OPENHANDS_DOCKER_TAG=${sanitized_ref_name}" >> $GITHUB_ENV
+          # Use the commit SHA to pin the exact app image built by ghcr_build_app,
+          # rather than a mutable branch tag like "main" which can serve stale cached layers.
+          echo "OPENHANDS_DOCKER_TAG=${RELEVANT_SHA}" >> $GITHUB_ENV
       - name: Build and push Docker image
         uses: useblacksmith/build-push-action@v1
         with:

.github/workflows/pr-artifacts.yml

@@ -0,0 +1,136 @@
+---
+name: PR Artifacts
+
+on:
+    workflow_dispatch: # Manual trigger for testing
+    pull_request:
+        types: [opened, synchronize, reopened]
+        branches: [main]
+    pull_request_review:
+        types: [submitted]
+
+jobs:
+  # Auto-remove .pr/ directory when a reviewer approves
+    cleanup-on-approval:
+        concurrency:
+            group: cleanup-pr-artifacts-${{ github.event.pull_request.number }}
+            cancel-in-progress: false
+        if: github.event_name == 'pull_request_review' && github.event.review.state == 'approved'
+        runs-on: ubuntu-latest
+        permissions:
+            contents: write
+            pull-requests: write
+        steps:
+            - name: Check if fork PR
+              id: check-fork
+              run: |
+                  if [ "${{ github.event.pull_request.head.repo.full_name }}" != "${{ github.event.pull_request.base.repo.full_name }}" ]; then
+                    echo "is_fork=true" >> $GITHUB_OUTPUT
+                    echo "::notice::Fork PR detected - skipping auto-cleanup (manual removal required)"
+                  else
+                    echo "is_fork=false" >> $GITHUB_OUTPUT
+                  fi
+
+            - uses: actions/checkout@v5
+              if: steps.check-fork.outputs.is_fork == 'false'
+              with:
+                  ref: ${{ github.event.pull_request.head.ref }}
+                  token: ${{ secrets.ALLHANDS_BOT_GITHUB_PAT }}
+
+            - name: Remove .pr/ directory
+              id: remove
+              if: steps.check-fork.outputs.is_fork == 'false'
+              run: |
+                  if [ -d ".pr" ]; then
+                    git config user.name "allhands-bot"
+                    git config user.email "allhands-bot@users.noreply.github.com"
+                    git rm -rf .pr/
+                    git commit -m "chore: Remove PR-only artifacts [automated]"
+                    git push || {
+                      echo "::error::Failed to push cleanup commit. Check branch protection rules."
+                      exit 1
+                    }
+                    echo "removed=true" >> $GITHUB_OUTPUT
+                    echo "::notice::Removed .pr/ directory"
+                  else
+                    echo "removed=false" >> $GITHUB_OUTPUT
+                    echo "::notice::No .pr/ directory to remove"
+                  fi
+
+            - name: Update PR comment after cleanup
+              if: steps.check-fork.outputs.is_fork == 'false' && steps.remove.outputs.removed == 'true'
+              uses: actions/github-script@v7
+              with:
+                  script: |
+                      const marker = '<!-- pr-artifacts-notice -->';
+                      const body = `${marker}
+                      ✅ **PR Artifacts Cleaned Up**
+
+                      The \`.pr/\` directory has been automatically removed.
+                      `;
+
+                      const { data: comments } = await github.rest.issues.listComments({
+                        owner: context.repo.owner,
+                        repo: context.repo.repo,
+                        issue_number: context.issue.number,
+                      });
+
+                      const existing = comments.find(c => c.body.includes(marker));
+                      if (existing) {
+                        await github.rest.issues.updateComment({
+                          owner: context.repo.owner,
+                          repo: context.repo.repo,
+                          comment_id: existing.id,
+                          body: body,
+                        });
+                      }
+
+  # Warn if .pr/ directory exists (will be auto-removed on approval)
+    check-pr-artifacts:
+        if: github.event_name == 'pull_request'
+        runs-on: ubuntu-latest
+        permissions:
+            contents: read
+            pull-requests: write
+        steps:
+            - uses: actions/checkout@v5
+
+            - name: Check for .pr/ directory
+              id: check
+              run: |
+                  if [ -d ".pr" ]; then
+                    echo "exists=true" >> $GITHUB_OUTPUT
+                    echo "::warning::.pr/ directory exists and will be automatically removed when the PR is approved. For fork PRs, manual removal is required before merging."
+                  else
+                    echo "exists=false" >> $GITHUB_OUTPUT
+                  fi
+
+            - name: Post or update PR comment
+              if: steps.check.outputs.exists == 'true'
+              uses: actions/github-script@v7
+              with:
+                  script: |
+                      const marker = '<!-- pr-artifacts-notice -->';
+                      const body = `${marker}
+                      📁 **PR Artifacts Notice**
+
+                      This PR contains a \`.pr/\` directory with PR-specific documents. This directory will be **automatically removed** when the PR is approved.
+
+                      > For fork PRs: Manual removal is required before merging.
+                      `;
+
+                      const { data: comments } = await github.rest.issues.listComments({
+                        owner: context.repo.owner,
+                        repo: context.repo.repo,
+                        issue_number: context.issue.number,
+                      });
+
+                      const existing = comments.find(c => c.body.includes(marker));
+                      if (!existing) {
+                        await github.rest.issues.createComment({
+                          owner: context.repo.owner,
+                          repo: context.repo.repo,
+                          issue_number: context.issue.number,
+                          body: body,
+                        });
+                      }

.gitignore

@@ -234,6 +234,8 @@ yarn-error.log*
 
 logs
 
+ralph/
+
 # agent
 .envrc
 /workspace

AGENTS.md

@@ -36,9 +36,45 @@ then re-run the command to ensure it passes. Common issues include:
 - Be especially careful with `git reset --hard` after staging files, as it will remove accidentally staged files
 - When remote has new changes, use `git fetch upstream && git rebase upstream/<branch>` on the same branch
 
+## PR-Specific Artifacts (`.pr/` directory)
+
+When working on a PR that requires design documents, scripts meant for development-only, or other temporary artifacts that should NOT be merged to main, store them in a `.pr/` directory at the repository root.
+
+### Usage
+
+```
+.pr/
+├── design.md       # Design decisions and architecture notes
+├── analysis.md     # Investigation or debugging notes
+├── logs/           # Test output or CI logs for reviewer reference
+└── notes.md        # Any other PR-specific content
+```
+
+### How It Works
+
+1. **Notification**: When `.pr/` exists, a comment is posted to the PR conversation alerting reviewers
+2. **Auto-cleanup**: When the PR is approved, the `.pr/` directory is automatically removed via `.github/workflows/pr-artifacts.yml`
+3. **Fork PRs**: Auto-cleanup cannot push to forks, so manual removal is required before merging
+
+### Important Notes
+
+- Do NOT put anything in `.pr/` that needs to be preserved after merge
+- The `.pr/` check passes (green ✅) during development — it only posts a notification, not a blocking error
+- For fork PRs: You must manually remove `.pr/` before the PR can be merged
+
+### When to Use
+
+- Complex refactoring that benefits from written design rationale
+- Debugging sessions where you want to document your investigation
+- E2E test results or logs that demonstrate a cross-repo feature works
+- Feature implementations that need temporary planning docs
+- Any analysis that helps reviewers understand the PR but isn't needed long-term
+
 ## Repository Structure
 Backend:
 - Located in the `openhands` directory
+- The current V1 application server lives in `openhands/app_server/`. `make start-backend` still launches `openhands.server.listen:app`, which includes the V1 routes by default unless `ENABLE_V1=0`.
+- For V1 web-app docs, LLM setup should point users to the Settings UI.
 - Testing:
   - All tests are in `tests/unit/test_*.py`
   - To test new code, run `poetry run pytest tests/unit/test_xxx.py` where `xxx` is the appropriate file for the current functionality
@@ -342,3 +378,30 @@ To add a new LLM model to OpenHands, you need to update multiple files across bo
 - Models appear in CLI provider selection based on the verified arrays
 - The `organize_models_and_providers` function groups models by provider
 - Default model selection prioritizes verified models for each provider
+
+### Sandbox Settings API (SDK Credential Inheritance)
+
+The sandbox settings API allows SDK-created conversations to inherit the user's SaaS credentials
+(LLM config, secrets) securely via `LookupSecret`. Raw secret values only flow SaaS→sandbox,
+never through the SDK client.
+
+#### User Credentials with Exposed Secrets (in `openhands/app_server/user/user_router.py`):
+- `GET /api/v1/users/me?expose_secrets=true` → Full user settings with unmasked secrets (e.g., `llm_api_key`)
+- `GET /api/v1/users/me` → Full user settings (secrets masked, Bearer only)
+
+Auth requirements for `expose_secrets=true`:
+- Bearer token (proves user identity via `OPENHANDS_API_KEY`)
+- `X-Session-API-Key` header (proves caller has an active sandbox owned by the authenticated user)
+
+Called by `workspace.get_llm()` in the SDK to retrieve LLM config with the API key.
+
+#### Sandbox-Scoped Secrets Endpoints (in `openhands/app_server/sandbox/sandbox_router.py`):
+- `GET /sandboxes/{id}/settings/secrets` → list secret names (no values)
+- `GET /sandboxes/{id}/settings/secrets/{name}` → raw secret value (called FROM sandbox)
+
+#### Auth: `X-Session-API-Key` header, validated via `SandboxService.get_sandbox_by_session_api_key()`
+
+#### Related SDK code (in `software-agent-sdk` repo):
+- `openhands/sdk/llm/llm.py`: `LLM.api_key` accepts `SecretSource` (including `LookupSecret`)
+- `openhands/workspace/cloud/workspace.py`: `get_llm()` and `get_secrets()` return LookupSecret-backed objects
+- Tests: `tests/sdk/llm/test_llm_secret_source_api_key.py`, `tests/workspace/test_cloud_workspace_sdk_settings.py`

CONTRIBUTING.md

@@ -1,83 +1,105 @@
 # Contributing
 
-Thanks for your interest in contributing to OpenHands! We welcome and appreciate contributions.
+Thanks for your interest in contributing to OpenHands! We're building the future of AI-powered software development, and we'd love for you to be part of this journey.
 
-## Understanding OpenHands's CodeBase
+## Our Vision
 
-To understand the codebase, please refer to the README in each module:
-- [frontend](./frontend/README.md)
-- [openhands](./openhands/README.md)
-   - [agenthub](./openhands/agenthub/README.md)
-   - [server](./openhands/server/README.md)
+The OpenHands community is built around the belief that AI and AI agents are going to fundamentally change the way we build software. If this is true, we should do everything we can to make sure that the benefits provided by such powerful technology are accessible to everyone.
 
-For benchmarks and evaluation, see the [OpenHands/benchmarks](https://github.com/OpenHands/benchmarks) repository.
+We believe in the power of open source to democratize access to cutting-edge AI technology. Just as the internet transformed how we share information, we envision a world where AI-powered development tools are available to every developer, regardless of their background or resources.
 
-## Setting up Your Development Environment
+## Getting Started
 
-We have a separate doc [Development.md](https://github.com/OpenHands/OpenHands/blob/main/Development.md) that tells
-you how to set up a development workflow.
+### Quick Ways to Contribute
 
-## How Can I Contribute?
+- **Use OpenHands** and [report issues](https://github.com/OpenHands/OpenHands/issues) you encounter
+- **Give feedback** using the thumbs-up/thumbs-down buttons after each session
+- **Star our repository** on [GitHub](https://github.com/OpenHands/OpenHands)
+- **Share OpenHands** with other developers
 
-There are many ways that you can contribute:
+### Set Up Your Development Environment
 
-1. **Download and use** OpenHands, and send [issues](https://github.com/OpenHands/OpenHands/issues) when you encounter something that isn't working or a feature that you'd like to see.
-2. **Send feedback** after each session by [clicking the thumbs-up thumbs-down buttons](https://docs.openhands.dev/usage/feedback), so we can see where things are working and failing, and also build an open dataset for training code agents.
-3. **Improve the Codebase** by sending [PRs](#sending-pull-requests-to-openhands) (see details below). In particular, we have some [good first issues](https://github.com/OpenHands/OpenHands/labels/good%20first%20issue) that may be ones to start on.
+- **Requirements**: Linux/Mac/WSL, Docker, Python 3.12, Node.js 22+, Poetry 1.8+
+- **Quick setup**: `make build`
+- **Run locally**: `make run`
+- **LLM setup (V1 web app)**: configure your model and API key in the Settings UI after the app starts
 
-## What Can I Build?
+Full details in our [Development Guide](./Development.md).
 
-Here are a few ways you can help improve the codebase.
+### Find Your First Issue
 
-#### UI/UX
+- Browse [good first issues](https://github.com/OpenHands/OpenHands/labels/good%20first%20issue)
+- Check our [project boards](https://github.com/OpenHands/OpenHands/projects) for organized tasks
+- Join our [Slack community](https://openhands.dev/joinslack) to ask what needs help
 
-We're always looking to improve the look and feel of the application. If you've got a small fix
-for something that's bugging you, feel free to open up a PR that changes the [`./frontend`](./frontend) directory.
+## Understanding the Codebase
 
-If you're looking to make a bigger change, add a new UI element, or significantly alter the style
-of the application, please open an issue first, or better, join the #dev-ui-ux channel in our Slack
-to gather consensus from our design team first.
+- **[Frontend](./frontend/README.md)** - React application
+- **[App Server (V1)](./openhands/app_server/README.md)** - Current FastAPI application server and REST API modules
+- **[Agents](./openhands/agenthub/README.md)** - AI agent implementations
+- **[Runtime](./openhands/runtime/README.md)** - Execution environments
+- **[Evaluation](https://github.com/OpenHands/benchmarks)** - Testing and benchmarks
 
-#### Improving the agent
+## What Can You Build?
 
-Our main agent is the CodeAct agent. You can [see its prompts here](https://github.com/OpenHands/OpenHands/tree/main/openhands/agenthub/codeact_agent).
+### Frontend & UI/UX
+- React & TypeScript development
+- UI/UX improvements
+- Mobile responsiveness
+- Component libraries
 
-Changes to these prompts, and to the underlying behavior in Python, can have a huge impact on user experience.
-You can try modifying the prompts to see how they change the behavior of the agent as you use the app
-locally, but we will need to do an end-to-end evaluation of any changes here to ensure that the agent
-is getting better over time.
+For bigger changes, join the #proj-gui channel in [Slack](https://openhands.dev/joinslack) first.
 
-We use the [SWE-bench](https://www.swebench.com/) benchmark to test our agent. You can join the #evaluation
-channel in Slack to learn more.
+### Agent Development
+- Prompt engineering
+- New agent types
+- Agent evaluation
+- Multi-agent systems
 
-#### Adding a new agent
+We use [SWE-bench](https://www.swebench.com/) to evaluate agents.
 
-You may want to experiment with building new types of agents. You can add an agent to [`openhands/agenthub`](./openhands/agenthub)
-to help expand the capabilities of OpenHands.
+### Backend & Infrastructure
+- Python development
+- Runtime systems (Docker containers, sandboxes)
+- Cloud integrations
+- Performance optimization
 
-#### Adding a new runtime
+### Testing & Quality Assurance
+- Unit testing
+- Integration testing
+- Bug hunting
+- Performance testing
 
-The agent needs a place to run code and commands. When you run OpenHands on your laptop, it uses a Docker container
-to do this by default. But there are other ways of creating a sandbox for the agent.
+### Documentation & Education
+- Technical documentation
+- Translation
+- Community support
 
-If you work for a company that provides a cloud-based runtime, you could help us add support for that runtime
-by implementing the [interface specified here](https://github.com/OpenHands/OpenHands/blob/main/openhands/runtime/base.py).
+## Pull Request Process
 
-#### Testing
+### Small Improvements
+- Quick review and approval
+- Ensure CI tests pass
+- Include clear description of changes
 
-When you write code, it is also good to write tests. Please navigate to the [`./tests`](./tests) folder to see existing
-test suites. At the moment, we have these kinds of tests: [`unit`](./tests/unit), [`runtime`](./tests/runtime), and [`end-to-end (e2e)`](./tests/e2e).
-Please refer to the README for each test suite. These tests also run on GitHub's continuous integration to ensure
-quality of the project.
+### Core Agent Changes
+These are evaluated based on:
+- **Accuracy** - Does it make the agent better at solving problems?
+- **Efficiency** - Does it improve speed or reduce resource usage?
+- **Code Quality** - Is the code maintainable and well-tested?
+
+Discuss major changes in [GitHub issues](https://github.com/OpenHands/OpenHands/issues) or [Slack](https://openhands.dev/joinslack) first.
 
 ## Sending Pull Requests to OpenHands
 
 You'll need to fork our repository to send us a Pull Request. You can learn more
 about how to fork a GitHub repo and open a PR with your changes in [this article](https://medium.com/swlh/forks-and-pull-requests-how-to-contribute-to-github-repos-8843fac34ce8).
 
-### Pull Request title
+You may also check out previous PRs in the [PR list](https://github.com/OpenHands/OpenHands/pulls).
 
-As described [here](https://github.com/commitizen/conventional-commit-types/blob/master/index.json), ideally a valid PR title should begin with one of the following prefixes:
+### Pull Request Title Format
+
+As described [here](https://github.com/commitizen/conventional-commit-types/blob/master/index.json), a valid PR title should begin with one of the following prefixes:
 
 - `feat`: A new feature
 - `fix`: A bug fix
@@ -95,45 +117,27 @@ For example, a PR title could be:
 - `refactor: modify package path`
 - `feat(frontend): xxxx`, where `(frontend)` means that this PR mainly focuses on the frontend component.
 
-You may also check out previous PRs in the [PR list](https://github.com/OpenHands/OpenHands/pulls).
+### Pull Request Description
 
-### Pull Request description
+- Explain what the PR does and why
+- Link to related issues
+- Include screenshots for UI changes
+- If your changes are user-facing (e.g. a new feature in the UI, a change in behavior, or a bugfix),
+  please include a short message that we can add to our changelog
 
-- If your PR is small (such as a typo fix), you can go brief.
-- If it contains a lot of changes, it's better to write more details.
+## Becoming a Maintainer
 
-If your changes are user-facing (e.g. a new feature in the UI, a change in behavior, or a bugfix)
-please include a short message that we can add to our changelog.
+For contributors who have made significant and sustained contributions to the project, there is a possibility of joining the maintainer team.
+The process for this is as follows:
 
-## How to Make Effective Contributions
+1. Any contributor who has made sustained and high-quality contributions to the codebase can be nominated by any maintainer. If you feel that you may qualify you can reach out to any of the maintainers that have reviewed your PRs and ask if you can be nominated.
+2. Once a maintainer nominates a new maintainer, there will be a discussion period among the maintainers for at least 3 days.
+3. If no concerns are raised the nomination will be accepted by acclamation, and if concerns are raised there will be a discussion and possible vote.
 
-### Opening Issues
+Note that just making many PRs does not immediately imply that you will become a maintainer. We will be looking at sustained high-quality contributions over a period of time, as well as good teamwork and adherence to our [Code of Conduct](./CODE_OF_CONDUCT.md).
 
-If you notice any bugs or have any feature requests please open them via the [issues page](https://github.com/OpenHands/OpenHands/issues). We will triage
-based on how critical the bug is or how potentially useful the improvement is, discuss, and implement the ones that
-the community has interest/effort for.
+## Need Help?
 
-Further, if you see an issue you like, please leave a "thumbs-up" or a comment, which will help us prioritize.
-
-### Making Pull Requests
-
-We're generally happy to consider all pull requests with the evaluation process varying based on the type of change:
-
-#### For Small Improvements
-
-Small improvements with few downsides are typically reviewed and approved quickly.
-One thing to check when making changes is to ensure that all continuous integration tests pass, which you can check
-before getting a review.
-
-#### For Core Agent Changes
-
-We need to be more careful with changes to the core agent, as it is imperative to maintain high quality. These PRs are
-evaluated based on three key metrics:
-
-1. **Accuracy**
-2. **Efficiency**
-3. **Code Complexity**
-
-If it improves accuracy, efficiency, or both with only a minimal change to code quality, that's great we're happy to merge it in!
-If there are bigger tradeoffs (e.g. helping efficiency a lot and hurting accuracy a little) we might want to put it behind a feature flag.
-Either way, please feel free to discuss on github issues or slack, and we will give guidance and preliminary feedback.
+- **Slack**: [Join our community](https://openhands.dev/joinslack)
+- **GitHub Issues**: [Open an issue](https://github.com/OpenHands/OpenHands/issues)
+- **Email**: contact@openhands.dev

Development.md

@@ -6,22 +6,196 @@ If you wish to contribute your changes, check out the
 on how to clone and setup the project initially before moving on. Otherwise,
 you can clone the OpenHands project directly.
 
-## Start the Server for Development
+## Choose Your Setup
 
-### 1. Requirements
+Select your operating system to see the specific setup instructions:
 
-- Linux, Mac OS, or [WSL on Windows](https://learn.microsoft.com/en-us/windows/wsl/install) [Ubuntu >= 22.04]
-- [Docker](https://docs.docker.com/engine/install/) (For those on MacOS, make sure to allow the default Docker socket to be used from advanced settings!)
-- [Python](https://www.python.org/downloads/) = 3.12
-- [NodeJS](https://nodejs.org/en/download/package-manager) >= 22.x
-- [Poetry](https://python-poetry.org/docs/#installing-with-the-official-installer) >= 1.8
-- OS-specific dependencies:
-  - Ubuntu: build-essential => `sudo apt-get install build-essential python3.12-dev`
-  - WSL: netcat => `sudo apt-get install netcat`
+- [macOS](#macos-setup)
+- [Linux](#linux-setup)
+- [Windows WSL](#windows-wsl-setup)
+- [Dev Container](#dev-container)
+- [Developing in Docker](#developing-in-docker)
+- [No sudo access?](#develop-without-sudo-access)
 
-Make sure you have all these dependencies installed before moving on to `make build`.
+---
 
-#### Dev container
+## macOS Setup
+
+### 1. Install Prerequisites
+
+You'll need the following installed:
+
+- **Python 3.12** — `brew install python@3.12` (see the [official Homebrew Python docs](https://docs.brew.sh/Homebrew-and-Python) for details). Make sure `python3.12` is available in your PATH (the `make build` step will verify this).
+- **Node.js >= 22** — `brew install node`
+- **Poetry >= 1.8** — `brew install poetry`
+- **Docker Desktop** — `brew install --cask docker`
+  - After installing, open Docker Desktop → **Settings → Advanced** → Enable **"Allow the default Docker socket to be used"**
+
+### 2. Build and Setup the Environment
+
+```bash
+make build
+```
+
+### 3. Configure the Language Model
+
+OpenHands supports a diverse array of Language Models (LMs) through the powerful [litellm](https://docs.litellm.ai) library.
+
+For the V1 web app, start OpenHands and configure your model and API key in the Settings UI.
+
+If you are running headless or CLI workflows, you can prepare local defaults with:
+
+```bash
+make setup-config
+```
+
+**Note on Alternative Models:**
+See [our documentation](https://docs.openhands.dev/usage/llms) for recommended models.
+
+### 4. Run the Application
+
+```bash
+# Run both backend and frontend
+make run
+
+# Or run separately:
+make start-backend  # Backend only on port 3000
+make start-frontend # Frontend only on port 3001
+```
+
+These targets serve the current OpenHands V1 API by default. In the codebase, `make start-backend` runs `openhands.server.listen:app`, and that app includes the `openhands/app_server` V1 routes unless `ENABLE_V1=0`.
+
+---
+
+## Linux Setup
+
+This guide covers Ubuntu/Debian. For other distributions, adapt the package manager commands accordingly.
+
+### 1. Install Prerequisites
+
+```bash
+# Update package list
+sudo apt update
+
+# Install system dependencies
+sudo apt install -y build-essential curl netcat software-properties-common
+
+# Install Python 3.12
+# Ubuntu 24.04+ and Debian 13+ ship with Python 3.12 — skip the PPA step if
+# python3.12 --version already works on your system.
+# The deadsnakes PPA is Ubuntu-only and needed for Ubuntu 22.04 or older:
+sudo add-apt-repository -y ppa:deadsnakes/ppa
+sudo apt update
+sudo apt install -y python3.12 python3.12-dev python3.12-venv
+
+# Install Node.js 22.x
+curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
+sudo apt install -y nodejs
+
+# Install Poetry
+curl -sSL https://install.python-poetry.org | python3 -
+
+# Add Poetry to your PATH
+echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.bashrc
+source ~/.bashrc
+
+# Install Docker
+# Follow the official guide: https://docs.docker.com/engine/install/ubuntu/
+# Quick version:
+sudo install -m 0755 -d /etc/apt/keyrings
+curl -fsSL https://download.docker.com/linux/ubuntu/gpg -o /etc/apt/keyrings/docker.asc
+sudo chmod a+r /etc/apt/keyrings/docker.asc
+echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu $(. /etc/os-release && echo "$VERSION_CODENAME") stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
+sudo apt update
+sudo apt install -y docker-ce docker-ce-cli containerd.io docker-buildx-plugin docker-compose-plugin
+sudo usermod -aG docker $USER
+# Log out and back in for Docker group changes to take effect
+```
+
+### 2. Build and Setup the Environment
+
+```bash
+make build
+```
+
+### 3. Configure the Language Model
+
+See the [macOS section above](#3-configure-the-language-model) for guidance: configure your model and API key in the Settings UI.
+
+### 4. Run the Application
+
+```bash
+# Run both backend and frontend
+make run
+
+# Or run separately:
+make start-backend  # Backend only on port 3000
+make start-frontend # Frontend only on port 3001
+```
+
+---
+
+## Windows WSL Setup
+
+WSL2 with Ubuntu is recommended. The setup is similar to Linux, with a few WSL-specific considerations.
+
+### 1. Install WSL2
+
+**Option A: Windows 11 (Microsoft Store)**
+The easiest way on Windows 11:
+1. Open the **Microsoft Store** app
+2. Search for **"Ubuntu 22.04 LTS"** or **"Ubuntu"**
+3. Click **Install**
+4. Launch Ubuntu from the Start menu
+
+**Option B: PowerShell**
+```powershell
+# Run this in PowerShell as Administrator
+wsl --install -d Ubuntu-22.04
+```
+
+After installation, restart your computer and open Ubuntu.
+
+### 2. Install Prerequisites (in WSL Ubuntu)
+
+Follow [Step 1 from the Linux setup](#1-install-prerequisites-1) to install system dependencies, Python 3.12, Node.js, and Poetry. Skip the Docker installation — Docker is provided through Docker Desktop below.
+
+### 3. Configure Docker for WSL2
+
+1. Install [Docker Desktop for Windows](https://www.docker.com/products/docker-desktop)
+2. Open Docker Desktop > Settings > General
+3. Enable: "Use the WSL 2 based engine"
+4. Go to Settings > Resources > WSL Integration
+5. Enable integration with your Ubuntu distribution
+
+**Important:** Keep your project files in the WSL filesystem (e.g., `~/workspace/openhands`), not in `/mnt/c`. Files accessed via `/mnt/c` will be significantly slower.
+
+### 4. Build and Setup the Environment
+
+```bash
+make build
+```
+
+### 5. Configure the Language Model
+
+See the [macOS section above](#3-configure-the-language-model) for the current V1 guidance: configure your model and API key in the Settings UI for the web app, and use `make setup-config` only for headless or CLI workflows.
+
+### 6. Run the Application
+
+```bash
+# Run both backend and frontend
+make run
+
+# Or run separately:
+make start-backend  # Backend only on port 3000
+make start-frontend # Frontend only on port 3001
+```
+
+Access the frontend at `http://localhost:3001` from your Windows browser.
+
+---
+
+## Dev Container
 
 There is a [dev container](https://containers.dev/) available which provides a
 pre-configured environment with all the necessary dependencies installed if you
@@ -32,7 +206,38 @@ extension installed, you can open the project in a dev container by using the
 _Dev Container: Reopen in Container_ command from the Command Palette
 (Ctrl+Shift+P).
 
-#### Develop without sudo access
+---
+
+## Developing in Docker
+
+If you don't want to install dependencies on your host machine, you can develop inside a Docker container.
+
+### Quick Start
+
+```bash
+make docker-dev
+```
+
+For more details, see the [dev container documentation](./containers/dev/README.md).
+
+### Alternative: Docker Run
+
+If you just want to run OpenHands without setting up a dev environment:
+
+```bash
+make docker-run
+```
+
+If you don't have `make` installed, run:
+
+```bash
+cd ./containers/dev
+./dev.sh
+```
+
+---
+
+## Develop without sudo access
 
 If you want to develop without system admin/sudo access to upgrade/install `Python` and/or `NodeJS`, you can use
 `conda` or `mamba` to manage the packages for you:
@@ -48,159 +253,90 @@ mamba install conda-forge::nodejs
 mamba install conda-forge::poetry
 ```
 
-### 2. Build and Setup The Environment
+---
 
-Begin by building the project which includes setting up the environment and installing dependencies. This step ensures
-that OpenHands is ready to run on your system:
+## Running OpenHands with OpenHands
+
+You can use OpenHands to develop and improve OpenHands itself!
+
+### Quick Start
 
 ```bash
-make build
+export INSTALL_DOCKER=0
+export RUNTIME=local
+make build && make run
 ```
 
-### 3. Configuring the Language Model
+Access the interface at:
+- Local development: http://localhost:3001
+- Remote/cloud environments: Use the appropriate external URL
 
-OpenHands supports a diverse array of Language Models (LMs) through the powerful [litellm](https://docs.litellm.ai) library.
+For external access:
+```bash
+make run FRONTEND_PORT=12000 FRONTEND_HOST=0.0.0.0 BACKEND_HOST=0.0.0.0
+```
 
-To configure the LM of your choice, run:
+---
+
+## LLM Debugging
+
+If you encounter issues with the Language Model, enable debug logging:
 
 ```bash
-make setup-config
+export DEBUG=1
+# Restart the backend
+make start-backend
 ```
 
-This command will prompt you to enter the LLM API key, model name, and other variables ensuring that OpenHands is
-tailored to your specific needs. Note that the model name will apply only when you run headless. If you use the UI,
-please set the model in the UI.
+Logs will be saved to `logs/llm/CURRENT_DATE/` for troubleshooting.
 
-Note: If you have previously run OpenHands using the docker command, you may have already set some environment
-variables in your terminal. The final configurations are set from highest to lowest priority:
-Environment variables > config.toml variables > default variables
+---
 
-**Note on Alternative Models:**
-See [our documentation](https://docs.openhands.dev/usage/llms) for recommended models.
+## Testing
 
-### 4. Running the application
-
-#### Option A: Run the Full Application
-
-Once the setup is complete, this command starts both the backend and frontend servers, allowing you to interact with OpenHands:
-
-```bash
-make run
-```
-
-#### Option B: Individual Server Startup
-
-- **Start the Backend Server:** If you prefer, you can start the backend server independently to focus on
-  backend-related tasks or configurations.
-
-  ```bash
-  make start-backend
-  ```
-
-- **Start the Frontend Server:** Similarly, you can start the frontend server on its own to work on frontend-related
-  components or interface enhancements.
-  ```bash
-  make start-frontend
-  ```
-
-### 5. Running OpenHands with OpenHands
-
-You can use OpenHands to develop and improve OpenHands itself! This is a powerful way to leverage AI assistance for contributing to the project.
-
-#### Quick Start
-
-1. **Build and run OpenHands:**
-
-   ```bash
-   export INSTALL_DOCKER=0
-   export RUNTIME=local
-   make build && make run
-   ```
-
-2. **Access the interface:**
-
-   - Local development: http://localhost:3001
-   - Remote/cloud environments: Use the appropriate external URL
-
-3. **Configure for external access (if needed):**
-   ```bash
-   # For external access (e.g., cloud environments)
-   make run FRONTEND_PORT=12000 FRONTEND_HOST=0.0.0.0 BACKEND_HOST=0.0.0.0
-   ```
-
-### 6. LLM Debugging
-
-If you encounter any issues with the Language Model (LM) or you're simply curious, export DEBUG=1 in the environment and restart the backend.
-OpenHands will log the prompts and responses in the logs/llm/CURRENT_DATE directory, allowing you to identify the causes.
-
-### 7. Help
-
-Need help or info on available targets and commands? Use the help command for all the guidance you need with OpenHands.
-
-```bash
-make help
-```
-
-### 8. Testing
-
-To run tests, refer to the following:
-
-#### Unit tests
+### Unit Tests
 
 ```bash
 poetry run pytest ./tests/unit/test_*.py
 ```
 
-### 9. Add or update dependency
+---
 
-1. Add your dependency in `pyproject.toml` or use `poetry add xxx`.
-2. Update the poetry.lock file via `poetry lock --no-update`.
+## Adding Dependencies
 
-### 10. Use existing Docker image
+1. Add your dependency in `pyproject.toml` or use `poetry add xxx`
+2. Update the lock file: `poetry lock --no-update`
 
-To reduce build time (e.g., if no changes were made to the client-runtime component), you can use an existing Docker
-container image by setting the SANDBOX_RUNTIME_CONTAINER_IMAGE environment variable to the desired Docker image.
+---
 
-Example: `export SANDBOX_RUNTIME_CONTAINER_IMAGE=ghcr.io/openhands/runtime:1.2-nikolaik`
+## Using Existing Docker Images
 
-## Develop inside Docker container
-
-TL;DR
+To reduce build time, you can use an existing runtime image:
 
 ```bash
-make docker-dev
+export SANDBOX_RUNTIME_CONTAINER_IMAGE=ghcr.io/openhands/runtime:1.2-nikolaik
 ```
 
-See more details [here](./containers/dev/README.md).
+---
 
-If you are just interested in running `OpenHands` without installing all the required tools on your host.
+## Help
 
 ```bash
-make docker-run
+make help
 ```
 
-If you do not have `make` on your host, run:
-
-```bash
-cd ./containers/dev
-./dev.sh
-```
-
-You do need [Docker](https://docs.docker.com/engine/install/) installed on your host though.
+---
 
 ## Key Documentation Resources
 
-Here's a guide to the important documentation files in the repository:
-
 - [/README.md](./README.md): Main project overview, features, and basic setup instructions
 - [/Development.md](./Development.md) (this file): Comprehensive guide for developers working on OpenHands
 - [/CONTRIBUTING.md](./CONTRIBUTING.md): Guidelines for contributing to the project, including code style and PR process
 - [DOC_STYLE_GUIDE.md](https://github.com/OpenHands/docs/blob/main/openhands/DOC_STYLE_GUIDE.md): Standards for writing and maintaining project documentation
-- [/openhands/README.md](./openhands/README.md): Details about the backend Python implementation
+- [/openhands/app_server/README.md](./openhands/app_server/README.md): Current V1 application server implementation and REST API modules
 - [/frontend/README.md](./frontend/README.md): Frontend React application setup and development guide
 - [/containers/README.md](./containers/README.md): Information about Docker containers and deployment
 - [/tests/unit/README.md](./tests/unit/README.md): Guide to writing and running unit tests
 - [OpenHands/benchmarks](https://github.com/OpenHands/benchmarks): Documentation for the evaluation framework and benchmarks
 - [/skills/README.md](./skills/README.md): Information about the skills architecture and implementation
-- [/openhands/server/README.md](./openhands/server/README.md): Server implementation details and API documentation
 - [/openhands/runtime/README.md](./openhands/runtime/README.md): Documentation for the runtime environment and execution model

config.template.toml

@@ -296,7 +296,7 @@ classpath = "my_package.my_module.MyCustomAgent"
 #user_id = 1000
 
 # Container image to use for the sandbox
-#base_container_image = "nikolaik/python-nodejs:python3.12-nodejs22"
+#base_container_image = "nikolaik/python-nodejs:python3.12-nodejs22-slim"
 
 # Use host network
 #use_host_network = false

containers/app/Dockerfile

@@ -1,5 +1,5 @@
 ARG OPENHANDS_BUILD_VERSION=dev
-FROM node:25.2-trixie-slim AS frontend-builder
+FROM node:25.8-trixie-slim AS frontend-builder
 
 WORKDIR /app
 
@@ -22,7 +22,7 @@ ENV POETRY_NO_INTERACTION=1 \
 
 RUN apt-get update -y \
     && apt-get install -y curl make git build-essential jq gettext \
-    && python3 -m pip install poetry --break-system-packages
+    && python3 -m pip install "poetry>=2.3.0" --break-system-packages
 
 COPY pyproject.toml poetry.lock ./
 RUN touch README.md

enterprise/Dockerfile

@@ -10,7 +10,7 @@ LABEL com.datadoghq.tags.env="${DD_ENV}"
 # Apply security updates to fix CVEs
 RUN apt-get update && \
     apt-get install -y curl && \
-    curl -fsSL https://deb.nodesource.com/setup_20.x | bash - && \
+    curl -fsSL https://deb.nodesource.com/setup_24.x | bash - && \
     apt-get install -y nodejs && \
     apt-get install -y jq gettext && \
     # Apply security updates for packages with available fixes
@@ -33,7 +33,8 @@ RUN cd /tmp/enterprise && \
     # Export only main dependencies with hashes for supply chain security
     /app/.venv/bin/poetry export --only main -o requirements.txt && \
     # Remove the local path dependency (openhands-ai is already in base image)
-    sed -i '/^-e /d; /openhands-ai/d' requirements.txt && \
+    # and git-based SDK dependencies (already installed via the base app image)
+    sed -i '/^-e /d; /openhands-ai/d; /^openhands-.*@ git+/d' requirements.txt && \
     # Install pinned dependencies from lock file
     /app/.venv/bin/pip install -r requirements.txt && \
     # Cleanup - return to /app before removing /tmp/enterprise

enterprise/README.md

@@ -51,6 +51,6 @@ NOTE: in the future we will simply replace the `GithubTokenManager` with keycloa
 ## User ID vs User Token
 
 - In OpenHands, the entire app revolves around the GitHub token the user sets. `openhands/server` uses `request.state.github_token` for the entire app
-- On Enterprise, the entire APP resolves around the Github User ID. This is because the cookie sets it, so `openhands/server` AND `enterprise/server` depend on it and completly ignore `request.state.github_token` (token is fetched from `GithubTokenManager` instead)
+- On Enterprise, the entire APP resolves around the Github User ID. This is because the cookie sets it, so `openhands/server` AND `enterprise/server` depend on it and completely ignore `request.state.github_token` (token is fetched from `GithubTokenManager` instead)
 
 Note that introducing GitHub User ID in OpenHands, for instance, will cause large breakages.

enterprise/doc/architecture/README.md

@@ -0,0 +1,13 @@
+# Enterprise Architecture Documentation
+
+Architecture diagrams specific to the OpenHands SaaS/Enterprise deployment.
+
+## Documentation
+
+- [Authentication Flow](./authentication.md) - Keycloak-based authentication for SaaS deployment
+- [External Integrations](./external-integrations.md) - GitHub, Slack, Jira, and other service integrations
+
+## Related Documentation
+
+For core OpenHands architecture (applicable to all deployments), see:
+- [Core Architecture Documentation](../../../openhands/architecture/README.md)

enterprise/doc/architecture/authentication.md

@@ -0,0 +1,58 @@
+# Authentication Flow (SaaS Deployment)
+
+OpenHands uses Keycloak for identity management in the SaaS deployment. The authentication flow involves multiple services:
+
+```mermaid
+sequenceDiagram
+    autonumber
+    participant User as User (Browser)
+    participant App as App Server
+    participant KC as Keycloak
+    participant IdP as Identity Provider<br/>(GitHub, Google, etc.)
+    participant DB as User Database
+
+    Note over User,DB: OAuth 2.0 / OIDC Authentication Flow
+
+    User->>App: Access OpenHands
+    App->>User: Redirect to Keycloak
+    User->>KC: Login request
+    KC->>User: Show login options
+    User->>KC: Select provider (e.g., GitHub)
+    KC->>IdP: OAuth redirect
+    User->>IdP: Authenticate
+    IdP-->>KC: OAuth callback + tokens
+    Note over KC: Create/update user session
+    KC-->>User: Redirect with auth code
+    User->>App: Auth code
+    App->>KC: Exchange code for tokens
+    KC-->>App: Access token + Refresh token
+    Note over App: Create signed JWT cookie
+    App->>DB: Store/update user record
+    App-->>User: Set keycloak_auth cookie
+
+    Note over User,DB: Subsequent Requests
+
+    User->>App: Request with cookie
+    Note over App: Verify JWT signature
+    App->>KC: Validate token (if needed)
+    KC-->>App: Token valid
+    Note over App: Extract user context
+    App-->>User: Authorized response
+```
+
+### Authentication Components
+
+| Component | Purpose | Location |
+|-----------|---------|----------|
+| **Keycloak** | Identity provider, SSO, token management | External service |
+| **UserAuth** | Abstract auth interface | `openhands/server/user_auth/user_auth.py` |
+| **SaasUserAuth** | Keycloak implementation | `enterprise/server/auth/saas_user_auth.py` |
+| **JWT Service** | Token signing/verification | `openhands/app_server/services/jwt_service.py` |
+| **Auth Routes** | Login/logout endpoints | `enterprise/server/routes/auth.py` |
+
+### Token Flow
+
+1. **Keycloak Access Token**: Short-lived token for API access
+2. **Keycloak Refresh Token**: Long-lived token to obtain new access tokens
+3. **Signed JWT Cookie**: App Server's session cookie containing encrypted Keycloak tokens
+4. **Provider Tokens**: OAuth tokens for GitHub, GitLab, etc. (stored separately for git operations)

enterprise/doc/architecture/external-integrations.md

@@ -0,0 +1,88 @@
+# External Integrations
+
+OpenHands integrates with external services (GitHub, Slack, Jira, etc.) through webhook-based event handling:
+
+```mermaid
+sequenceDiagram
+    autonumber
+    participant Ext as External Service<br/>(GitHub/Slack/Jira)
+    participant App as App Server
+    participant IntRouter as Integration Router
+    participant Manager as Integration Manager
+    participant Conv as Conversation Service
+    participant Sandbox as Sandbox
+
+    Note over Ext,Sandbox: Webhook Event Flow (e.g., GitHub Issue Created)
+
+    Ext->>App: POST /api/integration/{service}/events
+    App->>IntRouter: Route to service handler
+    Note over IntRouter: Verify signature (HMAC)
+
+    IntRouter->>Manager: Parse event payload
+    Note over Manager: Extract context (repo, issue, user)
+    Note over Manager: Map external user → OpenHands user
+
+    Manager->>Conv: Create conversation (with issue context)
+    Conv->>Sandbox: Provision sandbox
+    Sandbox-->>Conv: Ready
+
+    Manager->>Sandbox: Start agent with task
+
+    Note over Ext,Sandbox: Agent Works on Task...
+
+    Sandbox-->>Manager: Task complete
+    Manager->>Ext: POST result<br/>(PR, comment, etc.)
+
+    Note over Ext,Sandbox: Callback Flow (Agent → External Service)
+
+    Sandbox->>App: Webhook callback<br/>/api/v1/webhooks
+    App->>Manager: Process callback
+    Manager->>Ext: Update external service
+```
+
+### Supported Integrations
+
+| Integration | Trigger Events | Agent Actions |
+|-------------|----------------|---------------|
+| **GitHub** | Issue created, PR opened, @mention | Create PR, comment, push commits |
+| **GitLab** | Issue created, MR opened | Create MR, comment, push commits |
+| **Slack** | @mention in channel | Reply in thread, create tasks |
+| **Jira** | Issue created/updated | Update ticket, add comments |
+| **Linear** | Issue created | Update status, add comments |
+
+### Integration Components
+
+| Component | Purpose | Location |
+|-----------|---------|----------|
+| **Integration Routes** | Webhook endpoints per service | `enterprise/server/routes/integration/` |
+| **Integration Managers** | Business logic per service | `enterprise/integrations/{service}/` |
+| **Token Manager** | Store/retrieve OAuth tokens | `enterprise/server/auth/token_manager.py` |
+| **Callback Processor** | Handle agent → service updates | `enterprise/integrations/{service}/*_callback_processor.py` |
+
+### Integration Authentication
+
+```
+External Service (e.g., GitHub)
+        │
+        ▼
+┌─────────────────────────────────┐
+│ GitHub App Installation         │
+│ - Webhook secret for signature  │
+│ - App private key for API calls │
+└─────────────────────────────────┘
+        │
+        ▼
+┌─────────────────────────────────┐
+│ User Account Linking            │
+│ - Keycloak user ID              │
+│ - GitHub user ID                │
+│ - Stored OAuth tokens           │
+└─────────────────────────────────┘
+        │
+        ▼
+┌─────────────────────────────────┐
+│ Agent Execution                 │
+│ - Uses linked tokens for API    │
+│ - Can push, create PRs, comment │
+└─────────────────────────────────┘
+```

enterprise/integrations/github/github_solvability.py

@@ -106,16 +106,18 @@ async def summarize_issue_solvability(
             f'Solvability analysis disabled for user {github_view.user_info.user_id}'
         )
 
-    if user_settings.llm_api_key is None:
+    agent_settings = user_settings.agent_settings
+    llm_settings = agent_settings.llm
+    if llm_settings.api_key is None:
         raise ValueError(
             f'[Solvability] No LLM API key found for user {github_view.user_info.user_id}'
         )
 
     try:
         llm_config = LLMConfig(
-            model=user_settings.llm_model,
-            api_key=user_settings.llm_api_key.get_secret_value(),
-            base_url=user_settings.llm_base_url,
+            model=llm_settings.model,
+            api_key=llm_settings.api_key.get_secret_value(),
+            base_url=llm_settings.base_url,
         )
     except ValidationError as e:
         raise ValueError(

enterprise/integrations/github/github_v1_callback_processor.py

@@ -43,15 +43,20 @@ class GithubV1CallbackProcessor(EventCallbackProcessor):
         event: Event,
     ) -> EventCallbackResult | None:
         """Process events for GitHub V1 integration."""
-        # Only handle ConversationStateUpdateEvent
+        # Only handle ConversationStateUpdateEvent for execution_status
         if not isinstance(event, ConversationStateUpdateEvent):
             return None
 
-        # Only act when execution has finished
-        if not (event.key == 'execution_status' and event.value == 'finished'):
+        if event.key != 'execution_status':
             return None
 
+        # Log ALL terminal states for monitoring (finished, error, stuck)
         _logger.info('[GitHub V1] Callback agent state was %s', event)
+
+        # Only request summary when execution has finished successfully
+        if event.value != 'finished':
+            return None
+
         _logger.info(
             '[GitHub V1] Should request summary: %s', self.should_request_summary
         )

enterprise/integrations/gitlab/gitlab_v1_callback_processor.py

@@ -41,15 +41,20 @@ class GitlabV1CallbackProcessor(EventCallbackProcessor):
         event: Event,
     ) -> EventCallbackResult | None:
         """Process events for GitLab V1 integration."""
-        # Only handle ConversationStateUpdateEvent
+        # Only handle ConversationStateUpdateEvent for execution_status
         if not isinstance(event, ConversationStateUpdateEvent):
             return None
 
-        # Only act when execution has finished
-        if not (event.key == 'execution_status' and event.value == 'finished'):
+        if event.key != 'execution_status':
             return None
 
+        # Log ALL terminal states for monitoring (finished, error, stuck)
         _logger.info('[GitLab V1] Callback agent state was %s', event)
+
+        # Only request summary when execution has finished successfully
+        if event.value != 'finished':
+            return None
+
         _logger.info(
             '[GitLab V1] Should request summary: %s', self.should_request_summary
         )

enterprise/integrations/resolver_context.py

@@ -60,7 +60,9 @@ class ResolverUserContext(UserContext):
                 return provider_token.token.get_secret_value()
         return None
 
-    async def get_provider_tokens(self) -> PROVIDER_TOKEN_TYPE | None:
+    async def get_provider_tokens(
+        self, as_env_vars: bool = False
+    ) -> PROVIDER_TOKEN_TYPE | dict[str, str] | None:
         return await self.saas_user_auth.get_provider_tokens()
 
     async def get_secrets(self) -> dict[str, SecretSource]:

enterprise/integrations/slack/slack_v1_callback_processor.py

@@ -40,16 +40,20 @@ class SlackV1CallbackProcessor(EventCallbackProcessor):
         event: Event,
     ) -> EventCallbackResult | None:
         """Process events for Slack V1 integration."""
-        # Only handle ConversationStateUpdateEvent
+        # Only handle ConversationStateUpdateEvent for execution_status
         if not isinstance(event, ConversationStateUpdateEvent):
             return None
 
-        # Only act when execution has finished
-        if not (event.key == 'execution_status' and event.value == 'finished'):
+        if event.key != 'execution_status':
             return None
 
+        # Log ALL terminal states for monitoring (finished, error, stuck)
         _logger.info('[Slack V1] Callback agent state was %s', event)
 
+        # Only request summary when execution has finished successfully
+        if event.value != 'finished':
+            return None
+
         try:
             summary = await self._request_summary(conversation_id)
             await self._post_summary_to_slack(summary)

enterprise/integrations/stripe_service.py

@@ -100,27 +100,25 @@ async def has_payment_method_by_user_id(user_id: str) -> bool:
     return bool(payment_methods.data)
 
 
-async def migrate_customer(user_id: str, org: Org):
-    async with a_session_maker() as session:
-        result = await session.execute(
-            select(StripeCustomer).where(StripeCustomer.keycloak_user_id == user_id)
-        )
-        stripe_customer = result.scalar_one_or_none()
-        if stripe_customer is None:
-            return
-        stripe_customer.org_id = org.id
-        customer = await stripe.Customer.modify_async(
-            id=stripe_customer.stripe_customer_id,
-            email=org.contact_email,
-            metadata={'user_id': '', 'org_id': str(org.id)},
-        )
+async def migrate_customer(session, user_id: str, org: Org):
+    result = await session.execute(
+        select(StripeCustomer).where(StripeCustomer.keycloak_user_id == user_id)
+    )
+    stripe_customer = result.scalar_one_or_none()
+    if stripe_customer is None:
+        return
+    stripe_customer.org_id = org.id
+    customer = await stripe.Customer.modify_async(
+        id=stripe_customer.stripe_customer_id,
+        email=org.contact_email,
+        metadata={'user_id': '', 'org_id': str(org.id)},
+    )
 
-        logger.info(
-            'migrated_customer',
-            extra={
-                'user_id': user_id,
-                'org_id': str(org.id),
-                'stripe_customer_id': customer.id,
-            },
-        )
-        await session.commit()
+    logger.info(
+        'migrated_customer',
+        extra={
+            'user_id': user_id,
+            'org_id': str(org.id),
+            'stripe_customer_id': customer.id,
+        },
+    )

enterprise/migrations/versions/100_add_sandbox_grouping_strategy.py

@@ -0,0 +1,33 @@
+"""Add sandbox_grouping_strategy column to user, org, and user_settings tables.
+
+Revision ID: 100
+Revises: 099
+Create Date: 2025-03-12
+"""
+
+import sqlalchemy as sa
+from alembic import op
+
+revision = '100'
+down_revision = '099'
+
+
+def upgrade() -> None:
+    op.add_column(
+        'user',
+        sa.Column('sandbox_grouping_strategy', sa.String, nullable=True),
+    )
+    op.add_column(
+        'org',
+        sa.Column('sandbox_grouping_strategy', sa.String, nullable=True),
+    )
+    op.add_column(
+        'user_settings',
+        sa.Column('sandbox_grouping_strategy', sa.String, nullable=True),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column('user_settings', 'sandbox_grouping_strategy')
+    op.drop_column('org', 'sandbox_grouping_strategy')
+    op.drop_column('user', 'sandbox_grouping_strategy')

enterprise/migrations/versions/101_add_pending_messages_table.py

@@ -0,0 +1,39 @@
+"""Add pending_messages table for server-side message queuing
+
+Revision ID: 101
+Revises: 100
+Create Date: 2025-03-15 00:00:00.000000
+
+"""
+
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = '101'
+down_revision: Union[str, None] = '100'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    """Create pending_messages table for storing messages before conversation is ready.
+
+    Messages are stored temporarily until the conversation becomes ready, then
+    delivered and deleted regardless of success or failure.
+    """
+    op.create_table(
+        'pending_messages',
+        sa.Column('id', sa.String(), primary_key=True),
+        sa.Column('conversation_id', sa.String(), nullable=False, index=True),
+        sa.Column('role', sa.String(20), nullable=False, server_default='user'),
+        sa.Column('content', sa.JSON, nullable=False),
+        sa.Column('created_at', sa.DateTime(timezone=True), nullable=False),
+    )
+
+
+def downgrade() -> None:
+    """Remove pending_messages table."""
+    op.drop_table('pending_messages')

enterprise/migrations/versions/102_add_agent_settings_to_enterprise_settings.py

@@ -0,0 +1,296 @@
+"""Add agent_settings columns to enterprise settings tables.
+
+Revision ID: 102
+Revises: 101
+Create Date: 2026-03-22 00:00:00.000000
+
+"""
+
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = '102'
+down_revision: Union[str, None] = '101'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+_EMPTY_JSON = sa.text("'{}'::json")
+
+
+def upgrade() -> None:
+    op.add_column(
+        'user_settings',
+        sa.Column(
+            'agent_settings', sa.JSON(), nullable=False, server_default=_EMPTY_JSON
+        ),
+    )
+    op.add_column(
+        'org_member',
+        sa.Column(
+            'agent_settings', sa.JSON(), nullable=False, server_default=_EMPTY_JSON
+        ),
+    )
+    op.add_column(
+        'org',
+        sa.Column(
+            'agent_settings', sa.JSON(), nullable=False, server_default=_EMPTY_JSON
+        ),
+    )
+
+    op.execute(
+        sa.text(
+            """
+            UPDATE user_settings
+            SET agent_settings = jsonb_strip_nulls(
+                jsonb_build_object(
+                    'schema_version', 1,
+                    'agent', agent,
+                    'llm.model', llm_model,
+                    'llm.base_url', llm_base_url,
+                    'verification.confirmation_mode', confirmation_mode,
+                    'verification.security_analyzer', security_analyzer,
+                    'condenser.enabled', enable_default_condenser,
+                    'condenser.max_size', condenser_max_size,
+                    'max_iterations', max_iterations
+                ) || COALESCE(agent_settings::jsonb, '{}'::jsonb)
+            )::json
+            """
+        )
+    )
+    op.execute(
+        sa.text(
+            """
+            UPDATE org_member
+            SET agent_settings = jsonb_strip_nulls(
+                jsonb_build_object(
+                    'schema_version', 1,
+                    'llm.model', llm_model,
+                    'llm.base_url', llm_base_url,
+                    'max_iterations', max_iterations
+                ) || COALESCE(agent_settings::jsonb, '{}'::jsonb)
+            )::json
+            """
+        )
+    )
+    op.execute(
+        sa.text(
+            """
+            UPDATE org
+            SET agent_settings = jsonb_strip_nulls(
+                jsonb_build_object(
+                    'schema_version', 1,
+                    'agent', agent,
+                    'llm.model', default_llm_model,
+                    'llm.base_url', default_llm_base_url,
+                    'verification.confirmation_mode', confirmation_mode,
+                    'verification.security_analyzer', security_analyzer,
+                    'condenser.enabled', enable_default_condenser,
+                    'condenser.max_size', condenser_max_size,
+                    'max_iterations', default_max_iterations,
+                    'mcp_config', mcp_config
+                ) || COALESCE(agent_settings::jsonb, '{}'::jsonb)
+            )::json
+            """
+        )
+    )
+
+    op.execute(
+        sa.text(
+            """
+            UPDATE user_settings AS us
+            SET llm_api_key_for_byor = om._llm_api_key_for_byor
+            FROM "user" AS u
+            JOIN org_member AS om
+              ON om.user_id = u.id
+             AND om.org_id = u.current_org_id
+            WHERE us.keycloak_user_id = u.id::text
+              AND us.llm_api_key_for_byor IS NULL
+              AND om._llm_api_key_for_byor IS NOT NULL
+            """
+        )
+    )
+    op.execute(
+        sa.text(
+            """
+            INSERT INTO user_settings (keycloak_user_id, llm_api_key_for_byor, agent_settings)
+            SELECT u.id::text, om._llm_api_key_for_byor, '{}'::json
+            FROM "user" AS u
+            JOIN org_member AS om
+              ON om.user_id = u.id
+             AND om.org_id = u.current_org_id
+            LEFT JOIN user_settings AS us
+              ON us.keycloak_user_id = u.id::text
+            WHERE us.id IS NULL
+              AND om._llm_api_key_for_byor IS NOT NULL
+            """
+        )
+    )
+
+    op.alter_column('user_settings', 'agent_settings', server_default=None)
+    op.alter_column('org_member', 'agent_settings', server_default=None)
+    op.alter_column('org', 'agent_settings', server_default=None)
+    op.drop_column('user_settings', 'agent')
+    op.drop_column('user_settings', 'max_iterations')
+    op.drop_column('user_settings', 'security_analyzer')
+    op.drop_column('user_settings', 'confirmation_mode')
+    op.drop_column('user_settings', 'llm_model')
+    op.drop_column('user_settings', 'llm_base_url')
+    op.drop_column('user_settings', 'enable_default_condenser')
+    op.drop_column('user_settings', 'condenser_max_size')
+    op.drop_column('org_member', 'max_iterations')
+    op.drop_column('org_member', 'llm_model')
+    op.drop_column('org_member', '_llm_api_key_for_byor')
+    op.drop_column('org_member', 'llm_base_url')
+    op.drop_column('org', 'agent')
+    op.drop_column('org', 'default_max_iterations')
+    op.drop_column('org', 'security_analyzer')
+    op.drop_column('org', 'confirmation_mode')
+    op.drop_column('org', 'default_llm_model')
+    op.drop_column('org', 'default_llm_base_url')
+    op.drop_column('org', 'enable_default_condenser')
+    op.drop_column('org', 'mcp_config')
+    op.drop_column('org', 'condenser_max_size')
+
+
+def downgrade() -> None:
+    op.add_column('user_settings', sa.Column('agent', sa.String(), nullable=True))
+    op.add_column(
+        'user_settings', sa.Column('max_iterations', sa.Integer(), nullable=True)
+    )
+    op.add_column(
+        'user_settings', sa.Column('security_analyzer', sa.String(), nullable=True)
+    )
+    op.add_column(
+        'user_settings', sa.Column('confirmation_mode', sa.Boolean(), nullable=True)
+    )
+    op.add_column('user_settings', sa.Column('llm_model', sa.String(), nullable=True))
+    op.add_column(
+        'user_settings', sa.Column('llm_base_url', sa.String(), nullable=True)
+    )
+    op.add_column(
+        'user_settings',
+        sa.Column(
+            'enable_default_condenser',
+            sa.Boolean(),
+            nullable=False,
+            server_default=sa.true(),
+        ),
+    )
+    op.add_column(
+        'user_settings', sa.Column('condenser_max_size', sa.Integer(), nullable=True)
+    )
+    op.add_column('org_member', sa.Column('llm_base_url', sa.String(), nullable=True))
+    op.add_column(
+        'org_member', sa.Column('_llm_api_key_for_byor', sa.String(), nullable=True)
+    )
+    op.add_column('org_member', sa.Column('llm_model', sa.String(), nullable=True))
+    op.add_column(
+        'org_member', sa.Column('max_iterations', sa.Integer(), nullable=True)
+    )
+    op.add_column('org', sa.Column('agent', sa.String(), nullable=True))
+    op.add_column(
+        'org', sa.Column('default_max_iterations', sa.Integer(), nullable=True)
+    )
+    op.add_column('org', sa.Column('security_analyzer', sa.String(), nullable=True))
+    op.add_column('org', sa.Column('confirmation_mode', sa.Boolean(), nullable=True))
+    op.add_column('org', sa.Column('default_llm_model', sa.String(), nullable=True))
+    op.add_column('org', sa.Column('default_llm_base_url', sa.String(), nullable=True))
+    op.add_column(
+        'org',
+        sa.Column(
+            'enable_default_condenser',
+            sa.Boolean(),
+            nullable=False,
+            server_default=sa.true(),
+        ),
+    )
+    op.add_column('org', sa.Column('mcp_config', sa.JSON(), nullable=True))
+    op.add_column('org', sa.Column('condenser_max_size', sa.Integer(), nullable=True))
+
+    op.execute(
+        sa.text(
+            """
+            UPDATE user_settings
+            SET
+                agent = agent_settings ->> 'agent',
+                max_iterations = NULLIF(agent_settings ->> 'max_iterations', '')::integer,
+                security_analyzer =
+                    agent_settings ->> 'verification.security_analyzer',
+                confirmation_mode = CASE
+                    WHEN agent_settings::jsonb ? 'verification.confirmation_mode'
+                    THEN (agent_settings ->> 'verification.confirmation_mode')::boolean
+                    ELSE NULL
+                END,
+                llm_model = agent_settings ->> 'llm.model',
+                llm_base_url = agent_settings ->> 'llm.base_url',
+                enable_default_condenser = CASE
+                    WHEN agent_settings::jsonb ? 'condenser.enabled'
+                    THEN (agent_settings ->> 'condenser.enabled')::boolean
+                    ELSE TRUE
+                END,
+                condenser_max_size =
+                    NULLIF(agent_settings ->> 'condenser.max_size', '')::integer
+            """
+        )
+    )
+    op.execute(
+        sa.text(
+            """
+            UPDATE org_member
+            SET
+                llm_model = agent_settings ->> 'llm.model',
+                llm_base_url = agent_settings ->> 'llm.base_url',
+                max_iterations = NULLIF(agent_settings ->> 'max_iterations', '')::integer
+            """
+        )
+    )
+    op.execute(
+        sa.text(
+            """
+            UPDATE org
+            SET
+                agent = agent_settings ->> 'agent',
+                default_max_iterations =
+                    NULLIF(agent_settings ->> 'max_iterations', '')::integer,
+                security_analyzer =
+                    agent_settings ->> 'verification.security_analyzer',
+                confirmation_mode = CASE
+                    WHEN agent_settings::jsonb ? 'verification.confirmation_mode'
+                    THEN (agent_settings ->> 'verification.confirmation_mode')::boolean
+                    ELSE NULL
+                END,
+                default_llm_model = agent_settings ->> 'llm.model',
+                default_llm_base_url = agent_settings ->> 'llm.base_url',
+                enable_default_condenser = CASE
+                    WHEN agent_settings::jsonb ? 'condenser.enabled'
+                    THEN (agent_settings ->> 'condenser.enabled')::boolean
+                    ELSE TRUE
+                END,
+                mcp_config = agent_settings -> 'mcp_config',
+                condenser_max_size =
+                    NULLIF(agent_settings ->> 'condenser.max_size', '')::integer
+            """
+        )
+    )
+    op.execute(
+        sa.text(
+            """
+            UPDATE org_member AS om
+            SET _llm_api_key_for_byor = us.llm_api_key_for_byor
+            FROM "user" AS u
+            JOIN user_settings AS us
+              ON us.keycloak_user_id = u.id::text
+            WHERE om.user_id = u.id
+              AND om.org_id = u.current_org_id
+              AND us.llm_api_key_for_byor IS NOT NULL
+            """
+        )
+    )
+
+    op.drop_column('org', 'agent_settings')
+    op.drop_column('org_member', 'agent_settings')
+    op.drop_column('user_settings', 'agent_settings')

enterprise/migrations/versions/102_add_disabled_skills_to_user_settings.py

@@ -0,0 +1,28 @@
+"""Add disabled_skills to user_settings.
+
+Revision ID: 102
+Revises: 101
+Create Date: 2026-02-25
+
+"""
+
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = '102'
+down_revision: Union[str, None] = '101'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        'user_settings', sa.Column('disabled_skills', sa.JSON(), nullable=True)
+    )
+
+
+def downgrade() -> None:
+    op.drop_column('user_settings', 'disabled_skills')

enterprise/migrations/versions/103_add_mcp_config_to_org_member.py

@@ -0,0 +1,41 @@
+"""Add mcp_config to org_member for user-specific MCP settings.
+
+Revision ID: 103
+Revises: 102
+Create Date: 2026-03-26
+
+"""
+
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = '103'
+down_revision: Union[str, None] = '102'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column('org_member', sa.Column('mcp_config', sa.JSON(), nullable=True))
+
+    # Migrate existing org-level MCP configs to all members in each org.
+    # This preserves existing configurations while transitioning to user-specific settings.
+    conn = op.get_bind()
+    orgs_with_config = conn.execute(
+        sa.text('SELECT id, mcp_config FROM org WHERE mcp_config IS NOT NULL')
+    ).fetchall()
+
+    for org_id, mcp_config in orgs_with_config:
+        conn.execute(
+            sa.text(
+                'UPDATE org_member SET mcp_config = :config WHERE org_id = :org_id'
+            ),
+            {'config': mcp_config, 'org_id': str(org_id)},
+        )
+
+
+def downgrade() -> None:
+    op.drop_column('org_member', 'mcp_config')

enterprise/pyproject.toml

@@ -64,6 +64,7 @@ pytest-asyncio = "*"
 pytest-forked = "*"
 pytest-xdist = "*"
 flake8 = "*"
+freezegun = "^1.5.1"
 openai = "*"
 opencv-python = "*"
 pandas = "*"

enterprise/run_automation_executor.py

@@ -1,71 +0,0 @@
-"""Entry point for the automation executor.
-
-Usage: python -m run_automation_executor
-
-This runs as a Kubernetes Deployment (long-running). It polls the automation_events
-inbox, matches events to automations, claims and executes runs, and monitors
-conversation completion.
-
-Environment variables:
-    OPENHANDS_API_URL        Base URL for the V1 API (default: http://openhands-service:3000)
-    MAX_CONCURRENT_RUNS      Max concurrent runs per executor (default: 5)
-    RUN_TIMEOUT_SECONDS      Max time for a single run (default: 7200)
-    POLL_INTERVAL_SECONDS    Fallback poll interval (default: 30)
-    HEARTBEAT_INTERVAL_SECONDS  Heartbeat update interval (default: 60)
-"""
-
-import asyncio
-import logging
-import signal
-import sys
-
-logger = logging.getLogger('saas.automation.executor')
-
-
-def _setup_logging() -> None:
-    """Configure logging, deferring to enterprise logger if available."""
-    try:
-        from server.logger import setup_all_loggers
-
-        setup_all_loggers()
-    except ImportError:
-        logging.basicConfig(
-            level=logging.INFO,
-            format='%(asctime)s %(name)s %(levelname)s %(message)s',
-            stream=sys.stdout,
-        )
-
-
-def _install_signal_handlers(loop: asyncio.AbstractEventLoop) -> None:
-    """Install signal handlers for graceful shutdown."""
-    from services.automation_executor import request_shutdown
-
-    def _handle_signal(signum: int, _frame: object) -> None:
-        sig_name = signal.Signals(signum).name
-        logger.info('Received %s, initiating graceful shutdown...', sig_name)
-        request_shutdown()
-
-    for sig in (signal.SIGTERM, signal.SIGINT):
-        signal.signal(sig, _handle_signal)
-
-
-async def main() -> None:
-    from services.automation_executor import executor_main
-
-    await executor_main()
-
-
-if __name__ == '__main__':
-    _setup_logging()
-
-    loop = asyncio.new_event_loop()
-    _install_signal_handlers(loop)
-
-    logger.info('Starting automation executor')
-    try:
-        loop.run_until_complete(main())
-    except KeyboardInterrupt:
-        logger.info('Interrupted by user')
-    finally:
-        loop.close()
-        logger.info('Automation executor process exiting')

enterprise/saas_server.py

@@ -46,6 +46,7 @@ from server.routes.org_invitations import (  # noqa: E402
 )
 from server.routes.orgs import org_router  # noqa: E402
 from server.routes.readiness import readiness_router  # noqa: E402
+from server.routes.service import service_router  # noqa: E402
 from server.routes.user import saas_user_router  # noqa: E402
 from server.routes.user_app_settings import user_app_settings_router  # noqa: E402
 from server.sharing.shared_conversation_router import (  # noqa: E402
@@ -112,6 +113,7 @@ if GITLAB_APP_CLIENT_ID:
     base_app.include_router(gitlab_integration_router)
 
 base_app.include_router(api_keys_router)  # Add routes for API key management
+base_app.include_router(service_router)  # Add routes for internal service API
 base_app.include_router(org_router)  # Add routes for organization management
 base_app.include_router(
     verified_models_router

enterprise/server/auth/authorization.py

@@ -35,7 +35,7 @@ Usage:
 from enum import Enum
 from uuid import UUID
 
-from fastapi import Depends, HTTPException, status
+from fastapi import Depends, HTTPException, Request, status
 from storage.org_member_store import OrgMemberStore
 from storage.role import Role
 from storage.role_store import RoleStore
@@ -214,6 +214,19 @@ def has_permission(user_role: Role, permission: Permission) -> bool:
     return permission in permissions
 
 
+async def get_api_key_org_id_from_request(request: Request) -> UUID | None:
+    """Get the org_id bound to the API key used for authentication.
+
+    Returns None if:
+    - Not authenticated via API key (cookie auth)
+    - API key is a legacy key without org binding
+    """
+    user_auth = getattr(request.state, 'user_auth', None)
+    if user_auth and hasattr(user_auth, 'get_api_key_org_id'):
+        return user_auth.get_api_key_org_id()
+    return None
+
+
 def require_permission(permission: Permission):
     """
     Factory function that creates a dependency to require a specific permission.
@@ -221,8 +234,9 @@ def require_permission(permission: Permission):
     This creates a FastAPI dependency that:
     1. Extracts org_id from the path parameter
     2. Gets the authenticated user_id
-    3. Checks if the user has the required permission in the organization
-    4. Returns the user_id if authorized, raises HTTPException otherwise
+    3. Validates API key org binding (if using API key auth)
+    4. Checks if the user has the required permission in the organization
+    5. Returns the user_id if authorized, raises HTTPException otherwise
 
     Usage:
         @router.get('/{org_id}/settings')
@@ -240,6 +254,7 @@ def require_permission(permission: Permission):
     """
 
     async def permission_checker(
+        request: Request,
         org_id: UUID | None = None,
         user_id: str | None = Depends(get_user_id),
     ) -> str:
@@ -249,6 +264,23 @@ def require_permission(permission: Permission):
                 detail='User not authenticated',
             )
 
+        # Validate API key organization binding
+        api_key_org_id = await get_api_key_org_id_from_request(request)
+        if api_key_org_id is not None and org_id is not None:
+            if api_key_org_id != org_id:
+                logger.warning(
+                    'API key organization mismatch',
+                    extra={
+                        'user_id': user_id,
+                        'api_key_org_id': str(api_key_org_id),
+                        'target_org_id': str(org_id),
+                    },
+                )
+                raise HTTPException(
+                    status_code=status.HTTP_403_FORBIDDEN,
+                    detail='API key is not authorized for this organization',
+                )
+
         user_role = await get_user_org_role(user_id, org_id)
 
         if not user_role:

enterprise/server/auth/saas_user_auth.py

@@ -1,6 +1,7 @@
 import time
 from dataclasses import dataclass
 from types import MappingProxyType
+from uuid import UUID
 
 import jwt
 from fastapi import Request
@@ -59,6 +60,19 @@ class SaasUserAuth(UserAuth):
     _secrets: Secrets | None = None
     accepted_tos: bool | None = None
     auth_type: AuthType = AuthType.COOKIE
+    # API key context fields - populated when authenticated via API key
+    api_key_org_id: UUID | None = None  # Org bound to the API key used for auth
+    api_key_id: int | None = None
+    api_key_name: str | None = None
+
+    def get_api_key_org_id(self) -> UUID | None:
+        """Get the organization ID bound to the API key used for authentication.
+
+        Returns:
+            The org_id if authenticated via API key with org binding, None otherwise
+            (cookie auth or legacy API keys without org binding).
+        """
+        return self.api_key_org_id
 
     async def get_user_id(self) -> str | None:
         return self.user_id
@@ -283,14 +297,19 @@ async def saas_user_auth_from_bearer(request: Request) -> SaasUserAuth | None:
             return None
 
         api_key_store = ApiKeyStore.get_instance()
-        user_id = await api_key_store.validate_api_key(api_key)
-        if not user_id:
+        validation_result = await api_key_store.validate_api_key(api_key)
+        if not validation_result:
             return None
-        offline_token = await token_manager.load_offline_token(user_id)
+        offline_token = await token_manager.load_offline_token(
+            validation_result.user_id
+        )
         saas_user_auth = SaasUserAuth(
-            user_id=user_id,
+            user_id=validation_result.user_id,
             refresh_token=SecretStr(offline_token),
             auth_type=AuthType.BEARER,
+            api_key_org_id=validation_result.org_id,
+            api_key_id=validation_result.key_id,
+            api_key_name=validation_result.key_name,
         )
         await saas_user_auth.refresh()
         return saas_user_auth

enterprise/server/constants.py

@@ -77,6 +77,9 @@ PERMITTED_CORS_ORIGINS = [
     )
 ]
 
+# Controls whether new orgs/users default to V1 API (env: DEFAULT_V1_ENABLED)
+DEFAULT_V1_ENABLED = os.getenv('DEFAULT_V1_ENABLED', '1').lower() in ('1', 'true')
+
 
 def build_litellm_proxy_model_path(model_name: str) -> str:
     """Build the LiteLLM proxy model path based on model name.

enterprise/server/logger.py

@@ -84,6 +84,8 @@ def setup_json_logger(
         style='{',
         rename_fields={'levelname': 'severity'},
         json_serializer=custom_json_serializer,
+        # Use 'ts' for consistency with LOG_JSON_FOR_CONSOLE mode (skip when console mode to avoid duplicates)
+        timestamp='ts' if not LOG_JSON_FOR_CONSOLE else False,
     )
 
     handler.setFormatter(formatter)

enterprise/server/middleware.py

@@ -12,11 +12,8 @@ from server.auth.auth_error import (
 )
 from server.auth.gitlab_sync import schedule_gitlab_repo_sync
 from server.auth.saas_user_auth import SaasUserAuth, token_manager
-from server.routes.auth import (
-    get_cookie_domain,
-    get_cookie_samesite,
-    set_response_cookie,
-)
+from server.routes.auth import set_response_cookie
+from server.utils.url_utils import get_cookie_domain, get_cookie_samesite
 
 from openhands.core.logger import openhands_logger as logger
 from openhands.server.user_auth.user_auth import AuthType, UserAuth, get_user_auth
@@ -93,8 +90,8 @@ class SetAuthCookieMiddleware:
             if keycloak_auth_cookie:
                 response.delete_cookie(
                     key='keycloak_auth',
-                    domain=get_cookie_domain(request),
-                    samesite=get_cookie_samesite(request),
+                    domain=get_cookie_domain(),
+                    samesite=get_cookie_samesite(),
                 )
             return response
 
@@ -185,6 +182,10 @@ class SetAuthCookieMiddleware:
         if path.startswith('/api/v1/webhooks/'):
             return False
 
+        # Service API uses its own authentication (X-Service-API-Key header)
+        if path.startswith('/api/service/'):
+            return False
+
         is_mcp = path.startswith('/mcp')
         is_api_route = path.startswith('/api')
         return is_api_route or is_mcp

enterprise/server/routes/api_keys.py

@@ -1,55 +1,49 @@
 from datetime import UTC, datetime
+from typing import cast
 
-from fastapi import APIRouter, Depends, HTTPException, status
+from fastapi import APIRouter, Depends, HTTPException, Request, status
 from pydantic import BaseModel, field_validator
+from server.auth.saas_user_auth import SaasUserAuth
+from sqlalchemy import select
 from storage.api_key import ApiKey
 from storage.api_key_store import ApiKeyStore
+from storage.database import a_session_maker
 from storage.lite_llm_manager import LiteLlmManager
-from storage.org_member import OrgMember
-from storage.org_member_store import OrgMemberStore
 from storage.org_service import OrgService
+from storage.user_settings import UserSettings
 from storage.user_store import UserStore
 
 from openhands.core.logger import openhands_logger as logger
-from openhands.server.user_auth import get_user_id
+from openhands.server.user_auth import get_user_auth, get_user_id
+from openhands.server.user_auth.user_auth import AuthType
 
 
 # Helper functions for BYOR API key management
 async def get_byor_key_from_db(user_id: str) -> str | None:
     """Get the BYOR key from the database for a user."""
-    user = await UserStore.get_user_by_id(user_id)
-    if not user:
-        return None
+    async with a_session_maker() as session:
+        result = await session.execute(
+            select(UserSettings).filter(UserSettings.keycloak_user_id == user_id)
+        )
+        user_settings = result.scalars().first()
 
-    current_org_id = user.current_org_id
-    current_org_member: OrgMember | None = None
-    for org_member in user.org_members:
-        if org_member.org_id == current_org_id:
-            current_org_member = org_member
-            break
-    if not current_org_member:
-        return None
-    if current_org_member.llm_api_key_for_byor:
-        return current_org_member.llm_api_key_for_byor.get_secret_value()
-    return None
+    byor_key = user_settings.llm_api_key_for_byor_secret if user_settings else None
+    return byor_key.get_secret_value() if byor_key else None
 
 
 async def store_byor_key_in_db(user_id: str, key: str) -> None:
     """Store the BYOR key in the database for a user."""
-    user = await UserStore.get_user_by_id(user_id)
-    if not user:
-        return None
+    async with a_session_maker() as session:
+        result = await session.execute(
+            select(UserSettings).filter(UserSettings.keycloak_user_id == user_id)
+        )
+        user_settings = result.scalars().first()
+        if not user_settings:
+            user_settings = UserSettings(keycloak_user_id=user_id)
+            session.add(user_settings)
 
-    current_org_id = user.current_org_id
-    current_org_member: OrgMember | None = None
-    for org_member in user.org_members:
-        if org_member.org_id == current_org_id:
-            current_org_member = org_member
-            break
-    if not current_org_member:
-        return None
-    current_org_member.llm_api_key_for_byor = key
-    await OrgMemberStore.update_org_member(current_org_member)
+        user_settings.llm_api_key_for_byor_secret = key
+        await session.commit()
 
 
 async def generate_byor_key(user_id: str) -> str | None:
@@ -150,6 +144,16 @@ class MessageResponse(BaseModel):
     message: str
 
 
+class CurrentApiKeyResponse(BaseModel):
+    """Response model for the current API key endpoint."""
+
+    id: int
+    name: str | None
+    org_id: str
+    user_id: str
+    auth_type: str
+
+
 def api_key_to_response(key: ApiKey) -> ApiKeyResponse:
     """Convert an ApiKey model to an ApiKeyResponse."""
     return ApiKeyResponse(
@@ -262,6 +266,46 @@ async def delete_api_key(
         )
 
 
+@api_router.get('/current', tags=['Keys'])
+async def get_current_api_key(
+    request: Request,
+    user_id: str = Depends(get_user_id),
+) -> CurrentApiKeyResponse:
+    """Get information about the currently authenticated API key.
+
+    This endpoint returns metadata about the API key used for the current request,
+    including the org_id associated with the key. This is useful for API key
+    callers who need to know which organization context their key operates in.
+
+    Returns 400 if not authenticated via API key (e.g., using cookie auth).
+    """
+    user_auth = await get_user_auth(request)
+
+    # Check if authenticated via API key
+    if user_auth.get_auth_type() != AuthType.BEARER:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail='This endpoint requires API key authentication. Not available for cookie-based auth.',
+        )
+
+    # In SaaS context, bearer auth always produces SaasUserAuth
+    saas_user_auth = cast(SaasUserAuth, user_auth)
+
+    if saas_user_auth.api_key_org_id is None:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail='This API key was created before organization support. Please regenerate your API key to use this endpoint.',
+        )
+
+    return CurrentApiKeyResponse(
+        id=saas_user_auth.api_key_id,
+        name=saas_user_auth.api_key_name,
+        org_id=str(saas_user_auth.api_key_org_id),
+        user_id=user_id,
+        auth_type=saas_user_auth.auth_type.value,
+    )
+
+
 @api_router.get('/llm/byor', tags=['Keys'])
 async def get_llm_api_key_for_byor(
     user_id: str = Depends(get_user_id),

enterprise/server/routes/auth.py

@@ -3,7 +3,7 @@ import json
 import uuid
 import warnings
 from datetime import datetime, timezone
-from typing import Annotated, Literal, Optional, cast
+from typing import Annotated, Optional, cast
 from urllib.parse import quote, urlencode
 from uuid import UUID as parse_uuid
 
@@ -27,7 +27,7 @@ from server.auth.user.user_authorizer import (
     depends_user_authorizer,
 )
 from server.config import sign_token
-from server.constants import IS_FEATURE_ENV
+from server.constants import IS_FEATURE_ENV, IS_LOCAL_ENV
 from server.routes.event_webhook import _get_session_api_key, _get_user_id
 from server.services.org_invitation_service import (
     EmailMismatchError,
@@ -37,12 +37,12 @@ from server.services.org_invitation_service import (
     UserAlreadyMemberError,
 )
 from server.utils.rate_limit_utils import check_rate_limit_by_user_id
+from server.utils.url_utils import get_cookie_domain, get_cookie_samesite, get_web_url
 from sqlalchemy import select
 from storage.database import a_session_maker
 from storage.user import User
 from storage.user_store import UserStore
 
-from openhands.app_server.config import get_global_config
 from openhands.core.logger import openhands_logger as logger
 from openhands.integrations.provider import ProviderHandler
 from openhands.integrations.service_types import ProviderType, TokenResponse
@@ -77,7 +77,7 @@ def set_response_cookie(
     signed_token = sign_token(cookie_data, config.jwt_secret.get_secret_value())  # type: ignore
 
     # Set secure cookie with signed token
-    domain = get_cookie_domain(request)
+    domain = get_cookie_domain()
     if domain:
         response.set_cookie(
             key='keycloak_auth',
@@ -85,7 +85,7 @@ def set_response_cookie(
             domain=domain,
             httponly=True,
             secure=secure,
-            samesite=get_cookie_samesite(request),
+            samesite=get_cookie_samesite(),
         )
     else:
         response.set_cookie(
@@ -93,30 +93,10 @@ def set_response_cookie(
             value=signed_token,
             httponly=True,
             secure=secure,
-            samesite=get_cookie_samesite(request),
+            samesite=get_cookie_samesite(),
         )
 
 
-def get_cookie_domain(request: Request) -> str | None:
-    # for now just use the full hostname except for staging stacks.
-    return (
-        None
-        if not request.url.hostname
-        or request.url.hostname.endswith('staging.all-hands.dev')
-        else request.url.hostname
-    )
-
-
-def get_cookie_samesite(request: Request) -> Literal['lax', 'strict']:
-    # for localhost and feature/staging stacks we set it to 'lax' as the cookie domain won't allow 'strict'
-    return (
-        'lax'
-        if request.url.hostname == 'localhost'
-        or (request.url.hostname or '').endswith('staging.all-hands.dev')
-        else 'strict'
-    )
-
-
 def _extract_oauth_state(state: str | None) -> tuple[str, str | None, str | None]:
     """Extract redirect URL, reCAPTCHA token, and invitation token from OAuth state.
 
@@ -140,19 +120,6 @@ def _extract_oauth_state(state: str | None) -> tuple[str, str | None, str | None
         return state, None, None
 
 
-# Keep alias for backward compatibility
-def _extract_recaptcha_state(state: str | None) -> tuple[str, str | None]:
-    """Extract redirect URL and reCAPTCHA token from OAuth state.
-
-    Deprecated: Use _extract_oauth_state instead.
-
-    Returns:
-        Tuple of (redirect_url, recaptcha_token). Token may be None.
-    """
-    redirect_url, recaptcha_token, _ = _extract_oauth_state(state)
-    return redirect_url, recaptcha_token
-
-
 @oauth_router.get('/keycloak/callback')
 async def keycloak_callback(
     request: Request,
@@ -183,10 +150,7 @@ async def keycloak_callback(
             detail='Missing code in request params',
         )
 
-    web_url = get_global_config().web_url
-    if not web_url:
-        scheme = 'http' if request.url.hostname == 'localhost' else 'https'
-        web_url = f'{scheme}://{request.url.netloc}'
+    web_url = get_web_url(request)
     redirect_uri = web_url + request.url.path
 
     (
@@ -208,6 +172,23 @@ async def keycloak_callback(
 
     authorization = await user_authorizer.authorize_user(user_info)
     if not authorization.success:
+        # For duplicate_email errors, clean up the newly created Keycloak user
+        # (only if they're not already in our UserStore, i.e., they're a new user)
+        if authorization.error_detail == 'duplicate_email':
+            try:
+                existing_user = await UserStore.get_user_by_id(user_info.sub)
+                if not existing_user:
+                    # New user created during OAuth should be deleted from Keycloak
+                    await token_manager.delete_keycloak_user(user_info.sub)
+                    logger.info(
+                        f'Deleted orphaned Keycloak user {user_info.sub} '
+                        'after duplicate_email rejection'
+                    )
+            except Exception as e:
+                # Log but don't fail - user should still get 401 response
+                logger.warning(
+                    f'Failed to clean up orphaned Keycloak user {user_info.sub}: {e}'
+                )
         # Return unauthorized
         raise HTTPException(
             status_code=status.HTTP_401_UNAUTHORIZED,
@@ -313,7 +294,9 @@ async def keycloak_callback(
             else:
                 raise
 
-        verification_redirect_url = f'{request.base_url}login?email_verification_required=true&user_id={user_id}'
+        verification_redirect_url = (
+            f'{web_url}/login?email_verification_required=true&user_id={user_id}'
+        )
         if rate_limited:
             verification_redirect_url = f'{verification_redirect_url}&rate_limited=true'
 
@@ -474,9 +457,7 @@ async def keycloak_callback(
     # If the user hasn't accepted the TOS, redirect to the TOS page
     if not has_accepted_tos:
         encoded_redirect_url = quote(redirect_url, safe='')
-        tos_redirect_url = (
-            f'{request.base_url}accept-tos?redirect_url={encoded_redirect_url}'
-        )
+        tos_redirect_url = f'{web_url}/accept-tos?redirect_url={encoded_redirect_url}'
         if invitation_token:
             tos_redirect_url = f'{tos_redirect_url}&invitation_success=true'
         response = RedirectResponse(tos_redirect_url, status_code=302)
@@ -508,10 +489,9 @@ async def keycloak_offline_callback(code: str, state: str, request: Request):
             status_code=status.HTTP_400_BAD_REQUEST,
             content={'error': 'Missing code in request params'},
         )
-    scheme = 'https'
-    if request.url.hostname == 'localhost':
-        scheme = 'http'
-    redirect_uri = f'{scheme}://{request.url.netloc}{request.url.path}'
+
+    web_url = get_web_url(request)
+    redirect_uri = web_url + request.url.path
     logger.debug(f'code: {code}, redirect_uri: {redirect_uri}')
 
     (
@@ -533,15 +513,14 @@ async def keycloak_offline_callback(code: str, state: str, request: Request):
     )
 
     redirect_url, _, _ = _extract_oauth_state(state)
-    return RedirectResponse(
-        redirect_url if redirect_url else request.base_url, status_code=302
-    )
+    return RedirectResponse(redirect_url if redirect_url else web_url, status_code=302)
 
 
 @oauth_router.get('/github/callback')
 async def github_dummy_callback(request: Request):
     """Callback for GitHub that just forwards the user to the app base URL."""
-    return RedirectResponse(request.base_url, status_code=302)
+    web_url = get_web_url(request)
+    return RedirectResponse(web_url, status_code=302)
 
 
 @api_router.post('/authenticate')
@@ -563,8 +542,8 @@ async def authenticate(request: Request):
         if keycloak_auth_cookie:
             response.delete_cookie(
                 key='keycloak_auth',
-                domain=get_cookie_domain(request),
-                samesite=get_cookie_samesite(request),
+                domain=get_cookie_domain(),
+                samesite=get_cookie_samesite(),
             )
 
         return response
@@ -588,7 +567,8 @@ async def accept_tos(request: Request):
 
     # Get redirect URL from request body
     body = await request.json()
-    redirect_url = body.get('redirect_url', str(request.base_url))
+    web_url = get_web_url(request)
+    redirect_url = body.get('redirect_url', str(web_url))
 
     # Update user settings with TOS acceptance
     accepted_tos: datetime = datetime.now(timezone.utc).replace(tzinfo=None)
@@ -618,7 +598,7 @@ async def accept_tos(request: Request):
         response=response,
         keycloak_access_token=access_token.get_secret_value(),
         keycloak_refresh_token=refresh_token.get_secret_value(),
-        secure=False if request.url.hostname == 'localhost' else True,
+        secure=not IS_LOCAL_ENV,
         accepted_tos=True,
     )
     return response
@@ -635,8 +615,8 @@ async def logout(request: Request):
     # Always delete the cookie regardless of what happens
     response.delete_cookie(
         key='keycloak_auth',
-        domain=get_cookie_domain(request),
-        samesite=get_cookie_samesite(request),
+        domain=get_cookie_domain(),
+        samesite=get_cookie_samesite(),
     )
 
     # Try to properly logout from Keycloak, but don't fail if it doesn't work

enterprise/server/routes/billing.py

@@ -11,8 +11,8 @@ from integrations import stripe_service
 from pydantic import BaseModel
 from server.constants import STRIPE_API_KEY
 from server.logger import logger
+from server.utils.url_utils import get_web_url
 from sqlalchemy import select
-from starlette.datastructures import URL
 from storage.billing_session import BillingSession
 from storage.database import a_session_maker
 from storage.lite_llm_manager import LiteLlmManager
@@ -151,7 +151,7 @@ async def create_customer_setup_session(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail='Could not find or create customer for user',
         )
-    base_url = _get_base_url(request)
+    base_url = get_web_url(request)
     checkout_session = await stripe.checkout.Session.create_async(
         customer=customer_info['customer_id'],
         mode='setup',
@@ -170,7 +170,7 @@ async def create_checkout_session(
     user_id: str = Depends(get_user_id),
 ) -> CreateBillingSessionResponse:
     await validate_billing_enabled()
-    base_url = _get_base_url(request)
+    base_url = get_web_url(request)
     customer_info = await stripe_service.find_or_create_customer_by_user_id(user_id)
     if not customer_info:
         raise HTTPException(
@@ -198,8 +198,8 @@ async def create_checkout_session(
         saved_payment_method_options={
             'payment_method_save': 'enabled',
         },
-        success_url=f'{base_url}api/billing/success?session_id={{CHECKOUT_SESSION_ID}}',
-        cancel_url=f'{base_url}api/billing/cancel?session_id={{CHECKOUT_SESSION_ID}}',
+        success_url=f'{base_url}/api/billing/success?session_id={{CHECKOUT_SESSION_ID}}',
+        cancel_url=f'{base_url}/api/billing/cancel?session_id={{CHECKOUT_SESSION_ID}}',
     )
     logger.info(
         'created_stripe_checkout_session',
@@ -300,7 +300,7 @@ async def success_callback(session_id: str, request: Request):
         await session.commit()
 
     return RedirectResponse(
-        f'{_get_base_url(request)}settings/billing?checkout=success', status_code=302
+        f'{get_web_url(request)}/settings/billing?checkout=success', status_code=302
     )
 
 
@@ -325,17 +325,9 @@ async def cancel_callback(session_id: str, request: Request):
             )
             billing_session.status = 'cancelled'
             billing_session.updated_at = datetime.now(UTC)
-            session.merge(billing_session)
+            await session.merge(billing_session)
             await session.commit()
 
     return RedirectResponse(
-        f'{_get_base_url(request)}settings/billing?checkout=cancel', status_code=302
+        f'{get_web_url(request)}/settings/billing?checkout=cancel', status_code=302
     )
-
-
-def _get_base_url(request: Request) -> URL:
-    # Never send any part of the credit card process over a non secure connection
-    base_url = request.base_url
-    if base_url.hostname != 'localhost':
-        base_url = base_url.replace(scheme='https')
-    return base_url

enterprise/server/routes/email.py

@@ -7,8 +7,10 @@ from pydantic import BaseModel, field_validator
 from server.auth.constants import KEYCLOAK_CLIENT_ID
 from server.auth.keycloak_manager import get_keycloak_admin
 from server.auth.saas_user_auth import SaasUserAuth
+from server.constants import IS_LOCAL_ENV
 from server.routes.auth import set_response_cookie
 from server.utils.rate_limit_utils import check_rate_limit_by_user_id
+from server.utils.url_utils import get_web_url
 from storage.user_store import UserStore
 
 from openhands.core.logger import openhands_logger as logger
@@ -87,7 +89,7 @@ async def update_email(
             response=response,
             keycloak_access_token=user_auth.access_token.get_secret_value(),
             keycloak_refresh_token=user_auth.refresh_token.get_secret_value(),
-            secure=False if request.url.hostname == 'localhost' else True,
+            secure=not IS_LOCAL_ENV,
             accepted_tos=user_auth.accepted_tos or False,
         )
 
@@ -156,8 +158,8 @@ async def verified_email(request: Request):
     await user_auth.refresh()  # refresh so access token has updated email
     user_auth.email_verified = True
     await UserStore.update_user_email(user_id=user_auth.user_id, email_verified=True)
-    scheme = 'http' if request.url.hostname == 'localhost' else 'https'
-    redirect_uri = f'{scheme}://{request.url.netloc}/settings/user'
+
+    redirect_uri = f'{get_web_url(request)}/settings/user'
     response = RedirectResponse(redirect_uri, status_code=302)
 
     # need to set auth cookie to the new tokens
@@ -180,11 +182,10 @@ async def verified_email(request: Request):
 
 async def verify_email(request: Request, user_id: str, is_auth_flow: bool = False):
     keycloak_admin = get_keycloak_admin()
-    scheme = 'http' if request.url.hostname == 'localhost' else 'https'
     if is_auth_flow:
-        redirect_uri = f'{scheme}://{request.url.netloc}/login?email_verified=true'
+        redirect_uri = f'{get_web_url(request)}/login?email_verified=true'
     else:
-        redirect_uri = f'{scheme}://{request.url.netloc}/api/email/verified'
+        redirect_uri = f'{get_web_url(request)}/api/email/verified'
     logger.info(f'Redirect URI: {redirect_uri}')
     await keycloak_admin.a_send_verify_email(
         user_id=user_id,

enterprise/server/routes/oauth_device.py

@@ -6,6 +6,7 @@ from typing import Optional
 from fastapi import APIRouter, Depends, Form, HTTPException, Request, status
 from fastapi.responses import JSONResponse
 from pydantic import BaseModel
+from server.utils.url_utils import get_web_url
 from storage.api_key_store import ApiKeyStore
 from storage.device_code_store import DeviceCodeStore
 
@@ -93,7 +94,7 @@ async def device_authorization(
             expires_in=DEVICE_CODE_EXPIRES_IN,
         )
 
-        base_url = str(http_request.base_url).rstrip('/')
+        base_url = get_web_url(http_request)
         verification_uri = f'{base_url}/oauth/device/verify'
         verification_uri_complete = (
             f'{verification_uri}?user_code={device_code_entry.user_code}'

enterprise/server/routes/org_models.py

@@ -1,4 +1,4 @@
-from typing import Annotated
+from typing import Annotated, Any
 
 from pydantic import (
     BaseModel,
@@ -8,6 +8,10 @@ from pydantic import (
     StringConstraints,
     field_validator,
 )
+from storage.agent_settings_utils import (
+    get_org_agent_settings,
+    get_org_member_agent_settings,
+)
 from storage.org import Org
 from storage.org_member import OrgMember
 from storage.role import Role
@@ -144,21 +148,13 @@ class OrgResponse(BaseModel):
     contact_name: str
     contact_email: str
     conversation_expiration: int | None = None
-    agent: str | None = None
-    default_max_iterations: int | None = None
-    security_analyzer: str | None = None
-    confirmation_mode: bool | None = None
-    default_llm_model: str | None = None
-    default_llm_api_key_for_byor: str | None = None
-    default_llm_base_url: str | None = None
     remote_runtime_resource_factor: int | None = None
-    enable_default_condenser: bool = True
     billing_margin: float | None = None
     enable_proactive_conversation_starters: bool = True
     sandbox_base_container_image: str | None = None
     sandbox_runtime_container_image: str | None = None
     org_version: int = 0
-    mcp_config: dict | None = None
+    agent_settings: dict[str, Any] = Field(default_factory=dict)
     search_api_key: str | None = None
     sandbox_api_key: str | None = None
     max_budget_per_task: float | None = None
@@ -171,33 +167,14 @@ class OrgResponse(BaseModel):
     def from_org(
         cls, org: Org, credits: float | None = None, user_id: str | None = None
     ) -> 'OrgResponse':
-        """Create an OrgResponse from an Org entity.
-
-        Args:
-            org: The organization entity to convert
-            credits: Optional credits value (defaults to None)
-            user_id: Optional user ID to determine if org is personal (defaults to None)
-
-        Returns:
-            OrgResponse: The response model instance
-        """
+        """Create an OrgResponse from an Org entity."""
         return cls(
             id=str(org.id),
             name=org.name,
             contact_name=org.contact_name,
             contact_email=org.contact_email,
             conversation_expiration=org.conversation_expiration,
-            agent=org.agent,
-            default_max_iterations=org.default_max_iterations,
-            security_analyzer=org.security_analyzer,
-            confirmation_mode=org.confirmation_mode,
-            default_llm_model=org.default_llm_model,
-            default_llm_api_key_for_byor=None,
-            default_llm_base_url=org.default_llm_base_url,
             remote_runtime_resource_factor=org.remote_runtime_resource_factor,
-            enable_default_condenser=org.enable_default_condenser
-            if org.enable_default_condenser is not None
-            else True,
             billing_margin=org.billing_margin,
             enable_proactive_conversation_starters=org.enable_proactive_conversation_starters
             if org.enable_proactive_conversation_starters is not None
@@ -205,7 +182,7 @@ class OrgResponse(BaseModel):
             sandbox_base_container_image=org.sandbox_base_container_image,
             sandbox_runtime_container_image=org.sandbox_runtime_container_image,
             org_version=org.org_version if org.org_version is not None else 0,
-            mcp_config=org.mcp_config,
+            agent_settings=get_org_agent_settings(org),
             search_api_key=None,
             sandbox_api_key=None,
             max_budget_per_task=org.max_budget_per_task,
@@ -227,7 +204,6 @@ class OrgPage(BaseModel):
 class OrgUpdate(BaseModel):
     """Request model for updating an organization."""
 
-    # Basic organization information (any authenticated user can update)
     name: Annotated[
         str | None,
         StringConstraints(strip_whitespace=True, min_length=1, max_length=255),
@@ -235,42 +211,24 @@ class OrgUpdate(BaseModel):
     contact_name: str | None = None
     contact_email: EmailStr | None = None
     conversation_expiration: int | None = None
-    default_max_iterations: int | None = Field(default=None, gt=0)
     remote_runtime_resource_factor: int | None = Field(default=None, gt=0)
     billing_margin: float | None = Field(default=None, ge=0, le=1)
     enable_proactive_conversation_starters: bool | None = None
     sandbox_base_container_image: str | None = None
     sandbox_runtime_container_image: str | None = None
-    mcp_config: dict | None = None
     sandbox_api_key: str | None = None
     max_budget_per_task: float | None = Field(default=None, gt=0)
     enable_solvability_analysis: bool | None = None
     v1_enabled: bool | None = None
-
-    # LLM settings (require admin/owner role)
-    default_llm_model: str | None = None
-    default_llm_api_key_for_byor: str | None = None
-    default_llm_base_url: str | None = None
     search_api_key: str | None = None
-    security_analyzer: str | None = None
-    agent: str | None = None
-    confirmation_mode: bool | None = None
-    enable_default_condenser: bool | None = None
-    condenser_max_size: int | None = Field(default=None, ge=20)
+    agent_settings: dict[str, Any] | None = None
 
 
 class OrgLLMSettingsResponse(BaseModel):
     """Response model for organization LLM settings."""
 
-    default_llm_model: str | None = None
-    default_llm_base_url: str | None = None
+    agent_settings: dict[str, Any] = Field(default_factory=dict)
     search_api_key: str | None = None  # Masked in response
-    agent: str | None = None
-    confirmation_mode: bool | None = None
-    security_analyzer: str | None = None
-    enable_default_condenser: bool = True
-    condenser_max_size: int | None = None
-    default_max_iterations: int | None = None
 
     @staticmethod
     def _mask_key(secret: SecretStr | None) -> str | None:
@@ -288,83 +246,45 @@ class OrgLLMSettingsResponse(BaseModel):
     def from_org(cls, org: Org) -> 'OrgLLMSettingsResponse':
         """Create response from Org entity."""
         return cls(
-            default_llm_model=org.default_llm_model,
-            default_llm_base_url=org.default_llm_base_url,
+            agent_settings=get_org_agent_settings(org),
             search_api_key=cls._mask_key(org.search_api_key),
-            agent=org.agent,
-            confirmation_mode=org.confirmation_mode,
-            security_analyzer=org.security_analyzer,
-            enable_default_condenser=org.enable_default_condenser
-            if org.enable_default_condenser is not None
-            else True,
-            condenser_max_size=org.condenser_max_size,
-            default_max_iterations=org.default_max_iterations,
         )
 
 
 class OrgMemberLLMSettings(BaseModel):
-    """LLM settings to propagate to organization members.
+    """Shared LLM settings that may be propagated to organization members."""
 
-    Field names match OrgMember DB columns.
-    """
-
-    llm_model: str | None = None
-    llm_base_url: str | None = None
-    max_iterations: int | None = None
+    agent_settings: dict[str, Any] | None = None
     llm_api_key: str | None = None
 
     def has_updates(self) -> bool:
         """Check if any field is set (not None)."""
-        return any(getattr(self, field) is not None for field in self.model_fields)
+        return any(
+            getattr(self, field) is not None for field in type(self).model_fields
+        )
 
 
 class OrgLLMSettingsUpdate(BaseModel):
-    """Request model for updating organization LLM settings.
+    """Request model for updating organization LLM settings."""
 
-    Field names match Org DB columns exactly.
-    """
-
-    default_llm_model: str | None = None
-    default_llm_base_url: str | None = None
+    agent_settings: dict[str, Any] | None = None
     search_api_key: str | None = None
-    agent: str | None = None
-    confirmation_mode: bool | None = None
-    security_analyzer: str | None = None
-    enable_default_condenser: bool | None = None
-    condenser_max_size: int | None = Field(default=None, ge=20)
-    default_max_iterations: int | None = Field(default=None, gt=0)
     llm_api_key: str | None = None
 
     def has_updates(self) -> bool:
         """Check if any field is set (not None)."""
-        return any(getattr(self, field) is not None for field in self.model_fields)
+        return any(
+            getattr(self, field) is not None for field in type(self).model_fields
+        )
 
     def apply_to_org(self, org: Org) -> None:
-        """Apply non-None settings to the organization model.
-
-        Args:
-            org: Organization entity to update in place
-        """
-        for field_name in self.model_fields:
-            value = getattr(self, field_name)
-            # Skip llm_api_key - it's only for member propagation, not org-level
-            if value is not None and field_name != 'llm_api_key':
-                setattr(org, field_name, value)
+        """Apply non-None settings to the organization model."""
+        if self.search_api_key is not None:
+            org.search_api_key = self.search_api_key
 
     def get_member_updates(self) -> OrgMemberLLMSettings | None:
-        """Get updates that need to be propagated to org members.
-
-        Returns:
-            OrgMemberLLMSettings with mapped field values, or None if no member updates needed.
-            Maps: default_llm_model → llm_model, default_llm_base_url → llm_base_url,
-                  default_max_iterations → max_iterations, llm_api_key → llm_api_key
-        """
-        member_settings = OrgMemberLLMSettings(
-            llm_model=self.default_llm_model,
-            llm_base_url=self.default_llm_base_url,
-            max_iterations=self.default_max_iterations,
-            llm_api_key=self.llm_api_key,
-        )
+        """Get updates that need to be propagated to org members."""
+        member_settings = OrgMemberLLMSettings(llm_api_key=self.llm_api_key)
         return member_settings if member_settings.has_updates() else None
 
 
@@ -401,18 +321,15 @@ class MeResponse(BaseModel):
     email: str
     role: str
     llm_api_key: str
-    max_iterations: int | None = None
-    llm_model: str | None = None
-    llm_api_key_for_byor: str | None = None
-    llm_base_url: str | None = None
+    agent_settings: dict[str, Any] = Field(default_factory=dict)
     status: str | None = None
 
     @staticmethod
-    def _mask_key(secret: SecretStr | None) -> str:
+    def _mask_key(secret: str | SecretStr | None) -> str:
         """Mask an API key, showing only last 4 characters."""
         if secret is None:
             return ''
-        raw = secret.get_secret_value()
+        raw = secret.get_secret_value() if isinstance(secret, SecretStr) else secret
         if not raw:
             return ''
         if len(raw) <= 4:
@@ -420,27 +337,20 @@ class MeResponse(BaseModel):
         return '****' + raw[-4:]
 
     @classmethod
-    def from_org_member(cls, member: OrgMember, role: Role, email: str) -> 'MeResponse':
-        """Create a MeResponse from an OrgMember, Role, and user email.
-
-        Args:
-            member: The OrgMember entity
-            role: The Role entity (provides role name)
-            email: The user's email address
-
-        Returns:
-            MeResponse with masked API keys
-        """
+    def from_org_member(
+        cls,
+        member: OrgMember,
+        role: Role,
+        email: str,
+    ) -> 'MeResponse':
+        """Create a MeResponse from an OrgMember, Role, and user email."""
         return cls(
             org_id=str(member.org_id),
             user_id=str(member.user_id),
             email=email,
             role=role.name,
             llm_api_key=cls._mask_key(member.llm_api_key),
-            max_iterations=member.max_iterations,
-            llm_model=member.llm_model,
-            llm_api_key_for_byor=cls._mask_key(member.llm_api_key_for_byor) or None,
-            llm_base_url=member.llm_base_url,
+            agent_settings=get_org_member_agent_settings(member),
             status=member.status,
         )
 

enterprise/server/routes/orgs.py

@@ -68,7 +68,7 @@ async def list_user_orgs(
     ] = None,
     limit: Annotated[
         int,
-        Query(title='The max number of results in the page', gt=0, lte=100),
+        Query(title='The max number of results in the page', gt=0, le=100),
     ] = 100,
     user_id: str = Depends(get_user_id),
 ) -> OrgPage:
@@ -734,7 +734,7 @@ async def get_org_members(
         Query(
             title='The max number of results in the page',
             gt=0,
-            lte=100,
+            le=100,
         ),
     ] = 10,
     email: Annotated[

enterprise/server/routes/service.py

@@ -0,0 +1,270 @@
+"""
+Service API routes for internal service-to-service communication.
+
+This module provides endpoints for trusted internal services (e.g., automations service)
+to perform privileged operations like creating API keys on behalf of users.
+
+Authentication is via a shared secret (X-Service-API-Key header) configured
+through the AUTOMATIONS_SERVICE_KEY environment variable.
+"""
+
+import os
+from uuid import UUID
+
+from fastapi import APIRouter, Header, HTTPException, status
+from pydantic import BaseModel, field_validator
+from storage.api_key_store import ApiKeyStore
+from storage.org_member_store import OrgMemberStore
+from storage.user_store import UserStore
+
+from openhands.core.logger import openhands_logger as logger
+
+# Environment variable for the service API key
+AUTOMATIONS_SERVICE_KEY = os.getenv('AUTOMATIONS_SERVICE_KEY', '').strip()
+
+service_router = APIRouter(prefix='/api/service', tags=['Service'])
+
+
+class CreateUserApiKeyRequest(BaseModel):
+    """Request model for creating an API key on behalf of a user."""
+
+    name: str  # Required - used to identify the key
+
+    @field_validator('name')
+    @classmethod
+    def validate_name(cls, v: str) -> str:
+        if not v or not v.strip():
+            raise ValueError('name is required and cannot be empty')
+        return v.strip()
+
+
+class CreateUserApiKeyResponse(BaseModel):
+    """Response model for created API key."""
+
+    key: str
+    user_id: str
+    org_id: str
+    name: str
+
+
+class ServiceInfoResponse(BaseModel):
+    """Response model for service info endpoint."""
+
+    service: str
+    authenticated: bool
+
+
+async def validate_service_api_key(
+    x_service_api_key: str | None = Header(default=None, alias='X-Service-API-Key'),
+) -> str:
+    """
+    Validate the service API key from the request header.
+
+    Args:
+        x_service_api_key: The service API key from the X-Service-API-Key header
+
+    Returns:
+        str: Service identifier for audit logging
+
+    Raises:
+        HTTPException: 401 if key is missing or invalid
+        HTTPException: 503 if service auth is not configured
+    """
+    if not AUTOMATIONS_SERVICE_KEY:
+        logger.warning(
+            'Service authentication not configured (AUTOMATIONS_SERVICE_KEY not set)'
+        )
+        raise HTTPException(
+            status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
+            detail='Service authentication not configured',
+        )
+
+    if not x_service_api_key:
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED,
+            detail='X-Service-API-Key header is required',
+        )
+
+    if x_service_api_key != AUTOMATIONS_SERVICE_KEY:
+        logger.warning('Invalid service API key attempted')
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED,
+            detail='Invalid service API key',
+        )
+
+    return 'automations-service'
+
+
+@service_router.get('/health')
+async def service_health() -> dict:
+    """Health check endpoint for the service API.
+
+    This endpoint does not require authentication and can be used
+    to verify the service routes are accessible.
+    """
+    return {
+        'status': 'ok',
+        'service_auth_configured': bool(AUTOMATIONS_SERVICE_KEY),
+    }
+
+
+@service_router.post('/users/{user_id}/orgs/{org_id}/api-keys')
+async def get_or_create_api_key_for_user(
+    user_id: str,
+    org_id: UUID,
+    request: CreateUserApiKeyRequest,
+    x_service_api_key: str | None = Header(default=None, alias='X-Service-API-Key'),
+) -> CreateUserApiKeyResponse:
+    """
+    Get or create an API key for a user on behalf of the automations service.
+
+    If a key with the given name already exists for the user/org and is not expired,
+    returns the existing key. Otherwise, creates a new key.
+
+    The created/returned keys are system keys and are:
+    - Not visible to the user in their API keys list
+    - Not deletable by the user
+    - Never expire
+
+    Args:
+        user_id: The user ID
+        org_id: The organization ID
+        request: Request body containing name (required)
+        x_service_api_key: Service API key header for authentication
+
+    Returns:
+        CreateUserApiKeyResponse: The API key and metadata
+
+    Raises:
+        HTTPException: 401 if service key is invalid
+        HTTPException: 404 if user not found
+        HTTPException: 403 if user is not a member of the specified org
+    """
+    # Validate service API key
+    service_id = await validate_service_api_key(x_service_api_key)
+
+    # Verify user exists
+    user = await UserStore.get_user_by_id(user_id)
+    if not user:
+        logger.warning(
+            'Service attempted to create key for non-existent user',
+            extra={'user_id': user_id},
+        )
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=f'User {user_id} not found',
+        )
+
+    # Verify user is a member of the specified org
+    org_member = await OrgMemberStore.get_org_member(org_id, UUID(user_id))
+    if not org_member:
+        logger.warning(
+            'Service attempted to create key for user not in org',
+            extra={
+                'user_id': user_id,
+                'org_id': str(org_id),
+            },
+        )
+        raise HTTPException(
+            status_code=status.HTTP_403_FORBIDDEN,
+            detail=f'User {user_id} is not a member of org {org_id}',
+        )
+
+    # Get or create the system API key
+    api_key_store = ApiKeyStore.get_instance()
+
+    try:
+        api_key = await api_key_store.get_or_create_system_api_key(
+            user_id=user_id,
+            org_id=org_id,
+            name=request.name,
+        )
+    except Exception as e:
+        logger.exception(
+            'Failed to get or create system API key',
+            extra={
+                'user_id': user_id,
+                'org_id': str(org_id),
+                'error': str(e),
+            },
+        )
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+            detail='Failed to get or create API key',
+        )
+
+    logger.info(
+        'Service created API key for user',
+        extra={
+            'service_id': service_id,
+            'user_id': user_id,
+            'org_id': str(org_id),
+            'key_name': request.name,
+        },
+    )
+
+    return CreateUserApiKeyResponse(
+        key=api_key,
+        user_id=user_id,
+        org_id=str(org_id),
+        name=request.name,
+    )
+
+
+@service_router.delete('/users/{user_id}/orgs/{org_id}/api-keys/{key_name}')
+async def delete_user_api_key(
+    user_id: str,
+    org_id: UUID,
+    key_name: str,
+    x_service_api_key: str | None = Header(default=None, alias='X-Service-API-Key'),
+) -> dict:
+    """
+    Delete a system API key created by the service.
+
+    This endpoint allows the automations service to clean up API keys
+    it previously created for users.
+
+    Args:
+        user_id: The user ID
+        org_id: The organization ID
+        key_name: The name of the key to delete (without __SYSTEM__: prefix)
+        x_service_api_key: Service API key header for authentication
+
+    Returns:
+        dict: Success message
+
+    Raises:
+        HTTPException: 401 if service key is invalid
+        HTTPException: 404 if key not found
+    """
+    # Validate service API key
+    service_id = await validate_service_api_key(x_service_api_key)
+
+    api_key_store = ApiKeyStore.get_instance()
+
+    # Delete the key by name (wrap with system key prefix since service creates system keys)
+    system_key_name = api_key_store.make_system_key_name(key_name)
+    success = await api_key_store.delete_api_key_by_name(
+        user_id=user_id,
+        org_id=org_id,
+        name=system_key_name,
+        allow_system=True,
+    )
+
+    if not success:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=f'API key with name "{key_name}" not found for user {user_id} in org {org_id}',
+        )
+
+    logger.info(
+        'Service deleted API key for user',
+        extra={
+            'service_id': service_id,
+            'user_id': user_id,
+            'org_id': str(org_id),
+            'key_name': key_name,
+        },
+    )
+
+    return {'message': 'API key deleted successfully'}

enterprise/server/saas_nested_conversation_manager.py

@@ -522,10 +522,9 @@ class SaasNestedConversationManager(ConversationManager):
 
         mcp_config = await self._get_mcp_config(user_id)
         if mcp_config:
-            # Merge with any MCP config from settings
-            if settings.mcp_config:
-                mcp_config = mcp_config.merge(settings.mcp_config)
-            # Check again since theoretically merge could return None.
+            settings_mcp_config = settings.to_legacy_mcp_config()
+            if settings_mcp_config:
+                mcp_config = mcp_config.merge(settings_mcp_config)
             if mcp_config:
                 init_conversation['mcp_config'] = mcp_config.model_dump()
 
@@ -855,7 +854,7 @@ class SaasNestedConversationManager(ConversationManager):
             user_id=user_id,
         )
         llm_registry.retry_listner = session._notify_on_llm_retry
-        agent_cls = settings.agent or self.config.default_agent
+        agent_cls = settings.agent_settings.agent or self.config.default_agent
         agent_config = self.config.get_agent_config(agent_cls)
         agent = Agent.get_cls(agent_cls)(agent_config, llm_registry)
 

enterprise/server/services/org_invitation_service.py

@@ -365,17 +365,15 @@ class OrgInvitationService:
                 'Failed to set up organization access. Please try again.'
             )
 
-        # Step 5: Add user to organization
-        from storage.org_member_store import OrgMemberStore as OMS
+        # Step 4.5: Fetch organization to get its LLM settings
+        org = await OrgStore.get_org_by_id(invitation.org_id)
+        if not org:
+            raise InvitationInvalidError('Organization not found')
 
-        org_member_kwargs = OMS.get_kwargs_from_settings(settings)
-        # Don't override with org defaults - use invitation-specified role
-        org_member_kwargs.pop('llm_model', None)
-        org_member_kwargs.pop('llm_base_url', None)
-
-        # Get the llm_api_key as string (it's SecretStr | None in Settings)
+        # Step 5: Add user to organization with inherited org LLM settings
+        llm_api_key_secret = settings.get_secret_agent_setting('llm.api_key')
         llm_api_key = (
-            settings.llm_api_key.get_secret_value() if settings.llm_api_key else ''
+            llm_api_key_secret.get_secret_value() if llm_api_key_secret else ''
         )
 
         await OrgMemberStore.add_user_to_org(
@@ -384,6 +382,7 @@ class OrgInvitationService:
             role_id=invitation.role_id,
             llm_api_key=llm_api_key,
             status='active',
+            agent_settings=OrgStore.get_agent_settings_from_org(org),
         )
 
         # Step 6: Mark invitation as accepted

enterprise/server/services/org_member_service.py

@@ -55,7 +55,6 @@ class OrgMemberService:
         if role is None:
             raise RoleNotFoundError(org_member.role_id)
 
-        # Get user email
         user = await UserStore.get_user_by_id(str(user_id))
         email = user.email if user and user.email else ''
 

enterprise/server/sharing/aws_shared_event_service.py

@@ -0,0 +1,171 @@
+"""Implementation of SharedEventService for AWS S3.
+
+This implementation provides read-only access to events from shared conversations:
+- Validates that the conversation is shared before returning events
+- Uses existing EventService for actual event retrieval
+- Uses SharedConversationInfoService for shared conversation validation
+
+Uses role-based authentication (no credentials needed).
+"""
+
+from __future__ import annotations
+
+import logging
+import os
+from dataclasses import dataclass
+from datetime import datetime
+from pathlib import Path
+from typing import Any, AsyncGenerator
+from uuid import UUID
+
+import boto3
+from fastapi import Request
+from pydantic import Field
+from server.sharing.shared_conversation_info_service import (
+    SharedConversationInfoService,
+)
+from server.sharing.shared_event_service import (
+    SharedEventService,
+    SharedEventServiceInjector,
+)
+from server.sharing.sql_shared_conversation_info_service import (
+    SQLSharedConversationInfoService,
+)
+
+from openhands.agent_server.models import EventPage, EventSortOrder
+from openhands.app_server.event.aws_event_service import AwsEventService
+from openhands.app_server.event.event_service import EventService
+from openhands.app_server.event_callback.event_callback_models import EventKind
+from openhands.app_server.services.injector import InjectorState
+from openhands.sdk import Event
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class AwsSharedEventService(SharedEventService):
+    """Implementation of SharedEventService for AWS S3 that validates shared access.
+
+    Uses role-based authentication (no credentials needed).
+    """
+
+    shared_conversation_info_service: SharedConversationInfoService
+    s3_client: Any
+    bucket_name: str
+
+    async def get_event_service(self, conversation_id: UUID) -> EventService | None:
+        shared_conversation_info = (
+            await self.shared_conversation_info_service.get_shared_conversation_info(
+                conversation_id
+            )
+        )
+        if shared_conversation_info is None:
+            return None
+
+        return AwsEventService(
+            s3_client=self.s3_client,
+            bucket_name=self.bucket_name,
+            prefix=Path('users'),
+            user_id=shared_conversation_info.created_by_user_id,
+            app_conversation_info_service=None,
+            app_conversation_info_load_tasks={},
+        )
+
+    async def get_shared_event(
+        self, conversation_id: UUID, event_id: UUID
+    ) -> Event | None:
+        """Given a conversation_id and event_id, retrieve an event if the conversation is shared."""
+        # First check if the conversation is shared
+        event_service = await self.get_event_service(conversation_id)
+        if event_service is None:
+            return None
+
+        # If conversation is shared, get the event
+        return await event_service.get_event(conversation_id, event_id)
+
+    async def search_shared_events(
+        self,
+        conversation_id: UUID,
+        kind__eq: EventKind | None = None,
+        timestamp__gte: datetime | None = None,
+        timestamp__lt: datetime | None = None,
+        sort_order: EventSortOrder = EventSortOrder.TIMESTAMP,
+        page_id: str | None = None,
+        limit: int = 100,
+    ) -> EventPage:
+        """Search events for a specific shared conversation."""
+        # First check if the conversation is shared
+        event_service = await self.get_event_service(conversation_id)
+        if event_service is None:
+            # Return empty page if conversation is not shared
+            return EventPage(items=[], next_page_id=None)
+
+        # If conversation is shared, search events for this conversation
+        return await event_service.search_events(
+            conversation_id=conversation_id,
+            kind__eq=kind__eq,
+            timestamp__gte=timestamp__gte,
+            timestamp__lt=timestamp__lt,
+            sort_order=sort_order,
+            page_id=page_id,
+            limit=limit,
+        )
+
+    async def count_shared_events(
+        self,
+        conversation_id: UUID,
+        kind__eq: EventKind | None = None,
+        timestamp__gte: datetime | None = None,
+        timestamp__lt: datetime | None = None,
+    ) -> int:
+        """Count events for a specific shared conversation."""
+        # First check if the conversation is shared
+        event_service = await self.get_event_service(conversation_id)
+        if event_service is None:
+            # Return empty page if conversation is not shared
+            return 0
+
+        # If conversation is shared, count events for this conversation
+        return await event_service.count_events(
+            conversation_id=conversation_id,
+            kind__eq=kind__eq,
+            timestamp__gte=timestamp__gte,
+            timestamp__lt=timestamp__lt,
+        )
+
+
+class AwsSharedEventServiceInjector(SharedEventServiceInjector):
+    bucket_name: str | None = Field(
+        default_factory=lambda: os.environ.get('FILE_STORE_PATH')
+    )
+
+    async def inject(
+        self, state: InjectorState, request: Request | None = None
+    ) -> AsyncGenerator[SharedEventService, None]:
+        # Define inline to prevent circular lookup
+        from openhands.app_server.config import get_db_session
+
+        async with get_db_session(state, request) as db_session:
+            shared_conversation_info_service = SQLSharedConversationInfoService(
+                db_session=db_session
+            )
+
+            bucket_name = self.bucket_name
+            if bucket_name is None:
+                raise ValueError(
+                    'bucket_name is required. Set FILE_STORE_PATH environment variable.'
+                )
+
+            # Use role-based authentication - boto3 will automatically
+            # use IAM role credentials when running in AWS
+            s3_client = boto3.client(
+                's3',
+                endpoint_url=os.getenv('AWS_S3_ENDPOINT'),
+            )
+
+            service = AwsSharedEventService(
+                shared_conversation_info_service=shared_conversation_info_service,
+                s3_client=s3_client,
+                bucket_name=bucket_name,
+            )
+            yield service

enterprise/server/sharing/filesystem_shared_event_service.py

@@ -0,0 +1,143 @@
+"""Implementation of SharedEventService.
+
+This implementation provides read-only access to events from shared conversations:
+- Validates that the conversation is shared before returning events
+- Uses existing EventService for actual event retrieval
+- Uses SharedConversationInfoService for shared conversation validation
+"""
+
+from __future__ import annotations
+
+import logging
+from dataclasses import dataclass
+from datetime import datetime
+from pathlib import Path
+from typing import AsyncGenerator
+from uuid import UUID
+
+from fastapi import Request
+from server.sharing.shared_conversation_info_service import (
+    SharedConversationInfoService,
+)
+from server.sharing.shared_event_service import (
+    SharedEventService,
+    SharedEventServiceInjector,
+)
+from server.sharing.sql_shared_conversation_info_service import (
+    SQLSharedConversationInfoService,
+)
+
+from openhands.agent_server.models import EventPage, EventSortOrder
+from openhands.app_server.config import get_global_config
+from openhands.app_server.event.event_service import EventService
+from openhands.app_server.event.filesystem_event_service import FilesystemEventService
+from openhands.app_server.event_callback.event_callback_models import EventKind
+from openhands.app_server.services.injector import InjectorState
+from openhands.sdk import Event
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class FilesystemSharedEventService(SharedEventService):
+    """Implementation of SharedEventService that validates shared access."""
+
+    shared_conversation_info_service: SharedConversationInfoService
+    persistence_dir: Path
+
+    async def get_event_service(self, conversation_id: UUID) -> EventService | None:
+        shared_conversation_info = (
+            await self.shared_conversation_info_service.get_shared_conversation_info(
+                conversation_id
+            )
+        )
+        if shared_conversation_info is None:
+            return None
+
+        return FilesystemEventService(
+            prefix=self.persistence_dir,
+            user_id=shared_conversation_info.created_by_user_id,
+            app_conversation_info_service=None,
+            app_conversation_info_load_tasks={},
+        )
+
+    async def get_shared_event(
+        self, conversation_id: UUID, event_id: UUID
+    ) -> Event | None:
+        """Given a conversation_id and event_id, retrieve an event if the conversation is shared."""
+        # First check if the conversation is shared
+        event_service = await self.get_event_service(conversation_id)
+        if event_service is None:
+            return None
+
+        # If conversation is shared, get the event
+        return await event_service.get_event(conversation_id, event_id)
+
+    async def search_shared_events(
+        self,
+        conversation_id: UUID,
+        kind__eq: EventKind | None = None,
+        timestamp__gte: datetime | None = None,
+        timestamp__lt: datetime | None = None,
+        sort_order: EventSortOrder = EventSortOrder.TIMESTAMP,
+        page_id: str | None = None,
+        limit: int = 100,
+    ) -> EventPage:
+        """Search events for a specific shared conversation."""
+        # First check if the conversation is shared
+        event_service = await self.get_event_service(conversation_id)
+        if event_service is None:
+            # Return empty page if conversation is not shared
+            return EventPage(items=[], next_page_id=None)
+
+        # If conversation is shared, search events for this conversation
+        return await event_service.search_events(
+            conversation_id=conversation_id,
+            kind__eq=kind__eq,
+            timestamp__gte=timestamp__gte,
+            timestamp__lt=timestamp__lt,
+            sort_order=sort_order,
+            page_id=page_id,
+            limit=limit,
+        )
+
+    async def count_shared_events(
+        self,
+        conversation_id: UUID,
+        kind__eq: EventKind | None = None,
+        timestamp__gte: datetime | None = None,
+        timestamp__lt: datetime | None = None,
+    ) -> int:
+        """Count events for a specific shared conversation."""
+        # First check if the conversation is shared
+        event_service = await self.get_event_service(conversation_id)
+        if event_service is None:
+            # Return empty page if conversation is not shared
+            return 0
+
+        # If conversation is shared, count events for this conversation
+        return await event_service.count_events(
+            conversation_id=conversation_id,
+            kind__eq=kind__eq,
+            timestamp__gte=timestamp__gte,
+            timestamp__lt=timestamp__lt,
+        )
+
+
+class FilesystemSharedEventServiceInjector(SharedEventServiceInjector):
+    async def inject(
+        self, state: InjectorState, request: Request | None = None
+    ) -> AsyncGenerator[SharedEventService, None]:
+        # Define inline to prevent circular lookup
+        from openhands.app_server.config import get_db_session
+
+        async with get_db_session(state, request) as db_session:
+            shared_conversation_info_service = SQLSharedConversationInfoService(
+                db_session=db_session
+            )
+
+            service = FilesystemSharedEventService(
+                shared_conversation_info_service=shared_conversation_info_service,
+                persistence_dir=get_global_config().persistence_dir,
+            )
+            yield service

enterprise/server/sharing/shared_conversation_router.py

@@ -4,7 +4,7 @@ from datetime import datetime
 from typing import Annotated
 from uuid import UUID
 
-from fastapi import APIRouter, Depends, Query
+from fastapi import APIRouter, Depends, HTTPException, Query
 from server.sharing.shared_conversation_info_service import (
     SharedConversationInfoService,
 )
@@ -60,7 +60,7 @@ async def search_shared_conversations(
         Query(
             title='The max number of results in the page',
             gt=0,
-            lte=100,
+            le=100,
         ),
     ] = 100,
     include_sub_conversations: Annotated[
@@ -72,8 +72,6 @@ async def search_shared_conversations(
     shared_conversation_service: SharedConversationInfoService = shared_conversation_info_service_dependency,
 ) -> SharedConversationPage:
     """Search / List shared conversations."""
-    assert limit > 0
-    assert limit <= 100
     return await shared_conversation_service.search_shared_conversation_info(
         title__contains=title__contains,
         created_at__gte=created_at__gte,
@@ -127,7 +125,11 @@ async def batch_get_shared_conversations(
     shared_conversation_service: SharedConversationInfoService = shared_conversation_info_service_dependency,
 ) -> list[SharedConversation | None]:
     """Get a batch of shared conversations given their ids. Return None for any missing or non-shared."""
-    assert len(ids) <= 100
+    if len(ids) > 100:
+        raise HTTPException(
+            status_code=400,
+            detail=f'Cannot request more than 100 conversations at once, got {len(ids)}',
+        )
     uuids = [UUID(id_) for id_ in ids]
     shared_conversation_info = (
         await shared_conversation_service.batch_get_shared_conversation_info(uuids)

enterprise/server/sharing/shared_event_router.py

@@ -4,20 +4,52 @@ from datetime import datetime
 from typing import Annotated
 from uuid import UUID
 
-from fastapi import APIRouter, Depends, Query
-from server.sharing.google_cloud_shared_event_service import (
-    GoogleCloudSharedEventServiceInjector,
+from fastapi import APIRouter, Depends, HTTPException, Query
+from server.sharing.shared_event_service import (
+    SharedEventService,
+    SharedEventServiceInjector,
 )
-from server.sharing.shared_event_service import SharedEventService
 
 from openhands.agent_server.models import EventPage, EventSortOrder
 from openhands.app_server.event_callback.event_callback_models import EventKind
 from openhands.sdk import Event
+from openhands.utils.environment import StorageProvider, get_storage_provider
+
+
+def get_shared_event_service_injector() -> SharedEventServiceInjector:
+    """Get the appropriate SharedEventServiceInjector based on configuration.
+
+    Uses get_storage_provider() to determine the storage backend.
+    See openhands.utils.environment for supported environment variables.
+
+    Note: Shared events only support AWS and GCP storage. Filesystem storage
+    falls back to GCP for shared events.
+    """
+    provider = get_storage_provider()
+
+    if provider == StorageProvider.AWS:
+        from server.sharing.aws_shared_event_service import (
+            AwsSharedEventServiceInjector,
+        )
+
+        return AwsSharedEventServiceInjector()
+    elif provider == StorageProvider.FILESYSTEM:
+        from server.sharing.filesystem_shared_event_service import (
+            FilesystemSharedEventServiceInjector,
+        )
+
+        return FilesystemSharedEventServiceInjector()
+    else:
+        # GCP is the default for shared events (including filesystem fallback)
+        from server.sharing.google_cloud_shared_event_service import (
+            GoogleCloudSharedEventServiceInjector,
+        )
+
+        return GoogleCloudSharedEventServiceInjector()
+
 
 router = APIRouter(prefix='/api/shared-events', tags=['Sharing'])
-shared_event_service_dependency = Depends(
-    GoogleCloudSharedEventServiceInjector().depends
-)
+shared_event_service_dependency = Depends(get_shared_event_service_injector().depends)
 
 
 # Read methods
@@ -51,13 +83,11 @@ async def search_shared_events(
     ] = None,
     limit: Annotated[
         int,
-        Query(title='The max number of results in the page', gt=0, lte=100),
+        Query(title='The max number of results in the page', gt=0, le=100),
     ] = 100,
     shared_event_service: SharedEventService = shared_event_service_dependency,
 ) -> EventPage:
     """Search / List events for a shared conversation."""
-    assert limit > 0
-    assert limit <= 100
     return await shared_event_service.search_shared_events(
         conversation_id=UUID(conversation_id),
         kind__eq=kind__eq,
@@ -108,7 +138,11 @@ async def batch_get_shared_events(
     shared_event_service: SharedEventService = shared_event_service_dependency,
 ) -> list[Event | None]:
     """Get a batch of events for a shared conversation given their ids, returning null for any missing event."""
-    assert len(id) <= 100
+    if len(id) > 100:
+        raise HTTPException(
+            status_code=400,
+            detail=f'Cannot request more than 100 events at once, got {len(id)}',
+        )
     event_ids = [UUID(id_) for id_ in id]
     events = await shared_event_service.batch_get_shared_events(
         UUID(conversation_id), event_ids

enterprise/server/utils/saas_app_conversation_info_injector.py

@@ -119,6 +119,7 @@ class SaasSQLAppConversationInfoService(SQLAppConversationInfoService):
         created_at__lt: datetime | None = None,
         updated_at__gte: datetime | None = None,
         updated_at__lt: datetime | None = None,
+        sandbox_id__eq: str | None = None,
         sort_order: AppConversationSortOrder = AppConversationSortOrder.CREATED_AT_DESC,
         page_id: str | None = None,
         limit: int = 100,
@@ -141,6 +142,7 @@ class SaasSQLAppConversationInfoService(SQLAppConversationInfoService):
             created_at__lt=created_at__lt,
             updated_at__gte=updated_at__gte,
             updated_at__lt=updated_at__lt,
+            sandbox_id__eq=sandbox_id__eq,
         )
 
         # Add sort order
@@ -198,6 +200,7 @@ class SaasSQLAppConversationInfoService(SQLAppConversationInfoService):
         created_at__lt: datetime | None = None,
         updated_at__gte: datetime | None = None,
         updated_at__lt: datetime | None = None,
+        sandbox_id__eq: str | None = None,
     ) -> int:
         """Count conversations matching the given filters with SAAS metadata."""
         query = (
@@ -220,6 +223,7 @@ class SaasSQLAppConversationInfoService(SQLAppConversationInfoService):
             created_at__lt=created_at__lt,
             updated_at__gte=updated_at__gte,
             updated_at__lt=updated_at__lt,
+            sandbox_id__eq=sandbox_id__eq,
         )
 
         result = await self.db_session.execute(query)
@@ -234,6 +238,7 @@ class SaasSQLAppConversationInfoService(SQLAppConversationInfoService):
         created_at__lt: datetime | None = None,
         updated_at__gte: datetime | None = None,
         updated_at__lt: datetime | None = None,
+        sandbox_id__eq: str | None = None,
     ):
         """Apply filters to query that includes SAAS metadata."""
         # Apply the same filters as the base class
@@ -259,6 +264,9 @@ class SaasSQLAppConversationInfoService(SQLAppConversationInfoService):
                 StoredConversationMetadata.last_updated_at < updated_at__lt
             )
 
+        if sandbox_id__eq is not None:
+            conditions.append(StoredConversationMetadata.sandbox_id == sandbox_id__eq)
+
         if conditions:
             query = query.where(*conditions)
         return query
@@ -334,7 +342,10 @@ class SaasSQLAppConversationInfoService(SQLAppConversationInfoService):
         await super().save_app_conversation_info(info)
 
         # Get current user_id for SAAS metadata
+        # Fall back to info.created_by_user_id for webhook callbacks (which use ADMIN context)
         user_id_str = await self.user_context.get_user_id()
+        if not user_id_str and info.created_by_user_id:
+            user_id_str = info.created_by_user_id
         if user_id_str:
             # Convert string user_id to UUID
             user_id_uuid = UUID(user_id_str)
@@ -343,6 +354,15 @@ class SaasSQLAppConversationInfoService(SQLAppConversationInfoService):
             user = result.scalar_one_or_none()
             assert user
 
+            # Determine org_id: prefer API key's org_id if authenticated via API key
+            org_id = user.current_org_id  # Default fallback
+            if hasattr(self.user_context, 'user_auth'):
+                user_auth = self.user_context.user_auth
+                if hasattr(user_auth, 'get_api_key_org_id'):
+                    api_key_org_id = user_auth.get_api_key_org_id()
+                    if api_key_org_id is not None:
+                        org_id = api_key_org_id
+
             # Check if SAAS metadata already exists
             saas_query = select(StoredConversationMetadataSaas).where(
                 StoredConversationMetadataSaas.conversation_id == str(info.id)
@@ -351,16 +371,15 @@ class SaasSQLAppConversationInfoService(SQLAppConversationInfoService):
             existing_saas_metadata = result.scalar_one_or_none()
             assert existing_saas_metadata is None or (
                 existing_saas_metadata.user_id == user_id_uuid
-                and existing_saas_metadata.org_id == user.current_org_id
+                and existing_saas_metadata.org_id == org_id
             )
 
             if not existing_saas_metadata:
-                # Create new SAAS metadata
-                # Set org_id to user_id as specified in requirements
+                # Create new SAAS metadata with the determined org_id
                 saas_metadata = StoredConversationMetadataSaas(
                     conversation_id=str(info.id),
                     user_id=user_id_uuid,
-                    org_id=user.current_org_id,
+                    org_id=org_id,
                 )
                 self.db_session.add(saas_metadata)
 

enterprise/server/utils/saas_pending_message_injector.py

@@ -0,0 +1,172 @@
+"""Enterprise injector for PendingMessageService with SAAS filtering."""
+
+from typing import AsyncGenerator
+from uuid import UUID
+
+from fastapi import Request
+from sqlalchemy import select
+from storage.stored_conversation_metadata_saas import StoredConversationMetadataSaas
+from storage.user import User
+
+from openhands.agent_server.models import ImageContent, TextContent
+from openhands.app_server.errors import AuthError
+from openhands.app_server.pending_messages.pending_message_models import (
+    PendingMessageResponse,
+)
+from openhands.app_server.pending_messages.pending_message_service import (
+    PendingMessageService,
+    PendingMessageServiceInjector,
+    SQLPendingMessageService,
+)
+from openhands.app_server.services.injector import InjectorState
+from openhands.app_server.user.specifiy_user_context import ADMIN
+from openhands.app_server.user.user_context import UserContext
+
+
+class SaasSQLPendingMessageService(SQLPendingMessageService):
+    """Extended SQLPendingMessageService with user and organization-based filtering.
+
+    This enterprise version ensures that:
+    - Users can only queue messages for conversations they own
+    - Organization isolation is enforced for multi-tenant deployments
+    """
+
+    def __init__(self, db_session, user_context: UserContext):
+        super().__init__(db_session=db_session)
+        self.user_context = user_context
+
+    async def _get_current_user(self) -> User | None:
+        """Get the current user using the existing db_session.
+
+        Returns:
+            User object or None if no user_id is available
+        """
+        user_id_str = await self.user_context.get_user_id()
+        if not user_id_str:
+            return None
+
+        user_id_uuid = UUID(user_id_str)
+        result = await self.db_session.execute(
+            select(User).where(User.id == user_id_uuid)
+        )
+        return result.scalars().first()
+
+    async def _validate_conversation_ownership(self, conversation_id: str) -> None:
+        """Validate that the current user owns the conversation.
+
+        This ensures multi-tenant isolation by checking:
+        - The conversation belongs to the current user
+        - The conversation belongs to the user's current organization
+
+        Args:
+            conversation_id: The conversation ID to validate (can be task-id or UUID)
+
+        Raises:
+            AuthError: If user doesn't own the conversation or authentication fails
+        """
+        # For internal operations (e.g., processing pending messages during startup)
+        # we need a mode that bypasses filtering. The ADMIN context enables this.
+        if self.user_context == ADMIN:
+            return
+
+        user_id_str = await self.user_context.get_user_id()
+        if not user_id_str:
+            raise AuthError('User authentication required')
+
+        user_id_uuid = UUID(user_id_str)
+
+        # Check conversation ownership via SAAS metadata
+        query = select(StoredConversationMetadataSaas).where(
+            StoredConversationMetadataSaas.conversation_id == conversation_id
+        )
+        result = await self.db_session.execute(query)
+        saas_metadata = result.scalar_one_or_none()
+
+        # If no SAAS metadata exists, the conversation might be a new task-id
+        # that hasn't been linked to a conversation yet. Allow access in this case
+        # as the message will be validated when the conversation is created.
+        if saas_metadata is None:
+            return
+
+        # Verify user ownership
+        if saas_metadata.user_id != user_id_uuid:
+            raise AuthError('You do not have access to this conversation')
+
+        # Verify organization ownership if applicable
+        user = await self._get_current_user()
+        if user and user.current_org_id is not None:
+            if saas_metadata.org_id != user.current_org_id:
+                raise AuthError('Conversation belongs to a different organization')
+
+    async def add_message(
+        self,
+        conversation_id: str,
+        content: list[TextContent | ImageContent],
+        role: str = 'user',
+    ) -> PendingMessageResponse:
+        """Queue a message with ownership validation.
+
+        Args:
+            conversation_id: The conversation ID to queue the message for
+            content: Message content
+            role: Message role (default: 'user')
+
+        Returns:
+            PendingMessageResponse with the queued message info
+
+        Raises:
+            AuthError: If user doesn't own the conversation
+        """
+        await self._validate_conversation_ownership(conversation_id)
+        return await super().add_message(conversation_id, content, role)
+
+    async def get_pending_messages(self, conversation_id: str):
+        """Get pending messages with ownership validation.
+
+        Args:
+            conversation_id: The conversation ID to get messages for
+
+        Returns:
+            List of pending messages
+
+        Raises:
+            AuthError: If user doesn't own the conversation
+        """
+        await self._validate_conversation_ownership(conversation_id)
+        return await super().get_pending_messages(conversation_id)
+
+    async def count_pending_messages(self, conversation_id: str) -> int:
+        """Count pending messages with ownership validation.
+
+        Args:
+            conversation_id: The conversation ID to count messages for
+
+        Returns:
+            Number of pending messages
+
+        Raises:
+            AuthError: If user doesn't own the conversation
+        """
+        await self._validate_conversation_ownership(conversation_id)
+        return await super().count_pending_messages(conversation_id)
+
+
+class SaasPendingMessageServiceInjector(PendingMessageServiceInjector):
+    """Enterprise injector for PendingMessageService with SAAS filtering."""
+
+    async def inject(
+        self, state: InjectorState, request: Request | None = None
+    ) -> AsyncGenerator[PendingMessageService, None]:
+        from openhands.app_server.config import (
+            get_db_session,
+            get_user_context,
+        )
+
+        async with (
+            get_user_context(state, request) as user_context,
+            get_db_session(state, request) as db_session,
+        ):
+            service = SaasSQLPendingMessageService(
+                db_session=db_session, user_context=user_context
+            )
+            yield service

enterprise/server/utils/url_utils.py

@@ -0,0 +1,38 @@
+from typing import Literal
+
+from fastapi import Request
+from server.constants import IS_FEATURE_ENV, IS_LOCAL_ENV, IS_STAGING_ENV
+from starlette.datastructures import URL
+
+from openhands.app_server.config import get_global_config
+
+
+def get_web_url(request: Request):
+    web_url = get_global_config().web_url
+    if not web_url:
+        scheme = 'http' if request.url.hostname == 'localhost' else 'https'
+        web_url = f'{scheme}://{request.url.netloc}'
+    else:
+        web_url = web_url.rstrip('/')
+    return web_url
+
+
+def get_cookie_domain() -> str | None:
+    config = get_global_config()
+    web_url = config.web_url
+    # for now just use the full hostname except for staging stacks.
+    return (
+        URL(web_url).hostname
+        if web_url and not (IS_FEATURE_ENV or IS_STAGING_ENV or IS_LOCAL_ENV)
+        else None
+    )
+
+
+def get_cookie_samesite() -> Literal['lax', 'strict']:
+    # for localhost and feature/staging stacks we set it to 'lax' as the cookie domain won't allow 'strict'
+    web_url = get_global_config().web_url
+    return (
+        'strict'
+        if web_url and not (IS_FEATURE_ENV or IS_STAGING_ENV or IS_LOCAL_ENV)
+        else 'lax'
+    )

enterprise/services/automation_executor.py

@@ -1,555 +0,0 @@
-"""Automation executor — processes events, claims and executes runs.
-
-The executor is a long-running process with three phases:
-  1. Process inbox: match NEW events to automations, create PENDING runs
-  2. Claim and execute: claim PENDING runs, submit to V1 API, heartbeat
-  3. Stale recovery: recover RUNNING runs with expired heartbeats
-"""
-
-import asyncio
-import logging
-import os
-import socket
-from datetime import datetime, timedelta, timezone
-from uuid import uuid4
-
-from services.openhands_api_client import OpenHandsAPIClient
-from sqlalchemy import or_, select, update
-from sqlalchemy.ext.asyncio import AsyncSession
-from storage.automation import Automation, AutomationRun
-from storage.automation_event import AutomationEvent
-
-logger = logging.getLogger('saas.automation.executor')
-
-# Environment-configurable settings
-POLL_INTERVAL_SECONDS = float(os.getenv('POLL_INTERVAL_SECONDS', '30'))
-HEARTBEAT_INTERVAL_SECONDS = float(os.getenv('HEARTBEAT_INTERVAL_SECONDS', '60'))
-RUN_TIMEOUT_SECONDS = float(os.getenv('RUN_TIMEOUT_SECONDS', '7200'))
-MAX_CONCURRENT_RUNS = int(os.getenv('MAX_CONCURRENT_RUNS', '5'))
-STALE_THRESHOLD_MINUTES = 5
-MAX_EVENTS_PER_BATCH = 50
-MAX_RETRIES_DEFAULT = 3
-
-# Terminal conversation statuses
-TERMINAL_STATUSES = frozenset({'STOPPED', 'ERROR', 'COMPLETED', 'CANCELLED'})
-
-# Shutdown flag — set by signal handlers
-_shutdown_event: asyncio.Event | None = None
-
-# Background task tracking for graceful shutdown
-_pending_tasks: set[asyncio.Task] = set()
-
-
-def utc_now() -> datetime:
-    return datetime.now(timezone.utc)
-
-
-def get_shutdown_event() -> asyncio.Event:
-    global _shutdown_event
-    if _shutdown_event is None:
-        _shutdown_event = asyncio.Event()
-    return _shutdown_event
-
-
-def should_continue() -> bool:
-    return not get_shutdown_event().is_set()
-
-
-def request_shutdown() -> None:
-    get_shutdown_event().set()
-
-
-# ---------------------------------------------------------------------------
-# Phase 1: Process inbox (event matching)
-# ---------------------------------------------------------------------------
-
-
-async def find_matching_automations(
-    session: AsyncSession, event: AutomationEvent
-) -> list[Automation]:
-    """Find automations that match the given event.
-
-    Phase 1 supports cron and manual triggers only — both carry
-    ``automation_id`` in the event payload.
-    """
-    source_type = event.source_type
-    payload = event.payload
-    if payload is None:
-        logger.error('Event %s has None payload — possible data corruption', event.id)
-        return []
-
-    if source_type in ('cron', 'manual'):
-        automation_id = payload.get('automation_id')
-        if not automation_id:
-            logger.warning(
-                'Event %s (source=%s) missing automation_id in payload',
-                event.id,
-                source_type,
-            )
-            return []
-
-        result = await session.execute(
-            select(Automation).where(
-                Automation.id == automation_id,
-                Automation.enabled.is_(True),
-            )
-        )
-        automation = result.scalar_one_or_none()
-        return [automation] if automation else []
-
-    logger.debug('Unhandled event source_type=%s for event %s', source_type, event.id)
-    return []
-
-
-async def process_new_events(session: AsyncSession) -> int:
-    """Claim NEW events from inbox, match to automations, create runs.
-
-    Returns the number of events processed.
-    """
-    result = await session.execute(
-        select(AutomationEvent)
-        .where(AutomationEvent.status == 'NEW')
-        .order_by(AutomationEvent.created_at)
-        .limit(MAX_EVENTS_PER_BATCH)
-        .with_for_update(skip_locked=True)
-    )
-    events = list(result.scalars())
-
-    processed = 0
-    for event in events:
-        try:
-            automations = await find_matching_automations(session, event)
-            if not automations:
-                event.status = 'NO_MATCH'
-                event.processed_at = utc_now()
-            else:
-                for automation in automations:
-                    run = AutomationRun(
-                        id=uuid4().hex,
-                        automation_id=automation.id,
-                        event_id=event.id,
-                        status='PENDING',
-                        event_payload=event.payload,
-                    )
-                    session.add(run)
-                event.status = 'PROCESSED'
-                event.processed_at = utc_now()
-            processed += 1
-        except Exception as e:
-            logger.exception('Error processing event %s', event.id)
-            event.status = 'ERROR'
-            event.error_detail = f'Failed during event matching: {type(e).__name__}: {e}'
-            event.processed_at = utc_now()
-
-    if processed:
-        await session.commit()
-        logger.info('Processed %d events', processed)
-
-    return processed
-
-
-# ---------------------------------------------------------------------------
-# Phase 2: Claim and execute runs
-# ---------------------------------------------------------------------------
-
-
-async def resolve_user_api_key(session: AsyncSession, user_id: str) -> str | None:
-    """Look up a user's API key from the api_keys table.
-
-    Returns the first active key found, or None.
-    """
-    from storage.api_key import ApiKey
-
-    result = await session.execute(
-        select(ApiKey.key).where(ApiKey.user_id == user_id).limit(1)
-    )
-    row = result.scalar_one_or_none()
-    return row
-
-
-async def download_automation_file(file_store_key: str) -> bytes:
-    """Download the automation .py file from object storage."""
-    try:
-        from openhands.server.shared import file_store
-    except ImportError as exc:
-        raise RuntimeError(
-            'file_store is not available — ensure the enterprise server '
-            'has been initialised before calling download_automation_file'
-        ) from exc
-
-    content = file_store.read(file_store_key)
-    if isinstance(content, str):
-        return content.encode('utf-8')
-    return content
-
-
-def is_terminal(conversation: dict) -> bool:
-    """Check if a conversation has reached a terminal status."""
-    status = (conversation.get('status') or '').upper()
-    return status in TERMINAL_STATUSES
-
-
-async def _prepare_run(
-    run: AutomationRun,
-    automation: Automation,
-    session_factory: object,
-) -> tuple[str, bytes]:
-    """Resolve the user's API key and download the automation file.
-
-    Returns:
-        (api_key, automation_file) tuple ready for submission.
-
-    Raises:
-        ValueError: If no API key is found.
-        RuntimeError: If file_store is unavailable.
-    """
-    async with session_factory() as key_session:
-        api_key = await resolve_user_api_key(key_session, automation.user_id)
-
-    if not api_key:
-        raise ValueError(f'No API key found for user {automation.user_id}')
-
-    automation_file = await download_automation_file(automation.file_store_key)
-    return api_key, automation_file
-
-
-async def _monitor_conversation(
-    run: AutomationRun,
-    conversation_id: str,
-    api_client: OpenHandsAPIClient,
-    api_key: str,
-    session_factory: object,
-) -> bool:
-    """Monitor a conversation until completion or timeout.
-
-    Returns True if completed successfully, False if shutdown requested.
-
-    Raises:
-        TimeoutError: If the run exceeds RUN_TIMEOUT_SECONDS.
-    """
-    start_time = utc_now()
-    while should_continue():
-        elapsed = (utc_now() - start_time).total_seconds()
-        if elapsed > RUN_TIMEOUT_SECONDS:
-            raise TimeoutError(f'Run exceeded {RUN_TIMEOUT_SECONDS}s timeout')
-
-        await asyncio.sleep(HEARTBEAT_INTERVAL_SECONDS)
-
-        # Update heartbeat
-        async with session_factory() as session:
-            run_obj = await session.get(AutomationRun, run.id)
-            if run_obj:
-                run_obj.heartbeat_at = utc_now()
-                await session.commit()
-
-        # Check conversation status
-        conversation = (
-            await api_client.get_conversation(api_key, conversation_id) or {}
-        )
-        if is_terminal(conversation):
-            return True
-
-    return False  # shutdown requested
-
-
-async def _submit_and_monitor(
-    run: AutomationRun,
-    api_key: str,
-    automation_file: bytes,
-    automation: Automation,
-    api_client: OpenHandsAPIClient,
-    session_factory: object,
-) -> None:
-    """Submit the automation to the V1 API and monitor until completion.
-
-    Updates the run's conversation_id, sends heartbeats, and marks the
-    final status when the conversation reaches a terminal state.
-    """
-    conversation = await api_client.start_conversation(
-        api_key=api_key,
-        automation_file=automation_file,
-        title=f'Automation: {automation.name}',
-        event_payload=run.event_payload,
-    )
-
-    conversation_id = conversation.get('app_conversation_id') or conversation.get(
-        'conversation_id'
-    )
-
-    # Persist conversation ID
-    async with session_factory() as update_session:
-        run_obj = await update_session.get(AutomationRun, run.id)
-        if run_obj:
-            run_obj.conversation_id = conversation_id
-            await update_session.commit()
-
-    # Monitor with heartbeats
-    completed = await _monitor_conversation(
-        run, conversation_id, api_client, api_key, session_factory
-    )
-
-    # Update final status
-    async with session_factory() as final_session:
-        run_obj = await final_session.get(AutomationRun, run.id)
-        if run_obj:
-            if not completed:
-                # Leave as RUNNING — stale recovery will handle it if needed.
-                # The conversation may still be running on the API side.
-                logger.info(
-                    'Run %s left as RUNNING due to executor shutdown', run.id
-                )
-            else:
-                run_obj.status = 'COMPLETED'
-                run_obj.completed_at = utc_now()
-                logger.info('Run %s completed successfully', run.id)
-            await final_session.commit()
-
-
-async def execute_run(
-    run: AutomationRun,
-    automation: Automation,
-    api_client: OpenHandsAPIClient,
-    session_factory: object,
-) -> None:
-    """Execute a single automation run end-to-end.
-
-    Orchestrates preparation (API key + file download) and submission/monitoring.
-    On failure, marks the run for retry or dead-letter.
-    """
-    try:
-        api_key, automation_file = await _prepare_run(
-            run, automation, session_factory
-        )
-        await _submit_and_monitor(
-            run, api_key, automation_file, automation, api_client, session_factory
-        )
-    except Exception as e:
-        logger.exception('Run %s failed: %s', run.id, e)
-        await _mark_run_failed(run, str(e), session_factory)
-
-
-async def _mark_run_failed(
-    run: AutomationRun, error: str, session_factory: object
-) -> None:
-    """Mark a run as FAILED or return to PENDING for retry."""
-    async with session_factory() as session:
-        run_obj = await session.get(AutomationRun, run.id)
-        if not run_obj:
-            return
-
-        run_obj.retry_count = (run_obj.retry_count or 0) + 1
-        run_obj.error_detail = error
-
-        if run_obj.retry_count >= (run_obj.max_retries or MAX_RETRIES_DEFAULT):
-            run_obj.status = 'DEAD_LETTER'
-            run_obj.completed_at = utc_now()
-            logger.error(
-                'Run %s moved to DEAD_LETTER after %d retries',
-                run.id,
-                run_obj.retry_count,
-            )
-        else:
-            run_obj.status = 'PENDING'
-            run_obj.claimed_by = None
-            backoff_seconds = 30 * (2 ** (run_obj.retry_count - 1))
-            run_obj.next_retry_at = utc_now() + timedelta(seconds=backoff_seconds)
-            logger.warning(
-                'Run %s returned to PENDING, retry %d/%d in %ds',
-                run.id,
-                run_obj.retry_count,
-                run_obj.max_retries or MAX_RETRIES_DEFAULT,
-                backoff_seconds,
-            )
-
-        await session.commit()
-
-
-async def claim_and_execute_runs(
-    session: AsyncSession,
-    executor_id: str,
-    api_client: OpenHandsAPIClient,
-    session_factory: object,
-) -> bool:
-    """Claim a PENDING run and start executing it.
-
-    Returns True if a run was claimed, False otherwise.
-    """
-    result = await session.execute(
-        select(AutomationRun)
-        .where(
-            AutomationRun.status == 'PENDING',
-            or_(
-                AutomationRun.next_retry_at.is_(None),
-                AutomationRun.next_retry_at <= utc_now(),
-            ),
-        )
-        .order_by(AutomationRun.created_at)
-        .limit(1)
-        .with_for_update(skip_locked=True)
-    )
-    run = result.scalar_one_or_none()
-    if not run:
-        return False
-
-    # Claim the run
-    run.status = 'RUNNING'
-    run.claimed_by = executor_id
-    run.claimed_at = utc_now()
-    run.heartbeat_at = utc_now()
-    run.started_at = utc_now()
-    await session.commit()
-
-    # Load automation for the run
-    auto_result = await session.execute(
-        select(Automation).where(Automation.id == run.automation_id)
-    )
-    automation = auto_result.scalar_one_or_none()
-    if not automation:
-        logger.error('Automation %s not found for run %s', run.automation_id, run.id)
-        await _mark_run_failed(
-            run, f'Automation {run.automation_id} not found', session_factory
-        )
-        return True
-
-    # Execute in background (long-running) with task tracking
-    task = asyncio.create_task(
-        execute_run(run, automation, api_client, session_factory),
-        name=f'execute-run-{run.id}',
-    )
-    _pending_tasks.add(task)
-    task.add_done_callback(_pending_tasks.discard)
-
-    logger.info(
-        'Claimed run %s (automation=%s) by executor %s',
-        run.id,
-        run.automation_id,
-        executor_id,
-    )
-    return True
-
-
-# ---------------------------------------------------------------------------
-# Phase 3: Stale run recovery
-# ---------------------------------------------------------------------------
-
-
-async def recover_stale_runs(session: AsyncSession) -> int:
-    """Mark RUNNING runs with expired heartbeats as PENDING for retry.
-
-    Returns the number of recovered runs.
-    """
-    stale_threshold = utc_now() - timedelta(minutes=STALE_THRESHOLD_MINUTES)
-    timeout_threshold = utc_now() - timedelta(seconds=RUN_TIMEOUT_SECONDS)
-
-    # Recover stale runs (heartbeat expired)
-    result = await session.execute(
-        update(AutomationRun)
-        .where(
-            AutomationRun.status == 'RUNNING',
-            AutomationRun.heartbeat_at < stale_threshold,
-            AutomationRun.heartbeat_at >= timeout_threshold,
-        )
-        .values(
-            status='PENDING',
-            claimed_by=None,
-            retry_count=AutomationRun.retry_count + 1,
-            next_retry_at=utc_now() + timedelta(seconds=30),
-        )
-        .returning(AutomationRun.id)
-    )
-    recovered_rows = result.fetchall()
-
-    # Mark truly timed-out runs as DEAD_LETTER
-    timeout_result = await session.execute(
-        update(AutomationRun)
-        .where(
-            AutomationRun.status == 'RUNNING',
-            AutomationRun.heartbeat_at < timeout_threshold,
-        )
-        .values(
-            status='DEAD_LETTER',
-            error_detail='Run exceeded timeout',
-            completed_at=utc_now(),
-        )
-        .returning(AutomationRun.id)
-    )
-    timed_out_rows = timeout_result.fetchall()
-
-    await session.commit()
-
-    recovered_count = len(recovered_rows)
-    timed_out_count = len(timed_out_rows)
-
-    if recovered_count:
-        logger.warning('Recovered %d stale automation runs', recovered_count)
-    if timed_out_count:
-        logger.warning(
-            'Marked %d automation runs as DEAD_LETTER (timeout)', timed_out_count
-        )
-
-    return recovered_count + timed_out_count
-
-
-# ---------------------------------------------------------------------------
-# Main executor loop
-# ---------------------------------------------------------------------------
-
-
-async def executor_main(session_factory: object | None = None) -> None:
-    """Main executor loop.
-
-    Args:
-        session_factory: Async context manager that yields AsyncSession instances.
-            If None, uses the default ``a_session_maker`` from database module.
-    """
-    if session_factory is None:
-        from storage.database import a_session_maker
-
-        session_factory = a_session_maker
-
-    executor_id = f'executor-{socket.gethostname()}-{os.getpid()}'
-    api_url = os.getenv('OPENHANDS_API_URL', 'http://openhands-service:3000')
-    api_client = OpenHandsAPIClient(base_url=api_url)
-
-    logger.info(
-        'Automation executor %s starting (api_url=%s, poll=%ss, heartbeat=%ss)',
-        executor_id,
-        api_url,
-        POLL_INTERVAL_SECONDS,
-        HEARTBEAT_INTERVAL_SECONDS,
-    )
-
-    try:
-        while should_continue():
-            try:
-                async with session_factory() as session:
-                    await process_new_events(session)
-
-                async with session_factory() as session:
-                    await claim_and_execute_runs(
-                        session, executor_id, api_client, session_factory
-                    )
-
-                async with session_factory() as session:
-                    await recover_stale_runs(session)
-
-            except Exception:
-                logger.exception('Error in executor main loop iteration')
-
-            # Wait for next poll interval (or early wakeup on shutdown)
-            try:
-                await asyncio.wait_for(
-                    get_shutdown_event().wait(),
-                    timeout=POLL_INTERVAL_SECONDS,
-                )
-            except asyncio.TimeoutError:
-                pass  # Normal — poll interval elapsed
-
-    finally:
-        if _pending_tasks:
-            logger.info(
-                'Waiting for %d running tasks to complete...', len(_pending_tasks)
-            )
-            await asyncio.gather(*_pending_tasks, return_exceptions=True)
-        await api_client.close()
-        logger.info('Automation executor %s shut down', executor_id)

enterprise/services/openhands_api_client.py

@@ -1,93 +0,0 @@
-"""HTTP client for the main OpenHands V1 API (internal cluster calls).
-
-Used by the automation executor to create and monitor conversations
-in the main OpenHands server.
-"""
-
-import base64
-import logging
-
-import httpx
-
-logger = logging.getLogger('saas.automation.api_client')
-
-
-def _raise_with_body(resp: httpx.Response) -> None:
-    """Call raise_for_status, enriching the error with the response body."""
-    try:
-        resp.raise_for_status()
-    except httpx.HTTPStatusError as e:
-        error_body = resp.text[:500] if resp.text else 'no response body'
-        raise httpx.HTTPStatusError(
-            f'{e.args[0]} — Response: {error_body}',
-            request=e.request,
-            response=e.response,
-        ) from e
-
-
-class OpenHandsAPIClient:
-    """Async HTTP client for the OpenHands V1 API."""
-
-    def __init__(self, base_url: str = 'http://openhands-service:3000'):
-        self.base_url = base_url.rstrip('/')
-        self.client = httpx.AsyncClient(base_url=self.base_url, timeout=30.0)
-
-    async def start_conversation(
-        self,
-        api_key: str,
-        automation_file: bytes,
-        title: str,
-        event_payload: dict | None = None,
-    ) -> dict:
-        """Submit an SDK script for sandboxed execution via V1 API.
-
-        Args:
-            api_key: User's API key for authentication.
-            automation_file: Raw bytes of the .py automation script.
-            title: Display title for the conversation.
-            event_payload: Optional trigger event data (injected as env var).
-
-        Returns:
-            Parsed JSON response containing conversation details.
-
-        Raises:
-            httpx.HTTPStatusError: If the API returns a non-2xx status.
-        """
-        resp = await self.client.post(
-            '/api/v1/app-conversations',
-            json={
-                'automation_file': base64.b64encode(automation_file).decode(),
-                'trigger': 'automation',
-                'title': title,
-                'event_payload': event_payload,
-            },
-            headers={'Authorization': f'Bearer {api_key}'},
-        )
-        _raise_with_body(resp)
-        return resp.json()
-
-    async def get_conversation(self, api_key: str, conversation_id: str) -> dict | None:
-        """Get conversation status.
-
-        Args:
-            api_key: User's API key for authentication.
-            conversation_id: The conversation ID to look up.
-
-        Returns:
-            Conversation data dict, or None if not found.
-
-        Raises:
-            httpx.HTTPStatusError: If the API returns a non-2xx status.
-        """
-        resp = await self.client.get(
-            '/api/v1/app-conversations',
-            params={'ids': [conversation_id]},
-            headers={'Authorization': f'Bearer {api_key}'},
-        )
-        _raise_with_body(resp)
-        conversations = resp.json()
-        return conversations[0] if conversations else None
-
-    async def close(self) -> None:
-        """Close the underlying HTTP client."""
-        await self.client.aclose()

enterprise/storage/agent_settings_utils.py

@@ -0,0 +1,38 @@
+from __future__ import annotations
+
+from typing import Any, Mapping
+
+from storage.org import Org
+from storage.org_member import OrgMember
+
+_SCHEMA_VERSION = 1
+
+
+def ensure_schema_version(agent_settings: Mapping[str, Any] | None) -> dict[str, Any]:
+    normalized = dict(agent_settings or {})
+    if normalized and 'schema_version' not in normalized:
+        normalized['schema_version'] = _SCHEMA_VERSION
+    return normalized
+
+
+def merge_agent_settings(
+    base: Mapping[str, Any] | None,
+    updates: Mapping[str, Any] | None,
+) -> dict[str, Any]:
+    merged = dict(base or {})
+    for key, value in (updates or {}).items():
+        if key == 'schema_version':
+            continue
+        if value is None:
+            merged.pop(key, None)
+        else:
+            merged[key] = value
+    return ensure_schema_version(merged)
+
+
+def get_org_agent_settings(org: Org) -> dict[str, Any]:
+    return ensure_schema_version(dict(getattr(org, 'agent_settings', {}) or {}))
+
+
+def get_org_member_agent_settings(org_member: OrgMember) -> dict[str, Any]:
+    return ensure_schema_version(dict(getattr(org_member, 'agent_settings', {}) or {}))

enterprise/storage/api_key_store.py

@@ -4,6 +4,7 @@ import secrets
 import string
 from dataclasses import dataclass
 from datetime import UTC, datetime
+from uuid import UUID
 
 from sqlalchemy import select, update
 from storage.api_key import ApiKey
@@ -13,9 +14,22 @@ from storage.user_store import UserStore
 from openhands.core.logger import openhands_logger as logger
 
 
+@dataclass
+class ApiKeyValidationResult:
+    """Result of API key validation containing user and organization info."""
+
+    user_id: str
+    org_id: UUID | None  # None for legacy API keys without org binding
+    key_id: int
+    key_name: str | None
+
+
 @dataclass
 class ApiKeyStore:
     API_KEY_PREFIX = 'sk-oh-'
+    # Prefix for system keys created by internal services (e.g., automations)
+    # Keys with this prefix are hidden from users and cannot be deleted by users
+    SYSTEM_KEY_NAME_PREFIX = '__SYSTEM__:'
 
     def generate_api_key(self, length: int = 32) -> str:
         """Generate a random API key with the sk-oh- prefix."""
@@ -23,6 +37,19 @@ class ApiKeyStore:
         random_part = ''.join(secrets.choice(alphabet) for _ in range(length))
         return f'{self.API_KEY_PREFIX}{random_part}'
 
+    @classmethod
+    def is_system_key_name(cls, name: str | None) -> bool:
+        """Check if a key name indicates a system key."""
+        return name is not None and name.startswith(cls.SYSTEM_KEY_NAME_PREFIX)
+
+    @classmethod
+    def make_system_key_name(cls, name: str) -> str:
+        """Create a system key name with the appropriate prefix.
+
+        Format: __SYSTEM__:<name>
+        """
+        return f'{cls.SYSTEM_KEY_NAME_PREFIX}{name}'
+
     async def create_api_key(
         self, user_id: str, name: str | None = None, expires_at: datetime | None = None
     ) -> str:
@@ -60,8 +87,120 @@ class ApiKeyStore:
 
         return api_key
 
-    async def validate_api_key(self, api_key: str) -> str | None:
-        """Validate an API key and return the associated user_id if valid."""
+    async def get_or_create_system_api_key(
+        self,
+        user_id: str,
+        org_id: UUID,
+        name: str,
+    ) -> str:
+        """Get or create a system API key for a user on behalf of an internal service.
+
+        If a key with the given name already exists for this user/org and is not expired,
+        returns the existing key. Otherwise, creates a new key (and deletes any expired one).
+
+        System keys are:
+        - Not visible to users in their API keys list (filtered by name prefix)
+        - Not deletable by users (protected by name prefix check)
+        - Associated with a specific org (not the user's current org)
+        - Never expire (no expiration date)
+
+        Args:
+            user_id: The ID of the user to create the key for
+            org_id: The organization ID to associate the key with
+            name: Required name for the key (will be prefixed with __SYSTEM__:)
+
+        Returns:
+            The API key (existing or newly created)
+        """
+        # Create system key name with prefix
+        system_key_name = self.make_system_key_name(name)
+
+        async with a_session_maker() as session:
+            # Check if key already exists for this user/org/name
+            result = await session.execute(
+                select(ApiKey).filter(
+                    ApiKey.user_id == user_id,
+                    ApiKey.org_id == org_id,
+                    ApiKey.name == system_key_name,
+                )
+            )
+            existing_key = result.scalars().first()
+
+            if existing_key:
+                # Check if expired
+                if existing_key.expires_at:
+                    now = datetime.now(UTC)
+                    expires_at = existing_key.expires_at
+                    if expires_at.tzinfo is None:
+                        expires_at = expires_at.replace(tzinfo=UTC)
+
+                    if expires_at < now:
+                        # Key is expired, delete it and create new one
+                        logger.info(
+                            'System API key expired, re-issuing',
+                            extra={
+                                'user_id': user_id,
+                                'org_id': str(org_id),
+                                'key_name': system_key_name,
+                            },
+                        )
+                        await session.delete(existing_key)
+                        await session.commit()
+                    else:
+                        # Key exists and is not expired, return it
+                        logger.debug(
+                            'Returning existing system API key',
+                            extra={
+                                'user_id': user_id,
+                                'org_id': str(org_id),
+                                'key_name': system_key_name,
+                            },
+                        )
+                        return existing_key.key
+                else:
+                    # Key exists and has no expiration, return it
+                    logger.debug(
+                        'Returning existing system API key',
+                        extra={
+                            'user_id': user_id,
+                            'org_id': str(org_id),
+                            'key_name': system_key_name,
+                        },
+                    )
+                    return existing_key.key
+
+        # Create new key (no expiration)
+        api_key = self.generate_api_key()
+
+        async with a_session_maker() as session:
+            key_record = ApiKey(
+                key=api_key,
+                user_id=user_id,
+                org_id=org_id,
+                name=system_key_name,
+                expires_at=None,  # System keys never expire
+            )
+            session.add(key_record)
+            await session.commit()
+
+        logger.info(
+            'Created system API key',
+            extra={
+                'user_id': user_id,
+                'org_id': str(org_id),
+                'key_name': system_key_name,
+            },
+        )
+
+        return api_key
+
+    async def validate_api_key(self, api_key: str) -> ApiKeyValidationResult | None:
+        """Validate an API key and return the associated user_id and org_id if valid.
+
+        Returns:
+            ApiKeyValidationResult if the key is valid, None otherwise.
+            The org_id may be None for legacy API keys that weren't bound to an organization.
+        """
         now = datetime.now(UTC)
 
         async with a_session_maker() as session:
@@ -89,7 +228,12 @@ class ApiKeyStore:
             )
             await session.commit()
 
-            return key_record.user_id
+            return ApiKeyValidationResult(
+                user_id=key_record.user_id,
+                org_id=key_record.org_id,
+                key_id=key_record.id,
+                key_name=key_record.name,
+            )
 
     async def delete_api_key(self, api_key: str) -> bool:
         """Delete an API key by the key value."""
@@ -105,8 +249,18 @@ class ApiKeyStore:
 
             return True
 
-    async def delete_api_key_by_id(self, key_id: int) -> bool:
-        """Delete an API key by its ID."""
+    async def delete_api_key_by_id(
+        self, key_id: int, allow_system: bool = False
+    ) -> bool:
+        """Delete an API key by its ID.
+
+        Args:
+            key_id: The ID of the key to delete
+            allow_system: If False (default), system keys cannot be deleted
+
+        Returns:
+            True if the key was deleted, False if not found or is a protected system key
+        """
         async with a_session_maker() as session:
             result = await session.execute(select(ApiKey).filter(ApiKey.id == key_id))
             key_record = result.scalars().first()
@@ -114,13 +268,26 @@ class ApiKeyStore:
             if not key_record:
                 return False
 
+            # Protect system keys from deletion unless explicitly allowed
+            if self.is_system_key_name(key_record.name) and not allow_system:
+                logger.warning(
+                    'Attempted to delete system API key',
+                    extra={'key_id': key_id, 'user_id': key_record.user_id},
+                )
+                return False
+
             await session.delete(key_record)
             await session.commit()
 
             return True
 
     async def list_api_keys(self, user_id: str) -> list[ApiKey]:
-        """List all API keys for a user."""
+        """List all user-visible API keys for a user.
+
+        This excludes:
+        - System keys (name starts with __SYSTEM__:) - created by internal services
+        - MCP_API_KEY - internal MCP key
+        """
         user = await UserStore.get_user_by_id(user_id)
         if user is None:
             raise ValueError(f'User not found: {user_id}')
@@ -129,11 +296,17 @@ class ApiKeyStore:
         async with a_session_maker() as session:
             result = await session.execute(
                 select(ApiKey).filter(
-                    ApiKey.user_id == user_id, ApiKey.org_id == org_id
+                    ApiKey.user_id == user_id,
+                    ApiKey.org_id == org_id,
                 )
             )
             keys = result.scalars().all()
-            return [key for key in keys if key.name != 'MCP_API_KEY']
+            # Filter out system keys and MCP_API_KEY
+            return [
+                key
+                for key in keys
+                if key.name != 'MCP_API_KEY' and not self.is_system_key_name(key.name)
+            ]
 
     async def retrieve_mcp_api_key(self, user_id: str) -> str | None:
         user = await UserStore.get_user_by_id(user_id)
@@ -163,17 +336,44 @@ class ApiKeyStore:
             key_record = result.scalars().first()
             return key_record.key if key_record else None
 
-    async def delete_api_key_by_name(self, user_id: str, name: str) -> bool:
-        """Delete an API key by name for a specific user."""
+    async def delete_api_key_by_name(
+        self,
+        user_id: str,
+        name: str,
+        org_id: UUID | None = None,
+        allow_system: bool = False,
+    ) -> bool:
+        """Delete an API key by name for a specific user.
+
+        Args:
+            user_id: The ID of the user whose key to delete
+            name: The name of the key to delete
+            org_id: Optional organization ID to filter by (required for system keys)
+            allow_system: If False (default), system keys cannot be deleted
+
+        Returns:
+            True if the key was deleted, False if not found or is a protected system key
+        """
         async with a_session_maker() as session:
-            result = await session.execute(
-                select(ApiKey).filter(ApiKey.user_id == user_id, ApiKey.name == name)
-            )
+            # Build the query filters
+            filters = [ApiKey.user_id == user_id, ApiKey.name == name]
+            if org_id is not None:
+                filters.append(ApiKey.org_id == org_id)
+
+            result = await session.execute(select(ApiKey).filter(*filters))
             key_record = result.scalars().first()
 
             if not key_record:
                 return False
 
+            # Protect system keys from deletion unless explicitly allowed
+            if self.is_system_key_name(key_record.name) and not allow_system:
+                logger.warning(
+                    'Attempted to delete system API key',
+                    extra={'user_id': user_id, 'key_name': name},
+                )
+                return False
+
             await session.delete(key_record)
             await session.commit()
 

enterprise/storage/automation.py

@@ -1,77 +0,0 @@
-"""SQLAlchemy models for automations and automation runs.
-
-Stub for Task 1 (Data Foundation). These models will be replaced when Task 1
-is merged into automations-phase1.
-"""
-
-from sqlalchemy import (
-    Boolean,
-    Column,
-    DateTime,
-    ForeignKey,
-    Integer,
-    String,
-    Text,
-    text,
-)
-from sqlalchemy.orm import relationship
-from sqlalchemy.types import JSON
-from storage.base import Base
-
-
-class Automation(Base):
-    __tablename__ = 'automations'
-
-    id = Column(String, primary_key=True)
-    user_id = Column(String, nullable=False, index=True)
-    org_id = Column(String, nullable=True, index=True)
-
-    name = Column(String, nullable=False)
-    enabled = Column(Boolean, nullable=False, server_default=text('true'))
-
-    config = Column(JSON, nullable=False)
-    trigger_type = Column(String, nullable=False)
-    file_store_key = Column(String, nullable=False)
-
-    last_triggered_at = Column(DateTime(timezone=True), nullable=True)
-    created_at = Column(
-        DateTime(timezone=True),
-        nullable=False,
-        server_default=text('CURRENT_TIMESTAMP'),
-    )
-    updated_at = Column(
-        DateTime(timezone=True),
-        nullable=False,
-        server_default=text('CURRENT_TIMESTAMP'),
-    )
-
-    runs = relationship('AutomationRun', back_populates='automation')
-
-
-class AutomationRun(Base):
-    __tablename__ = 'automation_runs'
-
-    id = Column(String, primary_key=True)
-    automation_id = Column(
-        String, ForeignKey('automations.id', ondelete='CASCADE'), nullable=False
-    )
-    event_id = Column(Integer, ForeignKey('automation_events.id'), nullable=True)
-    conversation_id = Column(String, nullable=True)
-    status = Column(String, nullable=False, server_default=text("'PENDING'"))
-    claimed_by = Column(String, nullable=True)
-    claimed_at = Column(DateTime(timezone=True), nullable=True)
-    heartbeat_at = Column(DateTime(timezone=True), nullable=True)
-    retry_count = Column(Integer, nullable=False, server_default=text('0'))
-    max_retries = Column(Integer, nullable=False, server_default=text('3'))
-    next_retry_at = Column(DateTime(timezone=True), nullable=True)
-    event_payload = Column(JSON, nullable=True)
-    error_detail = Column(Text, nullable=True)
-    started_at = Column(DateTime(timezone=True), nullable=True)
-    completed_at = Column(DateTime(timezone=True), nullable=True)
-    created_at = Column(
-        DateTime(timezone=True),
-        nullable=False,
-        server_default=text('CURRENT_TIMESTAMP'),
-    )
-
-    automation = relationship('Automation', back_populates='runs')

enterprise/storage/automation_event.py

@@ -1,27 +0,0 @@
-"""SQLAlchemy model for automation events (the inbox).
-
-Stub for Task 1 (Data Foundation). This model will be replaced when Task 1
-is merged into automations-phase1.
-"""
-
-from sqlalchemy import Column, DateTime, Integer, String, Text, text
-from sqlalchemy.types import JSON
-from storage.base import Base
-
-
-class AutomationEvent(Base):
-    __tablename__ = 'automation_events'
-
-    id = Column(Integer, primary_key=True, autoincrement=True)
-    source_type = Column(String, nullable=False)
-    payload = Column(JSON, nullable=False)
-    metadata_ = Column('metadata', JSON, nullable=True)
-    dedup_key = Column(String, nullable=False, unique=True)
-    status = Column(String, nullable=False, server_default=text("'NEW'"))
-    error_detail = Column(Text, nullable=True)
-    created_at = Column(
-        DateTime(timezone=True),
-        nullable=False,
-        server_default=text('CURRENT_TIMESTAMP'),
-    )
-    processed_at = Column(DateTime(timezone=True), nullable=True)

enterprise/storage/lite_llm_manager.py

@@ -29,6 +29,38 @@ KEY_VERIFICATION_TIMEOUT = 5.0
 # A very large number to represent "unlimited" until LiteLLM fixes their unlimited update bug.
 UNLIMITED_BUDGET_SETTING = 1000000000.0
 
+# Check if billing is enabled (defaults to false for enterprise deployments)
+ENABLE_BILLING = os.environ.get('ENABLE_BILLING', 'false').lower() == 'true'
+
+
+def _get_default_initial_budget() -> float | None:
+    """Get the default initial budget for new teams.
+
+    When billing is disabled (ENABLE_BILLING=false), returns None to disable
+    budget enforcement in LiteLLM. When billing is enabled, returns the
+    DEFAULT_INITIAL_BUDGET environment variable value (default 0.0).
+
+    Returns:
+        float | None: The default budget, or None to disable budget enforcement.
+    """
+    if not ENABLE_BILLING:
+        return None
+
+    try:
+        budget = float(os.environ.get('DEFAULT_INITIAL_BUDGET', 0.0))
+        if budget < 0:
+            raise ValueError(
+                f'DEFAULT_INITIAL_BUDGET must be non-negative, got {budget}'
+            )
+        return budget
+    except ValueError as e:
+        raise ValueError(
+            f'Invalid DEFAULT_INITIAL_BUDGET environment variable: {e}'
+        ) from e
+
+
+DEFAULT_INITIAL_BUDGET: float | None = _get_default_initial_budget()
+
 
 def get_openhands_cloud_key_alias(keycloak_user_id: str, org_id: str) -> str:
     """Generate the key alias for OpenHands Cloud managed keys."""
@@ -101,12 +133,15 @@ class LiteLlmManager:
             ) as client:
                 # Check if team already exists and get its budget
                 # New users joining existing orgs should inherit the team's budget
-                team_budget = 0.0
+                # When billing is disabled, DEFAULT_INITIAL_BUDGET is None
+                team_budget: float | None = DEFAULT_INITIAL_BUDGET
                 try:
                     existing_team = await LiteLlmManager._get_team(client, org_id)
                     if existing_team:
                         team_info = existing_team.get('team_info', {})
-                        team_budget = team_info.get('max_budget', 0.0) or 0.0
+                        # Preserve None from existing team (no budget enforcement)
+                        existing_budget = team_info.get('max_budget')
+                        team_budget = existing_budget
                         logger.info(
                             'LiteLlmManager:create_entries:existing_team_budget',
                             extra={
@@ -129,9 +164,33 @@ class LiteLlmManager:
                 )
 
                 if create_user:
-                    await LiteLlmManager._create_user(
+                    user_created = await LiteLlmManager._create_user(
                         client, keycloak_user_info.get('email'), keycloak_user_id
                     )
+                    if not user_created:
+                        logger.error(
+                            'create_entries_failed_user_creation',
+                            extra={
+                                'org_id': org_id,
+                                'user_id': keycloak_user_id,
+                            },
+                        )
+                        return None
+
+                # Verify user exists before proceeding with key generation
+                user_exists = await LiteLlmManager._user_exists(
+                    client, keycloak_user_id
+                )
+                if not user_exists:
+                    logger.error(
+                        'create_entries_user_not_found_before_key_generation',
+                        extra={
+                            'org_id': org_id,
+                            'user_id': keycloak_user_id,
+                            'create_user_flag': create_user,
+                        },
+                    )
+                    return None
 
                 await LiteLlmManager._add_user_to_team(
                     client, keycloak_user_id, org_id, team_budget
@@ -157,11 +216,11 @@ class LiteLlmManager:
                     None,
                 )
 
-        oss_settings.agent = 'CodeActAgent'
+        oss_settings.set_agent_setting('agent', 'CodeActAgent')
         # Use the model corresponding to the current user settings version
-        oss_settings.llm_model = get_default_litellm_model()
-        oss_settings.llm_api_key = SecretStr(key)
-        oss_settings.llm_base_url = LITE_LLM_API_URL
+        oss_settings.set_agent_setting('llm.model', get_default_litellm_model())
+        oss_settings.set_agent_setting('llm.api_key', SecretStr(key))
+        oss_settings.set_agent_setting('llm.base_url', LITE_LLM_API_URL)
         return oss_settings
 
     @staticmethod
@@ -295,10 +354,15 @@ class LiteLlmManager:
                 # Check if the database key exists in LiteLLM
                 # If not, generate a new key to prevent verification failures later
                 db_key = None
+                llm_base_url = (
+                    user_settings.agent_settings.get('llm.base_url')
+                    if user_settings and user_settings.agent_settings
+                    else None
+                )
                 if (
                     user_settings
                     and user_settings.llm_api_key
-                    and user_settings.llm_base_url == LITE_LLM_API_URL
+                    and llm_base_url == LITE_LLM_API_URL
                 ):
                     db_key = user_settings.llm_api_key
                     if hasattr(db_key, 'get_secret_value'):
@@ -334,7 +398,7 @@ class LiteLlmManager:
                         )
                         # Update user_settings with the new key so it gets stored in org_member
                         user_settings.llm_api_key = SecretStr(new_key)
-                        user_settings.llm_api_key_for_byor = SecretStr(new_key)
+                        user_settings.llm_api_key_for_byor_secret = SecretStr(new_key)
 
         logger.info(
             'LiteLlmManager:migrate_lite_llm_entries:complete',
@@ -516,25 +580,40 @@ class LiteLlmManager:
         client: httpx.AsyncClient,
         team_alias: str,
         team_id: str,
-        max_budget: float,
+        max_budget: float | None,
     ):
+        """Create a new team in LiteLLM.
+
+        Args:
+            client: The HTTP client to use.
+            team_alias: The alias for the team.
+            team_id: The ID for the team.
+            max_budget: The maximum budget for the team. When None, budget
+                enforcement is disabled (unlimited usage).
+        """
         if LITE_LLM_API_KEY is None or LITE_LLM_API_URL is None:
             logger.warning('LiteLLM API configuration not found')
             return
+
+        json_data: dict[str, Any] = {
+            'team_id': team_id,
+            'team_alias': team_alias,
+            'models': [],
+            'spend': 0,
+            'metadata': {
+                'version': ORG_SETTINGS_VERSION,
+                'model': get_default_litellm_model(),
+            },
+        }
+
+        if max_budget is not None:
+            json_data['max_budget'] = max_budget
+
         response = await client.post(
             f'{LITE_LLM_API_URL}/team/new',
-            json={
-                'team_id': team_id,
-                'team_alias': team_alias,
-                'models': [],
-                'max_budget': max_budget,
-                'spend': 0,
-                'metadata': {
-                    'version': ORG_SETTINGS_VERSION,
-                    'model': get_default_litellm_model(),
-                },
-            },
+            json=json_data,
         )
+
         # Team failed to create in litellm - this is an unforseen error state...
         if not response.is_success:
             if (
@@ -611,15 +690,48 @@ class LiteLlmManager:
             )
         response.raise_for_status()
 
+    @staticmethod
+    async def _user_exists(
+        client: httpx.AsyncClient,
+        user_id: str,
+    ) -> bool:
+        """Check if a user exists in LiteLLM.
+
+        Returns True if the user exists, False otherwise.
+        """
+        if LITE_LLM_API_KEY is None or LITE_LLM_API_URL is None:
+            return False
+        try:
+            response = await client.get(
+                f'{LITE_LLM_API_URL}/user/info?user_id={user_id}',
+            )
+            if response.is_success:
+                user_data = response.json()
+                # Check that user_info exists and has the user_id
+                user_info = user_data.get('user_info', {})
+                return user_info.get('user_id') == user_id
+            return False
+        except Exception as e:
+            logger.warning(
+                'litellm_user_exists_check_failed',
+                extra={'user_id': user_id, 'error': str(e)},
+            )
+            return False
+
     @staticmethod
     async def _create_user(
         client: httpx.AsyncClient,
         email: str | None,
         keycloak_user_id: str,
-    ):
+    ) -> bool:
+        """Create a user in LiteLLM.
+
+        Returns True if the user was created or already exists and is verified,
+        False if creation failed and user does not exist.
+        """
         if LITE_LLM_API_KEY is None or LITE_LLM_API_URL is None:
             logger.warning('LiteLLM API configuration not found')
-            return
+            return False
         response = await client.post(
             f'{LITE_LLM_API_URL}/user/new',
             json={
@@ -672,17 +784,33 @@ class LiteLlmManager:
                             'user_id': keycloak_user_id,
                         },
                     )
-                    return
+                    # Verify the user actually exists before returning success
+                    user_exists = await LiteLlmManager._user_exists(
+                        client, keycloak_user_id
+                    )
+                    if not user_exists:
+                        logger.error(
+                            'litellm_user_claimed_exists_but_not_found',
+                            extra={
+                                'user_id': keycloak_user_id,
+                                'status_code': response.status_code,
+                                'text': response.text,
+                            },
+                        )
+                        return False
+                    return True
                 logger.error(
                     'error_creating_litellm_user',
                     extra={
                         'status_code': response.status_code,
                         'text': response.text,
-                        'user_id': [keycloak_user_id],
+                        'user_id': keycloak_user_id,
                         'email': None,
                     },
                 )
+                return False
             response.raise_for_status()
+        return True
 
     @staticmethod
     async def _get_user(client: httpx.AsyncClient, user_id: str) -> dict | None:
@@ -909,19 +1037,34 @@ class LiteLlmManager:
         client: httpx.AsyncClient,
         keycloak_user_id: str,
         team_id: str,
-        max_budget: float,
+        max_budget: float | None,
     ):
+        """Add a user to a team in LiteLLM.
+
+        Args:
+            client: The HTTP client to use.
+            keycloak_user_id: The user's Keycloak ID.
+            team_id: The team ID.
+            max_budget: The maximum budget for the user in the team. When None,
+                budget enforcement is disabled (unlimited usage).
+        """
         if LITE_LLM_API_KEY is None or LITE_LLM_API_URL is None:
             logger.warning('LiteLLM API configuration not found')
             return
+
+        json_data: dict[str, Any] = {
+            'team_id': team_id,
+            'member': {'user_id': keycloak_user_id, 'role': 'user'},
+        }
+
+        if max_budget is not None:
+            json_data['max_budget_in_team'] = max_budget
+
         response = await client.post(
             f'{LITE_LLM_API_URL}/team/member_add',
-            json={
-                'team_id': team_id,
-                'member': {'user_id': keycloak_user_id, 'role': 'user'},
-                'max_budget_in_team': max_budget,
-            },
+            json=json_data,
         )
+
         # Failed to add user to team - this is an unforseen error state...
         if not response.is_success:
             if (
@@ -989,19 +1132,34 @@ class LiteLlmManager:
         client: httpx.AsyncClient,
         keycloak_user_id: str,
         team_id: str,
-        max_budget: float,
+        max_budget: float | None,
     ):
+        """Update a user's budget in a team.
+
+        Args:
+            client: The HTTP client to use.
+            keycloak_user_id: The user's Keycloak ID.
+            team_id: The team ID.
+            max_budget: The maximum budget for the user in the team. When None,
+                budget enforcement is disabled (unlimited usage).
+        """
         if LITE_LLM_API_KEY is None or LITE_LLM_API_URL is None:
             logger.warning('LiteLLM API configuration not found')
             return
+
+        json_data: dict[str, Any] = {
+            'team_id': team_id,
+            'user_id': keycloak_user_id,
+        }
+
+        if max_budget is not None:
+            json_data['max_budget_in_team'] = max_budget
+
         response = await client.post(
             f'{LITE_LLM_API_URL}/team/member_update',
-            json={
-                'team_id': team_id,
-                'user_id': keycloak_user_id,
-                'max_budget_in_team': max_budget,
-            },
+            json=json_data,
         )
+
         # Failed to update user in team - this is an unforseen error state...
         if not response.is_success:
             logger.error(
@@ -1378,7 +1536,8 @@ class LiteLlmManager:
         @functools.wraps(internal_fn)
         async def wrapper(*args, **kwargs):
             async with httpx.AsyncClient(
-                headers={'x-goog-api-key': LITE_LLM_API_KEY}
+                headers={'x-goog-api-key': LITE_LLM_API_KEY},
+                timeout=httpx.Timeout(30.0),
             ) as client:
                 return await internal_fn(client, *args, **kwargs)
 
@@ -1388,6 +1547,7 @@ class LiteLlmManager:
     create_team = staticmethod(with_http_client(_create_team))
     get_team = staticmethod(with_http_client(_get_team))
     update_team = staticmethod(with_http_client(_update_team))
+    user_exists = staticmethod(with_http_client(_user_exists))
     create_user = staticmethod(with_http_client(_create_user))
     get_user = staticmethod(with_http_client(_get_user))
     update_user = staticmethod(with_http_client(_update_user))

enterprise/storage/org.py

@@ -21,14 +21,7 @@ class Org(Base):  # type: ignore
     name = Column(String, nullable=False, unique=True)
     contact_name = Column(String, nullable=True)
     contact_email = Column(String, nullable=True)
-    agent = Column(String, nullable=True)
-    default_max_iterations = Column(Integer, nullable=True)
-    security_analyzer = Column(String, nullable=True)
-    confirmation_mode = Column(Boolean, nullable=True, default=False)
-    default_llm_model = Column(String, nullable=True)
-    default_llm_base_url = Column(String, nullable=True)
     remote_runtime_resource_factor = Column(Integer, nullable=True)
-    enable_default_condenser = Column(Boolean, nullable=False, default=True)
     billing_margin = Column(Float, nullable=True, default=DEFAULT_BILLING_MARGIN)
     enable_proactive_conversation_starters = Column(
         Boolean, nullable=False, default=True
@@ -36,7 +29,7 @@ class Org(Base):  # type: ignore
     sandbox_base_container_image = Column(String, nullable=True)
     sandbox_runtime_container_image = Column(String, nullable=True)
     org_version = Column(Integer, nullable=False, default=0)
-    mcp_config = Column(JSON, nullable=True)
+    agent_settings = Column(JSON, nullable=False, default=dict)
     # encrypted column, don't set directly, set without the underscore
     _search_api_key = Column(String, nullable=True)
     # encrypted column, don't set directly, set without the underscore
@@ -45,8 +38,8 @@ class Org(Base):  # type: ignore
     enable_solvability_analysis = Column(Boolean, nullable=True, default=False)
     v1_enabled = Column(Boolean, nullable=True)
     conversation_expiration = Column(Integer, nullable=True)
-    condenser_max_size = Column(Integer, nullable=True)
     byor_export_enabled = Column(Boolean, nullable=False, default=False)
+    sandbox_grouping_strategy = Column(String, nullable=True)
 
     # Relationships
     org_members = relationship('OrgMember', back_populates='org')

enterprise/storage/org_app_settings_store.py

@@ -13,6 +13,7 @@ from server.constants import (
 from server.routes.org_models import OrgAppSettingsUpdate
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
+from storage.agent_settings_utils import get_org_agent_settings, merge_agent_settings
 from storage.org import Org
 from storage.user import User
 
@@ -65,8 +66,13 @@ class OrgAppSettingsStore:
         """
         if org.org_version < ORG_SETTINGS_VERSION:
             org.org_version = ORG_SETTINGS_VERSION
-            org.default_llm_model = get_default_litellm_model()
-            org.llm_base_url = LITE_LLM_API_URL
+            org.agent_settings = merge_agent_settings(
+                get_org_agent_settings(org),
+                {
+                    'llm.model': get_default_litellm_model(),
+                    'llm.base_url': LITE_LLM_API_URL,
+                },
+            )
             await self.db_session.flush()
             await self.db_session.refresh(org)
 

enterprise/storage/org_llm_settings_store.py

@@ -9,6 +9,7 @@ from uuid import UUID
 from server.routes.org_models import OrgLLMSettingsUpdate
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
+from storage.agent_settings_utils import get_org_agent_settings, merge_agent_settings
 from storage.org import Org
 from storage.org_member_store import OrgMemberStore
 from storage.user import User
@@ -67,8 +68,12 @@ class OrgLLMSettingsStore:
         if not org:
             return None
 
-        # Apply updates to org (excludes llm_api_key which is member-only)
         update_data.apply_to_org(org)
+        if update_data.agent_settings is not None:
+            org.agent_settings = merge_agent_settings(
+                get_org_agent_settings(org),
+                update_data.agent_settings,
+            )
 
         # Propagate relevant settings to all org members
         member_updates = update_data.get_member_updates()

enterprise/storage/org_member.py

@@ -1,9 +1,7 @@
-"""
-SQLAlchemy model for Organization-Member relationship.
-"""
+"""SQLAlchemy model for organization-member relationships."""
 
 from pydantic import SecretStr
-from sqlalchemy import UUID, Column, ForeignKey, Integer, String
+from sqlalchemy import JSON, UUID, Column, ForeignKey, Integer, String
 from sqlalchemy.orm import relationship
 from storage.base import Base
 from storage.encrypt_utils import decrypt_value, encrypt_value
@@ -18,50 +16,30 @@ class OrgMember(Base):  # type: ignore
     user_id = Column(UUID(as_uuid=True), ForeignKey('user.id'), primary_key=True)
     role_id = Column(Integer, ForeignKey('role.id'), nullable=False)
     _llm_api_key = Column(String, nullable=False)
-    max_iterations = Column(Integer, nullable=True)
-    llm_model = Column(String, nullable=True)
-    _llm_api_key_for_byor = Column(String, nullable=True)
-    llm_base_url = Column(String, nullable=True)
+    agent_settings = Column(JSON, nullable=False, default=dict)
     status = Column(String, nullable=True)
+    mcp_config = Column(JSON, nullable=True)
 
-    # Relationships
     org = relationship('Org', back_populates='org_members')
     user = relationship('User', back_populates='org_members')
     role = relationship('Role', back_populates='org_members')
 
     def __init__(self, **kwargs):
-        # Handle known SQLAlchemy columns directly
         for key in list(kwargs):
             if hasattr(self.__class__, key):
                 setattr(self, key, kwargs.pop(key))
 
-        # Handle custom property-style fields
         if 'llm_api_key' in kwargs:
             self.llm_api_key = kwargs.pop('llm_api_key')
-        if 'llm_api_key_for_byor' in kwargs:
-            self.llm_api_key_for_byor = kwargs.pop('llm_api_key_for_byor')
 
         if kwargs:
             raise TypeError(f'Unexpected keyword arguments: {list(kwargs.keys())}')
 
     @property
     def llm_api_key(self) -> SecretStr:
-        decrypted = decrypt_value(self._llm_api_key)
-        return SecretStr(decrypted)
+        return SecretStr(decrypt_value(self._llm_api_key))
 
     @llm_api_key.setter
     def llm_api_key(self, value: str | SecretStr):
         raw = value.get_secret_value() if isinstance(value, SecretStr) else value
         self._llm_api_key = encrypt_value(raw)
-
-    @property
-    def llm_api_key_for_byor(self) -> SecretStr | None:
-        if self._llm_api_key_for_byor:
-            decrypted = decrypt_value(self._llm_api_key_for_byor)
-            return SecretStr(decrypted)
-        return None
-
-    @llm_api_key_for_byor.setter
-    def llm_api_key_for_byor(self, value: str | SecretStr | None):
-        raw = value.get_secret_value() if isinstance(value, SecretStr) else value
-        self._llm_api_key_for_byor = encrypt_value(raw) if raw else None

enterprise/storage/org_member_store.py

@@ -6,11 +6,14 @@ from typing import Optional
 from uuid import UUID
 
 from server.routes.org_models import OrgMemberLLMSettings
-from sqlalchemy import func, select, update
+from sqlalchemy import func, select
 from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy.orm import joinedload
+from storage.agent_settings_utils import (
+    get_org_member_agent_settings,
+    merge_agent_settings,
+)
 from storage.database import a_session_maker
-from storage.encrypt_utils import encrypt_value
 from storage.org_member import OrgMember
 from storage.user import User
 from storage.user_settings import UserSettings
@@ -28,8 +31,11 @@ class OrgMemberStore:
         role_id: int,
         llm_api_key: str,
         status: Optional[str] = None,
+        agent_settings: Optional[dict] = None,
     ) -> OrgMember:
         """Add a user to an organization with a specific role."""
+        agent_settings = dict(agent_settings or {})
+
         async with a_session_maker() as session:
             org_member = OrgMember(
                 org_id=org_id,
@@ -37,6 +43,7 @@ class OrgMemberStore:
                 role_id=role_id,
                 llm_api_key=llm_api_key,
                 status=status,
+                agent_settings=agent_settings,
             )
             session.add(org_member)
             await session.commit()
@@ -142,23 +149,28 @@ class OrgMemberStore:
             await session.commit()
             return True
 
+    @staticmethod
+    def get_agent_settings_from_org_member(org_member: OrgMember) -> dict[str, object]:
+        return get_org_member_agent_settings(org_member)
+
     @staticmethod
     def get_kwargs_from_settings(settings: Settings):
-        kwargs = {
-            normalized: getattr(settings, normalized)
-            for c in OrgMember.__table__.columns
-            if (normalized := c.name.lstrip('_')) and hasattr(settings, normalized)
+        return {
+            'llm_api_key': settings.get_secret_agent_setting('llm.api_key'),
+            'agent_settings': settings.normalized_agent_settings(
+                strip_secret_values=True
+            ),
         }
-        return kwargs
 
     @staticmethod
     def get_kwargs_from_user_settings(user_settings: UserSettings):
-        kwargs = {
-            normalized: getattr(user_settings, normalized)
-            for c in OrgMember.__table__.columns
-            if (normalized := c.name.lstrip('_')) and hasattr(user_settings, normalized)
+        agent_settings = dict(user_settings.agent_settings or {})
+        if agent_settings and 'schema_version' not in agent_settings:
+            agent_settings['schema_version'] = 1
+        return {
+            'llm_api_key': user_settings.llm_api_key,
+            'agent_settings': agent_settings,
         }
-        return kwargs
 
     @staticmethod
     async def get_org_members_count(
@@ -238,21 +250,34 @@ class OrgMemberStore:
         org_id: UUID,
         member_settings: OrgMemberLLMSettings,
     ) -> None:
-        """Update LLM settings for all members of an organization.
+        """Update shared LLM settings for all members of an organization.
 
         Args:
             session: Database session (passed from caller for transaction)
             org_id: Organization ID
-            member_settings: Typed LLM settings to apply to all members
+            member_settings: Shared settings to apply to all members
         """
-        # Build update values from non-None fields
         values = member_settings.model_dump(exclude_none=True)
+        if not values:
+            return
 
-        # Handle encrypted llm_api_key field - map to _llm_api_key column with encryption
-        if 'llm_api_key' in values:
-            raw_key = values.pop('llm_api_key')
-            values['_llm_api_key'] = encrypt_value(raw_key)
+        result = await session.execute(
+            select(OrgMember).where(OrgMember.org_id == org_id)
+        )
+        org_members = list(result.scalars().all())
 
-        if values:
-            stmt = update(OrgMember).where(OrgMember.org_id == org_id).values(**values)
-            await session.execute(stmt)
+        raw_key = values.pop('llm_api_key', None)
+        agent_settings_updates = values.pop('agent_settings', None)
+
+        for org_member in org_members:
+            if raw_key is not None:
+                org_member.llm_api_key = raw_key
+
+            if agent_settings_updates is not None:
+                org_member.agent_settings = merge_agent_settings(
+                    get_org_member_agent_settings(org_member),
+                    agent_settings_updates,
+                )
+
+            for key, value in values.items():
+                setattr(org_member, key, value)

enterprise/storage/org_service.py

@@ -113,7 +113,10 @@ class OrgService:
             contact_name=contact_name,
             contact_email=contact_email,
             org_version=ORG_SETTINGS_VERSION,
-            default_llm_model=get_default_litellm_model(),
+            agent_settings={
+                'schema_version': 1,
+                'llm.model': get_default_litellm_model(),
+            },
         )
 
     @staticmethod
@@ -467,42 +470,6 @@ class OrgService:
             )
             return False
 
-    @staticmethod
-    def _get_llm_settings_fields() -> set[str]:
-        """
-        Get the set of organization fields that are considered LLM settings
-        and require admin/owner role to update.
-
-        Returns:
-            set[str]: Set of field names that require elevated permissions
-        """
-        return {
-            'default_llm_model',
-            'default_llm_api_key_for_byor',
-            'default_llm_base_url',
-            'search_api_key',
-            'security_analyzer',
-            'agent',
-            'confirmation_mode',
-            'enable_default_condenser',
-            'condenser_max_size',
-        }
-
-    @staticmethod
-    def _has_llm_settings_updates(update_data: OrgUpdate) -> set[str]:
-        """
-        Check if the update contains any LLM settings fields.
-
-        Args:
-            update_data: The organization update data
-
-        Returns:
-            set[str]: Set of LLM fields being updated (empty if none)
-        """
-        llm_fields = OrgService._get_llm_settings_fields()
-        update_dict = update_data.model_dump(exclude_none=True)
-        return llm_fields.intersection(update_dict.keys())
-
     @staticmethod
     async def update_org_with_permissions(
         org_id: UUID,
@@ -571,33 +538,6 @@ class OrgService:
                 )
                 raise OrgNameExistsError(update_data.name)
 
-        # Check if update contains any LLM settings
-        llm_fields_being_updated = OrgService._has_llm_settings_updates(update_data)
-        if llm_fields_being_updated:
-            # Verify user has admin or owner role
-            has_permission = await OrgService.has_admin_or_owner_role(user_id, org_id)
-            if not has_permission:
-                logger.warning(
-                    'User attempted to update LLM settings without permission',
-                    extra={
-                        'user_id': user_id,
-                        'org_id': str(org_id),
-                        'attempted_fields': list(llm_fields_being_updated),
-                    },
-                )
-                raise PermissionError(
-                    'Admin or owner role required to update LLM settings'
-                )
-
-            logger.debug(
-                'User has permission to update LLM settings',
-                extra={
-                    'user_id': user_id,
-                    'org_id': str(org_id),
-                    'llm_fields': list(llm_fields_being_updated),
-                },
-            )
-
         # Convert to dict for OrgStore (excluding None values)
         update_dict = update_data.model_dump(exclude_none=True)
         if not update_dict:
@@ -607,6 +547,24 @@ class OrgService:
             )
             return existing_org
 
+        restricted_fields = {'agent_settings', 'search_api_key', 'sandbox_api_key'}
+        if restricted_fields.intersection(
+            update_dict
+        ) and not await OrgService.has_admin_or_owner_role(user_id, org_id):
+            logger.warning(
+                'Insufficient role for restricted organization settings update',
+                extra={
+                    'user_id': user_id,
+                    'org_id': str(org_id),
+                    'restricted_fields': sorted(
+                        restricted_fields.intersection(update_dict)
+                    ),
+                },
+            )
+            raise PermissionError(
+                'Admin or owner role required to update organization agent settings'
+            )
+
         # Perform the update
         try:
             updated_org = await OrgStore.update_org(org_id, update_dict)

enterprise/storage/org_store.py

@@ -6,6 +6,7 @@ from typing import Optional
 from uuid import UUID
 
 from server.constants import (
+    DEFAULT_V1_ENABLED,
     LITE_LLM_API_URL,
     ORG_SETTINGS_VERSION,
     get_default_litellm_model,
@@ -13,6 +14,10 @@ from server.constants import (
 from server.routes.org_models import OrgLLMSettingsUpdate, OrphanedUserError
 from sqlalchemy import select, text
 from sqlalchemy.orm import joinedload
+from storage.agent_settings_utils import (
+    get_org_agent_settings,
+    merge_agent_settings,
+)
 from storage.database import a_session_maker
 from storage.lite_llm_manager import LiteLlmManager
 from storage.org import Org
@@ -23,10 +28,32 @@ from storage.user_settings import UserSettings
 from openhands.core.logger import openhands_logger as logger
 from openhands.storage.data_models.settings import Settings
 
+_ORG_SETTINGS_EXCLUDED_FIELDS = {
+    'id',
+    'name',
+    'contact_name',
+    'contact_email',
+    'org_version',
+    'agent_settings',
+}
+_ORG_SETTINGS_FIELDS = {
+    normalized
+    for column in Org.__table__.columns
+    if (normalized := column.name.lstrip('_')) not in _ORG_SETTINGS_EXCLUDED_FIELDS
+}
+
 
 class OrgStore:
     """Store for managing organizations."""
 
+    @staticmethod
+    def get_agent_settings_from_org(org: Org) -> dict[str, object]:
+        return get_org_agent_settings(org)
+
+    @staticmethod
+    def sync_agent_settings(org: Org) -> None:
+        org.agent_settings = get_org_agent_settings(org)
+
     @staticmethod
     async def create_org(
         kwargs: dict,
@@ -35,7 +62,15 @@ class OrgStore:
         async with a_session_maker() as session:
             org = Org(**kwargs)
             org.org_version = ORG_SETTINGS_VERSION
-            org.default_llm_model = get_default_litellm_model()
+            org.agent_settings = merge_agent_settings(
+                org.agent_settings,
+                {
+                    'llm.model': get_org_agent_settings(org).get('llm.model')
+                    or get_default_litellm_model()
+                },
+            )
+            if org.v1_enabled is None:
+                org.v1_enabled = DEFAULT_V1_ENABLED
             session.add(org)
             await session.commit()
             await session.refresh(org)
@@ -89,8 +124,10 @@ class OrgStore:
                 org.id,
                 {
                     'org_version': ORG_SETTINGS_VERSION,
-                    'default_llm_model': get_default_litellm_model(),
-                    'llm_base_url': LITE_LLM_API_URL,
+                    'agent_settings': {
+                        'llm.model': get_default_litellm_model(),
+                        'llm.base_url': LITE_LLM_API_URL,
+                    },
                 },
             )
         return org
@@ -177,57 +214,45 @@ class OrgStore:
 
             if 'id' in kwargs:
                 kwargs.pop('id')
+
+            agent_settings_updates = kwargs.pop('agent_settings', None)
             for key, value in kwargs.items():
                 if hasattr(org, key):
                     setattr(org, key, value)
 
+            if agent_settings_updates is not None:
+                org.agent_settings = merge_agent_settings(
+                    get_org_agent_settings(org),
+                    agent_settings_updates,
+                )
+
             await session.commit()
             await session.refresh(org)
             return org
 
     @staticmethod
     def get_kwargs_from_settings(settings: Settings):
-        kwargs = {}
-
-        for c in Org.__table__.columns:
-            # Normalize for lookup
-            normalized = (
-                c.name.removeprefix('_default_').removeprefix('default_').lstrip('_')
-            )
-
-            if not hasattr(settings, normalized):
-                continue
-
-            # ---- FIX: Output key should drop *only* leading "_" but preserve "default" ----
-            key = c.name
-            if key.startswith('_'):
-                key = key[1:]  # remove only the very first leading underscore
-
-            kwargs[key] = getattr(settings, normalized)
-
+        kwargs = {
+            field: getattr(settings, field)
+            for field in _ORG_SETTINGS_FIELDS
+            if hasattr(settings, field)
+        }
+        kwargs['agent_settings'] = settings.normalized_agent_settings(
+            strip_secret_values=True
+        )
         return kwargs
 
     @staticmethod
     def get_kwargs_from_user_settings(user_settings: UserSettings):
-        kwargs = {}
-
-        for c in Org.__table__.columns:
-            # Normalize for lookup
-            normalized = (
-                c.name.removeprefix('_default_').removeprefix('default_').lstrip('_')
-            )
-
-            if not hasattr(user_settings, normalized):
-                continue
-
-            # ---- FIX: Output key should drop *only* leading "_" but preserve "default" ----
-            key = c.name
-            if key.startswith('_'):
-                key = key[1:]  # remove only the very first leading underscore
-
-            kwargs[key] = getattr(user_settings, normalized)
-
+        kwargs = {
+            field: getattr(user_settings, field)
+            for field in _ORG_SETTINGS_FIELDS
+            if hasattr(user_settings, field)
+        }
         kwargs['org_version'] = user_settings.user_version
+        kwargs['agent_settings'] = (
+            user_settings.to_settings().normalized_agent_settings(strip_secret_values=True)
+        )
         return kwargs
 
     @staticmethod
@@ -428,8 +453,12 @@ class OrgStore:
             if not org:
                 return None
 
-            # Apply updates to org
             llm_settings.apply_to_org(org)
+            if llm_settings.agent_settings is not None:
+                org.agent_settings = merge_agent_settings(
+                    get_org_agent_settings(org),
+                    llm_settings.agent_settings,
+                )
 
             # Propagate relevant settings to all org members
             member_updates = llm_settings.get_member_updates()

enterprise/storage/saas_conversation_validator.py

@@ -15,25 +15,27 @@ class SaasConversationValidator(ConversationValidator):
 
     async def _validate_api_key(self, api_key: str) -> str | None:
         """
-        Validate an API key and return the user_id and github_user_id if valid.
+        Validate an API key and return the user_id if valid.
 
         Args:
             api_key: The API key to validate
 
         Returns:
-            A tuple of (user_id, github_user_id) if the API key is valid, None otherwise
+            The user_id if the API key is valid, None otherwise
         """
         try:
             token_manager = TokenManager()
 
             # Validate the API key and get the user_id
             api_key_store = ApiKeyStore.get_instance()
-            user_id = await api_key_store.validate_api_key(api_key)
+            validation_result = await api_key_store.validate_api_key(api_key)
 
-            if not user_id:
+            if not validation_result:
                 logger.warning('Invalid API key')
                 return None
 
+            user_id = validation_result.user_id
+
             # Get the offline token for the user
             offline_token = await token_manager.load_offline_token(user_id)
             if not offline_token:

enterprise/storage/saas_secrets_store.py

@@ -59,12 +59,15 @@ class SaasSecretsStore(SecretsStore):
 
         async with a_session_maker() as session:
             # Incoming secrets are always the most updated ones
-            # Delete all existing records and override with incoming ones
-            await session.execute(
-                delete(StoredCustomSecrets).filter(
-                    StoredCustomSecrets.keycloak_user_id == self.user_id
-                )
+            # Delete existing records for this user AND organization only
+            delete_query = delete(StoredCustomSecrets).filter(
+                StoredCustomSecrets.keycloak_user_id == self.user_id
             )
+            if org_id is not None:
+                delete_query = delete_query.filter(StoredCustomSecrets.org_id == org_id)
+            else:
+                delete_query = delete_query.filter(StoredCustomSecrets.org_id.is_(None))
+            await session.execute(delete_query)
 
             # Prepare the new secrets data
             kwargs = item.model_dump(context={'expose_secrets': True})

enterprise/storage/saas_settings_store.py

@@ -1,12 +1,8 @@
 from __future__ import annotations
 
-import binascii
-import hashlib
 import uuid
-from base64 import b64decode, b64encode
 from dataclasses import dataclass
 
-from cryptography.fernet import Fernet
 from pydantic import SecretStr
 from server.auth.token_manager import TokenManager
 from server.constants import LITE_LLM_API_URL
@@ -14,10 +10,10 @@ from server.logger import logger
 from sqlalchemy import select, update
 from sqlalchemy.orm import joinedload
 from storage.database import a_session_maker
-from storage.encrypt_utils import encrypt_value
 from storage.lite_llm_manager import LiteLlmManager, get_openhands_cloud_key_alias
 from storage.org import Org
 from storage.org_member import OrgMember
+from storage.org_member_store import OrgMemberStore
 from storage.org_store import OrgStore
 from storage.user import User
 from storage.user_settings import UserSettings
@@ -33,7 +29,6 @@ from openhands.utils.llm import is_openhands_model
 class SaasSettingsStore(SettingsStore):
     user_id: str
     config: OpenHandsConfig
-    ENCRYPT_VALUES = ['llm_api_key', 'llm_api_key_for_byor', 'search_api_key']
 
     async def _get_user_settings_by_keycloak_id_async(
         self, keycloak_user_id: str, session=None
@@ -69,6 +64,33 @@ class SaasSettingsStore(SettingsStore):
                 )
                 return result.scalars().first()
 
+    async def _persist_agent_settings_async(
+        self, org_id: uuid.UUID, agent_settings: dict
+    ) -> None:
+        async with a_session_maker() as session:
+            stmt = (
+                update(OrgMember)
+                .where(
+                    OrgMember.org_id == org_id,
+                    OrgMember.user_id == uuid.UUID(self.user_id),
+                )
+                .values(agent_settings=agent_settings)
+            )
+            await session.execute(stmt)
+            await session.commit()
+
+    async def _persist_org_agent_settings_async(
+        self, org_id: uuid.UUID, agent_settings: dict
+    ) -> None:
+        async with a_session_maker() as session:
+            stmt = (
+                update(Org)
+                .where(Org.id == org_id)
+                .values(agent_settings=agent_settings)
+            )
+            await session.execute(stmt)
+            await session.commit()
+
     async def load(self) -> Settings | None:
         user = await UserStore.get_user_by_id(self.user_id)
         if not user:
@@ -89,6 +111,11 @@ class SaasSettingsStore(SettingsStore):
                 f'Org not found for ID {org_id} as the current org for user {self.user_id}'
             )
             return None
+        org_agent_settings = OrgStore.get_agent_settings_from_org(org)
+        member_agent_settings = OrgMemberStore.get_agent_settings_from_org_member(
+            org_member
+        )
+
         kwargs = {
             **{
                 normalized: getattr(org, c.name)
@@ -107,18 +134,26 @@ class SaasSettingsStore(SettingsStore):
             },
         }
         kwargs['llm_api_key'] = org_member.llm_api_key
-        if org_member.max_iterations:
-            kwargs['max_iterations'] = org_member.max_iterations
-        if org_member.llm_model:
-            kwargs['llm_model'] = org_member.llm_model
-        if org_member.llm_api_key_for_byor:
-            kwargs['llm_api_key_for_byor'] = org_member.llm_api_key_for_byor
-        if org_member.llm_base_url:
-            kwargs['llm_base_url'] = org_member.llm_base_url
+        if org_member.mcp_config is not None:
+            kwargs['mcp_config'] = org_member.mcp_config
+        effective_member_agent_settings = {
+            **org_agent_settings,
+            **member_agent_settings,
+        }
+        kwargs['agent_settings'] = effective_member_agent_settings
         if org.v1_enabled is None:
             kwargs['v1_enabled'] = True
+        # Apply default if sandbox_grouping_strategy is None in the database
+        if kwargs.get('sandbox_grouping_strategy') is None:
+            kwargs.pop('sandbox_grouping_strategy', None)
 
         settings = Settings(**kwargs)
+        if org_agent_settings != (org.agent_settings or {}):
+            await self._persist_org_agent_settings_async(org_id, org_agent_settings)
+        if effective_member_agent_settings != (org_member.agent_settings or {}):
+            await self._persist_agent_settings_async(
+                org_id, effective_member_agent_settings
+            )
         return settings
 
     async def store(self, item: Settings):
@@ -175,41 +210,52 @@ class SaasSettingsStore(SettingsStore):
                 )
                 return None
 
+            llm_model = item.get_agent_setting('llm.model')
+            llm_base_url = item.get_agent_setting('llm.base_url')
+            uses_managed_llm_key = not llm_base_url or llm_base_url == LITE_LLM_API_URL
+
             # Check if we need to generate an LLM key.
-            if item.llm_base_url == LITE_LLM_API_URL:
+            if uses_managed_llm_key:
                 await self._ensure_api_key(
-                    item, str(org_id), openhands_type=is_openhands_model(item.llm_model)
+                    item, str(org_id), openhands_type=is_openhands_model(llm_model)
                 )
 
+            normalized_agent_settings = item.normalized_agent_settings(
+                strip_secret_values=True
+            )
+            shared_agent_settings = {
+                key: value
+                for key, value in normalized_agent_settings.items()
+                if key not in {'llm.api_key', 'mcp_config'}
+            }
+            current_member_llm_api_key = item.get_secret_agent_setting('llm.api_key')
+            shared_llm_api_key = (
+                current_member_llm_api_key.get_secret_value()
+                if current_member_llm_api_key and not uses_managed_llm_key
+                else None
+            )
+
             kwargs = item.model_dump(context={'expose_secrets': True})
-            for model in (user, org, org_member):
-                for key, value in kwargs.items():
-                    if hasattr(model, key):
-                        setattr(model, key, value)
+            kwargs.pop('agent_settings', None)
+            for key, value in kwargs.items():
+                if hasattr(user, key):
+                    setattr(user, key, value)
+                if key != 'mcp_config' and hasattr(org, key):
+                    setattr(org, key, value)
+                if key == 'mcp_config' and hasattr(org_member, key):
+                    setattr(org_member, key, value)
 
-            # Propagate LLM settings to all org members
-            # This ensures all members see the same LLM configuration when an admin saves
-            # Note: Concurrent saves by multiple admins will result in last-write-wins.
-            # Consider adding optimistic locking if this becomes a problem.
-            member_update_values: dict = {}
-            if item.llm_model is not None:
-                member_update_values['llm_model'] = item.llm_model
-            if item.llm_base_url is not None:
-                member_update_values['llm_base_url'] = item.llm_base_url
-            if item.max_iterations is not None:
-                member_update_values['max_iterations'] = item.max_iterations
-            if item.llm_api_key is not None:
-                member_update_values['_llm_api_key'] = encrypt_value(
-                    item.llm_api_key.get_secret_value()
-                )
+            org.agent_settings = shared_agent_settings
 
-            if member_update_values:
-                stmt = (
-                    update(OrgMember)
-                    .where(OrgMember.org_id == org_id)
-                    .values(**member_update_values)
-                )
-                await session.execute(stmt)
+            result = await session.execute(select(OrgMember).filter(OrgMember.org_id == org_id))
+            org_members = list(result.scalars().all())
+            for member in org_members:
+                member.agent_settings = dict(shared_agent_settings)
+                if shared_llm_api_key is not None:
+                    member.llm_api_key = shared_llm_api_key
+
+            if current_member_llm_api_key is not None:
+                org_member.llm_api_key = current_member_llm_api_key
 
             await session.commit()
 
@@ -222,52 +268,6 @@ class SaasSettingsStore(SettingsStore):
         logger.debug(f'saas_settings_store.get_instance::{user_id}')
         return SaasSettingsStore(user_id, config)
 
-    def _should_encrypt(self, key):
-        return key in self.ENCRYPT_VALUES
-
-    def _decrypt_kwargs(self, kwargs: dict):
-        fernet = self._fernet()
-        for key, value in kwargs.items():
-            try:
-                if value is None:
-                    continue
-                if self._should_encrypt(key):
-                    if isinstance(value, SecretStr):
-                        value = fernet.decrypt(
-                            b64decode(value.get_secret_value().encode())
-                        ).decode()
-                    else:
-                        value = fernet.decrypt(b64decode(value.encode())).decode()
-                    kwargs[key] = value
-            except binascii.Error:
-                pass  # Key is in legacy format...
-
-    def _encrypt_kwargs(self, kwargs: dict):
-        fernet = self._fernet()
-        for key, value in kwargs.items():
-            if value is None:
-                continue
-
-            if isinstance(value, dict):
-                self._encrypt_kwargs(value)
-                continue
-
-            if self._should_encrypt(key):
-                if isinstance(value, SecretStr):
-                    value = b64encode(
-                        fernet.encrypt(value.get_secret_value().encode())
-                    ).decode()
-                else:
-                    value = b64encode(fernet.encrypt(value.encode())).decode()
-                kwargs[key] = value
-
-    def _fernet(self):
-        if not self.config.jwt_secret:
-            raise ValueError('jwt_secret must be defined on config')
-        jwt_secret = self.config.jwt_secret.get_secret_value()
-        fernet_key = b64encode(hashlib.sha256(jwt_secret.encode()).digest())
-        return Fernet(fernet_key)
-
     async def _ensure_api_key(
         self, item: Settings, org_id: str, openhands_type: bool = False
     ) -> None:
@@ -277,9 +277,11 @@ class SaasSettingsStore(SettingsStore):
         is valid in LiteLLM. If valid, reuses it. Otherwise, generates a new key.
         """
 
+        llm_api_key = item.get_secret_agent_setting('llm.api_key')
+
         # First, check if our current key is valid
-        if item.llm_api_key and not await LiteLlmManager.verify_existing_key(
-            item.llm_api_key.get_secret_value(),
+        if llm_api_key and not await LiteLlmManager.verify_existing_key(
+            llm_api_key.get_secret_value(),
             self.user_id,
             org_id,
             openhands_type=openhands_type,
@@ -302,7 +304,7 @@ class SaasSettingsStore(SettingsStore):
                     None,
                 )
 
-            item.llm_api_key = SecretStr(generated_key)
+            item.set_agent_setting('llm.api_key', SecretStr(generated_key))
             logger.info(
                 'saas_settings_store:store:generated_openhands_key',
                 extra={'user_id': self.user_id},

enterprise/storage/slack_conversation_store.py

@@ -25,10 +25,10 @@ class SlackConversationStore:
             return result.scalar_one_or_none()
 
     async def create_slack_conversation(
-        self, slack_converstion: SlackConversation
+        self, slack_conversation: SlackConversation
     ) -> None:
         async with a_session_maker() as session:
-            session.merge(slack_converstion)
+            await session.merge(slack_conversation)
             await session.commit()
 
     @classmethod

enterprise/storage/user.py

@@ -33,6 +33,7 @@ class User(Base):  # type: ignore
     email_verified = Column(Boolean, nullable=True)
     git_user_name = Column(String, nullable=True)
     git_user_email = Column(String, nullable=True)
+    sandbox_grouping_strategy = Column(String, nullable=True)
 
     # Relationships
     role = relationship('Role', back_populates='users')

enterprise/storage/user_settings.py

@@ -1,6 +1,10 @@
+from __future__ import annotations
+
+from pydantic import SecretStr
 from server.constants import DEFAULT_BILLING_MARGIN
 from sqlalchemy import JSON, Boolean, Column, DateTime, Float, Identity, Integer, String
 from storage.base import Base
+from storage.encrypt_utils import decrypt_legacy_value, encrypt_legacy_value
 
 
 class UserSettings(Base):  # type: ignore
@@ -8,17 +12,9 @@ class UserSettings(Base):  # type: ignore
     id = Column(Integer, Identity(), primary_key=True)
     keycloak_user_id = Column(String, nullable=True, index=True)
     language = Column(String, nullable=True)
-    agent = Column(String, nullable=True)
-    max_iterations = Column(Integer, nullable=True)
-    security_analyzer = Column(String, nullable=True)
-    confirmation_mode = Column(Boolean, nullable=True, default=False)
-    llm_model = Column(String, nullable=True)
     llm_api_key = Column(String, nullable=True)
     llm_api_key_for_byor = Column(String, nullable=True)
-    llm_base_url = Column(String, nullable=True)
     remote_runtime_resource_factor = Column(Integer, nullable=True)
-    enable_default_condenser = Column(Boolean, nullable=False, default=True)
-    condenser_max_size = Column(Integer, nullable=True)
     user_consents_to_analytics = Column(Boolean, nullable=True)
     billing_margin = Column(Float, nullable=True, default=DEFAULT_BILLING_MARGIN)
     enable_sound_notifications = Column(Boolean, nullable=True, default=False)
@@ -27,9 +23,11 @@ class UserSettings(Base):  # type: ignore
     )
     sandbox_base_container_image = Column(String, nullable=True)
     sandbox_runtime_container_image = Column(String, nullable=True)
+    sandbox_grouping_strategy = Column(String, nullable=True)
     user_version = Column(Integer, nullable=False, default=0)
     accepted_tos = Column(DateTime, nullable=True)
     mcp_config = Column(JSON, nullable=True)
+    disabled_skills = Column(JSON, nullable=True)
     search_api_key = Column(String, nullable=True)
     sandbox_api_key = Column(String, nullable=True)
     max_budget_per_task = Column(Float, nullable=True)
@@ -39,6 +37,31 @@ class UserSettings(Base):  # type: ignore
     git_user_name = Column(String, nullable=True)
     git_user_email = Column(String, nullable=True)
     v1_enabled = Column(Boolean, nullable=True)
+    agent_settings = Column(JSON, nullable=False, default=dict)
+
+    @property
+    def llm_api_key_for_byor_secret(self) -> SecretStr | None:
+        raw = self.llm_api_key_for_byor
+        if not raw:
+            return None
+        try:
+            return SecretStr(decrypt_legacy_value(raw))
+        except Exception:
+            return SecretStr(raw)
+
+    @llm_api_key_for_byor_secret.setter
+    def llm_api_key_for_byor_secret(self, value: str | SecretStr | None) -> None:
+        if value is None:
+            self.llm_api_key_for_byor = None
+            return
+        raw = value.get_secret_value() if isinstance(value, SecretStr) else value
+        self.llm_api_key_for_byor = encrypt_legacy_value(raw)
+
     already_migrated = Column(
         Boolean, nullable=True, default=False
     )  # False = not migrated, True = migrated
+
+    def to_settings(self):
+        from openhands.storage.data_models.settings import Settings
+
+        return Settings(agent_settings=dict(self.agent_settings or {}))

enterprise/storage/user_store.py

@@ -7,6 +7,7 @@ from uuid import UUID
 
 from server.auth.token_manager import TokenManager
 from server.constants import (
+    DEFAULT_V1_ENABLED,
     LITE_LLM_API_URL,
     ORG_SETTINGS_VERSION,
     PERSONAL_WORKSPACE_VERSION_TO_MODEL,
@@ -90,9 +91,6 @@ class UserStore:
             from storage.org_member_store import OrgMemberStore
 
             org_member_kwargs = OrgMemberStore.get_kwargs_from_settings(settings)
-            # avoid setting org member llm fields to use org defaults on user creation
-            del org_member_kwargs['llm_model']
-            del org_member_kwargs['llm_base_url']
             org_member = OrgMember(
                 org_id=org.id,
                 user_id=user.id,
@@ -213,14 +211,15 @@ class UserStore:
                 decrypted_user_settings, user_settings.user_version
             )
 
-            # avoids circular reference. This migrate method is temprorary until all users are migrated.
+            # Migrate stripe customer (pass session to avoid FK violation)
+            # avoids circular reference. This migrate method is temporary until all users are migrated.
             from integrations.stripe_service import migrate_customer
 
             logger.debug(
                 'user_store:migrate_user:calling_stripe_migrate_customer',
                 extra={'user_id': user_id},
             )
-            await migrate_customer(user_id, org)
+            await migrate_customer(session, user_id, org)
             logger.debug(
                 'user_store:migrate_user:done_stripe_migrate_customer',
                 extra={'user_id': user_id},
@@ -231,16 +230,23 @@ class UserStore:
             org_kwargs = OrgStore.get_kwargs_from_user_settings(decrypted_user_settings)
             org_kwargs.pop('id', None)
 
-            # if user has custom settings, set org defaults to current version
+            # If the user has custom settings, keep the org defaults minimal.
             if custom_settings:
-                org_kwargs['default_llm_model'] = get_default_litellm_model()
-                org_kwargs['llm_base_url'] = LITE_LLM_API_URL
+                org_kwargs['agent_settings'] = {
+                    'schema_version': 1,
+                    'llm.model': get_default_litellm_model(),
+                    'llm.base_url': LITE_LLM_API_URL,
+                }
                 org_kwargs['org_version'] = ORG_SETTINGS_VERSION
 
             for key, value in org_kwargs.items():
                 if hasattr(org, key):
                     setattr(org, key, value)
 
+            # Apply DEFAULT_V1_ENABLED for migrated orgs if v1_enabled was not set
+            if org.v1_enabled is None:
+                org.v1_enabled = DEFAULT_V1_ENABLED
+
             user_kwargs = UserStore.get_kwargs_from_user_settings(
                 decrypted_user_settings
             )
@@ -270,12 +276,10 @@ class UserStore:
             org_member_kwargs = OrgMemberStore.get_kwargs_from_user_settings(
                 decrypted_user_settings
             )
-
-            # if the user did not have custom settings in the old model,
-            # then use the org defaults by not setting org_member fields
             if not custom_settings:
-                del org_member_kwargs['llm_model']
-                del org_member_kwargs['llm_base_url']
+                org_member_kwargs['agent_settings'] = (
+                    OrgStore.get_agent_settings_from_org(org)
+                )
 
             org_member = OrgMember(
                 org_id=org.id,
@@ -461,13 +465,6 @@ class UserStore:
                     user_settings.llm_api_key = encrypt_legacy_value(
                         org_member.llm_api_key.get_secret_value()
                     )
-                if (
-                    org_member.llm_api_key_for_byor
-                    and org_member.llm_api_key_for_byor.get_secret_value()
-                ):
-                    user_settings.llm_api_key_for_byor = encrypt_legacy_value(
-                        org_member.llm_api_key_for_byor.get_secret_value()
-                    )
                 logger.info(
                     'user_store:downgrade_user:updated_user_settings_from_org_member',
                     extra={'user_id': user_id},
@@ -892,6 +889,8 @@ class UserStore:
             language='en', enable_proactive_conversation_starters=True
         )
 
+        default_settings.v1_enabled = DEFAULT_V1_ENABLED
+
         from storage.lite_llm_manager import LiteLlmManager
 
         settings = await LiteLlmManager.create_entries(
@@ -943,44 +942,20 @@ class UserStore:
         Returns:
             A new UserSettings object populated from the entities
         """
-        # Mapping from OrgMember fields to corresponding Org "default_" fields
-        org_member_to_org_default = {
-            'llm_model': 'default_llm_model',
-            'llm_base_url': 'default_llm_base_url',
-            'max_iterations': 'default_max_iterations',
-        }
+        from storage.org_member_store import OrgMemberStore
+        from storage.org_store import OrgStore
 
-        def get_value_with_org_fallback(field_name: str, org_member_value):
-            """Get value from OrgMember, falling back to Org default if None."""
-            if org_member_value is not None:
-                return org_member_value
-            org_default_field = org_member_to_org_default.get(field_name)
-            if org_default_field and hasattr(org, org_default_field):
-                return getattr(org, org_default_field)
-            return None
-
-        # Get values from OrgMember with Org fallback for fields with default_ prefix
-        llm_model = get_value_with_org_fallback('llm_model', org_member.llm_model)
-        llm_base_url = get_value_with_org_fallback(
-            'llm_base_url', org_member.llm_base_url
-        )
-        max_iterations = get_value_with_org_fallback(
-            'max_iterations', org_member.max_iterations
+        member_agent_settings = OrgMemberStore.get_agent_settings_from_org_member(
+            org_member
         )
+        org_agent_settings = OrgStore.get_agent_settings_from_org(org)
+        agent_settings = {**org_agent_settings, **member_agent_settings}
 
         return UserSettings(
             keycloak_user_id=user_id,
-            # OrgMember fields
             llm_api_key=org_member.llm_api_key.get_secret_value()
             if org_member.llm_api_key
             else None,
-            llm_api_key_for_byor=org_member.llm_api_key_for_byor.get_secret_value()
-            if org_member.llm_api_key_for_byor
-            else None,
-            llm_model=llm_model,
-            llm_base_url=llm_base_url,
-            max_iterations=max_iterations,
-            # User fields
             accepted_tos=user.accepted_tos,
             enable_sound_notifications=user.enable_sound_notifications,
             language=user.language,
@@ -989,18 +964,12 @@ class UserStore:
             email_verified=user.email_verified,
             git_user_name=user.git_user_name,
             git_user_email=user.git_user_email,
-            # Org fields
-            agent=org.agent,
-            security_analyzer=org.security_analyzer,
-            confirmation_mode=org.confirmation_mode,
             remote_runtime_resource_factor=org.remote_runtime_resource_factor,
-            enable_default_condenser=org.enable_default_condenser,
             billing_margin=org.billing_margin,
             enable_proactive_conversation_starters=org.enable_proactive_conversation_starters,
             sandbox_base_container_image=org.sandbox_base_container_image,
             sandbox_runtime_container_image=org.sandbox_runtime_container_image,
             user_version=org.org_version,
-            mcp_config=org.mcp_config,
             search_api_key=org.search_api_key.get_secret_value()
             if org.search_api_key
             else None,
@@ -1010,7 +979,8 @@ class UserStore:
             max_budget_per_task=org.max_budget_per_task,
             enable_solvability_analysis=org.enable_solvability_analysis,
             v1_enabled=org.v1_enabled,
-            condenser_max_size=org.condenser_max_size,
+            sandbox_grouping_strategy=org.sandbox_grouping_strategy,
+            agent_settings=agent_settings,
             already_migrated=False,
         )
 
@@ -1028,16 +998,17 @@ class UserStore:
         Returns:
             True if user has custom settings, False if using old defaults
         """
-        # Normalize values
-        user_model = (
-            user_settings.llm_model.strip() or None if user_settings.llm_model else None
+        persisted_agent_settings = user_settings.agent_settings or {}
+        user_model = persisted_agent_settings.get('llm.model') or getattr(
+            user_settings, 'llm_model', None
         )
-        user_base_url = (
-            user_settings.llm_base_url.strip() or None
-            if user_settings.llm_base_url
-            else None
+        user_base_url = persisted_agent_settings.get('llm.base_url') or getattr(
+            user_settings, 'llm_base_url', None
         )
 
+        user_model = user_model.strip() or None if user_model else None
+        user_base_url = user_base_url.strip() or None if user_base_url else None
+
         # Custom base_url = definitely custom settings (BYOK)
         if user_base_url and user_base_url != LITE_LLM_API_URL:
             return True

enterprise/sync/common_room_sync.py

@@ -1,562 +0,0 @@
-#!/usr/bin/env python3
-"""
-Common Room Sync
-
-This script queries the database to count conversations created by each user,
-then creates or updates a signal in Common Room for each user with their
-conversation count.
-"""
-
-import asyncio
-import logging
-import os
-import sys
-import time
-from datetime import UTC, datetime
-from typing import Any, Dict, List, Optional, Set
-
-import requests
-from sqlalchemy import text
-
-# Add the parent directory to the path so we can import from storage
-sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-from server.auth.token_manager import get_keycloak_admin
-from storage.database import get_engine
-
-# Configure logging
-logging.basicConfig(
-    level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
-)
-logger = logging.getLogger('common_room_sync')
-
-# Common Room API configuration
-COMMON_ROOM_API_KEY = os.environ.get('COMMON_ROOM_API_KEY')
-COMMON_ROOM_DESTINATION_SOURCE_ID = os.environ.get('COMMON_ROOM_DESTINATION_SOURCE_ID')
-COMMON_ROOM_API_BASE_URL = 'https://api.commonroom.io/community/v1'
-
-# Sync configuration
-BATCH_SIZE = int(os.environ.get('BATCH_SIZE', '100'))
-KEYCLOAK_BATCH_SIZE = int(os.environ.get('KEYCLOAK_BATCH_SIZE', '20'))
-MAX_RETRIES = int(os.environ.get('MAX_RETRIES', '3'))
-INITIAL_BACKOFF_SECONDS = float(os.environ.get('INITIAL_BACKOFF_SECONDS', '1'))
-MAX_BACKOFF_SECONDS = float(os.environ.get('MAX_BACKOFF_SECONDS', '60'))
-BACKOFF_FACTOR = float(os.environ.get('BACKOFF_FACTOR', '2'))
-RATE_LIMIT = float(os.environ.get('RATE_LIMIT', '2'))  # Requests per second
-
-
-class CommonRoomSyncError(Exception):
-    """Base exception for Common Room sync errors."""
-
-
-class DatabaseError(CommonRoomSyncError):
-    """Exception for database errors."""
-
-
-class CommonRoomAPIError(CommonRoomSyncError):
-    """Exception for Common Room API errors."""
-
-
-class KeycloakClientError(CommonRoomSyncError):
-    """Exception for Keycloak client errors."""
-
-
-def get_recent_conversations(minutes: int = 60) -> List[Dict[str, Any]]:
-    """Get conversations created in the past N minutes.
-
-    Args:
-        minutes: Number of minutes to look back for new conversations.
-
-    Returns:
-        A list of dictionaries, each containing conversation details.
-
-    Raises:
-        DatabaseError: If the database query fails.
-    """
-    try:
-        # Use a different syntax for the interval that works with pg8000
-        query = text("""
-            SELECT
-                conversation_id, user_id, title, created_at
-            FROM
-                conversation_metadata
-            WHERE
-                created_at >= NOW() - (INTERVAL '1 minute' * :minutes)
-            ORDER BY
-                created_at DESC
-        """)
-
-        with get_engine().connect() as connection:
-            result = connection.execute(query, {'minutes': minutes})
-            conversations = [
-                {
-                    'conversation_id': row[0],
-                    'user_id': row[1],
-                    'title': row[2],
-                    'created_at': row[3].isoformat() if row[3] else None,
-                }
-                for row in result
-            ]
-
-        logger.info(
-            f'Retrieved {len(conversations)} conversations created in the past {minutes} minutes'
-        )
-        return conversations
-    except Exception as e:
-        logger.exception(f'Error querying recent conversations: {e}')
-        raise DatabaseError(f'Failed to query recent conversations: {e}')
-
-
-async def get_users_from_keycloak(user_ids: Set[str]) -> Dict[str, Dict[str, Any]]:
-    """Get user information from Keycloak for a set of user IDs.
-
-    Args:
-        user_ids: A set of user IDs to look up.
-
-    Returns:
-        A dictionary mapping user IDs to user information dictionaries.
-
-    Raises:
-        KeycloakClientError: If the Keycloak API call fails.
-    """
-    try:
-        # Get Keycloak admin client
-        keycloak_admin = get_keycloak_admin()
-
-        # Create a dictionary to store user information
-        user_info_dict = {}
-
-        # Convert set to list for easier batching
-        user_id_list = list(user_ids)
-
-        # Process user IDs in batches
-        for i in range(0, len(user_id_list), KEYCLOAK_BATCH_SIZE):
-            batch = user_id_list[i : i + KEYCLOAK_BATCH_SIZE]
-            batch_tasks = []
-
-            # Create tasks for each user ID in the batch
-            for user_id in batch:
-                # Use the Keycloak admin client to get user by ID
-                batch_tasks.append(get_user_by_id(keycloak_admin, user_id))
-
-            # Run the batch of tasks concurrently
-            batch_results = await asyncio.gather(*batch_tasks, return_exceptions=True)
-
-            # Process the results
-            for user_id, result in zip(batch, batch_results):
-                if isinstance(result, Exception):
-                    logger.warning(f'Error getting user {user_id}: {result}')
-                    continue
-
-                if result and isinstance(result, dict):
-                    user_info_dict[user_id] = {
-                        'username': result.get('username'),
-                        'email': result.get('email'),
-                        'id': result.get('id'),
-                    }
-
-        logger.info(
-            f'Retrieved information for {len(user_info_dict)} users from Keycloak'
-        )
-        return user_info_dict
-
-    except Exception as e:
-        error_msg = f'Error getting users from Keycloak: {e}'
-        logger.exception(error_msg)
-        raise KeycloakClientError(error_msg)
-
-
-async def get_user_by_id(keycloak_admin, user_id: str) -> Optional[Dict[str, Any]]:
-    """Get a user from Keycloak by ID.
-
-    Args:
-        keycloak_admin: The Keycloak admin client.
-        user_id: The user ID to look up.
-
-    Returns:
-        A dictionary with the user's information, or None if not found.
-    """
-    try:
-        # Use the Keycloak admin client to get user by ID
-        user = keycloak_admin.get_user(user_id)
-        if user:
-            logger.debug(
-                f"Found user in Keycloak: {user.get('username')}, {user.get('email')}"
-            )
-            return user
-        else:
-            logger.warning(f'User {user_id} not found in Keycloak')
-            return None
-    except Exception as e:
-        logger.warning(f'Error getting user {user_id} from Keycloak: {e}')
-        return None
-
-
-def get_user_info(
-    user_id: str, user_info_cache: Dict[str, Dict[str, Any]]
-) -> Optional[Dict[str, str]]:
-    """Get the email address and GitHub username for a user from the cache.
-
-    Args:
-        user_id: The user ID to look up.
-        user_info_cache: A dictionary mapping user IDs to user information.
-
-    Returns:
-        A dictionary with the user's email and username, or None if not found.
-    """
-    # Check if the user is in the cache
-    if user_id in user_info_cache:
-        user_info = user_info_cache[user_id]
-        logger.debug(
-            f"Found user info in cache: {user_info.get('username')}, {user_info.get('email')}"
-        )
-        return user_info
-    else:
-        logger.warning(f'User {user_id} not found in user info cache')
-        return None
-
-
-def register_user_in_common_room(
-    user_id: str, email: str, github_username: str
-) -> Dict[str, Any]:
-    """Create or update a user in Common Room.
-
-    Args:
-        user_id: The user ID.
-        email: The user's email address.
-        github_username: The user's GitHub username.
-
-    Returns:
-        The API response from Common Room.
-
-    Raises:
-        CommonRoomAPIError: If the Common Room API request fails.
-    """
-    if not COMMON_ROOM_API_KEY:
-        raise CommonRoomAPIError('COMMON_ROOM_API_KEY environment variable not set')
-
-    if not COMMON_ROOM_DESTINATION_SOURCE_ID:
-        raise CommonRoomAPIError(
-            'COMMON_ROOM_DESTINATION_SOURCE_ID environment variable not set'
-        )
-
-    try:
-        headers = {
-            'Authorization': f'Bearer {COMMON_ROOM_API_KEY}',
-            'Content-Type': 'application/json',
-        }
-
-        # Create or update user in Common Room
-        user_data = {
-            'id': user_id,
-            'email': email,
-            'username': github_username,
-            'github': {'type': 'handle', 'value': github_username},
-        }
-
-        user_url = f'{COMMON_ROOM_API_BASE_URL}/source/{COMMON_ROOM_DESTINATION_SOURCE_ID}/user'
-        user_response = requests.post(user_url, headers=headers, json=user_data)
-
-        if user_response.status_code not in (200, 202):
-            logger.error(
-                f'Failed to create/update user in Common Room: {user_response.text}'
-            )
-            logger.error(f'Response status code: {user_response.status_code}')
-            raise CommonRoomAPIError(
-                f'Failed to create/update user: {user_response.text}'
-            )
-
-        logger.info(
-            f'Registered/updated user {user_id} (GitHub: {github_username}) in Common Room'
-        )
-        return user_response.json()
-    except requests.RequestException as e:
-        logger.exception(f'Error communicating with Common Room API: {e}')
-        raise CommonRoomAPIError(f'Failed to communicate with Common Room API: {e}')
-
-
-def register_conversation_activity(
-    user_id: str,
-    conversation_id: str,
-    conversation_title: str,
-    created_at: datetime,
-    email: str,
-    github_username: str,
-) -> Dict[str, Any]:
-    """Create an activity in Common Room for a new conversation.
-
-    Args:
-        user_id: The user ID who created the conversation.
-        conversation_id: The ID of the conversation.
-        conversation_title: The title of the conversation.
-        created_at: The datetime object when the conversation was created.
-        email: The user's email address.
-        github_username: The user's GitHub username.
-
-    Returns:
-        The API response from Common Room.
-
-    Raises:
-        CommonRoomAPIError: If the Common Room API request fails.
-    """
-    if not COMMON_ROOM_API_KEY:
-        raise CommonRoomAPIError('COMMON_ROOM_API_KEY environment variable not set')
-
-    if not COMMON_ROOM_DESTINATION_SOURCE_ID:
-        raise CommonRoomAPIError(
-            'COMMON_ROOM_DESTINATION_SOURCE_ID environment variable not set'
-        )
-
-    try:
-        headers = {
-            'Authorization': f'Bearer {COMMON_ROOM_API_KEY}',
-            'Content-Type': 'application/json',
-        }
-
-        # Format the datetime object to the expected ISO format
-        formatted_timestamp = (
-            created_at.strftime('%Y-%m-%dT%H:%M:%SZ')
-            if created_at
-            else time.strftime('%Y-%m-%dT%H:%M:%SZ', time.gmtime())
-        )
-
-        # Create activity for the conversation
-        activity_data = {
-            'id': f'conversation_{conversation_id}',  # Use conversation ID to ensure uniqueness
-            'activityType': 'started_session',
-            'user': {
-                'id': user_id,
-                'email': email,
-                'github': {'type': 'handle', 'value': github_username},
-                'username': github_username,
-            },
-            'activityTitle': {
-                'type': 'text',
-                'value': conversation_title or 'New Conversation',
-            },
-            'content': {
-                'type': 'text',
-                'value': f'Started a new conversation: {conversation_title or "Untitled"}',
-            },
-            'timestamp': formatted_timestamp,
-            'url': f'https://app.all-hands.dev/conversations/{conversation_id}',
-        }
-
-        # Log the activity data for debugging
-        logger.info(f'Activity data payload: {activity_data}')
-
-        activity_url = f'{COMMON_ROOM_API_BASE_URL}/source/{COMMON_ROOM_DESTINATION_SOURCE_ID}/activity'
-        activity_response = requests.post(
-            activity_url, headers=headers, json=activity_data
-        )
-
-        if activity_response.status_code not in (200, 202):
-            logger.error(
-                f'Failed to create activity in Common Room: {activity_response.text}'
-            )
-            logger.error(f'Response status code: {activity_response.status_code}')
-            raise CommonRoomAPIError(
-                f'Failed to create activity: {activity_response.text}'
-            )
-
-        logger.info(
-            f'Registered conversation activity for user {user_id}, conversation {conversation_id}'
-        )
-        return activity_response.json()
-    except requests.RequestException as e:
-        logger.exception(f'Error communicating with Common Room API: {e}')
-        raise CommonRoomAPIError(f'Failed to communicate with Common Room API: {e}')
-
-
-def retry_with_backoff(func, *args, **kwargs):
-    """Retry a function with exponential backoff.
-
-    Args:
-        func: The function to retry.
-        *args: Positional arguments to pass to the function.
-        **kwargs: Keyword arguments to pass to the function.
-
-    Returns:
-        The result of the function call.
-
-    Raises:
-        The last exception raised by the function.
-    """
-    backoff = INITIAL_BACKOFF_SECONDS
-    last_exception = None
-
-    for attempt in range(MAX_RETRIES):
-        try:
-            return func(*args, **kwargs)
-        except Exception as e:
-            last_exception = e
-            logger.warning(f'Attempt {attempt + 1}/{MAX_RETRIES} failed: {e}')
-
-            if attempt < MAX_RETRIES - 1:
-                sleep_time = min(backoff, MAX_BACKOFF_SECONDS)
-                logger.info(f'Retrying in {sleep_time:.2f} seconds...')
-                time.sleep(sleep_time)
-                backoff *= BACKOFF_FACTOR
-            else:
-                logger.exception(f'All {MAX_RETRIES} attempts failed')
-                raise last_exception
-
-
-async def retry_with_backoff_async(func, *args, **kwargs):
-    """Retry an async function with exponential backoff.
-
-    Args:
-        func: The async function to retry.
-        *args: Positional arguments to pass to the function.
-        **kwargs: Keyword arguments to pass to the function.
-
-    Returns:
-        The result of the function call.
-
-    Raises:
-        The last exception raised by the function.
-    """
-    backoff = INITIAL_BACKOFF_SECONDS
-    last_exception = None
-
-    for attempt in range(MAX_RETRIES):
-        try:
-            return await func(*args, **kwargs)
-        except Exception as e:
-            last_exception = e
-            logger.warning(f'Attempt {attempt + 1}/{MAX_RETRIES} failed: {e}')
-
-            if attempt < MAX_RETRIES - 1:
-                sleep_time = min(backoff, MAX_BACKOFF_SECONDS)
-                logger.info(f'Retrying in {sleep_time:.2f} seconds...')
-                await asyncio.sleep(sleep_time)
-                backoff *= BACKOFF_FACTOR
-            else:
-                logger.exception(f'All {MAX_RETRIES} attempts failed')
-                raise last_exception
-
-
-async def async_sync_recent_conversations_to_common_room(minutes: int = 60):
-    """Async main function to sync recent conversations to Common Room.
-
-    Args:
-        minutes: Number of minutes to look back for new conversations.
-    """
-    logger.info(
-        f'Starting Common Room recent conversations sync (past {minutes} minutes)'
-    )
-
-    stats = {
-        'total_conversations': 0,
-        'registered_users': 0,
-        'registered_activities': 0,
-        'errors': 0,
-        'missing_user_info': 0,
-    }
-
-    try:
-        # Get conversations created in the past N minutes
-        recent_conversations = retry_with_backoff(get_recent_conversations, minutes)
-        stats['total_conversations'] = len(recent_conversations)
-
-        logger.info(f'Processing {len(recent_conversations)} recent conversations')
-
-        if not recent_conversations:
-            logger.info('No recent conversations found, exiting')
-            return
-
-        # Extract all unique user IDs
-        user_ids = {conv['user_id'] for conv in recent_conversations if conv['user_id']}
-
-        # Get user information for all users in batches
-        user_info_cache = await retry_with_backoff_async(
-            get_users_from_keycloak, user_ids
-        )
-
-        # Track registered users to avoid duplicate registrations
-        registered_users = set()
-
-        # Process each conversation
-        for conversation in recent_conversations:
-            conversation_id = conversation['conversation_id']
-            user_id = conversation['user_id']
-            title = conversation['title']
-            created_at = conversation[
-                'created_at'
-            ]  # This might be a string or datetime object
-
-            try:
-                # Get user info from cache
-                user_info = get_user_info(user_id, user_info_cache)
-                if not user_info:
-                    logger.warning(
-                        f'Could not find user info for user {user_id}, skipping conversation {conversation_id}'
-                    )
-                    stats['missing_user_info'] += 1
-                    continue
-
-                email = user_info['email']
-                github_username = user_info['username']
-
-                if not email:
-                    logger.warning(
-                        f'User {user_id} has no email, skipping conversation {conversation_id}'
-                    )
-                    stats['errors'] += 1
-                    continue
-
-                # Register user in Common Room if not already registered in this run
-                if user_id not in registered_users:
-                    register_user_in_common_room(user_id, email, github_username)
-                    registered_users.add(user_id)
-                    stats['registered_users'] += 1
-
-                # If created_at is a string, parse it to a datetime object
-                # If it's already a datetime object, use it as is
-                # If it's None, use current time
-                created_at_datetime = (
-                    created_at
-                    if isinstance(created_at, datetime)
-                    else datetime.fromisoformat(created_at.replace('Z', '+00:00'))
-                    if created_at
-                    else datetime.now(UTC)
-                )
-
-                # Register conversation activity with email and github username
-                register_conversation_activity(
-                    user_id,
-                    conversation_id,
-                    title,
-                    created_at_datetime,
-                    email,
-                    github_username,
-                )
-                stats['registered_activities'] += 1
-
-                # Sleep to respect rate limit
-                await asyncio.sleep(1 / RATE_LIMIT)
-            except Exception as e:
-                logger.exception(
-                    f'Error processing conversation {conversation_id} for user {user_id}: {e}'
-                )
-                stats['errors'] += 1
-    except Exception as e:
-        logger.exception(f'Sync failed: {e}')
-        raise
-    finally:
-        logger.info(f'Sync completed. Stats: {stats}')
-
-
-def sync_recent_conversations_to_common_room(minutes: int = 60):
-    """Main function to sync recent conversations to Common Room.
-
-    Args:
-        minutes: Number of minutes to look back for new conversations.
-    """
-    # Run the async function in the event loop
-    asyncio.run(async_sync_recent_conversations_to_common_room(minutes))
-
-
-if __name__ == '__main__':
-    # Default to looking back 60 minutes for new conversations
-    minutes = int(os.environ.get('SYNC_MINUTES', '60'))
-    sync_recent_conversations_to_common_room(minutes)

enterprise/sync/test_common_room_sync.py

@@ -1,51 +0,0 @@
-#!/usr/bin/env python3
-"""
-Test script for Common Room conversation count sync.
-
-This script tests the functionality of the Common Room sync script
-without making any API calls to Common Room or database connections.
-"""
-
-import os
-import sys
-import unittest
-from unittest.mock import MagicMock, patch
-
-sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-from sync.common_room_sync import (
-    retry_with_backoff,
-)
-
-
-class TestCommonRoomSync(unittest.TestCase):
-    """Test cases for Common Room sync functionality."""
-
-    def test_retry_with_backoff(self):
-        """Test the retry_with_backoff function."""
-        # Mock function that succeeds on the second attempt
-        mock_func = MagicMock(
-            side_effect=[Exception('First attempt failed'), 'success']
-        )
-
-        # Set environment variables for testing
-        with patch.dict(
-            os.environ,
-            {
-                'MAX_RETRIES': '3',
-                'INITIAL_BACKOFF_SECONDS': '0.01',
-                'BACKOFF_FACTOR': '2',
-                'MAX_BACKOFF_SECONDS': '1',
-            },
-        ):
-            result = retry_with_backoff(mock_func, 'arg1', 'arg2', kwarg1='kwarg1')
-
-            # Check that the function was called twice
-            self.assertEqual(mock_func.call_count, 2)
-            # Check that the function was called with the correct arguments
-            mock_func.assert_called_with('arg1', 'arg2', kwarg1='kwarg1')
-            # Check that the function returned the expected result
-            self.assertEqual(result, 'success')
-
-
-if __name__ == '__main__':
-    unittest.main()

enterprise/sync/test_conversation_count_query.py

@@ -1,83 +0,0 @@
-#!/usr/bin/env python3
-"""Test script to verify the conversation count query.
-
-This script tests the database query to count conversations by user,
-without making any API calls to Common Room.
-"""
-
-import os
-import sys
-
-from sqlalchemy import text
-
-# Add the parent directory to the path so we can import from storage
-sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-
-from storage.database import get_engine
-
-
-def test_conversation_count_query():
-    """Test the query to count conversations by user."""
-    try:
-        # Query to count conversations by user
-        count_query = text("""
-            SELECT
-                user_id, COUNT(*) as conversation_count
-            FROM
-                conversation_metadata
-            GROUP BY
-                user_id
-        """)
-
-        engine = get_engine()
-
-        with engine.connect() as connection:
-            count_result = connection.execute(count_query)
-            user_counts = [
-                {'user_id': row[0], 'conversation_count': row[1]}
-                for row in count_result
-            ]
-
-        print(f'Found {len(user_counts)} users with conversations')
-
-        # Print the first 5 results
-        for i, user_data in enumerate(user_counts[:5]):
-            print(
-                f"User {i+1}: {user_data['user_id']} - {user_data['conversation_count']} conversations"
-            )
-
-        # Test the user_entity query for the first user (if any)
-        if user_counts:
-            first_user_id = user_counts[0]['user_id']
-
-            user_query = text("""
-                SELECT username, email, id
-                FROM user_entity
-                WHERE id = :user_id
-            """)
-
-            with engine.connect() as connection:
-                user_result = connection.execute(user_query, {'user_id': first_user_id})
-                user_row = user_result.fetchone()
-
-                if user_row:
-                    print(f'\nUser details for {first_user_id}:')
-                    print(f'  GitHub Username: {user_row[0]}')
-                    print(f'  Email: {user_row[1]}')
-                    print(f'  ID: {user_row[2]}')
-                else:
-                    print(
-                        f'\nNo user details found for {first_user_id} in user_entity table'
-                    )
-
-        print('\nTest completed successfully')
-    except Exception as e:
-        print(f'Error: {str(e)}')
-        import traceback
-
-        traceback.print_exc()
-        sys.exit(1)
-
-
-if __name__ == '__main__':
-    test_conversation_count_query()

enterprise/tests/unit/conftest.py

@@ -1,3 +1,4 @@
+import os
 import uuid
 from datetime import datetime
 from uuid import UUID
@@ -28,6 +29,7 @@ from storage.org import Org
 from storage.org_invitation import OrgInvitation  # noqa: F401
 from storage.org_member import OrgMember
 from storage.role import Role
+from storage.slack_conversation import SlackConversation  # noqa: F401
 from storage.stored_conversation_metadata import StoredConversationMetadata
 from storage.stored_conversation_metadata_saas import (
     StoredConversationMetadataSaas,
@@ -35,6 +37,20 @@ from storage.stored_conversation_metadata_saas import (
 from storage.stored_offline_token import StoredOfflineToken
 from storage.stripe_customer import StripeCustomer
 from storage.user import User
+from storage.user_settings import UserSettings  # noqa: F401
+
+
+@pytest.fixture(autouse=True)
+def allow_short_context_windows():
+    old = os.environ.get('ALLOW_SHORT_CONTEXT_WINDOWS')
+    os.environ['ALLOW_SHORT_CONTEXT_WINDOWS'] = 'true'
+    try:
+        yield
+    finally:
+        if old is None:
+            os.environ.pop('ALLOW_SHORT_CONTEXT_WINDOWS', None)
+        else:
+            os.environ['ALLOW_SHORT_CONTEXT_WINDOWS'] = old
 
 
 @pytest.fixture
@@ -170,7 +186,6 @@ def add_minimal_fixtures(session_maker):
                 id=uuid.UUID('5594c7b6-f959-4b81-92e9-b09c206f5081'),
                 name='mock-org',
                 org_version=ORG_SETTINGS_VERSION,
-                enable_default_condenser=True,
                 enable_proactive_conversation_starters=True,
             )
         )

enterprise/tests/unit/routes/test_service.py

@@ -0,0 +1,325 @@
+"""Unit tests for service API routes."""
+
+import uuid
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+from fastapi import HTTPException
+from server.routes.service import (
+    CreateUserApiKeyRequest,
+    delete_user_api_key,
+    get_or_create_api_key_for_user,
+    validate_service_api_key,
+)
+
+
+class TestValidateServiceApiKey:
+    """Test cases for validate_service_api_key."""
+
+    @pytest.mark.asyncio
+    async def test_valid_service_key(self):
+        """Test validation with valid service API key."""
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', 'test-service-key'):
+            result = await validate_service_api_key('test-service-key')
+        assert result == 'automations-service'
+
+    @pytest.mark.asyncio
+    async def test_missing_service_key(self):
+        """Test validation with missing service API key header."""
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', 'test-service-key'):
+            with pytest.raises(HTTPException) as exc_info:
+                await validate_service_api_key(None)
+        assert exc_info.value.status_code == 401
+        assert 'X-Service-API-Key header is required' in exc_info.value.detail
+
+    @pytest.mark.asyncio
+    async def test_invalid_service_key(self):
+        """Test validation with invalid service API key."""
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', 'test-service-key'):
+            with pytest.raises(HTTPException) as exc_info:
+                await validate_service_api_key('wrong-key')
+        assert exc_info.value.status_code == 401
+        assert 'Invalid service API key' in exc_info.value.detail
+
+    @pytest.mark.asyncio
+    async def test_service_auth_not_configured(self):
+        """Test validation when service auth is not configured."""
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', ''):
+            with pytest.raises(HTTPException) as exc_info:
+                await validate_service_api_key('any-key')
+        assert exc_info.value.status_code == 503
+        assert 'Service authentication not configured' in exc_info.value.detail
+
+
+class TestCreateUserApiKeyRequest:
+    """Test cases for CreateUserApiKeyRequest validation."""
+
+    def test_valid_request(self):
+        """Test valid request with all fields."""
+        request = CreateUserApiKeyRequest(
+            name='automation',
+        )
+        assert request.name == 'automation'
+
+    def test_name_is_required(self):
+        """Test that name field is required."""
+        with pytest.raises(ValueError):
+            CreateUserApiKeyRequest(
+                name='',  # Empty name should fail
+            )
+
+    def test_name_is_stripped(self):
+        """Test that name field is stripped of whitespace."""
+        request = CreateUserApiKeyRequest(
+            name='  automation  ',
+        )
+        assert request.name == 'automation'
+
+    def test_whitespace_only_name_fails(self):
+        """Test that whitespace-only name fails validation."""
+        with pytest.raises(ValueError):
+            CreateUserApiKeyRequest(
+                name='   ',
+            )
+
+
+class TestGetOrCreateApiKeyForUser:
+    """Test cases for get_or_create_api_key_for_user endpoint."""
+
+    @pytest.fixture
+    def valid_user_id(self):
+        """Return a valid user ID."""
+        return '5594c7b6-f959-4b81-92e9-b09c206f5081'
+
+    @pytest.fixture
+    def valid_org_id(self):
+        """Return a valid org ID."""
+        return uuid.UUID('5594c7b6-f959-4b81-92e9-b09c206f5081')
+
+    @pytest.fixture
+    def valid_request(self):
+        """Create a valid request object."""
+        return CreateUserApiKeyRequest(
+            name='automation',
+        )
+
+    @pytest.mark.asyncio
+    async def test_user_not_found(self, valid_user_id, valid_org_id, valid_request):
+        """Test error when user doesn't exist."""
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', 'test-key'):
+            with patch(
+                'server.routes.service.UserStore.get_user_by_id', new_callable=AsyncMock
+            ) as mock_get_user:
+                mock_get_user.return_value = None
+                with pytest.raises(HTTPException) as exc_info:
+                    await get_or_create_api_key_for_user(
+                        user_id=valid_user_id,
+                        org_id=valid_org_id,
+                        request=valid_request,
+                        x_service_api_key='test-key',
+                    )
+        assert exc_info.value.status_code == 404
+        assert 'not found' in exc_info.value.detail
+
+    @pytest.mark.asyncio
+    async def test_user_not_in_org(self, valid_user_id, valid_org_id, valid_request):
+        """Test error when user is not a member of the org."""
+        mock_user = MagicMock()
+
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', 'test-key'):
+            with patch(
+                'server.routes.service.UserStore.get_user_by_id', new_callable=AsyncMock
+            ) as mock_get_user:
+                with patch(
+                    'server.routes.service.OrgMemberStore.get_org_member',
+                    new_callable=AsyncMock,
+                ) as mock_get_member:
+                    mock_get_user.return_value = mock_user
+                    mock_get_member.return_value = None
+                    with pytest.raises(HTTPException) as exc_info:
+                        await get_or_create_api_key_for_user(
+                            user_id=valid_user_id,
+                            org_id=valid_org_id,
+                            request=valid_request,
+                            x_service_api_key='test-key',
+                        )
+        assert exc_info.value.status_code == 403
+        assert 'not a member of org' in exc_info.value.detail
+
+    @pytest.mark.asyncio
+    async def test_successful_key_creation(
+        self, valid_user_id, valid_org_id, valid_request
+    ):
+        """Test successful API key creation."""
+        mock_user = MagicMock()
+        mock_org_member = MagicMock()
+        mock_api_key_store = MagicMock()
+        mock_api_key_store.get_or_create_system_api_key = AsyncMock(
+            return_value='sk-oh-test-key-12345678901234567890'
+        )
+
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', 'test-key'):
+            with patch(
+                'server.routes.service.UserStore.get_user_by_id', new_callable=AsyncMock
+            ) as mock_get_user:
+                with patch(
+                    'server.routes.service.OrgMemberStore.get_org_member',
+                    new_callable=AsyncMock,
+                ) as mock_get_member:
+                    with patch(
+                        'server.routes.service.ApiKeyStore.get_instance'
+                    ) as mock_get_store:
+                        mock_get_user.return_value = mock_user
+                        mock_get_member.return_value = mock_org_member
+                        mock_get_store.return_value = mock_api_key_store
+
+                        response = await get_or_create_api_key_for_user(
+                            user_id=valid_user_id,
+                            org_id=valid_org_id,
+                            request=valid_request,
+                            x_service_api_key='test-key',
+                        )
+
+        assert response.key == 'sk-oh-test-key-12345678901234567890'
+        assert response.user_id == valid_user_id
+        assert response.org_id == str(valid_org_id)
+        assert response.name == 'automation'
+
+        # Verify the store was called with correct arguments
+        mock_api_key_store.get_or_create_system_api_key.assert_called_once_with(
+            user_id=valid_user_id,
+            org_id=valid_org_id,
+            name='automation',
+        )
+
+    @pytest.mark.asyncio
+    async def test_store_exception_handling(
+        self, valid_user_id, valid_org_id, valid_request
+    ):
+        """Test error handling when store raises exception."""
+        mock_user = MagicMock()
+        mock_org_member = MagicMock()
+        mock_api_key_store = MagicMock()
+        mock_api_key_store.get_or_create_system_api_key = AsyncMock(
+            side_effect=Exception('Database error')
+        )
+
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', 'test-key'):
+            with patch(
+                'server.routes.service.UserStore.get_user_by_id', new_callable=AsyncMock
+            ) as mock_get_user:
+                with patch(
+                    'server.routes.service.OrgMemberStore.get_org_member',
+                    new_callable=AsyncMock,
+                ) as mock_get_member:
+                    with patch(
+                        'server.routes.service.ApiKeyStore.get_instance'
+                    ) as mock_get_store:
+                        mock_get_user.return_value = mock_user
+                        mock_get_member.return_value = mock_org_member
+                        mock_get_store.return_value = mock_api_key_store
+
+                        with pytest.raises(HTTPException) as exc_info:
+                            await get_or_create_api_key_for_user(
+                                user_id=valid_user_id,
+                                org_id=valid_org_id,
+                                request=valid_request,
+                                x_service_api_key='test-key',
+                            )
+
+        assert exc_info.value.status_code == 500
+        assert 'Failed to get or create API key' in exc_info.value.detail
+
+
+class TestDeleteUserApiKey:
+    """Test cases for delete_user_api_key endpoint."""
+
+    @pytest.fixture
+    def valid_org_id(self):
+        """Return a valid org ID."""
+        return uuid.UUID('5594c7b6-f959-4b81-92e9-b09c206f5081')
+
+    @pytest.mark.asyncio
+    async def test_successful_delete(self, valid_org_id):
+        """Test successful deletion of a system API key."""
+        mock_api_key_store = MagicMock()
+        mock_api_key_store.make_system_key_name.return_value = '__SYSTEM__:automation'
+        mock_api_key_store.delete_api_key_by_name = AsyncMock(return_value=True)
+
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', 'test-key'):
+            with patch(
+                'server.routes.service.ApiKeyStore.get_instance'
+            ) as mock_get_store:
+                mock_get_store.return_value = mock_api_key_store
+
+                response = await delete_user_api_key(
+                    user_id='user-123',
+                    org_id=valid_org_id,
+                    key_name='automation',
+                    x_service_api_key='test-key',
+                )
+
+        assert response == {'message': 'API key deleted successfully'}
+
+        # Verify the store was called with correct arguments
+        mock_api_key_store.make_system_key_name.assert_called_once_with('automation')
+        mock_api_key_store.delete_api_key_by_name.assert_called_once_with(
+            user_id='user-123',
+            org_id=valid_org_id,
+            name='__SYSTEM__:automation',
+            allow_system=True,
+        )
+
+    @pytest.mark.asyncio
+    async def test_delete_key_not_found(self, valid_org_id):
+        """Test error when key to delete is not found."""
+        mock_api_key_store = MagicMock()
+        mock_api_key_store.make_system_key_name.return_value = '__SYSTEM__:nonexistent'
+        mock_api_key_store.delete_api_key_by_name = AsyncMock(return_value=False)
+
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', 'test-key'):
+            with patch(
+                'server.routes.service.ApiKeyStore.get_instance'
+            ) as mock_get_store:
+                mock_get_store.return_value = mock_api_key_store
+
+                with pytest.raises(HTTPException) as exc_info:
+                    await delete_user_api_key(
+                        user_id='user-123',
+                        org_id=valid_org_id,
+                        key_name='nonexistent',
+                        x_service_api_key='test-key',
+                    )
+
+        assert exc_info.value.status_code == 404
+        assert 'not found' in exc_info.value.detail
+
+    @pytest.mark.asyncio
+    async def test_delete_invalid_service_key(self, valid_org_id):
+        """Test error when service API key is invalid."""
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', 'test-key'):
+            with pytest.raises(HTTPException) as exc_info:
+                await delete_user_api_key(
+                    user_id='user-123',
+                    org_id=valid_org_id,
+                    key_name='automation',
+                    x_service_api_key='wrong-key',
+                )
+
+        assert exc_info.value.status_code == 401
+        assert 'Invalid service API key' in exc_info.value.detail
+
+    @pytest.mark.asyncio
+    async def test_delete_missing_service_key(self, valid_org_id):
+        """Test error when service API key header is missing."""
+        with patch('server.routes.service.AUTOMATIONS_SERVICE_KEY', 'test-key'):
+            with pytest.raises(HTTPException) as exc_info:
+                await delete_user_api_key(
+                    user_id='user-123',
+                    org_id=valid_org_id,
+                    key_name='automation',
+                    x_service_api_key=None,
+                )
+
+        assert exc_info.value.status_code == 401
+        assert 'X-Service-API-Key header is required' in exc_info.value.detail

enterprise/tests/unit/server/routes/test_api_keys.py

@@ -1,19 +1,26 @@
 """Unit tests for API keys routes, focusing on BYOR key validation and retrieval."""
 
+import uuid
 from unittest.mock import AsyncMock, MagicMock, patch
 
 import httpx
 import pytest
 from fastapi import HTTPException
+from pydantic import SecretStr
+from server.auth.saas_user_auth import SaasUserAuth
 from server.routes.api_keys import (
     ByorPermittedResponse,
+    CurrentApiKeyResponse,
     LlmApiKeyResponse,
     check_byor_permitted,
     delete_byor_key_from_litellm,
+    get_current_api_key,
     get_llm_api_key_for_byor,
 )
 from storage.lite_llm_manager import LiteLlmManager
 
+from openhands.server.user_auth.user_auth import AuthType
+
 
 class TestVerifyByorKeyInLitellm:
     """Test the verify_byor_key_in_litellm function."""
@@ -512,3 +519,81 @@ class TestCheckByorPermitted:
 
         assert exc_info.value.status_code == 500
         assert 'Failed to check BYOR export permission' in exc_info.value.detail
+
+
+class TestGetCurrentApiKey:
+    """Test the get_current_api_key endpoint."""
+
+    @pytest.mark.asyncio
+    @patch('server.routes.api_keys.get_user_auth')
+    async def test_returns_api_key_info_for_bearer_auth(self, mock_get_user_auth):
+        """Test that API key metadata including org_id is returned for bearer token auth."""
+        # Arrange
+        user_id = 'user-123'
+        org_id = uuid.uuid4()
+        mock_request = MagicMock()
+
+        user_auth = SaasUserAuth(
+            refresh_token=SecretStr('mock-token'),
+            user_id=user_id,
+            auth_type=AuthType.BEARER,
+            api_key_org_id=org_id,
+            api_key_id=42,
+            api_key_name='My Production Key',
+        )
+        mock_get_user_auth.return_value = user_auth
+
+        # Act
+        result = await get_current_api_key(request=mock_request, user_id=user_id)
+
+        # Assert
+        assert isinstance(result, CurrentApiKeyResponse)
+        assert result.org_id == str(org_id)
+        assert result.id == 42
+        assert result.name == 'My Production Key'
+        assert result.user_id == user_id
+        assert result.auth_type == 'bearer'
+
+    @pytest.mark.asyncio
+    @patch('server.routes.api_keys.get_user_auth')
+    async def test_returns_400_for_cookie_auth(self, mock_get_user_auth):
+        """Test that 400 Bad Request is returned when using cookie authentication."""
+        # Arrange
+        user_id = 'user-123'
+        mock_request = MagicMock()
+
+        mock_user_auth = MagicMock()
+        mock_user_auth.get_auth_type.return_value = AuthType.COOKIE
+        mock_get_user_auth.return_value = mock_user_auth
+
+        # Act & Assert
+        with pytest.raises(HTTPException) as exc_info:
+            await get_current_api_key(request=mock_request, user_id=user_id)
+
+        assert exc_info.value.status_code == 400
+        assert 'API key authentication' in exc_info.value.detail
+
+    @pytest.mark.asyncio
+    @patch('server.routes.api_keys.get_user_auth')
+    async def test_returns_400_when_api_key_org_id_is_none(self, mock_get_user_auth):
+        """Test that 400 is returned when API key has no org_id (legacy key)."""
+        # Arrange
+        user_id = 'user-123'
+        mock_request = MagicMock()
+
+        user_auth = SaasUserAuth(
+            refresh_token=SecretStr('mock-token'),
+            user_id=user_id,
+            auth_type=AuthType.BEARER,
+            api_key_org_id=None,  # No org_id - legacy key
+            api_key_id=42,
+            api_key_name='Legacy Key',
+        )
+        mock_get_user_auth.return_value = user_auth
+
+        # Act & Assert
+        with pytest.raises(HTTPException) as exc_info:
+            await get_current_api_key(request=mock_request, user_id=user_id)
+
+        assert exc_info.value.status_code == 400
+        assert 'created before organization support' in exc_info.value.detail

enterprise/tests/unit/server/routes/test_orgs.py

@@ -106,8 +106,10 @@ async def test_create_org_success(mock_app):
         contact_name='John Doe',
         contact_email='john@example.com',
         org_version=5,
-        default_llm_model='claude-opus-4-5-20251101',
-        enable_default_condenser=True,
+        agent_settings={
+            'schema_version': 1,
+            'llm.model': 'claude-opus-4-5-20251101',
+        },
         enable_proactive_conversation_starters=True,
     )
 
@@ -140,7 +142,9 @@ async def test_create_org_success(mock_app):
         assert response_data['contact_email'] == 'john@example.com'
         assert response_data['credits'] == 100.0
         assert response_data['org_version'] == 5
-        assert response_data['default_llm_model'] == 'claude-opus-4-5-20251101'
+        assert (
+            response_data['agent_settings']['llm.model'] == 'claude-opus-4-5-20251101'
+        )
 
 
 @pytest.mark.asyncio
@@ -427,8 +431,11 @@ async def test_create_org_sensitive_fields_not_exposed(mock_app):
         contact_name='John Doe',
         contact_email='john@example.com',
         org_version=5,
-        default_llm_model='claude-opus-4-5-20251101',
-        enable_default_condenser=True,
+        agent_settings={
+            'schema_version': 1,
+            'llm.model': 'claude-opus-4-5-20251101',
+            'condenser.enabled': True,
+        },
         enable_proactive_conversation_starters=True,
     )
 
@@ -507,7 +514,10 @@ async def test_list_user_orgs_success(mock_app_list):
         contact_name='John Doe',
         contact_email='john@example.com',
         org_version=5,
-        default_llm_model='claude-opus-4-5-20251101',
+        agent_settings={
+            'schema_version': 1,
+            'llm.model': 'claude-opus-4-5-20251101',
+        },
     )
     mock_user = MagicMock()
     mock_user.current_org_id = org_id
@@ -918,20 +928,23 @@ async def test_list_user_orgs_all_fields_present(mock_app_list):
         contact_name='John Doe',
         contact_email='john@example.com',
         conversation_expiration=3600,
-        agent='CodeActAgent',
-        default_max_iterations=50,
-        security_analyzer='enabled',
-        confirmation_mode=True,
-        default_llm_model='claude-opus-4-5-20251101',
-        default_llm_base_url='https://api.example.com',
+        agent_settings={
+            'schema_version': 1,
+            'agent': 'CodeActAgent',
+            'max_iterations': 50,
+            'verification.security_analyzer': 'enabled',
+            'verification.confirmation_mode': True,
+            'llm.model': 'claude-opus-4-5-20251101',
+            'llm.base_url': 'https://api.example.com',
+            'condenser.enabled': True,
+            'mcp_config': {'key': 'value'},
+        },
         remote_runtime_resource_factor=2,
-        enable_default_condenser=True,
         billing_margin=0.15,
         enable_proactive_conversation_starters=True,
         sandbox_base_container_image='test-image',
         sandbox_runtime_container_image='test-runtime',
         org_version=5,
-        mcp_config={'key': 'value'},
         max_budget_per_task=1000.0,
         enable_solvability_analysis=True,
         v1_enabled=True,
@@ -962,20 +975,18 @@ async def test_list_user_orgs_all_fields_present(mock_app_list):
         assert org_data['contact_name'] == 'John Doe'
         assert org_data['contact_email'] == 'john@example.com'
         assert org_data['conversation_expiration'] == 3600
-        assert org_data['agent'] == 'CodeActAgent'
-        assert org_data['default_max_iterations'] == 50
-        assert org_data['security_analyzer'] == 'enabled'
-        assert org_data['confirmation_mode'] is True
-        assert org_data['default_llm_model'] == 'claude-opus-4-5-20251101'
-        assert org_data['default_llm_base_url'] == 'https://api.example.com'
+        assert org_data['agent_settings']['agent'] == 'CodeActAgent'
+        assert org_data['agent_settings']['max_iterations'] == 50
+        assert org_data['agent_settings']['verification.security_analyzer'] == 'enabled'
+        assert org_data['agent_settings']['verification.confirmation_mode'] is True
+        assert org_data['agent_settings']['llm.model'] == 'claude-opus-4-5-20251101'
+        assert org_data['agent_settings']['llm.base_url'] == 'https://api.example.com'
         assert org_data['remote_runtime_resource_factor'] == 2
-        assert org_data['enable_default_condenser'] is True
         assert org_data['billing_margin'] == 0.15
         assert org_data['enable_proactive_conversation_starters'] is True
         assert org_data['sandbox_base_container_image'] == 'test-image'
         assert org_data['sandbox_runtime_container_image'] == 'test-runtime'
         assert org_data['org_version'] == 5
-        assert org_data['mcp_config'] == {'key': 'value'}
         assert org_data['max_budget_per_task'] == 1000.0
         assert org_data['enable_solvability_analysis'] is True
         assert org_data['v1_enabled'] is True
@@ -1020,8 +1031,11 @@ async def test_get_org_success(mock_app_with_get_user_id, mock_owner_role):
         contact_name='John Doe',
         contact_email='john@example.com',
         org_version=5,
-        default_llm_model='claude-opus-4-5-20251101',
-        enable_default_condenser=True,
+        agent_settings={
+            'schema_version': 1,
+            'llm.model': 'claude-opus-4-5-20251101',
+            'condenser.enabled': True,
+        },
         enable_proactive_conversation_starters=True,
     )
 
@@ -1298,8 +1312,11 @@ async def test_get_org_with_credits_none(mock_app_with_get_user_id, mock_owner_r
         contact_name='John Doe',
         contact_email='john@example.com',
         org_version=5,
-        default_llm_model='claude-opus-4-5-20251101',
-        enable_default_condenser=True,
+        agent_settings={
+            'schema_version': 1,
+            'llm.model': 'claude-opus-4-5-20251101',
+            'condenser.enabled': True,
+        },
         enable_proactive_conversation_starters=True,
     )
 
@@ -1345,10 +1362,13 @@ async def test_get_org_sensitive_fields_not_exposed(
         contact_name='John Doe',
         contact_email='john@example.com',
         org_version=5,
-        default_llm_model='claude-opus-4-5-20251101',
+        agent_settings={
+            'schema_version': 1,
+            'llm.model': 'claude-opus-4-5-20251101',
+            'condenser.enabled': True,
+        },
         search_api_key='secret-search-key-123',  # Should not be exposed
         sandbox_api_key='secret-sandbox-key-123',  # Should not be exposed
-        enable_default_condenser=True,
         enable_proactive_conversation_starters=True,
     )
 
@@ -1872,7 +1892,7 @@ async def test_update_org_permission_denied_llm_settings(
     """
     # Arrange
     org_id = uuid.uuid4()
-    update_data = {'default_llm_model': 'claude-opus-4-5-20251101'}
+    update_data = {'agent_settings': {'llm.model': 'claude-opus-4-5-20251101'}}
 
     with (
         patch(
@@ -2032,13 +2052,13 @@ async def test_update_org_invalid_uuid_format(mock_update_app):
 @pytest.mark.asyncio
 async def test_update_org_invalid_field_values(mock_update_app, mock_owner_role):
     """
-    GIVEN: Update request with invalid field values (e.g., negative max_iterations)
+    GIVEN: Update request with invalid field values (e.g., negative billing margin)
     WHEN: PATCH /api/organizations/{org_id} is called
     THEN: 422 validation error is returned
     """
     # Arrange
     org_id = uuid.uuid4()
-    update_data = {'default_max_iterations': -1}  # Invalid: must be > 0
+    update_data = {'billing_margin': -1}  # Invalid: must be >= 0
 
     with patch(
         'server.auth.authorization.get_user_org_role',
@@ -2995,20 +3015,24 @@ class TestGetMeEndpoint:
         llm_model='gpt-4',
         llm_base_url='https://api.example.com',
         max_iterations=50,
-        llm_api_key_for_byor=None,
         status_val='active',
     ):
         """Create a MeResponse for testing."""
+        agent_settings = {'schema_version': 1}
+        if llm_model is not None:
+            agent_settings['llm.model'] = llm_model
+        if llm_base_url is not None:
+            agent_settings['llm.base_url'] = llm_base_url
+        if max_iterations is not None:
+            agent_settings['max_iterations'] = max_iterations
+
         return MeResponse(
             org_id=str(org_id),
             user_id=str(user_id),
             email=email,
             role=role,
             llm_api_key=llm_api_key,
-            llm_model=llm_model,
-            llm_base_url=llm_base_url,
-            max_iterations=max_iterations,
-            llm_api_key_for_byor=llm_api_key_for_byor,
+            agent_settings=agent_settings,
             status=status_val,
         )
 
@@ -3043,9 +3067,9 @@ class TestGetMeEndpoint:
         assert data['user_id'] == test_user_id
         assert data['email'] == 'owner@example.com'
         assert data['role'] == 'owner'
-        assert data['llm_model'] == 'gpt-4'
-        assert data['llm_base_url'] == 'https://api.example.com'
-        assert data['max_iterations'] == 50
+        assert data['agent_settings']['llm.model'] == 'gpt-4'
+        assert data['agent_settings']['llm.base_url'] == 'https://api.example.com'
+        assert data['agent_settings']['max_iterations'] == 50
         assert data['status'] == 'active'
 
     @pytest.mark.asyncio
@@ -3169,9 +3193,7 @@ class TestGetMeEndpoint:
 
         assert response.status_code == status.HTTP_200_OK
         data = response.json()
-        assert data['llm_model'] is None
-        assert data['llm_base_url'] is None
-        assert data['max_iterations'] is None
+        assert data['agent_settings'] == {'schema_version': 1}
 
     @pytest.mark.asyncio
     async def test_get_me_with_admin_role(self, mock_me_app, test_user_id, test_org_id):
@@ -3200,35 +3222,6 @@ class TestGetMeEndpoint:
         data = response.json()
         assert data['role'] == 'admin'
 
-    @pytest.mark.asyncio
-    async def test_get_me_masks_byor_api_key(
-        self, mock_me_app, test_user_id, test_org_id
-    ):
-        """GIVEN: User has an llm_api_key_for_byor set
-        WHEN: GET /api/organizations/{org_id}/me is called
-        THEN: The llm_api_key_for_byor field is also masked
-        """
-        me_response = self._make_me_response(
-            org_id=test_org_id,
-            user_id=test_user_id,
-            llm_api_key_for_byor='****-key',  # Masked key
-        )
-
-        with patch(
-            'server.routes.orgs.OrgMemberService.get_me',
-            new_callable=AsyncMock,
-            return_value=me_response,
-        ):
-            client = TestClient(mock_me_app)
-            response = client.get(f'/api/organizations/{test_org_id}/me')
-
-        assert response.status_code == status.HTTP_200_OK
-        data = response.json()
-        assert data['llm_api_key_for_byor'] != 'sk-byor-secret-key'
-        assert (
-            data['llm_api_key_for_byor'] is None or '**' in data['llm_api_key_for_byor']
-        )
-
     @pytest.mark.asyncio
     async def test_get_me_role_not_found_returns_500(self, mock_me_app, test_org_id):
         """GIVEN: Role lookup fails (data integrity issue)
@@ -3315,7 +3308,10 @@ async def test_switch_org_success(mock_app_with_get_user_id):
         contact_name='John Doe',
         contact_email='john@example.com',
         org_version=5,
-        default_llm_model='claude-opus-4-5-20251101',
+        agent_settings={
+            'schema_version': 1,
+            'llm.model': 'claude-opus-4-5-20251101',
+        },
     )
 
     with (

enterprise/tests/unit/server/services/test_org_llm_settings_service.py

@@ -34,15 +34,15 @@ def mock_org(org_id):
     """Create a mock organization with LLM settings."""
     org = MagicMock(spec=Org)
     org.id = org_id
-    org.default_llm_model = 'claude-3'
-    org.default_llm_base_url = 'https://api.anthropic.com'
+    org.agent_settings = {
+        'schema_version': 1,
+        'llm.model': 'claude-3',
+        'llm.base_url': 'https://api.anthropic.com',
+        'agent': 'CodeActAgent',
+        'verification.confirmation_mode': True,
+        'max_iterations': 50,
+    }
     org.search_api_key = None
-    org.agent = 'CodeActAgent'
-    org.confirmation_mode = True
-    org.security_analyzer = None
-    org.enable_default_condenser = True
-    org.condenser_max_size = None
-    org.default_max_iterations = 50
     return org
 
 
@@ -78,8 +78,8 @@ async def test_get_org_llm_settings_success(
 
     # Assert
     assert isinstance(result, OrgLLMSettingsResponse)
-    assert result.default_llm_model == 'claude-3'
-    assert result.agent == 'CodeActAgent'
+    assert result.agent_settings['llm.model'] == 'claude-3'
+    assert result.agent_settings['agent'] == 'CodeActAgent'
     mock_store.get_current_org_by_user_id.assert_called_once_with(user_id)
 
 
@@ -134,20 +134,21 @@ async def test_update_org_llm_settings_success(
     # Arrange
     updated_org = MagicMock(spec=Org)
     updated_org.id = mock_org.id
-    updated_org.default_llm_model = 'new-model'
-    updated_org.default_llm_base_url = None
+    updated_org.agent_settings = {
+        'schema_version': 1,
+        'llm.model': 'new-model',
+        'agent': 'CodeActAgent',
+        'verification.confirmation_mode': False,
+        'max_iterations': 100,
+    }
     updated_org.search_api_key = None
-    updated_org.agent = 'CodeActAgent'
-    updated_org.confirmation_mode = False
-    updated_org.security_analyzer = None
-    updated_org.enable_default_condenser = True
-    updated_org.condenser_max_size = None
-    updated_org.default_max_iterations = 100
 
     update_data = OrgLLMSettingsUpdate(
-        default_llm_model='new-model',
-        confirmation_mode=False,
-        default_max_iterations=100,
+        agent_settings={
+            'llm.model': 'new-model',
+            'verification.confirmation_mode': False,
+            'max_iterations': 100,
+        }
     )
 
     mock_store.get_current_org_by_user_id = AsyncMock(return_value=mock_org)
@@ -159,9 +160,9 @@ async def test_update_org_llm_settings_success(
 
     # Assert
     assert isinstance(result, OrgLLMSettingsResponse)
-    assert result.default_llm_model == 'new-model'
-    assert result.confirmation_mode is False
-    assert result.default_max_iterations == 100
+    assert result.agent_settings['llm.model'] == 'new-model'
+    assert result.agent_settings['verification.confirmation_mode'] is False
+    assert result.agent_settings['max_iterations'] == 100
     mock_store.update_org_llm_settings.assert_called_once_with(
         org_id=mock_org.id,
         update_data=update_data,
@@ -189,7 +190,7 @@ async def test_update_org_llm_settings_no_changes(
 
     # Assert
     assert isinstance(result, OrgLLMSettingsResponse)
-    assert result.default_llm_model == 'claude-3'
+    assert result.agent_settings['llm.model'] == 'claude-3'
     mock_store.update_org_llm_settings.assert_not_called()
 
 
@@ -203,7 +204,7 @@ async def test_update_org_llm_settings_org_not_found(
     THEN: OrgNotFoundError is raised
     """
     # Arrange
-    update_data = OrgLLMSettingsUpdate(default_llm_model='new-model')
+    update_data = OrgLLMSettingsUpdate(agent_settings={'llm.model': 'new-model'})
 
     mock_store.get_current_org_by_user_id = AsyncMock(return_value=None)
     service = OrgLLMSettingsService(store=mock_store, user_context=mock_user_context)

enterprise/tests/unit/server/services/test_org_member_service.py

@@ -2224,10 +2224,12 @@ class TestOrgMemberServiceGetMe:
         member.user_id = current_user_id
         member.role_id = 1
         member.llm_api_key = SecretStr('sk-test-key-12345')
-        member.llm_api_key_for_byor = None
-        member.llm_model = 'gpt-4'
-        member.llm_base_url = 'https://api.example.com'
-        member.max_iterations = 50
+        member.agent_settings = {
+            'schema_version': 1,
+            'llm.model': 'gpt-4',
+            'llm.base_url': 'https://api.example.com',
+            'max_iterations': 50,
+        }
         member.status = 'active'
         return member
 
@@ -2275,8 +2277,8 @@ class TestOrgMemberServiceGetMe:
             assert result.user_id == str(current_user_id)
             assert result.email == 'test@example.com'
             assert result.role == 'owner'
-            assert result.llm_model == 'gpt-4'
-            assert result.max_iterations == 50
+            assert result.agent_settings['llm.model'] == 'gpt-4'
+            assert result.agent_settings['max_iterations'] == 50
             assert result.status == 'active'
 
     @pytest.mark.asyncio

enterprise/tests/unit/services/conftest.py

@@ -1,68 +0,0 @@
-"""Shared fixtures for services tests.
-
-Note: We pre-load ``storage`` as a namespace package to avoid the heavy
-``storage/__init__.py`` that imports the entire enterprise model graph.
-This must happen *before* any ``from storage.…`` import.
-"""
-
-import contextlib
-import sys
-import types
-
-# Prevent storage/__init__.py from loading the full model graph.
-# We only need the lightweight automation models for these tests.
-if 'storage' not in sys.modules:
-    import pathlib
-
-    _storage_dir = str(pathlib.Path(__file__).resolve().parents[3] / 'storage')
-    _mod = types.ModuleType('storage')
-    _mod.__path__ = [_storage_dir]
-    sys.modules['storage'] = _mod
-
-import pytest
-from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker, create_async_engine
-from sqlalchemy.pool import StaticPool
-from storage.automation import Automation, AutomationRun  # noqa: F401
-from storage.automation_event import AutomationEvent  # noqa: F401
-from storage.base import Base
-
-
-@pytest.fixture
-async def async_engine():
-    """Create an async SQLite engine for testing."""
-    engine = create_async_engine(
-        'sqlite+aiosqlite:///:memory:',
-        poolclass=StaticPool,
-        connect_args={'check_same_thread': False},
-        echo=False,
-    )
-
-    async with engine.begin() as conn:
-        await conn.run_sync(Base.metadata.create_all)
-
-    yield engine
-    await engine.dispose()
-
-
-@pytest.fixture
-async def async_session_factory(async_engine):
-    """Create an async session factory that yields context-managed sessions."""
-    factory = async_sessionmaker(
-        bind=async_engine,
-        class_=AsyncSession,
-        expire_on_commit=False,
-    )
-
-    @contextlib.asynccontextmanager
-    async def _session_ctx():
-        async with factory() as session:
-            yield session
-
-    return _session_ctx
-
-
-@pytest.fixture
-async def async_session(async_session_factory):
-    """Create a single async session for testing."""
-    async with async_session_factory() as session:
-        yield session

enterprise/tests/unit/services/test_automation_executor.py

@@ -1,624 +0,0 @@
-"""Tests for the automation executor.
-
-Uses real SQLite database operations for event processing, run claiming,
-and stale run recovery. HTTP calls to the V1 API are mocked.
-"""
-
-from datetime import datetime, timedelta
-from unittest.mock import AsyncMock, patch
-from uuid import uuid4
-
-import pytest
-from services.automation_executor import (
-    _mark_run_failed,
-    claim_and_execute_runs,
-    find_matching_automations,
-    is_terminal,
-    process_new_events,
-    recover_stale_runs,
-    utc_now,
-)
-from sqlalchemy import select
-from storage.automation import Automation, AutomationRun
-from storage.automation_event import AutomationEvent
-
-# ---------------------------------------------------------------------------
-# Helpers
-# ---------------------------------------------------------------------------
-
-
-def make_automation(
-    automation_id: str = 'auto-1',
-    user_id: str = 'user-1',
-    enabled: bool = True,
-    trigger_type: str = 'cron',
-    name: str = 'Test Automation',
-) -> Automation:
-    return Automation(
-        id=automation_id,
-        user_id=user_id,
-        org_id='org-1',
-        name=name,
-        enabled=enabled,
-        config={'triggers': {'cron': {'schedule': '0 9 * * 5'}}},
-        trigger_type=trigger_type,
-        file_store_key=f'automations/{automation_id}/script.py',
-    )
-
-
-def make_event(
-    source_type: str = 'cron',
-    payload: dict | None = None,
-    status: str = 'NEW',
-    dedup_key: str | None = None,
-) -> AutomationEvent:
-    return AutomationEvent(
-        source_type=source_type,
-        payload=payload or {'automation_id': 'auto-1'},
-        dedup_key=dedup_key or f'dedup-{uuid4().hex[:8]}',
-        status=status,
-        created_at=utc_now(),
-    )
-
-
-def make_run(
-    run_id: str | None = None,
-    automation_id: str = 'auto-1',
-    status: str = 'PENDING',
-    claimed_by: str | None = None,
-    heartbeat_at: datetime | None = None,
-    retry_count: int = 0,
-    max_retries: int = 3,
-    next_retry_at: datetime | None = None,
-) -> AutomationRun:
-    return AutomationRun(
-        id=run_id or uuid4().hex,
-        automation_id=automation_id,
-        status=status,
-        claimed_by=claimed_by,
-        heartbeat_at=heartbeat_at,
-        retry_count=retry_count,
-        max_retries=max_retries,
-        next_retry_at=next_retry_at,
-        event_payload={'automation_id': automation_id},
-        created_at=utc_now(),
-    )
-
-
-# ---------------------------------------------------------------------------
-# find_matching_automations
-# ---------------------------------------------------------------------------
-
-
-@pytest.mark.asyncio
-async def test_find_matching_automations_cron_event(async_session):
-    """Cron events match by automation_id in payload."""
-    automation = make_automation()
-    async_session.add(automation)
-    await async_session.commit()
-
-    event = make_event(source_type='cron', payload={'automation_id': 'auto-1'})
-    async_session.add(event)
-    await async_session.commit()
-
-    result = await find_matching_automations(async_session, event)
-
-    assert len(result) == 1
-    assert result[0].id == 'auto-1'
-
-
-@pytest.mark.asyncio
-async def test_find_matching_automations_manual_event(async_session):
-    """Manual events also match by automation_id in payload."""
-    automation = make_automation()
-    async_session.add(automation)
-    await async_session.commit()
-
-    event = make_event(source_type='manual', payload={'automation_id': 'auto-1'})
-    async_session.add(event)
-    await async_session.commit()
-
-    result = await find_matching_automations(async_session, event)
-
-    assert len(result) == 1
-    assert result[0].id == 'auto-1'
-
-
-@pytest.mark.asyncio
-async def test_find_matching_automations_disabled_automation(async_session):
-    """Disabled automations are not matched."""
-    automation = make_automation(enabled=False)
-    async_session.add(automation)
-    await async_session.commit()
-
-    event = make_event(payload={'automation_id': 'auto-1'})
-    async_session.add(event)
-    await async_session.commit()
-
-    result = await find_matching_automations(async_session, event)
-
-    assert len(result) == 0
-
-
-@pytest.mark.asyncio
-async def test_find_matching_automations_missing_automation_id(async_session):
-    """Events without automation_id in payload return empty list."""
-    event = make_event(payload={'something_else': 'value'})
-    async_session.add(event)
-    await async_session.commit()
-
-    result = await find_matching_automations(async_session, event)
-
-    assert len(result) == 0
-
-
-@pytest.mark.asyncio
-async def test_find_matching_automations_nonexistent_automation(async_session):
-    """Events referencing a non-existent automation return empty list."""
-    event = make_event(payload={'automation_id': 'nonexistent'})
-    async_session.add(event)
-    await async_session.commit()
-
-    result = await find_matching_automations(async_session, event)
-
-    assert len(result) == 0
-
-
-@pytest.mark.asyncio
-async def test_find_matching_automations_unknown_source_type(async_session):
-    """Unknown source types return empty list."""
-    event = make_event(source_type='unknown', payload={'automation_id': 'auto-1'})
-    async_session.add(event)
-    await async_session.commit()
-
-    result = await find_matching_automations(async_session, event)
-
-    assert len(result) == 0
-
-
-# ---------------------------------------------------------------------------
-# process_new_events
-# ---------------------------------------------------------------------------
-
-
-@pytest.mark.asyncio
-async def test_process_new_events_creates_runs(async_session):
-    """Processing NEW events creates PENDING runs and marks events PROCESSED."""
-    automation = make_automation()
-    event = make_event(payload={'automation_id': 'auto-1'})
-    async_session.add_all([automation, event])
-    await async_session.commit()
-
-    count = await process_new_events(async_session)
-
-    assert count == 1
-
-    # Event should be PROCESSED
-    await async_session.refresh(event)
-    assert event.status == 'PROCESSED'
-    assert event.processed_at is not None
-
-    # A run should have been created
-    runs = (await async_session.execute(select(AutomationRun))).scalars().all()
-    assert len(runs) == 1
-    assert runs[0].automation_id == 'auto-1'
-    assert runs[0].status == 'PENDING'
-    assert runs[0].event_payload == {'automation_id': 'auto-1'}
-
-
-@pytest.mark.asyncio
-async def test_process_new_events_no_match(async_session):
-    """Events with no matching automation are marked NO_MATCH."""
-    event = make_event(payload={'automation_id': 'nonexistent'})
-    async_session.add(event)
-    await async_session.commit()
-
-    count = await process_new_events(async_session)
-
-    assert count == 1
-
-    await async_session.refresh(event)
-    assert event.status == 'NO_MATCH'
-    assert event.processed_at is not None
-
-    # No runs created
-    runs = (await async_session.execute(select(AutomationRun))).scalars().all()
-    assert len(runs) == 0
-
-
-@pytest.mark.asyncio
-async def test_process_new_events_skips_processed(async_session):
-    """Already processed events are not re-processed."""
-    event = make_event(status='PROCESSED')
-    async_session.add(event)
-    await async_session.commit()
-
-    count = await process_new_events(async_session)
-
-    assert count == 0
-
-
-@pytest.mark.asyncio
-async def test_process_new_events_multiple_events(async_session):
-    """Multiple NEW events are processed in one batch."""
-    auto1 = make_automation(automation_id='auto-1')
-    auto2 = make_automation(automation_id='auto-2', name='Auto 2')
-    event1 = make_event(payload={'automation_id': 'auto-1'}, dedup_key='dedup-1')
-    event2 = make_event(payload={'automation_id': 'auto-2'}, dedup_key='dedup-2')
-    event3 = make_event(payload={'automation_id': 'nonexistent'}, dedup_key='dedup-3')
-    async_session.add_all([auto1, auto2, event1, event2, event3])
-    await async_session.commit()
-
-    count = await process_new_events(async_session)
-
-    assert count == 3
-
-    # Two runs created (for auto-1 and auto-2), none for nonexistent
-    runs = (await async_session.execute(select(AutomationRun))).scalars().all()
-    assert len(runs) == 2
-
-    await async_session.refresh(event1)
-    await async_session.refresh(event2)
-    await async_session.refresh(event3)
-    assert event1.status == 'PROCESSED'
-    assert event2.status == 'PROCESSED'
-    assert event3.status == 'NO_MATCH'
-
-
-# ---------------------------------------------------------------------------
-# claim_and_execute_runs
-# ---------------------------------------------------------------------------
-
-
-@pytest.mark.asyncio
-async def test_claim_and_execute_runs_claims_pending(
-    async_session, async_session_factory
-):
-    """Claims a PENDING run and transitions to RUNNING."""
-    automation = make_automation()
-    run = make_run(run_id='run-1')
-    async_session.add_all([automation, run])
-    await async_session.commit()
-
-    api_client = AsyncMock()
-
-    with patch('services.automation_executor.execute_run', new_callable=AsyncMock):
-        claimed = await claim_and_execute_runs(
-            async_session, 'executor-test-1', api_client, async_session_factory
-        )
-
-    assert claimed is True
-
-    await async_session.refresh(run)
-    assert run.status == 'RUNNING'
-    assert run.claimed_by == 'executor-test-1'
-    assert run.claimed_at is not None
-    assert run.heartbeat_at is not None
-    assert run.started_at is not None
-
-
-@pytest.mark.asyncio
-async def test_claim_and_execute_runs_no_pending(async_session, async_session_factory):
-    """Returns False when no PENDING runs exist."""
-    api_client = AsyncMock()
-
-    claimed = await claim_and_execute_runs(
-        async_session, 'executor-test-1', api_client, async_session_factory
-    )
-
-    assert claimed is False
-
-
-@pytest.mark.asyncio
-async def test_claim_and_execute_runs_respects_next_retry_at(
-    async_session, async_session_factory
-):
-    """Runs with future next_retry_at are not claimed."""
-    automation = make_automation()
-    run = make_run(
-        run_id='run-retry',
-        next_retry_at=utc_now() + timedelta(hours=1),
-    )
-    async_session.add_all([automation, run])
-    await async_session.commit()
-
-    api_client = AsyncMock()
-
-    claimed = await claim_and_execute_runs(
-        async_session, 'executor-test-1', api_client, async_session_factory
-    )
-
-    assert claimed is False
-
-
-@pytest.mark.asyncio
-async def test_claim_and_execute_runs_past_retry_at(
-    async_session, async_session_factory
-):
-    """Runs with past next_retry_at are claimable."""
-    automation = make_automation()
-    run = make_run(
-        run_id='run-retry-past',
-        next_retry_at=utc_now() - timedelta(minutes=5),
-    )
-    async_session.add_all([automation, run])
-    await async_session.commit()
-
-    api_client = AsyncMock()
-
-    with patch('services.automation_executor.execute_run', new_callable=AsyncMock):
-        claimed = await claim_and_execute_runs(
-            async_session, 'executor-test-1', api_client, async_session_factory
-        )
-
-    assert claimed is True
-
-
-@pytest.mark.asyncio
-async def test_claim_skips_running_runs(async_session, async_session_factory):
-    """RUNNING runs are not claimed."""
-    automation = make_automation()
-    run = make_run(run_id='run-running', status='RUNNING', claimed_by='other-executor')
-    async_session.add_all([automation, run])
-    await async_session.commit()
-
-    api_client = AsyncMock()
-
-    claimed = await claim_and_execute_runs(
-        async_session, 'executor-test-1', api_client, async_session_factory
-    )
-
-    assert claimed is False
-
-
-# ---------------------------------------------------------------------------
-# recover_stale_runs
-# ---------------------------------------------------------------------------
-
-
-@pytest.mark.asyncio
-async def test_recover_stale_runs_recovers_stale(async_session):
-    """RUNNING runs with expired heartbeats are recovered to PENDING."""
-    automation = make_automation()
-    stale_run = make_run(
-        run_id='stale-1',
-        status='RUNNING',
-        claimed_by='crashed-executor',
-        heartbeat_at=utc_now() - timedelta(minutes=10),
-        retry_count=0,
-    )
-    async_session.add_all([automation, stale_run])
-    await async_session.commit()
-
-    count = await recover_stale_runs(async_session)
-
-    assert count >= 1
-
-    await async_session.refresh(stale_run)
-    assert stale_run.status == 'PENDING'
-    assert stale_run.claimed_by is None
-    assert stale_run.retry_count == 1
-    assert stale_run.next_retry_at is not None
-
-
-@pytest.mark.asyncio
-async def test_recover_stale_runs_ignores_fresh(async_session):
-    """RUNNING runs with recent heartbeats are not recovered."""
-    automation = make_automation()
-    fresh_run = make_run(
-        run_id='fresh-1',
-        status='RUNNING',
-        claimed_by='active-executor',
-        heartbeat_at=utc_now() - timedelta(seconds=30),
-    )
-    async_session.add_all([automation, fresh_run])
-    await async_session.commit()
-
-    count = await recover_stale_runs(async_session)
-
-    assert count == 0
-
-    await async_session.refresh(fresh_run)
-    assert fresh_run.status == 'RUNNING'
-    assert fresh_run.claimed_by == 'active-executor'
-
-
-@pytest.mark.asyncio
-async def test_recover_stale_runs_ignores_pending(async_session):
-    """PENDING runs are not affected by recovery."""
-    automation = make_automation()
-    pending_run = make_run(run_id='pending-1', status='PENDING')
-    async_session.add_all([automation, pending_run])
-    await async_session.commit()
-
-    count = await recover_stale_runs(async_session)
-
-    assert count == 0
-
-    await async_session.refresh(pending_run)
-    assert pending_run.status == 'PENDING'
-
-
-@pytest.mark.asyncio
-async def test_recover_stale_runs_increments_retry_count(async_session):
-    """Recovery increments the retry_count."""
-    automation = make_automation()
-    stale_run = make_run(
-        run_id='stale-retry',
-        status='RUNNING',
-        claimed_by='old-executor',
-        heartbeat_at=utc_now() - timedelta(minutes=10),
-        retry_count=2,
-    )
-    async_session.add_all([automation, stale_run])
-    await async_session.commit()
-
-    await recover_stale_runs(async_session)
-
-    await async_session.refresh(stale_run)
-    assert stale_run.retry_count == 3
-
-
-# ---------------------------------------------------------------------------
-# _mark_run_failed (error handling)
-# ---------------------------------------------------------------------------
-
-
-@pytest.mark.asyncio
-async def test_mark_run_failed_retries(async_session_factory):
-    """Failed runs with retries left return to PENDING."""
-    async with async_session_factory() as session:
-        automation = make_automation()
-        run = make_run(run_id='fail-retry', retry_count=0, max_retries=3)
-        session.add_all([automation, run])
-        await session.commit()
-
-    async with async_session_factory() as session:
-        run_obj = await session.get(AutomationRun, 'fail-retry')
-        await _mark_run_failed(run_obj, 'API error', async_session_factory)
-
-    async with async_session_factory() as session:
-        run_obj = await session.get(AutomationRun, 'fail-retry')
-        assert run_obj.status == 'PENDING'
-        assert run_obj.retry_count == 1
-        assert run_obj.error_detail == 'API error'
-        assert run_obj.next_retry_at is not None
-        assert run_obj.claimed_by is None
-
-
-@pytest.mark.asyncio
-async def test_mark_run_failed_dead_letter(async_session_factory):
-    """Failed runs that exceed max_retries go to DEAD_LETTER."""
-    async with async_session_factory() as session:
-        automation = make_automation()
-        run = make_run(run_id='fail-dead', retry_count=2, max_retries=3)
-        session.add_all([automation, run])
-        await session.commit()
-
-    async with async_session_factory() as session:
-        run_obj = await session.get(AutomationRun, 'fail-dead')
-        await _mark_run_failed(run_obj, 'Final failure', async_session_factory)
-
-    async with async_session_factory() as session:
-        run_obj = await session.get(AutomationRun, 'fail-dead')
-        assert run_obj.status == 'DEAD_LETTER'
-        assert run_obj.retry_count == 3
-        assert run_obj.error_detail == 'Final failure'
-        assert run_obj.completed_at is not None
-
-
-# ---------------------------------------------------------------------------
-# is_terminal
-# ---------------------------------------------------------------------------
-
-
-def test_is_terminal_stopped():
-    assert is_terminal({'status': 'STOPPED'}) is True
-
-
-def test_is_terminal_error():
-    assert is_terminal({'status': 'ERROR'}) is True
-
-
-def test_is_terminal_completed():
-    assert is_terminal({'status': 'COMPLETED'}) is True
-
-
-def test_is_terminal_cancelled():
-    assert is_terminal({'status': 'CANCELLED'}) is True
-
-
-def test_is_terminal_running():
-    assert is_terminal({'status': 'RUNNING'}) is False
-
-
-def test_is_terminal_empty():
-    assert is_terminal({}) is False
-
-
-def test_is_terminal_case_insensitive():
-    assert is_terminal({'status': 'stopped'}) is True
-    assert is_terminal({'status': 'Completed'}) is True
-
-
-# ---------------------------------------------------------------------------
-# find_matching_automations — None payload
-# ---------------------------------------------------------------------------
-
-
-@pytest.mark.asyncio
-async def test_find_matching_automations_none_payload(async_session):
-    """Events with None payload return empty list (data corruption guard)."""
-    event = make_event(source_type='cron')
-    event.payload = None
-    async_session.add(event)
-    await async_session.commit()
-
-    result = await find_matching_automations(async_session, event)
-
-    assert result == []
-
-
-# ---------------------------------------------------------------------------
-# Integration: event → run creation → claim
-# ---------------------------------------------------------------------------
-
-
-@pytest.mark.asyncio
-async def test_integration_event_to_run_to_claim(
-    async_session_factory,
-):
-    """Full flow: create event + automation → process_new_events → claim_and_execute_runs.
-
-    Uses a real SQLite database; only the external API client is mocked.
-    """
-    # 1. Seed an automation and a NEW event
-    async with async_session_factory() as session:
-        automation = make_automation(automation_id='integ-auto')
-        event = make_event(
-            source_type='cron',
-            payload={'automation_id': 'integ-auto'},
-            dedup_key='integ-dedup',
-        )
-        session.add_all([automation, event])
-        await session.commit()
-        event_id = event.id
-
-    # 2. Process inbox — should match and create a PENDING run
-    async with async_session_factory() as session:
-        processed = await process_new_events(session)
-
-    assert processed == 1
-
-    # Verify event is PROCESSED and run was created
-    async with async_session_factory() as session:
-        evt = await session.get(AutomationEvent, event_id)
-        assert evt.status == 'PROCESSED'
-
-        runs = (await session.execute(select(AutomationRun))).scalars().all()
-        assert len(runs) == 1
-        run = runs[0]
-        assert run.automation_id == 'integ-auto'
-        assert run.status == 'PENDING'
-        assert run.event_payload == {'automation_id': 'integ-auto'}
-
-    # 3. Claim the run — mock execute_run to avoid real API calls
-    api_client = AsyncMock()
-
-    with patch('services.automation_executor.execute_run', new_callable=AsyncMock):
-        async with async_session_factory() as session:
-            claimed = await claim_and_execute_runs(
-                session, 'executor-integ', api_client, async_session_factory
-            )
-
-    assert claimed is True
-
-    # 4. Verify the run moved to RUNNING with correct executor
-    async with async_session_factory() as session:
-        runs = (await session.execute(select(AutomationRun))).scalars().all()
-        assert len(runs) == 1
-        run = runs[0]
-        assert run.status == 'RUNNING'
-        assert run.claimed_by == 'executor-integ'
-        assert run.started_at is not None
-        assert run.heartbeat_at is not None

enterprise/tests/unit/services/test_openhands_api_client.py

@@ -1,185 +0,0 @@
-"""Tests for OpenHandsAPIClient with mocked HTTP responses."""
-
-import base64
-
-import httpx
-import pytest
-from services.openhands_api_client import OpenHandsAPIClient
-
-
-@pytest.fixture
-def api_client():
-    client = OpenHandsAPIClient(base_url='http://test-server:3000')
-    yield client
-    # close handled in tests that need it
-
-
-# ---------------------------------------------------------------------------
-# start_conversation
-# ---------------------------------------------------------------------------
-
-
-@pytest.mark.asyncio
-async def test_start_conversation_sends_correct_request(api_client, respx_mock):
-    """start_conversation sends properly formatted POST with auth header."""
-    automation_file = b'print("hello")'
-    expected_b64 = base64.b64encode(automation_file).decode()
-
-    route = respx_mock.post('http://test-server:3000/api/v1/app-conversations').mock(
-        return_value=httpx.Response(
-            200,
-            json={
-                'app_conversation_id': 'conv-123',
-                'status': 'RUNNING',
-            },
-        )
-    )
-
-    result = await api_client.start_conversation(
-        api_key='sk-oh-test123',
-        automation_file=automation_file,
-        title='Test Automation',
-        event_payload={'automation_id': 'auto-1'},
-    )
-
-    assert route.called
-    request = route.calls[0].request
-
-    assert request.headers['Authorization'] == 'Bearer sk-oh-test123'
-
-    import json
-
-    body = json.loads(request.content)
-    assert body['automation_file'] == expected_b64
-    assert body['trigger'] == 'automation'
-    assert body['title'] == 'Test Automation'
-    assert body['event_payload'] == {'automation_id': 'auto-1'}
-
-    assert result == {'app_conversation_id': 'conv-123', 'status': 'RUNNING'}
-
-
-@pytest.mark.asyncio
-async def test_start_conversation_without_event_payload(api_client, respx_mock):
-    """start_conversation works with event_payload=None."""
-    respx_mock.post('http://test-server:3000/api/v1/app-conversations').mock(
-        return_value=httpx.Response(200, json={'app_conversation_id': 'conv-456'})
-    )
-
-    result = await api_client.start_conversation(
-        api_key='sk-oh-test',
-        automation_file=b'code',
-        title='Test',
-        event_payload=None,
-    )
-
-    assert result['app_conversation_id'] == 'conv-456'
-
-
-@pytest.mark.asyncio
-async def test_start_conversation_http_error(api_client, respx_mock):
-    """start_conversation raises on HTTP errors."""
-    respx_mock.post('http://test-server:3000/api/v1/app-conversations').mock(
-        return_value=httpx.Response(500, json={'error': 'Internal Server Error'})
-    )
-
-    with pytest.raises(httpx.HTTPStatusError) as exc_info:
-        await api_client.start_conversation(
-            api_key='sk-oh-test',
-            automation_file=b'code',
-            title='Test',
-        )
-
-    assert exc_info.value.response.status_code == 500
-
-
-@pytest.mark.asyncio
-async def test_start_conversation_auth_error(api_client, respx_mock):
-    """start_conversation raises on 401 Unauthorized."""
-    respx_mock.post('http://test-server:3000/api/v1/app-conversations').mock(
-        return_value=httpx.Response(401, json={'error': 'Unauthorized'})
-    )
-
-    with pytest.raises(httpx.HTTPStatusError) as exc_info:
-        await api_client.start_conversation(
-            api_key='bad-key',
-            automation_file=b'code',
-            title='Test',
-        )
-
-    assert exc_info.value.response.status_code == 401
-
-
-# ---------------------------------------------------------------------------
-# get_conversation
-# ---------------------------------------------------------------------------
-
-
-@pytest.mark.asyncio
-async def test_get_conversation_returns_data(api_client, respx_mock):
-    """get_conversation returns the first conversation from the list."""
-    respx_mock.get('http://test-server:3000/api/v1/app-conversations').mock(
-        return_value=httpx.Response(
-            200,
-            json=[
-                {
-                    'conversation_id': 'conv-123',
-                    'status': 'RUNNING',
-                    'title': 'My Automation',
-                }
-            ],
-        )
-    )
-
-    result = await api_client.get_conversation('sk-oh-test', 'conv-123')
-
-    assert result is not None
-    assert result['conversation_id'] == 'conv-123'
-    assert result['status'] == 'RUNNING'
-
-
-@pytest.mark.asyncio
-async def test_get_conversation_returns_none_when_empty(api_client, respx_mock):
-    """get_conversation returns None when API returns empty list."""
-    respx_mock.get('http://test-server:3000/api/v1/app-conversations').mock(
-        return_value=httpx.Response(200, json=[])
-    )
-
-    result = await api_client.get_conversation('sk-oh-test', 'nonexistent')
-
-    assert result is None
-
-
-@pytest.mark.asyncio
-async def test_get_conversation_sends_auth_header(api_client, respx_mock):
-    """get_conversation sends the correct authorization header."""
-    route = respx_mock.get('http://test-server:3000/api/v1/app-conversations').mock(
-        return_value=httpx.Response(200, json=[])
-    )
-
-    await api_client.get_conversation('sk-oh-mykey', 'conv-1')
-
-    assert route.called
-    request = route.calls[0].request
-    assert request.headers['Authorization'] == 'Bearer sk-oh-mykey'
-
-
-@pytest.mark.asyncio
-async def test_get_conversation_http_error(api_client, respx_mock):
-    """get_conversation raises on HTTP errors."""
-    respx_mock.get('http://test-server:3000/api/v1/app-conversations').mock(
-        return_value=httpx.Response(503, text='Service Unavailable')
-    )
-
-    with pytest.raises(httpx.HTTPStatusError):
-        await api_client.get_conversation('sk-oh-test', 'conv-1')
-
-
-# ---------------------------------------------------------------------------
-# close
-# ---------------------------------------------------------------------------
-
-
-@pytest.mark.asyncio
-async def test_close(api_client):
-    """close() shuts down the HTTP client without errors."""
-    await api_client.close()

enterprise/tests/unit/storage/test_api_key_store.py

@@ -0,0 +1,314 @@
+"""Unit tests for ApiKeyStore system key functionality."""
+
+import uuid
+from datetime import UTC, datetime, timedelta
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+from sqlalchemy import select
+from storage.api_key import ApiKey
+from storage.api_key_store import ApiKeyStore
+
+
+@pytest.fixture
+def api_key_store():
+    """Create ApiKeyStore instance."""
+    return ApiKeyStore()
+
+
+class TestApiKeyStoreSystemKeys:
+    """Test cases for system API key functionality."""
+
+    def test_is_system_key_name_with_prefix(self, api_key_store):
+        """Test that names with __SYSTEM__: prefix are identified as system keys."""
+        assert api_key_store.is_system_key_name('__SYSTEM__:automation') is True
+        assert api_key_store.is_system_key_name('__SYSTEM__:test-key') is True
+        assert api_key_store.is_system_key_name('__SYSTEM__:') is True
+
+    def test_is_system_key_name_without_prefix(self, api_key_store):
+        """Test that names without __SYSTEM__: prefix are not system keys."""
+        assert api_key_store.is_system_key_name('my-key') is False
+        assert api_key_store.is_system_key_name('automation') is False
+        assert api_key_store.is_system_key_name('MCP_API_KEY') is False
+        assert api_key_store.is_system_key_name('') is False
+
+    def test_is_system_key_name_none(self, api_key_store):
+        """Test that None is not a system key."""
+        assert api_key_store.is_system_key_name(None) is False
+
+    def test_make_system_key_name(self, api_key_store):
+        """Test system key name generation."""
+        assert (
+            api_key_store.make_system_key_name('automation') == '__SYSTEM__:automation'
+        )
+        assert api_key_store.make_system_key_name('test-key') == '__SYSTEM__:test-key'
+
+    @pytest.mark.asyncio
+    async def test_get_or_create_system_api_key_creates_new(
+        self, api_key_store, async_session_maker
+    ):
+        """Test creating a new system API key when none exists."""
+        user_id = '5594c7b6-f959-4b81-92e9-b09c206f5081'
+        org_id = uuid.UUID('5594c7b6-f959-4b81-92e9-b09c206f5081')
+        key_name = 'automation'
+
+        with patch('storage.api_key_store.a_session_maker', async_session_maker):
+            api_key = await api_key_store.get_or_create_system_api_key(
+                user_id=user_id,
+                org_id=org_id,
+                name=key_name,
+            )
+
+        assert api_key.startswith('sk-oh-')
+        assert len(api_key) == len('sk-oh-') + 32
+
+        # Verify the key was created in the database
+        async with async_session_maker() as session:
+            result = await session.execute(select(ApiKey).filter(ApiKey.key == api_key))
+            key_record = result.scalars().first()
+            assert key_record is not None
+            assert key_record.user_id == user_id
+            assert key_record.org_id == org_id
+            assert key_record.name == '__SYSTEM__:automation'
+            assert key_record.expires_at is None  # System keys never expire
+
+    @pytest.mark.asyncio
+    async def test_get_or_create_system_api_key_returns_existing(
+        self, api_key_store, async_session_maker
+    ):
+        """Test that existing valid system key is returned."""
+        user_id = '5594c7b6-f959-4b81-92e9-b09c206f5081'
+        org_id = uuid.UUID('5594c7b6-f959-4b81-92e9-b09c206f5081')
+        key_name = 'automation'
+
+        with patch('storage.api_key_store.a_session_maker', async_session_maker):
+            # Create the first key
+            first_key = await api_key_store.get_or_create_system_api_key(
+                user_id=user_id,
+                org_id=org_id,
+                name=key_name,
+            )
+
+            # Request again - should return the same key
+            second_key = await api_key_store.get_or_create_system_api_key(
+                user_id=user_id,
+                org_id=org_id,
+                name=key_name,
+            )
+
+        assert first_key == second_key
+
+    @pytest.mark.asyncio
+    async def test_get_or_create_system_api_key_different_names(
+        self, api_key_store, async_session_maker
+    ):
+        """Test that different names create different keys."""
+        user_id = '5594c7b6-f959-4b81-92e9-b09c206f5081'
+        org_id = uuid.UUID('5594c7b6-f959-4b81-92e9-b09c206f5081')
+
+        with patch('storage.api_key_store.a_session_maker', async_session_maker):
+            key1 = await api_key_store.get_or_create_system_api_key(
+                user_id=user_id,
+                org_id=org_id,
+                name='automation-1',
+            )
+
+            key2 = await api_key_store.get_or_create_system_api_key(
+                user_id=user_id,
+                org_id=org_id,
+                name='automation-2',
+            )
+
+        assert key1 != key2
+
+    @pytest.mark.asyncio
+    async def test_get_or_create_system_api_key_reissues_expired(
+        self, api_key_store, async_session_maker
+    ):
+        """Test that expired system key is replaced with a new one."""
+        user_id = '5594c7b6-f959-4b81-92e9-b09c206f5081'
+        org_id = uuid.UUID('5594c7b6-f959-4b81-92e9-b09c206f5081')
+        key_name = 'automation'
+        system_key_name = '__SYSTEM__:automation'
+
+        # First, manually create an expired key
+        expired_time = datetime.now(UTC) - timedelta(hours=1)
+        async with async_session_maker() as session:
+            expired_key = ApiKey(
+                key='sk-oh-expired-key-12345678901234567890',
+                user_id=user_id,
+                org_id=org_id,
+                name=system_key_name,
+                expires_at=expired_time.replace(tzinfo=None),
+            )
+            session.add(expired_key)
+            await session.commit()
+
+        with patch('storage.api_key_store.a_session_maker', async_session_maker):
+            # Request the key - should create a new one
+            new_key = await api_key_store.get_or_create_system_api_key(
+                user_id=user_id,
+                org_id=org_id,
+                name=key_name,
+            )
+
+        assert new_key != 'sk-oh-expired-key-12345678901234567890'
+        assert new_key.startswith('sk-oh-')
+
+        # Verify old key was deleted and new key exists
+        async with async_session_maker() as session:
+            result = await session.execute(
+                select(ApiKey).filter(ApiKey.name == system_key_name)
+            )
+            keys = result.scalars().all()
+            assert len(keys) == 1
+            assert keys[0].key == new_key
+            assert keys[0].expires_at is None
+
+    @pytest.mark.asyncio
+    async def test_list_api_keys_excludes_system_keys(
+        self, api_key_store, async_session_maker
+    ):
+        """Test that list_api_keys excludes system keys."""
+        user_id = '5594c7b6-f959-4b81-92e9-b09c206f5081'
+        org_id = uuid.UUID('5594c7b6-f959-4b81-92e9-b09c206f5081')
+
+        # Create a user key and a system key
+        async with async_session_maker() as session:
+            user_key = ApiKey(
+                key='sk-oh-user-key-123456789012345678901',
+                user_id=user_id,
+                org_id=org_id,
+                name='my-user-key',
+            )
+            system_key = ApiKey(
+                key='sk-oh-system-key-12345678901234567890',
+                user_id=user_id,
+                org_id=org_id,
+                name='__SYSTEM__:automation',
+            )
+            mcp_key = ApiKey(
+                key='sk-oh-mcp-key-1234567890123456789012',
+                user_id=user_id,
+                org_id=org_id,
+                name='MCP_API_KEY',
+            )
+            session.add(user_key)
+            session.add(system_key)
+            session.add(mcp_key)
+            await session.commit()
+
+        # Mock UserStore.get_user_by_id to return a user with the correct org
+        mock_user = MagicMock()
+        mock_user.current_org_id = org_id
+
+        with patch('storage.api_key_store.a_session_maker', async_session_maker):
+            with patch(
+                'storage.api_key_store.UserStore.get_user_by_id', new_callable=AsyncMock
+            ) as mock_get_user:
+                mock_get_user.return_value = mock_user
+                keys = await api_key_store.list_api_keys(user_id)
+
+        # Should only return the user key
+        assert len(keys) == 1
+        assert keys[0].name == 'my-user-key'
+
+    @pytest.mark.asyncio
+    async def test_delete_api_key_by_id_protects_system_keys(
+        self, api_key_store, async_session_maker
+    ):
+        """Test that system keys cannot be deleted by users."""
+        user_id = '5594c7b6-f959-4b81-92e9-b09c206f5081'
+        org_id = uuid.UUID('5594c7b6-f959-4b81-92e9-b09c206f5081')
+
+        # Create a system key
+        async with async_session_maker() as session:
+            system_key = ApiKey(
+                key='sk-oh-system-key-12345678901234567890',
+                user_id=user_id,
+                org_id=org_id,
+                name='__SYSTEM__:automation',
+            )
+            session.add(system_key)
+            await session.commit()
+            key_id = system_key.id
+
+        with patch('storage.api_key_store.a_session_maker', async_session_maker):
+            # Attempt to delete without allow_system flag
+            result = await api_key_store.delete_api_key_by_id(
+                key_id, allow_system=False
+            )
+
+        assert result is False
+
+        # Verify the key still exists
+        async with async_session_maker() as session:
+            result = await session.execute(select(ApiKey).filter(ApiKey.id == key_id))
+            key_record = result.scalars().first()
+            assert key_record is not None
+
+    @pytest.mark.asyncio
+    async def test_delete_api_key_by_id_allows_system_with_flag(
+        self, api_key_store, async_session_maker
+    ):
+        """Test that system keys can be deleted with allow_system=True."""
+        user_id = '5594c7b6-f959-4b81-92e9-b09c206f5081'
+        org_id = uuid.UUID('5594c7b6-f959-4b81-92e9-b09c206f5081')
+
+        # Create a system key
+        async with async_session_maker() as session:
+            system_key = ApiKey(
+                key='sk-oh-system-key-12345678901234567890',
+                user_id=user_id,
+                org_id=org_id,
+                name='__SYSTEM__:automation',
+            )
+            session.add(system_key)
+            await session.commit()
+            key_id = system_key.id
+
+        with patch('storage.api_key_store.a_session_maker', async_session_maker):
+            # Delete with allow_system=True
+            result = await api_key_store.delete_api_key_by_id(key_id, allow_system=True)
+
+        assert result is True
+
+        # Verify the key was deleted
+        async with async_session_maker() as session:
+            result = await session.execute(select(ApiKey).filter(ApiKey.id == key_id))
+            key_record = result.scalars().first()
+            assert key_record is None
+
+    @pytest.mark.asyncio
+    async def test_delete_api_key_by_id_allows_regular_keys(
+        self, api_key_store, async_session_maker
+    ):
+        """Test that regular keys can be deleted normally."""
+        user_id = '5594c7b6-f959-4b81-92e9-b09c206f5081'
+        org_id = uuid.UUID('5594c7b6-f959-4b81-92e9-b09c206f5081')
+
+        # Create a regular key
+        async with async_session_maker() as session:
+            regular_key = ApiKey(
+                key='sk-oh-regular-key-1234567890123456789',
+                user_id=user_id,
+                org_id=org_id,
+                name='my-regular-key',
+            )
+            session.add(regular_key)
+            await session.commit()
+            key_id = regular_key.id
+
+        with patch('storage.api_key_store.a_session_maker', async_session_maker):
+            # Delete without allow_system flag - should work for regular keys
+            result = await api_key_store.delete_api_key_by_id(
+                key_id, allow_system=False
+            )
+
+        assert result is True
+
+        # Verify the key was deleted
+        async with async_session_maker() as session:
+            result = await session.execute(select(ApiKey).filter(ApiKey.id == key_id))
+            key_record = result.scalars().first()
+            assert key_record is None

enterprise/tests/unit/storage/test_org_llm_settings_store.py

@@ -47,7 +47,10 @@ async def test_get_current_org_by_user_id_success(async_session_maker):
     """
     # Arrange
     async with async_session_maker() as session:
-        org = Org(name='test-org', default_llm_model='claude-3')
+        org = Org(
+            name='test-org',
+            agent_settings={'schema_version': 1, 'llm.model': 'claude-3'},
+        )
         session.add(org)
         await session.flush()
 
@@ -63,7 +66,7 @@ async def test_get_current_org_by_user_id_success(async_session_maker):
     # Assert
     assert result is not None
     assert result.name == 'test-org'
-    assert result.default_llm_model == 'claude-3'
+    assert result.agent_settings['llm.model'] == 'claude-3'
 
 
 @pytest.mark.asyncio
@@ -94,15 +97,20 @@ async def test_update_org_llm_settings_success(async_session_maker):
     """
     # Arrange
     async with async_session_maker() as session:
-        org = Org(name='test-org', default_llm_model='old-model')
+        org = Org(
+            name='test-org',
+            agent_settings={'schema_version': 1, 'llm.model': 'old-model'},
+        )
         session.add(org)
         await session.commit()
         org_id = org.id
 
         update_data = OrgLLMSettingsUpdate(
-            default_llm_model='new-model',
-            agent='CodeActAgent',
-            confirmation_mode=True,
+            agent_settings={
+                'llm.model': 'new-model',
+                'agent': 'CodeActAgent',
+                'verification.confirmation_mode': True,
+            }
         )
 
         # Act
@@ -115,9 +123,9 @@ async def test_update_org_llm_settings_success(async_session_maker):
 
     # Assert
     assert result is not None
-    assert result.default_llm_model == 'new-model'
-    assert result.agent == 'CodeActAgent'
-    assert result.confirmation_mode is True
+    assert result.agent_settings['llm.model'] == 'new-model'
+    assert result.agent_settings['agent'] == 'CodeActAgent'
+    assert result.agent_settings['verification.confirmation_mode'] is True
 
 
 @pytest.mark.asyncio
@@ -129,7 +137,7 @@ async def test_update_org_llm_settings_org_not_found(async_session_maker):
     """
     # Arrange
     non_existent_org_id = uuid.uuid4()
-    update_data = OrgLLMSettingsUpdate(default_llm_model='new-model')
+    update_data = OrgLLMSettingsUpdate(agent_settings={'llm.model': 'new-model'})
 
     # Act
     async with async_session_maker() as session:
@@ -149,13 +157,16 @@ async def test_update_org_llm_settings_propagates_to_members(async_session_maker
     """
     # Arrange
     async with async_session_maker() as session:
-        org = Org(name='test-org', default_llm_model='old-model')
+        org = Org(
+            name='test-org',
+            agent_settings={'schema_version': 1, 'llm.model': 'old-model'},
+        )
         session.add(org)
         await session.commit()
         org_id = org.id
 
         update_data = OrgLLMSettingsUpdate(
-            default_llm_model='new-model',
+            agent_settings={'llm.model': 'new-model'},
             llm_api_key='new-api-key',
         )
 
@@ -171,5 +182,5 @@ async def test_update_org_llm_settings_propagates_to_members(async_session_maker
         mock_update_members.assert_called_once()
         call_args = mock_update_members.call_args
         member_settings = call_args[0][2]
-        assert member_settings.llm_model == 'new-model'
+        assert member_settings.agent_settings is None
         assert member_settings.llm_api_key == 'new-api-key'

enterprise/tests/unit/storage/test_saas_sql_app_conversation_info_service.py

@@ -10,6 +10,9 @@ from unittest.mock import AsyncMock, MagicMock
 from uuid import UUID, uuid4
 
 import pytest
+from server.utils.saas_app_conversation_info_injector import (
+    SaasSQLAppConversationInfoService,
+)
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker, create_async_engine
 from sqlalchemy.pool import StaticPool
@@ -17,9 +20,6 @@ from storage.base import Base
 from storage.org import Org
 from storage.user import User
 
-from enterprise.server.utils.saas_app_conversation_info_injector import (
-    SaasSQLAppConversationInfoService,
-)
 from openhands.app_server.app_conversation.app_conversation_models import (
     AppConversationInfo,
 )
@@ -76,13 +76,11 @@ async def async_session_with_users(async_engine) -> AsyncGenerator[AsyncSession,
         org1 = Org(
             id=ORG1_ID,
             name='test-org-1',
-            enable_default_condenser=True,
             enable_proactive_conversation_starters=True,
         )
         org2 = Org(
             id=ORG2_ID,
             name='test-org-2',
-            enable_default_condenser=True,
             enable_proactive_conversation_starters=True,
         )
         db_session.add(org1)
@@ -663,3 +661,644 @@ class TestSaasSQLAppConversationInfoServiceAdminContext:
 
         admin_page = await admin_service.search_app_conversation_info()
         assert len(admin_page.items) == 5
+
+
+class TestSaasSQLAppConversationInfoServiceWebhookFallback:
+    """Test suite for webhook callback fallback using info.created_by_user_id."""
+
+    @pytest.mark.asyncio
+    async def test_save_with_admin_context_uses_created_by_user_id_fallback(
+        self,
+        async_session_with_users: AsyncSession,
+    ):
+        """Test that save_app_conversation_info uses info.created_by_user_id when user_context returns None.
+
+        This is the key fix for SDK-created conversations: when the webhook endpoint
+        uses ADMIN context (user_id=None), the service should fall back to using
+        the created_by_user_id from the AppConversationInfo object.
+        """
+        from storage.stored_conversation_metadata_saas import (
+            StoredConversationMetadataSaas,
+        )
+
+        from openhands.app_server.user.specifiy_user_context import ADMIN
+
+        # Arrange: Create service with ADMIN context (user_id=None)
+        admin_service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=ADMIN,
+        )
+
+        # Create conversation info with created_by_user_id set (as would come from sandbox_info)
+        conv_id = uuid4()
+        conv_info = AppConversationInfo(
+            id=conv_id,
+            created_by_user_id=str(USER1_ID),  # This should be used as fallback
+            sandbox_id='sandbox_webhook_test',
+            title='Webhook Created Conversation',
+        )
+
+        # Act: Save using ADMIN context
+        await admin_service.save_app_conversation_info(conv_info)
+
+        # Assert: SAAS metadata should be created with user_id from info.created_by_user_id
+        saas_query = select(StoredConversationMetadataSaas).where(
+            StoredConversationMetadataSaas.conversation_id == str(conv_id)
+        )
+        result = await async_session_with_users.execute(saas_query)
+        saas_metadata = result.scalar_one_or_none()
+
+        assert saas_metadata is not None, 'SAAS metadata should be created'
+        assert (
+            saas_metadata.user_id == USER1_ID
+        ), 'user_id should match info.created_by_user_id'
+        assert saas_metadata.org_id == ORG1_ID, 'org_id should match user current org'
+
+    @pytest.mark.asyncio
+    async def test_save_with_admin_context_no_user_id_skips_saas_metadata(
+        self,
+        async_session_with_users: AsyncSession,
+    ):
+        """Test that save_app_conversation_info skips SAAS metadata when both user_context and info have no user_id."""
+        from storage.stored_conversation_metadata_saas import (
+            StoredConversationMetadataSaas,
+        )
+
+        from openhands.app_server.user.specifiy_user_context import ADMIN
+
+        # Arrange: Create service with ADMIN context (user_id=None)
+        admin_service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=ADMIN,
+        )
+
+        # Create conversation info without created_by_user_id
+        conv_id = uuid4()
+        conv_info = AppConversationInfo(
+            id=conv_id,
+            created_by_user_id=None,  # No user_id available
+            sandbox_id='sandbox_no_user',
+            title='No User Conversation',
+        )
+
+        # Act: Save using ADMIN context with no user_id fallback
+        await admin_service.save_app_conversation_info(conv_info)
+
+        # Assert: SAAS metadata should NOT be created
+        saas_query = select(StoredConversationMetadataSaas).where(
+            StoredConversationMetadataSaas.conversation_id == str(conv_id)
+        )
+        result = await async_session_with_users.execute(saas_query)
+        saas_metadata = result.scalar_one_or_none()
+
+        assert (
+            saas_metadata is None
+        ), 'SAAS metadata should not be created without user_id'
+
+    @pytest.mark.asyncio
+    async def test_webhook_created_conversation_visible_to_user(
+        self,
+        async_session_with_users: AsyncSession,
+    ):
+        """Test end-to-end: conversation saved via webhook is visible to the owning user."""
+        from openhands.app_server.user.specifiy_user_context import ADMIN
+
+        # Arrange: Save conversation using ADMIN context (simulating webhook)
+        admin_service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=ADMIN,
+        )
+
+        conv_id = uuid4()
+        conv_info = AppConversationInfo(
+            id=conv_id,
+            created_by_user_id=str(USER1_ID),
+            sandbox_id='sandbox_webhook_e2e',
+            title='E2E Webhook Conversation',
+        )
+        await admin_service.save_app_conversation_info(conv_info)
+
+        # Act: Query as the owning user
+        user1_service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=SpecifyUserContext(user_id=str(USER1_ID)),
+        )
+        user1_page = await user1_service.search_app_conversation_info()
+
+        # Assert: User should see the webhook-created conversation
+        assert len(user1_page.items) == 1
+        assert user1_page.items[0].id == conv_id
+        assert user1_page.items[0].title == 'E2E Webhook Conversation'
+
+
+class TestSandboxIdFilterSaas:
+    """Test suite for sandbox_id__eq filter parameter in SAAS service."""
+
+    @pytest.mark.asyncio
+    async def test_search_by_sandbox_id(
+        self,
+        async_session_with_users: AsyncSession,
+    ):
+        """Test searching conversations by exact sandbox_id match with SAAS user filtering."""
+        # Create service for user1
+        user1_service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=SpecifyUserContext(user_id=str(USER1_ID)),
+        )
+
+        # Create conversations with different sandbox IDs for user1
+        conv1 = AppConversationInfo(
+            id=uuid4(),
+            created_by_user_id=str(USER1_ID),
+            sandbox_id='sandbox_alpha',
+            title='Conversation Alpha',
+            created_at=datetime(2024, 1, 1, 12, 0, 0, tzinfo=timezone.utc),
+            updated_at=datetime(2024, 1, 1, 12, 30, 0, tzinfo=timezone.utc),
+        )
+        conv2 = AppConversationInfo(
+            id=uuid4(),
+            created_by_user_id=str(USER1_ID),
+            sandbox_id='sandbox_beta',
+            title='Conversation Beta',
+            created_at=datetime(2024, 1, 1, 13, 0, 0, tzinfo=timezone.utc),
+            updated_at=datetime(2024, 1, 1, 13, 30, 0, tzinfo=timezone.utc),
+        )
+        conv3 = AppConversationInfo(
+            id=uuid4(),
+            created_by_user_id=str(USER1_ID),
+            sandbox_id='sandbox_alpha',
+            title='Conversation Gamma',
+            created_at=datetime(2024, 1, 1, 14, 0, 0, tzinfo=timezone.utc),
+            updated_at=datetime(2024, 1, 1, 14, 30, 0, tzinfo=timezone.utc),
+        )
+
+        # Save all conversations
+        await user1_service.save_app_conversation_info(conv1)
+        await user1_service.save_app_conversation_info(conv2)
+        await user1_service.save_app_conversation_info(conv3)
+
+        # Search for sandbox_alpha - should return 2 conversations
+        page = await user1_service.search_app_conversation_info(
+            sandbox_id__eq='sandbox_alpha'
+        )
+        assert len(page.items) == 2
+        sandbox_ids = {item.sandbox_id for item in page.items}
+        assert sandbox_ids == {'sandbox_alpha'}
+        conversation_ids = {item.id for item in page.items}
+        assert conv1.id in conversation_ids
+        assert conv3.id in conversation_ids
+
+        # Search for sandbox_beta - should return 1 conversation
+        page = await user1_service.search_app_conversation_info(
+            sandbox_id__eq='sandbox_beta'
+        )
+        assert len(page.items) == 1
+        assert page.items[0].id == conv2.id
+
+        # Search for non-existent sandbox - should return 0 conversations
+        page = await user1_service.search_app_conversation_info(
+            sandbox_id__eq='sandbox_nonexistent'
+        )
+        assert len(page.items) == 0
+
+    @pytest.mark.asyncio
+    async def test_count_by_sandbox_id(
+        self,
+        async_session_with_users: AsyncSession,
+    ):
+        """Test counting conversations by exact sandbox_id match with SAAS user filtering."""
+        # Create service for user1
+        user1_service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=SpecifyUserContext(user_id=str(USER1_ID)),
+        )
+
+        # Create conversations with different sandbox IDs
+        conv1 = AppConversationInfo(
+            id=uuid4(),
+            created_by_user_id=str(USER1_ID),
+            sandbox_id='sandbox_x',
+            title='Conversation X1',
+            created_at=datetime(2024, 1, 1, 12, 0, 0, tzinfo=timezone.utc),
+            updated_at=datetime(2024, 1, 1, 12, 30, 0, tzinfo=timezone.utc),
+        )
+        conv2 = AppConversationInfo(
+            id=uuid4(),
+            created_by_user_id=str(USER1_ID),
+            sandbox_id='sandbox_y',
+            title='Conversation Y1',
+            created_at=datetime(2024, 1, 1, 13, 0, 0, tzinfo=timezone.utc),
+            updated_at=datetime(2024, 1, 1, 13, 30, 0, tzinfo=timezone.utc),
+        )
+        conv3 = AppConversationInfo(
+            id=uuid4(),
+            created_by_user_id=str(USER1_ID),
+            sandbox_id='sandbox_x',
+            title='Conversation X2',
+            created_at=datetime(2024, 1, 1, 14, 0, 0, tzinfo=timezone.utc),
+            updated_at=datetime(2024, 1, 1, 14, 30, 0, tzinfo=timezone.utc),
+        )
+
+        # Save all conversations
+        await user1_service.save_app_conversation_info(conv1)
+        await user1_service.save_app_conversation_info(conv2)
+        await user1_service.save_app_conversation_info(conv3)
+
+        # Count for sandbox_x - should be 2
+        count = await user1_service.count_app_conversation_info(
+            sandbox_id__eq='sandbox_x'
+        )
+        assert count == 2
+
+        # Count for sandbox_y - should be 1
+        count = await user1_service.count_app_conversation_info(
+            sandbox_id__eq='sandbox_y'
+        )
+        assert count == 1
+
+        # Count for non-existent sandbox - should be 0
+        count = await user1_service.count_app_conversation_info(
+            sandbox_id__eq='sandbox_nonexistent'
+        )
+        assert count == 0
+
+    @pytest.mark.asyncio
+    async def test_sandbox_id_filter_respects_user_isolation(
+        self,
+        async_session_with_users: AsyncSession,
+    ):
+        """Test that sandbox_id filter respects user isolation in SAAS environment."""
+        # Create services for both users
+        user1_service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=SpecifyUserContext(user_id=str(USER1_ID)),
+        )
+        user2_service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=SpecifyUserContext(user_id=str(USER2_ID)),
+        )
+
+        # Create conversation with same sandbox_id for both users
+        shared_sandbox_id = 'sandbox_shared'
+
+        conv_user1 = AppConversationInfo(
+            id=uuid4(),
+            created_by_user_id=str(USER1_ID),
+            sandbox_id=shared_sandbox_id,
+            title='User1 Conversation',
+            created_at=datetime(2024, 1, 1, 12, 0, 0, tzinfo=timezone.utc),
+            updated_at=datetime(2024, 1, 1, 12, 30, 0, tzinfo=timezone.utc),
+        )
+        conv_user2 = AppConversationInfo(
+            id=uuid4(),
+            created_by_user_id=str(USER2_ID),
+            sandbox_id=shared_sandbox_id,
+            title='User2 Conversation',
+            created_at=datetime(2024, 1, 1, 13, 0, 0, tzinfo=timezone.utc),
+            updated_at=datetime(2024, 1, 1, 13, 30, 0, tzinfo=timezone.utc),
+        )
+
+        # Save conversations
+        await user1_service.save_app_conversation_info(conv_user1)
+        await user2_service.save_app_conversation_info(conv_user2)
+
+        # User1 should only see their own conversation with this sandbox_id
+        page = await user1_service.search_app_conversation_info(
+            sandbox_id__eq=shared_sandbox_id
+        )
+        assert len(page.items) == 1
+        assert page.items[0].id == conv_user1.id
+        assert page.items[0].title == 'User1 Conversation'
+
+        # User2 should only see their own conversation with this sandbox_id
+        page = await user2_service.search_app_conversation_info(
+            sandbox_id__eq=shared_sandbox_id
+        )
+        assert len(page.items) == 1
+        assert page.items[0].id == conv_user2.id
+        assert page.items[0].title == 'User2 Conversation'
+
+        # Count should also respect user isolation
+        count = await user1_service.count_app_conversation_info(
+            sandbox_id__eq=shared_sandbox_id
+        )
+        assert count == 1
+
+        count = await user2_service.count_app_conversation_info(
+            sandbox_id__eq=shared_sandbox_id
+        )
+        assert count == 1
+
+
+class TestApiKeyOrgIdHandling:
+    """Test suite for API key organization ID handling in save_app_conversation_info.
+
+    These tests verify that when a conversation is created using API key authentication,
+    the conversation is associated with the API key's bound organization, not the user's
+    currently selected organization.
+    """
+
+    @pytest.mark.asyncio
+    async def test_api_key_org_id_used_when_available(
+        self,
+        async_session_with_users: AsyncSession,
+    ):
+        """Test that API key's org_id is used when saving conversation via API key auth.
+
+        This tests the main bug fix: when a user creates an API key in Personal Workspace,
+        then switches to OpenHands org in browser, and uses the API key to create a
+        conversation, the conversation should be saved in Personal Workspace (API key's org),
+        not OpenHands (user's current org).
+        """
+        from dataclasses import dataclass
+
+        from storage.stored_conversation_metadata_saas import (
+            StoredConversationMetadataSaas,
+        )
+
+        # Create a mock UserAuth with API key org_id
+        @dataclass
+        class MockUserAuth:
+            user_id: str
+            api_key_org_id: UUID | None = None
+
+            async def get_user_id(self) -> str:
+                return self.user_id
+
+            def get_api_key_org_id(self) -> UUID | None:
+                return self.api_key_org_id
+
+        # Create a mock UserContext that wraps the MockUserAuth
+        @dataclass
+        class MockAuthUserContext:
+            user_auth: MockUserAuth
+
+            async def get_user_id(self) -> str | None:
+                return await self.user_auth.get_user_id()
+
+        # Simulate: User1's current org is ORG2, but API key is bound to ORG1
+        # First, update user1's current_org_id to ORG2
+        result = await async_session_with_users.execute(
+            select(User).where(User.id == USER1_ID)
+        )
+        user_to_update = result.scalars().first()
+        user_to_update.current_org_id = ORG2_ID  # User is viewing ORG2
+        await async_session_with_users.commit()
+        async_session_with_users.expire_all()
+
+        # Create service with mock auth context where API key is bound to ORG1
+        mock_user_auth = MockUserAuth(
+            user_id=str(USER1_ID),
+            api_key_org_id=ORG1_ID,  # API key created in ORG1
+        )
+        mock_context = MockAuthUserContext(user_auth=mock_user_auth)
+
+        service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=mock_context,
+        )
+
+        # Create and save a conversation
+        conv_id = uuid4()
+        conv_info = AppConversationInfo(
+            id=conv_id,
+            created_by_user_id=str(USER1_ID),
+            sandbox_id='sandbox_api_key_test',
+            title='API Key Created Conversation',
+        )
+        await service.save_app_conversation_info(conv_info)
+
+        # Verify: SAAS metadata should have ORG1 (API key's org), not ORG2 (user's current org)
+        saas_query = select(StoredConversationMetadataSaas).where(
+            StoredConversationMetadataSaas.conversation_id == str(conv_id)
+        )
+        result = await async_session_with_users.execute(saas_query)
+        saas_metadata = result.scalar_one_or_none()
+
+        assert saas_metadata is not None, 'SAAS metadata should be created'
+        assert saas_metadata.user_id == USER1_ID
+        assert (
+            saas_metadata.org_id == ORG1_ID
+        ), 'Conversation should be in API key org (ORG1), not user current org (ORG2)'
+
+    @pytest.mark.asyncio
+    async def test_legacy_api_key_without_org_uses_user_current_org(
+        self,
+        async_session_with_users: AsyncSession,
+    ):
+        """Test that legacy API keys (without org_id) fall back to user's current org.
+
+        Legacy API keys created before the org_id feature was added will have
+        api_key_org_id = None. In this case, we should fall back to the user's
+        current_org_id.
+        """
+        from dataclasses import dataclass
+
+        from storage.stored_conversation_metadata_saas import (
+            StoredConversationMetadataSaas,
+        )
+
+        # Create a mock UserAuth with API key but NO org_id (legacy key)
+        @dataclass
+        class MockUserAuth:
+            user_id: str
+            api_key_org_id: UUID | None = None
+
+            async def get_user_id(self) -> str:
+                return self.user_id
+
+            def get_api_key_org_id(self) -> UUID | None:
+                return self.api_key_org_id
+
+        @dataclass
+        class MockAuthUserContext:
+            user_auth: MockUserAuth
+
+            async def get_user_id(self) -> str | None:
+                return await self.user_auth.get_user_id()
+
+        # Create service with mock auth context where API key has NO org_id
+        mock_user_auth = MockUserAuth(
+            user_id=str(USER1_ID),
+            api_key_org_id=None,  # Legacy key without org binding
+        )
+        mock_context = MockAuthUserContext(user_auth=mock_user_auth)
+
+        service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=mock_context,
+        )
+
+        # Create and save a conversation
+        conv_id = uuid4()
+        conv_info = AppConversationInfo(
+            id=conv_id,
+            created_by_user_id=str(USER1_ID),
+            sandbox_id='sandbox_legacy_key_test',
+            title='Legacy API Key Conversation',
+        )
+        await service.save_app_conversation_info(conv_info)
+
+        # Verify: SAAS metadata should use user's current org (ORG1) as fallback
+        saas_query = select(StoredConversationMetadataSaas).where(
+            StoredConversationMetadataSaas.conversation_id == str(conv_id)
+        )
+        result = await async_session_with_users.execute(saas_query)
+        saas_metadata = result.scalar_one_or_none()
+
+        assert saas_metadata is not None, 'SAAS metadata should be created'
+        assert saas_metadata.user_id == USER1_ID
+        assert (
+            saas_metadata.org_id == ORG1_ID
+        ), 'Legacy key should fall back to user current org (ORG1)'
+
+    @pytest.mark.asyncio
+    async def test_cookie_auth_without_api_key_uses_user_current_org(
+        self,
+        async_session_with_users: AsyncSession,
+    ):
+        """Test that cookie auth (no API key) uses user's current org.
+
+        When authenticated via browser cookie (no API key), there's no
+        get_api_key_org_id method, so we use user's current_org_id.
+        This is already tested by other tests using SpecifyUserContext,
+        but we explicitly test the case where user_context doesn't have user_auth.
+        """
+        from storage.stored_conversation_metadata_saas import (
+            StoredConversationMetadataSaas,
+        )
+
+        # Use SpecifyUserContext which doesn't have user_auth attribute
+        service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=SpecifyUserContext(user_id=str(USER1_ID)),
+        )
+
+        # Create and save a conversation
+        conv_id = uuid4()
+        conv_info = AppConversationInfo(
+            id=conv_id,
+            created_by_user_id=str(USER1_ID),
+            sandbox_id='sandbox_cookie_auth_test',
+            title='Cookie Auth Conversation',
+        )
+        await service.save_app_conversation_info(conv_info)
+
+        # Verify: SAAS metadata should use user's current org (ORG1)
+        saas_query = select(StoredConversationMetadataSaas).where(
+            StoredConversationMetadataSaas.conversation_id == str(conv_id)
+        )
+        result = await async_session_with_users.execute(saas_query)
+        saas_metadata = result.scalar_one_or_none()
+
+        assert saas_metadata is not None, 'SAAS metadata should be created'
+        assert saas_metadata.user_id == USER1_ID
+        assert (
+            saas_metadata.org_id == ORG1_ID
+        ), 'Cookie auth should use user current org (ORG1)'
+
+    @pytest.mark.asyncio
+    async def test_api_key_org_isolation_cross_org_visibility(
+        self,
+        async_session_with_users: AsyncSession,
+    ):
+        """Test end-to-end: conversation created via API key is visible in correct org.
+
+        Simulates the full bug scenario:
+        1. Create conversation via API key (bound to ORG1)
+        2. User switches to ORG2
+        3. User should NOT see the conversation in ORG2
+        4. User switches back to ORG1
+        5. User should see the conversation in ORG1
+        """
+        from dataclasses import dataclass
+
+        @dataclass
+        class MockUserAuth:
+            user_id: str
+            api_key_org_id: UUID | None = None
+
+            async def get_user_id(self) -> str:
+                return self.user_id
+
+            def get_api_key_org_id(self) -> UUID | None:
+                return self.api_key_org_id
+
+        @dataclass
+        class MockAuthUserContext:
+            user_auth: MockUserAuth
+
+            async def get_user_id(self) -> str | None:
+                return await self.user_auth.get_user_id()
+
+        # Step 1: Create conversation via API key bound to ORG1
+        mock_user_auth = MockUserAuth(
+            user_id=str(USER1_ID),
+            api_key_org_id=ORG1_ID,
+        )
+        mock_context = MockAuthUserContext(user_auth=mock_user_auth)
+
+        api_key_service = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=mock_context,
+        )
+
+        conv_id = uuid4()
+        conv_info = AppConversationInfo(
+            id=conv_id,
+            created_by_user_id=str(USER1_ID),
+            sandbox_id='sandbox_e2e_api_key',
+            title='E2E API Key Conversation',
+        )
+        await api_key_service.save_app_conversation_info(conv_info)
+
+        # Step 2: Switch user to ORG2 in browser session
+        result = await async_session_with_users.execute(
+            select(User).where(User.id == USER1_ID)
+        )
+        user_to_update = result.scalars().first()
+        user_to_update.current_org_id = ORG2_ID
+        await async_session_with_users.commit()
+        async_session_with_users.expire_all()
+
+        # Step 3: User in ORG2 should NOT see the conversation
+        user_service_org2 = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=SpecifyUserContext(user_id=str(USER1_ID)),
+        )
+        page_org2 = await user_service_org2.search_app_conversation_info()
+        assert (
+            len(page_org2.items) == 0
+        ), 'User in ORG2 should not see conversation created via API key in ORG1'
+
+        # Also verify get_app_conversation_info returns None
+        conv_from_org2 = await user_service_org2.get_app_conversation_info(conv_id)
+        assert (
+            conv_from_org2 is None
+        ), 'User in ORG2 should not access conversation from ORG1'
+
+        # Step 4: Switch user back to ORG1
+        result = await async_session_with_users.execute(
+            select(User).where(User.id == USER1_ID)
+        )
+        user_to_update = result.scalars().first()
+        user_to_update.current_org_id = ORG1_ID
+        await async_session_with_users.commit()
+        async_session_with_users.expire_all()
+
+        # Step 5: User in ORG1 should see the conversation
+        user_service_org1 = SaasSQLAppConversationInfoService(
+            db_session=async_session_with_users,
+            user_context=SpecifyUserContext(user_id=str(USER1_ID)),
+        )
+        page_org1 = await user_service_org1.search_app_conversation_info()
+        assert (
+            len(page_org1.items) == 1
+        ), 'User in ORG1 should see conversation created via API key in ORG1'
+        assert page_org1.items[0].id == conv_id
+        assert page_org1.items[0].title == 'E2E API Key Conversation'
+
+        # Also verify get_app_conversation_info works
+        conv_from_org1 = await user_service_org1.get_app_conversation_info(conv_id)
+        assert conv_from_org1 is not None
+        assert conv_from_org1.id == conv_id

enterprise/tests/unit/storage/test_slack_conversation_store.py

@@ -0,0 +1,75 @@
+"""Unit tests for SlackConversationStore."""
+
+from unittest.mock import patch
+
+import pytest
+from sqlalchemy import select
+from storage.slack_conversation import SlackConversation
+from storage.slack_conversation_store import SlackConversationStore
+
+
+@pytest.fixture
+def slack_conversation_store():
+    """Create SlackConversationStore instance."""
+    return SlackConversationStore()
+
+
+class TestSlackConversationStore:
+    """Test cases for SlackConversationStore."""
+
+    @pytest.mark.asyncio
+    async def test_create_slack_conversation_persists_to_database(
+        self, slack_conversation_store, async_session_maker
+    ):
+        """Test that create_slack_conversation actually stores data in the database.
+
+        This test verifies that the await statement is present before session.merge().
+        Without the await, the data won't be persisted and subsequent lookups will
+        return None even though we just created the conversation.
+        """
+        channel_id = 'C123456'
+        parent_id = '1234567890.123456'
+        conversation_id = 'conv-test-123'
+        keycloak_user_id = 'user-123'
+
+        slack_conversation = SlackConversation(
+            conversation_id=conversation_id,
+            channel_id=channel_id,
+            keycloak_user_id=keycloak_user_id,
+            parent_id=parent_id,
+        )
+
+        with patch(
+            'storage.slack_conversation_store.a_session_maker', async_session_maker
+        ):
+            # Create the slack conversation
+            await slack_conversation_store.create_slack_conversation(slack_conversation)
+
+            # Verify we can retrieve the conversation using the store method
+            result = await slack_conversation_store.get_slack_conversation(
+                channel_id=channel_id,
+                parent_id=parent_id,
+            )
+
+        # This assertion would fail if the await was missing before session.merge()
+        # because the data wouldn't be persisted to the database
+        assert result is not None, (
+            'Slack conversation was not persisted to the database. '
+            'Ensure await is used before session.merge() in create_slack_conversation.'
+        )
+        assert result.conversation_id == conversation_id
+        assert result.channel_id == channel_id
+        assert result.parent_id == parent_id
+        assert result.keycloak_user_id == keycloak_user_id
+
+        # Also verify directly in the database
+        async with async_session_maker() as session:
+            db_result = await session.execute(
+                select(SlackConversation).where(
+                    SlackConversation.channel_id == channel_id,
+                    SlackConversation.parent_id == parent_id,
+                )
+            )
+            db_conversation = db_result.scalar_one_or_none()
+            assert db_conversation is not None
+            assert db_conversation.conversation_id == conversation_id

enterprise/tests/unit/test_api_key_store.py

@@ -5,7 +5,7 @@ from unittest.mock import AsyncMock, MagicMock, patch
 import pytest
 from sqlalchemy import select
 from storage.api_key import ApiKey
-from storage.api_key_store import ApiKeyStore
+from storage.api_key_store import ApiKeyStore, ApiKeyValidationResult
 
 
 @pytest.fixture
@@ -110,8 +110,8 @@ async def test_create_api_key(
 
 @pytest.mark.asyncio
 async def test_validate_api_key_valid(api_key_store, async_session_maker):
-    """Test validating a valid API key."""
-    # Setup - create an API key in the database
+    """Test validating a valid API key returns user_id and org_id."""
+    # Arrange
     user_id = str(uuid.uuid4())
     org_id = uuid.uuid4()
     api_key_value = 'test-api-key'
@@ -126,13 +126,19 @@ async def test_validate_api_key_valid(api_key_store, async_session_maker):
         )
         session.add(key_record)
         await session.commit()
+        key_id = key_record.id
 
-    # Execute - patch a_session_maker to use test's async session maker
+    # Act
     with patch('storage.api_key_store.a_session_maker', async_session_maker):
         result = await api_key_store.validate_api_key(api_key_value)
 
-    # Verify
-    assert result == user_id
+    # Assert
+    assert isinstance(result, ApiKeyValidationResult)
+    assert result is not None
+    assert result.user_id == user_id
+    assert result.org_id == org_id
+    assert result.key_id == key_id
+    assert result.key_name == 'Test Key'
 
 
 @pytest.mark.asyncio
@@ -197,7 +203,7 @@ async def test_validate_api_key_valid_timezone_naive(
     api_key_store, async_session_maker
 ):
     """Test validating a valid API key with timezone-naive datetime from database."""
-    # Setup - create a valid API key with timezone-naive datetime (future date)
+    # Arrange
     user_id = str(uuid.uuid4())
     org_id = uuid.uuid4()
     api_key_value = 'test-valid-naive-key'
@@ -214,12 +220,44 @@ async def test_validate_api_key_valid_timezone_naive(
         session.add(key_record)
         await session.commit()
 
-    # Execute - patch a_session_maker to use test's async session maker
+    # Act
     with patch('storage.api_key_store.a_session_maker', async_session_maker):
         result = await api_key_store.validate_api_key(api_key_value)
 
-    # Verify
-    assert result == user_id
+    # Assert
+    assert isinstance(result, ApiKeyValidationResult)
+    assert result.user_id == user_id
+    assert result.org_id == org_id
+
+
+@pytest.mark.asyncio
+async def test_validate_api_key_legacy_without_org_id(
+    api_key_store, async_session_maker
+):
+    """Test validating a legacy API key without org_id returns None for org_id."""
+    # Arrange
+    user_id = str(uuid.uuid4())
+    api_key_value = 'test-legacy-key-no-org'
+
+    async with async_session_maker() as session:
+        key_record = ApiKey(
+            key=api_key_value,
+            user_id=user_id,
+            org_id=None,  # Legacy key without org binding
+            name='Legacy Key',
+        )
+        session.add(key_record)
+        await session.commit()
+
+    # Act
+    with patch('storage.api_key_store.a_session_maker', async_session_maker):
+        result = await api_key_store.validate_api_key(api_key_value)
+
+    # Assert
+    assert isinstance(result, ApiKeyValidationResult)
+    assert result is not None
+    assert result.user_id == user_id
+    assert result.org_id is None
 
 
 @pytest.mark.asyncio

enterprise/tests/unit/test_auth_routes.py

@@ -11,7 +11,6 @@ from server.auth.auth_error import AuthError
 from server.auth.saas_user_auth import SaasUserAuth
 from server.auth.user.user_authorizer import UserAuthorizationResponse, UserAuthorizer
 from server.routes.auth import (
-    _extract_recaptcha_state,
     accept_tos,
     authenticate,
     keycloak_callback,
@@ -55,11 +54,12 @@ def mock_response():
 def test_set_response_cookie(mock_response, mock_request):
     """Test setting the auth cookie on a response."""
 
-    with patch('server.routes.auth.config') as mock_config:
+    with (
+        patch('server.routes.auth.config') as mock_config,
+        patch('server.utils.url_utils.get_global_config') as get_global_config,
+    ):
         mock_config.jwt_secret.get_secret_value.return_value = 'test_secret'
-
-        # Configure mock_request.url.hostname
-        mock_request.url.hostname = 'example.com'
+        get_global_config.return_value = MagicMock(web_url='https://example.com')
 
         set_response_cookie(
             request=mock_request,
@@ -846,10 +846,108 @@ async def test_keycloak_callback_duplicate_email_detected(
         assert exc_info.value.detail == 'duplicate_email'
 
 
-# Note: test_keycloak_callback_duplicate_email_deletion_fails was removed as part of
-# the user authorization refactor. The Keycloak user deletion logic for duplicate emails
-# has been removed from keycloak_callback. If this behavior needs to be restored,
-# it should be implemented in the DefaultUserAuthorizer or handled separately.
+@pytest.mark.asyncio
+async def test_keycloak_callback_duplicate_email_deletes_new_keycloak_user(
+    mock_request, create_keycloak_user_info
+):
+    """Test that new Keycloak user is deleted when duplicate email is detected.
+
+    When a user attempts to sign up with a +modifier email (e.g., joe+1@example.com)
+    and an account with the base email already exists, the newly created Keycloak
+    user should be deleted to prevent orphaned accounts from blocking future sign-ins.
+    """
+    with (
+        patch('server.routes.auth.token_manager') as mock_token_manager,
+        patch('server.routes.auth.UserStore') as mock_user_store,
+    ):
+        # Arrange
+        mock_token_manager.get_keycloak_tokens = AsyncMock(
+            return_value=('test_access_token', 'test_refresh_token')
+        )
+        mock_token_manager.get_user_info = AsyncMock(
+            return_value=create_keycloak_user_info(
+                sub='new_user_id',
+                preferred_username='test_user',
+                email='joe+1@example.com',
+                identity_provider='github',
+            )
+        )
+        mock_token_manager.delete_keycloak_user = AsyncMock(return_value=True)
+
+        # User does NOT exist in UserStore (new signup attempt)
+        mock_user_store.get_user_by_id = AsyncMock(return_value=None)
+
+        # Create mock authorizer that returns duplicate_email error
+        mock_authorizer = create_mock_user_authorizer(
+            success=False, error_detail='duplicate_email'
+        )
+
+        # Act & Assert
+        with pytest.raises(HTTPException) as exc_info:
+            await keycloak_callback(
+                code='test_code',
+                state='test_state',
+                request=mock_request,
+                user_authorizer=mock_authorizer,
+            )
+
+        assert exc_info.value.status_code == status.HTTP_401_UNAUTHORIZED
+        assert exc_info.value.detail == 'duplicate_email'
+        # Keycloak user should be deleted since user doesn't exist in UserStore
+        mock_token_manager.delete_keycloak_user.assert_called_once_with('new_user_id')
+
+
+@pytest.mark.asyncio
+async def test_keycloak_callback_duplicate_email_preserves_existing_user(
+    mock_request, create_keycloak_user_info
+):
+    """Test that existing users are not deleted when duplicate email is detected.
+
+    When an existing user signs in and duplicate email is detected (e.g., because
+    another account with the same base email was created while duplicate checking
+    was disabled), the existing user's Keycloak account should NOT be deleted.
+    """
+    with (
+        patch('server.routes.auth.token_manager') as mock_token_manager,
+        patch('server.routes.auth.UserStore') as mock_user_store,
+    ):
+        # Arrange
+        mock_token_manager.get_keycloak_tokens = AsyncMock(
+            return_value=('test_access_token', 'test_refresh_token')
+        )
+        mock_token_manager.get_user_info = AsyncMock(
+            return_value=create_keycloak_user_info(
+                sub='existing_user_id',
+                preferred_username='test_user',
+                email='joe@example.com',
+                identity_provider='github',
+            )
+        )
+        mock_token_manager.delete_keycloak_user = AsyncMock(return_value=True)
+
+        # User EXISTS in UserStore (legitimate existing user)
+        mock_existing_user = MagicMock()
+        mock_existing_user.id = 'existing_user_id'
+        mock_user_store.get_user_by_id = AsyncMock(return_value=mock_existing_user)
+
+        # Create mock authorizer that returns duplicate_email error
+        mock_authorizer = create_mock_user_authorizer(
+            success=False, error_detail='duplicate_email'
+        )
+
+        # Act & Assert
+        with pytest.raises(HTTPException) as exc_info:
+            await keycloak_callback(
+                code='test_code',
+                state='test_state',
+                request=mock_request,
+                user_authorizer=mock_authorizer,
+            )
+
+        assert exc_info.value.status_code == status.HTTP_401_UNAUTHORIZED
+        assert exc_info.value.detail == 'duplicate_email'
+        # Keycloak user should NOT be deleted since user exists in UserStore
+        mock_token_manager.delete_keycloak_user.assert_not_called()
 
 
 @pytest.mark.asyncio
@@ -1036,79 +1134,6 @@ async def test_keycloak_callback_no_email_in_user_info(
         mock_token_manager.check_duplicate_base_email.assert_not_called()
 
 
-class TestExtractRecaptchaState:
-    """Tests for _extract_recaptcha_state() helper function."""
-
-    def test_should_extract_redirect_url_and_token_from_new_json_format(self):
-        """Test extraction from new base64-encoded JSON format."""
-        # Arrange
-        state_data = {
-            'redirect_url': 'https://example.com',
-            'recaptcha_token': 'test-token',
-        }
-        encoded_state = base64.urlsafe_b64encode(
-            json.dumps(state_data).encode()
-        ).decode()
-
-        # Act
-        redirect_url, token = _extract_recaptcha_state(encoded_state)
-
-        # Assert
-        assert redirect_url == 'https://example.com'
-        assert token == 'test-token'
-
-    def test_should_handle_old_format_plain_redirect_url(self):
-        """Test handling of old format (plain redirect URL string)."""
-        # Arrange
-        state = 'https://example.com'
-
-        # Act
-        redirect_url, token = _extract_recaptcha_state(state)
-
-        # Assert
-        assert redirect_url == 'https://example.com'
-        assert token is None
-
-    def test_should_handle_none_state(self):
-        """Test handling of None state."""
-        # Arrange
-        state = None
-
-        # Act
-        redirect_url, token = _extract_recaptcha_state(state)
-
-        # Assert
-        assert redirect_url == ''
-        assert token is None
-
-    def test_should_handle_invalid_base64_gracefully(self):
-        """Test handling of invalid base64/JSON (fallback to old format)."""
-        # Arrange
-        state = 'not-valid-base64!!!'
-
-        # Act
-        redirect_url, token = _extract_recaptcha_state(state)
-
-        # Assert
-        assert redirect_url == state
-        assert token is None
-
-    def test_should_handle_missing_redirect_url_in_json(self):
-        """Test handling when redirect_url is missing in JSON."""
-        # Arrange
-        state_data = {'recaptcha_token': 'test-token'}
-        encoded_state = base64.urlsafe_b64encode(
-            json.dumps(state_data).encode()
-        ).decode()
-
-        # Act
-        redirect_url, token = _extract_recaptcha_state(encoded_state)
-
-        # Assert
-        assert redirect_url == ''
-        assert token == 'test-token'
-
-
 class TestKeycloakCallbackRecaptcha:
     """Tests for reCAPTCHA integration in keycloak_callback()."""
 

enterprise/tests/unit/test_authorization.py

@@ -13,6 +13,7 @@ from server.auth.authorization import (
     ROLE_PERMISSIONS,
     Permission,
     RoleName,
+    get_api_key_org_id_from_request,
     get_role_permissions,
     get_user_org_role,
     has_permission,
@@ -444,6 +445,15 @@ class TestGetUserOrgRole:
 # =============================================================================
 
 
+def _create_mock_request(api_key_org_id=None):
+    """Helper to create a mock request with optional api_key_org_id."""
+    mock_request = MagicMock()
+    mock_user_auth = MagicMock()
+    mock_user_auth.get_api_key_org_id.return_value = api_key_org_id
+    mock_request.state.user_auth = mock_user_auth
+    return mock_request
+
+
 class TestRequirePermission:
     """Tests for require_permission dependency factory."""
 
@@ -456,6 +466,7 @@ class TestRequirePermission:
         """
         user_id = str(uuid4())
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         mock_role = MagicMock()
         mock_role.name = 'admin'
@@ -465,7 +476,9 @@ class TestRequirePermission:
             AsyncMock(return_value=mock_role),
         ):
             permission_checker = require_permission(Permission.VIEW_LLM_SETTINGS)
-            result = await permission_checker(org_id=org_id, user_id=user_id)
+            result = await permission_checker(
+                request=mock_request, org_id=org_id, user_id=user_id
+            )
             assert result == user_id
 
     @pytest.mark.asyncio
@@ -476,10 +489,11 @@ class TestRequirePermission:
         THEN: 401 Unauthorized is raised
         """
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         permission_checker = require_permission(Permission.VIEW_LLM_SETTINGS)
         with pytest.raises(HTTPException) as exc_info:
-            await permission_checker(org_id=org_id, user_id=None)
+            await permission_checker(request=mock_request, org_id=org_id, user_id=None)
 
         assert exc_info.value.status_code == 401
         assert 'not authenticated' in exc_info.value.detail.lower()
@@ -493,6 +507,7 @@ class TestRequirePermission:
         """
         user_id = str(uuid4())
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         with patch(
             'server.auth.authorization.get_user_org_role',
@@ -500,7 +515,9 @@ class TestRequirePermission:
         ):
             permission_checker = require_permission(Permission.VIEW_LLM_SETTINGS)
             with pytest.raises(HTTPException) as exc_info:
-                await permission_checker(org_id=org_id, user_id=user_id)
+                await permission_checker(
+                    request=mock_request, org_id=org_id, user_id=user_id
+                )
 
             assert exc_info.value.status_code == 403
             assert 'not a member' in exc_info.value.detail.lower()
@@ -514,6 +531,7 @@ class TestRequirePermission:
         """
         user_id = str(uuid4())
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         mock_role = MagicMock()
         mock_role.name = 'member'
@@ -524,7 +542,9 @@ class TestRequirePermission:
         ):
             permission_checker = require_permission(Permission.DELETE_ORGANIZATION)
             with pytest.raises(HTTPException) as exc_info:
-                await permission_checker(org_id=org_id, user_id=user_id)
+                await permission_checker(
+                    request=mock_request, org_id=org_id, user_id=user_id
+                )
 
             assert exc_info.value.status_code == 403
             assert 'delete_organization' in exc_info.value.detail.lower()
@@ -538,6 +558,7 @@ class TestRequirePermission:
         """
         user_id = str(uuid4())
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         mock_role = MagicMock()
         mock_role.name = 'owner'
@@ -547,7 +568,9 @@ class TestRequirePermission:
             AsyncMock(return_value=mock_role),
         ):
             permission_checker = require_permission(Permission.DELETE_ORGANIZATION)
-            result = await permission_checker(org_id=org_id, user_id=user_id)
+            result = await permission_checker(
+                request=mock_request, org_id=org_id, user_id=user_id
+            )
             assert result == user_id
 
     @pytest.mark.asyncio
@@ -559,6 +582,7 @@ class TestRequirePermission:
         """
         user_id = str(uuid4())
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         mock_role = MagicMock()
         mock_role.name = 'admin'
@@ -569,7 +593,9 @@ class TestRequirePermission:
         ):
             permission_checker = require_permission(Permission.DELETE_ORGANIZATION)
             with pytest.raises(HTTPException) as exc_info:
-                await permission_checker(org_id=org_id, user_id=user_id)
+                await permission_checker(
+                    request=mock_request, org_id=org_id, user_id=user_id
+                )
 
             assert exc_info.value.status_code == 403
 
@@ -582,6 +608,7 @@ class TestRequirePermission:
         """
         user_id = str(uuid4())
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         mock_role = MagicMock()
         mock_role.name = 'member'
@@ -595,7 +622,9 @@ class TestRequirePermission:
         ):
             permission_checker = require_permission(Permission.DELETE_ORGANIZATION)
             with pytest.raises(HTTPException):
-                await permission_checker(org_id=org_id, user_id=user_id)
+                await permission_checker(
+                    request=mock_request, org_id=org_id, user_id=user_id
+                )
 
             mock_logger.warning.assert_called()
             call_args = mock_logger.warning.call_args
@@ -611,6 +640,7 @@ class TestRequirePermission:
         THEN: User ID is returned
         """
         user_id = str(uuid4())
+        mock_request = _create_mock_request()
 
         mock_role = MagicMock()
         mock_role.name = 'admin'
@@ -620,7 +650,9 @@ class TestRequirePermission:
             AsyncMock(return_value=mock_role),
         ) as mock_get_role:
             permission_checker = require_permission(Permission.VIEW_LLM_SETTINGS)
-            result = await permission_checker(org_id=None, user_id=user_id)
+            result = await permission_checker(
+                request=mock_request, org_id=None, user_id=user_id
+            )
             assert result == user_id
             mock_get_role.assert_called_once_with(user_id, None)
 
@@ -632,6 +664,7 @@ class TestRequirePermission:
         THEN: HTTPException with 403 status is raised
         """
         user_id = str(uuid4())
+        mock_request = _create_mock_request()
 
         with patch(
             'server.auth.authorization.get_user_org_role',
@@ -639,7 +672,9 @@ class TestRequirePermission:
         ):
             permission_checker = require_permission(Permission.VIEW_LLM_SETTINGS)
             with pytest.raises(HTTPException) as exc_info:
-                await permission_checker(org_id=None, user_id=user_id)
+                await permission_checker(
+                    request=mock_request, org_id=None, user_id=user_id
+                )
 
             assert exc_info.value.status_code == 403
             assert 'not a member' in exc_info.value.detail
@@ -662,6 +697,7 @@ class TestPermissionScenarios:
         """
         user_id = str(uuid4())
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         mock_role = MagicMock()
         mock_role.name = 'member'
@@ -671,7 +707,9 @@ class TestPermissionScenarios:
             AsyncMock(return_value=mock_role),
         ):
             permission_checker = require_permission(Permission.MANAGE_SECRETS)
-            result = await permission_checker(org_id=org_id, user_id=user_id)
+            result = await permission_checker(
+                request=mock_request, org_id=org_id, user_id=user_id
+            )
             assert result == user_id
 
     @pytest.mark.asyncio
@@ -683,6 +721,7 @@ class TestPermissionScenarios:
         """
         user_id = str(uuid4())
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         mock_role = MagicMock()
         mock_role.name = 'member'
@@ -695,7 +734,9 @@ class TestPermissionScenarios:
                 Permission.INVITE_USER_TO_ORGANIZATION
             )
             with pytest.raises(HTTPException) as exc_info:
-                await permission_checker(org_id=org_id, user_id=user_id)
+                await permission_checker(
+                    request=mock_request, org_id=org_id, user_id=user_id
+                )
 
             assert exc_info.value.status_code == 403
 
@@ -708,6 +749,7 @@ class TestPermissionScenarios:
         """
         user_id = str(uuid4())
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         mock_role = MagicMock()
         mock_role.name = 'admin'
@@ -719,7 +761,9 @@ class TestPermissionScenarios:
             permission_checker = require_permission(
                 Permission.INVITE_USER_TO_ORGANIZATION
             )
-            result = await permission_checker(org_id=org_id, user_id=user_id)
+            result = await permission_checker(
+                request=mock_request, org_id=org_id, user_id=user_id
+            )
             assert result == user_id
 
     @pytest.mark.asyncio
@@ -731,6 +775,7 @@ class TestPermissionScenarios:
         """
         user_id = str(uuid4())
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         mock_role = MagicMock()
         mock_role.name = 'admin'
@@ -741,7 +786,9 @@ class TestPermissionScenarios:
         ):
             permission_checker = require_permission(Permission.CHANGE_USER_ROLE_OWNER)
             with pytest.raises(HTTPException) as exc_info:
-                await permission_checker(org_id=org_id, user_id=user_id)
+                await permission_checker(
+                    request=mock_request, org_id=org_id, user_id=user_id
+                )
 
             assert exc_info.value.status_code == 403
 
@@ -754,6 +801,7 @@ class TestPermissionScenarios:
         """
         user_id = str(uuid4())
         org_id = uuid4()
+        mock_request = _create_mock_request()
 
         mock_role = MagicMock()
         mock_role.name = 'owner'
@@ -763,5 +811,200 @@ class TestPermissionScenarios:
             AsyncMock(return_value=mock_role),
         ):
             permission_checker = require_permission(Permission.CHANGE_USER_ROLE_OWNER)
-            result = await permission_checker(org_id=org_id, user_id=user_id)
+            result = await permission_checker(
+                request=mock_request, org_id=org_id, user_id=user_id
+            )
             assert result == user_id
+
+
+# =============================================================================
+# Tests for API key organization validation
+# =============================================================================
+
+
+class TestApiKeyOrgValidation:
+    """Tests for API key organization binding validation in require_permission."""
+
+    @pytest.mark.asyncio
+    async def test_allows_access_when_api_key_org_matches_target_org(self):
+        """
+        GIVEN: API key with org_id that matches the target org_id in the request
+        WHEN: Permission checker is called
+        THEN: User ID is returned (access allowed)
+        """
+        # Arrange
+        user_id = str(uuid4())
+        org_id = uuid4()
+        mock_request = _create_mock_request(api_key_org_id=org_id)
+
+        mock_role = MagicMock()
+        mock_role.name = 'admin'
+
+        # Act & Assert
+        with patch(
+            'server.auth.authorization.get_user_org_role',
+            AsyncMock(return_value=mock_role),
+        ):
+            permission_checker = require_permission(Permission.VIEW_LLM_SETTINGS)
+            result = await permission_checker(
+                request=mock_request, org_id=org_id, user_id=user_id
+            )
+            assert result == user_id
+
+    @pytest.mark.asyncio
+    async def test_denies_access_when_api_key_org_mismatches_target_org(self):
+        """
+        GIVEN: API key created for Org A, but user tries to access Org B
+        WHEN: Permission checker is called
+        THEN: 403 Forbidden is raised with org mismatch message
+        """
+        # Arrange
+        user_id = str(uuid4())
+        api_key_org_id = uuid4()  # Org A - where API key was created
+        target_org_id = uuid4()  # Org B - where user is trying to access
+        mock_request = _create_mock_request(api_key_org_id=api_key_org_id)
+
+        # Act & Assert
+        permission_checker = require_permission(Permission.VIEW_LLM_SETTINGS)
+        with pytest.raises(HTTPException) as exc_info:
+            await permission_checker(
+                request=mock_request, org_id=target_org_id, user_id=user_id
+            )
+
+        assert exc_info.value.status_code == 403
+        assert (
+            'API key is not authorized for this organization' in exc_info.value.detail
+        )
+
+    @pytest.mark.asyncio
+    async def test_allows_access_for_legacy_api_key_without_org_binding(self):
+        """
+        GIVEN: Legacy API key without org_id binding (org_id is None)
+        WHEN: Permission checker is called
+        THEN: Falls through to normal permission check (backward compatible)
+        """
+        # Arrange
+        user_id = str(uuid4())
+        org_id = uuid4()
+        mock_request = _create_mock_request(api_key_org_id=None)
+
+        mock_role = MagicMock()
+        mock_role.name = 'admin'
+
+        # Act & Assert
+        with patch(
+            'server.auth.authorization.get_user_org_role',
+            AsyncMock(return_value=mock_role),
+        ):
+            permission_checker = require_permission(Permission.VIEW_LLM_SETTINGS)
+            result = await permission_checker(
+                request=mock_request, org_id=org_id, user_id=user_id
+            )
+            assert result == user_id
+
+    @pytest.mark.asyncio
+    async def test_allows_access_for_cookie_auth_without_api_key_org_id(self):
+        """
+        GIVEN: Cookie-based authentication (no api_key_org_id in user_auth)
+        WHEN: Permission checker is called
+        THEN: Falls through to normal permission check
+        """
+        # Arrange
+        user_id = str(uuid4())
+        org_id = uuid4()
+        mock_request = _create_mock_request(api_key_org_id=None)
+
+        mock_role = MagicMock()
+        mock_role.name = 'admin'
+
+        # Act & Assert
+        with patch(
+            'server.auth.authorization.get_user_org_role',
+            AsyncMock(return_value=mock_role),
+        ):
+            permission_checker = require_permission(Permission.VIEW_LLM_SETTINGS)
+            result = await permission_checker(
+                request=mock_request, org_id=org_id, user_id=user_id
+            )
+            assert result == user_id
+
+    @pytest.mark.asyncio
+    async def test_logs_warning_on_api_key_org_mismatch(self):
+        """
+        GIVEN: API key org_id doesn't match target org_id
+        WHEN: Permission checker is called
+        THEN: Warning is logged with org mismatch details
+        """
+        # Arrange
+        user_id = str(uuid4())
+        api_key_org_id = uuid4()
+        target_org_id = uuid4()
+        mock_request = _create_mock_request(api_key_org_id=api_key_org_id)
+
+        # Act & Assert
+        with patch('server.auth.authorization.logger') as mock_logger:
+            permission_checker = require_permission(Permission.VIEW_LLM_SETTINGS)
+            with pytest.raises(HTTPException):
+                await permission_checker(
+                    request=mock_request, org_id=target_org_id, user_id=user_id
+                )
+
+            mock_logger.warning.assert_called()
+            call_args = mock_logger.warning.call_args
+            assert call_args[1]['extra']['user_id'] == user_id
+            assert call_args[1]['extra']['api_key_org_id'] == str(api_key_org_id)
+            assert call_args[1]['extra']['target_org_id'] == str(target_org_id)
+
+
+class TestGetApiKeyOrgIdFromRequest:
+    """Tests for get_api_key_org_id_from_request helper function."""
+
+    @pytest.mark.asyncio
+    async def test_returns_org_id_when_user_auth_has_api_key_org_id(self):
+        """
+        GIVEN: Request with user_auth that has api_key_org_id
+        WHEN: get_api_key_org_id_from_request is called
+        THEN: Returns the api_key_org_id
+        """
+        # Arrange
+        org_id = uuid4()
+        mock_request = _create_mock_request(api_key_org_id=org_id)
+
+        # Act
+        result = await get_api_key_org_id_from_request(mock_request)
+
+        # Assert
+        assert result == org_id
+
+    @pytest.mark.asyncio
+    async def test_returns_none_when_user_auth_has_no_api_key_org_id(self):
+        """
+        GIVEN: Request with user_auth that has no api_key_org_id (cookie auth)
+        WHEN: get_api_key_org_id_from_request is called
+        THEN: Returns None
+        """
+        # Arrange
+        mock_request = _create_mock_request(api_key_org_id=None)
+
+        # Act
+        result = await get_api_key_org_id_from_request(mock_request)
+
+        # Assert
+        assert result is None
+
+    @pytest.mark.asyncio
+    async def test_returns_none_when_no_user_auth_in_request(self):
+        """
+        GIVEN: Request without user_auth in state
+        WHEN: get_api_key_org_id_from_request is called
+        THEN: Returns None
+        """
+        # Arrange
+        mock_request = MagicMock()
+        mock_request.state.user_auth = None
+
+        # Act
+        result = await get_api_key_org_id_from_request(mock_request)
+
+        # Assert
+        assert result is None

enterprise/tests/unit/test_billing.py

@@ -48,7 +48,7 @@ def mock_checkout_request():
             'server': ('test.com', 80),
         }
     )
-    request._base_url = URL('http://test.com/')
+    request._url = URL('http://test.com/')
     return request
 
 
@@ -62,7 +62,7 @@ def mock_subscription_request():
             'server': ('test.com', 80),
         }
     )
-    request._base_url = URL('http://test.com/')
+    request._url = URL('http://test.com/')
     return request
 
 
@@ -75,7 +75,6 @@ async def test_org(async_session_maker):
             id=org_id,
             name=f'test-org-{org_id}',
             org_version=ORG_SETTINGS_VERSION,
-            enable_default_condenser=True,
             enable_proactive_conversation_starters=True,
         )
         session.add(org)
@@ -264,7 +263,7 @@ async def test_create_checkout_session_success(
 async def test_success_callback_session_not_found(async_session_maker):
     """Test success callback when billing session is not found."""
     mock_request = Request(scope={'type': 'http'})
-    mock_request._base_url = URL('http://test.com/')
+    mock_request._url = URL('http://test.com/')
 
     with (
         patch('server.routes.billing.a_session_maker', async_session_maker),
@@ -281,7 +280,7 @@ async def test_success_callback_stripe_incomplete(
 ):
     """Test success callback when Stripe session is not complete."""
     mock_request = Request(scope={'type': 'http'})
-    mock_request._base_url = URL('http://test.com/')
+    mock_request._url = URL('http://test.com/')
 
     session_id = 'test_incomplete_session'
     async with async_session_maker() as session:
@@ -319,7 +318,7 @@ async def test_success_callback_stripe_incomplete(
 async def test_success_callback_success(async_session_maker, test_org, test_user):
     """Test successful payment completion and credit update."""
     mock_request = Request(scope={'type': 'http'})
-    mock_request._base_url = URL('http://test.com/')
+    mock_request._url = URL('http://test.com/')
 
     session_id = 'test_success_session'
     async with async_session_maker() as session:
@@ -391,7 +390,7 @@ async def test_success_callback_lite_llm_error(
 ):
     """Test handling of LiteLLM API errors during success callback."""
     mock_request = Request(scope={'type': 'http'})
-    mock_request._base_url = URL('http://test.com/')
+    mock_request._url = URL('http://test.com/')
 
     session_id = 'test_litellm_error_session'
     async with async_session_maker() as session:
@@ -445,7 +444,7 @@ async def test_success_callback_lite_llm_update_budget_error_rollback(
     the database transaction rolls back.
     """
     mock_request = Request(scope={'type': 'http'})
-    mock_request._base_url = URL('http://test.com/')
+    mock_request._url = URL('http://test.com/')
 
     session_id = 'test_budget_rollback_session'
     async with async_session_maker() as session:
@@ -502,7 +501,7 @@ async def test_success_callback_lite_llm_update_budget_error_rollback(
 async def test_cancel_callback_session_not_found(async_session_maker):
     """Test cancel callback when billing session is not found."""
     mock_request = Request(scope={'type': 'http'})
-    mock_request._base_url = URL('http://test.com/')
+    mock_request._url = URL('http://test.com/')
 
     with patch('server.routes.billing.a_session_maker', async_session_maker):
         response = await cancel_callback('nonexistent_session_id', mock_request)
@@ -517,7 +516,7 @@ async def test_cancel_callback_session_not_found(async_session_maker):
 async def test_cancel_callback_success(async_session_maker, test_org, test_user):
     """Test successful cancellation of billing session."""
     mock_request = Request(scope={'type': 'http'})
-    mock_request._base_url = URL('http://test.com/')
+    mock_request._url = URL('http://test.com/')
 
     session_id = 'test_cancel_session'
     async with async_session_maker() as session:
@@ -588,7 +587,7 @@ async def test_create_customer_setup_session_success():
             'headers': [],
         }
     )
-    mock_request._base_url = URL('http://test.com/')
+    mock_request._url = URL('http://test.com/')
 
     mock_customer_info = {'customer_id': 'mock-customer-id', 'org_id': 'mock-org-id'}
     mock_session = MagicMock()
@@ -613,6 +612,6 @@ async def test_create_customer_setup_session_success():
             customer='mock-customer-id',
             mode='setup',
             payment_method_types=['card'],
-            success_url='https://test.com/?setup=success',
-            cancel_url='https://test.com/',
+            success_url='https://test.com?setup=success',
+            cancel_url='https://test.com',
         )

enterprise/tests/unit/test_lite_llm_manager.py

@@ -2,7 +2,9 @@
 Unit tests for LiteLlmManager class.
 """
 
+import importlib
 import os
+import sys
 from unittest.mock import AsyncMock, MagicMock, patch
 
 import httpx
@@ -21,6 +23,107 @@ from storage.user_settings import UserSettings
 from openhands.server.settings import Settings
 
 
+def _agent_value(settings: Settings, key: str):
+    return settings.get_agent_setting(key)
+
+
+def _secret_value(settings: Settings, key: str):
+    secret = settings.get_secret_agent_setting(key)
+    return secret.get_secret_value() if secret else None
+
+
+class TestDefaultInitialBudget:
+    """Test cases for DEFAULT_INITIAL_BUDGET configuration."""
+
+    @pytest.fixture(autouse=True)
+    def restore_module_state(self):
+        """Ensure module is properly restored after each test."""
+        # Save original module if it exists
+        original_module = sys.modules.get('storage.lite_llm_manager')
+
+        yield
+
+        # Restore module state after each test
+        if 'storage.lite_llm_manager' in sys.modules:
+            del sys.modules['storage.lite_llm_manager']
+
+        # Clear the env vars
+        os.environ.pop('DEFAULT_INITIAL_BUDGET', None)
+        os.environ.pop('ENABLE_BILLING', None)
+
+        # Restore original module or reimport fresh
+        if original_module is not None:
+            sys.modules['storage.lite_llm_manager'] = original_module
+        else:
+            importlib.import_module('storage.lite_llm_manager')
+
+    def test_default_initial_budget_none_when_billing_disabled(self):
+        """Test that DEFAULT_INITIAL_BUDGET is None when billing is disabled."""
+        # Temporarily remove the module so we can reimport with different env vars
+        if 'storage.lite_llm_manager' in sys.modules:
+            del sys.modules['storage.lite_llm_manager']
+
+        # Ensure billing is disabled (default) and reimport
+        os.environ.pop('ENABLE_BILLING', None)
+        os.environ.pop('DEFAULT_INITIAL_BUDGET', None)
+        module = importlib.import_module('storage.lite_llm_manager')
+        assert module.DEFAULT_INITIAL_BUDGET is None
+
+    def test_default_initial_budget_defaults_to_zero_when_billing_enabled(self):
+        """Test that DEFAULT_INITIAL_BUDGET defaults to 0.0 when billing is enabled."""
+        # Temporarily remove the module so we can reimport with different env vars
+        if 'storage.lite_llm_manager' in sys.modules:
+            del sys.modules['storage.lite_llm_manager']
+
+        # Enable billing and reimport
+        os.environ['ENABLE_BILLING'] = 'true'
+        os.environ.pop('DEFAULT_INITIAL_BUDGET', None)
+        module = importlib.import_module('storage.lite_llm_manager')
+        assert module.DEFAULT_INITIAL_BUDGET == 0.0
+
+    def test_default_initial_budget_uses_env_var_when_billing_enabled(self):
+        """Test that DEFAULT_INITIAL_BUDGET uses value from environment variable when billing enabled."""
+        if 'storage.lite_llm_manager' in sys.modules:
+            del sys.modules['storage.lite_llm_manager']
+
+        os.environ['ENABLE_BILLING'] = 'true'
+        os.environ['DEFAULT_INITIAL_BUDGET'] = '100.0'
+        module = importlib.import_module('storage.lite_llm_manager')
+        assert module.DEFAULT_INITIAL_BUDGET == 100.0
+
+    def test_default_initial_budget_ignores_env_var_when_billing_disabled(self):
+        """Test that DEFAULT_INITIAL_BUDGET returns None when billing disabled, ignoring env var."""
+        if 'storage.lite_llm_manager' in sys.modules:
+            del sys.modules['storage.lite_llm_manager']
+
+        os.environ.pop('ENABLE_BILLING', None)  # billing disabled by default
+        os.environ['DEFAULT_INITIAL_BUDGET'] = '100.0'
+        module = importlib.import_module('storage.lite_llm_manager')
+        assert module.DEFAULT_INITIAL_BUDGET is None
+
+    def test_default_initial_budget_rejects_invalid_value(self):
+        """Test that DEFAULT_INITIAL_BUDGET raises ValueError for invalid values."""
+        if 'storage.lite_llm_manager' in sys.modules:
+            del sys.modules['storage.lite_llm_manager']
+
+        os.environ['ENABLE_BILLING'] = 'true'
+        os.environ['DEFAULT_INITIAL_BUDGET'] = 'abc'
+        with pytest.raises(ValueError) as exc_info:
+            importlib.import_module('storage.lite_llm_manager')
+        assert 'Invalid DEFAULT_INITIAL_BUDGET' in str(exc_info.value)
+
+    def test_default_initial_budget_rejects_negative_value(self):
+        """Test that DEFAULT_INITIAL_BUDGET raises ValueError for negative values."""
+        if 'storage.lite_llm_manager' in sys.modules:
+            del sys.modules['storage.lite_llm_manager']
+
+        os.environ['ENABLE_BILLING'] = 'true'
+        os.environ['DEFAULT_INITIAL_BUDGET'] = '-10.0'
+        with pytest.raises(ValueError) as exc_info:
+            importlib.import_module('storage.lite_llm_manager')
+        assert 'must be non-negative' in str(exc_info.value)
+
+
 class TestLiteLlmManager:
     """Test cases for LiteLlmManager class."""
 
@@ -28,20 +131,22 @@ class TestLiteLlmManager:
     def mock_settings(self):
         """Create a mock Settings object."""
         settings = Settings()
-        settings.agent = 'TestAgent'
-        settings.llm_model = 'test-model'
-        settings.llm_api_key = SecretStr('test-key')
-        settings.llm_base_url = 'http://test.com'
+        settings.set_agent_setting('agent', 'TestAgent')
+        settings.set_agent_setting('llm.model', 'test-model')
+        settings.set_agent_setting('llm.api_key', SecretStr('test-key'))
+        settings.set_agent_setting('llm.base_url', 'http://test.com')
         return settings
 
     @pytest.fixture
     def mock_user_settings(self):
         """Create a mock UserSettings object."""
         user_settings = UserSettings()
-        user_settings.agent = 'TestAgent'
-        user_settings.llm_model = 'test-model'
+        user_settings.agent_settings = {
+            'agent': 'TestAgent',
+            'llm.model': 'test-model',
+            'llm.base_url': 'http://test.com',
+        }
         user_settings.llm_api_key = SecretStr('test-key')
-        user_settings.llm_base_url = 'http://test.com'
         user_settings.user_version = 4  # Set version to avoid None comparison
         return user_settings
 
@@ -134,10 +239,14 @@ class TestLiteLlmManager:
                     )
 
                     assert result is not None
-                    assert result.agent == 'CodeActAgent'
-                    assert result.llm_model == get_default_litellm_model()
-                    assert result.llm_api_key.get_secret_value() == 'test-key'
-                    assert result.llm_base_url == 'http://test.com'
+                    assert _agent_value(result, 'agent') == 'CodeActAgent'
+                    assert _agent_value(
+                        result, 'llm.model'
+                    ) == get_default_litellm_model().replace(
+                        'litellm_proxy/', 'openhands/'
+                    )
+                    assert _secret_value(result, 'llm.api_key') == 'test-key'
+                    assert _agent_value(result, 'llm.base_url') == 'http://test.com'
 
     @pytest.mark.asyncio
     async def test_create_entries_cloud_deployment(self, mock_settings, mock_response):
@@ -145,6 +254,16 @@ class TestLiteLlmManager:
         mock_404_response = MagicMock()
         mock_404_response.status_code = 404
         mock_404_response.is_success = False
+        mock_404_response.raise_for_status.side_effect = httpx.HTTPStatusError(
+            message='Not Found', request=MagicMock(), response=mock_404_response
+        )
+
+        # Mock user exists check response
+        mock_user_exists_response = MagicMock()
+        mock_user_exists_response.is_success = True
+        mock_user_exists_response.json.return_value = {
+            'user_info': {'user_id': 'test-user-id'}
+        }
 
         mock_token_manager = MagicMock()
         mock_token_manager.return_value.get_user_info_from_user_id = AsyncMock(
@@ -152,12 +271,8 @@ class TestLiteLlmManager:
         )
 
         mock_client = AsyncMock()
-        mock_client.get.return_value = mock_404_response
-        mock_client.get.return_value.raise_for_status.side_effect = (
-            httpx.HTTPStatusError(
-                message='Not Found', request=MagicMock(), response=mock_404_response
-            )
-        )
+        # First GET is for _get_team (404), second GET is for _user_exists (success)
+        mock_client.get.side_effect = [mock_404_response, mock_user_exists_response]
         mock_client.post.return_value = mock_response
 
         mock_client_class = MagicMock()
@@ -175,13 +290,15 @@ class TestLiteLlmManager:
             )
 
             assert result is not None
-            assert result.agent == 'CodeActAgent'
-            assert result.llm_model == get_default_litellm_model()
-            assert result.llm_api_key.get_secret_value() == 'test-api-key'
-            assert result.llm_base_url == 'http://test.com'
+            assert _agent_value(result, 'agent') == 'CodeActAgent'
+            assert _agent_value(
+                result, 'llm.model'
+            ) == get_default_litellm_model().replace('litellm_proxy/', 'openhands/')
+            assert _secret_value(result, 'llm.api_key') == 'test-api-key'
+            assert _agent_value(result, 'llm.base_url') == 'http://test.com'
 
-            # Verify API calls were made (get_team + 4 posts)
-            assert mock_client.get.call_count == 1  # get_team
+            # Verify API calls were made (get_team + user_exists + 4 posts)
+            assert mock_client.get.call_count == 2  # get_team + user_exists
             assert (
                 mock_client.post.call_count == 4
             )  # create_team, add_user_to_team, delete_key_by_alias, generate_key
@@ -200,13 +317,21 @@ class TestLiteLlmManager:
         }
         mock_team_response.raise_for_status = MagicMock()
 
+        # Mock user exists check response
+        mock_user_exists_response = MagicMock()
+        mock_user_exists_response.is_success = True
+        mock_user_exists_response.json.return_value = {
+            'user_info': {'user_id': 'test-user-id'}
+        }
+
         mock_token_manager = MagicMock()
         mock_token_manager.return_value.get_user_info_from_user_id = AsyncMock(
             return_value={'email': 'test@example.com'}
         )
 
         mock_client = AsyncMock()
-        mock_client.get.return_value = mock_team_response
+        # First GET is for _get_team (success), second GET is for _user_exists (success)
+        mock_client.get.side_effect = [mock_team_response, mock_user_exists_response]
         mock_client.post.return_value = mock_response
 
         mock_client_class = MagicMock()
@@ -226,8 +351,8 @@ class TestLiteLlmManager:
             assert result is not None
 
             # Verify _get_team was called first
-            mock_client.get.assert_called_once()
-            get_call_url = mock_client.get.call_args[0][0]
+            assert mock_client.get.call_count == 2  # get_team + user_exists
+            get_call_url = mock_client.get.call_args_list[0][0][0]
             assert 'team/info' in get_call_url
             assert 'test-org-id' in get_call_url
 
@@ -242,26 +367,32 @@ class TestLiteLlmManager:
             assert add_user_call[1]['json']['max_budget_in_team'] == 30.0
 
     @pytest.mark.asyncio
-    async def test_create_entries_new_org_uses_zero_budget(
+    async def test_create_entries_new_org_uses_default_initial_budget(
         self, mock_settings, mock_response
     ):
-        """Test that create_entries uses budget=0 for new org (team doesn't exist)."""
+        """Test that create_entries uses DEFAULT_INITIAL_BUDGET for new org."""
         mock_404_response = MagicMock()
         mock_404_response.status_code = 404
         mock_404_response.is_success = False
+        mock_404_response.raise_for_status.side_effect = httpx.HTTPStatusError(
+            message='Not Found', request=MagicMock(), response=mock_404_response
+        )
 
         mock_token_manager = MagicMock()
         mock_token_manager.return_value.get_user_info_from_user_id = AsyncMock(
             return_value={'email': 'test@example.com'}
         )
 
+        # Mock user exists check response
+        mock_user_exists_response = MagicMock()
+        mock_user_exists_response.is_success = True
+        mock_user_exists_response.json.return_value = {
+            'user_info': {'user_id': 'test-user-id'}
+        }
+
         mock_client = AsyncMock()
-        mock_client.get.return_value = mock_404_response
-        mock_client.get.return_value.raise_for_status.side_effect = (
-            httpx.HTTPStatusError(
-                message='Not Found', request=MagicMock(), response=mock_404_response
-            )
-        )
+        # First GET is for _get_team (404), second GET is for _user_exists (success)
+        mock_client.get.side_effect = [mock_404_response, mock_user_exists_response]
         mock_client.post.return_value = mock_response
 
         mock_client_class = MagicMock()
@@ -273,6 +404,7 @@ class TestLiteLlmManager:
             patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com'),
             patch('storage.lite_llm_manager.TokenManager', mock_token_manager),
             patch('httpx.AsyncClient', mock_client_class),
+            patch('storage.lite_llm_manager.DEFAULT_INITIAL_BUDGET', 0.0),
         ):
             result = await LiteLlmManager.create_entries(
                 'test-org-id', 'test-user-id', mock_settings, create_user=False
@@ -280,16 +412,73 @@ class TestLiteLlmManager:
 
             assert result is not None
 
-            # Verify _create_team was called with budget=0
+            # Verify _create_team was called with DEFAULT_INITIAL_BUDGET (0.0)
             create_team_call = mock_client.post.call_args_list[0]
             assert 'team/new' in create_team_call[0][0]
             assert create_team_call[1]['json']['max_budget'] == 0.0
 
-            # Verify _add_user_to_team was called with budget=0
+            # Verify _add_user_to_team was called with DEFAULT_INITIAL_BUDGET (0.0)
             add_user_call = mock_client.post.call_args_list[1]
             assert 'team/member_add' in add_user_call[0][0]
             assert add_user_call[1]['json']['max_budget_in_team'] == 0.0
 
+    @pytest.mark.asyncio
+    async def test_create_entries_new_org_uses_custom_default_budget(
+        self, mock_settings, mock_response
+    ):
+        """Test that create_entries uses custom DEFAULT_INITIAL_BUDGET for new org."""
+        mock_404_response = MagicMock()
+        mock_404_response.status_code = 404
+        mock_404_response.is_success = False
+        mock_404_response.raise_for_status.side_effect = httpx.HTTPStatusError(
+            message='Not Found', request=MagicMock(), response=mock_404_response
+        )
+
+        # Mock user exists check response
+        mock_user_exists_response = MagicMock()
+        mock_user_exists_response.is_success = True
+        mock_user_exists_response.json.return_value = {
+            'user_info': {'user_id': 'test-user-id'}
+        }
+
+        mock_token_manager = MagicMock()
+        mock_token_manager.return_value.get_user_info_from_user_id = AsyncMock(
+            return_value={'email': 'test@example.com'}
+        )
+
+        mock_client = AsyncMock()
+        # First GET is for _get_team (404), second GET is for _user_exists (success)
+        mock_client.get.side_effect = [mock_404_response, mock_user_exists_response]
+        mock_client.post.return_value = mock_response
+
+        mock_client_class = MagicMock()
+        mock_client_class.return_value.__aenter__.return_value = mock_client
+
+        custom_budget = 50.0
+        with (
+            patch.dict(os.environ, {'LOCAL_DEPLOYMENT': ''}),
+            patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-key'),
+            patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com'),
+            patch('storage.lite_llm_manager.TokenManager', mock_token_manager),
+            patch('httpx.AsyncClient', mock_client_class),
+            patch('storage.lite_llm_manager.DEFAULT_INITIAL_BUDGET', custom_budget),
+        ):
+            result = await LiteLlmManager.create_entries(
+                'test-org-id', 'test-user-id', mock_settings, create_user=False
+            )
+
+            assert result is not None
+
+            # Verify _create_team was called with custom DEFAULT_INITIAL_BUDGET
+            create_team_call = mock_client.post.call_args_list[0]
+            assert 'team/new' in create_team_call[0][0]
+            assert create_team_call[1]['json']['max_budget'] == custom_budget
+
+            # Verify _add_user_to_team was called with custom DEFAULT_INITIAL_BUDGET
+            add_user_call = mock_client.post.call_args_list[1]
+            assert 'team/member_add' in add_user_call[0][0]
+            assert add_user_call[1]['json']['max_budget_in_team'] == custom_budget
+
     @pytest.mark.asyncio
     async def test_create_entries_propagates_non_404_errors(self, mock_settings):
         """Test that create_entries propagates non-404 errors from _get_team."""
@@ -358,10 +547,14 @@ class TestLiteLlmManager:
 
                     # migrate_entries returns the user_settings unchanged
                     assert result is not None
-                    assert result.agent == 'TestAgent'
-                    assert result.llm_model == 'test-model'
+                    effective_settings = result.to_settings()
+                    assert _agent_value(effective_settings, 'agent') == 'TestAgent'
+                    assert _agent_value(effective_settings, 'llm.model') == 'test-model'
                     assert result.llm_api_key.get_secret_value() == 'test-key'
-                    assert result.llm_base_url == 'http://test.com'
+                    assert (
+                        _agent_value(effective_settings, 'llm.base_url')
+                        == 'http://test.com'
+                    )
 
     @pytest.mark.asyncio
     async def test_migrate_entries_no_user_found(self, mock_user_settings):
@@ -482,10 +675,19 @@ class TestLiteLlmManager:
 
                             # migrate_entries returns the user_settings unchanged
                             assert result is not None
-                            assert result.agent == 'TestAgent'
-                            assert result.llm_model == 'test-model'
+                            effective_settings = result.to_settings()
+                            assert (
+                                _agent_value(effective_settings, 'agent') == 'TestAgent'
+                            )
+                            assert (
+                                _agent_value(effective_settings, 'llm.model')
+                                == 'test-model'
+                            )
                             assert result.llm_api_key.get_secret_value() == 'test-key'
-                            assert result.llm_base_url == 'http://test.com'
+                            assert (
+                                _agent_value(effective_settings, 'llm.base_url')
+                                == 'http://test.com'
+                            )
 
                             # Verify migration steps were called:
                             # - 2 GET requests: _get_user, _get_user_keys
@@ -567,8 +769,9 @@ class TestLiteLlmManager:
                                 result.llm_api_key.get_secret_value()
                                 == 'new-generated-key'
                             )
+                            assert result.llm_api_key_for_byor_secret is not None
                             assert (
-                                result.llm_api_key_for_byor.get_secret_value()
+                                result.llm_api_key_for_byor_secret.get_secret_value()
                                 == 'new-generated-key'
                             )
 
@@ -687,15 +890,16 @@ class TestLiteLlmManager:
 
     @pytest.mark.asyncio
     async def test_create_user_success(self, mock_http_client, mock_response):
-        """Test successful _create_user operation."""
+        """Test successful _create_user operation returns True."""
         mock_http_client.post.return_value = mock_response
 
         with patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-key'):
             with patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com'):
-                await LiteLlmManager._create_user(
+                result = await LiteLlmManager._create_user(
                     mock_http_client, 'test@example.com', 'test-user-id'
                 )
 
+                assert result is True
                 mock_http_client.post.assert_called_once()
                 call_args = mock_http_client.post.call_args
                 assert 'http://test.com/user/new' in call_args[0]
@@ -704,7 +908,7 @@ class TestLiteLlmManager:
 
     @pytest.mark.asyncio
     async def test_create_user_duplicate_email(self, mock_http_client, mock_response):
-        """Test _create_user with duplicate email handling."""
+        """Test _create_user with duplicate email handling returns True."""
         # First call fails with duplicate email
         error_response = MagicMock()
         error_response.is_success = False
@@ -716,23 +920,81 @@ class TestLiteLlmManager:
 
         with patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-key'):
             with patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com'):
-                await LiteLlmManager._create_user(
+                result = await LiteLlmManager._create_user(
                     mock_http_client, 'test@example.com', 'test-user-id'
                 )
 
+                assert result is True
                 assert mock_http_client.post.call_count == 2
                 # Second call should have None email
                 second_call_args = mock_http_client.post.call_args_list[1]
                 assert second_call_args[1]['json']['user_email'] is None
 
+    @pytest.mark.asyncio
+    @patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com')
+    @patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-key')
+    async def test_user_exists_returns_true(self, mock_http_client):
+        """Test _user_exists returns True when user exists in LiteLLM."""
+        # Arrange
+        user_response = MagicMock()
+        user_response.is_success = True
+        user_response.json.return_value = {
+            'user_info': {'user_id': 'test-user-id', 'email': 'test@example.com'}
+        }
+        mock_http_client.get.return_value = user_response
+
+        # Act
+        result = await LiteLlmManager._user_exists(mock_http_client, 'test-user-id')
+
+        # Assert
+        assert result is True
+        mock_http_client.get.assert_called_once()
+
+    @pytest.mark.asyncio
+    @patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com')
+    @patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-key')
+    async def test_user_exists_returns_false_when_not_found(self, mock_http_client):
+        """Test _user_exists returns False when user not found."""
+        # Arrange
+        user_response = MagicMock()
+        user_response.is_success = False
+        mock_http_client.get.return_value = user_response
+
+        # Act
+        result = await LiteLlmManager._user_exists(mock_http_client, 'test-user-id')
+
+        # Assert
+        assert result is False
+
+    @pytest.mark.asyncio
+    @patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com')
+    @patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-key')
+    async def test_user_exists_returns_false_on_mismatched_user_id(
+        self, mock_http_client
+    ):
+        """Test _user_exists returns False when returned user_id doesn't match."""
+        # Arrange
+        user_response = MagicMock()
+        user_response.is_success = True
+        user_response.json.return_value = {
+            'user_info': {'user_id': 'different-user-id'}
+        }
+        mock_http_client.get.return_value = user_response
+
+        # Act
+        result = await LiteLlmManager._user_exists(mock_http_client, 'test-user-id')
+
+        # Assert
+        assert result is False
+
     @pytest.mark.asyncio
     @patch('storage.lite_llm_manager.logger')
     @patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com')
     @patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-key')
-    async def test_create_user_already_exists_with_409_status_code(
+    async def test_create_user_already_exists_and_verified(
         self, mock_logger, mock_http_client
     ):
-        """Test _create_user handles 409 Conflict when user already exists."""
+        """Test _create_user returns True when user already exists and is verified."""
         # Arrange
         first_response = MagicMock()
         first_response.is_success = False
@@ -744,14 +1006,141 @@ class TestLiteLlmManager:
         second_response.status_code = 409
         second_response.text = 'User with id test-user-id already exists'
 
+        user_exists_response = MagicMock()
+        user_exists_response.is_success = True
+        user_exists_response.json.return_value = {
+            'user_info': {'user_id': 'test-user-id'}
+        }
+
         mock_http_client.post.side_effect = [first_response, second_response]
+        mock_http_client.get.return_value = user_exists_response
 
         # Act
-        await LiteLlmManager._create_user(
+        result = await LiteLlmManager._create_user(
             mock_http_client, 'test@example.com', 'test-user-id'
         )
 
         # Assert
+        assert result is True
+        mock_logger.warning.assert_any_call(
+            'litellm_user_already_exists',
+            extra={'user_id': 'test-user-id'},
+        )
+
+    @pytest.mark.asyncio
+    @patch('storage.lite_llm_manager.logger')
+    @patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com')
+    @patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-key')
+    async def test_create_user_already_exists_but_not_found_returns_false(
+        self, mock_logger, mock_http_client
+    ):
+        """Test _create_user returns False when LiteLLM claims user exists but verification fails."""
+        # Arrange
+        first_response = MagicMock()
+        first_response.is_success = False
+        first_response.status_code = 400
+        first_response.text = 'duplicate email'
+
+        second_response = MagicMock()
+        second_response.is_success = False
+        second_response.status_code = 409
+        second_response.text = 'User with id test-user-id already exists'
+
+        user_not_exists_response = MagicMock()
+        user_not_exists_response.is_success = False
+
+        mock_http_client.post.side_effect = [first_response, second_response]
+        mock_http_client.get.return_value = user_not_exists_response
+
+        # Act
+        result = await LiteLlmManager._create_user(
+            mock_http_client, 'test@example.com', 'test-user-id'
+        )
+
+        # Assert
+        assert result is False
+        mock_logger.error.assert_any_call(
+            'litellm_user_claimed_exists_but_not_found',
+            extra={
+                'user_id': 'test-user-id',
+                'status_code': 409,
+                'text': 'User with id test-user-id already exists',
+            },
+        )
+
+    @pytest.mark.asyncio
+    @patch('storage.lite_llm_manager.logger')
+    @patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com')
+    @patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-key')
+    async def test_create_user_failure_returns_false(
+        self, mock_logger, mock_http_client
+    ):
+        """Test _create_user returns False when creation fails with non-'already exists' error."""
+        # Arrange
+        first_response = MagicMock()
+        first_response.is_success = False
+        first_response.status_code = 400
+        first_response.text = 'duplicate email'
+
+        second_response = MagicMock()
+        second_response.is_success = False
+        second_response.status_code = 500
+        second_response.text = 'Internal server error'
+
+        mock_http_client.post.side_effect = [first_response, second_response]
+
+        # Act
+        result = await LiteLlmManager._create_user(
+            mock_http_client, 'test@example.com', 'test-user-id'
+        )
+
+        # Assert
+        assert result is False
+        mock_logger.error.assert_any_call(
+            'error_creating_litellm_user',
+            extra={
+                'status_code': 500,
+                'text': 'Internal server error',
+                'user_id': 'test-user-id',
+                'email': None,
+            },
+        )
+
+    @pytest.mark.asyncio
+    @patch('storage.lite_llm_manager.logger')
+    @patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com')
+    @patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-key')
+    async def test_create_user_already_exists_with_409_status_code(
+        self, mock_logger, mock_http_client
+    ):
+        """Test _create_user handles 409 Conflict when user already exists and verifies."""
+        # Arrange
+        first_response = MagicMock()
+        first_response.is_success = False
+        first_response.status_code = 400
+        first_response.text = 'duplicate email'
+
+        second_response = MagicMock()
+        second_response.is_success = False
+        second_response.status_code = 409
+        second_response.text = 'User with id test-user-id already exists'
+
+        user_exists_response = MagicMock()
+        user_exists_response.is_success = True
+        user_exists_response.json.return_value = {
+            'user_info': {'user_id': 'test-user-id'}
+        }
+
+        mock_http_client.post.side_effect = [first_response, second_response]
+        mock_http_client.get.return_value = user_exists_response
+
+        # Act
+        result = await LiteLlmManager._create_user(
+            mock_http_client, 'test@example.com', 'test-user-id'
+        )
+
+        # Assert
+        assert result is True
         mock_logger.warning.assert_any_call(
             'litellm_user_already_exists',
             extra={'user_id': 'test-user-id'},
@@ -764,7 +1153,7 @@ class TestLiteLlmManager:
     async def test_create_user_already_exists_with_400_status_code(
         self, mock_logger, mock_http_client
     ):
-        """Test _create_user handles 400 Bad Request when user already exists."""
+        """Test _create_user handles 400 Bad Request when user already exists and verifies."""
         # Arrange
         first_response = MagicMock()
         first_response.is_success = False
@@ -776,14 +1165,22 @@ class TestLiteLlmManager:
         second_response.status_code = 400
         second_response.text = 'User already exists'
 
+        user_exists_response = MagicMock()
+        user_exists_response.is_success = True
+        user_exists_response.json.return_value = {
+            'user_info': {'user_id': 'test-user-id'}
+        }
+
         mock_http_client.post.side_effect = [first_response, second_response]
+        mock_http_client.get.return_value = user_exists_response
 
         # Act
-        await LiteLlmManager._create_user(
+        result = await LiteLlmManager._create_user(
             mock_http_client, 'test@example.com', 'test-user-id'
         )
 
         # Assert
+        assert result is True
         mock_logger.warning.assert_any_call(
             'litellm_user_already_exists',
             extra={'user_id': 'test-user-id'},
@@ -1665,7 +2062,10 @@ class TestLiteLlmManager:
 
                             # downgrade_entries returns the user_settings
                             assert result is not None
-                            assert result.agent == 'TestAgent'
+                            assert (
+                                _agent_value(result.to_settings(), 'agent')
+                                == 'TestAgent'
+                            )
 
                             # Verify downgrade steps were called:
                             # GET requests:
@@ -1699,7 +2099,7 @@ class TestLiteLlmManager:
                     # In local deployment, should return user_settings without
                     # making any LiteLLM calls
                     assert result is not None
-                    assert result.agent == 'TestAgent'
+                    assert _agent_value(result.to_settings(), 'agent') == 'TestAgent'
 
 
 class TestGetAllKeysForUser:
@@ -2018,3 +2418,195 @@ class TestVerifyExistingKey:
                 openhands_type=True,
             )
             assert result is False
+
+
+class TestBudgetPayloadHandling:
+    """Test cases for budget field handling in API payloads.
+
+    These tests verify that when max_budget is None, the budget field is NOT
+    included in the JSON payload (which tells LiteLLM to disable budget
+    enforcement), and when max_budget has a value, it IS included.
+    """
+
+    @pytest.mark.asyncio
+    async def test_create_team_excludes_max_budget_when_none(self):
+        """Test that _create_team does NOT include max_budget when it is None."""
+        mock_client = AsyncMock(spec=httpx.AsyncClient)
+        mock_response = MagicMock()
+        mock_response.is_success = True
+        mock_response.status_code = 200
+        mock_client.post.return_value = mock_response
+
+        with patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-api-key'):
+            with patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com'):
+                await LiteLlmManager._create_team(
+                    mock_client,
+                    team_alias='test-team',
+                    team_id='test-team-id',
+                    max_budget=None,  # None = no budget limit
+                )
+
+        # Verify the call was made
+        mock_client.post.assert_called_once()
+        call_args = mock_client.post.call_args
+
+        # Verify URL
+        assert call_args[0][0] == 'http://test.com/team/new'
+
+        # Verify that max_budget is NOT in the JSON payload
+        json_payload = call_args[1]['json']
+        assert 'max_budget' not in json_payload, (
+            'max_budget should NOT be in payload when None '
+            '(omitting it tells LiteLLM to disable budget enforcement)'
+        )
+
+    @pytest.mark.asyncio
+    async def test_create_team_includes_max_budget_when_set(self):
+        """Test that _create_team includes max_budget when it has a value."""
+        mock_client = AsyncMock(spec=httpx.AsyncClient)
+        mock_response = MagicMock()
+        mock_response.is_success = True
+        mock_response.status_code = 200
+        mock_client.post.return_value = mock_response
+
+        with patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-api-key'):
+            with patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com'):
+                await LiteLlmManager._create_team(
+                    mock_client,
+                    team_alias='test-team',
+                    team_id='test-team-id',
+                    max_budget=100.0,  # Explicit budget limit
+                )
+
+        # Verify the call was made
+        mock_client.post.assert_called_once()
+        call_args = mock_client.post.call_args
+
+        # Verify that max_budget IS in the JSON payload with the correct value
+        json_payload = call_args[1]['json']
+        assert (
+            'max_budget' in json_payload
+        ), 'max_budget should be in payload when set to a value'
+        assert json_payload['max_budget'] == 100.0
+
+    @pytest.mark.asyncio
+    async def test_add_user_to_team_excludes_max_budget_when_none(self):
+        """Test that _add_user_to_team does NOT include max_budget_in_team when None."""
+        mock_client = AsyncMock(spec=httpx.AsyncClient)
+        mock_response = MagicMock()
+        mock_response.is_success = True
+        mock_response.status_code = 200
+        mock_client.post.return_value = mock_response
+
+        with patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-api-key'):
+            with patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com'):
+                await LiteLlmManager._add_user_to_team(
+                    mock_client,
+                    keycloak_user_id='test-user-id',
+                    team_id='test-team-id',
+                    max_budget=None,  # None = no budget limit
+                )
+
+        # Verify the call was made
+        mock_client.post.assert_called_once()
+        call_args = mock_client.post.call_args
+
+        # Verify URL
+        assert call_args[0][0] == 'http://test.com/team/member_add'
+
+        # Verify that max_budget_in_team is NOT in the JSON payload
+        json_payload = call_args[1]['json']
+        assert 'max_budget_in_team' not in json_payload, (
+            'max_budget_in_team should NOT be in payload when None '
+            '(omitting it tells LiteLLM to disable budget enforcement)'
+        )
+
+    @pytest.mark.asyncio
+    async def test_add_user_to_team_includes_max_budget_when_set(self):
+        """Test that _add_user_to_team includes max_budget_in_team when set."""
+        mock_client = AsyncMock(spec=httpx.AsyncClient)
+        mock_response = MagicMock()
+        mock_response.is_success = True
+        mock_response.status_code = 200
+        mock_client.post.return_value = mock_response
+
+        with patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-api-key'):
+            with patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com'):
+                await LiteLlmManager._add_user_to_team(
+                    mock_client,
+                    keycloak_user_id='test-user-id',
+                    team_id='test-team-id',
+                    max_budget=50.0,  # Explicit budget limit
+                )
+
+        # Verify the call was made
+        mock_client.post.assert_called_once()
+        call_args = mock_client.post.call_args
+
+        # Verify that max_budget_in_team IS in the JSON payload
+        json_payload = call_args[1]['json']
+        assert (
+            'max_budget_in_team' in json_payload
+        ), 'max_budget_in_team should be in payload when set to a value'
+        assert json_payload['max_budget_in_team'] == 50.0
+
+    @pytest.mark.asyncio
+    async def test_update_user_in_team_excludes_max_budget_when_none(self):
+        """Test that _update_user_in_team does NOT include max_budget_in_team when None."""
+        mock_client = AsyncMock(spec=httpx.AsyncClient)
+        mock_response = MagicMock()
+        mock_response.is_success = True
+        mock_response.status_code = 200
+        mock_client.post.return_value = mock_response
+
+        with patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-api-key'):
+            with patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com'):
+                await LiteLlmManager._update_user_in_team(
+                    mock_client,
+                    keycloak_user_id='test-user-id',
+                    team_id='test-team-id',
+                    max_budget=None,  # None = no budget limit
+                )
+
+        # Verify the call was made
+        mock_client.post.assert_called_once()
+        call_args = mock_client.post.call_args
+
+        # Verify URL
+        assert call_args[0][0] == 'http://test.com/team/member_update'
+
+        # Verify that max_budget_in_team is NOT in the JSON payload
+        json_payload = call_args[1]['json']
+        assert 'max_budget_in_team' not in json_payload, (
+            'max_budget_in_team should NOT be in payload when None '
+            '(omitting it tells LiteLLM to disable budget enforcement)'
+        )
+
+    @pytest.mark.asyncio
+    async def test_update_user_in_team_includes_max_budget_when_set(self):
+        """Test that _update_user_in_team includes max_budget_in_team when set."""
+        mock_client = AsyncMock(spec=httpx.AsyncClient)
+        mock_response = MagicMock()
+        mock_response.is_success = True
+        mock_response.status_code = 200
+        mock_client.post.return_value = mock_response
+
+        with patch('storage.lite_llm_manager.LITE_LLM_API_KEY', 'test-api-key'):
+            with patch('storage.lite_llm_manager.LITE_LLM_API_URL', 'http://test.com'):
+                await LiteLlmManager._update_user_in_team(
+                    mock_client,
+                    keycloak_user_id='test-user-id',
+                    team_id='test-team-id',
+                    max_budget=75.0,  # Explicit budget limit
+                )
+
+        # Verify the call was made
+        mock_client.post.assert_called_once()
+        call_args = mock_client.post.call_args
+
+        # Verify that max_budget_in_team IS in the JSON payload
+        json_payload = call_args[1]['json']
+        assert (
+            'max_budget_in_team' in json_payload
+        ), 'max_budget_in_team should be in payload when set to a value'
+        assert json_payload['max_budget_in_team'] == 75.0

enterprise/tests/unit/test_logger.py

@@ -5,10 +5,15 @@ from io import StringIO
 from unittest.mock import patch
 
 import pytest
+from freezegun import freeze_time
 from server.logger import format_stack, setup_json_logger
 
 from openhands.core.logger import openhands_logger
 
+FROZEN_TIMESTAMP = '2024-01-15T10:30:00+00:00'
+# datetime.now().isoformat() doesn't include timezone info
+FROZEN_TIMESTAMP_NO_TZ = '2024-01-15T10:30:00'
+
 
 @pytest.fixture
 def log_output():
@@ -21,20 +26,31 @@ def log_output():
 
 
 class TestLogOutput:
+    @freeze_time(FROZEN_TIMESTAMP)
     def test_info(self, log_output):
         logger, string_io = log_output
 
         logger.info('Test message')
         output = json.loads(string_io.getvalue())
-        assert output == {'message': 'Test message', 'severity': 'INFO'}
+        assert output == {
+            'message': 'Test message',
+            'severity': 'INFO',
+            'ts': FROZEN_TIMESTAMP,
+        }
 
+    @freeze_time(FROZEN_TIMESTAMP)
     def test_error(self, log_output):
         logger, string_io = log_output
 
         logger.error('Test message')
         output = json.loads(string_io.getvalue())
-        assert output == {'message': 'Test message', 'severity': 'ERROR'}
+        assert output == {
+            'message': 'Test message',
+            'severity': 'ERROR',
+            'ts': FROZEN_TIMESTAMP,
+        }
 
+    @freeze_time(FROZEN_TIMESTAMP)
     def test_extra_fields(self, log_output):
         logger, string_io = log_output
 
@@ -44,6 +60,7 @@ class TestLogOutput:
             'key': '..val..',
             'message': 'Test message',
             'severity': 'INFO',
+            'ts': FROZEN_TIMESTAMP,
         }
 
     def test_format_stack(self):
@@ -257,6 +274,7 @@ class TestLogOutput:
             ]
             assert formatted == expected
 
+    @freeze_time(FROZEN_TIMESTAMP)
     def test_filtering(self):
         # Ensure that secret values are still filtered
         string_io = StringIO()
@@ -266,4 +284,62 @@ class TestLogOutput:
         ):
             openhands_logger.info('The secret key was supersecretvalue')
         output = json.loads(string_io.getvalue())
-        assert output == {'message': 'The secret key was ******', 'severity': 'INFO'}
+        assert output == {
+            'message': 'The secret key was ******',
+            'severity': 'INFO',
+            'ts': FROZEN_TIMESTAMP,
+        }
+
+    @freeze_time(FROZEN_TIMESTAMP)
+    def test_console_serializer_uses_ts_not_timestamp(self):
+        """When LOG_JSON_FOR_CONSOLE=1, use 'ts' from custom_json_serializer, not 'timestamp'."""
+        import server.logger as logger_module
+
+        string_io = StringIO()
+        logger = logging.Logger('test_console')
+
+        # Patch LOG_JSON_FOR_CONSOLE to 1 for both setup_json_logger and custom_json_serializer
+        with patch.object(logger_module, 'LOG_JSON_FOR_CONSOLE', 1):
+            setup_json_logger(logger, 'INFO', _out=string_io)
+            logger.info('Test console message')
+
+        # Parse output - LOG_JSON_FOR_CONSOLE pretty-prints JSON across multiple lines
+        output = json.loads(string_io.getvalue())
+
+        # Should have 'ts' from custom_json_serializer but NOT 'timestamp'
+        assert 'ts' in output
+        assert 'timestamp' not in output
+        assert output['message'] == 'Test console message'
+        assert output['severity'] == 'INFO'
+
+    @freeze_time(FROZEN_TIMESTAMP)
+    def test_ts_not_duplicated_when_both_json_modes_enabled(self):
+        """When both LOG_JSON=1 and LOG_JSON_FOR_CONSOLE=1, 'ts' should appear only once."""
+        import server.logger as logger_module
+
+        string_io = StringIO()
+        logger = logging.Logger('test_both_modes')
+
+        # Patch both LOG_JSON and LOG_JSON_FOR_CONSOLE to 1
+        with (
+            patch.object(logger_module, 'LOG_JSON', True),
+            patch.object(logger_module, 'LOG_JSON_FOR_CONSOLE', 1),
+        ):
+            setup_json_logger(logger, 'INFO', _out=string_io)
+            logger.info('Test both modes message')
+
+        raw_output = string_io.getvalue()
+        output = json.loads(raw_output)
+
+        # Should have exactly one 'ts' field (not duplicated)
+        assert 'ts' in output
+        assert 'timestamp' not in output
+        # Verify 'ts' appears only once in the raw output (not duplicated as key)
+        assert (
+            raw_output.count('"ts"') == 1
+        ), f"'ts' should appear exactly once, found in: {raw_output}"
+        assert output['message'] == 'Test both modes message'
+        assert output['severity'] == 'INFO'
+        # When LOG_JSON_FOR_CONSOLE=1, custom_json_serializer uses datetime.now().isoformat()
+        # which doesn't include timezone info
+        assert output['ts'] == FROZEN_TIMESTAMP_NO_TZ

enterprise/tests/unit/test_org_invitation_service.py

@@ -10,6 +10,7 @@ from server.routes.org_invitation_models import (
 )
 from server.services.org_invitation_service import OrgInvitationService
 from storage.org_invitation import OrgInvitation
+from storage.org_store import OrgStore
 
 
 class TestAcceptInvitationEmailValidation:
@@ -98,6 +99,9 @@ class TestAcceptInvitationEmailValidation:
 
         mock_keycloak_user_info = {'email': 'alice@example.com'}  # Email from Keycloak
 
+        mock_org = MagicMock()
+        mock_org.agent_settings = {'schema_version': 1, 'llm.model': 'test-model'}
+
         with (
             patch(
                 'server.services.org_invitation_service.OrgInvitationStore.get_invitation_by_token',
@@ -121,6 +125,10 @@ class TestAcceptInvitationEmailValidation:
                 'server.services.org_invitation_service.OrgService.create_litellm_integration',
                 new_callable=AsyncMock,
             ) as mock_create_litellm,
+            patch(
+                'server.services.org_invitation_service.OrgStore.get_org_by_id',
+                new_callable=AsyncMock,
+            ) as mock_get_org,
             patch(
                 'server.services.org_invitation_service.OrgMemberStore.add_user_to_org',
                 new_callable=AsyncMock,
@@ -145,6 +153,7 @@ class TestAcceptInvitationEmailValidation:
             mock_settings = MagicMock()
             mock_settings.llm_api_key = SecretStr('test-key')
             mock_create_litellm.return_value = mock_settings
+            mock_get_org.return_value = mock_org
             mock_update_status.return_value = mock_invitation
 
             # Act - should not raise error because Keycloak email matches
@@ -214,6 +223,9 @@ class TestAcceptInvitationEmailValidation:
 
         mock_invitation.email = 'alice@example.com'  # Lowercase in invitation
 
+        mock_org = MagicMock()
+        mock_org.agent_settings = {'schema_version': 1, 'llm.model': 'test-model'}
+
         with (
             patch(
                 'server.services.org_invitation_service.OrgInvitationStore.get_invitation_by_token',
@@ -234,6 +246,10 @@ class TestAcceptInvitationEmailValidation:
                 'server.services.org_invitation_service.OrgService.create_litellm_integration',
                 new_callable=AsyncMock,
             ) as mock_create_litellm,
+            patch(
+                'server.services.org_invitation_service.OrgStore.get_org_by_id',
+                new_callable=AsyncMock,
+            ) as mock_get_org,
             patch(
                 'server.services.org_invitation_service.OrgMemberStore.add_user_to_org',
                 new_callable=AsyncMock,
@@ -250,6 +266,7 @@ class TestAcceptInvitationEmailValidation:
             mock_settings = MagicMock()
             mock_settings.llm_api_key = SecretStr('test-key')
             mock_create_litellm.return_value = mock_settings
+            mock_get_org.return_value = mock_org
             mock_update_status.return_value = mock_invitation
 
             # Act - should not raise error because emails match case-insensitively
@@ -258,6 +275,80 @@ class TestAcceptInvitationEmailValidation:
             # Assert - invitation was accepted (update_invitation_status was called)
             mock_update_status.assert_called_once()
 
+    @pytest.mark.asyncio
+    async def test_accept_invitation_inherits_org_llm_settings(self, mock_invitation):
+        """Test that new members inherit the organization's LLM settings when accepting invitation."""
+        # Arrange
+        user_id = UUID('87654321-4321-8765-4321-876543218765')
+        token = 'inv-test-token-12345'
+
+        mock_user = MagicMock()
+        mock_user.id = user_id
+        mock_user.email = 'alice@example.com'
+
+        mock_org = MagicMock()
+        mock_org.agent_settings = {
+            'schema_version': 1,
+            'llm.model': 'claude-sonnet-4',
+            'llm.base_url': 'https://api.anthropic.com',
+            'max_iterations': 100,
+        }
+
+        with (
+            patch(
+                'server.services.org_invitation_service.OrgInvitationStore.get_invitation_by_token',
+                new_callable=AsyncMock,
+            ) as mock_get_invitation,
+            patch(
+                'server.services.org_invitation_service.OrgInvitationStore.is_token_expired'
+            ) as mock_is_expired,
+            patch(
+                'server.services.org_invitation_service.UserStore.get_user_by_id',
+                new_callable=AsyncMock,
+            ) as mock_get_user,
+            patch(
+                'server.services.org_invitation_service.OrgMemberStore.get_org_member',
+                new_callable=AsyncMock,
+            ) as mock_get_member,
+            patch(
+                'server.services.org_invitation_service.OrgService.create_litellm_integration',
+                new_callable=AsyncMock,
+            ) as mock_create_litellm,
+            patch(
+                'server.services.org_invitation_service.OrgStore.get_org_by_id',
+                new_callable=AsyncMock,
+            ) as mock_get_org,
+            patch(
+                'server.services.org_invitation_service.OrgMemberStore.add_user_to_org',
+                new_callable=AsyncMock,
+            ) as mock_add_user,
+            patch(
+                'server.services.org_invitation_service.OrgInvitationStore.update_invitation_status',
+                new_callable=AsyncMock,
+            ) as mock_update_status,
+        ):
+            mock_get_invitation.return_value = mock_invitation
+            mock_is_expired.return_value = False
+            mock_get_user.return_value = mock_user
+            mock_get_member.return_value = None
+            mock_settings = MagicMock()
+            mock_settings.get_secret_agent_setting.return_value = SecretStr('test-key')
+            mock_create_litellm.return_value = mock_settings
+            mock_get_org.return_value = mock_org
+            mock_update_status.return_value = mock_invitation
+
+            # Act
+            await OrgInvitationService.accept_invitation(token, user_id)
+
+            # Assert - verify add_user_to_org snapshots the org defaults onto
+            # the new membership row's canonical agent_settings blob.
+            mock_add_user.assert_called_once()
+            call_kwargs = mock_add_user.call_args.kwargs
+            assert call_kwargs['llm_api_key'] == 'test-key'
+            assert call_kwargs[
+                'agent_settings'
+            ] == OrgStore.get_agent_settings_from_org(mock_org)
+
 
 class TestCreateInvitationsBatch:
     """Test cases for batch invitation creation."""

enterprise/tests/unit/test_org_member_store.py

@@ -10,6 +10,69 @@ from storage.org_member import OrgMember
 from storage.org_member_store import OrgMemberStore
 from storage.role import Role
 from storage.user import User
+from storage.user_settings import UserSettings
+
+
+def test_get_kwargs_from_user_settings_uses_agent_settings_as_source_of_truth():
+    user_settings = UserSettings(
+        llm_api_key='legacy-secret',
+        agent_settings={
+            'schema_version': 1,
+            'agent': 'CodeActAgent',
+            'verification.confirmation_mode': True,
+            'verification.security_analyzer': 'llm',
+            'condenser.enabled': False,
+            'condenser.max_size': 128,
+            'llm.model': 'anthropic/claude-sonnet-4-5-20250929',
+            'llm.base_url': 'https://api.example.com',
+            'max_iterations': 42,
+        },
+    )
+
+    kwargs = OrgMemberStore.get_kwargs_from_user_settings(user_settings)
+
+    assert kwargs['llm_api_key'] == 'legacy-secret'
+    assert (
+        kwargs['agent_settings']
+        | {
+            'schema_version': 1,
+            'agent': 'CodeActAgent',
+            'verification.confirmation_mode': True,
+            'verification.security_analyzer': 'llm',
+            'condenser.enabled': False,
+            'condenser.max_size': 128,
+            'llm.model': 'anthropic/claude-sonnet-4-5-20250929',
+            'llm.base_url': 'https://api.example.com',
+            'max_iterations': 42,
+        }
+        == kwargs['agent_settings']
+    )
+
+
+def test_get_agent_settings_from_org_member_uses_canonical_snapshot_json():
+    org_member = OrgMember(
+        org_id=uuid.uuid4(),
+        user_id=uuid.uuid4(),
+        role_id=1,
+        llm_api_key='legacy-secret',
+        agent_settings={
+            'schema_version': 1,
+            'agent': 'CodeActAgent',
+            'llm.model': 'member-model',
+            'llm.base_url': 'https://member.example.com',
+            'max_iterations': 42,
+            'verification.confirmation_mode': True,
+        },
+    )
+
+    assert OrgMemberStore.get_agent_settings_from_org_member(org_member) == {
+        'schema_version': 1,
+        'agent': 'CodeActAgent',
+        'llm.model': 'member-model',
+        'llm.base_url': 'https://member.example.com',
+        'max_iterations': 42,
+        'verification.confirmation_mode': True,
+    }
 
 
 @pytest.fixture
@@ -246,6 +309,46 @@ async def test_add_user_to_org(async_session_maker):
         assert org_member.status == 'active'
 
 
+@pytest.mark.asyncio
+async def test_add_user_to_org_with_llm_settings(async_session_maker):
+    """Test that add_user_to_org correctly sets inherited LLM settings from organization."""
+    # Arrange
+    async with async_session_maker() as session:
+        org = Org(name='test-org-llm')
+        session.add(org)
+        await session.flush()
+
+        user = User(id=uuid.uuid4(), current_org_id=org.id)
+        role = Role(name='member', rank=2)
+        session.add_all([user, role])
+        await session.commit()
+        org_id = org.id
+        user_id = user.id
+        role_id = role.id
+
+    # Act
+    with patch('storage.org_member_store.a_session_maker', async_session_maker):
+        org_member = await OrgMemberStore.add_user_to_org(
+            org_id=org_id,
+            user_id=user_id,
+            role_id=role_id,
+            llm_api_key='test-api-key',
+            status='active',
+            agent_settings={
+                'schema_version': 1,
+                'llm.model': 'claude-sonnet-4',
+                'llm.base_url': 'https://api.example.com',
+                'max_iterations': 50,
+            },
+        )
+
+    # Assert
+    assert org_member is not None
+    assert org_member.agent_settings['llm.model'] == 'claude-sonnet-4'
+    assert org_member.agent_settings['llm.base_url'] == 'https://api.example.com'
+    assert org_member.agent_settings['max_iterations'] == 50
+
+
 @pytest.mark.asyncio
 async def test_update_user_role_in_org(async_session_maker):
     # Test updating user role in org
@@ -940,8 +1043,11 @@ async def test_update_all_members_llm_settings_async_with_non_encrypted_fields(
             user_id=user.id,
             role_id=role.id,
             llm_api_key='test-key',
-            llm_model='old-model',
-            max_iterations=10,
+            agent_settings={
+                'schema_version': 1,
+                'llm.model': 'old-model',
+                'max_iterations': 10,
+            },
             status='active',
         )
         session.add(org_member)
@@ -950,9 +1056,11 @@ async def test_update_all_members_llm_settings_async_with_non_encrypted_fields(
 
     # Act
     member_settings = OrgMemberLLMSettings(
-        llm_model='new-model',
-        llm_base_url='https://new-url.com',
-        max_iterations=50,
+        agent_settings={
+            'llm.model': 'new-model',
+            'llm.base_url': 'https://new-url.com',
+            'max_iterations': 50,
+        }
     )
 
     async with async_session_maker() as session:
@@ -970,9 +1078,9 @@ async def test_update_all_members_llm_settings_async_with_non_encrypted_fields(
         )
         updated_member = result.scalars().first()
 
-        assert updated_member.llm_model == 'new-model'
-        assert updated_member.llm_base_url == 'https://new-url.com'
-        assert updated_member.max_iterations == 50
+        assert updated_member.agent_settings['llm.model'] == 'new-model'
+        assert updated_member.agent_settings['llm.base_url'] == 'https://new-url.com'
+        assert updated_member.agent_settings['max_iterations'] == 50
 
 
 @pytest.mark.asyncio
@@ -1005,7 +1113,10 @@ async def test_update_all_members_llm_settings_async_with_empty_settings(
             user_id=user.id,
             role_id=role.id,
             llm_api_key='original-key',
-            llm_model='original-model',
+            agent_settings={
+                'schema_version': 1,
+                'llm.model': 'original-model',
+            },
             status='active',
         )
         session.add(org_member)
@@ -1030,7 +1141,7 @@ async def test_update_all_members_llm_settings_async_with_empty_settings(
         )
         member = result.scalars().first()
 
-        assert member.llm_model == 'original-model'
+        assert member.agent_settings['llm.model'] == 'original-model'
         # Original key should still be there (encrypted)
         assert member._llm_api_key is not None
 
@@ -1078,9 +1189,9 @@ def test_org_member_llm_settings_has_updates_empty():
 
 def test_org_llm_settings_update_apply_to_org_skips_llm_api_key():
     """
-    GIVEN: OrgLLMSettingsUpdate with llm_api_key and other fields set
+    GIVEN: OrgLLMSettingsUpdate with search_api_key and llm_api_key set
     WHEN: apply_to_org() is called
-    THEN: llm_api_key is NOT applied to org, but other fields are
+    THEN: search_api_key is applied to org, but llm_api_key is not
     """
     from unittest.mock import MagicMock
 
@@ -1088,18 +1199,17 @@ def test_org_llm_settings_update_apply_to_org_skips_llm_api_key():
 
     # Arrange
     settings = OrgLLMSettingsUpdate(
-        default_llm_model='claude-3',
+        search_api_key='applied-to-org',
         llm_api_key='should-not-be-applied',
     )
     mock_org = MagicMock()
-    mock_org.default_llm_model = None
+    mock_org.search_api_key = None
 
     # Act
     settings.apply_to_org(mock_org)
 
     # Assert
-    assert mock_org.default_llm_model == 'claude-3'
-    # llm_api_key should NOT be set on org (it's member-only)
+    assert mock_org.search_api_key == 'applied-to-org'
     assert (
         not hasattr(mock_org, 'llm_api_key')
         or mock_org.llm_api_key != 'should-not-be-applied'
@@ -1116,7 +1226,7 @@ def test_org_llm_settings_update_get_member_updates_includes_llm_api_key():
 
     # Arrange
     settings = OrgLLMSettingsUpdate(
-        default_llm_model='claude-3',
+        agent_settings={'llm.model': 'claude-3'},
         llm_api_key='new-member-key',
     )
 
@@ -1126,7 +1236,7 @@ def test_org_llm_settings_update_get_member_updates_includes_llm_api_key():
     # Assert
     assert member_updates is not None
     assert member_updates.llm_api_key == 'new-member-key'
-    assert member_updates.llm_model == 'claude-3'
+    assert member_updates.agent_settings is None
 
 
 def test_org_llm_settings_update_get_member_updates_only_llm_api_key():
@@ -1146,7 +1256,7 @@ def test_org_llm_settings_update_get_member_updates_only_llm_api_key():
     # Assert
     assert member_updates is not None
     assert member_updates.llm_api_key == 'member-key-only'
-    assert member_updates.llm_model is None
+    assert member_updates.agent_settings is None
 
 
 def test_org_llm_settings_update_has_updates_with_llm_api_key():