Merge remote-tracking branch 'origin/dev' into fix/copilot-progress-bar

I'm getting circular import issues because there is a lot of
cross-importing between `backend.data`, `backend.blocks`, and other
modules. This change reduces block-related cross-imports and thus risk
of breaking circular imports.

### Changes 🏗️

- Strip down `backend.data.block`
- Move `Block` base class and related class/enum defs to
`backend.blocks._base`
  - Move `is_block_auth_configured` to `backend.blocks._utils`
- Move `get_blocks()`, `get_io_block_ids()` etc. to `backend.blocks`
(`__init__.py`)
  - Update imports everywhere
- Remove unused and poorly typed `Block.create()`
  - Change usages from `block_cls.create()` to `block_cls()`
- Improve typing of `load_all_blocks` and `get_blocks`
- Move cross-import of `backend.api.features.library.model` from
`backend/data/__init__.py` to `backend/data/integrations.py`
- Remove deprecated attribute `NodeModel.webhook`
  - Re-generate OpenAPI spec and fix frontend usage
- Eliminate module-level `backend.blocks` import from `blocks/agent.py`
- Eliminate module-level `backend.data.execution` and
`backend.executor.manager` imports from `blocks/helpers/review.py`
- Replace `BlockInput` with `GraphInput` for graph inputs

### Checklist 📋

#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
  - CI static type-checking + tests should be sufficient for this

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

@@ -49,7 +49,7 @@ jobs:
 
       - name: Create PR ${{ env.BUILD_BRANCH }} -> ${{ github.ref_name }}
         if: github.event_name == 'push'
-        uses: peter-evans/create-pull-request@v7
+        uses: peter-evans/create-pull-request@v8
         with:
           add-paths: classic/frontend/build/web
           base: ${{ github.ref_name }}

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

@@ -22,7 +22,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout code
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           ref: ${{ github.event.workflow_run.head_branch }}
           fetch-depth: 0
@@ -42,7 +42,7 @@ jobs:
 
       - name: Get CI failure details
         id: failure_details
-        uses: actions/github-script@v7
+        uses: actions/github-script@v8
         with:
           script: |
             const run = await github.rest.actions.getWorkflowRun({

.github/workflows/claude-dependabot.yml

@@ -30,7 +30,7 @@ jobs:
       actions: read # Required for CI access
     steps:
       - name: Checkout code
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           fetch-depth: 1
 
@@ -41,7 +41,7 @@ jobs:
           python-version: "3.11"  # Use standard version matching CI
 
       - name: Set up Python dependency cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.cache/pypoetry
           key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
@@ -78,7 +78,7 @@ jobs:
 
       # Frontend Node.js/pnpm setup (mirrors platform-frontend-ci.yml)
       - name: Set up Node.js
-        uses: actions/setup-node@v4
+        uses: actions/setup-node@v6
         with:
           node-version: "22"
 
@@ -91,7 +91,7 @@ jobs:
           echo "PNPM_HOME=$HOME/.pnpm-store" >> $GITHUB_ENV
 
       - name: Cache frontend dependencies
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.pnpm-store
           key: ${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml', 'autogpt_platform/frontend/package.json') }}
@@ -124,7 +124,7 @@ jobs:
       # Phase 1: Cache and load Docker images for faster setup
       - name: Set up Docker image cache
         id: docker-cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/docker-cache
           # Use a versioned key for cache invalidation when image list changes
@@ -309,6 +309,7 @@ jobs:
         uses: anthropics/claude-code-action@v1
         with:
           claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
+          allowed_bots: "dependabot[bot]"
           claude_args: |
             --allowedTools "Bash(npm:*),Bash(pnpm:*),Bash(poetry:*),Bash(git:*),Edit,Replace,NotebookEditCell,mcp__github_inline_comment__create_inline_comment,Bash(gh pr comment:*), Bash(gh pr diff:*), Bash(gh pr view:*)"
           prompt: |

.github/workflows/claude.yml

@@ -40,7 +40,7 @@ jobs:
       actions: read # Required for CI access
     steps:
       - name: Checkout code
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           fetch-depth: 1
 
@@ -57,7 +57,7 @@ jobs:
           python-version: "3.11"  # Use standard version matching CI
 
       - name: Set up Python dependency cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.cache/pypoetry
           key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
@@ -94,7 +94,7 @@ jobs:
 
       # Frontend Node.js/pnpm setup (mirrors platform-frontend-ci.yml)
       - name: Set up Node.js
-        uses: actions/setup-node@v4
+        uses: actions/setup-node@v6
         with:
           node-version: "22"
 
@@ -107,7 +107,7 @@ jobs:
           echo "PNPM_HOME=$HOME/.pnpm-store" >> $GITHUB_ENV
 
       - name: Cache frontend dependencies
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.pnpm-store
           key: ${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml', 'autogpt_platform/frontend/package.json') }}
@@ -140,7 +140,7 @@ jobs:
       # Phase 1: Cache and load Docker images for faster setup
       - name: Set up Docker image cache
         id: docker-cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/docker-cache
           # Use a versioned key for cache invalidation when image list changes

.github/workflows/codeql.yml

@@ -58,7 +58,7 @@ jobs:
         # your codebase is analyzed, see https://docs.github.com/en/code-security/code-scanning/creating-an-advanced-setup-for-code-scanning/codeql-code-scanning-for-compiled-languages
     steps:
     - name: Checkout repository
-      uses: actions/checkout@v4
+      uses: actions/checkout@v6
 
     # Initializes the CodeQL tools for scanning.
     - name: Initialize CodeQL

.github/workflows/copilot-setup-steps.yml

@@ -27,7 +27,7 @@ jobs:
     # If you do not check out your code, Copilot will do this for you.
     steps:
       - name: Checkout code
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           fetch-depth: 0
           submodules: true
@@ -39,7 +39,7 @@ jobs:
           python-version: "3.11"  # Use standard version matching CI
 
       - name: Set up Python dependency cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.cache/pypoetry
           key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
@@ -76,7 +76,7 @@ jobs:
 
       # Frontend Node.js/pnpm setup (mirrors platform-frontend-ci.yml)
       - name: Set up Node.js
-        uses: actions/setup-node@v4
+        uses: actions/setup-node@v6
         with:
           node-version: "22"
 
@@ -89,7 +89,7 @@ jobs:
           echo "PNPM_HOME=$HOME/.pnpm-store" >> $GITHUB_ENV
 
       - name: Cache frontend dependencies
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.pnpm-store
           key: ${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml', 'autogpt_platform/frontend/package.json') }}
@@ -132,7 +132,7 @@ jobs:
       # Phase 1: Cache and load Docker images for faster setup
       - name: Set up Docker image cache
         id: docker-cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/docker-cache
           # Use a versioned key for cache invalidation when image list changes

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

@@ -23,7 +23,7 @@ jobs:
 
     steps:
       - name: Checkout code
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           fetch-depth: 1
 
@@ -33,7 +33,7 @@ jobs:
           python-version: "3.11"
 
       - name: Set up Python dependency cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.cache/pypoetry
           key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}

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

@@ -23,7 +23,7 @@ jobs:
 
     steps:
       - name: Checkout code
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           fetch-depth: 0
 
@@ -33,7 +33,7 @@ jobs:
           python-version: "3.11"
 
       - name: Set up Python dependency cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.cache/pypoetry
           key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}

.github/workflows/docs-enhance.yml

@@ -28,7 +28,7 @@ jobs:
 
     steps:
       - name: Checkout code
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           fetch-depth: 1
 
@@ -38,7 +38,7 @@ jobs:
           python-version: "3.11"
 
       - name: Set up Python dependency cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.cache/pypoetry
           key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}

.github/workflows/platform-autogpt-deploy-dev.yaml

@@ -25,7 +25,7 @@ jobs:
 
     steps:
       - name: Checkout code
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           ref: ${{ github.event.inputs.git_ref || github.ref_name }}
 
@@ -52,7 +52,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Trigger deploy workflow
-        uses: peter-evans/repository-dispatch@v3
+        uses: peter-evans/repository-dispatch@v4
         with:
           token: ${{ secrets.DEPLOY_TOKEN }}
           repository: Significant-Gravitas/AutoGPT_cloud_infrastructure

.github/workflows/platform-autogpt-deploy-prod.yml

@@ -17,7 +17,7 @@ jobs:
 
     steps:
       - name: Checkout code
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           ref: ${{ github.ref_name || 'master' }}
 
@@ -45,7 +45,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Trigger deploy workflow
-        uses: peter-evans/repository-dispatch@v3
+        uses: peter-evans/repository-dispatch@v4
         with:
           token: ${{ secrets.DEPLOY_TOKEN }}
           repository: Significant-Gravitas/AutoGPT_cloud_infrastructure

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

@@ -68,7 +68,7 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           fetch-depth: 0
           submodules: true
@@ -88,7 +88,7 @@ jobs:
         run: echo "date=$(date +'%Y-%m-%d')" >> $GITHUB_OUTPUT
 
       - name: Set up Python dependency cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.cache/pypoetry
           key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}

.github/workflows/platform-dev-deploy-event-dispatcher.yml

@@ -17,7 +17,7 @@ jobs:
       - name: Check comment permissions and deployment status
         id: check_status
         if: github.event_name == 'issue_comment' && github.event.issue.pull_request
-        uses: actions/github-script@v7
+        uses: actions/github-script@v8
         with:
           script: |
             const commentBody = context.payload.comment.body.trim();
@@ -55,7 +55,7 @@ jobs:
 
       - name: Post permission denied comment
         if: steps.check_status.outputs.permission_denied == 'true'
-        uses: actions/github-script@v7
+        uses: actions/github-script@v8
         with:
           script: |
             await github.rest.issues.createComment({
@@ -68,7 +68,7 @@ jobs:
       - name: Get PR details for deployment
         id: pr_details
         if: steps.check_status.outputs.should_deploy == 'true' || steps.check_status.outputs.should_undeploy == 'true'
-        uses: actions/github-script@v7
+        uses: actions/github-script@v8
         with:
           script: |
             const pr = await github.rest.pulls.get({
@@ -82,7 +82,7 @@ jobs:
           
       - name: Dispatch Deploy Event
         if: steps.check_status.outputs.should_deploy == 'true'
-        uses: peter-evans/repository-dispatch@v3
+        uses: peter-evans/repository-dispatch@v4
         with:
           token: ${{ secrets.DISPATCH_TOKEN }}
           repository: Significant-Gravitas/AutoGPT_cloud_infrastructure
@@ -98,7 +98,7 @@ jobs:
 
       - name: Post deploy success comment
         if: steps.check_status.outputs.should_deploy == 'true'
-        uses: actions/github-script@v7
+        uses: actions/github-script@v8
         with:
           script: |
             await github.rest.issues.createComment({
@@ -110,7 +110,7 @@ jobs:
 
       - name: Dispatch Undeploy Event (from comment)
         if: steps.check_status.outputs.should_undeploy == 'true'
-        uses: peter-evans/repository-dispatch@v3
+        uses: peter-evans/repository-dispatch@v4
         with:
           token: ${{ secrets.DISPATCH_TOKEN }}
           repository: Significant-Gravitas/AutoGPT_cloud_infrastructure
@@ -126,7 +126,7 @@ jobs:
 
       - name: Post undeploy success comment
         if: steps.check_status.outputs.should_undeploy == 'true'
-        uses: actions/github-script@v7
+        uses: actions/github-script@v8
         with:
           script: |
             await github.rest.issues.createComment({
@@ -139,7 +139,7 @@ jobs:
       - name: Check deployment status on PR close
         id: check_pr_close
         if: github.event_name == 'pull_request' && github.event.action == 'closed'
-        uses: actions/github-script@v7
+        uses: actions/github-script@v8
         with:
           script: |
             const comments = await github.rest.issues.listComments({
@@ -168,7 +168,7 @@ jobs:
           github.event_name == 'pull_request' &&
           github.event.action == 'closed' &&
           steps.check_pr_close.outputs.should_undeploy == 'true'
-        uses: peter-evans/repository-dispatch@v3
+        uses: peter-evans/repository-dispatch@v4
         with:
           token: ${{ secrets.DISPATCH_TOKEN }}
           repository: Significant-Gravitas/AutoGPT_cloud_infrastructure
@@ -187,7 +187,7 @@ jobs:
           github.event_name == 'pull_request' &&
           github.event.action == 'closed' &&
           steps.check_pr_close.outputs.should_undeploy == 'true'
-        uses: actions/github-script@v7
+        uses: actions/github-script@v8
         with:
           script: |
             await github.rest.issues.createComment({

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

@@ -27,13 +27,22 @@ jobs:
     runs-on: ubuntu-latest
     outputs:
       cache-key: ${{ steps.cache-key.outputs.key }}
+      components-changed: ${{ steps.filter.outputs.components }}
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
+
+      - name: Check for component changes
+        uses: dorny/paths-filter@v3
+        id: filter
+        with:
+          filters: |
+            components:
+              - 'autogpt_platform/frontend/src/components/**'
 
       - name: Set up Node.js
-        uses: actions/setup-node@v4
+        uses: actions/setup-node@v6
         with:
           node-version: "22.18.0"
 
@@ -45,7 +54,7 @@ jobs:
         run: echo "key=${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml', 'autogpt_platform/frontend/package.json') }}" >> $GITHUB_OUTPUT
 
       - name: Cache dependencies
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.pnpm-store
           key: ${{ steps.cache-key.outputs.key }}
@@ -62,10 +71,10 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
 
       - name: Set up Node.js
-        uses: actions/setup-node@v4
+        uses: actions/setup-node@v6
         with:
           node-version: "22.18.0"
 
@@ -73,7 +82,7 @@ jobs:
         run: corepack enable
 
       - name: Restore dependencies cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.pnpm-store
           key: ${{ needs.setup.outputs.cache-key }}
@@ -90,17 +99,20 @@ jobs:
   chromatic:
     runs-on: ubuntu-latest
     needs: setup
-    # Only run on dev branch pushes or PRs targeting dev
+    # Disabled: to re-enable, remove 'false &&' from the condition below
-    if: github.ref == 'refs/heads/dev' || github.base_ref == 'dev'
+    if: >-
+      false
+      && (github.ref == 'refs/heads/dev' || github.base_ref == 'dev')
+      && needs.setup.outputs.components-changed == 'true'
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           fetch-depth: 0
 
       - name: Set up Node.js
-        uses: actions/setup-node@v4
+        uses: actions/setup-node@v6
         with:
           node-version: "22.18.0"
 
@@ -108,7 +120,7 @@ jobs:
         run: corepack enable
 
       - name: Restore dependencies cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.pnpm-store
           key: ${{ needs.setup.outputs.cache-key }}
@@ -136,12 +148,12 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           submodules: recursive
 
       - name: Set up Node.js
-        uses: actions/setup-node@v4
+        uses: actions/setup-node@v6
         with:
           node-version: "22.18.0"
 
@@ -164,7 +176,7 @@ jobs:
         uses: docker/setup-buildx-action@v3
 
       - name: Cache Docker layers
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: /tmp/.buildx-cache
           key: ${{ runner.os }}-buildx-frontend-test-${{ hashFiles('autogpt_platform/docker-compose.yml', 'autogpt_platform/backend/Dockerfile', 'autogpt_platform/backend/pyproject.toml', 'autogpt_platform/backend/poetry.lock') }}
@@ -219,7 +231,7 @@ jobs:
           fi
 
       - name: Restore dependencies cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.pnpm-store
           key: ${{ needs.setup.outputs.cache-key }}
@@ -265,12 +277,12 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           submodules: recursive
 
       - name: Set up Node.js
-        uses: actions/setup-node@v4
+        uses: actions/setup-node@v6
         with:
           node-version: "22.18.0"
 
@@ -278,7 +290,7 @@ jobs:
         run: corepack enable
 
       - name: Restore dependencies cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.pnpm-store
           key: ${{ needs.setup.outputs.cache-key }}

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

@@ -29,10 +29,10 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
 
       - name: Set up Node.js
-        uses: actions/setup-node@v4
+        uses: actions/setup-node@v6
         with:
           node-version: "22.18.0"
 
@@ -44,7 +44,7 @@ jobs:
         run: echo "key=${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml', 'autogpt_platform/frontend/package.json') }}" >> $GITHUB_OUTPUT
 
       - name: Cache dependencies
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.pnpm-store
           key: ${{ steps.cache-key.outputs.key }}
@@ -56,19 +56,19 @@ jobs:
         run: pnpm install --frozen-lockfile
 
   types:
-    runs-on: ubuntu-latest
+    runs-on: big-boi
     needs: setup
     strategy:
       fail-fast: false
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
           submodules: recursive
 
       - name: Set up Node.js
-        uses: actions/setup-node@v4
+        uses: actions/setup-node@v6
         with:
           node-version: "22.18.0"
 
@@ -85,10 +85,10 @@ jobs:
 
       - name: Run docker compose
         run: |
-          docker compose -f ../docker-compose.yml --profile local --profile deps_backend up -d
+          docker compose -f ../docker-compose.yml --profile local up -d deps_backend
 
       - name: Restore dependencies cache
-        uses: actions/cache@v4
+        uses: actions/cache@v5
         with:
           path: ~/.pnpm-store
           key: ${{ needs.setup.outputs.cache-key }}

.github/workflows/repo-workflow-checker.yml

@@ -11,7 +11,7 @@ jobs:
     steps:
       # - name: Wait some time for all actions to start
       #   run: sleep 30
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
         # with:
           # fetch-depth: 0
       - name: Set up Python

autogpt_platform/autogpt_libs/pyproject.toml

@@ -9,25 +9,25 @@ packages = [{ include = "autogpt_libs" }]
 [tool.poetry.dependencies]
 python = ">=3.10,<4.0"
 colorama = "^0.4.6"
-cryptography = "^45.0"
+cryptography = "^46.0"
 expiringdict = "^1.2.2"
-fastapi = "^0.116.1"
+fastapi = "^0.128.0"
-google-cloud-logging = "^3.12.1"
+google-cloud-logging = "^3.13.0"
-launchdarkly-server-sdk = "^9.12.0"
+launchdarkly-server-sdk = "^9.14.1"
-pydantic = "^2.11.7"
+pydantic = "^2.12.5"
-pydantic-settings = "^2.10.1"
+pydantic-settings = "^2.12.0"
-pyjwt = { version = "^2.10.1", extras = ["crypto"] }
+pyjwt = { version = "^2.11.0", extras = ["crypto"] }
 redis = "^6.2.0"
-supabase = "^2.16.0"
+supabase = "^2.27.2"
-uvicorn = "^0.35.0"
+uvicorn = "^0.40.0"
 
 [tool.poetry.group.dev.dependencies]
-pyright = "^1.1.404"
+pyright = "^1.1.408"
 pytest = "^8.4.1"
-pytest-asyncio = "^1.1.0"
+pytest-asyncio = "^1.3.0"
-pytest-mock = "^3.14.1"
+pytest-mock = "^3.15.1"
-pytest-cov = "^6.2.1"
+pytest-cov = "^7.0.0"
-ruff = "^0.12.11"
+ruff = "^0.15.0"
 
 [build-system]
 requires = ["poetry-core"]

autogpt_platform/backend/.env.default

@@ -152,6 +152,7 @@ REPLICATE_API_KEY=
 REVID_API_KEY=
 SCREENSHOTONE_API_KEY=
 UNREAL_SPEECH_API_KEY=
+ELEVENLABS_API_KEY=
 
 # Data & Search Services
 E2B_API_KEY=

autogpt_platform/backend/.gitignore

@@ -19,3 +19,6 @@ load-tests/*.json
 load-tests/*.log
 load-tests/node_modules/*
 migrations/*/rollback*.sql
+
+# Workspace files
+workspaces/

autogpt_platform/backend/Dockerfile

@@ -62,10 +62,12 @@ ENV POETRY_HOME=/opt/poetry \
     DEBIAN_FRONTEND=noninteractive
 ENV PATH=/opt/poetry/bin:$PATH
 
-# Install Python without upgrading system-managed packages
+# Install Python, FFmpeg, and ImageMagick (required for video processing blocks)
 RUN apt-get update && apt-get install -y \
     python3.13 \
     python3-pip \
+    ffmpeg \
+    imagemagick \
     && rm -rf /var/lib/apt/lists/*
 
 # Copy only necessary files from builder

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

@@ -10,7 +10,7 @@ from typing_extensions import TypedDict
 
 import backend.api.features.store.cache as store_cache
 import backend.api.features.store.model as store_model
-import backend.data.block
+import backend.blocks
 from backend.api.external.middleware import require_permission
 from backend.data import execution as execution_db
 from backend.data import graph as graph_db
@@ -67,7 +67,7 @@ async def get_user_info(
     dependencies=[Security(require_permission(APIKeyPermission.READ_BLOCK))],
 )
 async def get_graph_blocks() -> Sequence[dict[Any, Any]]:
-    blocks = [block() for block in backend.data.block.get_blocks().values()]
+    blocks = [block() for block in backend.blocks.get_blocks().values()]
     return [b.to_dict() for b in blocks if not b.disabled]
 
 
@@ -83,7 +83,7 @@ async def execute_graph_block(
         require_permission(APIKeyPermission.EXECUTE_BLOCK)
     ),
 ) -> CompletedBlockOutput:
-    obj = backend.data.block.get_block(block_id)
+    obj = backend.blocks.get_block(block_id)
     if not obj:
         raise HTTPException(status_code=404, detail=f"Block #{block_id} not found.")
     if obj.disabled:

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

@@ -10,10 +10,15 @@ import backend.api.features.library.db as library_db
 import backend.api.features.library.model as library_model
 import backend.api.features.store.db as store_db
 import backend.api.features.store.model as store_model
-import backend.data.block
 from backend.blocks import load_all_blocks
+from backend.blocks._base import (
+    AnyBlockSchema,
+    BlockCategory,
+    BlockInfo,
+    BlockSchema,
+    BlockType,
+)
 from backend.blocks.llm import LlmModel
-from backend.data.block import AnyBlockSchema, BlockCategory, BlockInfo, BlockSchema
 from backend.data.db import query_raw_with_schema
 from backend.integrations.providers import ProviderName
 from backend.util.cache import cached
@@ -22,7 +27,7 @@ from backend.util.models import Pagination
 from .model import (
     BlockCategoryResponse,
     BlockResponse,
-    BlockType,
+    BlockTypeFilter,
     CountResponse,
     FilterType,
     Provider,
@@ -88,7 +93,7 @@ def get_block_categories(category_blocks: int = 3) -> list[BlockCategoryResponse
 def get_blocks(
     *,
     category: str | None = None,
-    type: BlockType | None = None,
+    type: BlockTypeFilter | None = None,
     provider: ProviderName | None = None,
     page: int = 1,
     page_size: int = 50,
@@ -669,9 +674,9 @@ async def get_suggested_blocks(count: int = 5) -> list[BlockInfo]:
     for block_type in load_all_blocks().values():
         block: AnyBlockSchema = block_type()
         if block.disabled or block.block_type in (
-            backend.data.block.BlockType.INPUT,
+            BlockType.INPUT,
-            backend.data.block.BlockType.OUTPUT,
+            BlockType.OUTPUT,
-            backend.data.block.BlockType.AGENT,
+            BlockType.AGENT,
         ):
             continue
         # Find the execution count for this block

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

@@ -4,7 +4,7 @@ from pydantic import BaseModel
 
 import backend.api.features.library.model as library_model
 import backend.api.features.store.model as store_model
-from backend.data.block import BlockInfo
+from backend.blocks._base import BlockInfo
 from backend.integrations.providers import ProviderName
 from backend.util.models import Pagination
 
@@ -15,7 +15,7 @@ FilterType = Literal[
     "my_agents",
 ]
 
-BlockType = Literal["all", "input", "action", "output"]
+BlockTypeFilter = Literal["all", "input", "action", "output"]
 
 
 class SearchEntry(BaseModel):

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

@@ -88,7 +88,7 @@ async def get_block_categories(
 )
 async def get_blocks(
     category: Annotated[str | None, fastapi.Query()] = None,
-    type: Annotated[builder_model.BlockType | None, fastapi.Query()] = None,
+    type: Annotated[builder_model.BlockTypeFilter | None, fastapi.Query()] = None,
     provider: Annotated[ProviderName | None, fastapi.Query()] = None,
     page: Annotated[int, fastapi.Query()] = 1,
     page_size: Annotated[int, fastapi.Query()] = 50,

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

@@ -11,7 +11,7 @@ class ChatConfig(BaseSettings):
 
     # OpenAI API Configuration
     model: str = Field(
-        default="anthropic/claude-opus-4.5", description="Default model to use"
+        default="anthropic/claude-opus-4.6", description="Default model to use"
     )
     title_model: str = Field(
         default="openai/gpt-4o-mini",
@@ -93,6 +93,12 @@ class ChatConfig(BaseSettings):
         description="Name of the prompt in Langfuse to fetch",
     )
 
+    # Extended thinking configuration for Claude models
+    thinking_enabled: bool = Field(
+        default=True,
+        description="Enable adaptive thinking for Claude models via OpenRouter",
+    )
+
     @field_validator("api_key", mode="before")
     @classmethod
     def get_api_key(cls, v):

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

@@ -45,10 +45,7 @@ async def create_chat_session(
         successfulAgentRuns=SafeJson({}),
         successfulAgentSchedules=SafeJson({}),
     )
-    return await PrismaChatSession.prisma().create(
+    return await PrismaChatSession.prisma().create(data=data)
-        data=data,
-        include={"Messages": True},
-    )
 
 
 async def update_chat_session(

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

@@ -2,7 +2,7 @@ import asyncio
 import logging
 import uuid
 from datetime import UTC, datetime
-from typing import Any
+from typing import Any, cast
 from weakref import WeakValueDictionary
 
 from openai.types.chat import (
@@ -104,6 +104,26 @@ class ChatSession(BaseModel):
     successful_agent_runs: dict[str, int] = {}
     successful_agent_schedules: dict[str, int] = {}
 
+    def add_tool_call_to_current_turn(self, tool_call: dict) -> None:
+        """Attach a tool_call to the current turn's assistant message.
+
+        Searches backwards for the most recent assistant message (stopping at
+        any user message boundary). If found, appends the tool_call to it.
+        Otherwise creates a new assistant message with the tool_call.
+        """
+        for msg in reversed(self.messages):
+            if msg.role == "user":
+                break
+            if msg.role == "assistant":
+                if not msg.tool_calls:
+                    msg.tool_calls = []
+                msg.tool_calls.append(tool_call)
+                return
+
+        self.messages.append(
+            ChatMessage(role="assistant", content="", tool_calls=[tool_call])
+        )
+
     @staticmethod
     def new(user_id: str) -> "ChatSession":
         return ChatSession(
@@ -172,6 +192,47 @@ class ChatSession(BaseModel):
             successful_agent_schedules=successful_agent_schedules,
         )
 
+    @staticmethod
+    def _merge_consecutive_assistant_messages(
+        messages: list[ChatCompletionMessageParam],
+    ) -> list[ChatCompletionMessageParam]:
+        """Merge consecutive assistant messages into single messages.
+
+        Long-running tool flows can create split assistant messages: one with
+        text content and another with tool_calls. Anthropic's API requires
+        tool_result blocks to reference a tool_use in the immediately preceding
+        assistant message, so these splits cause 400 errors via OpenRouter.
+        """
+        if len(messages) < 2:
+            return messages
+
+        result: list[ChatCompletionMessageParam] = [messages[0]]
+        for msg in messages[1:]:
+            prev = result[-1]
+            if prev.get("role") != "assistant" or msg.get("role") != "assistant":
+                result.append(msg)
+                continue
+
+            prev = cast(ChatCompletionAssistantMessageParam, prev)
+            curr = cast(ChatCompletionAssistantMessageParam, msg)
+
+            curr_content = curr.get("content") or ""
+            if curr_content:
+                prev_content = prev.get("content") or ""
+                prev["content"] = (
+                    f"{prev_content}\n{curr_content}" if prev_content else curr_content
+                )
+
+            curr_tool_calls = curr.get("tool_calls")
+            if curr_tool_calls:
+                prev_tool_calls = prev.get("tool_calls")
+                prev["tool_calls"] = (
+                    list(prev_tool_calls) + list(curr_tool_calls)
+                    if prev_tool_calls
+                    else list(curr_tool_calls)
+                )
+        return result
+
     def to_openai_messages(self) -> list[ChatCompletionMessageParam]:
         messages = []
         for message in self.messages:
@@ -258,7 +319,7 @@ class ChatSession(BaseModel):
                         name=message.name or "",
                     )
                 )
-        return messages
+        return self._merge_consecutive_assistant_messages(messages)
 
 
 async def _get_session_from_cache(session_id: str) -> ChatSession | None:

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

@@ -1,4 +1,16 @@
+from typing import cast
+
 import pytest
+from openai.types.chat import (
+    ChatCompletionAssistantMessageParam,
+    ChatCompletionMessageParam,
+    ChatCompletionToolMessageParam,
+    ChatCompletionUserMessageParam,
+)
+from openai.types.chat.chat_completion_message_tool_call_param import (
+    ChatCompletionMessageToolCallParam,
+    Function,
+)
 
 from .model import (
     ChatMessage,
@@ -117,3 +129,205 @@ async def test_chatsession_db_storage(setup_test_user, test_user_id):
                 loaded.tool_calls is not None
             ), f"Tool calls missing for {orig.role} message"
             assert len(orig.tool_calls) == len(loaded.tool_calls)
+
+
+# --------------------------------------------------------------------------- #
+#  _merge_consecutive_assistant_messages                                       #
+# --------------------------------------------------------------------------- #
+
+_tc = ChatCompletionMessageToolCallParam(
+    id="tc1", type="function", function=Function(name="do_stuff", arguments="{}")
+)
+_tc2 = ChatCompletionMessageToolCallParam(
+    id="tc2", type="function", function=Function(name="other", arguments="{}")
+)
+
+
+def test_merge_noop_when_no_consecutive_assistants():
+    """Messages without consecutive assistants are returned unchanged."""
+    msgs = [
+        ChatCompletionUserMessageParam(role="user", content="hi"),
+        ChatCompletionAssistantMessageParam(role="assistant", content="hello"),
+        ChatCompletionUserMessageParam(role="user", content="bye"),
+    ]
+    merged = ChatSession._merge_consecutive_assistant_messages(msgs)
+    assert len(merged) == 3
+    assert [m["role"] for m in merged] == ["user", "assistant", "user"]
+
+
+def test_merge_splits_text_and_tool_calls():
+    """The exact bug scenario: text-only assistant followed by tool_calls-only assistant."""
+    msgs = [
+        ChatCompletionUserMessageParam(role="user", content="build agent"),
+        ChatCompletionAssistantMessageParam(
+            role="assistant", content="Let me build that"
+        ),
+        ChatCompletionAssistantMessageParam(
+            role="assistant", content="", tool_calls=[_tc]
+        ),
+        ChatCompletionToolMessageParam(role="tool", content="ok", tool_call_id="tc1"),
+    ]
+    merged = ChatSession._merge_consecutive_assistant_messages(msgs)
+
+    assert len(merged) == 3
+    assert merged[0]["role"] == "user"
+    assert merged[2]["role"] == "tool"
+    a = cast(ChatCompletionAssistantMessageParam, merged[1])
+    assert a["role"] == "assistant"
+    assert a.get("content") == "Let me build that"
+    assert a.get("tool_calls") == [_tc]
+
+
+def test_merge_combines_tool_calls_from_both():
+    """Both consecutive assistants have tool_calls — they get merged."""
+    msgs: list[ChatCompletionAssistantMessageParam] = [
+        ChatCompletionAssistantMessageParam(
+            role="assistant", content="text", tool_calls=[_tc]
+        ),
+        ChatCompletionAssistantMessageParam(
+            role="assistant", content="", tool_calls=[_tc2]
+        ),
+    ]
+    merged = ChatSession._merge_consecutive_assistant_messages(msgs)  # type: ignore[arg-type]
+
+    assert len(merged) == 1
+    a = cast(ChatCompletionAssistantMessageParam, merged[0])
+    assert a.get("tool_calls") == [_tc, _tc2]
+    assert a.get("content") == "text"
+
+
+def test_merge_three_consecutive_assistants():
+    """Three consecutive assistants collapse into one."""
+    msgs: list[ChatCompletionAssistantMessageParam] = [
+        ChatCompletionAssistantMessageParam(role="assistant", content="a"),
+        ChatCompletionAssistantMessageParam(role="assistant", content="b"),
+        ChatCompletionAssistantMessageParam(
+            role="assistant", content="", tool_calls=[_tc]
+        ),
+    ]
+    merged = ChatSession._merge_consecutive_assistant_messages(msgs)  # type: ignore[arg-type]
+
+    assert len(merged) == 1
+    a = cast(ChatCompletionAssistantMessageParam, merged[0])
+    assert a.get("content") == "a\nb"
+    assert a.get("tool_calls") == [_tc]
+
+
+def test_merge_empty_and_single_message():
+    """Edge cases: empty list and single message."""
+    assert ChatSession._merge_consecutive_assistant_messages([]) == []
+
+    single: list[ChatCompletionMessageParam] = [
+        ChatCompletionUserMessageParam(role="user", content="hi")
+    ]
+    assert ChatSession._merge_consecutive_assistant_messages(single) == single
+
+
+# --------------------------------------------------------------------------- #
+#  add_tool_call_to_current_turn                                               #
+# --------------------------------------------------------------------------- #
+
+_raw_tc = {
+    "id": "tc1",
+    "type": "function",
+    "function": {"name": "f", "arguments": "{}"},
+}
+_raw_tc2 = {
+    "id": "tc2",
+    "type": "function",
+    "function": {"name": "g", "arguments": "{}"},
+}
+
+
+def test_add_tool_call_appends_to_existing_assistant():
+    """When the last assistant is from the current turn, tool_call is added to it."""
+    session = ChatSession.new(user_id="u")
+    session.messages = [
+        ChatMessage(role="user", content="hi"),
+        ChatMessage(role="assistant", content="working on it"),
+    ]
+    session.add_tool_call_to_current_turn(_raw_tc)
+
+    assert len(session.messages) == 2  # no new message created
+    assert session.messages[1].tool_calls == [_raw_tc]
+
+
+def test_add_tool_call_creates_assistant_when_none_exists():
+    """When there's no current-turn assistant, a new one is created."""
+    session = ChatSession.new(user_id="u")
+    session.messages = [
+        ChatMessage(role="user", content="hi"),
+    ]
+    session.add_tool_call_to_current_turn(_raw_tc)
+
+    assert len(session.messages) == 2
+    assert session.messages[1].role == "assistant"
+    assert session.messages[1].tool_calls == [_raw_tc]
+
+
+def test_add_tool_call_does_not_cross_user_boundary():
+    """A user message acts as a boundary — previous assistant is not modified."""
+    session = ChatSession.new(user_id="u")
+    session.messages = [
+        ChatMessage(role="assistant", content="old turn"),
+        ChatMessage(role="user", content="new message"),
+    ]
+    session.add_tool_call_to_current_turn(_raw_tc)
+
+    assert len(session.messages) == 3  # new assistant was created
+    assert session.messages[0].tool_calls is None  # old assistant untouched
+    assert session.messages[2].role == "assistant"
+    assert session.messages[2].tool_calls == [_raw_tc]
+
+
+def test_add_tool_call_multiple_times():
+    """Multiple long-running tool calls accumulate on the same assistant."""
+    session = ChatSession.new(user_id="u")
+    session.messages = [
+        ChatMessage(role="user", content="hi"),
+        ChatMessage(role="assistant", content="doing stuff"),
+    ]
+    session.add_tool_call_to_current_turn(_raw_tc)
+    # Simulate a pending tool result in between (like _yield_tool_call does)
+    session.messages.append(
+        ChatMessage(role="tool", content="pending", tool_call_id="tc1")
+    )
+    session.add_tool_call_to_current_turn(_raw_tc2)
+
+    assert len(session.messages) == 3  # user, assistant, tool — no extra assistant
+    assert session.messages[1].tool_calls == [_raw_tc, _raw_tc2]
+
+
+def test_to_openai_messages_merges_split_assistants():
+    """End-to-end: session with split assistants produces valid OpenAI messages."""
+    session = ChatSession.new(user_id="u")
+    session.messages = [
+        ChatMessage(role="user", content="build agent"),
+        ChatMessage(role="assistant", content="Let me build that"),
+        ChatMessage(
+            role="assistant",
+            content="",
+            tool_calls=[
+                {
+                    "id": "tc1",
+                    "type": "function",
+                    "function": {"name": "create_agent", "arguments": "{}"},
+                }
+            ],
+        ),
+        ChatMessage(role="tool", content="done", tool_call_id="tc1"),
+        ChatMessage(role="assistant", content="Saved!"),
+        ChatMessage(role="user", content="show me an example run"),
+    ]
+    openai_msgs = session.to_openai_messages()
+
+    # The two consecutive assistants at index 1,2 should be merged
+    roles = [m["role"] for m in openai_msgs]
+    assert roles == ["user", "assistant", "tool", "assistant", "user"]
+
+    # The merged assistant should have both content and tool_calls
+    merged = cast(ChatCompletionAssistantMessageParam, openai_msgs[1])
+    assert merged.get("content") == "Let me build that"
+    tc_list = merged.get("tool_calls")
+    assert tc_list is not None and len(list(tc_list)) == 1
+    assert list(tc_list)[0]["id"] == "tc1"

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

@@ -10,6 +10,8 @@ from typing import Any
 
 from pydantic import BaseModel, Field
 
+from backend.util.json import dumps as json_dumps
+
 
 class ResponseType(str, Enum):
     """Types of streaming responses following AI SDK protocol."""
@@ -18,6 +20,10 @@ class ResponseType(str, Enum):
     START = "start"
     FINISH = "finish"
 
+    # Step lifecycle (one LLM API call within a message)
+    START_STEP = "start-step"
+    FINISH_STEP = "finish-step"
+
     # Text streaming
     TEXT_START = "text-start"
     TEXT_DELTA = "text-delta"
@@ -57,6 +63,16 @@ class StreamStart(StreamBaseResponse):
         description="Task ID for SSE reconnection. Clients can reconnect using GET /tasks/{taskId}/stream",
     )
 
+    def to_sse(self) -> str:
+        """Convert to SSE format, excluding non-protocol fields like taskId."""
+        import json
+
+        data: dict[str, Any] = {
+            "type": self.type.value,
+            "messageId": self.messageId,
+        }
+        return f"data: {json.dumps(data)}\n\n"
+
 
 class StreamFinish(StreamBaseResponse):
     """End of message/stream."""
@@ -64,6 +80,26 @@ class StreamFinish(StreamBaseResponse):
     type: ResponseType = ResponseType.FINISH
 
 
+class StreamStartStep(StreamBaseResponse):
+    """Start of a step (one LLM API call within a message).
+
+    The AI SDK uses this to add a step-start boundary to message.parts,
+    enabling visual separation between multiple LLM calls in a single message.
+    """
+
+    type: ResponseType = ResponseType.START_STEP
+
+
+class StreamFinishStep(StreamBaseResponse):
+    """End of a step (one LLM API call within a message).
+
+    The AI SDK uses this to reset activeTextParts and activeReasoningParts,
+    so the next LLM call in a tool-call continuation starts with clean state.
+    """
+
+    type: ResponseType = ResponseType.FINISH_STEP
+
+
 # ========== Text Streaming ==========
 
 
@@ -117,7 +153,7 @@ class StreamToolOutputAvailable(StreamBaseResponse):
     type: ResponseType = ResponseType.TOOL_OUTPUT_AVAILABLE
     toolCallId: str = Field(..., description="Tool call ID this responds to")
     output: str | dict[str, Any] = Field(..., description="Tool execution output")
-    # Additional fields for internal use (not part of AI SDK spec but useful)
+    # Keep these for internal backend use
     toolName: str | None = Field(
         default=None, description="Name of the tool that was executed"
     )
@@ -125,6 +161,17 @@ class StreamToolOutputAvailable(StreamBaseResponse):
         default=True, description="Whether the tool execution succeeded"
     )
 
+    def to_sse(self) -> str:
+        """Convert to SSE format, excluding non-spec fields."""
+        import json
+
+        data = {
+            "type": self.type.value,
+            "toolCallId": self.toolCallId,
+            "output": self.output,
+        }
+        return f"data: {json.dumps(data)}\n\n"
+
 
 # ========== Other ==========
 
@@ -148,6 +195,18 @@ class StreamError(StreamBaseResponse):
         default=None, description="Additional error details"
     )
 
+    def to_sse(self) -> str:
+        """Convert to SSE format, only emitting fields required by AI SDK protocol.
+
+        The AI SDK uses z.strictObject({type, errorText}) which rejects
+        any extra fields like `code` or `details`.
+        """
+        data = {
+            "type": self.type.value,
+            "errorText": self.errorText,
+        }
+        return f"data: {json_dumps(data)}\n\n"
+
 
 class StreamHeartbeat(StreamBaseResponse):
     """Heartbeat to keep SSE connection alive during long-running operations.

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

@@ -6,7 +6,7 @@ from collections.abc import AsyncGenerator
 from typing import Annotated
 
 from autogpt_libs import auth
-from fastapi import APIRouter, Depends, Header, HTTPException, Query, Security
+from fastapi import APIRouter, Depends, Header, HTTPException, Query, Response, Security
 from fastapi.responses import StreamingResponse
 from pydantic import BaseModel
 
@@ -17,7 +17,29 @@ from . import stream_registry
 from .completion_handler import process_operation_failure, process_operation_success
 from .config import ChatConfig
 from .model import ChatSession, create_chat_session, get_chat_session, get_user_sessions
-from .response_model import StreamFinish, StreamHeartbeat, StreamStart
+from .response_model import StreamFinish, StreamHeartbeat
+from .tools.models import (
+    AgentDetailsResponse,
+    AgentOutputResponse,
+    AgentPreviewResponse,
+    AgentSavedResponse,
+    AgentsFoundResponse,
+    BlockListResponse,
+    BlockOutputResponse,
+    ClarificationNeededResponse,
+    DocPageResponse,
+    DocSearchResultsResponse,
+    ErrorResponse,
+    ExecutionStartedResponse,
+    InputValidationErrorResponse,
+    NeedLoginResponse,
+    NoResultsResponse,
+    OperationInProgressResponse,
+    OperationPendingResponse,
+    OperationStartedResponse,
+    SetupRequirementsResponse,
+    UnderstandingUpdatedResponse,
+)
 
 config = ChatConfig()
 
@@ -266,12 +288,36 @@ async def stream_chat_post(
 
     """
     import asyncio
+    import time
+
+    stream_start_time = time.perf_counter()
+    log_meta = {"component": "ChatStream", "session_id": session_id}
+    if user_id:
+        log_meta["user_id"] = user_id
+
+    logger.info(
+        f"[TIMING] stream_chat_post STARTED, session={session_id}, "
+        f"user={user_id}, message_len={len(request.message)}",
+        extra={"json_fields": log_meta},
+    )
 
     session = await _validate_and_get_session(session_id, user_id)
+    logger.info(
+        f"[TIMING] session validated in {(time.perf_counter() - stream_start_time)*1000:.1f}ms",
+        extra={
+            "json_fields": {
+                **log_meta,
+                "duration_ms": (time.perf_counter() - stream_start_time) * 1000,
+            }
+        },
+    )
 
     # Create a task in the stream registry for reconnection support
     task_id = str(uuid_module.uuid4())
     operation_id = str(uuid_module.uuid4())
+    log_meta["task_id"] = task_id
+
+    task_create_start = time.perf_counter()
     await stream_registry.create_task(
         task_id=task_id,
         session_id=session_id,
@@ -280,14 +326,28 @@ async def stream_chat_post(
         tool_name="chat",
         operation_id=operation_id,
     )
+    logger.info(
+        f"[TIMING] create_task completed in {(time.perf_counter() - task_create_start)*1000:.1f}ms",
+        extra={
+            "json_fields": {
+                **log_meta,
+                "duration_ms": (time.perf_counter() - task_create_start) * 1000,
+            }
+        },
+    )
 
     # Background task that runs the AI generation independently of SSE connection
     async def run_ai_generation():
-        try:
+        import time as time_module
-            # Emit a start event with task_id for reconnection
-            start_chunk = StreamStart(messageId=task_id, taskId=task_id)
-            await stream_registry.publish_chunk(task_id, start_chunk)
 
+        gen_start_time = time_module.perf_counter()
+        logger.info(
+            f"[TIMING] run_ai_generation STARTED, task={task_id}, session={session_id}, user={user_id}",
+            extra={"json_fields": log_meta},
+        )
+        first_chunk_time, ttfc = None, None
+        chunk_count = 0
+        try:
             async for chunk in chat_service.stream_chat_completion(
                 session_id,
                 request.message,
@@ -295,25 +355,79 @@ async def stream_chat_post(
                 user_id=user_id,
                 session=session,  # Pass pre-fetched session to avoid double-fetch
                 context=request.context,
+                _task_id=task_id,  # Pass task_id so service emits start with taskId for reconnection
             ):
+                chunk_count += 1
+                if first_chunk_time is None:
+                    first_chunk_time = time_module.perf_counter()
+                    ttfc = first_chunk_time - gen_start_time
+                    logger.info(
+                        f"[TIMING] FIRST AI CHUNK at {ttfc:.2f}s, type={type(chunk).__name__}",
+                        extra={
+                            "json_fields": {
+                                **log_meta,
+                                "chunk_type": type(chunk).__name__,
+                                "time_to_first_chunk_ms": ttfc * 1000,
+                            }
+                        },
+                    )
                 # Write to Redis (subscribers will receive via XREAD)
                 await stream_registry.publish_chunk(task_id, chunk)
 
-            # Mark task as completed
+            gen_end_time = time_module.perf_counter()
+            total_time = (gen_end_time - gen_start_time) * 1000
+            logger.info(
+                f"[TIMING] run_ai_generation FINISHED in {total_time/1000:.1f}s; "
+                f"task={task_id}, session={session_id}, "
+                f"ttfc={ttfc or -1:.2f}s, n_chunks={chunk_count}",
+                extra={
+                    "json_fields": {
+                        **log_meta,
+                        "total_time_ms": total_time,
+                        "time_to_first_chunk_ms": (
+                            ttfc * 1000 if ttfc is not None else None
+                        ),
+                        "n_chunks": chunk_count,
+                    }
+                },
+            )
             await stream_registry.mark_task_completed(task_id, "completed")
         except Exception as e:
+            elapsed = time_module.perf_counter() - gen_start_time
             logger.error(
-                f"Error in background AI generation for session {session_id}: {e}"
+                f"[TIMING] run_ai_generation ERROR after {elapsed:.2f}s: {e}",
+                extra={
+                    "json_fields": {
+                        **log_meta,
+                        "elapsed_ms": elapsed * 1000,
+                        "error": str(e),
+                    }
+                },
             )
             await stream_registry.mark_task_completed(task_id, "failed")
 
     # Start the AI generation in a background task
     bg_task = asyncio.create_task(run_ai_generation())
     await stream_registry.set_task_asyncio_task(task_id, bg_task)
+    setup_time = (time.perf_counter() - stream_start_time) * 1000
+    logger.info(
+        f"[TIMING] Background task started, setup={setup_time:.1f}ms",
+        extra={"json_fields": {**log_meta, "setup_time_ms": setup_time}},
+    )
 
     # SSE endpoint that subscribes to the task's stream
     async def event_generator() -> AsyncGenerator[str, None]:
+        import time as time_module
+
+        event_gen_start = time_module.perf_counter()
+        logger.info(
+            f"[TIMING] event_generator STARTED, task={task_id}, session={session_id}, "
+            f"user={user_id}",
+            extra={"json_fields": log_meta},
+        )
         subscriber_queue = None
+        first_chunk_yielded = False
+        chunks_yielded = 0
         try:
             # Subscribe to the task stream (this replays existing messages + live updates)
             subscriber_queue = await stream_registry.subscribe_to_task(
@@ -328,22 +442,70 @@ async def stream_chat_post(
                 return
 
             # Read from the subscriber queue and yield to SSE
+            logger.info(
+                "[TIMING] Starting to read from subscriber_queue",
+                extra={"json_fields": log_meta},
+            )
             while True:
                 try:
                     chunk = await asyncio.wait_for(subscriber_queue.get(), timeout=30.0)
+                    chunks_yielded += 1
+
+                    if not first_chunk_yielded:
+                        first_chunk_yielded = True
+                        elapsed = time_module.perf_counter() - event_gen_start
+                        logger.info(
+                            f"[TIMING] FIRST CHUNK from queue at {elapsed:.2f}s, "
+                            f"type={type(chunk).__name__}",
+                            extra={
+                                "json_fields": {
+                                    **log_meta,
+                                    "chunk_type": type(chunk).__name__,
+                                    "elapsed_ms": elapsed * 1000,
+                                }
+                            },
+                        )
+
                     yield chunk.to_sse()
 
                     # Check for finish signal
                     if isinstance(chunk, StreamFinish):
+                        total_time = time_module.perf_counter() - event_gen_start
+                        logger.info(
+                            f"[TIMING] StreamFinish received in {total_time:.2f}s; "
+                            f"n_chunks={chunks_yielded}",
+                            extra={
+                                "json_fields": {
+                                    **log_meta,
+                                    "chunks_yielded": chunks_yielded,
+                                    "total_time_ms": total_time * 1000,
+                                }
+                            },
+                        )
                         break
                 except asyncio.TimeoutError:
-                    # Send heartbeat to keep connection alive
                     yield StreamHeartbeat().to_sse()
 
         except GeneratorExit:
+            logger.info(
+                f"[TIMING] GeneratorExit (client disconnected), chunks={chunks_yielded}",
+                extra={
+                    "json_fields": {
+                        **log_meta,
+                        "chunks_yielded": chunks_yielded,
+                        "reason": "client_disconnect",
+                    }
+                },
+            )
             pass  # Client disconnected - background task continues
         except Exception as e:
-            logger.error(f"Error in SSE stream for task {task_id}: {e}")
+            elapsed = (time_module.perf_counter() - event_gen_start) * 1000
+            logger.error(
+                f"[TIMING] event_generator ERROR after {elapsed:.1f}ms: {e}",
+                extra={
+                    "json_fields": {**log_meta, "elapsed_ms": elapsed, "error": str(e)}
+                },
+            )
         finally:
             # Unsubscribe when client disconnects or stream ends to prevent resource leak
             if subscriber_queue is not None:
@@ -357,6 +519,18 @@ async def stream_chat_post(
                         exc_info=True,
                     )
             # AI SDK protocol termination - always yield even if unsubscribe fails
+            total_time = time_module.perf_counter() - event_gen_start
+            logger.info(
+                f"[TIMING] event_generator FINISHED in {total_time:.2f}s; "
+                f"task={task_id}, session={session_id}, n_chunks={chunks_yielded}",
+                extra={
+                    "json_fields": {
+                        **log_meta,
+                        "total_time_ms": total_time * 1000,
+                        "chunks_yielded": chunks_yielded,
+                    }
+                },
+            )
             yield "data: [DONE]\n\n"
 
     return StreamingResponse(
@@ -374,63 +548,90 @@ async def stream_chat_post(
 @router.get(
     "/sessions/{session_id}/stream",
 )
-async def stream_chat_get(
+async def resume_session_stream(
     session_id: str,
-    message: Annotated[str, Query(min_length=1, max_length=10000)],
     user_id: str | None = Depends(auth.get_user_id),
-    is_user_message: bool = Query(default=True),
 ):
     """
-    Stream chat responses for a session (GET - legacy endpoint).
+    Resume an active stream for a session.
 
-    Streams the AI/completion responses in real time over Server-Sent Events (SSE), including:
+    Called by the AI SDK's ``useChat(resume: true)`` on page load.
-      - Text fragments as they are generated
+    Checks for an active (in-progress) task on the session and either replays
-      - Tool call UI elements (if invoked)
+    the full SSE stream or returns 204 No Content if nothing is running.
-      - Tool execution results
 
     Args:
-        session_id: The chat session identifier to associate with the streamed messages.
+        session_id: The chat session identifier.
-        message: The user's new message to process.
         user_id: Optional authenticated user ID.
-        is_user_message: Whether the message is a user message.
-    Returns:
-        StreamingResponse: SSE-formatted response chunks.
 
+    Returns:
+        StreamingResponse (SSE) when an active stream exists,
+        or 204 No Content when there is nothing to resume.
     """
-    session = await _validate_and_get_session(session_id, user_id)
+    import asyncio
+
+    active_task, _last_id = await stream_registry.get_active_task_for_session(
+        session_id, user_id
+    )
+
+    if not active_task:
+        return Response(status_code=204)
+
+    subscriber_queue = await stream_registry.subscribe_to_task(
+        task_id=active_task.task_id,
+        user_id=user_id,
+        last_message_id="0-0",  # Full replay so useChat rebuilds the message
+    )
+
+    if subscriber_queue is None:
+        return Response(status_code=204)
 
     async def event_generator() -> AsyncGenerator[str, None]:
         chunk_count = 0
         first_chunk_type: str | None = None
-        async for chunk in chat_service.stream_chat_completion(
+        try:
-            session_id,
+            while True:
-            message,
+                try:
-            is_user_message=is_user_message,
+                    chunk = await asyncio.wait_for(subscriber_queue.get(), timeout=30.0)
-            user_id=user_id,
+                    if chunk_count < 3:
-            session=session,  # Pass pre-fetched session to avoid double-fetch
+                        logger.info(
-        ):
+                            "Resume stream chunk",
-            if chunk_count < 3:
+                            extra={
-                logger.info(
+                                "session_id": session_id,
-                    "Chat stream chunk",
+                                "chunk_type": str(chunk.type),
-                    extra={
+                            },
-                        "session_id": session_id,
+                        )
-                        "chunk_type": str(chunk.type),
+                    if not first_chunk_type:
-                    },
+                        first_chunk_type = str(chunk.type)
+                    chunk_count += 1
+                    yield chunk.to_sse()
+
+                    if isinstance(chunk, StreamFinish):
+                        break
+                except asyncio.TimeoutError:
+                    yield StreamHeartbeat().to_sse()
+        except GeneratorExit:
+            pass
+        except Exception as e:
+            logger.error(f"Error in resume stream for session {session_id}: {e}")
+        finally:
+            try:
+                await stream_registry.unsubscribe_from_task(
+                    active_task.task_id, subscriber_queue
                 )
-            if not first_chunk_type:
+            except Exception as unsub_err:
-                first_chunk_type = str(chunk.type)
+                logger.error(
-            chunk_count += 1
+                    f"Error unsubscribing from task {active_task.task_id}: {unsub_err}",
-            yield chunk.to_sse()
+                    exc_info=True,
-        logger.info(
+                )
-            "Chat stream completed",
+            logger.info(
-            extra={
+                "Resume stream completed",
-                "session_id": session_id,
+                extra={
-                "chunk_count": chunk_count,
+                    "session_id": session_id,
-                "first_chunk_type": first_chunk_type,
+                    "n_chunks": chunk_count,
-            },
+                    "first_chunk_type": first_chunk_type,
-        )
+                },
-        # AI SDK protocol termination
+            )
-        yield "data: [DONE]\n\n"
+            yield "data: [DONE]\n\n"
 
     return StreamingResponse(
         event_generator(),
@@ -438,8 +639,8 @@ async def stream_chat_get(
         headers={
             "Cache-Control": "no-cache",
             "Connection": "keep-alive",
-            "X-Accel-Buffering": "no",  # Disable nginx buffering
+            "X-Accel-Buffering": "no",
-            "x-vercel-ai-ui-message-stream": "v1",  # AI SDK protocol header
+            "x-vercel-ai-ui-message-stream": "v1",
         },
     )
 
@@ -751,3 +952,42 @@ async def health_check() -> dict:
         "service": "chat",
         "version": "0.1.0",
     }
+
+
+# ========== Schema Export (for OpenAPI / Orval codegen) ==========
+
+ToolResponseUnion = (
+    AgentsFoundResponse
+    | NoResultsResponse
+    | AgentDetailsResponse
+    | SetupRequirementsResponse
+    | ExecutionStartedResponse
+    | NeedLoginResponse
+    | ErrorResponse
+    | InputValidationErrorResponse
+    | AgentOutputResponse
+    | UnderstandingUpdatedResponse
+    | AgentPreviewResponse
+    | AgentSavedResponse
+    | ClarificationNeededResponse
+    | BlockListResponse
+    | BlockOutputResponse
+    | DocSearchResultsResponse
+    | DocPageResponse
+    | OperationStartedResponse
+    | OperationPendingResponse
+    | OperationInProgressResponse
+)
+
+
+@router.get(
+    "/schema/tool-responses",
+    response_model=ToolResponseUnion,
+    include_in_schema=True,
+    summary="[Dummy] Tool response type export for codegen",
+    description="This endpoint is not meant to be called. It exists solely to "
+    "expose tool response models in the OpenAPI schema for frontend codegen.",
+)
+async def _tool_response_schema() -> ToolResponseUnion:  # type: ignore[return]
+    """Never called at runtime. Exists only so Orval generates TS types."""
+    raise HTTPException(status_code=501, detail="Schema-only endpoint")

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

@@ -33,7 +33,7 @@ from backend.data.understanding import (
     get_business_understanding,
 )
 from backend.util.exceptions import NotFoundError
-from backend.util.settings import Settings
+from backend.util.settings import AppEnvironment, Settings
 
 from . import db as chat_db
 from . import stream_registry
@@ -52,8 +52,10 @@ from .response_model import (
     StreamBaseResponse,
     StreamError,
     StreamFinish,
+    StreamFinishStep,
     StreamHeartbeat,
     StreamStart,
+    StreamStartStep,
     StreamTextDelta,
     StreamTextEnd,
     StreamTextStart,
@@ -222,8 +224,18 @@ async def _get_system_prompt_template(context: str) -> str:
         try:
             # cache_ttl_seconds=0 disables SDK caching to always get the latest prompt
             # Use asyncio.to_thread to avoid blocking the event loop
+            # In non-production environments, fetch the latest prompt version
+            # instead of the production-labeled version for easier testing
+            label = (
+                None
+                if settings.config.app_env == AppEnvironment.PRODUCTION
+                else "latest"
+            )
             prompt = await asyncio.to_thread(
-                langfuse.get_prompt, config.langfuse_prompt_name, cache_ttl_seconds=0
+                langfuse.get_prompt,
+                config.langfuse_prompt_name,
+                label=label,
+                cache_ttl_seconds=0,
             )
             return prompt.compile(users_information=context)
         except Exception as e:
@@ -341,6 +353,10 @@ async def stream_chat_completion(
     retry_count: int = 0,
     session: ChatSession | None = None,
     context: dict[str, str] | None = None,  # {url: str, content: str}
+    _continuation_message_id: (
+        str | None
+    ) = None,  # Internal: reuse message ID for tool call continuations
+    _task_id: str | None = None,  # Internal: task ID for SSE reconnection support
 ) -> AsyncGenerator[StreamBaseResponse, None]:
     """Main entry point for streaming chat completions with database handling.
 
@@ -361,21 +377,45 @@ async def stream_chat_completion(
         ValueError: If max_context_messages is exceeded
 
     """
+    completion_start = time.monotonic()
+
+    # Build log metadata for structured logging
+    log_meta = {"component": "ChatService", "session_id": session_id}
+    if user_id:
+        log_meta["user_id"] = user_id
+
     logger.info(
-        f"Streaming chat completion for session {session_id} for message {message} and user id {user_id}. Message is user message: {is_user_message}"
+        f"[TIMING] stream_chat_completion STARTED, session={session_id}, user={user_id}, "
+        f"message_len={len(message) if message else 0}, is_user={is_user_message}",
+        extra={
+            "json_fields": {
+                **log_meta,
+                "message_len": len(message) if message else 0,
+                "is_user_message": is_user_message,
+            }
+        },
     )
 
     # Only fetch from Redis if session not provided (initial call)
     if session is None:
+        fetch_start = time.monotonic()
         session = await get_chat_session(session_id, user_id)
+        fetch_time = (time.monotonic() - fetch_start) * 1000
         logger.info(
-            f"Fetched session from Redis: {session.session_id if session else 'None'}, "
+            f"[TIMING] get_chat_session took {fetch_time:.1f}ms, "
-            f"message_count={len(session.messages) if session else 0}"
+            f"n_messages={len(session.messages) if session else 0}",
+            extra={
+                "json_fields": {
+                    **log_meta,
+                    "duration_ms": fetch_time,
+                    "n_messages": len(session.messages) if session else 0,
+                }
+            },
         )
     else:
         logger.info(
-            f"Using provided session object: {session.session_id}, "
+            f"[TIMING] Using provided session, messages={len(session.messages)}",
-            f"message_count={len(session.messages)}"
+            extra={"json_fields": {**log_meta, "n_messages": len(session.messages)}},
         )
 
     if not session:
@@ -396,17 +436,25 @@ async def stream_chat_completion(
 
         # Track user message in PostHog
         if is_user_message:
+            posthog_start = time.monotonic()
             track_user_message(
                 user_id=user_id,
                 session_id=session_id,
                 message_length=len(message),
             )
+            posthog_time = (time.monotonic() - posthog_start) * 1000
+            logger.info(
+                f"[TIMING] track_user_message took {posthog_time:.1f}ms",
+                extra={"json_fields": {**log_meta, "duration_ms": posthog_time}},
+            )
 
-    logger.info(
+    upsert_start = time.monotonic()
-        f"Upserting session: {session.session_id} with user id {session.user_id}, "
-        f"message_count={len(session.messages)}"
-    )
     session = await upsert_chat_session(session)
+    upsert_time = (time.monotonic() - upsert_start) * 1000
+    logger.info(
+        f"[TIMING] upsert_chat_session took {upsert_time:.1f}ms",
+        extra={"json_fields": {**log_meta, "duration_ms": upsert_time}},
+    )
     assert session, "Session not found"
 
     # Generate title for new sessions on first user message (non-blocking)
@@ -444,7 +492,13 @@ async def stream_chat_completion(
             asyncio.create_task(_update_title())
 
     # Build system prompt with business understanding
+    prompt_start = time.monotonic()
     system_prompt, understanding = await _build_system_prompt(user_id)
+    prompt_time = (time.monotonic() - prompt_start) * 1000
+    logger.info(
+        f"[TIMING] _build_system_prompt took {prompt_time:.1f}ms",
+        extra={"json_fields": {**log_meta, "duration_ms": prompt_time}},
+    )
 
     # Initialize variables for streaming
     assistant_response = ChatMessage(
@@ -469,13 +523,27 @@ async def stream_chat_completion(
     # Generate unique IDs for AI SDK protocol
     import uuid as uuid_module
 
-    message_id = str(uuid_module.uuid4())
+    is_continuation = _continuation_message_id is not None
+    message_id = _continuation_message_id or str(uuid_module.uuid4())
     text_block_id = str(uuid_module.uuid4())
 
-    # Yield message start
+    # Only yield message start for the initial call, not for continuations.
-    yield StreamStart(messageId=message_id)
+    setup_time = (time.monotonic() - completion_start) * 1000
+    logger.info(
+        f"[TIMING] Setup complete, yielding StreamStart at {setup_time:.1f}ms",
+        extra={"json_fields": {**log_meta, "setup_time_ms": setup_time}},
+    )
+    if not is_continuation:
+        yield StreamStart(messageId=message_id, taskId=_task_id)
+
+    # Emit start-step before each LLM call (AI SDK uses this to add step boundaries)
+    yield StreamStartStep()
 
     try:
+        logger.info(
+            "[TIMING] Calling _stream_chat_chunks",
+            extra={"json_fields": log_meta},
+        )
         async for chunk in _stream_chat_chunks(
             session=session,
             tools=tools,
@@ -575,6 +643,10 @@ async def stream_chat_completion(
                     )
                 yield chunk
             elif isinstance(chunk, StreamFinish):
+                if has_done_tool_call:
+                    # Tool calls happened — close the step but don't send message-level finish.
+                    # The continuation will open a new step, and finish will come at the end.
+                    yield StreamFinishStep()
                 if not has_done_tool_call:
                     # Emit text-end before finish if we received text but haven't closed it
                     if has_received_text and not text_streaming_ended:
@@ -606,6 +678,8 @@ async def stream_chat_completion(
                             has_saved_assistant_message = True
 
                     has_yielded_end = True
+                    # Emit finish-step before finish (resets AI SDK text/reasoning state)
+                    yield StreamFinishStep()
                     yield chunk
             elif isinstance(chunk, StreamError):
                 has_yielded_error = True
@@ -618,6 +692,9 @@ async def stream_chat_completion(
                         total_tokens=chunk.totalTokens,
                     )
                 )
+            elif isinstance(chunk, StreamHeartbeat):
+                # Pass through heartbeat to keep SSE connection alive
+                yield chunk
             else:
                 logger.error(f"Unknown chunk type: {type(chunk)}", exc_info=True)
 
@@ -652,6 +729,10 @@ async def stream_chat_completion(
             logger.info(
                 f"Retryable error encountered. Attempt {retry_count + 1}/{config.max_retries}"
             )
+            # Close the current step before retrying so the recursive call's
+            # StreamStartStep doesn't produce unbalanced step events.
+            if not has_yielded_end:
+                yield StreamFinishStep()
             should_retry = True
         else:
             # Non-retryable error or max retries exceeded
@@ -687,6 +768,7 @@ async def stream_chat_completion(
                 error_response = StreamError(errorText=error_message)
                 yield error_response
             if not has_yielded_end:
+                yield StreamFinishStep()
                 yield StreamFinish()
             return
 
@@ -701,6 +783,8 @@ async def stream_chat_completion(
             retry_count=retry_count + 1,
             session=session,
             context=context,
+            _continuation_message_id=message_id,  # Reuse message ID since start was already sent
+            _task_id=_task_id,
         ):
             yield chunk
         return  # Exit after retry to avoid double-saving in finally block
@@ -716,9 +800,13 @@ async def stream_chat_completion(
         # Build the messages list in the correct order
         messages_to_save: list[ChatMessage] = []
 
-        # Add assistant message with tool_calls if any
+        # Add assistant message with tool_calls if any.
+        # Use extend (not assign) to preserve tool_calls already added by
+        # _yield_tool_call for long-running tools.
         if accumulated_tool_calls:
-            assistant_response.tool_calls = accumulated_tool_calls
+            if not assistant_response.tool_calls:
+                assistant_response.tool_calls = []
+            assistant_response.tool_calls.extend(accumulated_tool_calls)
             logger.info(
                 f"Added {len(accumulated_tool_calls)} tool calls to assistant message"
             )
@@ -770,6 +858,8 @@ async def stream_chat_completion(
             session=session,  # Pass session object to avoid Redis refetch
             context=context,
             tool_call_response=str(tool_response_messages),
+            _continuation_message_id=message_id,  # Reuse message ID to avoid duplicates
+            _task_id=_task_id,
         ):
             yield chunk
 
@@ -880,9 +970,21 @@ async def _stream_chat_chunks(
         SSE formatted JSON response objects
 
     """
+    import time as time_module
+
+    stream_chunks_start = time_module.perf_counter()
     model = config.model
 
-    logger.info("Starting pure chat stream")
+    # Build log metadata for structured logging
+    log_meta = {"component": "ChatService", "session_id": session.session_id}
+    if session.user_id:
+        log_meta["user_id"] = session.user_id
+
+    logger.info(
+        f"[TIMING] _stream_chat_chunks STARTED, session={session.session_id}, "
+        f"user={session.user_id}, n_messages={len(session.messages)}",
+        extra={"json_fields": {**log_meta, "n_messages": len(session.messages)}},
+    )
 
     messages = session.to_openai_messages()
     if system_prompt:
@@ -893,12 +995,18 @@ async def _stream_chat_chunks(
         messages = [system_message] + messages
 
     # Apply context window management
+    context_start = time_module.perf_counter()
     context_result = await _manage_context_window(
         messages=messages,
         model=model,
         api_key=config.api_key,
         base_url=config.base_url,
     )
+    context_time = (time_module.perf_counter() - context_start) * 1000
+    logger.info(
+        f"[TIMING] _manage_context_window took {context_time:.1f}ms",
+        extra={"json_fields": {**log_meta, "duration_ms": context_time}},
+    )
 
     if context_result.error:
         if "System prompt dropped" in context_result.error:
@@ -933,9 +1041,19 @@ async def _stream_chat_chunks(
 
         while retry_count <= MAX_RETRIES:
             try:
+                elapsed = (time_module.perf_counter() - stream_chunks_start) * 1000
+                retry_info = (
+                    f" (retry {retry_count}/{MAX_RETRIES})" if retry_count > 0 else ""
+                )
                 logger.info(
-                    f"Creating OpenAI chat completion stream..."
+                    f"[TIMING] Creating OpenAI stream at {elapsed:.1f}ms{retry_info}",
-                    f"{f' (retry {retry_count}/{MAX_RETRIES})' if retry_count > 0 else ''}"
+                    extra={
+                        "json_fields": {
+                            **log_meta,
+                            "elapsed_ms": elapsed,
+                            "retry_count": retry_count,
+                        }
+                    },
                 )
 
                 # Build extra_body for OpenRouter tracing and PostHog analytics
@@ -952,6 +1070,11 @@ async def _stream_chat_chunks(
                         :128
                     ]  # OpenRouter limit
 
+                # Enable adaptive thinking for Anthropic models via OpenRouter
+                if config.thinking_enabled and "anthropic" in model.lower():
+                    extra_body["reasoning"] = {"enabled": True}
+
+                api_call_start = time_module.perf_counter()
                 stream = await client.chat.completions.create(
                     model=model,
                     messages=cast(list[ChatCompletionMessageParam], messages),
@@ -961,6 +1084,11 @@ async def _stream_chat_chunks(
                     stream_options=ChatCompletionStreamOptionsParam(include_usage=True),
                     extra_body=extra_body,
                 )
+                api_init_time = (time_module.perf_counter() - api_call_start) * 1000
+                logger.info(
+                    f"[TIMING] OpenAI stream object returned in {api_init_time:.1f}ms",
+                    extra={"json_fields": {**log_meta, "duration_ms": api_init_time}},
+                )
 
                 # Variables to accumulate tool calls
                 tool_calls: list[dict[str, Any]] = []
@@ -971,10 +1099,13 @@ async def _stream_chat_chunks(
 
                 # Track if we've started the text block
                 text_started = False
+                first_content_chunk = True
+                chunk_count = 0
 
                 # Process the stream
                 chunk: ChatCompletionChunk
                 async for chunk in stream:
+                    chunk_count += 1
                     if chunk.usage:
                         yield StreamUsage(
                             promptTokens=chunk.usage.prompt_tokens,
@@ -997,6 +1128,23 @@ async def _stream_chat_chunks(
                             if not text_started and text_block_id:
                                 yield StreamTextStart(id=text_block_id)
                                 text_started = True
+                            # Log timing for first content chunk
+                            if first_content_chunk:
+                                first_content_chunk = False
+                                ttfc = (
+                                    time_module.perf_counter() - api_call_start
+                                ) * 1000
+                                logger.info(
+                                    f"[TIMING] FIRST CONTENT CHUNK at {ttfc:.1f}ms "
+                                    f"(since API call), n_chunks={chunk_count}",
+                                    extra={
+                                        "json_fields": {
+                                            **log_meta,
+                                            "time_to_first_chunk_ms": ttfc,
+                                            "n_chunks": chunk_count,
+                                        }
+                                    },
+                                )
                             # Stream the text delta
                             text_response = StreamTextDelta(
                                 id=text_block_id or "",
@@ -1053,7 +1201,21 @@ async def _stream_chat_chunks(
                                         toolName=tool_calls[idx]["function"]["name"],
                                     )
                                     emitted_start_for_idx.add(idx)
-                logger.info(f"Stream complete. Finish reason: {finish_reason}")
+                stream_duration = time_module.perf_counter() - api_call_start
+                logger.info(
+                    f"[TIMING] OpenAI stream COMPLETE, finish_reason={finish_reason}, "
+                    f"duration={stream_duration:.2f}s, "
+                    f"n_chunks={chunk_count}, n_tool_calls={len(tool_calls)}",
+                    extra={
+                        "json_fields": {
+                            **log_meta,
+                            "stream_duration_ms": stream_duration * 1000,
+                            "finish_reason": finish_reason,
+                            "n_chunks": chunk_count,
+                            "n_tool_calls": len(tool_calls),
+                        }
+                    },
+                )
 
                 # Yield all accumulated tool calls after the stream is complete
                 # This ensures all tool call arguments have been fully received
@@ -1073,6 +1235,12 @@ async def _stream_chat_chunks(
                         # Re-raise to trigger retry logic in the parent function
                         raise
 
+                total_time = (time_module.perf_counter() - stream_chunks_start) * 1000
+                logger.info(
+                    f"[TIMING] _stream_chat_chunks COMPLETED in {total_time/1000:.1f}s; "
+                    f"session={session.session_id}, user={session.user_id}",
+                    extra={"json_fields": {**log_meta, "total_time_ms": total_time}},
+                )
                 yield StreamFinish()
                 return
             except Exception as e:
@@ -1240,13 +1408,9 @@ async def _yield_tool_call(
                 operation_id=operation_id,
             )
 
-            # Save assistant message with tool_call FIRST (required by LLM)
+            # Attach the tool_call to the current turn's assistant message
-            assistant_message = ChatMessage(
+            # (or create one if this is a tool-only response with no text).
-                role="assistant",
+            session.add_tool_call_to_current_turn(tool_calls[yield_idx])
-                content="",
-                tool_calls=[tool_calls[yield_idx]],
-            )
-            session.messages.append(assistant_message)
 
             # Then save pending tool result
             pending_message = ChatMessage(
@@ -1552,6 +1716,7 @@ async def _execute_long_running_tool_with_streaming(
             task_id,
             StreamError(errorText=str(e)),
         )
+        await stream_registry.publish_chunk(task_id, StreamFinishStep())
         await stream_registry.publish_chunk(task_id, StreamFinish())
 
         await _update_pending_operation(
@@ -1668,6 +1833,10 @@ async def _generate_llm_continuation(
         if session_id:
             extra_body["session_id"] = session_id[:128]
 
+        # Enable adaptive thinking for Anthropic models via OpenRouter
+        if config.thinking_enabled and "anthropic" in config.model.lower():
+            extra_body["reasoning"] = {"enabled": True}
+
         retry_count = 0
         last_error: Exception | None = None
         response = None
@@ -1798,6 +1967,10 @@ async def _generate_llm_continuation_with_streaming(
         if session_id:
             extra_body["session_id"] = session_id[:128]
 
+        # Enable adaptive thinking for Anthropic models via OpenRouter
+        if config.thinking_enabled and "anthropic" in config.model.lower():
+            extra_body["reasoning"] = {"enabled": True}
+
         # Make streaming LLM call (no tools - just text response)
         from typing import cast
 
@@ -1809,6 +1982,7 @@ async def _generate_llm_continuation_with_streaming(
 
         # Publish start event
         await stream_registry.publish_chunk(task_id, StreamStart(messageId=message_id))
+        await stream_registry.publish_chunk(task_id, StreamStartStep())
         await stream_registry.publish_chunk(task_id, StreamTextStart(id=text_block_id))
 
         # Stream the response
@@ -1832,6 +2006,7 @@ async def _generate_llm_continuation_with_streaming(
 
         # Publish end events
         await stream_registry.publish_chunk(task_id, StreamTextEnd(id=text_block_id))
+        await stream_registry.publish_chunk(task_id, StreamFinishStep())
 
         if assistant_content:
             # Reload session from DB to avoid race condition with user messages
@@ -1873,4 +2048,5 @@ async def _generate_llm_continuation_with_streaming(
             task_id,
             StreamError(errorText=f"Failed to generate response: {e}"),
         )
+        await stream_registry.publish_chunk(task_id, StreamFinishStep())
         await stream_registry.publish_chunk(task_id, StreamFinish())

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

@@ -104,6 +104,24 @@ async def create_task(
     Returns:
         The created ActiveTask instance (metadata only)
     """
+    import time
+
+    start_time = time.perf_counter()
+
+    # Build log metadata for structured logging
+    log_meta = {
+        "component": "StreamRegistry",
+        "task_id": task_id,
+        "session_id": session_id,
+    }
+    if user_id:
+        log_meta["user_id"] = user_id
+
+    logger.info(
+        f"[TIMING] create_task STARTED, task={task_id}, session={session_id}, user={user_id}",
+        extra={"json_fields": log_meta},
+    )
+
     task = ActiveTask(
         task_id=task_id,
         session_id=session_id,
@@ -114,10 +132,18 @@ async def create_task(
     )
 
     # Store metadata in Redis
+    redis_start = time.perf_counter()
     redis = await get_redis_async()
+    redis_time = (time.perf_counter() - redis_start) * 1000
+    logger.info(
+        f"[TIMING] get_redis_async took {redis_time:.1f}ms",
+        extra={"json_fields": {**log_meta, "duration_ms": redis_time}},
+    )
+
     meta_key = _get_task_meta_key(task_id)
     op_key = _get_operation_mapping_key(operation_id)
 
+    hset_start = time.perf_counter()
     await redis.hset(  # type: ignore[misc]
         meta_key,
         mapping={
@@ -131,12 +157,22 @@ async def create_task(
             "created_at": task.created_at.isoformat(),
         },
     )
+    hset_time = (time.perf_counter() - hset_start) * 1000
+    logger.info(
+        f"[TIMING] redis.hset took {hset_time:.1f}ms",
+        extra={"json_fields": {**log_meta, "duration_ms": hset_time}},
+    )
+
     await redis.expire(meta_key, config.stream_ttl)
 
     # Create operation_id -> task_id mapping for webhook lookups
     await redis.set(op_key, task_id, ex=config.stream_ttl)
 
-    logger.debug(f"Created task {task_id} for session {session_id}")
+    total_time = (time.perf_counter() - start_time) * 1000
+    logger.info(
+        f"[TIMING] create_task COMPLETED in {total_time:.1f}ms; task={task_id}, session={session_id}",
+        extra={"json_fields": {**log_meta, "total_time_ms": total_time}},
+    )
 
     return task
 
@@ -156,26 +192,60 @@ async def publish_chunk(
     Returns:
         The Redis Stream message ID
     """
+    import time
+
+    start_time = time.perf_counter()
+    chunk_type = type(chunk).__name__
     chunk_json = chunk.model_dump_json()
     message_id = "0-0"
 
+    # Build log metadata
+    log_meta = {
+        "component": "StreamRegistry",
+        "task_id": task_id,
+        "chunk_type": chunk_type,
+    }
+
     try:
         redis = await get_redis_async()
         stream_key = _get_task_stream_key(task_id)
 
         # Write to Redis Stream for persistence and real-time delivery
+        xadd_start = time.perf_counter()
         raw_id = await redis.xadd(
             stream_key,
             {"data": chunk_json},
             maxlen=config.stream_max_length,
         )
+        xadd_time = (time.perf_counter() - xadd_start) * 1000
         message_id = raw_id if isinstance(raw_id, str) else raw_id.decode()
 
         # Set TTL on stream to match task metadata TTL
         await redis.expire(stream_key, config.stream_ttl)
+
+        total_time = (time.perf_counter() - start_time) * 1000
+        # Only log timing for significant chunks or slow operations
+        if (
+            chunk_type
+            in ("StreamStart", "StreamFinish", "StreamTextStart", "StreamTextEnd")
+            or total_time > 50
+        ):
+            logger.info(
+                f"[TIMING] publish_chunk {chunk_type} in {total_time:.1f}ms (xadd={xadd_time:.1f}ms)",
+                extra={
+                    "json_fields": {
+                        **log_meta,
+                        "total_time_ms": total_time,
+                        "xadd_time_ms": xadd_time,
+                        "message_id": message_id,
+                    }
+                },
+            )
     except Exception as e:
+        elapsed = (time.perf_counter() - start_time) * 1000
         logger.error(
-            f"Failed to publish chunk for task {task_id}: {e}",
+            f"[TIMING] Failed to publish chunk {chunk_type} after {elapsed:.1f}ms: {e}",
+            extra={"json_fields": {**log_meta, "elapsed_ms": elapsed, "error": str(e)}},
             exc_info=True,
         )
 
@@ -200,24 +270,61 @@ async def subscribe_to_task(
         An asyncio Queue that will receive stream chunks, or None if task not found
         or user doesn't have access
     """
+    import time
+
+    start_time = time.perf_counter()
+
+    # Build log metadata
+    log_meta = {"component": "StreamRegistry", "task_id": task_id}
+    if user_id:
+        log_meta["user_id"] = user_id
+
+    logger.info(
+        f"[TIMING] subscribe_to_task STARTED, task={task_id}, user={user_id}, last_msg={last_message_id}",
+        extra={"json_fields": {**log_meta, "last_message_id": last_message_id}},
+    )
+
+    redis_start = time.perf_counter()
     redis = await get_redis_async()
     meta_key = _get_task_meta_key(task_id)
     meta: dict[Any, Any] = await redis.hgetall(meta_key)  # type: ignore[misc]
+    hgetall_time = (time.perf_counter() - redis_start) * 1000
+    logger.info(
+        f"[TIMING] Redis hgetall took {hgetall_time:.1f}ms",
+        extra={"json_fields": {**log_meta, "duration_ms": hgetall_time}},
+    )
 
     if not meta:
-        logger.debug(f"Task {task_id} not found in Redis")
+        elapsed = (time.perf_counter() - start_time) * 1000
+        logger.info(
+            f"[TIMING] Task not found in Redis after {elapsed:.1f}ms",
+            extra={
+                "json_fields": {
+                    **log_meta,
+                    "elapsed_ms": elapsed,
+                    "reason": "task_not_found",
+                }
+            },
+        )
         return None
 
     # Note: Redis client uses decode_responses=True, so keys are strings
     task_status = meta.get("status", "")
     task_user_id = meta.get("user_id", "") or None
+    log_meta["session_id"] = meta.get("session_id", "")
 
     # Validate ownership - if task has an owner, requester must match
     if task_user_id:
         if user_id != task_user_id:
             logger.warning(
-                f"User {user_id} denied access to task {task_id} "
+                f"[TIMING] Access denied: user {user_id} tried to access task owned by {task_user_id}",
-                f"owned by {task_user_id}"
+                extra={
+                    "json_fields": {
+                        **log_meta,
+                        "task_owner": task_user_id,
+                        "reason": "access_denied",
+                    }
+                },
             )
             return None
 
@@ -225,7 +332,19 @@ async def subscribe_to_task(
     stream_key = _get_task_stream_key(task_id)
 
     # Step 1: Replay messages from Redis Stream
+    xread_start = time.perf_counter()
     messages = await redis.xread({stream_key: last_message_id}, block=0, count=1000)
+    xread_time = (time.perf_counter() - xread_start) * 1000
+    logger.info(
+        f"[TIMING] Redis xread (replay) took {xread_time:.1f}ms, status={task_status}",
+        extra={
+            "json_fields": {
+                **log_meta,
+                "duration_ms": xread_time,
+                "task_status": task_status,
+            }
+        },
+    )
 
     replayed_count = 0
     replay_last_id = last_message_id
@@ -244,19 +363,48 @@ async def subscribe_to_task(
                     except Exception as e:
                         logger.warning(f"Failed to replay message: {e}")
 
-    logger.debug(f"Task {task_id}: replayed {replayed_count} messages")
+    logger.info(
+        f"[TIMING] Replayed {replayed_count} messages, last_id={replay_last_id}",
+        extra={
+            "json_fields": {
+                **log_meta,
+                "n_messages_replayed": replayed_count,
+                "replay_last_id": replay_last_id,
+            }
+        },
+    )
 
     # Step 2: If task is still running, start stream listener for live updates
     if task_status == "running":
+        logger.info(
+            "[TIMING] Task still running, starting _stream_listener",
+            extra={"json_fields": {**log_meta, "task_status": task_status}},
+        )
         listener_task = asyncio.create_task(
-            _stream_listener(task_id, subscriber_queue, replay_last_id)
+            _stream_listener(task_id, subscriber_queue, replay_last_id, log_meta)
         )
         # Track listener task for cleanup on unsubscribe
         _listener_tasks[id(subscriber_queue)] = (task_id, listener_task)
     else:
         # Task is completed/failed - add finish marker
+        logger.info(
+            f"[TIMING] Task already {task_status}, adding StreamFinish",
+            extra={"json_fields": {**log_meta, "task_status": task_status}},
+        )
         await subscriber_queue.put(StreamFinish())
 
+    total_time = (time.perf_counter() - start_time) * 1000
+    logger.info(
+        f"[TIMING] subscribe_to_task COMPLETED in {total_time:.1f}ms; task={task_id}, "
+        f"n_messages_replayed={replayed_count}",
+        extra={
+            "json_fields": {
+                **log_meta,
+                "total_time_ms": total_time,
+                "n_messages_replayed": replayed_count,
+            }
+        },
+    )
     return subscriber_queue
 
 
@@ -264,6 +412,7 @@ async def _stream_listener(
     task_id: str,
     subscriber_queue: asyncio.Queue[StreamBaseResponse],
     last_replayed_id: str,
+    log_meta: dict | None = None,
 ) -> None:
     """Listen to Redis Stream for new messages using blocking XREAD.
 
@@ -274,10 +423,27 @@ async def _stream_listener(
         task_id: Task ID to listen for
         subscriber_queue: Queue to deliver messages to
         last_replayed_id: Last message ID from replay (continue from here)
+        log_meta: Structured logging metadata
     """
+    import time
+
+    start_time = time.perf_counter()
+
+    # Use provided log_meta or build minimal one
+    if log_meta is None:
+        log_meta = {"component": "StreamRegistry", "task_id": task_id}
+
+    logger.info(
+        f"[TIMING] _stream_listener STARTED, task={task_id}, last_id={last_replayed_id}",
+        extra={"json_fields": {**log_meta, "last_replayed_id": last_replayed_id}},
+    )
+
     queue_id = id(subscriber_queue)
     # Track the last successfully delivered message ID for recovery hints
     last_delivered_id = last_replayed_id
+    messages_delivered = 0
+    first_message_time = None
+    xread_count = 0
 
     try:
         redis = await get_redis_async()
@@ -287,9 +453,39 @@ async def _stream_listener(
         while True:
             # Block for up to 30 seconds waiting for new messages
             # This allows periodic checking if task is still running
+            xread_start = time.perf_counter()
+            xread_count += 1
             messages = await redis.xread(
                 {stream_key: current_id}, block=30000, count=100
             )
+            xread_time = (time.perf_counter() - xread_start) * 1000
+
+            if messages:
+                msg_count = sum(len(msgs) for _, msgs in messages)
+                logger.info(
+                    f"[TIMING] xread #{xread_count} returned {msg_count} messages in {xread_time:.1f}ms",
+                    extra={
+                        "json_fields": {
+                            **log_meta,
+                            "xread_count": xread_count,
+                            "n_messages": msg_count,
+                            "duration_ms": xread_time,
+                        }
+                    },
+                )
+            elif xread_time > 1000:
+                # Only log timeouts (30s blocking)
+                logger.info(
+                    f"[TIMING] xread #{xread_count} timeout after {xread_time:.1f}ms",
+                    extra={
+                        "json_fields": {
+                            **log_meta,
+                            "xread_count": xread_count,
+                            "duration_ms": xread_time,
+                            "reason": "timeout",
+                        }
+                    },
+                )
 
             if not messages:
                 # Timeout - check if task is still running
@@ -326,10 +522,30 @@ async def _stream_listener(
                                 )
                                 # Update last delivered ID on successful delivery
                                 last_delivered_id = current_id
+                                messages_delivered += 1
+                                if first_message_time is None:
+                                    first_message_time = time.perf_counter()
+                                    elapsed = (first_message_time - start_time) * 1000
+                                    logger.info(
+                                        f"[TIMING] FIRST live message at {elapsed:.1f}ms, type={type(chunk).__name__}",
+                                        extra={
+                                            "json_fields": {
+                                                **log_meta,
+                                                "elapsed_ms": elapsed,
+                                                "chunk_type": type(chunk).__name__,
+                                            }
+                                        },
+                                    )
                             except asyncio.TimeoutError:
                                 logger.warning(
-                                    f"Subscriber queue full for task {task_id}, "
+                                    f"[TIMING] Subscriber queue full, delivery timed out after {QUEUE_PUT_TIMEOUT}s",
-                                    f"message delivery timed out after {QUEUE_PUT_TIMEOUT}s"
+                                    extra={
+                                        "json_fields": {
+                                            **log_meta,
+                                            "timeout_s": QUEUE_PUT_TIMEOUT,
+                                            "reason": "queue_full",
+                                        }
+                                    },
                                 )
                                 # Send overflow error with recovery info
                                 try:
@@ -351,15 +567,44 @@ async def _stream_listener(
 
                             # Stop listening on finish
                             if isinstance(chunk, StreamFinish):
+                                total_time = (time.perf_counter() - start_time) * 1000
+                                logger.info(
+                                    f"[TIMING] StreamFinish received in {total_time/1000:.1f}s; delivered={messages_delivered}",
+                                    extra={
+                                        "json_fields": {
+                                            **log_meta,
+                                            "total_time_ms": total_time,
+                                            "messages_delivered": messages_delivered,
+                                        }
+                                    },
+                                )
                                 return
                     except Exception as e:
-                        logger.warning(f"Error processing stream message: {e}")
+                        logger.warning(
+                            f"Error processing stream message: {e}",
+                            extra={"json_fields": {**log_meta, "error": str(e)}},
+                        )
 
     except asyncio.CancelledError:
-        logger.debug(f"Stream listener cancelled for task {task_id}")
+        elapsed = (time.perf_counter() - start_time) * 1000
+        logger.info(
+            f"[TIMING] _stream_listener CANCELLED after {elapsed:.1f}ms, delivered={messages_delivered}",
+            extra={
+                "json_fields": {
+                    **log_meta,
+                    "elapsed_ms": elapsed,
+                    "messages_delivered": messages_delivered,
+                    "reason": "cancelled",
+                }
+            },
+        )
         raise  # Re-raise to propagate cancellation
     except Exception as e:
-        logger.error(f"Stream listener error for task {task_id}: {e}")
+        elapsed = (time.perf_counter() - start_time) * 1000
+        logger.error(
+            f"[TIMING] _stream_listener ERROR after {elapsed:.1f}ms: {e}",
+            extra={"json_fields": {**log_meta, "elapsed_ms": elapsed, "error": str(e)}},
+        )
         # On error, send finish to unblock subscriber
         try:
             await asyncio.wait_for(
@@ -368,10 +613,24 @@ async def _stream_listener(
             )
         except (asyncio.TimeoutError, asyncio.QueueFull):
             logger.warning(
-                f"Could not deliver finish event for task {task_id} after error"
+                "Could not deliver finish event after error",
+                extra={"json_fields": log_meta},
             )
     finally:
         # Clean up listener task mapping on exit
+        total_time = (time.perf_counter() - start_time) * 1000
+        logger.info(
+            f"[TIMING] _stream_listener FINISHED in {total_time/1000:.1f}s; task={task_id}, "
+            f"delivered={messages_delivered}, xread_count={xread_count}",
+            extra={
+                "json_fields": {
+                    **log_meta,
+                    "total_time_ms": total_time,
+                    "messages_delivered": messages_delivered,
+                    "xread_count": xread_count,
+                }
+            },
+        )
         _listener_tasks.pop(queue_id, None)
 
 
@@ -598,8 +857,10 @@ def _reconstruct_chunk(chunk_data: dict) -> StreamBaseResponse | None:
         ResponseType,
         StreamError,
         StreamFinish,
+        StreamFinishStep,
         StreamHeartbeat,
         StreamStart,
+        StreamStartStep,
         StreamTextDelta,
         StreamTextEnd,
         StreamTextStart,
@@ -613,6 +874,8 @@ def _reconstruct_chunk(chunk_data: dict) -> StreamBaseResponse | None:
     type_to_class: dict[str, type[StreamBaseResponse]] = {
         ResponseType.START.value: StreamStart,
         ResponseType.FINISH.value: StreamFinish,
+        ResponseType.START_STEP.value: StreamStartStep,
+        ResponseType.FINISH_STEP.value: StreamFinishStep,
         ResponseType.TEXT_START.value: StreamTextStart,
         ResponseType.TEXT_DELTA.value: StreamTextDelta,
         ResponseType.TEXT_END.value: StreamTextEnd,

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

@@ -7,15 +7,7 @@ from typing import Any, NotRequired, TypedDict
 
 from backend.api.features.library import db as library_db
 from backend.api.features.store import db as store_db
-from backend.data.graph import (
+from backend.data.graph import Graph, Link, Node, get_graph, get_store_listed_graphs
-    Graph,
-    Link,
-    Node,
-    create_graph,
-    get_graph,
-    get_graph_all_versions,
-    get_store_listed_graphs,
-)
 from backend.util.exceptions import DatabaseError, NotFoundError
 
 from .service import (
@@ -28,8 +20,6 @@ from .service import (
 
 logger = logging.getLogger(__name__)
 
-AGENT_EXECUTOR_BLOCK_ID = "e189baac-8c20-45a1-94a7-55177ea42565"
-
 
 class ExecutionSummary(TypedDict):
     """Summary of a single execution for quality assessment."""
@@ -669,45 +659,6 @@ def json_to_graph(agent_json: dict[str, Any]) -> Graph:
     )
 
 
-def _reassign_node_ids(graph: Graph) -> None:
-    """Reassign all node and link IDs to new UUIDs.
-
-    This is needed when creating a new version to avoid unique constraint violations.
-    """
-    id_map = {node.id: str(uuid.uuid4()) for node in graph.nodes}
-
-    for node in graph.nodes:
-        node.id = id_map[node.id]
-
-    for link in graph.links:
-        link.id = str(uuid.uuid4())
-        if link.source_id in id_map:
-            link.source_id = id_map[link.source_id]
-        if link.sink_id in id_map:
-            link.sink_id = id_map[link.sink_id]
-
-
-def _populate_agent_executor_user_ids(agent_json: dict[str, Any], user_id: str) -> None:
-    """Populate user_id in AgentExecutorBlock nodes.
-
-    The external agent generator creates AgentExecutorBlock nodes with empty user_id.
-    This function fills in the actual user_id so sub-agents run with correct permissions.
-
-    Args:
-        agent_json: Agent JSON dict (modified in place)
-        user_id: User ID to set
-    """
-    for node in agent_json.get("nodes", []):
-        if node.get("block_id") == AGENT_EXECUTOR_BLOCK_ID:
-            input_default = node.get("input_default") or {}
-            if not input_default.get("user_id"):
-                input_default["user_id"] = user_id
-                node["input_default"] = input_default
-                logger.debug(
-                    f"Set user_id for AgentExecutorBlock node {node.get('id')}"
-                )
-
-
 async def save_agent_to_library(
     agent_json: dict[str, Any], user_id: str, is_update: bool = False
 ) -> tuple[Graph, Any]:
@@ -721,35 +672,10 @@ async def save_agent_to_library(
     Returns:
         Tuple of (created Graph, LibraryAgent)
     """
-    # Populate user_id in AgentExecutorBlock nodes before conversion
-    _populate_agent_executor_user_ids(agent_json, user_id)
-
     graph = json_to_graph(agent_json)
-
     if is_update:
-        if graph.id:
+        return await library_db.update_graph_in_library(graph, user_id)
-            existing_versions = await get_graph_all_versions(graph.id, user_id)
+    return await library_db.create_graph_in_library(graph, user_id)
-            if existing_versions:
-                latest_version = max(v.version for v in existing_versions)
-                graph.version = latest_version + 1
-                _reassign_node_ids(graph)
-                logger.info(f"Updating agent {graph.id} to version {graph.version}")
-    else:
-        graph.id = str(uuid.uuid4())
-        graph.version = 1
-        _reassign_node_ids(graph)
-        logger.info(f"Creating new agent with ID {graph.id}")
-
-    created_graph = await create_graph(graph, user_id)
-
-    library_agents = await library_db.create_library_agent(
-        graph=created_graph,
-        user_id=user_id,
-        sensitive_action_safe_mode=True,
-        create_library_agents_for_sub_graphs=False,
-    )
-
-    return created_graph, library_agents[0]
 
 
 def graph_to_json(graph: Graph) -> dict[str, Any]:

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

@@ -206,9 +206,9 @@ async def search_agents(
             ]
         )
         no_results_msg = (
-            f"No agents found matching '{query}'. Try different keywords or browse the marketplace."
+            f"No agents found matching '{query}'. Let the user know they can try different keywords or browse the marketplace. Also let them know you can create a custom agent for them based on their needs."
             if source == "marketplace"
-            else f"No agents matching '{query}' found in your library."
+            else f"No agents matching '{query}' found in your library. Let the user know you can create a custom agent for them based on their needs."
         )
         return NoResultsResponse(
             message=no_results_msg, session_id=session_id, suggestions=suggestions
@@ -224,10 +224,10 @@ async def search_agents(
     message = (
         "Now you have found some options for the user to choose from. "
         "You can add a link to a recommended agent at: /marketplace/agent/agent_id "
-        "Please ask the user if they would like to use any of these agents."
+        "Please ask the user if they would like to use any of these agents. Let the user know we can create a custom agent for them based on their needs."
         if source == "marketplace"
         else "Found agents in the user's library. You can provide a link to view an agent at: "
-        "/library/agents/{agent_id}. Use agent_output to get execution results, or run_agent to execute."
+        "/library/agents/{agent_id}. Use agent_output to get execution results, or run_agent to execute. Let the user know we can create a custom agent for them based on their needs."
     )
 
     return AgentsFoundResponse(

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

@@ -13,10 +13,33 @@ from backend.api.features.chat.tools.models import (
     NoResultsResponse,
 )
 from backend.api.features.store.hybrid_search import unified_hybrid_search
-from backend.data.block import get_block
+from backend.blocks import get_block
+from backend.blocks._base import BlockType
 
 logger = logging.getLogger(__name__)
 
+_TARGET_RESULTS = 10
+# Over-fetch to compensate for post-hoc filtering of graph-only blocks.
+# 40 is 2x current removed; speed of query 10 vs 40 is minimial
+_OVERFETCH_PAGE_SIZE = 40
+
+# Block types that only work within graphs and cannot run standalone in CoPilot.
+COPILOT_EXCLUDED_BLOCK_TYPES = {
+    BlockType.INPUT,  # Graph interface definition - data enters via chat, not graph inputs
+    BlockType.OUTPUT,  # Graph interface definition - data exits via chat, not graph outputs
+    BlockType.WEBHOOK,  # Wait for external events - would hang forever in CoPilot
+    BlockType.WEBHOOK_MANUAL,  # Same as WEBHOOK
+    BlockType.NOTE,  # Visual annotation only - no runtime behavior
+    BlockType.HUMAN_IN_THE_LOOP,  # Pauses for human approval - CoPilot IS human-in-the-loop
+    BlockType.AGENT,  # AgentExecutorBlock requires execution_context - use run_agent tool
+}
+
+# Specific block IDs excluded from CoPilot (STANDARD type but still require graph context)
+COPILOT_EXCLUDED_BLOCK_IDS = {
+    # SmartDecisionMakerBlock - dynamically discovers downstream blocks via graph topology
+    "3b191d9f-356f-482d-8238-ba04b6d18381",
+}
+
 
 class FindBlockTool(BaseTool):
     """Tool for searching available blocks."""
@@ -88,7 +111,7 @@ class FindBlockTool(BaseTool):
                 query=query,
                 content_types=[ContentType.BLOCK],
                 page=1,
-                page_size=10,
+                page_size=_OVERFETCH_PAGE_SIZE,
             )
 
             if not results:
@@ -108,60 +131,90 @@ class FindBlockTool(BaseTool):
                 block = get_block(block_id)
 
                 # Skip disabled blocks
-                if block and not block.disabled:
+                if not block or block.disabled:
-                    # Get input/output schemas
+                    continue
-                    input_schema = {}
-                    output_schema = {}
-                    try:
-                        input_schema = block.input_schema.jsonschema()
-                    except Exception:
-                        pass
-                    try:
-                        output_schema = block.output_schema.jsonschema()
-                    except Exception:
-                        pass
 
-                    # Get categories from block instance
+                # Skip blocks excluded from CoPilot (graph-only blocks)
-                    categories = []
+                if (
-                    if hasattr(block, "categories") and block.categories:
+                    block.block_type in COPILOT_EXCLUDED_BLOCK_TYPES
-                        categories = [cat.value for cat in block.categories]
+                    or block.id in COPILOT_EXCLUDED_BLOCK_IDS
+                ):
+                    continue
 
-                    # Extract required inputs for easier use
+                # Get input/output schemas
-                    required_inputs: list[BlockInputFieldInfo] = []
+                input_schema = {}
-                    if input_schema:
+                output_schema = {}
-                        properties = input_schema.get("properties", {})
+                try:
-                        required_fields = set(input_schema.get("required", []))
+                    input_schema = block.input_schema.jsonschema()
-                        # Get credential field names to exclude from required inputs
+                except Exception as e:
-                        credentials_fields = set(
+                    logger.debug(
-                            block.input_schema.get_credentials_fields().keys()
+                        "Failed to generate input schema for block %s: %s",
-                        )
+                        block_id,
-
+                        e,
-                        for field_name, field_schema in properties.items():
-                            # Skip credential fields - they're handled separately
-                            if field_name in credentials_fields:
-                                continue
-
-                            required_inputs.append(
-                                BlockInputFieldInfo(
-                                    name=field_name,
-                                    type=field_schema.get("type", "string"),
-                                    description=field_schema.get("description", ""),
-                                    required=field_name in required_fields,
-                                    default=field_schema.get("default"),
-                                )
-                            )
-
-                    blocks.append(
-                        BlockInfoSummary(
-                            id=block_id,
-                            name=block.name,
-                            description=block.description or "",
-                            categories=categories,
-                            input_schema=input_schema,
-                            output_schema=output_schema,
-                            required_inputs=required_inputs,
-                        )
                     )
+                try:
+                    output_schema = block.output_schema.jsonschema()
+                except Exception as e:
+                    logger.debug(
+                        "Failed to generate output schema for block %s: %s",
+                        block_id,
+                        e,
+                    )
+
+                # Get categories from block instance
+                categories = []
+                if hasattr(block, "categories") and block.categories:
+                    categories = [cat.value for cat in block.categories]
+
+                # Extract required inputs for easier use
+                required_inputs: list[BlockInputFieldInfo] = []
+                if input_schema:
+                    properties = input_schema.get("properties", {})
+                    required_fields = set(input_schema.get("required", []))
+                    # Get credential field names to exclude from required inputs
+                    credentials_fields = set(
+                        block.input_schema.get_credentials_fields().keys()
+                    )
+
+                    for field_name, field_schema in properties.items():
+                        # Skip credential fields - they're handled separately
+                        if field_name in credentials_fields:
+                            continue
+
+                        required_inputs.append(
+                            BlockInputFieldInfo(
+                                name=field_name,
+                                type=field_schema.get("type", "string"),
+                                description=field_schema.get("description", ""),
+                                required=field_name in required_fields,
+                                default=field_schema.get("default"),
+                            )
+                        )
+
+                blocks.append(
+                    BlockInfoSummary(
+                        id=block_id,
+                        name=block.name,
+                        description=block.description or "",
+                        categories=categories,
+                        input_schema=input_schema,
+                        output_schema=output_schema,
+                        required_inputs=required_inputs,
+                    )
+                )
+
+                if len(blocks) >= _TARGET_RESULTS:
+                    break
+
+            if blocks and len(blocks) < _TARGET_RESULTS:
+                logger.debug(
+                    "find_block returned %d/%d results for query '%s' "
+                    "(filtered %d excluded/disabled blocks)",
+                    len(blocks),
+                    _TARGET_RESULTS,
+                    query,
+                    len(results) - len(blocks),
+                )
 
             if not blocks:
                 return NoResultsResponse(

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

@@ -0,0 +1,139 @@
+"""Tests for block filtering in FindBlockTool."""
+
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+
+from backend.api.features.chat.tools.find_block import (
+    COPILOT_EXCLUDED_BLOCK_IDS,
+    COPILOT_EXCLUDED_BLOCK_TYPES,
+    FindBlockTool,
+)
+from backend.api.features.chat.tools.models import BlockListResponse
+from backend.blocks._base import BlockType
+
+from ._test_data import make_session
+
+_TEST_USER_ID = "test-user-find-block"
+
+
+def make_mock_block(
+    block_id: str, name: str, block_type: BlockType, disabled: bool = False
+):
+    """Create a mock block for testing."""
+    mock = MagicMock()
+    mock.id = block_id
+    mock.name = name
+    mock.description = f"{name} description"
+    mock.block_type = block_type
+    mock.disabled = disabled
+    mock.input_schema = MagicMock()
+    mock.input_schema.jsonschema.return_value = {"properties": {}, "required": []}
+    mock.input_schema.get_credentials_fields.return_value = {}
+    mock.output_schema = MagicMock()
+    mock.output_schema.jsonschema.return_value = {}
+    mock.categories = []
+    return mock
+
+
+class TestFindBlockFiltering:
+    """Tests for block filtering in FindBlockTool."""
+
+    def test_excluded_block_types_contains_expected_types(self):
+        """Verify COPILOT_EXCLUDED_BLOCK_TYPES contains all graph-only types."""
+        assert BlockType.INPUT in COPILOT_EXCLUDED_BLOCK_TYPES
+        assert BlockType.OUTPUT in COPILOT_EXCLUDED_BLOCK_TYPES
+        assert BlockType.WEBHOOK in COPILOT_EXCLUDED_BLOCK_TYPES
+        assert BlockType.WEBHOOK_MANUAL in COPILOT_EXCLUDED_BLOCK_TYPES
+        assert BlockType.NOTE in COPILOT_EXCLUDED_BLOCK_TYPES
+        assert BlockType.HUMAN_IN_THE_LOOP in COPILOT_EXCLUDED_BLOCK_TYPES
+        assert BlockType.AGENT in COPILOT_EXCLUDED_BLOCK_TYPES
+
+    def test_excluded_block_ids_contains_smart_decision_maker(self):
+        """Verify SmartDecisionMakerBlock is in COPILOT_EXCLUDED_BLOCK_IDS."""
+        assert "3b191d9f-356f-482d-8238-ba04b6d18381" in COPILOT_EXCLUDED_BLOCK_IDS
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_excluded_block_type_filtered_from_results(self):
+        """Verify blocks with excluded BlockTypes are filtered from search results."""
+        session = make_session(user_id=_TEST_USER_ID)
+
+        # Mock search returns an INPUT block (excluded) and a STANDARD block (included)
+        search_results = [
+            {"content_id": "input-block-id", "score": 0.9},
+            {"content_id": "standard-block-id", "score": 0.8},
+        ]
+
+        input_block = make_mock_block("input-block-id", "Input Block", BlockType.INPUT)
+        standard_block = make_mock_block(
+            "standard-block-id", "HTTP Request", BlockType.STANDARD
+        )
+
+        def mock_get_block(block_id):
+            return {
+                "input-block-id": input_block,
+                "standard-block-id": standard_block,
+            }.get(block_id)
+
+        with patch(
+            "backend.api.features.chat.tools.find_block.unified_hybrid_search",
+            new_callable=AsyncMock,
+            return_value=(search_results, 2),
+        ):
+            with patch(
+                "backend.api.features.chat.tools.find_block.get_block",
+                side_effect=mock_get_block,
+            ):
+                tool = FindBlockTool()
+                response = await tool._execute(
+                    user_id=_TEST_USER_ID, session=session, query="test"
+                )
+
+        # Should only return the standard block, not the INPUT block
+        assert isinstance(response, BlockListResponse)
+        assert len(response.blocks) == 1
+        assert response.blocks[0].id == "standard-block-id"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_excluded_block_id_filtered_from_results(self):
+        """Verify SmartDecisionMakerBlock is filtered from search results."""
+        session = make_session(user_id=_TEST_USER_ID)
+
+        smart_decision_id = "3b191d9f-356f-482d-8238-ba04b6d18381"
+        search_results = [
+            {"content_id": smart_decision_id, "score": 0.9},
+            {"content_id": "normal-block-id", "score": 0.8},
+        ]
+
+        # SmartDecisionMakerBlock has STANDARD type but is excluded by ID
+        smart_block = make_mock_block(
+            smart_decision_id, "Smart Decision Maker", BlockType.STANDARD
+        )
+        normal_block = make_mock_block(
+            "normal-block-id", "Normal Block", BlockType.STANDARD
+        )
+
+        def mock_get_block(block_id):
+            return {
+                smart_decision_id: smart_block,
+                "normal-block-id": normal_block,
+            }.get(block_id)
+
+        with patch(
+            "backend.api.features.chat.tools.find_block.unified_hybrid_search",
+            new_callable=AsyncMock,
+            return_value=(search_results, 2),
+        ):
+            with patch(
+                "backend.api.features.chat.tools.find_block.get_block",
+                side_effect=mock_get_block,
+            ):
+                tool = FindBlockTool()
+                response = await tool._execute(
+                    user_id=_TEST_USER_ID, session=session, query="decision"
+                )
+
+        # Should only return normal block, not SmartDecisionMakerBlock
+        assert isinstance(response, BlockListResponse)
+        assert len(response.blocks) == 1
+        assert response.blocks[0].id == "normal-block-id"

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

@@ -0,0 +1,29 @@
+"""Shared helpers for chat tools."""
+
+from typing import Any
+
+
+def get_inputs_from_schema(
+    input_schema: dict[str, Any],
+    exclude_fields: set[str] | None = None,
+) -> list[dict[str, Any]]:
+    """Extract input field info from JSON schema."""
+    if not isinstance(input_schema, dict):
+        return []
+
+    exclude = exclude_fields or set()
+    properties = input_schema.get("properties", {})
+    required = set(input_schema.get("required", []))
+
+    return [
+        {
+            "name": name,
+            "title": schema.get("title", name),
+            "type": schema.get("type", "string"),
+            "description": schema.get("description", ""),
+            "required": name in required,
+            "default": schema.get("default"),
+        }
+        for name, schema in properties.items()
+        if name not in exclude
+    ]

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

@@ -24,6 +24,7 @@ from backend.util.timezone_utils import (
 )
 
 from .base import BaseTool
+from .helpers import get_inputs_from_schema
 from .models import (
     AgentDetails,
     AgentDetailsResponse,
@@ -261,7 +262,7 @@ class RunAgentTool(BaseTool):
                         ),
                         requirements={
                             "credentials": requirements_creds_list,
-                            "inputs": self._get_inputs_list(graph.input_schema),
+                            "inputs": get_inputs_from_schema(graph.input_schema),
                             "execution_modes": self._get_execution_modes(graph),
                         },
                     ),
@@ -369,22 +370,6 @@ class RunAgentTool(BaseTool):
                 session_id=session_id,
             )
 
-    def _get_inputs_list(self, input_schema: dict[str, Any]) -> list[dict[str, Any]]:
-        """Extract inputs list from schema."""
-        inputs_list = []
-        if isinstance(input_schema, dict) and "properties" in input_schema:
-            for field_name, field_schema in input_schema["properties"].items():
-                inputs_list.append(
-                    {
-                        "name": field_name,
-                        "title": field_schema.get("title", field_name),
-                        "type": field_schema.get("type", "string"),
-                        "description": field_schema.get("description", ""),
-                        "required": field_name in input_schema.get("required", []),
-                    }
-                )
-        return inputs_list
-
     def _get_execution_modes(self, graph: GraphModel) -> list[str]:
         """Get available execution modes for the graph."""
         trigger_info = graph.trigger_setup_info
@@ -398,7 +383,7 @@ class RunAgentTool(BaseTool):
         suffix: str,
     ) -> str:
         """Build a message describing available inputs for an agent."""
-        inputs_list = self._get_inputs_list(graph.input_schema)
+        inputs_list = get_inputs_from_schema(graph.input_schema)
         required_names = [i["name"] for i in inputs_list if i["required"]]
         optional_names = [i["name"] for i in inputs_list if not i["required"]]
 

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

@@ -8,14 +8,20 @@ from typing import Any
 from pydantic_core import PydanticUndefined
 
 from backend.api.features.chat.model import ChatSession
-from backend.data.block import get_block
+from backend.api.features.chat.tools.find_block import (
+    COPILOT_EXCLUDED_BLOCK_IDS,
+    COPILOT_EXCLUDED_BLOCK_TYPES,
+)
+from backend.blocks import get_block
+from backend.blocks._base import AnyBlockSchema
 from backend.data.execution import ExecutionContext
-from backend.data.model import CredentialsMetaInput
+from backend.data.model import CredentialsFieldInfo, CredentialsMetaInput
 from backend.data.workspace import get_or_create_workspace
 from backend.integrations.creds_manager import IntegrationCredentialsManager
 from backend.util.exceptions import BlockError
 
 from .base import BaseTool
+from .helpers import get_inputs_from_schema
 from .models import (
     BlockOutputResponse,
     ErrorResponse,
@@ -24,7 +30,10 @@ from .models import (
     ToolResponseBase,
     UserReadiness,
 )
-from .utils import build_missing_credentials_from_field_info
+from .utils import (
+    build_missing_credentials_from_field_info,
+    match_credentials_to_requirements,
+)
 
 logger = logging.getLogger(__name__)
 
@@ -73,91 +82,6 @@ class RunBlockTool(BaseTool):
     def requires_auth(self) -> bool:
         return True
 
-    async def _check_block_credentials(
-        self,
-        user_id: str,
-        block: Any,
-        input_data: dict[str, Any] | None = None,
-    ) -> tuple[dict[str, CredentialsMetaInput], list[CredentialsMetaInput]]:
-        """
-        Check if user has required credentials for a block.
-
-        Args:
-            user_id: User ID
-            block: Block to check credentials for
-            input_data: Input data for the block (used to determine provider via discriminator)
-
-        Returns:
-            tuple[matched_credentials, missing_credentials]
-        """
-        matched_credentials: dict[str, CredentialsMetaInput] = {}
-        missing_credentials: list[CredentialsMetaInput] = []
-        input_data = input_data or {}
-
-        # Get credential field info from block's input schema
-        credentials_fields_info = block.input_schema.get_credentials_fields_info()
-
-        if not credentials_fields_info:
-            return matched_credentials, missing_credentials
-
-        # Get user's available credentials
-        creds_manager = IntegrationCredentialsManager()
-        available_creds = await creds_manager.store.get_all_creds(user_id)
-
-        for field_name, field_info in credentials_fields_info.items():
-            effective_field_info = field_info
-            if field_info.discriminator and field_info.discriminator_mapping:
-                # Get discriminator from input, falling back to schema default
-                discriminator_value = input_data.get(field_info.discriminator)
-                if discriminator_value is None:
-                    field = block.input_schema.model_fields.get(
-                        field_info.discriminator
-                    )
-                    if field and field.default is not PydanticUndefined:
-                        discriminator_value = field.default
-
-                if (
-                    discriminator_value
-                    and discriminator_value in field_info.discriminator_mapping
-                ):
-                    effective_field_info = field_info.discriminate(discriminator_value)
-                    logger.debug(
-                        f"Discriminated provider for {field_name}: "
-                        f"{discriminator_value} -> {effective_field_info.provider}"
-                    )
-
-            matching_cred = next(
-                (
-                    cred
-                    for cred in available_creds
-                    if cred.provider in effective_field_info.provider
-                    and cred.type in effective_field_info.supported_types
-                ),
-                None,
-            )
-
-            if matching_cred:
-                matched_credentials[field_name] = CredentialsMetaInput(
-                    id=matching_cred.id,
-                    provider=matching_cred.provider,  # type: ignore
-                    type=matching_cred.type,
-                    title=matching_cred.title,
-                )
-            else:
-                # Create a placeholder for the missing credential
-                provider = next(iter(effective_field_info.provider), "unknown")
-                cred_type = next(iter(effective_field_info.supported_types), "api_key")
-                missing_credentials.append(
-                    CredentialsMetaInput(
-                        id=field_name,
-                        provider=provider,  # type: ignore
-                        type=cred_type,  # type: ignore
-                        title=field_name.replace("_", " ").title(),
-                    )
-                )
-
-        return matched_credentials, missing_credentials
-
     async def _execute(
         self,
         user_id: str | None,
@@ -212,11 +136,24 @@ class RunBlockTool(BaseTool):
                 session_id=session_id,
             )
 
+        # Check if block is excluded from CoPilot (graph-only blocks)
+        if (
+            block.block_type in COPILOT_EXCLUDED_BLOCK_TYPES
+            or block.id in COPILOT_EXCLUDED_BLOCK_IDS
+        ):
+            return ErrorResponse(
+                message=(
+                    f"Block '{block.name}' cannot be run directly in CoPilot. "
+                    "This block is designed for use within graphs only."
+                ),
+                session_id=session_id,
+            )
+
         logger.info(f"Executing block {block.name} ({block_id}) for user {user_id}")
 
         creds_manager = IntegrationCredentialsManager()
-        matched_credentials, missing_credentials = await self._check_block_credentials(
+        matched_credentials, missing_credentials = (
-            user_id, block, input_data
+            await self._resolve_block_credentials(user_id, block, input_data)
         )
 
         if missing_credentials:
@@ -345,29 +282,75 @@ class RunBlockTool(BaseTool):
                 session_id=session_id,
             )
 
-    def _get_inputs_list(self, block: Any) -> list[dict[str, Any]]:
+    async def _resolve_block_credentials(
+        self,
+        user_id: str,
+        block: AnyBlockSchema,
+        input_data: dict[str, Any] | None = None,
+    ) -> tuple[dict[str, CredentialsMetaInput], list[CredentialsMetaInput]]:
+        """
+        Resolve credentials for a block by matching user's available credentials.
+
+        Args:
+            user_id: User ID
+            block: Block to resolve credentials for
+            input_data: Input data for the block (used to determine provider via discriminator)
+
+        Returns:
+            tuple of (matched_credentials, missing_credentials) - matched credentials
+            are used for block execution, missing ones indicate setup requirements.
+        """
+        input_data = input_data or {}
+        requirements = self._resolve_discriminated_credentials(block, input_data)
+
+        if not requirements:
+            return {}, []
+
+        return await match_credentials_to_requirements(user_id, requirements)
+
+    def _get_inputs_list(self, block: AnyBlockSchema) -> list[dict[str, Any]]:
         """Extract non-credential inputs from block schema."""
-        inputs_list = []
         schema = block.input_schema.jsonschema()
-        properties = schema.get("properties", {})
-        required_fields = set(schema.get("required", []))
-
-        # Get credential field names to exclude
         credentials_fields = set(block.input_schema.get_credentials_fields().keys())
+        return get_inputs_from_schema(schema, exclude_fields=credentials_fields)
 
-        for field_name, field_schema in properties.items():
+    def _resolve_discriminated_credentials(
-            # Skip credential fields
+        self,
-            if field_name in credentials_fields:
+        block: AnyBlockSchema,
-                continue
+        input_data: dict[str, Any],
+    ) -> dict[str, CredentialsFieldInfo]:
+        """Resolve credential requirements, applying discriminator logic where needed."""
+        credentials_fields_info = block.input_schema.get_credentials_fields_info()
+        if not credentials_fields_info:
+            return {}
 
-            inputs_list.append(
+        resolved: dict[str, CredentialsFieldInfo] = {}
-                {
-                    "name": field_name,
-                    "title": field_schema.get("title", field_name),
-                    "type": field_schema.get("type", "string"),
-                    "description": field_schema.get("description", ""),
-                    "required": field_name in required_fields,
-                }
-            )
 
-        return inputs_list
+        for field_name, field_info in credentials_fields_info.items():
+            effective_field_info = field_info
+
+            if field_info.discriminator and field_info.discriminator_mapping:
+                discriminator_value = input_data.get(field_info.discriminator)
+                if discriminator_value is None:
+                    field = block.input_schema.model_fields.get(
+                        field_info.discriminator
+                    )
+                    if field and field.default is not PydanticUndefined:
+                        discriminator_value = field.default
+
+                if (
+                    discriminator_value
+                    and discriminator_value in field_info.discriminator_mapping
+                ):
+                    effective_field_info = field_info.discriminate(discriminator_value)
+                    # For host-scoped credentials, add the discriminator value
+                    # (e.g., URL) so _credential_is_for_host can match it
+                    effective_field_info.discriminator_values.add(discriminator_value)
+                    logger.debug(
+                        f"Discriminated provider for {field_name}: "
+                        f"{discriminator_value} -> {effective_field_info.provider}"
+                    )
+
+            resolved[field_name] = effective_field_info
+
+        return resolved

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

@@ -0,0 +1,106 @@
+"""Tests for block execution guards in RunBlockTool."""
+
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+from backend.api.features.chat.tools.models import ErrorResponse
+from backend.api.features.chat.tools.run_block import RunBlockTool
+from backend.blocks._base import BlockType
+
+from ._test_data import make_session
+
+_TEST_USER_ID = "test-user-run-block"
+
+
+def make_mock_block(
+    block_id: str, name: str, block_type: BlockType, disabled: bool = False
+):
+    """Create a mock block for testing."""
+    mock = MagicMock()
+    mock.id = block_id
+    mock.name = name
+    mock.block_type = block_type
+    mock.disabled = disabled
+    mock.input_schema = MagicMock()
+    mock.input_schema.jsonschema.return_value = {"properties": {}, "required": []}
+    mock.input_schema.get_credentials_fields_info.return_value = []
+    return mock
+
+
+class TestRunBlockFiltering:
+    """Tests for block execution guards in RunBlockTool."""
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_excluded_block_type_returns_error(self):
+        """Attempting to execute a block with excluded BlockType returns error."""
+        session = make_session(user_id=_TEST_USER_ID)
+
+        input_block = make_mock_block("input-block-id", "Input Block", BlockType.INPUT)
+
+        with patch(
+            "backend.api.features.chat.tools.run_block.get_block",
+            return_value=input_block,
+        ):
+            tool = RunBlockTool()
+            response = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                block_id="input-block-id",
+                input_data={},
+            )
+
+        assert isinstance(response, ErrorResponse)
+        assert "cannot be run directly in CoPilot" in response.message
+        assert "designed for use within graphs only" in response.message
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_excluded_block_id_returns_error(self):
+        """Attempting to execute SmartDecisionMakerBlock returns error."""
+        session = make_session(user_id=_TEST_USER_ID)
+
+        smart_decision_id = "3b191d9f-356f-482d-8238-ba04b6d18381"
+        smart_block = make_mock_block(
+            smart_decision_id, "Smart Decision Maker", BlockType.STANDARD
+        )
+
+        with patch(
+            "backend.api.features.chat.tools.run_block.get_block",
+            return_value=smart_block,
+        ):
+            tool = RunBlockTool()
+            response = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                block_id=smart_decision_id,
+                input_data={},
+            )
+
+        assert isinstance(response, ErrorResponse)
+        assert "cannot be run directly in CoPilot" in response.message
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_non_excluded_block_passes_guard(self):
+        """Non-excluded blocks pass the filtering guard (may fail later for other reasons)."""
+        session = make_session(user_id=_TEST_USER_ID)
+
+        standard_block = make_mock_block(
+            "standard-id", "HTTP Request", BlockType.STANDARD
+        )
+
+        with patch(
+            "backend.api.features.chat.tools.run_block.get_block",
+            return_value=standard_block,
+        ):
+            tool = RunBlockTool()
+            response = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                block_id="standard-id",
+                input_data={},
+            )
+
+        # Should NOT be an ErrorResponse about CoPilot exclusion
+        # (may be other errors like missing credentials, but not the exclusion guard)
+        if isinstance(response, ErrorResponse):
+            assert "cannot be run directly in CoPilot" not in response.message

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

@@ -6,9 +6,14 @@ from typing import Any
 from backend.api.features.library import db as library_db
 from backend.api.features.library import model as library_model
 from backend.api.features.store import db as store_db
-from backend.data import graph as graph_db
 from backend.data.graph import GraphModel
-from backend.data.model import Credentials, CredentialsFieldInfo, CredentialsMetaInput
+from backend.data.model import (
+    Credentials,
+    CredentialsFieldInfo,
+    CredentialsMetaInput,
+    HostScopedCredentials,
+    OAuth2Credentials,
+)
 from backend.integrations.creds_manager import IntegrationCredentialsManager
 from backend.util.exceptions import NotFoundError
 
@@ -39,14 +44,8 @@ async def fetch_graph_from_store_slug(
         return None, None
 
     # Get the graph from store listing version
-    graph_meta = await store_db.get_available_graph(
+    graph = await store_db.get_available_graph(
-        store_agent.store_listing_version_id
+        store_agent.store_listing_version_id, hide_nodes=False
-    )
-    graph = await graph_db.get_graph(
-        graph_id=graph_meta.id,
-        version=graph_meta.version,
-        user_id=None,  # Public access
-        include_subgraphs=True,
     )
     return graph, store_agent
 
@@ -123,7 +122,7 @@ def build_missing_credentials_from_graph(
 
     return {
         field_key: _serialize_missing_credential(field_key, field_info)
-        for field_key, (field_info, _node_fields) in aggregated_fields.items()
+        for field_key, (field_info, _, _) in aggregated_fields.items()
         if field_key not in matched_keys
     }
 
@@ -225,6 +224,99 @@ async def get_or_create_library_agent(
     return library_agents[0]
 
 
+async def match_credentials_to_requirements(
+    user_id: str,
+    requirements: dict[str, CredentialsFieldInfo],
+) -> tuple[dict[str, CredentialsMetaInput], list[CredentialsMetaInput]]:
+    """
+    Match user's credentials against a dictionary of credential requirements.
+
+    This is the core matching logic shared by both graph and block credential matching.
+    """
+    matched: dict[str, CredentialsMetaInput] = {}
+    missing: list[CredentialsMetaInput] = []
+
+    if not requirements:
+        return matched, missing
+
+    available_creds = await get_user_credentials(user_id)
+
+    for field_name, field_info in requirements.items():
+        matching_cred = find_matching_credential(available_creds, field_info)
+
+        if matching_cred:
+            try:
+                matched[field_name] = create_credential_meta_from_match(matching_cred)
+            except Exception as e:
+                logger.error(
+                    f"Failed to create CredentialsMetaInput for field '{field_name}': "
+                    f"provider={matching_cred.provider}, type={matching_cred.type}, "
+                    f"credential_id={matching_cred.id}",
+                    exc_info=True,
+                )
+                provider = next(iter(field_info.provider), "unknown")
+                cred_type = next(iter(field_info.supported_types), "api_key")
+                missing.append(
+                    CredentialsMetaInput(
+                        id=field_name,
+                        provider=provider,  # type: ignore
+                        type=cred_type,  # type: ignore
+                        title=f"{field_name} (validation failed: {e})",
+                    )
+                )
+        else:
+            provider = next(iter(field_info.provider), "unknown")
+            cred_type = next(iter(field_info.supported_types), "api_key")
+            missing.append(
+                CredentialsMetaInput(
+                    id=field_name,
+                    provider=provider,  # type: ignore
+                    type=cred_type,  # type: ignore
+                    title=field_name.replace("_", " ").title(),
+                )
+            )
+
+    return matched, missing
+
+
+async def get_user_credentials(user_id: str) -> list[Credentials]:
+    """Get all available credentials for a user."""
+    creds_manager = IntegrationCredentialsManager()
+    return await creds_manager.store.get_all_creds(user_id)
+
+
+def find_matching_credential(
+    available_creds: list[Credentials],
+    field_info: CredentialsFieldInfo,
+) -> Credentials | None:
+    """Find a credential that matches the required provider, type, scopes, and host."""
+    for cred in available_creds:
+        if cred.provider not in field_info.provider:
+            continue
+        if cred.type not in field_info.supported_types:
+            continue
+        if cred.type == "oauth2" and not _credential_has_required_scopes(
+            cred, field_info
+        ):
+            continue
+        if cred.type == "host_scoped" and not _credential_is_for_host(cred, field_info):
+            continue
+        return cred
+    return None
+
+
+def create_credential_meta_from_match(
+    matching_cred: Credentials,
+) -> CredentialsMetaInput:
+    """Create a CredentialsMetaInput from a matched credential."""
+    return CredentialsMetaInput(
+        id=matching_cred.id,
+        provider=matching_cred.provider,  # type: ignore
+        type=matching_cred.type,
+        title=matching_cred.title,
+    )
+
+
 async def match_user_credentials_to_graph(
     user_id: str,
     graph: GraphModel,
@@ -264,7 +356,8 @@ async def match_user_credentials_to_graph(
     # provider is in the set of acceptable providers.
     for credential_field_name, (
         credential_requirements,
-        _node_fields,
+        _,
+        _,
     ) in aggregated_creds.items():
         # Find first matching credential by provider, type, and scopes
         matching_cred = next(
@@ -273,7 +366,14 @@ async def match_user_credentials_to_graph(
                 for cred in available_creds
                 if cred.provider in credential_requirements.provider
                 and cred.type in credential_requirements.supported_types
-                and _credential_has_required_scopes(cred, credential_requirements)
+                and (
+                    cred.type != "oauth2"
+                    or _credential_has_required_scopes(cred, credential_requirements)
+                )
+                and (
+                    cred.type != "host_scoped"
+                    or _credential_is_for_host(cred, credential_requirements)
+                )
             ),
             None,
         )
@@ -318,27 +418,32 @@ async def match_user_credentials_to_graph(
 
 
 def _credential_has_required_scopes(
-    credential: Credentials,
+    credential: OAuth2Credentials,
     requirements: CredentialsFieldInfo,
 ) -> bool:
-    """
+    """Check if an OAuth2 credential has all the scopes required by the input."""
-    Check if a credential has all the scopes required by the block.
-
-    For OAuth2 credentials, verifies that the credential's scopes are a superset
-    of the required scopes. For other credential types, returns True (no scope check).
-    """
-    # Only OAuth2 credentials have scopes to check
-    if credential.type != "oauth2":
-        return True
-
     # If no scopes are required, any credential matches
     if not requirements.required_scopes:
         return True
-
-    # Check that credential scopes are a superset of required scopes
     return set(credential.scopes).issuperset(requirements.required_scopes)
 
 
+def _credential_is_for_host(
+    credential: HostScopedCredentials,
+    requirements: CredentialsFieldInfo,
+) -> bool:
+    """Check if a host-scoped credential matches the host required by the input."""
+    # We need to know the host to match host-scoped credentials to.
+    # Graph.aggregate_credentials_inputs() adds the node's set URL value (if any)
+    # to discriminator_values. No discriminator_values -> no host to match against.
+    if not requirements.discriminator_values:
+        return True
+
+    # Check that credential host matches required host.
+    # Host-scoped credential inputs are grouped by host, so any item from the set works.
+    return credential.matches_url(list(requirements.discriminator_values)[0])
+
+
 async def check_user_has_required_credentials(
     user_id: str,
     required_credentials: list[CredentialsMetaInput],

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

@@ -12,14 +12,16 @@ import backend.api.features.store.image_gen as store_image_gen
 import backend.api.features.store.media as store_media
 import backend.data.graph as graph_db
 import backend.data.integrations as integrations_db
-from backend.data.block import BlockInput
 from backend.data.db import transaction
 from backend.data.execution import get_graph_execution
 from backend.data.graph import GraphSettings
 from backend.data.includes import AGENT_PRESET_INCLUDE, library_agent_include
-from backend.data.model import CredentialsMetaInput
+from backend.data.model import CredentialsMetaInput, GraphInput
 from backend.integrations.creds_manager import IntegrationCredentialsManager
-from backend.integrations.webhooks.graph_lifecycle_hooks import on_graph_activate
+from backend.integrations.webhooks.graph_lifecycle_hooks import (
+    on_graph_activate,
+    on_graph_deactivate,
+)
 from backend.util.clients import get_scheduler_client
 from backend.util.exceptions import DatabaseError, InvalidInputError, NotFoundError
 from backend.util.json import SafeJson
@@ -371,7 +373,7 @@ async def get_library_agent_by_graph_id(
 
 
 async def add_generated_agent_image(
-    graph: graph_db.BaseGraph,
+    graph: graph_db.GraphBaseMeta,
     user_id: str,
     library_agent_id: str,
 ) -> Optional[prisma.models.LibraryAgent]:
@@ -537,6 +539,92 @@ async def update_agent_version_in_library(
     return library_model.LibraryAgent.from_db(lib)
 
 
+async def create_graph_in_library(
+    graph: graph_db.Graph,
+    user_id: str,
+) -> tuple[graph_db.GraphModel, library_model.LibraryAgent]:
+    """Create a new graph and add it to the user's library."""
+    graph.version = 1
+    graph_model = graph_db.make_graph_model(graph, user_id)
+    graph_model.reassign_ids(user_id=user_id, reassign_graph_id=True)
+
+    created_graph = await graph_db.create_graph(graph_model, user_id)
+
+    library_agents = await create_library_agent(
+        graph=created_graph,
+        user_id=user_id,
+        sensitive_action_safe_mode=True,
+        create_library_agents_for_sub_graphs=False,
+    )
+
+    if created_graph.is_active:
+        created_graph = await on_graph_activate(created_graph, user_id=user_id)
+
+    return created_graph, library_agents[0]
+
+
+async def update_graph_in_library(
+    graph: graph_db.Graph,
+    user_id: str,
+) -> tuple[graph_db.GraphModel, library_model.LibraryAgent]:
+    """Create a new version of an existing graph and update the library entry."""
+    existing_versions = await graph_db.get_graph_all_versions(graph.id, user_id)
+    current_active_version = (
+        next((v for v in existing_versions if v.is_active), None)
+        if existing_versions
+        else None
+    )
+    graph.version = (
+        max(v.version for v in existing_versions) + 1 if existing_versions else 1
+    )
+
+    graph_model = graph_db.make_graph_model(graph, user_id)
+    graph_model.reassign_ids(user_id=user_id, reassign_graph_id=False)
+
+    created_graph = await graph_db.create_graph(graph_model, user_id)
+
+    library_agent = await get_library_agent_by_graph_id(user_id, created_graph.id)
+    if not library_agent:
+        raise NotFoundError(f"Library agent not found for graph {created_graph.id}")
+
+    library_agent = await update_library_agent_version_and_settings(
+        user_id, created_graph
+    )
+
+    if created_graph.is_active:
+        created_graph = await on_graph_activate(created_graph, user_id=user_id)
+        await graph_db.set_graph_active_version(
+            graph_id=created_graph.id,
+            version=created_graph.version,
+            user_id=user_id,
+        )
+        if current_active_version:
+            await on_graph_deactivate(current_active_version, user_id=user_id)
+
+    return created_graph, library_agent
+
+
+async def update_library_agent_version_and_settings(
+    user_id: str, agent_graph: graph_db.GraphModel
+) -> library_model.LibraryAgent:
+    """Update library agent to point to new graph version and sync settings."""
+    library = await update_agent_version_in_library(
+        user_id, agent_graph.id, agent_graph.version
+    )
+    updated_settings = GraphSettings.from_graph(
+        graph=agent_graph,
+        hitl_safe_mode=library.settings.human_in_the_loop_safe_mode,
+        sensitive_action_safe_mode=library.settings.sensitive_action_safe_mode,
+    )
+    if updated_settings != library.settings:
+        library = await update_library_agent(
+            library_agent_id=library.id,
+            user_id=user_id,
+            settings=updated_settings,
+        )
+    return library
+
+
 async def update_library_agent(
     library_agent_id: str,
     user_id: str,
@@ -1041,7 +1129,7 @@ async def create_preset_from_graph_execution(
 async def update_preset(
     user_id: str,
     preset_id: str,
-    inputs: Optional[BlockInput] = None,
+    inputs: Optional[GraphInput] = None,
     credentials: Optional[dict[str, CredentialsMetaInput]] = None,
     name: Optional[str] = None,
     description: Optional[str] = None,

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

@@ -6,9 +6,12 @@ import prisma.enums
 import prisma.models
 import pydantic
 
-from backend.data.block import BlockInput
 from backend.data.graph import GraphModel, GraphSettings, GraphTriggerInfo
-from backend.data.model import CredentialsMetaInput, is_credentials_field_name
+from backend.data.model import (
+    CredentialsMetaInput,
+    GraphInput,
+    is_credentials_field_name,
+)
 from backend.util.json import loads as json_loads
 from backend.util.models import Pagination
 
@@ -323,7 +326,7 @@ class LibraryAgentPresetCreatable(pydantic.BaseModel):
     graph_id: str
     graph_version: int
 
-    inputs: BlockInput
+    inputs: GraphInput
     credentials: dict[str, CredentialsMetaInput]
 
     name: str
@@ -352,7 +355,7 @@ class LibraryAgentPresetUpdatable(pydantic.BaseModel):
     Request model used when updating a preset for a library agent.
     """
 
-    inputs: Optional[BlockInput] = None
+    inputs: Optional[GraphInput] = None
     credentials: Optional[dict[str, CredentialsMetaInput]] = None
 
     name: Optional[str] = None
@@ -395,7 +398,7 @@ class LibraryAgentPreset(LibraryAgentPresetCreatable):
                 "Webhook must be included in AgentPreset query when webhookId is set"
             )
 
-        input_data: BlockInput = {}
+        input_data: GraphInput = {}
         input_credentials: dict[str, CredentialsMetaInput] = {}
 
         for preset_input in preset.InputPresets:

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

@@ -5,8 +5,8 @@ from typing import Optional
 import aiohttp
 from fastapi import HTTPException
 
+from backend.blocks import get_block
 from backend.data import graph as graph_db
-from backend.data.block import get_block
 from backend.util.settings import Settings
 
 from .models import ApiResponse, ChatRequest, GraphData

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

@@ -152,7 +152,7 @@ class BlockHandler(ContentHandler):
 
     async def get_missing_items(self, batch_size: int) -> list[ContentItem]:
         """Fetch blocks without embeddings."""
-        from backend.data.block import get_blocks
+        from backend.blocks import get_blocks
 
         # Get all available blocks
         all_blocks = get_blocks()
@@ -249,7 +249,7 @@ class BlockHandler(ContentHandler):
 
     async def get_stats(self) -> dict[str, int]:
         """Get statistics about block embedding coverage."""
-        from backend.data.block import get_blocks
+        from backend.blocks import get_blocks
 
         all_blocks = get_blocks()
 

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

@@ -93,7 +93,7 @@ async def test_block_handler_get_missing_items(mocker):
     mock_existing = []
 
     with patch(
-        "backend.data.block.get_blocks",
+        "backend.blocks.get_blocks",
         return_value=mock_blocks,
     ):
         with patch(
@@ -135,7 +135,7 @@ async def test_block_handler_get_stats(mocker):
     mock_embedded = [{"count": 2}]
 
     with patch(
-        "backend.data.block.get_blocks",
+        "backend.blocks.get_blocks",
         return_value=mock_blocks,
     ):
         with patch(
@@ -327,7 +327,7 @@ async def test_block_handler_handles_missing_attributes():
     mock_blocks = {"block-minimal": mock_block_class}
 
     with patch(
-        "backend.data.block.get_blocks",
+        "backend.blocks.get_blocks",
         return_value=mock_blocks,
     ):
         with patch(
@@ -360,7 +360,7 @@ async def test_block_handler_skips_failed_blocks():
     mock_blocks = {"good-block": good_block, "bad-block": bad_block}
 
     with patch(
-        "backend.data.block.get_blocks",
+        "backend.blocks.get_blocks",
         return_value=mock_blocks,
     ):
         with patch(

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

@@ -1,7 +1,7 @@
 import asyncio
 import logging
 from datetime import datetime, timezone
-from typing import Any, Literal
+from typing import Any, Literal, overload
 
 import fastapi
 import prisma.enums
@@ -11,8 +11,8 @@ import prisma.types
 
 from backend.data.db import transaction
 from backend.data.graph import (
-    GraphMeta,
     GraphModel,
+    GraphModelWithoutNodes,
     get_graph,
     get_graph_as_admin,
     get_sub_graphs,
@@ -334,7 +334,22 @@ async def get_store_agent_details(
         raise DatabaseError("Failed to fetch agent details") from e
 
 
-async def get_available_graph(store_listing_version_id: str) -> GraphMeta:
+@overload
+async def get_available_graph(
+    store_listing_version_id: str, hide_nodes: Literal[False]
+) -> GraphModel: ...
+
+
+@overload
+async def get_available_graph(
+    store_listing_version_id: str, hide_nodes: Literal[True] = True
+) -> GraphModelWithoutNodes: ...
+
+
+async def get_available_graph(
+    store_listing_version_id: str,
+    hide_nodes: bool = True,
+) -> GraphModelWithoutNodes | GraphModel:
     try:
         # Get avaialble, non-deleted store listing version
         store_listing_version = (
@@ -344,7 +359,7 @@ async def get_available_graph(store_listing_version_id: str) -> GraphMeta:
                     "isAvailable": True,
                     "isDeleted": False,
                 },
-                include={"AgentGraph": {"include": {"Nodes": True}}},
+                include={"AgentGraph": {"include": AGENT_GRAPH_INCLUDE}},
             )
         )
 
@@ -354,7 +369,9 @@ async def get_available_graph(store_listing_version_id: str) -> GraphMeta:
                 detail=f"Store listing version {store_listing_version_id} not found",
             )
 
-        return GraphModel.from_db(store_listing_version.AgentGraph).meta()
+        return (GraphModelWithoutNodes if hide_nodes else GraphModel).from_db(
+            store_listing_version.AgentGraph
+        )
 
     except Exception as e:
         logger.error(f"Error getting agent: {e}")

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

@@ -662,7 +662,7 @@ async def cleanup_orphaned_embeddings() -> dict[str, Any]:
                 )
                 current_ids = {row["id"] for row in valid_agents}
             elif content_type == ContentType.BLOCK:
-                from backend.data.block import get_blocks
+                from backend.blocks import get_blocks
 
                 current_ids = set(get_blocks().keys())
             elif content_type == ContentType.DOCUMENTATION:

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

@@ -8,6 +8,7 @@ Includes BM25 reranking for improved lexical relevance.
 
 import logging
 import re
+import time
 from dataclasses import dataclass
 from typing import Any, Literal
 
@@ -362,7 +363,11 @@ async def unified_hybrid_search(
         LIMIT {limit_param} OFFSET {offset_param}
     """
 
-    results = await query_raw_with_schema(sql_query, *params)
+    try:
+        results = await query_raw_with_schema(sql_query, *params)
+    except Exception as e:
+        await _log_vector_error_diagnostics(e)
+        raise
 
     total = results[0]["total_count"] if results else 0
     # Apply BM25 reranking
@@ -686,7 +691,11 @@ async def hybrid_search(
         LIMIT {limit_param} OFFSET {offset_param}
     """
 
-    results = await query_raw_with_schema(sql_query, *params)
+    try:
+        results = await query_raw_with_schema(sql_query, *params)
+    except Exception as e:
+        await _log_vector_error_diagnostics(e)
+        raise
 
     total = results[0]["total_count"] if results else 0
 
@@ -718,6 +727,87 @@ async def hybrid_search_simple(
     return await hybrid_search(query=query, page=page, page_size=page_size)
 
 
+# ============================================================================
+# Diagnostics
+# ============================================================================
+
+# Rate limit: only log vector error diagnostics once per this interval
+_VECTOR_DIAG_INTERVAL_SECONDS = 60
+_last_vector_diag_time: float = 0
+
+
+async def _log_vector_error_diagnostics(error: Exception) -> None:
+    """Log diagnostic info when 'type vector does not exist' error occurs.
+
+    Note: Diagnostic queries use query_raw_with_schema which may run on a different
+    pooled connection than the one that failed. Session-level search_path can differ,
+    so these diagnostics show cluster-wide state, not necessarily the failed session.
+
+    Includes rate limiting to avoid log spam - only logs once per minute.
+    Caller should re-raise the error after calling this function.
+    """
+    global _last_vector_diag_time
+
+    # Check if this is the vector type error
+    error_str = str(error).lower()
+    if not (
+        "type" in error_str and "vector" in error_str and "does not exist" in error_str
+    ):
+        return
+
+    # Rate limit: only log once per interval
+    now = time.time()
+    if now - _last_vector_diag_time < _VECTOR_DIAG_INTERVAL_SECONDS:
+        return
+    _last_vector_diag_time = now
+
+    try:
+        diagnostics: dict[str, object] = {}
+
+        try:
+            search_path_result = await query_raw_with_schema("SHOW search_path")
+            diagnostics["search_path"] = search_path_result
+        except Exception as e:
+            diagnostics["search_path"] = f"Error: {e}"
+
+        try:
+            schema_result = await query_raw_with_schema("SELECT current_schema()")
+            diagnostics["current_schema"] = schema_result
+        except Exception as e:
+            diagnostics["current_schema"] = f"Error: {e}"
+
+        try:
+            user_result = await query_raw_with_schema(
+                "SELECT current_user, session_user, current_database()"
+            )
+            diagnostics["user_info"] = user_result
+        except Exception as e:
+            diagnostics["user_info"] = f"Error: {e}"
+
+        try:
+            # Check pgvector extension installation (cluster-wide, stable info)
+            ext_result = await query_raw_with_schema(
+                "SELECT extname, extversion, nspname as schema "
+                "FROM pg_extension e "
+                "JOIN pg_namespace n ON e.extnamespace = n.oid "
+                "WHERE extname = 'vector'"
+            )
+            diagnostics["pgvector_extension"] = ext_result
+        except Exception as e:
+            diagnostics["pgvector_extension"] = f"Error: {e}"
+
+        logger.error(
+            f"Vector type error diagnostics:\n"
+            f"  Error: {error}\n"
+            f"  search_path: {diagnostics.get('search_path')}\n"
+            f"  current_schema: {diagnostics.get('current_schema')}\n"
+            f"  user_info: {diagnostics.get('user_info')}\n"
+            f"  pgvector_extension: {diagnostics.get('pgvector_extension')}"
+        )
+    except Exception as diag_error:
+        logger.error(f"Failed to collect vector error diagnostics: {diag_error}")
+
+
 # Backward compatibility alias - HybridSearchWeights maps to StoreAgentSearchWeights
 # for existing code that expects the popularity parameter
 HybridSearchWeights = StoreAgentSearchWeights

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

@@ -7,16 +7,7 @@ from replicate.client import Client as ReplicateClient
 from replicate.exceptions import ReplicateError
 from replicate.helpers import FileOutput
 
-from backend.blocks.ideogram import (
+from backend.data.graph import GraphBaseMeta
-    AspectRatio,
-    ColorPalettePreset,
-    IdeogramModelBlock,
-    IdeogramModelName,
-    MagicPromptOption,
-    StyleType,
-    UpscaleOption,
-)
-from backend.data.graph import BaseGraph
 from backend.data.model import CredentialsMetaInput, ProviderName
 from backend.integrations.credentials_store import ideogram_credentials
 from backend.util.request import Requests
@@ -34,14 +25,14 @@ class ImageStyle(str, Enum):
     DIGITAL_ART = "digital art"
 
 
-async def generate_agent_image(agent: BaseGraph | AgentGraph) -> io.BytesIO:
+async def generate_agent_image(agent: GraphBaseMeta | AgentGraph) -> io.BytesIO:
     if settings.config.use_agent_image_generation_v2:
         return await generate_agent_image_v2(graph=agent)
     else:
         return await generate_agent_image_v1(agent=agent)
 
 
-async def generate_agent_image_v2(graph: BaseGraph | AgentGraph) -> io.BytesIO:
+async def generate_agent_image_v2(graph: GraphBaseMeta | AgentGraph) -> io.BytesIO:
     """
     Generate an image for an agent using Ideogram model.
     Returns:
@@ -50,18 +41,31 @@ async def generate_agent_image_v2(graph: BaseGraph | AgentGraph) -> io.BytesIO:
     if not ideogram_credentials.api_key:
         raise ValueError("Missing Ideogram API key")
 
+    from backend.blocks.ideogram import (
+        AspectRatio,
+        ColorPalettePreset,
+        IdeogramModelBlock,
+        IdeogramModelName,
+        MagicPromptOption,
+        StyleType,
+        UpscaleOption,
+    )
+
     name = graph.name
     description = f"{name} ({graph.description})" if graph.description else name
 
     prompt = (
-        f"Create a visually striking retro-futuristic vector pop art illustration prominently featuring "
+        "Create a visually striking retro-futuristic vector pop art illustration "
-        f'"{name}" in bold typography. The image clearly and literally depicts a {description}, '
+        f'prominently featuring "{name}" in bold typography. The image clearly and '
-        f"along with recognizable objects directly associated with the primary function of a {name}. "
+        f"literally depicts a {description}, along with recognizable objects directly "
-        f"Ensure the imagery is concrete, intuitive, and immediately understandable, clearly conveying the "
+        f"associated with the primary function of a {name}. "
-        f"purpose of a {name}. Maintain vibrant, limited-palette colors, sharp vector lines, geometric "
+        f"Ensure the imagery is concrete, intuitive, and immediately understandable, "
-        f"shapes, flat illustration techniques, and solid colors without gradients or shading. Preserve a "
+        f"clearly conveying the purpose of a {name}. "
-        f"retro-futuristic aesthetic influenced by mid-century futurism and 1960s psychedelia, "
+        "Maintain vibrant, limited-palette colors, sharp vector lines, "
-        f"prioritizing clear visual storytelling and thematic clarity above all else."
+        "geometric shapes, flat illustration techniques, and solid colors "
+        "without gradients or shading. Preserve a retro-futuristic aesthetic "
+        "influenced by mid-century futurism and 1960s psychedelia, "
+        "prioritizing clear visual storytelling and thematic clarity above all else."
     )
 
     custom_colors = [
@@ -99,12 +103,12 @@ async def generate_agent_image_v2(graph: BaseGraph | AgentGraph) -> io.BytesIO:
     return io.BytesIO(response.content)
 
 
-async def generate_agent_image_v1(agent: BaseGraph | AgentGraph) -> io.BytesIO:
+async def generate_agent_image_v1(agent: GraphBaseMeta | AgentGraph) -> io.BytesIO:
     """
     Generate an image for an agent using Flux model via Replicate API.
 
     Args:
-        agent (Graph): The agent to generate an image for
+        agent (GraphBaseMeta | AgentGraph): The agent to generate an image for
 
     Returns:
         io.BytesIO: The generated image as bytes
@@ -114,7 +118,13 @@ async def generate_agent_image_v1(agent: BaseGraph | AgentGraph) -> io.BytesIO:
             raise ValueError("Missing Replicate API key in settings")
 
         # Construct prompt from agent details
-        prompt = f"Create a visually engaging app store thumbnail for the AI agent that highlights what it does in a clear and captivating way:\n- **Name**: {agent.name}\n- **Description**: {agent.description}\nFocus on showcasing its core functionality with an appealing design."
+        prompt = (
+            "Create a visually engaging app store thumbnail for the AI agent "
+            "that highlights what it does in a clear and captivating way:\n"
+            f"- **Name**: {agent.name}\n"
+            f"- **Description**: {agent.description}\n"
+            f"Focus on showcasing its core functionality with an appealing design."
+        )
 
         # Set up Replicate client
         client = ReplicateClient(api_token=settings.secrets.replicate_api_key)

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

@@ -278,7 +278,7 @@ async def get_agent(
 )
 async def get_graph_meta_by_store_listing_version_id(
     store_listing_version_id: str,
-) -> backend.data.graph.GraphMeta:
+) -> backend.data.graph.GraphModelWithoutNodes:
     """
     Get Agent Graph from Store Listing Version ID.
     """

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

@@ -40,10 +40,11 @@ from backend.api.model import (
     UpdateTimezoneRequest,
     UploadFileResponse,
 )
+from backend.blocks import get_block, get_blocks
 from backend.data import execution as execution_db
 from backend.data import graph as graph_db
 from backend.data.auth import api_key as api_key_db
-from backend.data.block import BlockInput, CompletedBlockOutput, get_block, get_blocks
+from backend.data.block import BlockInput, CompletedBlockOutput
 from backend.data.credit import (
     AutoTopUpConfig,
     RefundRequest,
@@ -101,7 +102,6 @@ from backend.util.timezone_utils import (
 from backend.util.virus_scanner import scan_content_safe
 
 from .library import db as library_db
-from .library import model as library_model
 from .store.model import StoreAgentDetails
 
 
@@ -823,18 +823,16 @@ async def update_graph(
     graph: graph_db.Graph,
     user_id: Annotated[str, Security(get_user_id)],
 ) -> graph_db.GraphModel:
-    # Sanity check
     if graph.id and graph.id != graph_id:
         raise HTTPException(400, detail="Graph ID does not match ID in URI")
 
-    # Determine new version
     existing_versions = await graph_db.get_graph_all_versions(graph_id, user_id=user_id)
     if not existing_versions:
         raise HTTPException(404, detail=f"Graph #{graph_id} not found")
-    latest_version_number = max(g.version for g in existing_versions)
-    graph.version = latest_version_number + 1
 
+    graph.version = max(g.version for g in existing_versions) + 1
     current_active_version = next((v for v in existing_versions if v.is_active), None)
+
     graph = graph_db.make_graph_model(graph, user_id)
     graph.reassign_ids(user_id=user_id, reassign_graph_id=False)
     graph.validate_graph(for_run=False)
@@ -842,27 +840,23 @@ async def update_graph(
     new_graph_version = await graph_db.create_graph(graph, user_id=user_id)
 
     if new_graph_version.is_active:
-        # Keep the library agent up to date with the new active version
+        await library_db.update_library_agent_version_and_settings(
-        await _update_library_agent_version_and_settings(user_id, new_graph_version)
+            user_id, new_graph_version
-
+        )
-        # Handle activation of the new graph first to ensure continuity
         new_graph_version = await on_graph_activate(new_graph_version, user_id=user_id)
-        # Ensure new version is the only active version
         await graph_db.set_graph_active_version(
             graph_id=graph_id, version=new_graph_version.version, user_id=user_id
         )
         if current_active_version:
-            # Handle deactivation of the previously active version
             await on_graph_deactivate(current_active_version, user_id=user_id)
 
-    # Fetch new graph version *with sub-graphs* (needed for credentials input schema)
     new_graph_version_with_subgraphs = await graph_db.get_graph(
         graph_id,
         new_graph_version.version,
         user_id=user_id,
         include_subgraphs=True,
     )
-    assert new_graph_version_with_subgraphs  # make type checker happy
+    assert new_graph_version_with_subgraphs
     return new_graph_version_with_subgraphs
 
 
@@ -900,33 +894,15 @@ async def set_graph_active_version(
     )
 
     # Keep the library agent up to date with the new active version
-    await _update_library_agent_version_and_settings(user_id, new_active_graph)
+    await library_db.update_library_agent_version_and_settings(
+        user_id, new_active_graph
+    )
 
     if current_active_graph and current_active_graph.version != new_active_version:
         # Handle deactivation of the previously active version
         await on_graph_deactivate(current_active_graph, user_id=user_id)
 
 
-async def _update_library_agent_version_and_settings(
-    user_id: str, agent_graph: graph_db.GraphModel
-) -> library_model.LibraryAgent:
-    library = await library_db.update_agent_version_in_library(
-        user_id, agent_graph.id, agent_graph.version
-    )
-    updated_settings = GraphSettings.from_graph(
-        graph=agent_graph,
-        hitl_safe_mode=library.settings.human_in_the_loop_safe_mode,
-        sensitive_action_safe_mode=library.settings.sensitive_action_safe_mode,
-    )
-    if updated_settings != library.settings:
-        library = await library_db.update_library_agent(
-            library_agent_id=library.id,
-            user_id=user_id,
-            settings=updated_settings,
-        )
-    return library
-
-
 @v1_router.patch(
     path="/graphs/{graph_id}/settings",
     summary="Update graph settings",

autogpt_platform/backend/backend/blocks/__init__.py

@@ -3,22 +3,19 @@ import logging
 import os
 import re
 from pathlib import Path
-from typing import TYPE_CHECKING, TypeVar
+from typing import Sequence, Type, TypeVar
 
+from backend.blocks._base import AnyBlockSchema, BlockType
 from backend.util.cache import cached
 
 logger = logging.getLogger(__name__)
 
-
-if TYPE_CHECKING:
-    from backend.data.block import Block
-
 T = TypeVar("T")
 
 
 @cached(ttl_seconds=3600)
-def load_all_blocks() -> dict[str, type["Block"]]:
+def load_all_blocks() -> dict[str, type["AnyBlockSchema"]]:
-    from backend.data.block import Block
+    from backend.blocks._base import Block
     from backend.util.settings import Config
 
     # Check if example blocks should be loaded from settings
@@ -50,8 +47,8 @@ def load_all_blocks() -> dict[str, type["Block"]]:
         importlib.import_module(f".{module}", package=__name__)
 
     # Load all Block instances from the available modules
-    available_blocks: dict[str, type["Block"]] = {}
+    available_blocks: dict[str, type["AnyBlockSchema"]] = {}
-    for block_cls in all_subclasses(Block):
+    for block_cls in _all_subclasses(Block):
         class_name = block_cls.__name__
 
         if class_name.endswith("Base"):
@@ -64,7 +61,7 @@ def load_all_blocks() -> dict[str, type["Block"]]:
                 "please name the class with 'Base' at the end"
             )
 
-        block = block_cls.create()
+        block = block_cls()  # pyright: ignore[reportAbstractUsage]
 
         if not isinstance(block.id, str) or len(block.id) != 36:
             raise ValueError(
@@ -105,7 +102,7 @@ def load_all_blocks() -> dict[str, type["Block"]]:
         available_blocks[block.id] = block_cls
 
     # Filter out blocks with incomplete auth configs, e.g. missing OAuth server secrets
-    from backend.data.block import is_block_auth_configured
+    from ._utils import is_block_auth_configured
 
     filtered_blocks = {}
     for block_id, block_cls in available_blocks.items():
@@ -115,11 +112,48 @@ def load_all_blocks() -> dict[str, type["Block"]]:
     return filtered_blocks
 
 
-__all__ = ["load_all_blocks"]
+def _all_subclasses(cls: type[T]) -> list[type[T]]:
-
-
-def all_subclasses(cls: type[T]) -> list[type[T]]:
     subclasses = cls.__subclasses__()
     for subclass in subclasses:
-        subclasses += all_subclasses(subclass)
+        subclasses += _all_subclasses(subclass)
     return subclasses
+
+
+# ============== Block access helper functions ============== #
+
+
+def get_blocks() -> dict[str, Type["AnyBlockSchema"]]:
+    return load_all_blocks()
+
+
+# Note on the return type annotation: https://github.com/microsoft/pyright/issues/10281
+def get_block(block_id: str) -> "AnyBlockSchema | None":
+    cls = get_blocks().get(block_id)
+    return cls() if cls else None
+
+
+@cached(ttl_seconds=3600)
+def get_webhook_block_ids() -> Sequence[str]:
+    return [
+        id
+        for id, B in get_blocks().items()
+        if B().block_type in (BlockType.WEBHOOK, BlockType.WEBHOOK_MANUAL)
+    ]
+
+
+@cached(ttl_seconds=3600)
+def get_io_block_ids() -> Sequence[str]:
+    return [
+        id
+        for id, B in get_blocks().items()
+        if B().block_type in (BlockType.INPUT, BlockType.OUTPUT)
+    ]
+
+
+@cached(ttl_seconds=3600)
+def get_human_in_the_loop_block_ids() -> Sequence[str]:
+    return [
+        id
+        for id, B in get_blocks().items()
+        if B().block_type == BlockType.HUMAN_IN_THE_LOOP
+    ]

autogpt_platform/backend/backend/blocks/_base.py

@@ -0,0 +1,739 @@
+import inspect
+import logging
+from abc import ABC, abstractmethod
+from enum import Enum
+from typing import (
+    TYPE_CHECKING,
+    Any,
+    Callable,
+    ClassVar,
+    Generic,
+    Optional,
+    Type,
+    TypeAlias,
+    TypeVar,
+    cast,
+    get_origin,
+)
+
+import jsonref
+import jsonschema
+from pydantic import BaseModel
+
+from backend.data.block import BlockInput, BlockOutput, BlockOutputEntry
+from backend.data.model import (
+    Credentials,
+    CredentialsFieldInfo,
+    CredentialsMetaInput,
+    SchemaField,
+    is_credentials_field_name,
+)
+from backend.integrations.providers import ProviderName
+from backend.util import json
+from backend.util.exceptions import (
+    BlockError,
+    BlockExecutionError,
+    BlockInputError,
+    BlockOutputError,
+    BlockUnknownError,
+)
+from backend.util.settings import Config
+
+logger = logging.getLogger(__name__)
+
+if TYPE_CHECKING:
+    from backend.data.execution import ExecutionContext
+    from backend.data.model import ContributorDetails, NodeExecutionStats
+
+    from ..data.graph import Link
+
+app_config = Config()
+
+
+BlockTestOutput = BlockOutputEntry | tuple[str, Callable[[Any], bool]]
+
+
+class BlockType(Enum):
+    STANDARD = "Standard"
+    INPUT = "Input"
+    OUTPUT = "Output"
+    NOTE = "Note"
+    WEBHOOK = "Webhook"
+    WEBHOOK_MANUAL = "Webhook (manual)"
+    AGENT = "Agent"
+    AI = "AI"
+    AYRSHARE = "Ayrshare"
+    HUMAN_IN_THE_LOOP = "Human In The Loop"
+
+
+class BlockCategory(Enum):
+    AI = "Block that leverages AI to perform a task."
+    SOCIAL = "Block that interacts with social media platforms."
+    TEXT = "Block that processes text data."
+    SEARCH = "Block that searches or extracts information from the internet."
+    BASIC = "Block that performs basic operations."
+    INPUT = "Block that interacts with input of the graph."
+    OUTPUT = "Block that interacts with output of the graph."
+    LOGIC = "Programming logic to control the flow of your agent"
+    COMMUNICATION = "Block that interacts with communication platforms."
+    DEVELOPER_TOOLS = "Developer tools such as GitHub blocks."
+    DATA = "Block that interacts with structured data."
+    HARDWARE = "Block that interacts with hardware."
+    AGENT = "Block that interacts with other agents."
+    CRM = "Block that interacts with CRM services."
+    SAFETY = (
+        "Block that provides AI safety mechanisms such as detecting harmful content"
+    )
+    PRODUCTIVITY = "Block that helps with productivity"
+    ISSUE_TRACKING = "Block that helps with issue tracking"
+    MULTIMEDIA = "Block that interacts with multimedia content"
+    MARKETING = "Block that helps with marketing"
+
+    def dict(self) -> dict[str, str]:
+        return {"category": self.name, "description": self.value}
+
+
+class BlockCostType(str, Enum):
+    RUN = "run"  # cost X credits per run
+    BYTE = "byte"  # cost X credits per byte
+    SECOND = "second"  # cost X credits per second
+
+
+class BlockCost(BaseModel):
+    cost_amount: int
+    cost_filter: BlockInput
+    cost_type: BlockCostType
+
+    def __init__(
+        self,
+        cost_amount: int,
+        cost_type: BlockCostType = BlockCostType.RUN,
+        cost_filter: Optional[BlockInput] = None,
+        **data: Any,
+    ) -> None:
+        super().__init__(
+            cost_amount=cost_amount,
+            cost_filter=cost_filter or {},
+            cost_type=cost_type,
+            **data,
+        )
+
+
+class BlockInfo(BaseModel):
+    id: str
+    name: str
+    inputSchema: dict[str, Any]
+    outputSchema: dict[str, Any]
+    costs: list[BlockCost]
+    description: str
+    categories: list[dict[str, str]]
+    contributors: list[dict[str, Any]]
+    staticOutput: bool
+    uiType: str
+
+
+class BlockSchema(BaseModel):
+    cached_jsonschema: ClassVar[dict[str, Any]]
+
+    @classmethod
+    def jsonschema(cls) -> dict[str, Any]:
+        if cls.cached_jsonschema:
+            return cls.cached_jsonschema
+
+        model = jsonref.replace_refs(cls.model_json_schema(), merge_props=True)
+
+        def ref_to_dict(obj):
+            if isinstance(obj, dict):
+                # OpenAPI <3.1 does not support sibling fields that has a $ref key
+                # So sometimes, the schema has an "allOf"/"anyOf"/"oneOf" with 1 item.
+                keys = {"allOf", "anyOf", "oneOf"}
+                one_key = next((k for k in keys if k in obj and len(obj[k]) == 1), None)
+                if one_key:
+                    obj.update(obj[one_key][0])
+
+                return {
+                    key: ref_to_dict(value)
+                    for key, value in obj.items()
+                    if not key.startswith("$") and key != one_key
+                }
+            elif isinstance(obj, list):
+                return [ref_to_dict(item) for item in obj]
+
+            return obj
+
+        cls.cached_jsonschema = cast(dict[str, Any], ref_to_dict(model))
+
+        return cls.cached_jsonschema
+
+    @classmethod
+    def validate_data(cls, data: BlockInput) -> str | None:
+        return json.validate_with_jsonschema(
+            schema=cls.jsonschema(),
+            data={k: v for k, v in data.items() if v is not None},
+        )
+
+    @classmethod
+    def get_mismatch_error(cls, data: BlockInput) -> str | None:
+        return cls.validate_data(data)
+
+    @classmethod
+    def get_field_schema(cls, field_name: str) -> dict[str, Any]:
+        model_schema = cls.jsonschema().get("properties", {})
+        if not model_schema:
+            raise ValueError(f"Invalid model schema {cls}")
+
+        property_schema = model_schema.get(field_name)
+        if not property_schema:
+            raise ValueError(f"Invalid property name {field_name}")
+
+        return property_schema
+
+    @classmethod
+    def validate_field(cls, field_name: str, data: BlockInput) -> str | None:
+        """
+        Validate the data against a specific property (one of the input/output name).
+        Returns the validation error message if the data does not match the schema.
+        """
+        try:
+            property_schema = cls.get_field_schema(field_name)
+            jsonschema.validate(json.to_dict(data), property_schema)
+            return None
+        except jsonschema.ValidationError as e:
+            return str(e)
+
+    @classmethod
+    def get_fields(cls) -> set[str]:
+        return set(cls.model_fields.keys())
+
+    @classmethod
+    def get_required_fields(cls) -> set[str]:
+        return {
+            field
+            for field, field_info in cls.model_fields.items()
+            if field_info.is_required()
+        }
+
+    @classmethod
+    def __pydantic_init_subclass__(cls, **kwargs):
+        """Validates the schema definition. Rules:
+        - Fields with annotation `CredentialsMetaInput` MUST be
+          named `credentials` or `*_credentials`
+        - Fields named `credentials` or `*_credentials` MUST be
+          of type `CredentialsMetaInput`
+        """
+        super().__pydantic_init_subclass__(**kwargs)
+
+        # Reset cached JSON schema to prevent inheriting it from parent class
+        cls.cached_jsonschema = {}
+
+        credentials_fields = cls.get_credentials_fields()
+
+        for field_name in cls.get_fields():
+            if is_credentials_field_name(field_name):
+                if field_name not in credentials_fields:
+                    raise TypeError(
+                        f"Credentials field '{field_name}' on {cls.__qualname__} "
+                        f"is not of type {CredentialsMetaInput.__name__}"
+                    )
+
+                CredentialsMetaInput.validate_credentials_field_schema(
+                    cls.get_field_schema(field_name), field_name
+                )
+
+            elif field_name in credentials_fields:
+                raise KeyError(
+                    f"Credentials field '{field_name}' on {cls.__qualname__} "
+                    "has invalid name: must be 'credentials' or *_credentials"
+                )
+
+    @classmethod
+    def get_credentials_fields(cls) -> dict[str, type[CredentialsMetaInput]]:
+        return {
+            field_name: info.annotation
+            for field_name, info in cls.model_fields.items()
+            if (
+                inspect.isclass(info.annotation)
+                and issubclass(
+                    get_origin(info.annotation) or info.annotation,
+                    CredentialsMetaInput,
+                )
+            )
+        }
+
+    @classmethod
+    def get_auto_credentials_fields(cls) -> dict[str, dict[str, Any]]:
+        """
+        Get fields that have auto_credentials metadata (e.g., GoogleDriveFileInput).
+
+        Returns a dict mapping kwarg_name -> {field_name, auto_credentials_config}
+
+        Raises:
+            ValueError: If multiple fields have the same kwarg_name, as this would
+                cause silent overwriting and only the last field would be processed.
+        """
+        result: dict[str, dict[str, Any]] = {}
+        schema = cls.jsonschema()
+        properties = schema.get("properties", {})
+
+        for field_name, field_schema in properties.items():
+            auto_creds = field_schema.get("auto_credentials")
+            if auto_creds:
+                kwarg_name = auto_creds.get("kwarg_name", "credentials")
+                if kwarg_name in result:
+                    raise ValueError(
+                        f"Duplicate auto_credentials kwarg_name '{kwarg_name}' "
+                        f"in fields '{result[kwarg_name]['field_name']}' and "
+                        f"'{field_name}' on {cls.__qualname__}"
+                    )
+                result[kwarg_name] = {
+                    "field_name": field_name,
+                    "config": auto_creds,
+                }
+        return result
+
+    @classmethod
+    def get_credentials_fields_info(cls) -> dict[str, CredentialsFieldInfo]:
+        result = {}
+
+        # Regular credentials fields
+        for field_name in cls.get_credentials_fields().keys():
+            result[field_name] = CredentialsFieldInfo.model_validate(
+                cls.get_field_schema(field_name), by_alias=True
+            )
+
+        # Auto-generated credentials fields (from GoogleDriveFileInput etc.)
+        for kwarg_name, info in cls.get_auto_credentials_fields().items():
+            config = info["config"]
+            # Build a schema-like dict that CredentialsFieldInfo can parse
+            auto_schema = {
+                "credentials_provider": [config.get("provider", "google")],
+                "credentials_types": [config.get("type", "oauth2")],
+                "credentials_scopes": config.get("scopes"),
+            }
+            result[kwarg_name] = CredentialsFieldInfo.model_validate(
+                auto_schema, by_alias=True
+            )
+
+        return result
+
+    @classmethod
+    def get_input_defaults(cls, data: BlockInput) -> BlockInput:
+        return data  # Return as is, by default.
+
+    @classmethod
+    def get_missing_links(cls, data: BlockInput, links: list["Link"]) -> set[str]:
+        input_fields_from_nodes = {link.sink_name for link in links}
+        return input_fields_from_nodes - set(data)
+
+    @classmethod
+    def get_missing_input(cls, data: BlockInput) -> set[str]:
+        return cls.get_required_fields() - set(data)
+
+
+class BlockSchemaInput(BlockSchema):
+    """
+    Base schema class for block inputs.
+    All block input schemas should extend this class for consistency.
+    """
+
+    pass
+
+
+class BlockSchemaOutput(BlockSchema):
+    """
+    Base schema class for block outputs that includes a standard error field.
+    All block output schemas should extend this class to ensure consistent error handling.
+    """
+
+    error: str = SchemaField(
+        description="Error message if the operation failed", default=""
+    )
+
+
+BlockSchemaInputType = TypeVar("BlockSchemaInputType", bound=BlockSchemaInput)
+BlockSchemaOutputType = TypeVar("BlockSchemaOutputType", bound=BlockSchemaOutput)
+
+
+class EmptyInputSchema(BlockSchemaInput):
+    pass
+
+
+class EmptyOutputSchema(BlockSchemaOutput):
+    pass
+
+
+# For backward compatibility - will be deprecated
+EmptySchema = EmptyOutputSchema
+
+
+# --8<-- [start:BlockWebhookConfig]
+class BlockManualWebhookConfig(BaseModel):
+    """
+    Configuration model for webhook-triggered blocks on which
+    the user has to manually set up the webhook at the provider.
+    """
+
+    provider: ProviderName
+    """The service provider that the webhook connects to"""
+
+    webhook_type: str
+    """
+    Identifier for the webhook type. E.g. GitHub has repo and organization level hooks.
+
+    Only for use in the corresponding `WebhooksManager`.
+    """
+
+    event_filter_input: str = ""
+    """
+    Name of the block's event filter input.
+    Leave empty if the corresponding webhook doesn't have distinct event/payload types.
+    """
+
+    event_format: str = "{event}"
+    """
+    Template string for the event(s) that a block instance subscribes to.
+    Applied individually to each event selected in the event filter input.
+
+    Example: `"pull_request.{event}"` -> `"pull_request.opened"`
+    """
+
+
+class BlockWebhookConfig(BlockManualWebhookConfig):
+    """
+    Configuration model for webhook-triggered blocks for which
+    the webhook can be automatically set up through the provider's API.
+    """
+
+    resource_format: str
+    """
+    Template string for the resource that a block instance subscribes to.
+    Fields will be filled from the block's inputs (except `payload`).
+
+    Example: `f"{repo}/pull_requests"` (note: not how it's actually implemented)
+
+    Only for use in the corresponding `WebhooksManager`.
+    """
+    # --8<-- [end:BlockWebhookConfig]
+
+
+class Block(ABC, Generic[BlockSchemaInputType, BlockSchemaOutputType]):
+    def __init__(
+        self,
+        id: str = "",
+        description: str = "",
+        contributors: list["ContributorDetails"] = [],
+        categories: set[BlockCategory] | None = None,
+        input_schema: Type[BlockSchemaInputType] = EmptyInputSchema,
+        output_schema: Type[BlockSchemaOutputType] = EmptyOutputSchema,
+        test_input: BlockInput | list[BlockInput] | None = None,
+        test_output: BlockTestOutput | list[BlockTestOutput] | None = None,
+        test_mock: dict[str, Any] | None = None,
+        test_credentials: Optional[Credentials | dict[str, Credentials]] = None,
+        disabled: bool = False,
+        static_output: bool = False,
+        block_type: BlockType = BlockType.STANDARD,
+        webhook_config: Optional[BlockWebhookConfig | BlockManualWebhookConfig] = None,
+        is_sensitive_action: bool = False,
+    ):
+        """
+        Initialize the block with the given schema.
+
+        Args:
+            id: The unique identifier for the block, this value will be persisted in the
+                DB. So it should be a unique and constant across the application run.
+                Use the UUID format for the ID.
+            description: The description of the block, explaining what the block does.
+            contributors: The list of contributors who contributed to the block.
+            input_schema: The schema, defined as a Pydantic model, for the input data.
+            output_schema: The schema, defined as a Pydantic model, for the output data.
+            test_input: The list or single sample input data for the block, for testing.
+            test_output: The list or single expected output if the test_input is run.
+            test_mock: function names on the block implementation to mock on test run.
+            disabled: If the block is disabled, it will not be available for execution.
+            static_output: Whether the output links of the block are static by default.
+        """
+        from backend.data.model import NodeExecutionStats
+
+        self.id = id
+        self.input_schema = input_schema
+        self.output_schema = output_schema
+        self.test_input = test_input
+        self.test_output = test_output
+        self.test_mock = test_mock
+        self.test_credentials = test_credentials
+        self.description = description
+        self.categories = categories or set()
+        self.contributors = contributors or set()
+        self.disabled = disabled
+        self.static_output = static_output
+        self.block_type = block_type
+        self.webhook_config = webhook_config
+        self.is_sensitive_action = is_sensitive_action
+        self.execution_stats: "NodeExecutionStats" = NodeExecutionStats()
+
+        if self.webhook_config:
+            if isinstance(self.webhook_config, BlockWebhookConfig):
+                # Enforce presence of credentials field on auto-setup webhook blocks
+                if not (cred_fields := self.input_schema.get_credentials_fields()):
+                    raise TypeError(
+                        "credentials field is required on auto-setup webhook blocks"
+                    )
+                # Disallow multiple credentials inputs on webhook blocks
+                elif len(cred_fields) > 1:
+                    raise ValueError(
+                        "Multiple credentials inputs not supported on webhook blocks"
+                    )
+
+                self.block_type = BlockType.WEBHOOK
+            else:
+                self.block_type = BlockType.WEBHOOK_MANUAL
+
+            # Enforce shape of webhook event filter, if present
+            if self.webhook_config.event_filter_input:
+                event_filter_field = self.input_schema.model_fields[
+                    self.webhook_config.event_filter_input
+                ]
+                if not (
+                    isinstance(event_filter_field.annotation, type)
+                    and issubclass(event_filter_field.annotation, BaseModel)
+                    and all(
+                        field.annotation is bool
+                        for field in event_filter_field.annotation.model_fields.values()
+                    )
+                ):
+                    raise NotImplementedError(
+                        f"{self.name} has an invalid webhook event selector: "
+                        "field must be a BaseModel and all its fields must be boolean"
+                    )
+
+            # Enforce presence of 'payload' input
+            if "payload" not in self.input_schema.model_fields:
+                raise TypeError(
+                    f"{self.name} is webhook-triggered but has no 'payload' input"
+                )
+
+            # Disable webhook-triggered block if webhook functionality not available
+            if not app_config.platform_base_url:
+                self.disabled = True
+
+    @abstractmethod
+    async def run(self, input_data: BlockSchemaInputType, **kwargs) -> BlockOutput:
+        """
+        Run the block with the given input data.
+        Args:
+            input_data: The input data with the structure of input_schema.
+
+        Kwargs: Currently 14/02/2025 these include
+            graph_id: The ID of the graph.
+            node_id: The ID of the node.
+            graph_exec_id: The ID of the graph execution.
+            node_exec_id: The ID of the node execution.
+            user_id: The ID of the user.
+
+        Returns:
+            A Generator that yields (output_name, output_data).
+            output_name: One of the output name defined in Block's output_schema.
+            output_data: The data for the output_name, matching the defined schema.
+        """
+        # --- satisfy the type checker, never executed -------------
+        if False:  # noqa: SIM115
+            yield "name", "value"  # pyright: ignore[reportMissingYield]
+        raise NotImplementedError(f"{self.name} does not implement the run method.")
+
+    async def run_once(
+        self, input_data: BlockSchemaInputType, output: str, **kwargs
+    ) -> Any:
+        async for item in self.run(input_data, **kwargs):
+            name, data = item
+            if name == output:
+                return data
+        raise ValueError(f"{self.name} did not produce any output for {output}")
+
+    def merge_stats(self, stats: "NodeExecutionStats") -> "NodeExecutionStats":
+        self.execution_stats += stats
+        return self.execution_stats
+
+    @property
+    def name(self):
+        return self.__class__.__name__
+
+    def to_dict(self):
+        return {
+            "id": self.id,
+            "name": self.name,
+            "inputSchema": self.input_schema.jsonschema(),
+            "outputSchema": self.output_schema.jsonschema(),
+            "description": self.description,
+            "categories": [category.dict() for category in self.categories],
+            "contributors": [
+                contributor.model_dump() for contributor in self.contributors
+            ],
+            "staticOutput": self.static_output,
+            "uiType": self.block_type.value,
+        }
+
+    def get_info(self) -> BlockInfo:
+        from backend.data.credit import get_block_cost
+
+        return BlockInfo(
+            id=self.id,
+            name=self.name,
+            inputSchema=self.input_schema.jsonschema(),
+            outputSchema=self.output_schema.jsonschema(),
+            costs=get_block_cost(self),
+            description=self.description,
+            categories=[category.dict() for category in self.categories],
+            contributors=[
+                contributor.model_dump() for contributor in self.contributors
+            ],
+            staticOutput=self.static_output,
+            uiType=self.block_type.value,
+        )
+
+    async def execute(self, input_data: BlockInput, **kwargs) -> BlockOutput:
+        try:
+            async for output_name, output_data in self._execute(input_data, **kwargs):
+                yield output_name, output_data
+        except Exception as ex:
+            if isinstance(ex, BlockError):
+                raise ex
+            else:
+                raise (
+                    BlockExecutionError
+                    if isinstance(ex, ValueError)
+                    else BlockUnknownError
+                )(
+                    message=str(ex),
+                    block_name=self.name,
+                    block_id=self.id,
+                ) from ex
+
+    async def is_block_exec_need_review(
+        self,
+        input_data: BlockInput,
+        *,
+        user_id: str,
+        node_id: str,
+        node_exec_id: str,
+        graph_exec_id: str,
+        graph_id: str,
+        graph_version: int,
+        execution_context: "ExecutionContext",
+        **kwargs,
+    ) -> tuple[bool, BlockInput]:
+        """
+        Check if this block execution needs human review and handle the review process.
+
+        Returns:
+            Tuple of (should_pause, input_data_to_use)
+            - should_pause: True if execution should be paused for review
+            - input_data_to_use: The input data to use (may be modified by reviewer)
+        """
+        if not (
+            self.is_sensitive_action and execution_context.sensitive_action_safe_mode
+        ):
+            return False, input_data
+
+        from backend.blocks.helpers.review import HITLReviewHelper
+
+        # Handle the review request and get decision
+        decision = await HITLReviewHelper.handle_review_decision(
+            input_data=input_data,
+            user_id=user_id,
+            node_id=node_id,
+            node_exec_id=node_exec_id,
+            graph_exec_id=graph_exec_id,
+            graph_id=graph_id,
+            graph_version=graph_version,
+            block_name=self.name,
+            editable=True,
+        )
+
+        if decision is None:
+            # We're awaiting review - pause execution
+            return True, input_data
+
+        if not decision.should_proceed:
+            # Review was rejected, raise an error to stop execution
+            raise BlockExecutionError(
+                message=f"Block execution rejected by reviewer: {decision.message}",
+                block_name=self.name,
+                block_id=self.id,
+            )
+
+        # Review was approved - use the potentially modified data
+        # ReviewResult.data must be a dict for block inputs
+        reviewed_data = decision.review_result.data
+        if not isinstance(reviewed_data, dict):
+            raise BlockExecutionError(
+                message=f"Review data must be a dict for block input, got {type(reviewed_data).__name__}",
+                block_name=self.name,
+                block_id=self.id,
+            )
+        return False, reviewed_data
+
+    async def _execute(self, input_data: BlockInput, **kwargs) -> BlockOutput:
+        # Check for review requirement only if running within a graph execution context
+        # Direct block execution (e.g., from chat) skips the review process
+        has_graph_context = all(
+            key in kwargs
+            for key in (
+                "node_exec_id",
+                "graph_exec_id",
+                "graph_id",
+                "execution_context",
+            )
+        )
+        if has_graph_context:
+            should_pause, input_data = await self.is_block_exec_need_review(
+                input_data, **kwargs
+            )
+            if should_pause:
+                return
+
+        # Validate the input data (original or reviewer-modified) once
+        if error := self.input_schema.validate_data(input_data):
+            raise BlockInputError(
+                message=f"Unable to execute block with invalid input data: {error}",
+                block_name=self.name,
+                block_id=self.id,
+            )
+
+        # Use the validated input data
+        async for output_name, output_data in self.run(
+            self.input_schema(**{k: v for k, v in input_data.items() if v is not None}),
+            **kwargs,
+        ):
+            if output_name == "error":
+                raise BlockExecutionError(
+                    message=output_data, block_name=self.name, block_id=self.id
+                )
+            if self.block_type == BlockType.STANDARD and (
+                error := self.output_schema.validate_field(output_name, output_data)
+            ):
+                raise BlockOutputError(
+                    message=f"Block produced an invalid output data: {error}",
+                    block_name=self.name,
+                    block_id=self.id,
+                )
+            yield output_name, output_data
+
+    def is_triggered_by_event_type(
+        self, trigger_config: dict[str, Any], event_type: str
+    ) -> bool:
+        if not self.webhook_config:
+            raise TypeError("This method can't be used on non-trigger blocks")
+        if not self.webhook_config.event_filter_input:
+            return True
+        event_filter = trigger_config.get(self.webhook_config.event_filter_input)
+        if not event_filter:
+            raise ValueError("Event filter is not configured on trigger")
+        return event_type in [
+            self.webhook_config.event_format.format(event=k)
+            for k in event_filter
+            if event_filter[k] is True
+        ]
+
+
+# Type alias for any block with standard input/output schemas
+AnyBlockSchema: TypeAlias = Block[BlockSchemaInput, BlockSchemaOutput]

autogpt_platform/backend/backend/blocks/_utils.py

@@ -0,0 +1,122 @@
+import logging
+import os
+
+from backend.integrations.providers import ProviderName
+
+from ._base import AnyBlockSchema
+
+logger = logging.getLogger(__name__)
+
+
+def is_block_auth_configured(
+    block_cls: type[AnyBlockSchema],
+) -> bool:
+    """
+    Check if a block has a valid authentication method configured at runtime.
+
+    For example if a block is an OAuth-only block and there env vars are not set,
+    do not show it in the UI.
+
+    """
+    from backend.sdk.registry import AutoRegistry
+
+    # Create an instance to access input_schema
+    try:
+        block = block_cls()
+    except Exception as e:
+        # If we can't create a block instance, assume it's not OAuth-only
+        logger.error(f"Error creating block instance for {block_cls.__name__}: {e}")
+        return True
+    logger.debug(
+        f"Checking if block {block_cls.__name__} has a valid provider configured"
+    )
+
+    # Get all credential inputs from input schema
+    credential_inputs = block.input_schema.get_credentials_fields_info()
+    required_inputs = block.input_schema.get_required_fields()
+    if not credential_inputs:
+        logger.debug(
+            f"Block {block_cls.__name__} has no credential inputs - Treating as valid"
+        )
+        return True
+
+    # Check credential inputs
+    if len(required_inputs.intersection(credential_inputs.keys())) == 0:
+        logger.debug(
+            f"Block {block_cls.__name__} has only optional credential inputs"
+            " - will work without credentials configured"
+        )
+
+    # Check if the credential inputs for this block are correctly configured
+    for field_name, field_info in credential_inputs.items():
+        provider_names = field_info.provider
+        if not provider_names:
+            logger.warning(
+                f"Block {block_cls.__name__} "
+                f"has credential input '{field_name}' with no provider options"
+                " - Disabling"
+            )
+            return False
+
+        # If a field has multiple possible providers, each one needs to be usable to
+        # prevent breaking the UX
+        for _provider_name in provider_names:
+            provider_name = _provider_name.value
+            if provider_name in ProviderName.__members__.values():
+                logger.debug(
+                    f"Block {block_cls.__name__} credential input '{field_name}' "
+                    f"provider '{provider_name}' is part of the legacy provider system"
+                    " - Treating as valid"
+                )
+                break
+
+            provider = AutoRegistry.get_provider(provider_name)
+            if not provider:
+                logger.warning(
+                    f"Block {block_cls.__name__} credential input '{field_name}' "
+                    f"refers to unknown provider '{provider_name}' - Disabling"
+                )
+                return False
+
+            # Check the provider's supported auth types
+            if field_info.supported_types != provider.supported_auth_types:
+                logger.warning(
+                    f"Block {block_cls.__name__} credential input '{field_name}' "
+                    f"has mismatched supported auth types (field <> Provider): "
+                    f"{field_info.supported_types} != {provider.supported_auth_types}"
+                )
+
+            if not (supported_auth_types := provider.supported_auth_types):
+                # No auth methods are been configured for this provider
+                logger.warning(
+                    f"Block {block_cls.__name__} credential input '{field_name}' "
+                    f"provider '{provider_name}' "
+                    "has no authentication methods configured - Disabling"
+                )
+                return False
+
+            # Check if provider supports OAuth
+            if "oauth2" in supported_auth_types:
+                # Check if OAuth environment variables are set
+                if (oauth_config := provider.oauth_config) and bool(
+                    os.getenv(oauth_config.client_id_env_var)
+                    and os.getenv(oauth_config.client_secret_env_var)
+                ):
+                    logger.debug(
+                        f"Block {block_cls.__name__} credential input '{field_name}' "
+                        f"provider '{provider_name}' is configured for OAuth"
+                    )
+                else:
+                    logger.error(
+                        f"Block {block_cls.__name__} credential input '{field_name}' "
+                        f"provider '{provider_name}' "
+                        "is missing OAuth client ID or secret - Disabling"
+                    )
+                    return False
+
+        logger.debug(
+            f"Block {block_cls.__name__} credential input '{field_name}' is valid; "
+            f"supported credential types: {', '.join(field_info.supported_types)}"
+        )
+
+    return True

autogpt_platform/backend/backend/blocks/agent.py

@@ -1,7 +1,7 @@
 import logging
-from typing import Any, Optional
+from typing import TYPE_CHECKING, Any, Optional
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockInput,
@@ -9,13 +9,15 @@ from backend.data.block import (
     BlockSchema,
     BlockSchemaInput,
     BlockType,
-    get_block,
 )
 from backend.data.execution import ExecutionContext, ExecutionStatus, NodesInputMasks
 from backend.data.model import NodeExecutionStats, SchemaField
 from backend.util.json import validate_with_jsonschema
 from backend.util.retry import func_retry
 
+if TYPE_CHECKING:
+    from backend.executor.utils import LogMetadata
+
 _logger = logging.getLogger(__name__)
 
 
@@ -124,9 +126,10 @@ class AgentExecutorBlock(Block):
         graph_version: int,
         graph_exec_id: str,
         user_id: str,
-        logger,
+        logger: "LogMetadata",
     ) -> BlockOutput:
 
+        from backend.blocks import get_block
         from backend.data.execution import ExecutionEventType
         from backend.executor import utils as execution_utils
 
@@ -198,7 +201,7 @@ class AgentExecutorBlock(Block):
         self,
         graph_exec_id: str,
         user_id: str,
-        logger,
+        logger: "LogMetadata",
     ) -> None:
         from backend.executor import utils as execution_utils
 

autogpt_platform/backend/backend/blocks/ai_condition.py

@@ -1,5 +1,11 @@
 from typing import Any
 
+from backend.blocks._base import (
+    BlockCategory,
+    BlockOutput,
+    BlockSchemaInput,
+    BlockSchemaOutput,
+)
 from backend.blocks.llm import (
     DEFAULT_LLM_MODEL,
     TEST_CREDENTIALS,
@@ -11,12 +17,6 @@ from backend.blocks.llm import (
     LLMResponse,
     llm_call,
 )
-from backend.data.block import (
-    BlockCategory,
-    BlockOutput,
-    BlockSchemaInput,
-    BlockSchemaOutput,
-)
 from backend.data.model import APIKeyCredentials, NodeExecutionStats, SchemaField
 
 

autogpt_platform/backend/backend/blocks/ai_image_customizer.py

@@ -6,7 +6,7 @@ from pydantic import SecretStr
 from replicate.client import Client as ReplicateClient
 from replicate.helpers import FileOutput
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/ai_image_generator_block.py

@@ -5,7 +5,12 @@ from pydantic import SecretStr
 from replicate.client import Client as ReplicateClient
 from replicate.helpers import FileOutput
 
-from backend.data.block import Block, BlockCategory, BlockSchemaInput, BlockSchemaOutput
+from backend.blocks._base import (
+    Block,
+    BlockCategory,
+    BlockSchemaInput,
+    BlockSchemaOutput,
+)
 from backend.data.execution import ExecutionContext
 from backend.data.model import (
     APIKeyCredentials,

autogpt_platform/backend/backend/blocks/ai_music_generator.py

@@ -6,7 +6,7 @@ from typing import Literal
 from pydantic import SecretStr
 from replicate.client import Client as ReplicateClient
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/ai_shortform_video_block.py

@@ -6,7 +6,7 @@ from typing import Literal
 
 from pydantic import SecretStr
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

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

@@ -1,3 +1,10 @@
+from backend.blocks._base import (
+    Block,
+    BlockCategory,
+    BlockOutput,
+    BlockSchemaInput,
+    BlockSchemaOutput,
+)
 from backend.blocks.apollo._api import ApolloClient
 from backend.blocks.apollo._auth import (
     TEST_CREDENTIALS,
@@ -10,13 +17,6 @@ from backend.blocks.apollo.models import (
     PrimaryPhone,
     SearchOrganizationsRequest,
 )
-from backend.data.block import (
-    Block,
-    BlockCategory,
-    BlockOutput,
-    BlockSchemaInput,
-    BlockSchemaOutput,
-)
 from backend.data.model import CredentialsField, SchemaField
 
 

autogpt_platform/backend/backend/blocks/apollo/people.py

@@ -1,5 +1,12 @@
 import asyncio
 
+from backend.blocks._base import (
+    Block,
+    BlockCategory,
+    BlockOutput,
+    BlockSchemaInput,
+    BlockSchemaOutput,
+)
 from backend.blocks.apollo._api import ApolloClient
 from backend.blocks.apollo._auth import (
     TEST_CREDENTIALS,
@@ -14,13 +21,6 @@ from backend.blocks.apollo.models import (
     SearchPeopleRequest,
     SenorityLevels,
 )
-from backend.data.block import (
-    Block,
-    BlockCategory,
-    BlockOutput,
-    BlockSchemaInput,
-    BlockSchemaOutput,
-)
 from backend.data.model import CredentialsField, SchemaField
 
 

autogpt_platform/backend/backend/blocks/apollo/person.py

@@ -1,3 +1,10 @@
+from backend.blocks._base import (
+    Block,
+    BlockCategory,
+    BlockOutput,
+    BlockSchemaInput,
+    BlockSchemaOutput,
+)
 from backend.blocks.apollo._api import ApolloClient
 from backend.blocks.apollo._auth import (
     TEST_CREDENTIALS,
@@ -6,13 +13,6 @@ from backend.blocks.apollo._auth import (
     ApolloCredentialsInput,
 )
 from backend.blocks.apollo.models import Contact, EnrichPersonRequest
-from backend.data.block import (
-    Block,
-    BlockCategory,
-    BlockOutput,
-    BlockSchemaInput,
-    BlockSchemaOutput,
-)
 from backend.data.model import CredentialsField, SchemaField
 
 

autogpt_platform/backend/backend/blocks/ayrshare/_util.py

@@ -3,7 +3,7 @@ from typing import Optional
 
 from pydantic import BaseModel, Field
 
-from backend.data.block import BlockSchemaInput
+from backend.blocks._base import BlockSchemaInput
 from backend.data.model import SchemaField, UserIntegrations
 from backend.integrations.ayrshare import AyrshareClient
 from backend.util.clients import get_database_manager_async_client

autogpt_platform/backend/backend/blocks/basic.py

@@ -1,7 +1,7 @@
 import enum
 from typing import Any
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/block.py

@@ -2,7 +2,7 @@ import os
 import re
 from typing import Type
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/branching.py

@@ -1,7 +1,7 @@
 from enum import Enum
 from typing import Any
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/claude_code.py

@@ -6,7 +6,7 @@ from typing import Literal, Optional
 from e2b import AsyncSandbox as BaseAsyncSandbox
 from pydantic import BaseModel, SecretStr
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/code_executor.py

@@ -6,7 +6,7 @@ from e2b_code_interpreter import Result as E2BExecutionResult
 from e2b_code_interpreter.charts import Chart as E2BExecutionResultChart
 from pydantic import BaseModel, Field, JsonValue, SecretStr
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/code_extraction_block.py

@@ -1,6 +1,6 @@
 import re
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/codex.py

@@ -6,7 +6,7 @@ from openai import AsyncOpenAI
 from openai.types.responses import Response as OpenAIResponse
 from pydantic import SecretStr
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/compass/triggers.py

@@ -1,6 +1,6 @@
 from pydantic import BaseModel
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockManualWebhookConfig,

autogpt_platform/backend/backend/blocks/count_words_and_char_block.py

@@ -1,4 +1,4 @@
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/data_manipulation.py

@@ -1,6 +1,6 @@
 from typing import Any, List
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/decoder_block.py

@@ -1,6 +1,6 @@
 import codecs
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/discord/bot_blocks.py

@@ -8,7 +8,7 @@ from typing import Any, Literal, cast
 import discord
 from pydantic import SecretStr
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/discord/oauth_blocks.py

@@ -2,7 +2,7 @@
 Discord OAuth-based blocks.
 """
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/elevenlabs/_auth.py

@@ -0,0 +1,28 @@
+"""ElevenLabs integration blocks - test credentials and shared utilities."""
+
+from typing import Literal
+
+from pydantic import SecretStr
+
+from backend.data.model import APIKeyCredentials, CredentialsMetaInput
+from backend.integrations.providers import ProviderName
+
+TEST_CREDENTIALS = APIKeyCredentials(
+    id="01234567-89ab-cdef-0123-456789abcdef",
+    provider="elevenlabs",
+    api_key=SecretStr("mock-elevenlabs-api-key"),
+    title="Mock ElevenLabs API key",
+    expires_at=None,
+)
+
+TEST_CREDENTIALS_INPUT = {
+    "provider": TEST_CREDENTIALS.provider,
+    "id": TEST_CREDENTIALS.id,
+    "type": TEST_CREDENTIALS.type,
+    "title": TEST_CREDENTIALS.title,
+}
+
+ElevenLabsCredentials = APIKeyCredentials
+ElevenLabsCredentialsInput = CredentialsMetaInput[
+    Literal[ProviderName.ELEVENLABS], Literal["api_key"]
+]

autogpt_platform/backend/backend/blocks/email_block.py

@@ -7,7 +7,7 @@ from typing import Literal
 
 from pydantic import BaseModel, ConfigDict, SecretStr
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/encoder_block.py

@@ -0,0 +1,77 @@
+"""Text encoding block for converting special characters to escape sequences."""
+
+import codecs
+
+from backend.blocks._base import (
+    Block,
+    BlockCategory,
+    BlockOutput,
+    BlockSchemaInput,
+    BlockSchemaOutput,
+)
+from backend.data.model import SchemaField
+
+
+class TextEncoderBlock(Block):
+    """
+    Encodes a string by converting special characters into escape sequences.
+
+    This block is the inverse of TextDecoderBlock. It takes text containing
+    special characters (like newlines, tabs, etc.) and converts them into
+    their escape sequence representations (e.g., newline becomes \\n).
+    """
+
+    class Input(BlockSchemaInput):
+        """Input schema for TextEncoderBlock."""
+
+        text: str = SchemaField(
+            description="A string containing special characters to be encoded",
+            placeholder="Your text with newlines and quotes to encode",
+        )
+
+    class Output(BlockSchemaOutput):
+        """Output schema for TextEncoderBlock."""
+
+        encoded_text: str = SchemaField(
+            description="The encoded text with special characters converted to escape sequences"
+        )
+        error: str = SchemaField(description="Error message if encoding fails")
+
+    def __init__(self):
+        super().__init__(
+            id="5185f32e-4b65-4ecf-8fbb-873f003f09d6",
+            description="Encodes a string by converting special characters into escape sequences",
+            categories={BlockCategory.TEXT},
+            input_schema=TextEncoderBlock.Input,
+            output_schema=TextEncoderBlock.Output,
+            test_input={
+                "text": """Hello
+World!
+This is a "quoted" string."""
+            },
+            test_output=[
+                (
+                    "encoded_text",
+                    """Hello\\nWorld!\\nThis is a "quoted" string.""",
+                )
+            ],
+        )
+
+    async def run(self, input_data: Input, **kwargs) -> BlockOutput:
+        """
+        Encode the input text by converting special characters to escape sequences.
+
+        Args:
+            input_data: The input containing the text to encode.
+            **kwargs: Additional keyword arguments (unused).
+
+        Yields:
+            The encoded text with escape sequences, or an error message if encoding fails.
+        """
+        try:
+            encoded_text = codecs.encode(input_data.text, "unicode_escape").decode(
+                "utf-8"
+            )
+            yield "encoded_text", encoded_text
+        except Exception as e:
+            yield "error", f"Encoding error: {str(e)}"

autogpt_platform/backend/backend/blocks/enrichlayer/linkedin.py

@@ -8,7 +8,7 @@ which provides access to LinkedIn profile data and related information.
 import logging
 from typing import Optional
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/exa/websets.py

@@ -478,7 +478,7 @@ class ExaCreateOrFindWebsetBlock(Block):
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
         try:
-            webset = aexa.websets.get(id=input_data.external_id)
+            webset = await aexa.websets.get(id=input_data.external_id)
             webset_result = Webset.model_validate(webset.model_dump(by_alias=True))
 
             yield "webset", webset_result
@@ -494,7 +494,7 @@ class ExaCreateOrFindWebsetBlock(Block):
                         count=input_data.search_count,
                     )
 
-                webset = aexa.websets.create(
+                webset = await aexa.websets.create(
                     params=CreateWebsetParameters(
                         search=search_params,
                         external_id=input_data.external_id,
@@ -554,7 +554,7 @@ class ExaUpdateWebsetBlock(Block):
         if input_data.metadata is not None:
             payload["metadata"] = input_data.metadata
 
-        sdk_webset = aexa.websets.update(id=input_data.webset_id, params=payload)
+        sdk_webset = await aexa.websets.update(id=input_data.webset_id, params=payload)
 
         status_str = (
             sdk_webset.status.value
@@ -617,7 +617,7 @@ class ExaListWebsetsBlock(Block):
     ) -> BlockOutput:
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        response = aexa.websets.list(
+        response = await aexa.websets.list(
             cursor=input_data.cursor,
             limit=input_data.limit,
         )
@@ -678,7 +678,7 @@ class ExaGetWebsetBlock(Block):
     ) -> BlockOutput:
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        sdk_webset = aexa.websets.get(id=input_data.webset_id)
+        sdk_webset = await aexa.websets.get(id=input_data.webset_id)
 
         status_str = (
             sdk_webset.status.value
@@ -748,7 +748,7 @@ class ExaDeleteWebsetBlock(Block):
     ) -> BlockOutput:
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        deleted_webset = aexa.websets.delete(id=input_data.webset_id)
+        deleted_webset = await aexa.websets.delete(id=input_data.webset_id)
 
         status_str = (
             deleted_webset.status.value
@@ -798,7 +798,7 @@ class ExaCancelWebsetBlock(Block):
     ) -> BlockOutput:
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        canceled_webset = aexa.websets.cancel(id=input_data.webset_id)
+        canceled_webset = await aexa.websets.cancel(id=input_data.webset_id)
 
         status_str = (
             canceled_webset.status.value
@@ -968,7 +968,7 @@ class ExaPreviewWebsetBlock(Block):
                 entity["description"] = input_data.entity_description
             payload["entity"] = entity
 
-        sdk_preview = aexa.websets.preview(params=payload)
+        sdk_preview = await aexa.websets.preview(params=payload)
 
         preview = PreviewWebsetModel.from_sdk(sdk_preview)
 
@@ -1051,7 +1051,7 @@ class ExaWebsetStatusBlock(Block):
     ) -> BlockOutput:
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        webset = aexa.websets.get(id=input_data.webset_id)
+        webset = await aexa.websets.get(id=input_data.webset_id)
 
         status = (
             webset.status.value
@@ -1185,7 +1185,7 @@ class ExaWebsetSummaryBlock(Block):
     ) -> BlockOutput:
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        webset = aexa.websets.get(id=input_data.webset_id)
+        webset = await aexa.websets.get(id=input_data.webset_id)
 
         # Extract basic info
         webset_id = webset.id
@@ -1211,7 +1211,7 @@ class ExaWebsetSummaryBlock(Block):
         total_items = 0
 
         if input_data.include_sample_items and input_data.sample_size > 0:
-            items_response = aexa.websets.items.list(
+            items_response = await aexa.websets.items.list(
                 webset_id=input_data.webset_id, limit=input_data.sample_size
             )
             sample_items_data = [
@@ -1362,7 +1362,7 @@ class ExaWebsetReadyCheckBlock(Block):
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
         # Get webset details
-        webset = aexa.websets.get(id=input_data.webset_id)
+        webset = await aexa.websets.get(id=input_data.webset_id)
 
         status = (
             webset.status.value

autogpt_platform/backend/backend/blocks/exa/websets_enrichment.py

@@ -202,7 +202,7 @@ class ExaCreateEnrichmentBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        sdk_enrichment = aexa.websets.enrichments.create(
+        sdk_enrichment = await aexa.websets.enrichments.create(
             webset_id=input_data.webset_id, params=payload
         )
 
@@ -223,7 +223,7 @@ class ExaCreateEnrichmentBlock(Block):
             items_enriched = 0
 
             while time.time() - poll_start < input_data.polling_timeout:
-                current_enrich = aexa.websets.enrichments.get(
+                current_enrich = await aexa.websets.enrichments.get(
                     webset_id=input_data.webset_id, id=enrichment_id
                 )
                 current_status = (
@@ -234,7 +234,7 @@ class ExaCreateEnrichmentBlock(Block):
 
                 if current_status in ["completed", "failed", "cancelled"]:
                     # Estimate items from webset searches
-                    webset = aexa.websets.get(id=input_data.webset_id)
+                    webset = await aexa.websets.get(id=input_data.webset_id)
                     if webset.searches:
                         for search in webset.searches:
                             if search.progress:
@@ -329,7 +329,7 @@ class ExaGetEnrichmentBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        sdk_enrichment = aexa.websets.enrichments.get(
+        sdk_enrichment = await aexa.websets.enrichments.get(
             webset_id=input_data.webset_id, id=input_data.enrichment_id
         )
 
@@ -474,7 +474,7 @@ class ExaDeleteEnrichmentBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        deleted_enrichment = aexa.websets.enrichments.delete(
+        deleted_enrichment = await aexa.websets.enrichments.delete(
             webset_id=input_data.webset_id, id=input_data.enrichment_id
         )
 
@@ -525,13 +525,13 @@ class ExaCancelEnrichmentBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        canceled_enrichment = aexa.websets.enrichments.cancel(
+        canceled_enrichment = await aexa.websets.enrichments.cancel(
             webset_id=input_data.webset_id, id=input_data.enrichment_id
         )
 
         # Try to estimate how many items were enriched before cancellation
         items_enriched = 0
-        items_response = aexa.websets.items.list(
+        items_response = await aexa.websets.items.list(
             webset_id=input_data.webset_id, limit=100
         )
 

autogpt_platform/backend/backend/blocks/exa/websets_import_export.py

@@ -222,7 +222,7 @@ class ExaCreateImportBlock(Block):
     def _create_test_mock():
         """Create test mocks for the AsyncExa SDK."""
         from datetime import datetime
-        from unittest.mock import MagicMock
+        from unittest.mock import AsyncMock, MagicMock
 
         # Create mock SDK import object
         mock_import = MagicMock()
@@ -247,7 +247,7 @@ class ExaCreateImportBlock(Block):
         return {
             "_get_client": lambda *args, **kwargs: MagicMock(
                 websets=MagicMock(
-                    imports=MagicMock(create=lambda *args, **kwargs: mock_import)
+                    imports=MagicMock(create=AsyncMock(return_value=mock_import))
                 )
             )
         }
@@ -294,7 +294,7 @@ class ExaCreateImportBlock(Block):
         if input_data.metadata:
             payload["metadata"] = input_data.metadata
 
-        sdk_import = aexa.websets.imports.create(
+        sdk_import = await aexa.websets.imports.create(
             params=payload, csv_data=input_data.csv_data
         )
 
@@ -360,7 +360,7 @@ class ExaGetImportBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        sdk_import = aexa.websets.imports.get(import_id=input_data.import_id)
+        sdk_import = await aexa.websets.imports.get(import_id=input_data.import_id)
 
         import_obj = ImportModel.from_sdk(sdk_import)
 
@@ -426,7 +426,7 @@ class ExaListImportsBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        response = aexa.websets.imports.list(
+        response = await aexa.websets.imports.list(
             cursor=input_data.cursor,
             limit=input_data.limit,
         )
@@ -474,7 +474,9 @@ class ExaDeleteImportBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        deleted_import = aexa.websets.imports.delete(import_id=input_data.import_id)
+        deleted_import = await aexa.websets.imports.delete(
+            import_id=input_data.import_id
+        )
 
         yield "import_id", deleted_import.id
         yield "success", "true"
@@ -573,14 +575,14 @@ class ExaExportWebsetBlock(Block):
             }
         )
 
-        # Create mock iterator
+        # Create async iterator for list_all
-        mock_items = [mock_item1, mock_item2]
+        async def async_item_iterator(*args, **kwargs):
+            for item in [mock_item1, mock_item2]:
+                yield item
 
         return {
             "_get_client": lambda *args, **kwargs: MagicMock(
-                websets=MagicMock(
+                websets=MagicMock(items=MagicMock(list_all=async_item_iterator))
-                    items=MagicMock(list_all=lambda *args, **kwargs: iter(mock_items))
-                )
             )
         }
 
@@ -602,7 +604,7 @@ class ExaExportWebsetBlock(Block):
                 webset_id=input_data.webset_id, limit=input_data.max_items
             )
 
-            for sdk_item in item_iterator:
+            async for sdk_item in item_iterator:
                 if len(all_items) >= input_data.max_items:
                     break
 

autogpt_platform/backend/backend/blocks/exa/websets_items.py

@@ -178,7 +178,7 @@ class ExaGetWebsetItemBlock(Block):
     ) -> BlockOutput:
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        sdk_item = aexa.websets.items.get(
+        sdk_item = await aexa.websets.items.get(
             webset_id=input_data.webset_id, id=input_data.item_id
         )
 
@@ -269,7 +269,7 @@ class ExaListWebsetItemsBlock(Block):
             response = None
 
             while time.time() - start_time < input_data.wait_timeout:
-                response = aexa.websets.items.list(
+                response = await aexa.websets.items.list(
                     webset_id=input_data.webset_id,
                     cursor=input_data.cursor,
                     limit=input_data.limit,
@@ -282,13 +282,13 @@ class ExaListWebsetItemsBlock(Block):
                 interval = min(interval * 1.2, 10)
 
             if not response:
-                response = aexa.websets.items.list(
+                response = await aexa.websets.items.list(
                     webset_id=input_data.webset_id,
                     cursor=input_data.cursor,
                     limit=input_data.limit,
                 )
         else:
-            response = aexa.websets.items.list(
+            response = await aexa.websets.items.list(
                 webset_id=input_data.webset_id,
                 cursor=input_data.cursor,
                 limit=input_data.limit,
@@ -340,7 +340,7 @@ class ExaDeleteWebsetItemBlock(Block):
     ) -> BlockOutput:
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        deleted_item = aexa.websets.items.delete(
+        deleted_item = await aexa.websets.items.delete(
             webset_id=input_data.webset_id, id=input_data.item_id
         )
 
@@ -408,7 +408,7 @@ class ExaBulkWebsetItemsBlock(Block):
             webset_id=input_data.webset_id, limit=input_data.max_items
         )
 
-        for sdk_item in item_iterator:
+        async for sdk_item in item_iterator:
             if len(all_items) >= input_data.max_items:
                 break
 
@@ -475,7 +475,7 @@ class ExaWebsetItemsSummaryBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        webset = aexa.websets.get(id=input_data.webset_id)
+        webset = await aexa.websets.get(id=input_data.webset_id)
 
         entity_type = "unknown"
         if webset.searches:
@@ -495,7 +495,7 @@ class ExaWebsetItemsSummaryBlock(Block):
         # Get sample items if requested
         sample_items: List[WebsetItemModel] = []
         if input_data.sample_size > 0:
-            items_response = aexa.websets.items.list(
+            items_response = await aexa.websets.items.list(
                 webset_id=input_data.webset_id, limit=input_data.sample_size
             )
             # Convert to our stable models
@@ -569,7 +569,7 @@ class ExaGetNewItemsBlock(Block):
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
         # Get items starting from cursor
-        response = aexa.websets.items.list(
+        response = await aexa.websets.items.list(
             webset_id=input_data.webset_id,
             cursor=input_data.since_cursor,
             limit=input_data.max_items,

autogpt_platform/backend/backend/blocks/exa/websets_monitor.py

@@ -233,7 +233,7 @@ class ExaCreateMonitorBlock(Block):
     def _create_test_mock():
         """Create test mocks for the AsyncExa SDK."""
         from datetime import datetime
-        from unittest.mock import MagicMock
+        from unittest.mock import AsyncMock, MagicMock
 
         # Create mock SDK monitor object
         mock_monitor = MagicMock()
@@ -263,7 +263,7 @@ class ExaCreateMonitorBlock(Block):
         return {
             "_get_client": lambda *args, **kwargs: MagicMock(
                 websets=MagicMock(
-                    monitors=MagicMock(create=lambda *args, **kwargs: mock_monitor)
+                    monitors=MagicMock(create=AsyncMock(return_value=mock_monitor))
                 )
             )
         }
@@ -320,7 +320,7 @@ class ExaCreateMonitorBlock(Block):
         if input_data.metadata:
             payload["metadata"] = input_data.metadata
 
-        sdk_monitor = aexa.websets.monitors.create(params=payload)
+        sdk_monitor = await aexa.websets.monitors.create(params=payload)
 
         monitor = MonitorModel.from_sdk(sdk_monitor)
 
@@ -384,7 +384,7 @@ class ExaGetMonitorBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        sdk_monitor = aexa.websets.monitors.get(monitor_id=input_data.monitor_id)
+        sdk_monitor = await aexa.websets.monitors.get(monitor_id=input_data.monitor_id)
 
         monitor = MonitorModel.from_sdk(sdk_monitor)
 
@@ -476,7 +476,7 @@ class ExaUpdateMonitorBlock(Block):
         if input_data.metadata is not None:
             payload["metadata"] = input_data.metadata
 
-        sdk_monitor = aexa.websets.monitors.update(
+        sdk_monitor = await aexa.websets.monitors.update(
             monitor_id=input_data.monitor_id, params=payload
         )
 
@@ -522,7 +522,9 @@ class ExaDeleteMonitorBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        deleted_monitor = aexa.websets.monitors.delete(monitor_id=input_data.monitor_id)
+        deleted_monitor = await aexa.websets.monitors.delete(
+            monitor_id=input_data.monitor_id
+        )
 
         yield "monitor_id", deleted_monitor.id
         yield "success", "true"
@@ -579,7 +581,7 @@ class ExaListMonitorsBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        response = aexa.websets.monitors.list(
+        response = await aexa.websets.monitors.list(
             cursor=input_data.cursor,
             limit=input_data.limit,
             webset_id=input_data.webset_id,

autogpt_platform/backend/backend/blocks/exa/websets_polling.py

@@ -121,7 +121,7 @@ class ExaWaitForWebsetBlock(Block):
                 WebsetTargetStatus.IDLE,
                 WebsetTargetStatus.ANY_COMPLETE,
             ]:
-                final_webset = aexa.websets.wait_until_idle(
+                final_webset = await aexa.websets.wait_until_idle(
                     id=input_data.webset_id,
                     timeout=input_data.timeout,
                     poll_interval=input_data.check_interval,
@@ -164,7 +164,7 @@ class ExaWaitForWebsetBlock(Block):
                 interval = input_data.check_interval
                 while time.time() - start_time < input_data.timeout:
                     # Get current webset status
-                    webset = aexa.websets.get(id=input_data.webset_id)
+                    webset = await aexa.websets.get(id=input_data.webset_id)
                     current_status = (
                         webset.status.value
                         if hasattr(webset.status, "value")
@@ -209,7 +209,7 @@ class ExaWaitForWebsetBlock(Block):
 
                 # Timeout reached
                 elapsed = time.time() - start_time
-                webset = aexa.websets.get(id=input_data.webset_id)
+                webset = await aexa.websets.get(id=input_data.webset_id)
                 final_status = (
                     webset.status.value
                     if hasattr(webset.status, "value")
@@ -345,7 +345,7 @@ class ExaWaitForSearchBlock(Block):
         try:
             while time.time() - start_time < input_data.timeout:
                 # Get current search status using SDK
-                search = aexa.websets.searches.get(
+                search = await aexa.websets.searches.get(
                     webset_id=input_data.webset_id, id=input_data.search_id
                 )
 
@@ -401,7 +401,7 @@ class ExaWaitForSearchBlock(Block):
             elapsed = time.time() - start_time
 
             # Get last known status
-            search = aexa.websets.searches.get(
+            search = await aexa.websets.searches.get(
                 webset_id=input_data.webset_id, id=input_data.search_id
             )
             final_status = (
@@ -503,7 +503,7 @@ class ExaWaitForEnrichmentBlock(Block):
         try:
             while time.time() - start_time < input_data.timeout:
                 # Get current enrichment status using SDK
-                enrichment = aexa.websets.enrichments.get(
+                enrichment = await aexa.websets.enrichments.get(
                     webset_id=input_data.webset_id, id=input_data.enrichment_id
                 )
 
@@ -548,7 +548,7 @@ class ExaWaitForEnrichmentBlock(Block):
             elapsed = time.time() - start_time
 
             # Get last known status
-            enrichment = aexa.websets.enrichments.get(
+            enrichment = await aexa.websets.enrichments.get(
                 webset_id=input_data.webset_id, id=input_data.enrichment_id
             )
             final_status = (
@@ -575,7 +575,7 @@ class ExaWaitForEnrichmentBlock(Block):
     ) -> tuple[list[SampleEnrichmentModel], int]:
         """Get sample enriched data and count."""
         # Get a few items to see enrichment results using SDK
-        response = aexa.websets.items.list(webset_id=webset_id, limit=5)
+        response = await aexa.websets.items.list(webset_id=webset_id, limit=5)
 
         sample_data: list[SampleEnrichmentModel] = []
         enriched_count = 0

autogpt_platform/backend/backend/blocks/exa/websets_search.py

@@ -317,7 +317,7 @@ class ExaCreateWebsetSearchBlock(Block):
 
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        sdk_search = aexa.websets.searches.create(
+        sdk_search = await aexa.websets.searches.create(
             webset_id=input_data.webset_id, params=payload
         )
 
@@ -350,7 +350,7 @@ class ExaCreateWebsetSearchBlock(Block):
             poll_start = time.time()
 
             while time.time() - poll_start < input_data.polling_timeout:
-                current_search = aexa.websets.searches.get(
+                current_search = await aexa.websets.searches.get(
                     webset_id=input_data.webset_id, id=search_id
                 )
                 current_status = (
@@ -442,7 +442,7 @@ class ExaGetWebsetSearchBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        sdk_search = aexa.websets.searches.get(
+        sdk_search = await aexa.websets.searches.get(
             webset_id=input_data.webset_id, id=input_data.search_id
         )
 
@@ -523,7 +523,7 @@ class ExaCancelWebsetSearchBlock(Block):
         # Use AsyncExa SDK
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
-        canceled_search = aexa.websets.searches.cancel(
+        canceled_search = await aexa.websets.searches.cancel(
             webset_id=input_data.webset_id, id=input_data.search_id
         )
 
@@ -604,7 +604,7 @@ class ExaFindOrCreateSearchBlock(Block):
         aexa = AsyncExa(api_key=credentials.api_key.get_secret_value())
 
         # Get webset to check existing searches
-        webset = aexa.websets.get(id=input_data.webset_id)
+        webset = await aexa.websets.get(id=input_data.webset_id)
 
         # Look for existing search with same query
         existing_search = None
@@ -636,7 +636,7 @@ class ExaFindOrCreateSearchBlock(Block):
             if input_data.entity_type != SearchEntityType.AUTO:
                 payload["entity"] = {"type": input_data.entity_type.value}
 
-            sdk_search = aexa.websets.searches.create(
+            sdk_search = await aexa.websets.searches.create(
                 webset_id=input_data.webset_id, params=payload
             )
 

autogpt_platform/backend/backend/blocks/fal/ai_video_generator.py

@@ -3,6 +3,13 @@ import logging
 from enum import Enum
 from typing import Any
 
+from backend.blocks._base import (
+    Block,
+    BlockCategory,
+    BlockOutput,
+    BlockSchemaInput,
+    BlockSchemaOutput,
+)
 from backend.blocks.fal._auth import (
     TEST_CREDENTIALS,
     TEST_CREDENTIALS_INPUT,
@@ -10,13 +17,6 @@ from backend.blocks.fal._auth import (
     FalCredentialsField,
     FalCredentialsInput,
 )
-from backend.data.block import (
-    Block,
-    BlockCategory,
-    BlockOutput,
-    BlockSchemaInput,
-    BlockSchemaOutput,
-)
 from backend.data.execution import ExecutionContext
 from backend.data.model import SchemaField
 from backend.util.file import store_media_file

autogpt_platform/backend/backend/blocks/flux_kontext.py

@@ -5,7 +5,7 @@ from pydantic import SecretStr
 from replicate.client import Client as ReplicateClient
 from replicate.helpers import FileOutput
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/github/checks.py

@@ -3,7 +3,7 @@ from typing import Optional
 
 from pydantic import BaseModel
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/github/ci.py

@@ -5,7 +5,7 @@ from typing import Optional
 
 from typing_extensions import TypedDict
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

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

@@ -3,7 +3,7 @@ from urllib.parse import urlparse
 
 from typing_extensions import TypedDict
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/github/pull_requests.py

@@ -2,7 +2,7 @@ import re
 
 from typing_extensions import TypedDict
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/github/repo.py

@@ -2,7 +2,7 @@ import base64
 
 from typing_extensions import TypedDict
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/github/reviews.py

@@ -4,7 +4,7 @@ from typing import Any, List, Optional
 
 from typing_extensions import TypedDict
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/github/statuses.py

@@ -3,7 +3,7 @@ from typing import Optional
 
 from pydantic import BaseModel
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,

autogpt_platform/backend/backend/blocks/github/triggers.py

@@ -4,7 +4,7 @@ from pathlib import Path
 
 from pydantic import BaseModel
 
-from backend.data.block import (
+from backend.blocks._base import (
     Block,
     BlockCategory,
     BlockOutput,