Debug test

* add o1 support + bump litellm/openai

* ports

* update exception message for testing

.github/actions/custom-build-and-push/action.yml

@@ -0,0 +1,76 @@
+name: 'Build and Push Docker Image with Retry'
+description: 'Attempts to build and push a Docker image, with a retry on failure'
+inputs:
+  context:
+    description: 'Build context'
+    required: true
+  file:
+    description: 'Dockerfile location'
+    required: true
+  platforms:
+    description: 'Target platforms'
+    required: true
+  pull:
+    description: 'Always attempt to pull a newer version of the image'
+    required: false
+    default: 'true'
+  push:
+    description: 'Push the image to registry'
+    required: false
+    default: 'true'
+  load:
+    description: 'Load the image into Docker daemon'
+    required: false
+    default: 'true'
+  tags:
+    description: 'Image tags'
+    required: true
+  cache-from:
+    description: 'Cache sources'
+    required: false
+  cache-to:
+    description: 'Cache destinations'
+    required: false
+  retry-wait-time:
+    description: 'Time to wait before retry in seconds'
+    required: false
+    default: '5'
+
+runs:
+  using: "composite"
+  steps:
+    - name: Build and push Docker image (First Attempt)
+      id: buildx1
+      uses: docker/build-push-action@v5
+      continue-on-error: true
+      with:
+        context: ${{ inputs.context }}
+        file: ${{ inputs.file }}
+        platforms: ${{ inputs.platforms }}
+        pull: ${{ inputs.pull }}
+        push: ${{ inputs.push }}
+        load: ${{ inputs.load }}
+        tags: ${{ inputs.tags }}
+        cache-from: ${{ inputs.cache-from }}
+        cache-to: ${{ inputs.cache-to }}
+
+    - name: Wait to retry
+      if: steps.buildx1.outcome != 'success'
+      run: |
+        echo "First attempt failed. Waiting ${{ inputs.retry-wait-time }} seconds before retry..."
+        sleep ${{ inputs.retry-wait-time }}
+      shell: bash
+
+    - name: Build and push Docker image (Retry Attempt)
+      if: steps.buildx1.outcome != 'success'
+      uses: docker/build-push-action@v5
+      with:
+        context: ${{ inputs.context }}
+        file: ${{ inputs.file }}
+        platforms: ${{ inputs.platforms }}
+        pull: ${{ inputs.pull }}
+        push: ${{ inputs.push }}
+        load: ${{ inputs.load }}
+        tags: ${{ inputs.tags }}
+        cache-from: ${{ inputs.cache-from }}
+        cache-to: ${{ inputs.cache-to }}

.github/workflows/docker-build-backend-container-on-merge-group.yml

@@ -1,33 +0,0 @@
-name: Build Backend Image on Merge Group
-
-on:
-  merge_group:
-    types: [checks_requested]
-
-env:
-  REGISTRY_IMAGE: danswer/danswer-backend
-
-jobs:
-  build:
-    # TODO: make this a matrix build like the web containers
-    runs-on: 
-      group: amd64-image-builders
-      
-    steps:
-    - name: Checkout code
-      uses: actions/checkout@v2
-
-    - name: Set up Docker Buildx
-      uses: docker/setup-buildx-action@v3
-
-    - name: Backend Image Docker Build
-      uses: docker/build-push-action@v5
-      with:
-        context: ./backend
-        file: ./backend/Dockerfile
-        platforms: linux/amd64,linux/arm64
-        push: false
-        tags: |
-          ${{ env.REGISTRY_IMAGE }}:latest
-        build-args: |
-          DANSWER_VERSION=v0.0.1

.github/workflows/docker-build-push-backend-container-on-tag.yml

@@ -27,6 +27,11 @@ jobs:
         username: ${{ secrets.DOCKER_USERNAME }}
         password: ${{ secrets.DOCKER_TOKEN }}
 
+    - name: Install build-essential
+      run: |
+        sudo apt-get update
+        sudo apt-get install -y build-essential
+        
     - name: Backend Image Docker Build and Push
       uses: docker/build-push-action@v5
       with:

.github/workflows/docker-build-web-container-on-merge-group.yml

@@ -1,53 +0,0 @@
-name: Build Web Image on Merge Group
-
-on:
-  merge_group:
-    types: [checks_requested]
-
-env:
-  REGISTRY_IMAGE: danswer/danswer-web-server
-
-jobs:
-  build:
-    runs-on: 
-      group: ${{ matrix.platform == 'linux/amd64' && 'amd64-image-builders' || 'arm64-image-builders' }}
-    strategy:
-      fail-fast: false
-      matrix:
-        platform:
-          - linux/amd64
-          - linux/arm64
-
-    steps:
-      - name: Prepare
-        run: |
-          platform=${{ matrix.platform }}
-          echo "PLATFORM_PAIR=${platform//\//-}" >> $GITHUB_ENV          
-      
-      - name: Checkout
-        uses: actions/checkout@v4
-      
-      - name: Docker meta
-        id: meta
-        uses: docker/metadata-action@v5
-        with:
-          images: ${{ env.REGISTRY_IMAGE }}
-          tags: |
-            type=raw,value=${{ env.REGISTRY_IMAGE }}:latest
-      
-      - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-    
-      - name: Build by digest
-        id: build
-        uses: docker/build-push-action@v5
-        with:
-          context: ./web
-          file: ./web/Dockerfile
-          platforms: ${{ matrix.platform }}
-          push: false
-          build-args: |
-            DANSWER_VERSION=v0.0.1
-          # needed due to weird interactions with the builds for different platforms  
-          no-cache: true
-          labels: ${{ steps.meta.outputs.labels }}

.github/workflows/pr-helm-chart-testing.yml.disabled.txt

@@ -0,0 +1,67 @@
+# This workflow is intentionally disabled while we're still working on it
+# It's close to ready, but a race condition needs to be fixed with
+# API server and Vespa startup, and it needs to have a way to build/test against
+# local containers
+
+name: Helm - Lint and Test Charts
+
+on:
+  merge_group:
+  pull_request:
+    branches: [ main ]
+
+jobs:
+  lint-test:
+    runs-on: Amd64
+
+    # fetch-depth 0 is required for helm/chart-testing-action
+    steps:
+    - name: Checkout code
+      uses: actions/checkout@v3
+      with:
+        fetch-depth: 0
+        
+    - name: Set up Helm
+      uses: azure/setup-helm@v4.2.0
+      with:
+        version: v3.14.4
+      
+    - name: Set up Python
+      uses: actions/setup-python@v4
+      with:
+        python-version: '3.11'
+        cache: 'pip'
+        cache-dependency-path: |
+          backend/requirements/default.txt
+          backend/requirements/dev.txt
+          backend/requirements/model_server.txt
+    - run: |
+        python -m pip install --upgrade pip
+        pip install --retries 5 --timeout 30 -r backend/requirements/default.txt
+        pip install --retries 5 --timeout 30 -r backend/requirements/dev.txt
+        pip install --retries 5 --timeout 30 -r backend/requirements/model_server.txt
+
+    - name: Set up chart-testing
+      uses: helm/chart-testing-action@v2.6.1
+
+    - name: Run chart-testing (list-changed)
+      id: list-changed
+      run: |
+        changed=$(ct list-changed --target-branch ${{ github.event.repository.default_branch }})
+        if [[ -n "$changed" ]]; then
+          echo "changed=true" >> "$GITHUB_OUTPUT"
+        fi
+
+    - name: Run chart-testing (lint)
+#       if: steps.list-changed.outputs.changed == 'true'
+      run: ct lint --all --config ct.yaml --target-branch ${{ github.event.repository.default_branch }}
+
+    - name: Create kind cluster
+#       if: steps.list-changed.outputs.changed == 'true'
+      uses: helm/kind-action@v1.10.0
+
+    - name: Run chart-testing (install)
+#       if: steps.list-changed.outputs.changed == 'true'
+      run: ct install --all --config ct.yaml
+#       run: ct install --target-branch ${{ github.event.repository.default_branch }}
+      

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

@@ -1,6 +1,7 @@
 name: Python Checks
 
 on:
+  merge_group:
   pull_request:
     branches: [ main ]
 
@@ -23,9 +24,9 @@ jobs:
           backend/requirements/model_server.txt
     - run: |
         python -m pip install --upgrade pip
-        pip install -r backend/requirements/default.txt
-        pip install -r backend/requirements/dev.txt
-        pip install -r backend/requirements/model_server.txt
+        pip install --retries 5 --timeout 30 -r backend/requirements/default.txt
+        pip install --retries 5 --timeout 30 -r backend/requirements/dev.txt
+        pip install --retries 5 --timeout 30 -r backend/requirements/model_server.txt
 
     - name: Run MyPy
       run: |

.github/workflows/pr-python-connector-tests.yml

@@ -0,0 +1,57 @@
+name: Connector Tests
+
+on:
+  pull_request:
+    branches: [main]
+  schedule:
+    # This cron expression runs the job daily at 16:00 UTC (9am PT)
+    - cron: "0 16 * * *"
+
+env:
+  # Confluence
+  CONFLUENCE_TEST_SPACE_URL: ${{ secrets.CONFLUENCE_TEST_SPACE_URL }}
+  CONFLUENCE_TEST_SPACE: ${{ secrets.CONFLUENCE_TEST_SPACE }}
+  CONFLUENCE_IS_CLOUD: ${{ secrets.CONFLUENCE_IS_CLOUD }}
+  CONFLUENCE_TEST_PAGE_ID: ${{ secrets.CONFLUENCE_TEST_PAGE_ID }}
+  CONFLUENCE_USER_NAME: ${{ secrets.CONFLUENCE_USER_NAME }}
+  CONFLUENCE_ACCESS_TOKEN: ${{ secrets.CONFLUENCE_ACCESS_TOKEN }}
+
+jobs:
+  connectors-check:
+    runs-on: ubuntu-latest
+
+    env:
+      PYTHONPATH: ./backend
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: "3.11"
+          cache: "pip"
+          cache-dependency-path: |
+            backend/requirements/default.txt
+            backend/requirements/dev.txt
+
+      - name: Install Dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install --retries 5 --timeout 30 -r backend/requirements/default.txt
+          pip install --retries 5 --timeout 30 -r backend/requirements/dev.txt
+
+      - name: Run Tests
+        shell: script -q -e -c "bash --noprofile --norc -eo pipefail {0}"
+        run: py.test -o junit_family=xunit2 -xv --ff backend/tests/daily/connectors
+
+      - name: Alert on Failure
+        if: failure() && github.event_name == 'schedule'
+        env:
+          SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }}
+        run: |
+          curl -X POST \
+            -H 'Content-type: application/json' \
+            --data '{"text":"Scheduled Connector Tests failed! Check the run at: https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }}"}' \
+            $SLACK_WEBHOOK

.github/workflows/pr-python-tests.yml

@@ -1,6 +1,7 @@
 name: Python Unit Tests
 
 on:
+  merge_group:
   pull_request:
     branches: [ main ]
 
@@ -10,7 +11,8 @@ jobs:
 
     env:
       PYTHONPATH: ./backend
-
+      REDIS_CLOUD_PYTEST_PASSWORD: ${{ secrets.REDIS_CLOUD_PYTEST_PASSWORD }}
+      
     steps:
     - name: Checkout code
       uses: actions/checkout@v4
@@ -27,8 +29,8 @@ jobs:
     - name: Install Dependencies
       run: |
         python -m pip install --upgrade pip
-        pip install -r backend/requirements/default.txt
-        pip install -r backend/requirements/dev.txt
+        pip install --retries 5 --timeout 30 -r backend/requirements/default.txt
+        pip install --retries 5 --timeout 30 -r backend/requirements/dev.txt
 
     - name: Run Tests
       shell: script -q -e -c "bash --noprofile --norc -eo pipefail {0}"

.github/workflows/pr-quality-checks.yml

@@ -4,18 +4,19 @@ concurrency:
   cancel-in-progress: true
 
 on:
+  merge_group:
   pull_request: null
 
 jobs:
   quality-checks:
     runs-on: ubuntu-latest
     steps:
-    - uses: actions/checkout@v4
-      with:
-        fetch-depth: 0
-    - uses: actions/setup-python@v5
-      with:
-        python-version: '3.11'
-    - uses: pre-commit/action@v3.0.0
-      with:
-        extra_args: --from-ref ${{ github.event.pull_request.base.sha }} --to-ref ${{ github.event.pull_request.head.sha }}
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+      - uses: pre-commit/action@v3.0.0
+        with:
+          extra_args: ${{ github.event_name == 'pull_request' && format('--from-ref {0} --to-ref {1}', github.event.pull_request.base.sha, github.event.pull_request.head.sha) || '' }}

.github/workflows/run-it.yml

@@ -0,0 +1,161 @@
+name: Run Integration Tests
+concurrency:
+  group: Run-Integration-Tests-${{ github.head_ref }}
+  cancel-in-progress: true
+
+on:
+  merge_group:
+  pull_request:
+    branches: [ main ]
+
+env:
+  OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+
+jobs:
+  integration-tests:
+    runs-on: Amd64
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKER_USERNAME }}
+          password: ${{ secrets.DOCKER_TOKEN }}
+
+      # NOTE: we don't need to build the Web Docker image since it's not used
+      # during the IT for now. We have a separate action to verify it builds 
+      # succesfully
+      - name: Pull Web Docker image
+        run: |
+          docker pull danswer/danswer-web-server:latest
+          docker tag danswer/danswer-web-server:latest danswer/danswer-web-server:it
+
+      - name: Build Backend Docker image
+        uses: ./.github/actions/custom-build-and-push
+        with:
+          context: ./backend
+          file: ./backend/Dockerfile
+          platforms: linux/amd64
+          tags: danswer/danswer-backend:it
+          cache-from: type=registry,ref=danswer/danswer-backend:it
+          cache-to: |
+            type=registry,ref=danswer/danswer-backend:it,mode=max
+            type=inline
+
+      - name: Build Model Server Docker image
+        uses: ./.github/actions/custom-build-and-push
+        with:
+          context: ./backend
+          file: ./backend/Dockerfile.model_server
+          platforms: linux/amd64
+          tags: danswer/danswer-model-server:it
+          cache-from: type=registry,ref=danswer/danswer-model-server:it
+          cache-to: |
+            type=registry,ref=danswer/danswer-model-server:it,mode=max
+            type=inline
+
+      - name: Build integration test Docker image
+        uses: ./.github/actions/custom-build-and-push
+        with:
+          context: ./backend
+          file: ./backend/tests/integration/Dockerfile
+          platforms: linux/amd64
+          tags: danswer/integration-test-runner:it
+          cache-from: type=registry,ref=danswer/integration-test-runner:it
+          cache-to: |
+            type=registry,ref=danswer/integration-test-runner:it,mode=max
+            type=inline
+
+      - name: Start Docker containers
+        run: |
+          cd deployment/docker_compose
+          ENABLE_PAID_ENTERPRISE_EDITION_FEATURES=true \
+          AUTH_TYPE=basic \
+          REQUIRE_EMAIL_VERIFICATION=false \
+          DISABLE_TELEMETRY=true \
+          IMAGE_TAG=it \
+          docker compose -f docker-compose.dev.yml -p danswer-stack up -d
+        id: start_docker
+
+      - name: Wait for service to be ready
+        run: |
+          echo "Starting wait-for-service script..."
+          
+          docker logs -f danswer-stack-api_server-1 &
+
+          start_time=$(date +%s)
+          timeout=300  # 5 minutes in seconds
+          
+          while true; do
+            current_time=$(date +%s)
+            elapsed_time=$((current_time - start_time))
+            
+            if [ $elapsed_time -ge $timeout ]; then
+              echo "Timeout reached. Service did not become ready in 5 minutes."
+              exit 1
+            fi
+            
+            # Use curl with error handling to ignore specific exit code 56
+            response=$(curl -s -o /dev/null -w "%{http_code}" http://localhost:8080/health || echo "curl_error")
+            
+            if [ "$response" = "200" ]; then
+              echo "Service is ready!"
+              break
+            elif [ "$response" = "curl_error" ]; then
+              echo "Curl encountered an error, possibly exit code 56. Continuing to retry..."
+            else
+              echo "Service not ready yet (HTTP status $response). Retrying in 5 seconds..."
+            fi
+            
+            sleep 5
+          done
+          echo "Finished waiting for service."
+
+      - name: Run integration tests
+        run: |
+          echo "Running integration tests..."
+          docker run --rm --network danswer-stack_default \
+            -e POSTGRES_HOST=relational_db \
+            -e POSTGRES_USER=postgres \
+            -e POSTGRES_PASSWORD=password \
+            -e POSTGRES_DB=postgres \
+            -e VESPA_HOST=index \
+            -e REDIS_HOST=cache \
+            -e API_SERVER_HOST=api_server \
+            -e OPENAI_API_KEY=${OPENAI_API_KEY} \
+            danswer/integration-test-runner:it
+        continue-on-error: true
+        id: run_tests
+
+      - name: Check test results
+        run: |
+          if [ ${{ steps.run_tests.outcome }} == 'failure' ]; then
+            echo "Integration tests failed. Exiting with error."
+            exit 1
+          else
+            echo "All integration tests passed successfully."
+          fi
+
+      - name: Save Docker logs
+        if: success() || failure()
+        run: |
+          cd deployment/docker_compose
+          docker compose -f docker-compose.dev.yml -p danswer-stack logs > docker-compose.log
+          mv docker-compose.log ${{ github.workspace }}/docker-compose.log
+      
+      - name: Upload logs
+        if: success() || failure()
+        uses: actions/upload-artifact@v3
+        with:
+          name: docker-logs
+          path: ${{ github.workspace }}/docker-compose.log
+
+      - name: Stop Docker containers
+        run: |
+          cd deployment/docker_compose
+          docker compose -f docker-compose.dev.yml -p danswer-stack down -v

.gitignore

@@ -4,6 +4,6 @@
 .mypy_cache
 .idea
 /deployment/data/nginx/app.conf
-.vscode/launch.json
+.vscode/
 *.sw?
 /backend/tests/regression/answer_quality/search_test_config.yaml

.vscode/env_template.txt

@@ -1,5 +1,5 @@
-# Copy this file to .env at the base of the repo and fill in the <REPLACE THIS> values
-# This will help with development iteration speed and reduce repeat tasks for dev
+# Copy this file to .env in the .vscode folder
+# Fill in the <REPLACE THIS> values as needed, it is recommended to set the GEN_AI_API_KEY value to avoid having to set up an LLM in the UI
 # Also check out danswer/backend/scripts/restart_containers.sh for a script to restart the containers which Danswer relies on outside of VSCode/Cursor processes
 
 # For local dev, often user Authentication is not needed
@@ -15,7 +15,7 @@ LOG_LEVEL=debug
 
 # This passes top N results to LLM an additional time for reranking prior to answer generation
 # This step is quite heavy on token usage so we disable it for dev generally
-DISABLE_LLM_DOC_RELEVANCE=True
+DISABLE_LLM_DOC_RELEVANCE=False
 
 
 # Useful if you want to toggle auth on/off (google_oauth/OIDC specifically)
@@ -27,9 +27,9 @@ REQUIRE_EMAIL_VERIFICATION=False
 
 # Set these so if you wipe the DB, you don't end up having to go through the UI every time
 GEN_AI_API_KEY=<REPLACE THIS>
-# If answer quality isn't important for dev, use 3.5 turbo due to it being cheaper
-GEN_AI_MODEL_VERSION=gpt-3.5-turbo
-FAST_GEN_AI_MODEL_VERSION=gpt-3.5-turbo
+# If answer quality isn't important for dev, use gpt-4o-mini since it's cheaper
+GEN_AI_MODEL_VERSION=gpt-4o
+FAST_GEN_AI_MODEL_VERSION=gpt-4o
 
 # For Danswer Slack Bot, overrides the UI values so no need to set this up via UI every time
 # Only needed if using DanswerBot
@@ -38,7 +38,7 @@ FAST_GEN_AI_MODEL_VERSION=gpt-3.5-turbo
 
 
 # Python stuff
-PYTHONPATH=./backend
+PYTHONPATH=../backend
 PYTHONUNBUFFERED=1
 
 
@@ -49,4 +49,3 @@ BING_API_KEY=<REPLACE THIS>
 # Enable the full set of Danswer Enterprise Edition features
 # NOTE: DO NOT ENABLE THIS UNLESS YOU HAVE A PAID ENTERPRISE LICENSE (or if you are using this for local testing/development)
 ENABLE_PAID_ENTERPRISE_EDITION_FEATURES=False
-

.vscode/launch.template.jsonc

@@ -1,15 +1,23 @@
-/*
-
-  Copy this file into '.vscode/launch.json' or merge its
-  contents into your existing configurations.
-
-*/
+/* Copy this file into '.vscode/launch.json' or merge its contents into your existing configurations. */
 
 {
     // Use IntelliSense to learn about possible attributes.
     // Hover to view descriptions of existing attributes.
     // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
     "version": "0.2.0",
+    "compounds": [
+        {
+            "name": "Run All Danswer Services",
+            "configurations": [
+                "Web Server",
+                "Model Server",
+                "API Server",
+                "Indexing",
+                "Background Jobs",
+                "Slack Bot"
+            ]
+        }
+    ],
     "configurations": [
         {
             "name": "Web Server",
@@ -17,7 +25,7 @@
             "request": "launch",
             "cwd": "${workspaceRoot}/web",
             "runtimeExecutable": "npm",
-            "envFile": "${workspaceFolder}/.env",
+            "envFile": "${workspaceFolder}/.vscode/.env",
             "runtimeArgs": [
                 "run", "dev"
             ],
@@ -25,11 +33,12 @@
         },
         {
             "name": "Model Server",
-            "type": "python",
+            "consoleName": "Model Server",
+            "type": "debugpy",
             "request": "launch",
             "module": "uvicorn",
             "cwd": "${workspaceFolder}/backend",
-            "envFile": "${workspaceFolder}/.env",
+            "envFile": "${workspaceFolder}/.vscode/.env",
             "env": {
                 "LOG_LEVEL": "DEBUG",
                 "PYTHONUNBUFFERED": "1"
@@ -39,16 +48,16 @@
                 "--reload",
                 "--port",
                 "9000"
-            ],
-            "consoleTitle": "Model Server"
+            ]
         },
         {
             "name": "API Server",
-            "type": "python",
+            "consoleName": "API Server",
+            "type": "debugpy",
             "request": "launch",
             "module": "uvicorn",
             "cwd": "${workspaceFolder}/backend",
-            "envFile": "${workspaceFolder}/.env",
+            "envFile": "${workspaceFolder}/.vscode/.env",
             "env": {
                 "LOG_DANSWER_MODEL_INTERACTIONS": "True",
                 "LOG_LEVEL": "DEBUG",
@@ -59,32 +68,32 @@
                 "--reload",
                 "--port",
                 "8080"
-            ],
-            "consoleTitle": "API Server"
+            ]
         },
         {
             "name": "Indexing",
-            "type": "python",
+            "consoleName": "Indexing",
+            "type": "debugpy",
             "request": "launch",
             "program": "danswer/background/update.py",
             "cwd": "${workspaceFolder}/backend",
-            "envFile": "${workspaceFolder}/.env",
+            "envFile": "${workspaceFolder}/.vscode/.env",
             "env": {
                 "ENABLE_MULTIPASS_INDEXING": "false",
                 "LOG_LEVEL": "DEBUG",
                 "PYTHONUNBUFFERED": "1",
                 "PYTHONPATH": "."
-            },
-            "consoleTitle": "Indexing"
+            }
         },
         // Celery and all async jobs, usually would include indexing as well but this is handled separately above for dev
         {
             "name": "Background Jobs",
-            "type": "python",
+            "consoleName": "Background Jobs",
+            "type": "debugpy",
             "request": "launch",
             "program": "scripts/dev_run_background_jobs.py",
             "cwd": "${workspaceFolder}/backend",
-            "envFile": "${workspaceFolder}/.env",
+            "envFile": "${workspaceFolder}/.vscode/.env",
             "env": {
                 "LOG_DANSWER_MODEL_INTERACTIONS": "True",
                 "LOG_LEVEL": "DEBUG",
@@ -93,18 +102,18 @@
             },
             "args": [
                 "--no-indexing"
-            ],
-            "consoleTitle": "Background Jobs"
+            ]
         },
         // For the listner to access the Slack API,
         // DANSWER_BOT_SLACK_APP_TOKEN & DANSWER_BOT_SLACK_BOT_TOKEN need to be set in .env file located in the root of the project
         {
             "name": "Slack Bot",
-            "type": "python",
+            "consoleName": "Slack Bot",
+            "type": "debugpy",
             "request": "launch",
             "program": "danswer/danswerbot/slack/listener.py",
             "cwd": "${workspaceFolder}/backend",
-            "envFile": "${workspaceFolder}/.env",
+            "envFile": "${workspaceFolder}/.vscode/.env",
             "env": {
                 "LOG_LEVEL": "DEBUG",
                 "PYTHONUNBUFFERED": "1",
@@ -113,11 +122,12 @@
         },
         {
             "name": "Pytest",
-            "type": "python",
+            "consoleName": "Pytest",
+            "type": "debugpy",
             "request": "launch",
             "module": "pytest",
             "cwd": "${workspaceFolder}/backend",
-            "envFile": "${workspaceFolder}/.env",
+            "envFile": "${workspaceFolder}/.vscode/.env",
             "env": {
                 "LOG_LEVEL": "DEBUG",
                 "PYTHONUNBUFFERED": "1",
@@ -128,18 +138,16 @@
                 // Specify a sepcific module/test to run or provide nothing to run all tests
                 //"tests/unit/danswer/llm/answering/test_prune_and_merge.py"
             ]
-        }
-    ],
-    "compounds": [
+        },
         {
-            "name": "Run Danswer",
-            "configurations": [
-                "Web Server",
-                "Model Server",
-                "API Server",
-                "Indexing",
-                "Background Jobs",
-            ]
+            "name": "Clear and Restart External Volumes and Containers",
+            "type": "node",
+            "request": "launch",
+            "runtimeExecutable": "bash",
+            "runtimeArgs": ["${workspaceFolder}/backend/scripts/restart_containers.sh"],
+            "cwd": "${workspaceFolder}",
+            "console": "integratedTerminal",
+            "stopOnEntry": true
         }
     ]
 }

CONTRIBUTING.md

@@ -48,23 +48,26 @@ We would love to see you there!
 
 
 ## Get Started 🚀
-Danswer being a fully functional app, relies on some external pieces of software, specifically:
+Danswer being a fully functional app, relies on some external software, specifically:
 - [Postgres](https://www.postgresql.org/) (Relational DB)
 - [Vespa](https://vespa.ai/) (Vector DB/Search Engine)
+- [Redis](https://redis.io/) (Cache)
+- [Nginx](https://nginx.org/) (Not needed for development flows generally)
 
-This guide provides instructions to set up the Danswer specific services outside of Docker because it's easier for
-development purposes but also feel free to just use the containers and update with local changes by providing the
-`--build` flag.
+
+> **Note:**
+> This guide provides instructions to build and run Danswer locally from source with Docker containers providing the above external software. We believe this combination is easier for
+> development purposes. If you prefer to use pre-built container images, we provide instructions on running the full Danswer stack within Docker below.
 
 
 ### Local Set Up
-It is recommended to use Python version 3.11
+Be sure to use Python version 3.11. For instructions on installing Python 3.11 on macOS, refer to the [CONTRIBUTING_MACOS.md](./CONTRIBUTING_MACOS.md) readme.
 
 If using a lower version, modifications will have to be made to the code.
-If using a higher version, the version of Tensorflow we use may not be available for your platform.
+If using a higher version, sometimes some libraries will not be available (i.e. we had problems with Tensorflow in the past with higher versions of python).
 
 
-#### Installing Requirements
+#### Backend: Python requirements
 Currently, we use pip and recommend creating a virtual environment.
 
 For convenience here's a command for it:
@@ -73,8 +76,9 @@ python -m venv .venv
 source .venv/bin/activate
 ```
 
---> Note that this virtual environment MUST NOT be set up WITHIN the danswer
-directory
+> **Note:**
+> This virtual environment MUST NOT be set up WITHIN the danswer directory if you plan on using mypy within certain IDEs.
+> For simplicity, we recommend setting up the virtual environment outside of the danswer directory.
 
 _For Windows, activate the virtual environment using Command Prompt:_
 ```bash
@@ -89,34 +93,38 @@ Install the required python dependencies:
 ```bash
 pip install -r danswer/backend/requirements/default.txt
 pip install -r danswer/backend/requirements/dev.txt
+pip install -r danswer/backend/requirements/ee.txt
 pip install -r danswer/backend/requirements/model_server.txt
 ```
 
+Install Playwright for Python (headless browser required by the Web Connector)
+
+In the activated Python virtualenv, install Playwright for Python by running:
+```bash
+playwright install
+```
+
+You may have to deactivate and reactivate your virtualenv for `playwright` to appear on your path.
+
+#### Frontend: Node dependencies
+
 Install [Node.js and npm](https://docs.npmjs.com/downloading-and-installing-node-js-and-npm) for the frontend.
 Once the above is done, navigate to `danswer/web` run:
 ```bash
 npm i
 ```
 
-Install Playwright (required by the Web Connector)
+#### Docker containers for external software
+You will need Docker installed to run these containers.
 
-> Note: If you have just done the pip install, open a new terminal and source the python virtual-env again.
-This will update the path to include playwright
-
-Then install Playwright by running:
+First navigate to `danswer/deployment/docker_compose`, then start up Postgres/Vespa/Redis with:
 ```bash
-playwright install
+docker compose -f docker-compose.dev.yml -p danswer-stack up -d index relational_db cache
 ```
+(index refers to Vespa, relational_db refers to Postgres, and cache refers to Redis)
 
 
-#### Dependent Docker Containers
-First navigate to `danswer/deployment/docker_compose`, then start up Vespa and Postgres with:
-```bash
-docker compose -f docker-compose.dev.yml -p danswer-stack up -d index relational_db
-```
-(index refers to Vespa and relational_db refers to Postgres)
-
-#### Running Danswer
+#### Running Danswer locally
 To start the frontend, navigate to `danswer/web` and run:
 ```bash
 npm run dev
@@ -127,11 +135,10 @@ Navigate to `danswer/backend` and run:
 ```bash
 uvicorn model_server.main:app --reload --port 9000
 ```
+
 _For Windows (for compatibility with both PowerShell and Command Prompt):_
 ```bash
-powershell -Command "
-    uvicorn model_server.main:app --reload --port 9000
-"
+powershell -Command "uvicorn model_server.main:app --reload --port 9000"
 ```
 
 The first time running Danswer, you will need to run the DB migrations for Postgres.
@@ -154,6 +161,7 @@ To run the backend API server, navigate back to `danswer/backend` and run:
 ```bash
 AUTH_TYPE=disabled uvicorn danswer.main:app --reload --port 8080
 ```
+
 _For Windows (for compatibility with both PowerShell and Command Prompt):_
 ```bash
 powershell -Command "
@@ -162,20 +170,58 @@ powershell -Command "
 "
 ```
 
-Note: if you need finer logging, add the additional environment variable `LOG_LEVEL=DEBUG` to the relevant services.
+> **Note:**
+> If you need finer logging, add the additional environment variable `LOG_LEVEL=DEBUG` to the relevant services.
+
+#### Wrapping up
+
+You should now have 4 servers running:
+
+- Web server
+- Backend API
+- Model server
+- Background jobs
+
+Now, visit `http://localhost:3000` in your browser. You should see the Danswer onboarding wizard where you can connect your external LLM provider to Danswer.
+
+You've successfully set up a local Danswer instance! 🏁
+
+#### Running the Danswer application in a container
+
+You can run the full Danswer application stack from pre-built images including all external software dependencies.
+
+Navigate to `danswer/deployment/docker_compose` and run:
+
+```bash
+docker compose -f docker-compose.dev.yml -p danswer-stack up -d
+```
+
+After Docker pulls and starts these containers, navigate to `http://localhost:3000` to use Danswer.
+
+If you want to make changes to Danswer and run those changes in Docker, you can also build a local version of the Danswer container images that incorporates your changes like so:
+
+```bash
+docker compose -f docker-compose.dev.yml -p danswer-stack up -d --build
+```
 
 ### Formatting and Linting
 #### Backend
 For the backend, you'll need to setup pre-commit hooks (black / reorder-python-imports).
 First, install pre-commit (if you don't have it already) following the instructions
 [here](https://pre-commit.com/#installation).
+
+With the virtual environment active, install the pre-commit library with:
+```bash
+pip install pre-commit
+```
+
 Then, from the `danswer/backend` directory, run:
 ```bash
 pre-commit install
 ```
 
 Additionally, we use `mypy` for static type checking.
-Danswer is fully type-annotated, and we would like to keep it that way! 
+Danswer is fully type-annotated, and we want to keep it that way! 
 To run the mypy checks manually, run `python -m mypy .` from the `danswer/backend` directory.
 
 
@@ -186,6 +232,7 @@ Please double check that prettier passes before creating a pull request.
 
 
 ### Release Process
-Danswer follows the semver versioning standard.
+Danswer loosely follows the SemVer versioning standard.
+Major changes are released with a "minor" version bump. Currently we use patch release versions to indicate small feature changes.
 A set of Docker containers will be pushed automatically to DockerHub with every tag.
 You can see the containers [here](https://hub.docker.com/search?q=danswer%2F).

CONTRIBUTING_MACOS.md

@@ -0,0 +1,31 @@
+## Some additional notes for Mac Users
+The base instructions to set up the development environment are located in [CONTRIBUTING.md](https://github.com/danswer-ai/danswer/blob/main/CONTRIBUTING.md).
+
+### Setting up Python
+Ensure [Homebrew](https://brew.sh/) is already set up.
+
+Then install python 3.11.
+```bash
+brew install python@3.11
+```
+
+Add python 3.11 to your path: add the following line to ~/.zshrc
+```
+export PATH="$(brew --prefix)/opt/python@3.11/libexec/bin:$PATH"
+```
+
+> **Note:**
+> You will need to open a new terminal for the path change above to take effect.
+
+
+### Setting up Docker
+On macOS, you will need to install [Docker Desktop](https://www.docker.com/products/docker-desktop/) and 
+ensure it is running before continuing with the docker commands.
+
+
+### Formatting and Linting
+MacOS will likely require you to remove some quarantine attributes on some of the hooks for them to execute properly.
+After installing pre-commit, run the following command:
+```bash
+sudo xattr -r -d com.apple.quarantine ~/.cache/pre-commit
+```

backend/Dockerfile

@@ -9,7 +9,8 @@ founders@danswer.ai for more information. Please visit https://github.com/danswe
 
 # Default DANSWER_VERSION, typically overriden during builds by GitHub Actions.
 ARG DANSWER_VERSION=0.3-dev
-ENV DANSWER_VERSION=${DANSWER_VERSION}
+ENV DANSWER_VERSION=${DANSWER_VERSION} \
+    DANSWER_RUNNING_IN_DOCKER="true"
 
 RUN echo "DANSWER_VERSION: ${DANSWER_VERSION}"
 # Install system dependencies
@@ -40,6 +41,8 @@ RUN apt-get update && \
 COPY ./requirements/default.txt /tmp/requirements.txt
 COPY ./requirements/ee.txt /tmp/ee-requirements.txt
 RUN pip install --no-cache-dir --upgrade \
+        --retries 5 \
+        --timeout 30 \
         -r /tmp/requirements.txt \
         -r /tmp/ee-requirements.txt && \
     pip uninstall -y py && \
@@ -75,8 +78,8 @@ Tokenizer.from_pretrained('nomic-ai/nomic-embed-text-v1')"
 # Pre-downloading NLTK for setups with limited egress
 RUN python -c "import nltk; \
 nltk.download('stopwords', quiet=True); \
-nltk.download('wordnet', quiet=True); \
 nltk.download('punkt', quiet=True);"
+# nltk.download('wordnet', quiet=True); introduce this back if lemmatization is needed
 
 # Set up application files
 WORKDIR /app

backend/Dockerfile.model_server

@@ -8,11 +8,17 @@ visit https://github.com/danswer-ai/danswer."
 
 # Default DANSWER_VERSION, typically overriden during builds by GitHub Actions.
 ARG DANSWER_VERSION=0.3-dev
-ENV DANSWER_VERSION=${DANSWER_VERSION}
+ENV DANSWER_VERSION=${DANSWER_VERSION} \
+    DANSWER_RUNNING_IN_DOCKER="true"
+
+
 RUN echo "DANSWER_VERSION: ${DANSWER_VERSION}"
 
 COPY ./requirements/model_server.txt /tmp/requirements.txt
-RUN pip install --no-cache-dir --upgrade -r /tmp/requirements.txt
+RUN pip install --no-cache-dir --upgrade \
+        --retries 5 \
+        --timeout 30 \
+        -r /tmp/requirements.txt
 
 RUN apt-get remove -y --allow-remove-essential perl-base && \
     apt-get autoremove -y
@@ -22,14 +28,18 @@ RUN apt-get remove -y --allow-remove-essential perl-base && \
 # Download model weights
 # Run Nomic to pull in the custom architecture and have it cached locally
 RUN python -c "from transformers import AutoTokenizer; \
-AutoTokenizer.from_pretrained('distilbert-base-uncased', cache_folder='/root/.cache/temp_huggingface/hub/'); \
-AutoTokenizer.from_pretrained('mixedbread-ai/mxbai-rerank-xsmall-v1', cache_folder='/root/.cache/temp_huggingface/hub/'); \
+AutoTokenizer.from_pretrained('distilbert-base-uncased'); \
+AutoTokenizer.from_pretrained('mixedbread-ai/mxbai-rerank-xsmall-v1'); \
 from huggingface_hub import snapshot_download; \
-snapshot_download(repo_id='danswer/hybrid-intent-token-classifier', revision='v1.0.3', cache_dir='/root/.cache/temp_huggingface/hub/'); \
-snapshot_download('nomic-ai/nomic-embed-text-v1', cache_dir='/root/.cache/temp_huggingface/hub/'); \
-snapshot_download('mixedbread-ai/mxbai-rerank-xsmall-v1', cache_dir='/root/.cache/temp_huggingface/hub/'); \
+snapshot_download(repo_id='danswer/hybrid-intent-token-classifier', revision='v1.0.3'); \
+snapshot_download('nomic-ai/nomic-embed-text-v1'); \
+snapshot_download('mixedbread-ai/mxbai-rerank-xsmall-v1'); \
 from sentence_transformers import SentenceTransformer; \
-SentenceTransformer(model_name_or_path='nomic-ai/nomic-embed-text-v1', trust_remote_code=True, cache_folder='/root/.cache/temp_huggingface/hub/');"
+SentenceTransformer(model_name_or_path='nomic-ai/nomic-embed-text-v1', trust_remote_code=True);"
+
+# In case the user has volumes mounted to /root/.cache/huggingface that they've downloaded while
+# running Danswer, don't overwrite it with the built in cache folder
+RUN mv /root/.cache/huggingface /root/.cache/temp_huggingface
 
 WORKDIR /app
 

backend/alembic/env.py

@@ -8,6 +8,7 @@ from sqlalchemy import pool
 from sqlalchemy.engine import Connection
 from sqlalchemy.ext.asyncio import create_async_engine
 from celery.backends.database.session import ResultModelBase  # type: ignore
+from sqlalchemy.schema import SchemaItem
 
 # this is the Alembic Config object, which provides
 # access to the values within the .ini file in use.
@@ -15,7 +16,9 @@ config = context.config
 
 # Interpret the config file for Python logging.
 # This line sets up loggers basically.
-if config.config_file_name is not None:
+if config.config_file_name is not None and config.attributes.get(
+    "configure_logger", True
+):
     fileConfig(config.config_file_name)
 
 # add your model's MetaData object here
@@ -29,6 +32,20 @@ target_metadata = [Base.metadata, ResultModelBase.metadata]
 # my_important_option = config.get_main_option("my_important_option")
 # ... etc.
 
+EXCLUDE_TABLES = {"kombu_queue", "kombu_message"}
+
+
+def include_object(
+    object: SchemaItem,
+    name: str,
+    type_: str,
+    reflected: bool,
+    compare_to: SchemaItem | None,
+) -> bool:
+    if type_ == "table" and name in EXCLUDE_TABLES:
+        return False
+    return True
+
 
 def run_migrations_offline() -> None:
     """Run migrations in 'offline' mode.
@@ -55,7 +72,11 @@ def run_migrations_offline() -> None:
 
 
 def do_run_migrations(connection: Connection) -> None:
-    context.configure(connection=connection, target_metadata=target_metadata)  # type: ignore
+    context.configure(
+        connection=connection,
+        target_metadata=target_metadata,  # type: ignore
+        include_object=include_object,
+    )  # type: ignore
 
     with context.begin_transaction():
         context.run_migrations()

backend/alembic/versions/0ebb1d516877_add_ccpair_deletion_failure_message.py

@@ -0,0 +1,27 @@
+"""add ccpair deletion failure message
+
+Revision ID: 0ebb1d516877
+Revises: 52a219fb5233
+Create Date: 2024-09-10 15:03:48.233926
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = "0ebb1d516877"
+down_revision = "52a219fb5233"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "connector_credential_pair",
+        sa.Column("deletion_failure_message", sa.String(), nullable=True),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("connector_credential_pair", "deletion_failure_message")

backend/alembic/versions/1b8206b29c5d_add_user_delete_cascades.py

@@ -0,0 +1,102 @@
+"""add_user_delete_cascades
+
+Revision ID: 1b8206b29c5d
+Revises: 35e6853a51d5
+Create Date: 2024-09-18 11:48:59.418726
+
+"""
+from alembic import op
+
+
+# revision identifiers, used by Alembic.
+revision = "1b8206b29c5d"
+down_revision = "35e6853a51d5"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.drop_constraint("credential_user_id_fkey", "credential", type_="foreignkey")
+    op.create_foreign_key(
+        "credential_user_id_fkey",
+        "credential",
+        "user",
+        ["user_id"],
+        ["id"],
+        ondelete="CASCADE",
+    )
+
+    op.drop_constraint("chat_session_user_id_fkey", "chat_session", type_="foreignkey")
+    op.create_foreign_key(
+        "chat_session_user_id_fkey",
+        "chat_session",
+        "user",
+        ["user_id"],
+        ["id"],
+        ondelete="CASCADE",
+    )
+
+    op.drop_constraint("chat_folder_user_id_fkey", "chat_folder", type_="foreignkey")
+    op.create_foreign_key(
+        "chat_folder_user_id_fkey",
+        "chat_folder",
+        "user",
+        ["user_id"],
+        ["id"],
+        ondelete="CASCADE",
+    )
+
+    op.drop_constraint("prompt_user_id_fkey", "prompt", type_="foreignkey")
+    op.create_foreign_key(
+        "prompt_user_id_fkey", "prompt", "user", ["user_id"], ["id"], ondelete="CASCADE"
+    )
+
+    op.drop_constraint("notification_user_id_fkey", "notification", type_="foreignkey")
+    op.create_foreign_key(
+        "notification_user_id_fkey",
+        "notification",
+        "user",
+        ["user_id"],
+        ["id"],
+        ondelete="CASCADE",
+    )
+
+    op.drop_constraint("inputprompt_user_id_fkey", "inputprompt", type_="foreignkey")
+    op.create_foreign_key(
+        "inputprompt_user_id_fkey",
+        "inputprompt",
+        "user",
+        ["user_id"],
+        ["id"],
+        ondelete="CASCADE",
+    )
+
+
+def downgrade() -> None:
+    op.drop_constraint("credential_user_id_fkey", "credential", type_="foreignkey")
+    op.create_foreign_key(
+        "credential_user_id_fkey", "credential", "user", ["user_id"], ["id"]
+    )
+
+    op.drop_constraint("chat_session_user_id_fkey", "chat_session", type_="foreignkey")
+    op.create_foreign_key(
+        "chat_session_user_id_fkey", "chat_session", "user", ["user_id"], ["id"]
+    )
+
+    op.drop_constraint("chat_folder_user_id_fkey", "chat_folder", type_="foreignkey")
+    op.create_foreign_key(
+        "chat_folder_user_id_fkey", "chat_folder", "user", ["user_id"], ["id"]
+    )
+
+    op.drop_constraint("prompt_user_id_fkey", "prompt", type_="foreignkey")
+    op.create_foreign_key("prompt_user_id_fkey", "prompt", "user", ["user_id"], ["id"])
+
+    op.drop_constraint("notification_user_id_fkey", "notification", type_="foreignkey")
+    op.create_foreign_key(
+        "notification_user_id_fkey", "notification", "user", ["user_id"], ["id"]
+    )
+
+    op.drop_constraint("inputprompt_user_id_fkey", "inputprompt", type_="foreignkey")
+    op.create_foreign_key(
+        "inputprompt_user_id_fkey", "inputprompt", "user", ["user_id"], ["id"]
+    )

backend/alembic/versions/1f60f60c3401_embedding_model_search_settings.py

@@ -0,0 +1,135 @@
+"""embedding model -> search settings
+
+Revision ID: 1f60f60c3401
+Revises: f17bf3b0d9f1
+Create Date: 2024-08-25 12:39:51.731632
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+from danswer.configs.chat_configs import NUM_POSTPROCESSED_RESULTS
+
+# revision identifiers, used by Alembic.
+revision = "1f60f60c3401"
+down_revision = "f17bf3b0d9f1"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    op.drop_constraint(
+        "index_attempt__embedding_model_fk", "index_attempt", type_="foreignkey"
+    )
+    # Rename the table
+    op.rename_table("embedding_model", "search_settings")
+
+    # Add new columns
+    op.add_column(
+        "search_settings",
+        sa.Column(
+            "multipass_indexing", sa.Boolean(), nullable=False, server_default="false"
+        ),
+    )
+    op.add_column(
+        "search_settings",
+        sa.Column(
+            "multilingual_expansion",
+            postgresql.ARRAY(sa.String()),
+            nullable=False,
+            server_default="{}",
+        ),
+    )
+    op.add_column(
+        "search_settings",
+        sa.Column(
+            "disable_rerank_for_streaming",
+            sa.Boolean(),
+            nullable=False,
+            server_default="false",
+        ),
+    )
+    op.add_column(
+        "search_settings", sa.Column("rerank_model_name", sa.String(), nullable=True)
+    )
+    op.add_column(
+        "search_settings", sa.Column("rerank_provider_type", sa.String(), nullable=True)
+    )
+    op.add_column(
+        "search_settings", sa.Column("rerank_api_key", sa.String(), nullable=True)
+    )
+    op.add_column(
+        "search_settings",
+        sa.Column(
+            "num_rerank",
+            sa.Integer(),
+            nullable=False,
+            server_default=str(NUM_POSTPROCESSED_RESULTS),
+        ),
+    )
+
+    # Add the new column as nullable initially
+    op.add_column(
+        "index_attempt", sa.Column("search_settings_id", sa.Integer(), nullable=True)
+    )
+
+    # Populate the new column with data from the existing embedding_model_id
+    op.execute("UPDATE index_attempt SET search_settings_id = embedding_model_id")
+
+    # Create the foreign key constraint
+    op.create_foreign_key(
+        "fk_index_attempt_search_settings",
+        "index_attempt",
+        "search_settings",
+        ["search_settings_id"],
+        ["id"],
+    )
+
+    # Make the new column non-nullable
+    op.alter_column("index_attempt", "search_settings_id", nullable=False)
+
+    # Drop the old embedding_model_id column
+    op.drop_column("index_attempt", "embedding_model_id")
+
+
+def downgrade() -> None:
+    # Add back the embedding_model_id column
+    op.add_column(
+        "index_attempt", sa.Column("embedding_model_id", sa.Integer(), nullable=True)
+    )
+
+    # Populate the old column with data from search_settings_id
+    op.execute("UPDATE index_attempt SET embedding_model_id = search_settings_id")
+
+    # Make the old column non-nullable
+    op.alter_column("index_attempt", "embedding_model_id", nullable=False)
+
+    # Drop the foreign key constraint
+    op.drop_constraint(
+        "fk_index_attempt_search_settings", "index_attempt", type_="foreignkey"
+    )
+
+    # Drop the new search_settings_id column
+    op.drop_column("index_attempt", "search_settings_id")
+
+    # Rename the table back
+    op.rename_table("search_settings", "embedding_model")
+
+    # Remove added columns
+    op.drop_column("embedding_model", "num_rerank")
+    op.drop_column("embedding_model", "rerank_api_key")
+    op.drop_column("embedding_model", "rerank_provider_type")
+    op.drop_column("embedding_model", "rerank_model_name")
+    op.drop_column("embedding_model", "disable_rerank_for_streaming")
+    op.drop_column("embedding_model", "multilingual_expansion")
+    op.drop_column("embedding_model", "multipass_indexing")
+
+    op.create_foreign_key(
+        "index_attempt__embedding_model_fk",
+        "index_attempt",
+        "embedding_model",
+        ["embedding_model_id"],
+        ["id"],
+    )

backend/alembic/versions/2d2304e27d8c_add_above_below_to_persona.py

@@ -0,0 +1,32 @@
+"""Add Above Below to Persona
+
+Revision ID: 2d2304e27d8c
+Revises: 4b08d97e175a
+Create Date: 2024-08-21 19:15:15.762948
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "2d2304e27d8c"
+down_revision = "4b08d97e175a"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    op.add_column("persona", sa.Column("chunks_above", sa.Integer(), nullable=True))
+    op.add_column("persona", sa.Column("chunks_below", sa.Integer(), nullable=True))
+
+    op.execute(
+        "UPDATE persona SET chunks_above = 1, chunks_below = 1 WHERE chunks_above IS NULL AND chunks_below IS NULL"
+    )
+
+    op.alter_column("persona", "chunks_above", nullable=False)
+    op.alter_column("persona", "chunks_below", nullable=False)
+
+
+def downgrade() -> None:
+    op.drop_column("persona", "chunks_below")
+    op.drop_column("persona", "chunks_above")

backend/alembic/versions/351faebd379d_add_curator_fields.py

@@ -0,0 +1,90 @@
+"""Add curator fields
+
+Revision ID: 351faebd379d
+Revises: ee3f4b47fad5
+Create Date: 2024-08-15 22:37:08.397052
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "351faebd379d"
+down_revision = "ee3f4b47fad5"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    # Add is_curator column to User__UserGroup table
+    op.add_column(
+        "user__user_group",
+        sa.Column("is_curator", sa.Boolean(), nullable=False, server_default="false"),
+    )
+
+    # Use batch mode to modify the enum type
+    with op.batch_alter_table("user", schema=None) as batch_op:
+        batch_op.alter_column(  # type: ignore[attr-defined]
+            "role",
+            type_=sa.Enum(
+                "BASIC",
+                "ADMIN",
+                "CURATOR",
+                "GLOBAL_CURATOR",
+                name="userrole",
+                native_enum=False,
+            ),
+            existing_type=sa.Enum("BASIC", "ADMIN", name="userrole", native_enum=False),
+            existing_nullable=False,
+        )
+    # Create the association table
+    op.create_table(
+        "credential__user_group",
+        sa.Column("credential_id", sa.Integer(), nullable=False),
+        sa.Column("user_group_id", sa.Integer(), nullable=False),
+        sa.ForeignKeyConstraint(
+            ["credential_id"],
+            ["credential.id"],
+        ),
+        sa.ForeignKeyConstraint(
+            ["user_group_id"],
+            ["user_group.id"],
+        ),
+        sa.PrimaryKeyConstraint("credential_id", "user_group_id"),
+    )
+    op.add_column(
+        "credential",
+        sa.Column(
+            "curator_public", sa.Boolean(), nullable=False, server_default="false"
+        ),
+    )
+
+
+def downgrade() -> None:
+    # Update existing records to ensure they fit within the BASIC/ADMIN roles
+    op.execute(
+        "UPDATE \"user\" SET role = 'ADMIN' WHERE role IN ('CURATOR', 'GLOBAL_CURATOR')"
+    )
+
+    # Remove is_curator column from User__UserGroup table
+    op.drop_column("user__user_group", "is_curator")
+
+    with op.batch_alter_table("user", schema=None) as batch_op:
+        batch_op.alter_column(  # type: ignore[attr-defined]
+            "role",
+            type_=sa.Enum(
+                "BASIC", "ADMIN", name="userrole", native_enum=False, length=20
+            ),
+            existing_type=sa.Enum(
+                "BASIC",
+                "ADMIN",
+                "CURATOR",
+                "GLOBAL_CURATOR",
+                name="userrole",
+                native_enum=False,
+            ),
+            existing_nullable=False,
+        )
+    # Drop the association table
+    op.drop_table("credential__user_group")
+    op.drop_column("credential", "curator_public")

backend/alembic/versions/35e6853a51d5_server_default_chosen_assistants.py

@@ -0,0 +1,64 @@
+"""server default chosen assistants
+
+Revision ID: 35e6853a51d5
+Revises: c99d76fcd298
+Create Date: 2024-09-13 13:20:32.885317
+
+"""
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision = "35e6853a51d5"
+down_revision = "c99d76fcd298"
+branch_labels = None
+depends_on = None
+
+DEFAULT_ASSISTANTS = [-2, -1, 0]
+
+
+def upgrade() -> None:
+    # Step 1: Update any NULL values to the default value
+    # This upgrades existing users without ordered assistant
+    # to have default assistants set to visible assistants which are
+    # accessible by them.
+    op.execute(
+        """
+        UPDATE "user" u
+        SET chosen_assistants = (
+            SELECT jsonb_agg(
+                p.id ORDER BY
+                    COALESCE(p.display_priority, 2147483647) ASC,
+                    p.id ASC
+            )
+            FROM persona p
+            LEFT JOIN persona__user pu ON p.id = pu.persona_id AND pu.user_id = u.id
+            WHERE p.is_visible = true
+            AND (p.is_public = true OR pu.user_id IS NOT NULL)
+        )
+        WHERE chosen_assistants IS NULL
+        OR chosen_assistants = 'null'
+        OR jsonb_typeof(chosen_assistants) = 'null'
+        OR (jsonb_typeof(chosen_assistants) = 'string' AND chosen_assistants = '"null"')
+    """
+    )
+
+    # Step 2: Alter the column to make it non-nullable
+    op.alter_column(
+        "user",
+        "chosen_assistants",
+        type_=postgresql.JSONB(astext_type=sa.Text()),
+        nullable=False,
+        server_default=sa.text(f"'{DEFAULT_ASSISTANTS}'::jsonb"),
+    )
+
+
+def downgrade() -> None:
+    op.alter_column(
+        "user",
+        "chosen_assistants",
+        type_=postgresql.JSONB(astext_type=sa.Text()),
+        nullable=True,
+        server_default=None,
+    )

backend/alembic/versions/4a951134c801_moved_status_to_connector_credential_.py

@@ -11,8 +11,8 @@ import sqlalchemy as sa
 # revision identifiers, used by Alembic.
 revision = "4a951134c801"
 down_revision = "7477a5f5d728"
-branch_labels = None
-depends_on = None
+branch_labels: None = None
+depends_on: None = None
 
 
 def upgrade() -> None:

backend/alembic/versions/4b08d97e175a_change_default_prune_freq.py

@@ -0,0 +1,34 @@
+"""change default prune_freq
+
+Revision ID: 4b08d97e175a
+Revises: d9ec13955951
+Create Date: 2024-08-20 15:28:52.993827
+
+"""
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision = "4b08d97e175a"
+down_revision = "d9ec13955951"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    op.execute(
+        """
+        UPDATE connector
+        SET prune_freq = 2592000
+        WHERE prune_freq = 86400
+        """
+    )
+
+
+def downgrade() -> None:
+    op.execute(
+        """
+        UPDATE connector
+        SET prune_freq = 86400
+        WHERE prune_freq = 2592000
+        """
+    )

backend/alembic/versions/52a219fb5233_add_last_synced_and_last_modified_to_document_table.py

@@ -0,0 +1,66 @@
+"""Add last synced and last modified to document table
+
+Revision ID: 52a219fb5233
+Revises: f7e58d357687
+Create Date: 2024-08-28 17:40:46.077470
+
+"""
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.sql import func
+
+# revision identifiers, used by Alembic.
+revision = "52a219fb5233"
+down_revision = "f7e58d357687"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # last modified represents the last time anything needing syncing to vespa changed
+    # including row metadata and the document itself. This obviously does not include
+    # the last_synced column.
+    op.add_column(
+        "document",
+        sa.Column(
+            "last_modified",
+            sa.DateTime(timezone=True),
+            nullable=False,
+            server_default=func.now(),
+        ),
+    )
+
+    # last synced represents the last time this document was synced to Vespa
+    op.add_column(
+        "document",
+        sa.Column("last_synced", sa.DateTime(timezone=True), nullable=True),
+    )
+
+    # Set last_synced to the same value as last_modified for existing rows
+    op.execute(
+        """
+        UPDATE document
+        SET last_synced = last_modified
+        """
+    )
+
+    op.create_index(
+        op.f("ix_document_last_modified"),
+        "document",
+        ["last_modified"],
+        unique=False,
+    )
+
+    op.create_index(
+        op.f("ix_document_last_synced"),
+        "document",
+        ["last_synced"],
+        unique=False,
+    )
+
+
+def downgrade() -> None:
+    op.drop_index(op.f("ix_document_last_synced"), table_name="document")
+    op.drop_index(op.f("ix_document_last_modified"), table_name="document")
+    op.drop_column("document", "last_synced")
+    op.drop_column("document", "last_modified")

backend/alembic/versions/55546a7967ee_assistant_rework.py

@@ -0,0 +1,79 @@
+"""assistant_rework
+
+Revision ID: 55546a7967ee
+Revises: 61ff3651add4
+Create Date: 2024-09-18 17:00:23.755399
+
+"""
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+
+# revision identifiers, used by Alembic.
+revision = "55546a7967ee"
+down_revision = "61ff3651add4"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # Reworking persona and user tables for new assistant features
+    # keep track of user's chosen assistants separate from their `ordering`
+    op.add_column("persona", sa.Column("builtin_persona", sa.Boolean(), nullable=True))
+    op.execute("UPDATE persona SET builtin_persona = default_persona")
+    op.alter_column("persona", "builtin_persona", nullable=False)
+    op.drop_index("_default_persona_name_idx", table_name="persona")
+    op.create_index(
+        "_builtin_persona_name_idx",
+        "persona",
+        ["name"],
+        unique=True,
+        postgresql_where=sa.text("builtin_persona = true"),
+    )
+
+    op.add_column(
+        "user", sa.Column("visible_assistants", postgresql.JSONB(), nullable=True)
+    )
+    op.add_column(
+        "user", sa.Column("hidden_assistants", postgresql.JSONB(), nullable=True)
+    )
+    op.execute(
+        "UPDATE \"user\" SET visible_assistants = '[]'::jsonb, hidden_assistants = '[]'::jsonb"
+    )
+    op.alter_column(
+        "user",
+        "visible_assistants",
+        nullable=False,
+        server_default=sa.text("'[]'::jsonb"),
+    )
+    op.alter_column(
+        "user",
+        "hidden_assistants",
+        nullable=False,
+        server_default=sa.text("'[]'::jsonb"),
+    )
+    op.drop_column("persona", "default_persona")
+    op.add_column(
+        "persona", sa.Column("is_default_persona", sa.Boolean(), nullable=True)
+    )
+
+
+def downgrade() -> None:
+    # Reverting changes made in upgrade
+    op.drop_column("user", "hidden_assistants")
+    op.drop_column("user", "visible_assistants")
+    op.drop_index("_builtin_persona_name_idx", table_name="persona")
+
+    op.drop_column("persona", "is_default_persona")
+    op.add_column("persona", sa.Column("default_persona", sa.Boolean(), nullable=True))
+    op.execute("UPDATE persona SET default_persona = builtin_persona")
+    op.alter_column("persona", "default_persona", nullable=False)
+    op.drop_column("persona", "builtin_persona")
+    op.create_index(
+        "_default_persona_name_idx",
+        "persona",
+        ["name"],
+        unique=True,
+        postgresql_where=sa.text("default_persona = true"),
+    )

backend/alembic/versions/5c7fdadae813_match_any_keywords_flag_for_standard_.py

@@ -0,0 +1,35 @@
+"""match_any_keywords flag for standard answers
+
+Revision ID: 5c7fdadae813
+Revises: efb35676026c
+Create Date: 2024-09-13 18:52:59.256478
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "5c7fdadae813"
+down_revision = "efb35676026c"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.add_column(
+        "standard_answer",
+        sa.Column(
+            "match_any_keywords",
+            sa.Boolean(),
+            nullable=False,
+            server_default=sa.false(),
+        ),
+    )
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_column("standard_answer", "match_any_keywords")
+    # ### end Alembic commands ###

backend/alembic/versions/61ff3651add4_add_permission_syncing.py

@@ -0,0 +1,162 @@
+"""Add Permission Syncing
+
+Revision ID: 61ff3651add4
+Revises: 1b8206b29c5d
+Create Date: 2024-09-05 13:57:11.770413
+
+"""
+import fastapi_users_db_sqlalchemy
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision = "61ff3651add4"
+down_revision = "1b8206b29c5d"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # Admin user who set up connectors will lose access to the docs temporarily
+    # only way currently to give back access is to rerun from beginning
+    op.add_column(
+        "connector_credential_pair",
+        sa.Column(
+            "access_type",
+            sa.String(),
+            nullable=True,
+        ),
+    )
+    op.execute(
+        "UPDATE connector_credential_pair SET access_type = 'PUBLIC' WHERE is_public = true"
+    )
+    op.execute(
+        "UPDATE connector_credential_pair SET access_type = 'PRIVATE' WHERE is_public = false"
+    )
+    op.alter_column("connector_credential_pair", "access_type", nullable=False)
+
+    op.add_column(
+        "connector_credential_pair",
+        sa.Column(
+            "auto_sync_options",
+            postgresql.JSONB(astext_type=sa.Text()),
+            nullable=True,
+        ),
+    )
+    op.add_column(
+        "connector_credential_pair",
+        sa.Column("last_time_perm_sync", sa.DateTime(timezone=True), nullable=True),
+    )
+    op.drop_column("connector_credential_pair", "is_public")
+
+    op.add_column(
+        "document",
+        sa.Column("external_user_emails", postgresql.ARRAY(sa.String()), nullable=True),
+    )
+    op.add_column(
+        "document",
+        sa.Column(
+            "external_user_group_ids", postgresql.ARRAY(sa.String()), nullable=True
+        ),
+    )
+    op.add_column(
+        "document",
+        sa.Column("is_public", sa.Boolean(), nullable=True),
+    )
+
+    op.create_table(
+        "user__external_user_group_id",
+        sa.Column(
+            "user_id", fastapi_users_db_sqlalchemy.generics.GUID(), nullable=False
+        ),
+        sa.Column("external_user_group_id", sa.String(), nullable=False),
+        sa.Column("cc_pair_id", sa.Integer(), nullable=False),
+        sa.PrimaryKeyConstraint("user_id"),
+    )
+
+    op.drop_column("external_permission", "user_id")
+    op.drop_column("email_to_external_user_cache", "user_id")
+    op.drop_table("permission_sync_run")
+    op.drop_table("external_permission")
+    op.drop_table("email_to_external_user_cache")
+
+
+def downgrade() -> None:
+    op.add_column(
+        "connector_credential_pair",
+        sa.Column("is_public", sa.BOOLEAN(), nullable=True),
+    )
+    op.execute(
+        "UPDATE connector_credential_pair SET is_public = (access_type = 'PUBLIC')"
+    )
+    op.alter_column("connector_credential_pair", "is_public", nullable=False)
+
+    op.drop_column("connector_credential_pair", "auto_sync_options")
+    op.drop_column("connector_credential_pair", "access_type")
+    op.drop_column("connector_credential_pair", "last_time_perm_sync")
+    op.drop_column("document", "external_user_emails")
+    op.drop_column("document", "external_user_group_ids")
+    op.drop_column("document", "is_public")
+
+    op.drop_table("user__external_user_group_id")
+
+    # Drop the enum type at the end of the downgrade
+    op.create_table(
+        "permission_sync_run",
+        sa.Column("id", sa.Integer(), nullable=False),
+        sa.Column(
+            "source_type",
+            sa.String(),
+            nullable=False,
+        ),
+        sa.Column("update_type", sa.String(), nullable=False),
+        sa.Column("cc_pair_id", sa.Integer(), nullable=True),
+        sa.Column(
+            "status",
+            sa.String(),
+            nullable=False,
+        ),
+        sa.Column("error_msg", sa.Text(), nullable=True),
+        sa.Column(
+            "updated_at",
+            sa.DateTime(timezone=True),
+            server_default=sa.text("now()"),
+            nullable=False,
+        ),
+        sa.ForeignKeyConstraint(
+            ["cc_pair_id"],
+            ["connector_credential_pair.id"],
+        ),
+        sa.PrimaryKeyConstraint("id"),
+    )
+    op.create_table(
+        "external_permission",
+        sa.Column("id", sa.Integer(), nullable=False),
+        sa.Column("user_id", sa.UUID(), nullable=True),
+        sa.Column("user_email", sa.String(), nullable=False),
+        sa.Column(
+            "source_type",
+            sa.String(),
+            nullable=False,
+        ),
+        sa.Column("external_permission_group", sa.String(), nullable=False),
+        sa.ForeignKeyConstraint(
+            ["user_id"],
+            ["user.id"],
+        ),
+        sa.PrimaryKeyConstraint("id"),
+    )
+    op.create_table(
+        "email_to_external_user_cache",
+        sa.Column("id", sa.Integer(), nullable=False),
+        sa.Column("external_user_id", sa.String(), nullable=False),
+        sa.Column("user_id", sa.UUID(), nullable=True),
+        sa.Column("user_email", sa.String(), nullable=False),
+        sa.ForeignKeyConstraint(
+            ["user_id"],
+            ["user.id"],
+        ),
+        sa.PrimaryKeyConstraint("id"),
+    )

backend/alembic/versions/776b3bbe9092_remove_remaining_enums.py

@@ -10,7 +10,7 @@ import sqlalchemy as sa
 
 from danswer.db.models import IndexModelStatus
 from danswer.search.enums import RecencyBiasSetting
-from danswer.search.models import SearchType
+from danswer.search.enums import SearchType
 
 # revision identifiers, used by Alembic.
 revision = "776b3bbe9092"

backend/alembic/versions/797089dfb4d2_persona_start_date.py

@@ -0,0 +1,27 @@
+"""persona_start_date
+
+Revision ID: 797089dfb4d2
+Revises: 55546a7967ee
+Create Date: 2024-09-11 14:51:49.785835
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = "797089dfb4d2"
+down_revision = "55546a7967ee"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "persona",
+        sa.Column("search_start_date", sa.DateTime(timezone=True), nullable=True),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("persona", "search_start_date")

backend/alembic/versions/8a87bd6ec550_associate_index_attempts_with_ccpair.py

@@ -35,18 +35,22 @@ def upgrade() -> None:
     op.execute(
         """
         UPDATE index_attempt ia
-        SET connector_credential_pair_id =
-            CASE
-                WHEN ia.credential_id IS NULL THEN
-                    (SELECT id FROM connector_credential_pair
-                     WHERE connector_id = ia.connector_id
-                     LIMIT 1)
-                ELSE
-                    (SELECT id FROM connector_credential_pair
-                     WHERE connector_id = ia.connector_id
-                     AND credential_id = ia.credential_id)
-            END
-        WHERE ia.connector_id IS NOT NULL
+        SET connector_credential_pair_id = (
+            SELECT id FROM connector_credential_pair ccp
+            WHERE
+                (ia.connector_id IS NULL OR ccp.connector_id = ia.connector_id)
+                AND (ia.credential_id IS NULL OR ccp.credential_id = ia.credential_id)
+            LIMIT 1
+        )
+        WHERE ia.connector_id IS NOT NULL OR ia.credential_id IS NOT NULL
+        """
+    )
+
+    # For good measure
+    op.execute(
+        """
+        DELETE FROM index_attempt
+        WHERE connector_credential_pair_id IS NULL
         """
     )
 

backend/alembic/versions/a3795dce87be_migration_confluence_to_be_explicit.py

@@ -0,0 +1,158 @@
+"""migration confluence to be explicit
+
+Revision ID: a3795dce87be
+Revises: 1f60f60c3401
+Create Date: 2024-09-01 13:52:12.006740
+
+"""
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+from sqlalchemy.sql import table, column
+
+revision = "a3795dce87be"
+down_revision = "1f60f60c3401"
+branch_labels: None = None
+depends_on: None = None
+
+
+def extract_confluence_keys_from_url(wiki_url: str) -> tuple[str, str, str, bool]:
+    from urllib.parse import urlparse
+
+    def _extract_confluence_keys_from_cloud_url(wiki_url: str) -> tuple[str, str, str]:
+        parsed_url = urlparse(wiki_url)
+        wiki_base = f"{parsed_url.scheme}://{parsed_url.netloc}{parsed_url.path.split('/spaces')[0]}"
+        path_parts = parsed_url.path.split("/")
+        space = path_parts[3]
+        page_id = path_parts[5] if len(path_parts) > 5 else ""
+        return wiki_base, space, page_id
+
+    def _extract_confluence_keys_from_datacenter_url(
+        wiki_url: str,
+    ) -> tuple[str, str, str]:
+        DISPLAY = "/display/"
+        PAGE = "/pages/"
+        parsed_url = urlparse(wiki_url)
+        wiki_base = f"{parsed_url.scheme}://{parsed_url.netloc}{parsed_url.path.split(DISPLAY)[0]}"
+        space = DISPLAY.join(parsed_url.path.split(DISPLAY)[1:]).split("/")[0]
+        page_id = ""
+        if (content := parsed_url.path.split(PAGE)) and len(content) > 1:
+            page_id = content[1]
+        return wiki_base, space, page_id
+
+    is_confluence_cloud = (
+        ".atlassian.net/wiki/spaces/" in wiki_url
+        or ".jira.com/wiki/spaces/" in wiki_url
+    )
+
+    if is_confluence_cloud:
+        wiki_base, space, page_id = _extract_confluence_keys_from_cloud_url(wiki_url)
+    else:
+        wiki_base, space, page_id = _extract_confluence_keys_from_datacenter_url(
+            wiki_url
+        )
+
+    return wiki_base, space, page_id, is_confluence_cloud
+
+
+def reconstruct_confluence_url(
+    wiki_base: str, space: str, page_id: str, is_cloud: bool
+) -> str:
+    if is_cloud:
+        url = f"{wiki_base}/spaces/{space}"
+        if page_id:
+            url += f"/pages/{page_id}"
+    else:
+        url = f"{wiki_base}/display/{space}"
+        if page_id:
+            url += f"/pages/{page_id}"
+    return url
+
+
+def upgrade() -> None:
+    connector = table(
+        "connector",
+        column("id", sa.Integer),
+        column("source", sa.String()),
+        column("input_type", sa.String()),
+        column("connector_specific_config", postgresql.JSONB),
+    )
+
+    # Fetch all Confluence connectors
+    connection = op.get_bind()
+    confluence_connectors = connection.execute(
+        sa.select(connector).where(
+            sa.and_(
+                connector.c.source == "CONFLUENCE", connector.c.input_type == "POLL"
+            )
+        )
+    ).fetchall()
+
+    for row in confluence_connectors:
+        config = row.connector_specific_config
+        wiki_page_url = config["wiki_page_url"]
+        wiki_base, space, page_id, is_cloud = extract_confluence_keys_from_url(
+            wiki_page_url
+        )
+
+        new_config = {
+            "wiki_base": wiki_base,
+            "space": space,
+            "page_id": page_id,
+            "is_cloud": is_cloud,
+        }
+
+        for key, value in config.items():
+            if key not in ["wiki_page_url"]:
+                new_config[key] = value
+
+        op.execute(
+            connector.update()
+            .where(connector.c.id == row.id)
+            .values(connector_specific_config=new_config)
+        )
+
+
+def downgrade() -> None:
+    connector = table(
+        "connector",
+        column("id", sa.Integer),
+        column("source", sa.String()),
+        column("input_type", sa.String()),
+        column("connector_specific_config", postgresql.JSONB),
+    )
+
+    confluence_connectors = (
+        op.get_bind()
+        .execute(
+            sa.select(connector).where(
+                connector.c.source == "CONFLUENCE", connector.c.input_type == "POLL"
+            )
+        )
+        .fetchall()
+    )
+
+    for row in confluence_connectors:
+        config = row.connector_specific_config
+        if all(key in config for key in ["wiki_base", "space", "is_cloud"]):
+            wiki_page_url = reconstruct_confluence_url(
+                config["wiki_base"],
+                config["space"],
+                config.get("page_id", ""),
+                config["is_cloud"],
+            )
+
+            new_config = {"wiki_page_url": wiki_page_url}
+            new_config.update(
+                {
+                    k: v
+                    for k, v in config.items()
+                    if k not in ["wiki_base", "space", "page_id", "is_cloud"]
+                }
+            )
+
+            op.execute(
+                connector.update()
+                .where(connector.c.id == row.id)
+                .values(connector_specific_config=new_config)
+            )

backend/alembic/versions/ba98eba0f66a_add_support_for_litellm_proxy_in_.py

@@ -0,0 +1,26 @@
+"""add support for litellm proxy in reranking
+
+Revision ID: ba98eba0f66a
+Revises: bceb1e139447
+Create Date: 2024-09-06 10:36:04.507332
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = "ba98eba0f66a"
+down_revision = "bceb1e139447"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "search_settings", sa.Column("rerank_api_url", sa.String(), nullable=True)
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("search_settings", "rerank_api_url")

backend/alembic/versions/bceb1e139447_add_base_url_to_cloudembeddingprovider.py

@@ -0,0 +1,26 @@
+"""Add base_url to CloudEmbeddingProvider
+
+Revision ID: bceb1e139447
+Revises: a3795dce87be
+Create Date: 2024-08-28 17:00:52.554580
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = "bceb1e139447"
+down_revision = "a3795dce87be"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "embedding_provider", sa.Column("api_url", sa.String(), nullable=True)
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("embedding_provider", "api_url")

backend/alembic/versions/bd2921608c3a_non_nullable_default_persona.py

@@ -0,0 +1,43 @@
+"""non nullable default persona
+
+Revision ID: bd2921608c3a
+Revises: 797089dfb4d2
+Create Date: 2024-09-20 10:28:37.992042
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = "bd2921608c3a"
+down_revision = "797089dfb4d2"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # Set existing NULL values to False
+    op.execute(
+        "UPDATE persona SET is_default_persona = FALSE WHERE is_default_persona IS NULL"
+    )
+
+    # Alter the column to be not nullable with a default value of False
+    op.alter_column(
+        "persona",
+        "is_default_persona",
+        existing_type=sa.Boolean(),
+        nullable=False,
+        server_default=sa.text("false"),
+    )
+
+
+def downgrade() -> None:
+    # Revert the changes
+    op.alter_column(
+        "persona",
+        "is_default_persona",
+        existing_type=sa.Boolean(),
+        nullable=True,
+        server_default=None,
+    )

backend/alembic/versions/c5b692fa265c_add_index_attempt_errors_table.py

@@ -0,0 +1,57 @@
+"""Add index_attempt_errors table
+
+Revision ID: c5b692fa265c
+Revises: 4a951134c801
+Create Date: 2024-08-08 14:06:39.581972
+
+"""
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision = "c5b692fa265c"
+down_revision = "4a951134c801"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "index_attempt_errors",
+        sa.Column("id", sa.Integer(), nullable=False),
+        sa.Column("index_attempt_id", sa.Integer(), nullable=True),
+        sa.Column("batch", sa.Integer(), nullable=True),
+        sa.Column(
+            "doc_summaries",
+            postgresql.JSONB(astext_type=sa.Text()),
+            nullable=False,
+        ),
+        sa.Column("error_msg", sa.Text(), nullable=True),
+        sa.Column("traceback", sa.Text(), nullable=True),
+        sa.Column(
+            "time_created",
+            sa.DateTime(timezone=True),
+            server_default=sa.text("now()"),
+            nullable=False,
+        ),
+        sa.ForeignKeyConstraint(
+            ["index_attempt_id"],
+            ["index_attempt.id"],
+        ),
+        sa.PrimaryKeyConstraint("id"),
+    )
+    op.create_index(
+        "index_attempt_id",
+        "index_attempt_errors",
+        ["time_created"],
+        unique=False,
+    )
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_index("index_attempt_id", table_name="index_attempt_errors")
+    op.drop_table("index_attempt_errors")
+    # ### end Alembic commands ###

backend/alembic/versions/c99d76fcd298_add_nullable_to_persona_id_in_chat_.py

@@ -0,0 +1,31 @@
+"""add nullable to persona id in Chat Session
+
+Revision ID: c99d76fcd298
+Revises: 5c7fdadae813
+Create Date: 2024-07-09 19:27:01.579697
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "c99d76fcd298"
+down_revision = "5c7fdadae813"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.alter_column(
+        "chat_session", "persona_id", existing_type=sa.INTEGER(), nullable=True
+    )
+
+
+def downgrade() -> None:
+    op.alter_column(
+        "chat_session",
+        "persona_id",
+        existing_type=sa.INTEGER(),
+        nullable=False,
+    )

backend/alembic/versions/d9ec13955951_remove__dim_suffix_from_model_name.py

@@ -0,0 +1,31 @@
+"""Remove _alt suffix from model_name
+
+Revision ID: d9ec13955951
+Revises: da4c21c69164
+Create Date: 2024-08-20 16:31:32.955686
+
+"""
+
+from alembic import op
+
+
+# revision identifiers, used by Alembic.
+revision = "d9ec13955951"
+down_revision = "da4c21c69164"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    op.execute(
+        """
+        UPDATE embedding_model
+        SET model_name = regexp_replace(model_name, '__danswer_alt_index$', '')
+        WHERE model_name LIKE '%__danswer_alt_index'
+    """
+    )
+
+
+def downgrade() -> None:
+    # We can't reliably add the __danswer_alt_index suffix back, so we'll leave this empty
+    pass

backend/alembic/versions/da4c21c69164_chosen_assistants_changed_to_jsonb.py

@@ -0,0 +1,65 @@
+"""chosen_assistants changed to jsonb
+
+Revision ID: da4c21c69164
+Revises: c5b692fa265c
+Create Date: 2024-08-18 19:06:47.291491
+
+"""
+import json
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision = "da4c21c69164"
+down_revision = "c5b692fa265c"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    conn = op.get_bind()
+    existing_ids_and_chosen_assistants = conn.execute(
+        sa.text("select id, chosen_assistants from public.user")
+    )
+    op.drop_column(
+        "user",
+        "chosen_assistants",
+    )
+    op.add_column(
+        "user",
+        sa.Column(
+            "chosen_assistants",
+            postgresql.JSONB(astext_type=sa.Text()),
+            nullable=True,
+        ),
+    )
+    for id, chosen_assistants in existing_ids_and_chosen_assistants:
+        conn.execute(
+            sa.text(
+                "update public.user set chosen_assistants = :chosen_assistants where id = :id"
+            ),
+            {"chosen_assistants": json.dumps(chosen_assistants), "id": id},
+        )
+
+
+def downgrade() -> None:
+    conn = op.get_bind()
+    existing_ids_and_chosen_assistants = conn.execute(
+        sa.text("select id, chosen_assistants from public.user")
+    )
+    op.drop_column(
+        "user",
+        "chosen_assistants",
+    )
+    op.add_column(
+        "user",
+        sa.Column("chosen_assistants", postgresql.ARRAY(sa.Integer()), nullable=True),
+    )
+    for id, chosen_assistants in existing_ids_and_chosen_assistants:
+        conn.execute(
+            sa.text(
+                "update public.user set chosen_assistants = :chosen_assistants where id = :id"
+            ),
+            {"chosen_assistants": chosen_assistants, "id": id},
+        )

backend/alembic/versions/dbaa756c2ccf_embedding_models.py

@@ -9,7 +9,7 @@ from alembic import op
 import sqlalchemy as sa
 from sqlalchemy import table, column, String, Integer, Boolean
 
-from danswer.db.embedding_model import (
+from danswer.db.search_settings import (
     get_new_default_embedding_model,
     get_old_default_embedding_model,
     user_has_overridden_embedding_model,
@@ -71,14 +71,14 @@ def upgrade() -> None:
                 "query_prefix": old_embedding_model.query_prefix,
                 "passage_prefix": old_embedding_model.passage_prefix,
                 "index_name": old_embedding_model.index_name,
-                "status": old_embedding_model.status,
+                "status": IndexModelStatus.PRESENT,
             }
         ],
     )
     # if the user has not overridden the default embedding model via env variables,
     # insert the new default model into the database to auto-upgrade them
     if not user_has_overridden_embedding_model():
-        new_embedding_model = get_new_default_embedding_model(is_present=False)
+        new_embedding_model = get_new_default_embedding_model()
         op.bulk_insert(
             EmbeddingModel,
             [

backend/alembic/versions/ee3f4b47fad5_added_alternate_model_to_chat_message.py

@@ -0,0 +1,28 @@
+"""Added alternate model to chat message
+
+Revision ID: ee3f4b47fad5
+Revises: 2d2304e27d8c
+Create Date: 2024-08-12 00:11:50.915845
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = "ee3f4b47fad5"
+down_revision = "2d2304e27d8c"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "chat_message",
+        sa.Column("overridden_model", sa.String(length=255), nullable=True),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("chat_message", "overridden_model")

backend/alembic/versions/efb35676026c_standard_answer_match_regex_flag.py

@@ -0,0 +1,32 @@
+"""standard answer match_regex flag
+
+Revision ID: efb35676026c
+Revises: 0ebb1d516877
+Create Date: 2024-09-11 13:55:46.101149
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "efb35676026c"
+down_revision = "0ebb1d516877"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.add_column(
+        "standard_answer",
+        sa.Column(
+            "match_regex", sa.Boolean(), nullable=False, server_default=sa.false()
+        ),
+    )
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_column("standard_answer", "match_regex")
+    # ### end Alembic commands ###

backend/alembic/versions/f17bf3b0d9f1_embedding_provider_by_provider_type.py

@@ -0,0 +1,172 @@
+"""embedding provider by provider type
+
+Revision ID: f17bf3b0d9f1
+Revises: 351faebd379d
+Create Date: 2024-08-21 13:13:31.120460
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = "f17bf3b0d9f1"
+down_revision = "351faebd379d"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    # Add provider_type column to embedding_provider
+    op.add_column(
+        "embedding_provider",
+        sa.Column("provider_type", sa.String(50), nullable=True),
+    )
+
+    # Update provider_type with existing name values
+    op.execute("UPDATE embedding_provider SET provider_type = UPPER(name)")
+
+    # Make provider_type not nullable
+    op.alter_column("embedding_provider", "provider_type", nullable=False)
+
+    # Drop the foreign key constraint in embedding_model table
+    op.drop_constraint(
+        "fk_embedding_model_cloud_provider", "embedding_model", type_="foreignkey"
+    )
+
+    # Drop the existing primary key constraint
+    op.drop_constraint("embedding_provider_pkey", "embedding_provider", type_="primary")
+
+    # Create a new primary key constraint on provider_type
+    op.create_primary_key(
+        "embedding_provider_pkey", "embedding_provider", ["provider_type"]
+    )
+
+    # Add provider_type column to embedding_model
+    op.add_column(
+        "embedding_model",
+        sa.Column("provider_type", sa.String(50), nullable=True),
+    )
+
+    # Update provider_type for existing embedding models
+    op.execute(
+        """
+        UPDATE embedding_model
+        SET provider_type = (
+            SELECT provider_type
+            FROM embedding_provider
+            WHERE embedding_provider.id = embedding_model.cloud_provider_id
+        )
+    """
+    )
+
+    # Drop the old id column from embedding_provider
+    op.drop_column("embedding_provider", "id")
+
+    # Drop the name column from embedding_provider
+    op.drop_column("embedding_provider", "name")
+
+    # Drop the default_model_id column from embedding_provider
+    op.drop_column("embedding_provider", "default_model_id")
+
+    # Drop the old cloud_provider_id column from embedding_model
+    op.drop_column("embedding_model", "cloud_provider_id")
+
+    # Create the new foreign key constraint
+    op.create_foreign_key(
+        "fk_embedding_model_cloud_provider",
+        "embedding_model",
+        "embedding_provider",
+        ["provider_type"],
+        ["provider_type"],
+    )
+
+
+def downgrade() -> None:
+    # Drop the foreign key constraint in embedding_model table
+    op.drop_constraint(
+        "fk_embedding_model_cloud_provider", "embedding_model", type_="foreignkey"
+    )
+
+    # Add back the cloud_provider_id column to embedding_model
+    op.add_column(
+        "embedding_model", sa.Column("cloud_provider_id", sa.Integer(), nullable=True)
+    )
+    op.add_column("embedding_provider", sa.Column("id", sa.Integer(), nullable=True))
+
+    # Assign incrementing IDs to embedding providers
+    op.execute(
+        """
+        CREATE SEQUENCE IF NOT EXISTS embedding_provider_id_seq;"""
+    )
+    op.execute(
+        """
+        UPDATE embedding_provider SET id = nextval('embedding_provider_id_seq');
+    """
+    )
+
+    # Update cloud_provider_id based on provider_type
+    op.execute(
+        """
+        UPDATE embedding_model
+        SET cloud_provider_id = CASE
+            WHEN provider_type IS NULL THEN NULL
+            ELSE (
+                SELECT id
+                FROM embedding_provider
+                WHERE embedding_provider.provider_type = embedding_model.provider_type
+            )
+        END
+    """
+    )
+
+    # Drop the provider_type column from embedding_model
+    op.drop_column("embedding_model", "provider_type")
+
+    # Add back the columns to embedding_provider
+    op.add_column("embedding_provider", sa.Column("name", sa.String(50), nullable=True))
+    op.add_column(
+        "embedding_provider", sa.Column("default_model_id", sa.Integer(), nullable=True)
+    )
+
+    # Drop the existing primary key constraint on provider_type
+    op.drop_constraint("embedding_provider_pkey", "embedding_provider", type_="primary")
+
+    # Create the original primary key constraint on id
+    op.create_primary_key("embedding_provider_pkey", "embedding_provider", ["id"])
+
+    # Update name with existing provider_type values
+    op.execute(
+        """
+        UPDATE embedding_provider
+        SET name = CASE
+            WHEN provider_type = 'OPENAI' THEN 'OpenAI'
+            WHEN provider_type = 'COHERE' THEN 'Cohere'
+            WHEN provider_type = 'GOOGLE' THEN 'Google'
+            WHEN provider_type = 'VOYAGE' THEN 'Voyage'
+            ELSE provider_type
+        END
+    """
+    )
+
+    # Drop the provider_type column from embedding_provider
+    op.drop_column("embedding_provider", "provider_type")
+
+    # Recreate the foreign key constraint in embedding_model table
+    op.create_foreign_key(
+        "fk_embedding_model_cloud_provider",
+        "embedding_model",
+        "embedding_provider",
+        ["cloud_provider_id"],
+        ["id"],
+    )
+
+    # Recreate the foreign key constraint in embedding_model table
+    op.create_foreign_key(
+        "fk_embedding_provider_default_model",
+        "embedding_provider",
+        "embedding_model",
+        ["default_model_id"],
+        ["id"],
+    )

backend/alembic/versions/f32615f71aeb_add_custom_headers_to_tools.py

@@ -0,0 +1,26 @@
+"""add custom headers to tools
+
+Revision ID: f32615f71aeb
+Revises: bd2921608c3a
+Create Date: 2024-09-12 20:26:38.932377
+
+"""
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision = "f32615f71aeb"
+down_revision = "bd2921608c3a"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "tool", sa.Column("custom_headers", postgresql.JSONB(), nullable=True)
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("tool", "custom_headers")

backend/alembic/versions/f7e58d357687_add_has_web_column_to_user.py

@@ -0,0 +1,26 @@
+"""add has_web_login column to user
+
+Revision ID: f7e58d357687
+Revises: ba98eba0f66a
+Create Date: 2024-09-07 20:20:54.522620
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "f7e58d357687"
+down_revision = "ba98eba0f66a"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "user",
+        sa.Column("has_web_login", sa.Boolean(), nullable=False, server_default="true"),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("user", "has_web_login")

backend/danswer/access/access.py

@@ -1,26 +1,81 @@
 from sqlalchemy.orm import Session
 
 from danswer.access.models import DocumentAccess
-from danswer.access.utils import prefix_user
+from danswer.access.utils import prefix_user_email
 from danswer.configs.constants import PUBLIC_DOC_PAT
-from danswer.db.document import get_acccess_info_for_documents
+from danswer.db.document import get_access_info_for_document
+from danswer.db.document import get_access_info_for_documents
 from danswer.db.models import User
 from danswer.utils.variable_functionality import fetch_versioned_implementation
 
 
+def _get_access_for_document(
+    document_id: str,
+    db_session: Session,
+) -> DocumentAccess:
+    info = get_access_info_for_document(
+        db_session=db_session,
+        document_id=document_id,
+    )
+
+    return DocumentAccess.build(
+        user_emails=info[1] if info and info[1] else [],
+        user_groups=[],
+        external_user_emails=[],
+        external_user_group_ids=[],
+        is_public=info[2] if info else False,
+    )
+
+
+def get_access_for_document(
+    document_id: str,
+    db_session: Session,
+) -> DocumentAccess:
+    versioned_get_access_for_document_fn = fetch_versioned_implementation(
+        "danswer.access.access", "_get_access_for_document"
+    )
+    return versioned_get_access_for_document_fn(document_id, db_session)  # type: ignore
+
+
+def get_null_document_access() -> DocumentAccess:
+    return DocumentAccess(
+        user_emails=set(),
+        user_groups=set(),
+        is_public=False,
+        external_user_emails=set(),
+        external_user_group_ids=set(),
+    )
+
+
 def _get_access_for_documents(
     document_ids: list[str],
     db_session: Session,
 ) -> dict[str, DocumentAccess]:
-    document_access_info = get_acccess_info_for_documents(
+    document_access_info = get_access_info_for_documents(
         db_session=db_session,
         document_ids=document_ids,
     )
-    return {
-        document_id: DocumentAccess.build(user_ids, [], is_public)
-        for document_id, user_ids, is_public in document_access_info
+    doc_access = {
+        document_id: DocumentAccess(
+            user_emails=set([email for email in user_emails if email]),
+            # MIT version will wipe all groups and external groups on update
+            user_groups=set(),
+            is_public=is_public,
+            external_user_emails=set(),
+            external_user_group_ids=set(),
+        )
+        for document_id, user_emails, is_public in document_access_info
     }
 
+    # Sometimes the document has not be indexed by the indexing job yet, in those cases
+    # the document does not exist and so we use least permissive. Specifically the EE version
+    # checks the MIT version permissions and creates a superset. This ensures that this flow
+    # does not fail even if the Document has not yet been indexed.
+    for doc_id in document_ids:
+        if doc_id not in doc_access:
+            doc_access[doc_id] = get_null_document_access()
+    return doc_access
+
 
 def get_access_for_documents(
     document_ids: list[str],
@@ -42,7 +97,7 @@ def _get_acl_for_user(user: User | None, db_session: Session) -> set[str]:
     matches one entry in the returned set.
     """
     if user:
-        return {prefix_user(str(user.id)), PUBLIC_DOC_PAT}
+        return {prefix_user_email(user.email), PUBLIC_DOC_PAT}
     return {PUBLIC_DOC_PAT}
 
 

backend/danswer/access/models.py

@@ -1,30 +1,72 @@
 from dataclasses import dataclass
-from uuid import UUID
 
-from danswer.access.utils import prefix_user
+from danswer.access.utils import prefix_external_group
+from danswer.access.utils import prefix_user_email
 from danswer.access.utils import prefix_user_group
 from danswer.configs.constants import PUBLIC_DOC_PAT
 
 
 @dataclass(frozen=True)
-class DocumentAccess:
-    user_ids: set[str]  # stringified UUIDs
-    user_groups: set[str]  # names of user groups associated with this document
+class ExternalAccess:
+    # Emails of external users with access to the doc externally
+    external_user_emails: set[str]
+    # Names or external IDs of groups with access to the doc
+    external_user_group_ids: set[str]
+    # Whether the document is public in the external system or Danswer
     is_public: bool
 
-    def to_acl(self) -> list[str]:
-        return (
-            [prefix_user(user_id) for user_id in self.user_ids]
+
+@dataclass(frozen=True)
+class DocumentAccess(ExternalAccess):
+    # User emails for Danswer users, None indicates admin
+    user_emails: set[str | None]
+    # Names of user groups associated with this document
+    user_groups: set[str]
+
+    def to_acl(self) -> set[str]:
+        return set(
+            [
+                prefix_user_email(user_email)
+                for user_email in self.user_emails
+                if user_email
+            ]
             + [prefix_user_group(group_name) for group_name in self.user_groups]
+            + [
+                prefix_user_email(user_email)
+                for user_email in self.external_user_emails
+            ]
+            + [
+                # The group names are already prefixed by the source type
+                # This adds an additional prefix of "external_group:"
+                prefix_external_group(group_name)
+                for group_name in self.external_user_group_ids
+            ]
             + ([PUBLIC_DOC_PAT] if self.is_public else [])
         )
 
     @classmethod
     def build(
-        cls, user_ids: list[UUID | None], user_groups: list[str], is_public: bool
+        cls,
+        user_emails: list[str | None],
+        user_groups: list[str],
+        external_user_emails: list[str],
+        external_user_group_ids: list[str],
+        is_public: bool,
     ) -> "DocumentAccess":
         return cls(
-            user_ids={str(user_id) for user_id in user_ids if user_id},
+            external_user_emails={
+                prefix_user_email(external_email)
+                for external_email in external_user_emails
+            },
+            external_user_group_ids={
+                prefix_external_group(external_group_id)
+                for external_group_id in external_user_group_ids
+            },
+            user_emails={
+                prefix_user_email(user_email)
+                for user_email in user_emails
+                if user_email
+            },
             user_groups=set(user_groups),
             is_public=is_public,
         )

backend/danswer/access/utils.py

@@ -1,10 +1,24 @@
-def prefix_user(user_id: str) -> str:
-    """Prefixes a user ID to eliminate collision with group names.
-    This assumes that groups are prefixed with a different prefix."""
-    return f"user_id:{user_id}"
+from danswer.configs.constants import DocumentSource
+
+
+def prefix_user_email(user_email: str) -> str:
+    """Prefixes a user email to eliminate collision with group names.
+    This applies to both a Danswer user and an External user, this is to make the query time
+    more efficient"""
+    return f"user_email:{user_email}"
 
 
 def prefix_user_group(user_group_name: str) -> str:
-    """Prefixes a user group name to eliminate collision with user IDs.
+    """Prefixes a user group name to eliminate collision with user emails.
     This assumes that user ids are prefixed with a different prefix."""
     return f"group:{user_group_name}"
+
+
+def prefix_external_group(ext_group_name: str) -> str:
+    """Prefixes an external group name to eliminate collision with user emails / Danswer groups."""
+    return f"external_group:{ext_group_name}"
+
+
+def prefix_group_w_source(ext_group_name: str, source: DocumentSource) -> str:
+    """External groups may collide across sources, every source needs its own prefix."""
+    return f"{source.value.upper()}_{ext_group_name}"

backend/danswer/auth/noauth_user.py

@@ -13,7 +13,7 @@ from danswer.server.manage.models import UserPreferences
 def set_no_auth_user_preferences(
     store: DynamicConfigStore, preferences: UserPreferences
 ) -> None:
-    store.store(KV_NO_AUTH_USER_PREFERENCES_KEY, preferences.dict())
+    store.store(KV_NO_AUTH_USER_PREFERENCES_KEY, preferences.model_dump())
 
 
 def load_no_auth_user_preferences(store: DynamicConfigStore) -> UserPreferences:

backend/danswer/auth/schemas.py

@@ -5,8 +5,20 @@ from fastapi_users import schemas
 
 
 class UserRole(str, Enum):
+    """
+    User roles
+    - Basic can't perform any admin actions
+    - Admin can perform all admin actions
+    - Curator can perform admin actions for
+        groups they are curators of
+    - Global Curator can perform admin actions
+        for all groups they are a member of
+    """
+
     BASIC = "basic"
     ADMIN = "admin"
+    CURATOR = "curator"
+    GLOBAL_CURATOR = "global_curator"
 
 
 class UserStatus(str, Enum):
@@ -21,7 +33,9 @@ class UserRead(schemas.BaseUser[uuid.UUID]):
 
 class UserCreate(schemas.BaseUserCreate):
     role: UserRole = UserRole.BASIC
+    has_web_login: bool | None = True
 
 
 class UserUpdate(schemas.BaseUserUpdate):
     role: UserRole
+    has_web_login: bool | None = True

backend/danswer/auth/users.py

@@ -8,13 +8,17 @@ from email.mime.text import MIMEText
 from typing import Optional
 from typing import Tuple
 
+from email_validator import EmailNotValidError
+from email_validator import validate_email
 from fastapi import APIRouter
 from fastapi import Depends
 from fastapi import HTTPException
 from fastapi import Request
 from fastapi import Response
 from fastapi import status
+from fastapi.security import OAuth2PasswordRequestForm
 from fastapi_users import BaseUserManager
+from fastapi_users import exceptions
 from fastapi_users import FastAPIUsers
 from fastapi_users import models
 from fastapi_users import schemas
@@ -31,6 +35,7 @@ from sqlalchemy.orm import Session
 from danswer.auth.invited_users import get_invited_users
 from danswer.auth.schemas import UserCreate
 from danswer.auth.schemas import UserRole
+from danswer.auth.schemas import UserUpdate
 from danswer.configs.app_configs import AUTH_TYPE
 from danswer.configs.app_configs import DISABLE_AUTH
 from danswer.configs.app_configs import EMAIL_FROM
@@ -40,6 +45,7 @@ from danswer.configs.app_configs import SMTP_PASS
 from danswer.configs.app_configs import SMTP_PORT
 from danswer.configs.app_configs import SMTP_SERVER
 from danswer.configs.app_configs import SMTP_USER
+from danswer.configs.app_configs import TRACK_EXTERNAL_IDP_EXPIRY
 from danswer.configs.app_configs import USER_AUTH_SECRET
 from danswer.configs.app_configs import VALID_EMAIL_DOMAINS
 from danswer.configs.app_configs import WEB_DOMAIN
@@ -59,10 +65,7 @@ from danswer.db.users import get_user_by_email
 from danswer.utils.logger import setup_logger
 from danswer.utils.telemetry import optional_telemetry
 from danswer.utils.telemetry import RecordType
-from danswer.utils.variable_functionality import (
-    fetch_versioned_implementation,
-)
-
+from danswer.utils.variable_functionality import fetch_versioned_implementation
 
 logger = setup_logger()
 
@@ -81,7 +84,7 @@ def verify_auth_setting() -> None:
             "User must choose a valid user authentication method: "
             "disabled, basic, or google_oauth"
         )
-    logger.info(f"Using Auth Type: {AUTH_TYPE.value}")
+    logger.notice(f"Using Auth Type: {AUTH_TYPE.value}")
 
 
 def get_display_email(email: str | None, space_less: bool = False) -> str:
@@ -106,8 +109,28 @@ def user_needs_to_be_verified() -> bool:
 
 def verify_email_is_invited(email: str) -> None:
     whitelist = get_invited_users()
-    if (whitelist and email not in whitelist) or not email:
-        raise PermissionError("User not on allowed user whitelist")
+    if not whitelist:
+        return
+
+    if not email:
+        raise PermissionError("Email must be specified")
+
+    email_info = validate_email(email)  # can raise EmailNotValidError
+
+    for email_whitelist in whitelist:
+        try:
+            # normalized emails are now being inserted into the db
+            # we can remove this normalization on read after some time has passed
+            email_info_whitelist = validate_email(email_whitelist)
+        except EmailNotValidError:
+            continue
+
+        # oddly, normalization does not include lowercasing the user part of the
+        # email address ... which we want to allow
+        if email_info.normalized.lower() == email_info_whitelist.normalized.lower():
+            return
+
+    raise PermissionError("User not on allowed user whitelist")
 
 
 def verify_email_in_whitelist(email: str) -> None:
@@ -164,7 +187,7 @@ class UserManager(UUIDIDMixin, BaseUserManager[User, uuid.UUID]):
         user_create: schemas.UC | UserCreate,
         safe: bool = False,
         request: Optional[Request] = None,
-    ) -> models.UP:
+    ) -> User:
         verify_email_is_invited(user_create.email)
         verify_email_domain(user_create.email)
         if hasattr(user_create, "role"):
@@ -173,7 +196,27 @@ class UserManager(UUIDIDMixin, BaseUserManager[User, uuid.UUID]):
                 user_create.role = UserRole.ADMIN
             else:
                 user_create.role = UserRole.BASIC
-        return await super().create(user_create, safe=safe, request=request)  # type: ignore
+        user = None
+        try:
+            user = await super().create(user_create, safe=safe, request=request)  # type: ignore
+        except exceptions.UserAlreadyExists:
+            user = await self.get_by_email(user_create.email)
+            # Handle case where user has used product outside of web and is now creating an account through web
+            if (
+                not user.has_web_login
+                and hasattr(user_create, "has_web_login")
+                and user_create.has_web_login
+            ):
+                user_update = UserUpdate(
+                    password=user_create.password,
+                    has_web_login=True,
+                    role=user_create.role,
+                    is_verified=user_create.is_verified,
+                )
+                user = await self.update(user_update, user)
+            else:
+                raise exceptions.UserAlreadyExists()
+        return user
 
     async def oauth_callback(
         self: "BaseUserManager[models.UOAP, models.ID]",
@@ -203,18 +246,35 @@ class UserManager(UUIDIDMixin, BaseUserManager[User, uuid.UUID]):
             is_verified_by_default=is_verified_by_default,
         )
 
-        # NOTE: google oauth expires after 1hr. We don't want to force the user to
-        # re-authenticate that frequently, so for now we'll just ignore this for
-        # google oauth users
-        if expires_at and AUTH_TYPE != AuthType.GOOGLE_OAUTH:
+        # NOTE: Most IdPs have very short expiry times, and we don't want to force the user to
+        # re-authenticate that frequently, so by default this is disabled
+        if expires_at and TRACK_EXTERNAL_IDP_EXPIRY:
             oidc_expiry = datetime.fromtimestamp(expires_at, tz=timezone.utc)
             await self.user_db.update(user, update_dict={"oidc_expiry": oidc_expiry})
+
+        # this is needed if an organization goes from `TRACK_EXTERNAL_IDP_EXPIRY=true` to `false`
+        # otherwise, the oidc expiry will always be old, and the user will never be able to login
+        if user.oidc_expiry and not TRACK_EXTERNAL_IDP_EXPIRY:
+            await self.user_db.update(user, update_dict={"oidc_expiry": None})
+
+        # Handle case where user has used product outside of web and is now creating an account through web
+        if not user.has_web_login:
+            await self.user_db.update(
+                user,
+                update_dict={
+                    "is_verified": is_verified_by_default,
+                    "has_web_login": True,
+                },
+            )
+            user.is_verified = is_verified_by_default
+            user.has_web_login = True
+
         return user
 
     async def on_after_register(
         self, user: User, request: Optional[Request] = None
     ) -> None:
-        logger.info(f"User {user.id} has registered.")
+        logger.notice(f"User {user.id} has registered.")
         optional_telemetry(
             record_type=RecordType.SIGN_UP,
             data={"action": "create"},
@@ -224,19 +284,45 @@ class UserManager(UUIDIDMixin, BaseUserManager[User, uuid.UUID]):
     async def on_after_forgot_password(
         self, user: User, token: str, request: Optional[Request] = None
     ) -> None:
-        logger.info(f"User {user.id} has forgot their password. Reset token: {token}")
+        logger.notice(f"User {user.id} has forgot their password. Reset token: {token}")
 
     async def on_after_request_verify(
         self, user: User, token: str, request: Optional[Request] = None
     ) -> None:
         verify_email_domain(user.email)
 
-        logger.info(
+        logger.notice(
             f"Verification requested for user {user.id}. Verification token: {token}"
         )
 
         send_user_verification_email(user.email, token)
 
+    async def authenticate(
+        self, credentials: OAuth2PasswordRequestForm
+    ) -> Optional[User]:
+        try:
+            user = await self.get_by_email(credentials.username)
+        except exceptions.UserNotExists:
+            self.password_helper.hash(credentials.password)
+            return None
+
+        if not user.has_web_login:
+            raise HTTPException(
+                status_code=status.HTTP_403_FORBIDDEN,
+                detail="NO_WEB_LOGIN_AND_HAS_NO_PASSWORD",
+            )
+
+        verified, updated_password_hash = self.password_helper.verify_and_update(
+            credentials.password, user.hashed_password
+        )
+        if not verified:
+            return None
+
+        if updated_password_hash is not None:
+            await self.user_db.update(user, {"hashed_password": updated_password_hash})
+
+        return user
+
 
 async def get_user_manager(
     user_db: SQLAlchemyUserDatabase = Depends(get_user_db),
@@ -339,6 +425,7 @@ async def optional_user(
 async def double_check_user(
     user: User | None,
     optional: bool = DISABLE_AUTH,
+    include_expired: bool = False,
 ) -> User | None:
     if optional:
         return None
@@ -355,7 +442,11 @@ async def double_check_user(
             detail="Access denied. User is not verified.",
         )
 
-    if user.oidc_expiry and user.oidc_expiry < datetime.now(timezone.utc):
+    if (
+        user.oidc_expiry
+        and user.oidc_expiry < datetime.now(timezone.utc)
+        and not include_expired
+    ):
         raise HTTPException(
             status_code=status.HTTP_403_FORBIDDEN,
             detail="Access denied. User's OIDC token has expired.",
@@ -364,12 +455,40 @@ async def double_check_user(
     return user
 
 
+async def current_user_with_expired_token(
+    user: User | None = Depends(optional_user),
+) -> User | None:
+    return await double_check_user(user, include_expired=True)
+
+
 async def current_user(
     user: User | None = Depends(optional_user),
 ) -> User | None:
     return await double_check_user(user)
 
 
+async def current_curator_or_admin_user(
+    user: User | None = Depends(current_user),
+) -> User | None:
+    if DISABLE_AUTH:
+        return None
+
+    if not user or not hasattr(user, "role"):
+        raise HTTPException(
+            status_code=status.HTTP_403_FORBIDDEN,
+            detail="Access denied. User is not authenticated or lacks role information.",
+        )
+
+    allowed_roles = {UserRole.GLOBAL_CURATOR, UserRole.CURATOR, UserRole.ADMIN}
+    if user.role not in allowed_roles:
+        raise HTTPException(
+            status_code=status.HTTP_403_FORBIDDEN,
+            detail="Access denied. User is not a curator or admin.",
+        )
+
+    return user
+
+
 async def current_admin_user(user: User | None = Depends(current_user)) -> User | None:
     if DISABLE_AUTH:
         return None
@@ -377,7 +496,12 @@ async def current_admin_user(user: User | None = Depends(current_user)) -> User
     if not user or not hasattr(user, "role") or user.role != UserRole.ADMIN:
         raise HTTPException(
             status_code=status.HTTP_403_FORBIDDEN,
-            detail="Access denied. User is not an admin.",
+            detail="Access denied. User must be an admin to perform this action.",
         )
 
     return user
+
+
+def get_default_admin_user_emails_() -> list[str]:
+    # No default seeding available for Danswer MIT
+    return []

backend/danswer/background/celery/celery_redis.py

@@ -0,0 +1,361 @@
+# These are helper objects for tracking the keys we need to write in redis
+import time
+from abc import ABC
+from abc import abstractmethod
+from typing import cast
+from uuid import uuid4
+
+import redis
+from celery import Celery
+from redis import Redis
+from sqlalchemy.orm import Session
+
+from danswer.background.celery.celeryconfig import CELERY_SEPARATOR
+from danswer.configs.constants import CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT
+from danswer.configs.constants import DanswerCeleryPriority
+from danswer.configs.constants import DanswerCeleryQueues
+from danswer.db.connector_credential_pair import get_connector_credential_pair_from_id
+from danswer.db.document import construct_document_select_for_connector_credential_pair
+from danswer.db.document import (
+    construct_document_select_for_connector_credential_pair_by_needs_sync,
+)
+from danswer.db.document_set import construct_document_select_by_docset
+from danswer.utils.variable_functionality import fetch_versioned_implementation
+
+
+class RedisObjectHelper(ABC):
+    PREFIX = "base"
+    FENCE_PREFIX = PREFIX + "_fence"
+    TASKSET_PREFIX = PREFIX + "_taskset"
+
+    def __init__(self, id: int):
+        self._id: int = id
+
+    @property
+    def task_id_prefix(self) -> str:
+        return f"{self.PREFIX}_{self._id}"
+
+    @property
+    def fence_key(self) -> str:
+        # example: documentset_fence_1
+        return f"{self.FENCE_PREFIX}_{self._id}"
+
+    @property
+    def taskset_key(self) -> str:
+        # example: documentset_taskset_1
+        return f"{self.TASKSET_PREFIX}_{self._id}"
+
+    @staticmethod
+    def get_id_from_fence_key(key: str) -> int | None:
+        """
+        Extracts the object ID from a fence key in the format `PREFIX_fence_X`.
+
+        Args:
+            key (str): The fence key string.
+
+        Returns:
+            Optional[int]: The extracted ID if the key is in the correct format, otherwise None.
+        """
+        parts = key.split("_")
+        if len(parts) != 3:
+            return None
+
+        try:
+            object_id = int(parts[2])
+        except ValueError:
+            return None
+
+        return object_id
+
+    @staticmethod
+    def get_id_from_task_id(task_id: str) -> int | None:
+        """
+        Extracts the object ID from a task ID string.
+
+        This method assumes the task ID is formatted as `prefix_objectid_suffix`, where:
+        - `prefix` is an arbitrary string (e.g., the name of the task or entity),
+        - `objectid` is the ID you want to extract,
+        - `suffix` is another arbitrary string (e.g., a UUID).
+
+        Example:
+            If the input `task_id` is `documentset_1_cbfdc96a-80ca-4312-a242-0bb68da3c1dc`,
+            this method will return the string `"1"`.
+
+        Args:
+            task_id (str): The task ID string from which to extract the object ID.
+
+        Returns:
+            str | None: The extracted object ID if the task ID is in the correct format, otherwise None.
+        """
+        # example: task_id=documentset_1_cbfdc96a-80ca-4312-a242-0bb68da3c1dc
+        parts = task_id.split("_")
+        if len(parts) != 3:
+            return None
+
+        try:
+            object_id = int(parts[1])
+        except ValueError:
+            return None
+
+        return object_id
+
+    @abstractmethod
+    def generate_tasks(
+        self,
+        celery_app: Celery,
+        db_session: Session,
+        redis_client: Redis,
+        lock: redis.lock.Lock,
+    ) -> int | None:
+        pass
+
+
+class RedisDocumentSet(RedisObjectHelper):
+    PREFIX = "documentset"
+    FENCE_PREFIX = PREFIX + "_fence"
+    TASKSET_PREFIX = PREFIX + "_taskset"
+
+    def generate_tasks(
+        self,
+        celery_app: Celery,
+        db_session: Session,
+        redis_client: Redis,
+        lock: redis.lock.Lock,
+    ) -> int | None:
+        last_lock_time = time.monotonic()
+
+        async_results = []
+        stmt = construct_document_select_by_docset(self._id, current_only=False)
+        for doc in db_session.scalars(stmt).yield_per(1):
+            current_time = time.monotonic()
+            if current_time - last_lock_time >= (
+                CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT / 4
+            ):
+                lock.reacquire()
+                last_lock_time = current_time
+
+            # celery's default task id format is "dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            # the key for the result is "celery-task-meta-dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            # we prefix the task id so it's easier to keep track of who created the task
+            # aka "documentset_1_6dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            custom_task_id = f"{self.task_id_prefix}_{uuid4()}"
+
+            # add to the set BEFORE creating the task.
+            redis_client.sadd(self.taskset_key, custom_task_id)
+
+            result = celery_app.send_task(
+                "vespa_metadata_sync_task",
+                kwargs=dict(document_id=doc.id),
+                queue=DanswerCeleryQueues.VESPA_METADATA_SYNC,
+                task_id=custom_task_id,
+                priority=DanswerCeleryPriority.LOW,
+            )
+
+            async_results.append(result)
+
+        return len(async_results)
+
+
+class RedisUserGroup(RedisObjectHelper):
+    PREFIX = "usergroup"
+    FENCE_PREFIX = PREFIX + "_fence"
+    TASKSET_PREFIX = PREFIX + "_taskset"
+
+    def generate_tasks(
+        self,
+        celery_app: Celery,
+        db_session: Session,
+        redis_client: Redis,
+        lock: redis.lock.Lock,
+    ) -> int | None:
+        last_lock_time = time.monotonic()
+
+        async_results = []
+
+        try:
+            construct_document_select_by_usergroup = fetch_versioned_implementation(
+                "danswer.db.user_group",
+                "construct_document_select_by_usergroup",
+            )
+        except ModuleNotFoundError:
+            return 0
+
+        stmt = construct_document_select_by_usergroup(self._id)
+        for doc in db_session.scalars(stmt).yield_per(1):
+            current_time = time.monotonic()
+            if current_time - last_lock_time >= (
+                CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT / 4
+            ):
+                lock.reacquire()
+                last_lock_time = current_time
+
+            # celery's default task id format is "dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            # the key for the result is "celery-task-meta-dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            # we prefix the task id so it's easier to keep track of who created the task
+            # aka "documentset_1_6dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            custom_task_id = f"{self.task_id_prefix}_{uuid4()}"
+
+            # add to the set BEFORE creating the task.
+            redis_client.sadd(self.taskset_key, custom_task_id)
+
+            result = celery_app.send_task(
+                "vespa_metadata_sync_task",
+                kwargs=dict(document_id=doc.id),
+                queue=DanswerCeleryQueues.VESPA_METADATA_SYNC,
+                task_id=custom_task_id,
+                priority=DanswerCeleryPriority.LOW,
+            )
+
+            async_results.append(result)
+
+        return len(async_results)
+
+
+class RedisConnectorCredentialPair(RedisObjectHelper):
+    """This class differs from the default in that the taskset used spans
+    all connectors and is not per connector."""
+
+    PREFIX = "connectorsync"
+    FENCE_PREFIX = PREFIX + "_fence"
+    TASKSET_PREFIX = PREFIX + "_taskset"
+
+    @classmethod
+    def get_fence_key(cls) -> str:
+        return RedisConnectorCredentialPair.FENCE_PREFIX
+
+    @classmethod
+    def get_taskset_key(cls) -> str:
+        return RedisConnectorCredentialPair.TASKSET_PREFIX
+
+    @property
+    def taskset_key(self) -> str:
+        """Notice that this is intentionally reusing the same taskset for all
+        connector syncs"""
+        # example: connector_taskset
+        return f"{self.TASKSET_PREFIX}"
+
+    def generate_tasks(
+        self,
+        celery_app: Celery,
+        db_session: Session,
+        redis_client: Redis,
+        lock: redis.lock.Lock,
+    ) -> int | None:
+        last_lock_time = time.monotonic()
+
+        async_results = []
+        cc_pair = get_connector_credential_pair_from_id(self._id, db_session)
+        if not cc_pair:
+            return None
+
+        stmt = construct_document_select_for_connector_credential_pair_by_needs_sync(
+            cc_pair.connector_id, cc_pair.credential_id
+        )
+        for doc in db_session.scalars(stmt).yield_per(1):
+            current_time = time.monotonic()
+            if current_time - last_lock_time >= (
+                CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT / 4
+            ):
+                lock.reacquire()
+                last_lock_time = current_time
+
+            # celery's default task id format is "dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            # the key for the result is "celery-task-meta-dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            # we prefix the task id so it's easier to keep track of who created the task
+            # aka "documentset_1_6dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            custom_task_id = f"{self.task_id_prefix}_{uuid4()}"
+
+            # add to the tracking taskset in redis BEFORE creating the celery task.
+            # note that for the moment we are using a single taskset key, not differentiated by cc_pair id
+            redis_client.sadd(
+                RedisConnectorCredentialPair.get_taskset_key(), custom_task_id
+            )
+
+            # Priority on sync's triggered by new indexing should be medium
+            result = celery_app.send_task(
+                "vespa_metadata_sync_task",
+                kwargs=dict(document_id=doc.id),
+                queue=DanswerCeleryQueues.VESPA_METADATA_SYNC,
+                task_id=custom_task_id,
+                priority=DanswerCeleryPriority.MEDIUM,
+            )
+
+            async_results.append(result)
+
+        return len(async_results)
+
+
+class RedisConnectorDeletion(RedisObjectHelper):
+    PREFIX = "connectordeletion"
+    FENCE_PREFIX = PREFIX + "_fence"
+    TASKSET_PREFIX = PREFIX + "_taskset"
+
+    def generate_tasks(
+        self,
+        celery_app: Celery,
+        db_session: Session,
+        redis_client: Redis,
+        lock: redis.lock.Lock,
+    ) -> int | None:
+        last_lock_time = time.monotonic()
+
+        async_results = []
+        cc_pair = get_connector_credential_pair_from_id(self._id, db_session)
+        if not cc_pair:
+            return None
+
+        stmt = construct_document_select_for_connector_credential_pair(
+            cc_pair.connector_id, cc_pair.credential_id
+        )
+        for doc in db_session.scalars(stmt).yield_per(1):
+            current_time = time.monotonic()
+            if current_time - last_lock_time >= (
+                CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT / 4
+            ):
+                lock.reacquire()
+                last_lock_time = current_time
+
+            # celery's default task id format is "dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            # the actual redis key is "celery-task-meta-dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            # we prefix the task id so it's easier to keep track of who created the task
+            # aka "documentset_1_6dd32ded3-00aa-4884-8b21-42f8332e7fac"
+            custom_task_id = f"{self.task_id_prefix}_{uuid4()}"
+
+            # add to the tracking taskset in redis BEFORE creating the celery task.
+            # note that for the moment we are using a single taskset key, not differentiated by cc_pair id
+            redis_client.sadd(self.taskset_key, custom_task_id)
+
+            # Priority on sync's triggered by new indexing should be medium
+            result = celery_app.send_task(
+                "document_by_cc_pair_cleanup_task",
+                kwargs=dict(
+                    document_id=doc.id,
+                    connector_id=cc_pair.connector_id,
+                    credential_id=cc_pair.credential_id,
+                ),
+                queue=DanswerCeleryQueues.CONNECTOR_DELETION,
+                task_id=custom_task_id,
+                priority=DanswerCeleryPriority.MEDIUM,
+            )
+
+            async_results.append(result)
+
+        return len(async_results)
+
+
+def celery_get_queue_length(queue: str, r: Redis) -> int:
+    """This is a redis specific way to get the length of a celery queue.
+    It is priority aware and knows how to count across the multiple redis lists
+    used to implement task prioritization.
+    This operation is not atomic."""
+    total_length = 0
+    for i in range(len(DanswerCeleryPriority)):
+        queue_name = queue
+        if i > 0:
+            queue_name += CELERY_SEPARATOR
+            queue_name += str(i)
+
+        length = r.llen(queue_name)
+        total_length += cast(int, length)
+
+    return total_length

backend/danswer/background/celery/celery_utils.py

@@ -3,9 +3,8 @@ from datetime import timezone
 
 from sqlalchemy.orm import Session
 
-from danswer.background.task_utils import name_cc_cleanup_task
+from danswer.background.celery.celery_redis import RedisConnectorDeletion
 from danswer.background.task_utils import name_cc_prune_task
-from danswer.background.task_utils import name_document_set_sync_task
 from danswer.configs.app_configs import ALLOW_SIMULTANEOUS_PRUNING
 from danswer.configs.app_configs import MAX_PRUNING_DOCUMENT_RETRIEVAL_PER_MINUTE
 from danswer.connectors.cross_connector_utils.rate_limit_wrapper import (
@@ -16,36 +15,50 @@ from danswer.connectors.interfaces import IdConnector
 from danswer.connectors.interfaces import LoadConnector
 from danswer.connectors.interfaces import PollConnector
 from danswer.connectors.models import Document
-from danswer.db.deletion_attempt import check_deletion_attempt_is_allowed
+from danswer.db.connector_credential_pair import get_connector_credential_pair
 from danswer.db.engine import get_db_current_time
-from danswer.db.enums import ConnectorCredentialPairStatus
+from danswer.db.enums import TaskStatus
 from danswer.db.models import Connector
-from danswer.db.models import ConnectorCredentialPair
 from danswer.db.models import Credential
-from danswer.db.models import DocumentSet
 from danswer.db.models import TaskQueueState
 from danswer.db.tasks import check_task_is_live_and_not_timed_out
 from danswer.db.tasks import get_latest_task
 from danswer.db.tasks import get_latest_task_by_type
+from danswer.redis.redis_pool import RedisPool
 from danswer.server.documents.models import DeletionAttemptSnapshot
 from danswer.utils.logger import setup_logger
 
 logger = setup_logger()
+redis_pool = RedisPool()
 
 
-def get_deletion_status(
+def _get_deletion_status(
     connector_id: int, credential_id: int, db_session: Session
 ) -> TaskQueueState | None:
-    cleanup_task_name = name_cc_cleanup_task(
-        connector_id=connector_id, credential_id=credential_id
+    """We no longer store TaskQueueState in the DB for a deletion attempt.
+    This function populates TaskQueueState by just checking redis.
+    """
+    cc_pair = get_connector_credential_pair(
+        connector_id=connector_id, credential_id=credential_id, db_session=db_session
+    )
+    if not cc_pair:
+        return None
+
+    rcd = RedisConnectorDeletion(cc_pair.id)
+
+    r = redis_pool.get_client()
+    if not r.exists(rcd.fence_key):
+        return None
+
+    return TaskQueueState(
+        task_id="", task_name=rcd.fence_key, status=TaskStatus.STARTED
     )
-    return get_latest_task(task_name=cleanup_task_name, db_session=db_session)
 
 
 def get_deletion_attempt_snapshot(
     connector_id: int, credential_id: int, db_session: Session
 ) -> DeletionAttemptSnapshot | None:
-    deletion_task = get_deletion_status(connector_id, credential_id, db_session)
+    deletion_task = _get_deletion_status(connector_id, credential_id, db_session)
     if not deletion_task:
         return None
 
@@ -56,46 +69,6 @@ def get_deletion_attempt_snapshot(
     )
 
 
-def should_kick_off_deletion_of_cc_pair(
-    cc_pair: ConnectorCredentialPair, db_session: Session
-) -> bool:
-    if cc_pair.status != ConnectorCredentialPairStatus.DELETING:
-        return False
-
-    if check_deletion_attempt_is_allowed(cc_pair, db_session):
-        return False
-
-    deletion_task = get_deletion_status(
-        connector_id=cc_pair.connector_id,
-        credential_id=cc_pair.credential_id,
-        db_session=db_session,
-    )
-    if deletion_task and check_task_is_live_and_not_timed_out(
-        deletion_task,
-        db_session,
-        # 1 hour timeout
-        timeout=60 * 60,
-    ):
-        return False
-
-    return True
-
-
-def should_sync_doc_set(document_set: DocumentSet, db_session: Session) -> bool:
-    if document_set.is_up_to_date:
-        return False
-
-    task_name = name_document_set_sync_task(document_set.id)
-    latest_sync = get_latest_task(task_name, db_session)
-
-    if latest_sync and check_task_is_live_and_not_timed_out(latest_sync, db_session):
-        logger.info(f"Document set '{document_set.id}' is already syncing. Skipping.")
-        return False
-
-    logger.info(f"Document set {document_set.id} syncing now!")
-    return True
-
-
 def should_prune_cc_pair(
     connector: Connector, credential: Credential, db_session: Session
 ) -> bool:

backend/danswer/background/celery/celeryconfig.py

@@ -0,0 +1,76 @@
+# docs: https://docs.celeryq.dev/en/stable/userguide/configuration.html
+from danswer.configs.app_configs import CELERY_RESULT_EXPIRES
+from danswer.configs.app_configs import REDIS_DB_NUMBER_CELERY
+from danswer.configs.app_configs import REDIS_DB_NUMBER_CELERY_RESULT_BACKEND
+from danswer.configs.app_configs import REDIS_HOST
+from danswer.configs.app_configs import REDIS_PASSWORD
+from danswer.configs.app_configs import REDIS_PORT
+from danswer.configs.app_configs import REDIS_SSL
+from danswer.configs.app_configs import REDIS_SSL_CA_CERTS
+from danswer.configs.app_configs import REDIS_SSL_CERT_REQS
+from danswer.configs.constants import DanswerCeleryPriority
+
+CELERY_SEPARATOR = ":"
+
+CELERY_PASSWORD_PART = ""
+if REDIS_PASSWORD:
+    CELERY_PASSWORD_PART = f":{REDIS_PASSWORD}@"
+
+REDIS_SCHEME = "redis"
+
+# SSL-specific query parameters for Redis URL
+SSL_QUERY_PARAMS = ""
+if REDIS_SSL:
+    REDIS_SCHEME = "rediss"
+    SSL_QUERY_PARAMS = f"?ssl_cert_reqs={REDIS_SSL_CERT_REQS}"
+    if REDIS_SSL_CA_CERTS:
+        SSL_QUERY_PARAMS += f"&ssl_ca_certs={REDIS_SSL_CA_CERTS}"
+
+# example celery_broker_url: "redis://:password@localhost:6379/15"
+broker_url = f"{REDIS_SCHEME}://{CELERY_PASSWORD_PART}{REDIS_HOST}:{REDIS_PORT}/{REDIS_DB_NUMBER_CELERY}{SSL_QUERY_PARAMS}"
+
+result_backend = f"{REDIS_SCHEME}://{CELERY_PASSWORD_PART}{REDIS_HOST}:{REDIS_PORT}/{REDIS_DB_NUMBER_CELERY_RESULT_BACKEND}{SSL_QUERY_PARAMS}"
+
+# NOTE: prefetch 4 is significantly faster than prefetch 1 for small tasks
+# however, prefetching is bad when tasks are lengthy as those tasks
+# can stall other tasks.
+worker_prefetch_multiplier = 4
+
+broker_transport_options = {
+    "priority_steps": list(range(len(DanswerCeleryPriority))),
+    "sep": CELERY_SEPARATOR,
+    "queue_order_strategy": "priority",
+}
+
+task_default_priority = DanswerCeleryPriority.MEDIUM
+task_acks_late = True
+
+# It's possible we don't even need celery's result backend, in which case all of the optimization below
+# might be irrelevant
+result_expires = CELERY_RESULT_EXPIRES  # 86400 seconds is the default
+
+# Option 0: Defaults (json serializer, no compression)
+# about 1.5 KB per queued task. 1KB in queue, 400B for result, 100 as a child entry in generator result
+
+# Option 1: Reduces generator task result sizes by roughly 20%
+# task_compression = "bzip2"
+# task_serializer = "pickle"
+# result_compression = "bzip2"
+# result_serializer = "pickle"
+# accept_content=["pickle"]
+
+# Option 2: this significantly reduces the size of the result for generator tasks since the list of children
+# can be large. small tasks change very little
+# def pickle_bz2_encoder(data):
+#     return bz2.compress(pickle.dumps(data))
+
+# def pickle_bz2_decoder(data):
+#     return pickle.loads(bz2.decompress(data))
+
+# from kombu import serialization  # To register custom serialization with Celery/Kombu
+
+# serialization.register('pickle-bzip2', pickle_bz2_encoder, pickle_bz2_decoder, 'application/x-pickle-bz2', 'binary')
+
+# task_serializer = "pickle-bzip2"
+# result_serializer = "pickle-bzip2"
+# accept_content=["pickle", "pickle-bzip2"]

backend/danswer/background/connector_deletion.py

@@ -13,28 +13,16 @@ connector / credential pair from the access list
 from sqlalchemy.orm import Session
 
 from danswer.access.access import get_access_for_documents
-from danswer.db.connector import fetch_connector_by_id
-from danswer.db.connector_credential_pair import (
-    delete_connector_credential_pair__no_commit,
-)
-from danswer.db.document import delete_document_by_connector_credential_pair__no_commit
+from danswer.db.document import delete_documents_by_connector_credential_pair__no_commit
 from danswer.db.document import delete_documents_complete__no_commit
-from danswer.db.document import get_document_connector_cnts
-from danswer.db.document import get_documents_for_connector_credential_pair
+from danswer.db.document import get_document_connector_counts
 from danswer.db.document import prepare_to_modify_documents
-from danswer.db.document_set import delete_document_set_cc_pair_relationship__no_commit
 from danswer.db.document_set import fetch_document_sets_for_documents
 from danswer.db.engine import get_sqlalchemy_engine
-from danswer.db.index_attempt import delete_index_attempts
-from danswer.db.models import ConnectorCredentialPair
 from danswer.document_index.interfaces import DocumentIndex
 from danswer.document_index.interfaces import UpdateRequest
 from danswer.server.documents.models import ConnectorCredentialPairIdentifier
 from danswer.utils.logger import setup_logger
-from danswer.utils.variable_functionality import (
-    fetch_versioned_implementation_with_fallback,
-)
-from danswer.utils.variable_functionality import noop_fallback
 
 logger = setup_logger()
 
@@ -57,13 +45,15 @@ def delete_connector_credential_pair_batch(
         with prepare_to_modify_documents(
             db_session=db_session, document_ids=document_ids
         ):
-            document_connector_cnts = get_document_connector_cnts(
+            document_connector_counts = get_document_connector_counts(
                 db_session=db_session, document_ids=document_ids
             )
 
             # figure out which docs need to be completely deleted
             document_ids_to_delete = [
-                document_id for document_id, cnt in document_connector_cnts if cnt == 1
+                document_id
+                for document_id, cnt in document_connector_counts
+                if cnt == 1
             ]
             logger.debug(f"Deleting documents: {document_ids_to_delete}")
 
@@ -76,7 +66,7 @@ def delete_connector_credential_pair_batch(
 
             # figure out which docs need to be updated
             document_ids_to_update = [
-                document_id for document_id, cnt in document_connector_cnts if cnt > 1
+                document_id for document_id, cnt in document_connector_counts if cnt > 1
             ]
 
             # maps document id to list of document set names
@@ -109,7 +99,7 @@ def delete_connector_credential_pair_batch(
             document_index.update(update_requests=update_requests)
 
             # clean up Postgres
-            delete_document_by_connector_credential_pair__no_commit(
+            delete_documents_by_connector_credential_pair__no_commit(
                 db_session=db_session,
                 document_ids=document_ids_to_update,
                 connector_credential_pair_identifier=ConnectorCredentialPairIdentifier(
@@ -118,79 +108,3 @@ def delete_connector_credential_pair_batch(
                 ),
             )
             db_session.commit()
-
-
-def delete_connector_credential_pair(
-    db_session: Session,
-    document_index: DocumentIndex,
-    cc_pair: ConnectorCredentialPair,
-) -> int:
-    connector_id = cc_pair.connector_id
-    credential_id = cc_pair.credential_id
-
-    num_docs_deleted = 0
-    while True:
-        documents = get_documents_for_connector_credential_pair(
-            db_session=db_session,
-            connector_id=connector_id,
-            credential_id=credential_id,
-            limit=_DELETION_BATCH_SIZE,
-        )
-        if not documents:
-            break
-
-        delete_connector_credential_pair_batch(
-            document_ids=[document.id for document in documents],
-            connector_id=connector_id,
-            credential_id=credential_id,
-            document_index=document_index,
-        )
-        num_docs_deleted += len(documents)
-
-    # clean up the rest of the related Postgres entities
-    # index attempts
-    delete_index_attempts(
-        db_session=db_session,
-        connector_id=connector_id,
-        credential_id=credential_id,
-    )
-
-    # document sets
-    delete_document_set_cc_pair_relationship__no_commit(
-        db_session=db_session,
-        connector_id=connector_id,
-        credential_id=credential_id,
-    )
-
-    # user groups
-    cleanup_user_groups = fetch_versioned_implementation_with_fallback(
-        "danswer.db.user_group",
-        "delete_user_group_cc_pair_relationship__no_commit",
-        noop_fallback,
-    )
-    cleanup_user_groups(
-        cc_pair_id=cc_pair.id,
-        db_session=db_session,
-    )
-
-    # finally, delete the cc-pair
-    delete_connector_credential_pair__no_commit(
-        db_session=db_session,
-        connector_id=connector_id,
-        credential_id=credential_id,
-    )
-    # if there are no credentials left, delete the connector
-    connector = fetch_connector_by_id(
-        db_session=db_session,
-        connector_id=connector_id,
-    )
-    if not connector or not len(connector.credentials):
-        logger.debug("Found no credentials left for connector, deleting connector")
-        db_session.delete(connector)
-    db_session.commit()
-
-    logger.info(
-        "Successfully deleted connector_credential_pair with connector_id:"
-        f" '{connector_id}' and credential_id: '{credential_id}'. Deleted {num_docs_deleted} docs."
-    )
-    return num_docs_deleted

backend/danswer/background/indexing/run_indexing.py

@@ -7,14 +7,13 @@ from datetime import timezone
 from sqlalchemy.orm import Session
 
 from danswer.background.indexing.checkpointing import get_time_windows_for_index_attempt
+from danswer.background.indexing.tracer import DanswerTracer
 from danswer.configs.app_configs import INDEXING_SIZE_WARNING_THRESHOLD
+from danswer.configs.app_configs import INDEXING_TRACER_INTERVAL
 from danswer.configs.app_configs import POLL_CONNECTOR_OFFSET
+from danswer.connectors.connector_runner import ConnectorRunner
 from danswer.connectors.factory import instantiate_connector
-from danswer.connectors.interfaces import GenerateDocumentsOutput
-from danswer.connectors.interfaces import LoadConnector
-from danswer.connectors.interfaces import PollConnector
 from danswer.connectors.models import IndexAttemptMetadata
-from danswer.connectors.models import InputType
 from danswer.db.connector_credential_pair import get_last_successful_attempt_time
 from danswer.db.connector_credential_pair import update_connector_credential_pair
 from danswer.db.engine import get_sqlalchemy_engine
@@ -22,6 +21,7 @@ from danswer.db.enums import ConnectorCredentialPairStatus
 from danswer.db.index_attempt import get_index_attempt
 from danswer.db.index_attempt import mark_attempt_failed
 from danswer.db.index_attempt import mark_attempt_in_progress
+from danswer.db.index_attempt import mark_attempt_partially_succeeded
 from danswer.db.index_attempt import mark_attempt_succeeded
 from danswer.db.index_attempt import update_docs_indexed
 from danswer.db.models import IndexAttempt
@@ -36,13 +36,15 @@ from danswer.utils.variable_functionality import global_version
 
 logger = setup_logger()
 
+INDEXING_TRACER_NUM_PRINT_ENTRIES = 5
 
-def _get_document_generator(
+
+def _get_connector_runner(
     db_session: Session,
     attempt: IndexAttempt,
     start_time: datetime,
     end_time: datetime,
-) -> GenerateDocumentsOutput:
+) -> ConnectorRunner:
     """
     NOTE: `start_time` and `end_time` are only used for poll connectors
 
@@ -54,11 +56,11 @@ def _get_document_generator(
 
     try:
         runnable_connector = instantiate_connector(
-            attempt.connector_credential_pair.connector.source,
-            task,
-            attempt.connector_credential_pair.connector.connector_specific_config,
-            attempt.connector_credential_pair.credential,
-            db_session,
+            db_session=db_session,
+            source=attempt.connector_credential_pair.connector.source,
+            input_type=task,
+            connector_specific_config=attempt.connector_credential_pair.connector.connector_specific_config,
+            credential=attempt.connector_credential_pair.credential,
         )
     except Exception as e:
         logger.exception(f"Unable to instantiate connector due to {e}")
@@ -72,31 +74,9 @@ def _get_document_generator(
         )
         raise e
 
-    if task == InputType.LOAD_STATE:
-        assert isinstance(runnable_connector, LoadConnector)
-        doc_batch_generator = runnable_connector.load_from_state()
-
-    elif task == InputType.POLL:
-        assert isinstance(runnable_connector, PollConnector)
-        if (
-            attempt.connector_credential_pair.connector_id is None
-            or attempt.connector_credential_pair.connector_id is None
-        ):
-            raise ValueError(
-                f"Polling attempt {attempt.id} is missing connector_id or credential_id, "
-                f"can't fetch time range."
-            )
-
-        logger.info(f"Polling for updates between {start_time} and {end_time}")
-        doc_batch_generator = runnable_connector.poll_source(
-            start=start_time.timestamp(), end=end_time.timestamp()
-        )
-
-    else:
-        # Event types cannot be handled by a background type
-        raise RuntimeError(f"Invalid task type: {task}")
-
-    return doc_batch_generator
+    return ConnectorRunner(
+        connector=runnable_connector, time_range=(start_time, end_time)
+    )
 
 
 def _run_indexing(
@@ -109,54 +89,63 @@ def _run_indexing(
     3. Updates Postgres to record the indexed documents + the outcome of this run
     """
     start_time = time.time()
-    db_embedding_model = index_attempt.embedding_model
-    index_name = db_embedding_model.index_name
+
+    search_settings = index_attempt.search_settings
+    index_name = search_settings.index_name
 
     # Only update cc-pair status for primary index jobs
     # Secondary index syncs at the end when swapping
-    is_primary = index_attempt.embedding_model.status == IndexModelStatus.PRESENT
+    is_primary = search_settings.status == IndexModelStatus.PRESENT
 
     # Indexing is only done into one index at a time
     document_index = get_default_document_index(
         primary_index_name=index_name, secondary_index_name=None
     )
 
-    embedding_model = DefaultIndexingEmbedder(
-        model_name=db_embedding_model.model_name,
-        normalize=db_embedding_model.normalize,
-        query_prefix=db_embedding_model.query_prefix,
-        passage_prefix=db_embedding_model.passage_prefix,
-        api_key=db_embedding_model.api_key,
-        provider_type=db_embedding_model.provider_type,
+    embedding_model = DefaultIndexingEmbedder.from_db_search_settings(
+        search_settings=search_settings
     )
 
     indexing_pipeline = build_indexing_pipeline(
+        attempt_id=index_attempt.id,
         embedder=embedding_model,
         document_index=document_index,
         ignore_time_skip=index_attempt.from_beginning
-        or (db_embedding_model.status == IndexModelStatus.FUTURE),
+        or (search_settings.status == IndexModelStatus.FUTURE),
         db_session=db_session,
     )
 
     db_cc_pair = index_attempt.connector_credential_pair
     db_connector = index_attempt.connector_credential_pair.connector
     db_credential = index_attempt.connector_credential_pair.credential
+    earliest_index_time = (
+        db_connector.indexing_start.timestamp() if db_connector.indexing_start else 0
+    )
 
     last_successful_index_time = (
-        db_connector.indexing_start.timestamp()
-        if index_attempt.from_beginning and db_connector.indexing_start is not None
-        else (
-            0.0
-            if index_attempt.from_beginning
-            else get_last_successful_attempt_time(
-                connector_id=db_connector.id,
-                credential_id=db_credential.id,
-                embedding_model=index_attempt.embedding_model,
-                db_session=db_session,
-            )
+        earliest_index_time
+        if index_attempt.from_beginning
+        else get_last_successful_attempt_time(
+            connector_id=db_connector.id,
+            credential_id=db_credential.id,
+            earliest_index=earliest_index_time,
+            search_settings=index_attempt.search_settings,
+            db_session=db_session,
         )
     )
 
+    if INDEXING_TRACER_INTERVAL > 0:
+        logger.debug(f"Memory tracer starting: interval={INDEXING_TRACER_INTERVAL}")
+        tracer = DanswerTracer()
+        tracer.start()
+        tracer.snap()
+
+    index_attempt_md = IndexAttemptMetadata(
+        connector_id=db_connector.id,
+        credential_id=db_credential.id,
+    )
+
+    batch_num = 0
     net_doc_change = 0
     document_count = 0
     chunk_count = 0
@@ -175,7 +164,7 @@ def _run_indexing(
                 datetime(1970, 1, 1, tzinfo=timezone.utc),
             )
 
-            doc_batch_generator = _get_document_generator(
+            connector_runner = _get_connector_runner(
                 db_session=db_session,
                 attempt=index_attempt,
                 start_time=window_start,
@@ -183,15 +172,23 @@ def _run_indexing(
             )
 
             all_connector_doc_ids: set[str] = set()
-            for doc_batch in doc_batch_generator:
+
+            tracer_counter = 0
+            if INDEXING_TRACER_INTERVAL > 0:
+                tracer.snap()
+            for doc_batch in connector_runner.run():
                 # Check if connector is disabled mid run and stop if so unless it's the secondary
                 # index being built. We want to populate it even for paused connectors
                 # Often paused connectors are sources that aren't updated frequently but the
                 # contents still need to be initially pulled.
                 db_session.refresh(db_connector)
                 if (
-                    db_cc_pair.status == ConnectorCredentialPairStatus.PAUSED
-                    and db_embedding_model.status != IndexModelStatus.FUTURE
+                    (
+                        db_cc_pair.status == ConnectorCredentialPairStatus.PAUSED
+                        and search_settings.status != IndexModelStatus.FUTURE
+                    )
+                    # if it's deleting, we don't care if this is a secondary index
+                    or db_cc_pair.status == ConnectorCredentialPairStatus.DELETING
                 ):
                     # let the `except` block handle this
                     raise RuntimeError("Connector was disabled mid run")
@@ -218,13 +215,13 @@ def _run_indexing(
 
                 logger.debug(f"Indexing batch of documents: {batch_description}")
 
+                index_attempt_md.batch_num = batch_num + 1  # use 1-index for this
                 new_docs, total_batch_chunks = indexing_pipeline(
                     document_batch=doc_batch,
-                    index_attempt_metadata=IndexAttemptMetadata(
-                        connector_id=db_connector.id,
-                        credential_id=db_credential.id,
-                    ),
+                    index_attempt_metadata=index_attempt_md,
                 )
+
+                batch_num += 1
                 net_doc_change += new_docs
                 chunk_count += total_batch_chunks
                 document_count += len(doc_batch)
@@ -246,6 +243,17 @@ def _run_indexing(
                     docs_removed_from_index=0,
                 )
 
+                tracer_counter += 1
+                if (
+                    INDEXING_TRACER_INTERVAL > 0
+                    and tracer_counter % INDEXING_TRACER_INTERVAL == 0
+                ):
+                    logger.debug(
+                        f"Running trace comparison for batch {tracer_counter}. interval={INDEXING_TRACER_INTERVAL}"
+                    )
+                    tracer.snap()
+                    tracer.log_previous_diff(INDEXING_TRACER_NUM_PRINT_ENTRIES)
+
             run_end_dt = window_end
             if is_primary:
                 update_connector_credential_pair(
@@ -256,7 +264,7 @@ def _run_indexing(
                     run_dt=run_end_dt,
                 )
         except Exception as e:
-            logger.info(
+            logger.exception(
                 f"Connector run ran into exception after elapsed time: {time.time() - start_time} seconds"
             )
             # Only mark the attempt as a complete failure if this is the first indexing window.
@@ -268,7 +276,7 @@ def _run_indexing(
             # to give better clarity in the UI, as the next run will never happen.
             if (
                 ind == 0
-                or db_cc_pair.status == ConnectorCredentialPairStatus.PAUSED
+                or not db_cc_pair.status.is_active()
                 or index_attempt.status != IndexingStatus.IN_PROGRESS
             ):
                 mark_attempt_failed(
@@ -284,13 +292,62 @@ def _run_indexing(
                         credential_id=db_credential.id,
                         net_docs=net_doc_change,
                     )
+
+                if INDEXING_TRACER_INTERVAL > 0:
+                    tracer.stop()
                 raise e
 
             # break => similar to success case. As mentioned above, if the next run fails for the same
             # reason it will then be marked as a failure
             break
 
-    mark_attempt_succeeded(index_attempt, db_session)
+    if INDEXING_TRACER_INTERVAL > 0:
+        logger.debug(
+            f"Running trace comparison between start and end of indexing. {tracer_counter} batches processed."
+        )
+        tracer.snap()
+        tracer.log_first_diff(INDEXING_TRACER_NUM_PRINT_ENTRIES)
+        tracer.stop()
+        logger.debug("Memory tracer stopped.")
+
+    if (
+        index_attempt_md.num_exceptions > 0
+        and index_attempt_md.num_exceptions >= batch_num
+    ):
+        mark_attempt_failed(
+            index_attempt,
+            db_session,
+            failure_reason="All batches exceptioned.",
+        )
+        if is_primary:
+            update_connector_credential_pair(
+                db_session=db_session,
+                connector_id=index_attempt.connector_credential_pair.connector.id,
+                credential_id=index_attempt.connector_credential_pair.credential.id,
+            )
+        raise Exception(
+            f"Connector failed - All batches exceptioned: batches={batch_num}"
+        )
+
+    elapsed_time = time.time() - start_time
+
+    if index_attempt_md.num_exceptions == 0:
+        mark_attempt_succeeded(index_attempt, db_session)
+        logger.info(
+            f"Connector succeeded: "
+            f"docs={document_count} chunks={chunk_count} elapsed={elapsed_time:.2f}s"
+        )
+    else:
+        mark_attempt_partially_succeeded(index_attempt, db_session)
+        logger.info(
+            f"Connector completed with some errors: "
+            f"exceptions={index_attempt_md.num_exceptions} "
+            f"batches={batch_num} "
+            f"docs={document_count} "
+            f"chunks={chunk_count} "
+            f"elapsed={elapsed_time:.2f}s"
+        )
+
     if is_primary:
         update_connector_credential_pair(
             db_session=db_session,
@@ -299,11 +356,6 @@ def _run_indexing(
             run_dt=run_end_dt,
         )
 
-    elapsed_time = time.time() - start_time
-    logger.info(
-        f"Connector succeeded: docs={document_count} chunks={chunk_count} elapsed={elapsed_time:.2f}s"
-    )
-
 
 def _prepare_index_attempt(db_session: Session, index_attempt_id: int) -> IndexAttempt:
     # make sure that the index attempt can't change in between checking the
@@ -332,17 +384,22 @@ def _prepare_index_attempt(db_session: Session, index_attempt_id: int) -> IndexA
     return attempt
 
 
-def run_indexing_entrypoint(index_attempt_id: int, is_ee: bool = False) -> None:
+def run_indexing_entrypoint(
+    index_attempt_id: int, connector_credential_pair_id: int, is_ee: bool = False
+) -> None:
     """Entrypoint for indexing run when using dask distributed.
     Wraps the actual logic in a `try` block so that we can catch any exceptions
     and mark the attempt as failed."""
+
     try:
         if is_ee:
             global_version.set_ee()
 
         # set the indexing attempt ID so that all log messages from this process
         # will have it added as a prefix
-        IndexAttemptSingleton.set_index_attempt_id(index_attempt_id)
+        IndexAttemptSingleton.set_cc_and_index_id(
+            index_attempt_id, connector_credential_pair_id
+        )
 
         with Session(get_sqlalchemy_engine()) as db_session:
             # make sure that it is valid to run this indexing attempt + mark it

backend/danswer/background/indexing/tracer.py

@@ -0,0 +1,77 @@
+import tracemalloc
+
+from danswer.utils.logger import setup_logger
+
+logger = setup_logger()
+
+DANSWER_TRACEMALLOC_FRAMES = 10
+
+
+class DanswerTracer:
+    def __init__(self) -> None:
+        self.snapshot_first: tracemalloc.Snapshot | None = None
+        self.snapshot_prev: tracemalloc.Snapshot | None = None
+        self.snapshot: tracemalloc.Snapshot | None = None
+
+    def start(self) -> None:
+        tracemalloc.start(DANSWER_TRACEMALLOC_FRAMES)
+
+    def stop(self) -> None:
+        tracemalloc.stop()
+
+    def snap(self) -> None:
+        snapshot = tracemalloc.take_snapshot()
+        # Filter out irrelevant frames (e.g., from tracemalloc itself or importlib)
+        snapshot = snapshot.filter_traces(
+            (
+                tracemalloc.Filter(False, tracemalloc.__file__),  # Exclude tracemalloc
+                tracemalloc.Filter(
+                    False, "<frozen importlib._bootstrap>"
+                ),  # Exclude importlib
+                tracemalloc.Filter(
+                    False, "<frozen importlib._bootstrap_external>"
+                ),  # Exclude external importlib
+            )
+        )
+
+        if not self.snapshot_first:
+            self.snapshot_first = snapshot
+
+        if self.snapshot:
+            self.snapshot_prev = self.snapshot
+
+        self.snapshot = snapshot
+
+    def log_snapshot(self, numEntries: int) -> None:
+        if not self.snapshot:
+            return
+
+        stats = self.snapshot.statistics("traceback")
+        for s in stats[:numEntries]:
+            logger.debug(f"Tracer snap: {s}")
+            for line in s.traceback:
+                logger.debug(f"* {line}")
+
+    @staticmethod
+    def log_diff(
+        snap_current: tracemalloc.Snapshot,
+        snap_previous: tracemalloc.Snapshot,
+        numEntries: int,
+    ) -> None:
+        stats = snap_current.compare_to(snap_previous, "traceback")
+        for s in stats[:numEntries]:
+            logger.debug(f"Tracer diff: {s}")
+            for line in s.traceback.format():
+                logger.debug(f"* {line}")
+
+    def log_previous_diff(self, numEntries: int) -> None:
+        if not self.snapshot or not self.snapshot_prev:
+            return
+
+        DanswerTracer.log_diff(self.snapshot, self.snapshot_prev, numEntries)
+
+    def log_first_diff(self, numEntries: int) -> None:
+        if not self.snapshot or not self.snapshot_first:
+            return
+
+        DanswerTracer.log_diff(self.snapshot, self.snapshot_first, numEntries)

backend/danswer/background/task_utils.py

@@ -14,14 +14,6 @@ from danswer.db.tasks import mark_task_start
 from danswer.db.tasks import register_task
 
 
-def name_cc_cleanup_task(connector_id: int, credential_id: int) -> str:
-    return f"cleanup_connector_credential_pair_{connector_id}_{credential_id}"
-
-
-def name_document_set_sync_task(document_set_id: int) -> str:
-    return f"sync_doc_set_{document_set_id}"
-
-
 def name_cc_prune_task(
     connector_id: int | None = None, credential_id: int | None = None
 ) -> str:
@@ -93,9 +85,16 @@ def build_apply_async_wrapper(build_name_fn: Callable[..., str]) -> Callable[[AA
             kwargs_for_build_name = kwargs or {}
             task_name = build_name_fn(*args_for_build_name, **kwargs_for_build_name)
             with Session(get_sqlalchemy_engine()) as db_session:
-                # mark the task as started
+                # register_task must come before fn = apply_async or else the task
+                # might run mark_task_start (and crash) before the task row exists
+                db_task = register_task(task_name, db_session)
+
                 task = fn(args, kwargs, *other_args, **other_kwargs)
-                register_task(task.id, task_name, db_session)
+
+                # we update the celery task id for diagnostic purposes
+                # but it isn't currently used by any code
+                db_task.task_id = task.id
+                db_session.commit()
 
             return task
 

backend/danswer/background/update.py

@@ -17,15 +17,13 @@ from danswer.configs.app_configs import DASK_JOB_CLIENT_ENABLED
 from danswer.configs.app_configs import DISABLE_INDEX_UPDATE_ON_SWAP
 from danswer.configs.app_configs import NUM_INDEXING_WORKERS
 from danswer.configs.app_configs import NUM_SECONDARY_INDEXING_WORKERS
+from danswer.configs.constants import DocumentSource
 from danswer.configs.constants import POSTGRES_INDEXER_APP_NAME
 from danswer.db.connector import fetch_connectors
 from danswer.db.connector_credential_pair import fetch_connector_credential_pairs
-from danswer.db.embedding_model import get_current_db_embedding_model
-from danswer.db.embedding_model import get_secondary_db_embedding_model
 from danswer.db.engine import get_db_current_time
 from danswer.db.engine import get_sqlalchemy_engine
 from danswer.db.engine import init_sqlalchemy_engine
-from danswer.db.enums import ConnectorCredentialPairStatus
 from danswer.db.index_attempt import create_index_attempt
 from danswer.db.index_attempt import get_index_attempt
 from danswer.db.index_attempt import get_inprogress_index_attempts
@@ -33,11 +31,14 @@ from danswer.db.index_attempt import get_last_attempt_for_cc_pair
 from danswer.db.index_attempt import get_not_started_index_attempts
 from danswer.db.index_attempt import mark_attempt_failed
 from danswer.db.models import ConnectorCredentialPair
-from danswer.db.models import EmbeddingModel
 from danswer.db.models import IndexAttempt
 from danswer.db.models import IndexingStatus
 from danswer.db.models import IndexModelStatus
+from danswer.db.models import SearchSettings
+from danswer.db.search_settings import get_current_search_settings
+from danswer.db.search_settings import get_secondary_search_settings
 from danswer.db.swap_index import check_index_swap
+from danswer.natural_language_processing.search_nlp_models import EmbeddingModel
 from danswer.natural_language_processing.search_nlp_models import warm_up_bi_encoder
 from danswer.utils.logger import setup_logger
 from danswer.utils.variable_functionality import global_version
@@ -60,20 +61,27 @@ _UNEXPECTED_STATE_FAILURE_REASON = (
 def _should_create_new_indexing(
     cc_pair: ConnectorCredentialPair,
     last_index: IndexAttempt | None,
-    model: EmbeddingModel,
+    search_settings_instance: SearchSettings,
     secondary_index_building: bool,
     db_session: Session,
 ) -> bool:
     connector = cc_pair.connector
 
+    # don't kick off indexing for `NOT_APPLICABLE` sources
+    if connector.source == DocumentSource.NOT_APPLICABLE:
+        return False
+
     # User can still manually create single indexing attempts via the UI for the
     # currently in use index
     if DISABLE_INDEX_UPDATE_ON_SWAP:
-        if model.status == IndexModelStatus.PRESENT and secondary_index_building:
+        if (
+            search_settings_instance.status == IndexModelStatus.PRESENT
+            and secondary_index_building
+        ):
             return False
 
     # When switching over models, always index at least once
-    if model.status == IndexModelStatus.FUTURE:
+    if search_settings_instance.status == IndexModelStatus.FUTURE:
         if last_index:
             # No new index if the last index attempt succeeded
             # Once is enough. The model will never be able to swap otherwise.
@@ -95,7 +103,7 @@ def _should_create_new_indexing(
     # If the connector is paused or is the ingestion API, don't index
     # NOTE: during an embedding model switch over, the following logic
     # is bypassed by the above check for a future model
-    if cc_pair.status == ConnectorCredentialPairStatus.PAUSED or connector.id == 0:
+    if not cc_pair.status.is_active() or connector.id == 0:
         return False
 
     if not last_index:
@@ -120,16 +128,6 @@ def _should_create_new_indexing(
     return time_since_index.total_seconds() >= connector.refresh_freq
 
 
-def _is_indexing_job_marked_as_finished(index_attempt: IndexAttempt | None) -> bool:
-    if index_attempt is None:
-        return False
-
-    return (
-        index_attempt.status == IndexingStatus.FAILED
-        or index_attempt.status == IndexingStatus.SUCCESS
-    )
-
-
 def _mark_run_failed(
     db_session: Session, index_attempt: IndexAttempt, failure_reason: str
 ) -> None:
@@ -170,35 +168,42 @@ def create_indexing_jobs(existing_jobs: dict[int, Future | SimpleJob]) -> None:
             ongoing.add(
                 (
                     attempt.connector_credential_pair_id,
-                    attempt.embedding_model_id,
+                    attempt.search_settings_id,
                 )
             )
 
-        embedding_models = [get_current_db_embedding_model(db_session)]
-        secondary_embedding_model = get_secondary_db_embedding_model(db_session)
-        if secondary_embedding_model is not None:
-            embedding_models.append(secondary_embedding_model)
+        # Get the primary search settings
+        primary_search_settings = get_current_search_settings(db_session)
+        search_settings = [primary_search_settings]
+
+        # Check for secondary search settings
+        secondary_search_settings = get_secondary_search_settings(db_session)
+        if secondary_search_settings is not None:
+            # If secondary settings exist, add them to the list
+            search_settings.append(secondary_search_settings)
 
         all_connector_credential_pairs = fetch_connector_credential_pairs(db_session)
         for cc_pair in all_connector_credential_pairs:
-            for model in embedding_models:
+            for search_settings_instance in search_settings:
                 # Check if there is an ongoing indexing attempt for this connector credential pair
-                if (cc_pair.id, model.id) in ongoing:
+                if (cc_pair.id, search_settings_instance.id) in ongoing:
                     continue
 
                 last_attempt = get_last_attempt_for_cc_pair(
-                    cc_pair.id, model.id, db_session
+                    cc_pair.id, search_settings_instance.id, db_session
                 )
                 if not _should_create_new_indexing(
                     cc_pair=cc_pair,
                     last_index=last_attempt,
-                    model=model,
-                    secondary_index_building=len(embedding_models) > 1,
+                    search_settings_instance=search_settings_instance,
+                    secondary_index_building=len(search_settings) > 1,
                     db_session=db_session,
                 ):
                     continue
 
-                create_index_attempt(cc_pair.id, model.id, db_session)
+                create_index_attempt(
+                    cc_pair.id, search_settings_instance.id, db_session
+                )
 
 
 def cleanup_indexing_jobs(
@@ -206,7 +211,6 @@ def cleanup_indexing_jobs(
     timeout_hours: int = CLEANUP_INDEXING_JOBS_TIMEOUT,
 ) -> dict[int, Future | SimpleJob]:
     existing_jobs_copy = existing_jobs.copy()
-
     # clean up completed jobs
     with Session(get_sqlalchemy_engine()) as db_session:
         for attempt_id, job in existing_jobs.items():
@@ -215,10 +219,12 @@ def cleanup_indexing_jobs(
             )
 
             # do nothing for ongoing jobs that haven't been stopped
-            if not job.done() and not _is_indexing_job_marked_as_finished(
-                index_attempt
-            ):
-                continue
+            if not job.done():
+                if not index_attempt:
+                    continue
+
+                if not index_attempt.is_finished():
+                    continue
 
             if job.status == "error":
                 logger.error(job.exception())
@@ -293,7 +299,7 @@ def kickoff_indexing_jobs(
         # get_not_started_index_attempts orders its returned results from oldest to newest
         # we must process attempts in a FIFO manner to prevent connector starvation
         new_indexing_attempts = [
-            (attempt, attempt.embedding_model)
+            (attempt, attempt.search_settings)
             for attempt in get_not_started_index_attempts(db_session)
             if attempt.id not in existing_jobs
         ]
@@ -305,10 +311,15 @@ def kickoff_indexing_jobs(
 
     indexing_attempt_count = 0
 
-    for attempt, embedding_model in new_indexing_attempts:
+    primary_client_full = False
+    secondary_client_full = False
+    for attempt, search_settings in new_indexing_attempts:
+        if primary_client_full and secondary_client_full:
+            break
+
         use_secondary_index = (
-            embedding_model.status == IndexModelStatus.FUTURE
-            if embedding_model is not None
+            search_settings.status == IndexModelStatus.FUTURE
+            if search_settings is not None
             else False
         )
         if attempt.connector_credential_pair.connector is None:
@@ -330,20 +341,28 @@ def kickoff_indexing_jobs(
                 )
             continue
 
-        if use_secondary_index:
-            run = secondary_client.submit(
-                run_indexing_entrypoint,
-                attempt.id,
-                global_version.get_is_ee_version(),
-                pure=False,
-            )
+        if not use_secondary_index:
+            if not primary_client_full:
+                run = client.submit(
+                    run_indexing_entrypoint,
+                    attempt.id,
+                    attempt.connector_credential_pair_id,
+                    global_version.get_is_ee_version(),
+                    pure=False,
+                )
+                if not run:
+                    primary_client_full = True
         else:
-            run = client.submit(
-                run_indexing_entrypoint,
-                attempt.id,
-                global_version.get_is_ee_version(),
-                pure=False,
-            )
+            if not secondary_client_full:
+                run = secondary_client.submit(
+                    run_indexing_entrypoint,
+                    attempt.id,
+                    attempt.connector_credential_pair_id,
+                    global_version.get_is_ee_version(),
+                    pure=False,
+                )
+                if not run:
+                    secondary_client_full = True
 
         if run:
             if indexing_attempt_count == 0:
@@ -355,6 +374,7 @@ def kickoff_indexing_jobs(
             secondary_str = " (secondary index)" if use_secondary_index else ""
             logger.info(
                 f"Indexing dispatched{secondary_str}: "
+                f"attempt_id={attempt.id} "
                 f"connector='{attempt.connector_credential_pair.connector.name}' "
                 f"config='{attempt.connector_credential_pair.connector.connector_specific_config}' "
                 f"credentials='{attempt.connector_credential_pair.credential_id}'"
@@ -380,17 +400,21 @@ def update_loop(
     engine = get_sqlalchemy_engine()
     with Session(engine) as db_session:
         check_index_swap(db_session=db_session)
-        db_embedding_model = get_current_db_embedding_model(db_session)
+        search_settings = get_current_search_settings(db_session)
 
         # So that the first time users aren't surprised by really slow speed of first
         # batch of documents indexed
 
-        if db_embedding_model.cloud_provider_id is None:
-            logger.debug("Running a first inference to warm up embedding model")
+        if search_settings.provider_type is None:
+            logger.notice("Running a first inference to warm up embedding model")
+            embedding_model = EmbeddingModel.from_db_model(
+                search_settings=search_settings,
+                server_host=INDEXING_MODEL_SERVER_HOST,
+                server_port=MODEL_SERVER_PORT,
+            )
+
             warm_up_bi_encoder(
-                embedding_model=db_embedding_model,
-                model_server_host=INDEXING_MODEL_SERVER_HOST,
-                model_server_port=MODEL_SERVER_PORT,
+                embedding_model=embedding_model,
             )
 
     client_primary: Client | SimpleJobClient
@@ -453,7 +477,7 @@ def update__main() -> None:
     set_is_ee_based_on_env_variable()
     init_sqlalchemy_engine(POSTGRES_INDEXER_APP_NAME)
 
-    logger.info("Starting indexing service")
+    logger.notice("Starting indexing service")
     update_loop()
 
 

backend/danswer/chat/chat_utils.py

@@ -36,9 +36,12 @@ def create_chat_chain(
     chat_session_id: int,
     db_session: Session,
     prefetch_tool_calls: bool = True,
+    # Optional id at which we finish processing
+    stop_at_message_id: int | None = None,
 ) -> tuple[ChatMessage, list[ChatMessage]]:
     """Build the linear chain of messages without including the root message"""
     mainline_messages: list[ChatMessage] = []
+
     all_chat_messages = get_chat_messages_by_session(
         chat_session_id=chat_session_id,
         user_id=None,
@@ -60,7 +63,12 @@ def create_chat_chain(
     current_message: ChatMessage | None = root_message
     while current_message is not None:
         child_msg = current_message.latest_child_message
-        if not child_msg:
+
+        # Break if at the end of the chain
+        # or have reached the `final_id` of the submitted message
+        if not child_msg or (
+            stop_at_message_id and current_message.id == stop_at_message_id
+        ):
             break
         current_message = id_to_msg.get(child_msg)
 

backend/danswer/chat/load_yamls.py

@@ -122,7 +122,7 @@ def load_personas_from_yaml(
                 prompt_ids=prompt_ids,
                 document_set_ids=doc_set_ids,
                 tool_ids=tool_ids,
-                default_persona=True,
+                builtin_persona=True,
                 is_public=True,
                 display_priority=existing_persona.display_priority
                 if existing_persona is not None

backend/danswer/chat/models.py

@@ -1,5 +1,6 @@
 from collections.abc import Iterator
 from datetime import datetime
+from enum import Enum
 from typing import Any
 
 from pydantic import BaseModel
@@ -9,6 +10,7 @@ from danswer.search.enums import QueryFlow
 from danswer.search.enums import SearchType
 from danswer.search.models import RetrievalDocs
 from danswer.search.models import SearchResponse
+from danswer.tools.custom.base_tool_types import ToolResultType
 
 
 class LlmDoc(BaseModel):
@@ -34,16 +36,35 @@ class QADocsResponse(RetrievalDocs):
     applied_time_cutoff: datetime | None
     recency_bias_multiplier: float
 
-    def dict(self, *args: list, **kwargs: dict[str, Any]) -> dict[str, Any]:  # type: ignore
-        initial_dict = super().dict(*args, **kwargs)  # type: ignore
+    def model_dump(self, *args: list, **kwargs: dict[str, Any]) -> dict[str, Any]:  # type: ignore
+        initial_dict = super().model_dump(mode="json", *args, **kwargs)  # type: ignore
         initial_dict["applied_time_cutoff"] = (
             self.applied_time_cutoff.isoformat() if self.applied_time_cutoff else None
         )
+
         return initial_dict
 
 
+class StreamStopReason(Enum):
+    CONTEXT_LENGTH = "context_length"
+    CANCELLED = "cancelled"
+
+
+class StreamStopInfo(BaseModel):
+    stop_reason: StreamStopReason
+
+    def model_dump(self, *args: list, **kwargs: dict[str, Any]) -> dict[str, Any]:  # type: ignore
+        data = super().model_dump(mode="json", *args, **kwargs)  # type: ignore
+        data["stop_reason"] = self.stop_reason.name
+        return data
+
+
 class LLMRelevanceFilterResponse(BaseModel):
-    relevant_chunk_indices: list[int]
+    llm_selected_doc_indices: list[int]
+
+
+class FinalUsedContextDocsResponse(BaseModel):
+    final_context_docs: list[LlmDoc]
 
 
 class RelevanceAnalysis(BaseModel):
@@ -76,8 +97,24 @@ class CitationInfo(BaseModel):
     document_id: str
 
 
+class AllCitations(BaseModel):
+    citations: list[CitationInfo]
+
+
+# This is a mapping of the citation number to the document index within
+# the result search doc set
+class MessageSpecificCitations(BaseModel):
+    citation_map: dict[int, int]
+
+
+class MessageResponseIDInfo(BaseModel):
+    user_message_id: int | None
+    reserved_assistant_message_id: int
+
+
 class StreamingError(BaseModel):
     error: str
+    stack_trace: str | None = None
 
 
 class DanswerQuote(BaseModel):
@@ -115,7 +152,7 @@ class QAResponse(SearchResponse, DanswerAnswer):
     predicted_flow: QueryFlow
     predicted_search: SearchType
     eval_res_valid: bool | None = None
-    llm_chunks_indices: list[int] | None = None
+    llm_selected_doc_indices: list[int] | None = None
     error_msg: str | None = None
 
 
@@ -124,7 +161,7 @@ class ImageGenerationDisplay(BaseModel):
 
 
 class CustomToolResponse(BaseModel):
-    response: dict
+    response: ToolResultType
     tool_name: str
 
 
@@ -136,6 +173,7 @@ AnswerQuestionPossibleReturn = (
     | ImageGenerationDisplay
     | CustomToolResponse
     | StreamingError
+    | StreamStopInfo
 )
 
 

backend/danswer/chat/process_message.py

@@ -1,3 +1,4 @@
+import traceback
 from collections.abc import Callable
 from collections.abc import Iterator
 from functools import partial
@@ -6,11 +7,15 @@ from typing import cast
 from sqlalchemy.orm import Session
 
 from danswer.chat.chat_utils import create_chat_chain
+from danswer.chat.models import AllCitations
 from danswer.chat.models import CitationInfo
 from danswer.chat.models import CustomToolResponse
 from danswer.chat.models import DanswerAnswerPiece
+from danswer.chat.models import FinalUsedContextDocsResponse
 from danswer.chat.models import ImageGenerationDisplay
 from danswer.chat.models import LLMRelevanceFilterResponse
+from danswer.chat.models import MessageResponseIDInfo
+from danswer.chat.models import MessageSpecificCitations
 from danswer.chat.models import QADocsResponse
 from danswer.chat.models import StreamingError
 from danswer.configs.chat_configs import BING_API_KEY
@@ -27,15 +32,16 @@ from danswer.db.chat import get_chat_session_by_id
 from danswer.db.chat import get_db_search_doc_by_id
 from danswer.db.chat import get_doc_query_identifiers_from_model
 from danswer.db.chat import get_or_create_root_message
+from danswer.db.chat import reserve_message_id
 from danswer.db.chat import translate_db_message_to_chat_message_detail
 from danswer.db.chat import translate_db_search_doc_to_server_search_doc
-from danswer.db.embedding_model import get_current_db_embedding_model
 from danswer.db.engine import get_session_context_manager
 from danswer.db.llm import fetch_existing_llm_providers
 from danswer.db.models import SearchDoc as DbSearchDoc
 from danswer.db.models import ToolCall
 from danswer.db.models import User
 from danswer.db.persona import get_persona_by_id
+from danswer.db.search_settings import get_current_search_settings
 from danswer.document_index.factory import get_default_document_index
 from danswer.file_store.models import ChatFileType
 from danswer.file_store.models import FileDescriptor
@@ -51,6 +57,7 @@ from danswer.llm.exceptions import GenAIDisabledException
 from danswer.llm.factory import get_llms_for_persona
 from danswer.llm.factory import get_main_llm_from_tuple
 from danswer.llm.interfaces import LLMConfig
+from danswer.llm.utils import litellm_exception_to_error_msg
 from danswer.natural_language_processing.utils import get_tokenizer
 from danswer.search.enums import LLMEvaluationType
 from danswer.search.enums import OptionalSearchSetting
@@ -66,7 +73,9 @@ from danswer.server.query_and_chat.models import ChatMessageDetail
 from danswer.server.query_and_chat.models import CreateChatMessageRequest
 from danswer.server.utils import get_json_line
 from danswer.tools.built_in_tools import get_built_in_tool_by_id
-from danswer.tools.custom.custom_tool import build_custom_tools_from_openapi_schema
+from danswer.tools.custom.custom_tool import (
+    build_custom_tools_from_openapi_schema_and_headers,
+)
 from danswer.tools.custom.custom_tool import CUSTOM_TOOL_RESPONSE_ID
 from danswer.tools.custom.custom_tool import CustomToolCallSummary
 from danswer.tools.force import ForceUseTool
@@ -81,6 +90,8 @@ from danswer.tools.internet_search.internet_search_tool import (
 )
 from danswer.tools.internet_search.internet_search_tool import InternetSearchResponse
 from danswer.tools.internet_search.internet_search_tool import InternetSearchTool
+from danswer.tools.models import DynamicSchemaInfo
+from danswer.tools.search.search_tool import FINAL_CONTEXT_DOCUMENTS_ID
 from danswer.tools.search.search_tool import SEARCH_RESPONSE_SUMMARY_ID
 from danswer.tools.search.search_tool import SearchResponseSummary
 from danswer.tools.search.search_tool import SearchTool
@@ -96,9 +107,9 @@ from danswer.utils.timing import log_generator_function_time
 logger = setup_logger()
 
 
-def translate_citations(
+def _translate_citations(
     citations_list: list[CitationInfo], db_docs: list[DbSearchDoc]
-) -> dict[int, int]:
+) -> MessageSpecificCitations:
     """Always cites the first instance of the document_id, assumes the db_docs
     are sorted in the order displayed in the UI"""
     doc_id_to_saved_doc_id_map: dict[str, int] = {}
@@ -113,7 +124,7 @@ def translate_citations(
                 citation.citation_num
             ] = doc_id_to_saved_doc_id_map[citation.document_id]
 
-    return citation_to_saved_doc_id_map
+    return MessageSpecificCitations(citation_map=citation_to_saved_doc_id_map)
 
 
 def _handle_search_tool_response_summary(
@@ -235,11 +246,15 @@ ChatPacket = (
     StreamingError
     | QADocsResponse
     | LLMRelevanceFilterResponse
+    | FinalUsedContextDocsResponse
     | ChatMessageDetail
     | DanswerAnswerPiece
+    | AllCitations
     | CitationInfo
     | ImageGenerationDisplay
     | CustomToolResponse
+    | MessageSpecificCitations
+    | MessageResponseIDInfo
 )
 ChatPacketStream = Iterator[ChatPacket]
 
@@ -255,9 +270,10 @@ def stream_chat_message_objects(
     max_document_percentage: float = CHAT_TARGET_CHUNK_PERCENTAGE,
     # if specified, uses the last user message and does not create a new user message based
     # on the `new_msg_req.message`. Currently, requires a state where the last message is a
-    # user message (e.g. this can only be used for the chat-seeding flow).
     use_existing_user_message: bool = False,
     litellm_additional_headers: dict[str, str] | None = None,
+    is_connected: Callable[[], bool] | None = None,
+    enforce_chat_session_id_for_search_docs: bool = True,
 ) -> ChatPacketStream:
     """Streams in order:
     1. [conditional] Retrieved documents if a search needs to be run
@@ -265,6 +281,11 @@ def stream_chat_message_objects(
     3. [always] A set of streamed LLM tokens or an error anywhere along the line if something fails
     4. [always] Details on the final AI response message that is created
     """
+    # Currently surrounding context is not supported for chat
+    # Chat is already token heavy and harder for the model to process plus it would roll history over much faster
+    new_msg_req.chunks_above = 0
+    new_msg_req.chunks_below = 0
+
     try:
         user_id = user.id if user is not None else None
 
@@ -321,9 +342,9 @@ def stream_chat_message_objects(
             Callable[[str], list[int]], llm_tokenizer.encode
         )
 
-        embedding_model = get_current_db_embedding_model(db_session)
+        search_settings = get_current_search_settings(db_session)
         document_index = get_default_document_index(
-            primary_index_name=embedding_model.index_name, secondary_index_name=None
+            primary_index_name=search_settings.index_name, secondary_index_name=None
         )
 
         # Every chat Session begins with an empty root message
@@ -341,7 +362,15 @@ def stream_chat_message_objects(
             parent_message = root_message
 
         user_message = None
-        if not use_existing_user_message:
+
+        if new_msg_req.regenerate:
+            final_msg, history_msgs = create_chat_chain(
+                stop_at_message_id=parent_id,
+                chat_session_id=chat_session_id,
+                db_session=db_session,
+            )
+
+        elif not use_existing_user_message:
             # Create new message at the right place in the tree and update the parent's child pointer
             # Don't commit yet until we verify the chat message chain
             user_message = create_new_chat_message(
@@ -416,6 +445,7 @@ def stream_chat_message_objects(
                 chat_session=chat_session,
                 user_id=user_id,
                 db_session=db_session,
+                enforce_chat_session_id_for_search_docs=enforce_chat_session_id_for_search_docs,
             )
 
             # Generates full documents currently
@@ -447,9 +477,23 @@ def stream_chat_message_objects(
                     else default_num_chunks
                 ),
                 max_window_percentage=max_document_percentage,
-                use_sections=new_msg_req.chunks_above > 0
-                or new_msg_req.chunks_below > 0,
             )
+        reserved_message_id = reserve_message_id(
+            db_session=db_session,
+            chat_session_id=chat_session_id,
+            parent_message=user_message.id
+            if user_message is not None
+            else parent_message.id,
+            message_type=MessageType.ASSISTANT,
+        )
+        yield MessageResponseIDInfo(
+            user_message_id=user_message.id if user_message else None,
+            reserved_assistant_message_id=reserved_message_id,
+        )
+
+        overridden_model = (
+            new_msg_req.llm_override.model_version if new_msg_req.llm_override else None
+        )
 
         # Cannot determine these without the LLM step or breaking out early
         partial_response = partial(
@@ -457,6 +501,7 @@ def stream_chat_message_objects(
             chat_session_id=chat_session_id,
             parent_message=final_msg,
             prompt_id=prompt_id,
+            overridden_model=overridden_model,
             # message=,
             # rephrased_query=,
             # token_count=,
@@ -564,8 +609,13 @@ def stream_chat_message_objects(
             if db_tool_model.openapi_schema:
                 tool_dict[db_tool_model.id] = cast(
                     list[Tool],
-                    build_custom_tools_from_openapi_schema(
-                        db_tool_model.openapi_schema
+                    build_custom_tools_from_openapi_schema_and_headers(
+                        db_tool_model.openapi_schema,
+                        dynamic_schema_info=DynamicSchemaInfo(
+                            chat_session_id=chat_session_id,
+                            message_id=user_message.id if user_message else None,
+                        ),
+                        custom_headers=db_tool_model.custom_headers,
                     ),
                 )
 
@@ -583,6 +633,7 @@ def stream_chat_message_objects(
 
         # LLM prompt building, response capturing, etc.
         answer = Answer(
+            is_connected=is_connected,
             question=final_msg.message,
             latest_query_files=latest_query_files,
             answer_style_config=AnswerStyleConfig(
@@ -616,6 +667,7 @@ def stream_chat_message_objects(
         ai_message_files = None  # any files to associate with the AI message e.g. dall-e generated images
         dropped_indices = None
         tool_result = None
+
         for packet in answer.processed_streamed_output:
             if isinstance(packet, ToolResponse):
                 if packet.id == SEARCH_RESPONSE_SUMMARY_ID:
@@ -628,9 +680,11 @@ def stream_chat_message_objects(
                         db_session=db_session,
                         selected_search_docs=selected_db_search_docs,
                         # Deduping happens at the last step to avoid harming quality by dropping content early on
-                        dedupe_docs=retrieval_options.dedupe_docs
-                        if retrieval_options
-                        else False,
+                        dedupe_docs=(
+                            retrieval_options.dedupe_docs
+                            if retrieval_options
+                            else False
+                        ),
                     )
                     yield qa_docs_response
                 elif packet.id == SECTION_RELEVANCE_LIST_ID:
@@ -653,9 +707,14 @@ def stream_chat_message_objects(
                             )
 
                         yield LLMRelevanceFilterResponse(
-                            relevant_chunk_indices=llm_indices
+                            llm_selected_doc_indices=llm_indices
                         )
 
+                elif packet.id == FINAL_CONTEXT_DOCUMENTS_ID:
+                    yield FinalUsedContextDocsResponse(
+                        final_context_docs=packet.response
+                    )
+
                 elif packet.id == IMAGE_GENERATION_RESPONSE_ID:
                     img_generation_response = cast(
                         list[ImageGenerationResponse], packet.response
@@ -691,41 +750,38 @@ def stream_chat_message_objects(
                 if isinstance(packet, ToolCallFinalResult):
                     tool_result = packet
                 yield cast(ChatPacket, packet)
+        logger.debug("Reached end of stream")
+    except ValueError as e:
+        logger.exception("Failed to process chat message.")
+
+        error_msg = str(e)
+        yield StreamingError(error=error_msg)
+        db_session.rollback()
+        return
 
     except Exception as e:
+        logger.exception("Failed to process chat message.")
+
         error_msg = str(e)
-        logger.exception(f"Failed to process chat message: {error_msg}")
+        stack_trace = traceback.format_exc()
+        client_error_msg = litellm_exception_to_error_msg(e, llm)
+        if llm.config.api_key and len(llm.config.api_key) > 2:
+            error_msg = error_msg.replace(llm.config.api_key, "[REDACTED_API_KEY]")
+            stack_trace = stack_trace.replace(llm.config.api_key, "[REDACTED_API_KEY]")
 
-        if "Illegal header value b'Bearer  '" in error_msg:
-            error_msg = (
-                f"Authentication error: Invalid or empty API key provided for '{llm.config.model_provider}'. "
-                "Please check your API key configuration."
-            )
-        elif (
-            "Invalid leading whitespace, reserved character(s), or return character(s) in header value"
-            in error_msg
-        ):
-            error_msg = (
-                f"Authentication error: Invalid API key format for '{llm.config.model_provider}'. "
-                "Please ensure your API key does not contain leading/trailing whitespace or invalid characters."
-            )
-        elif llm.config.api_key and llm.config.api_key.lower() in error_msg.lower():
-            error_msg = f"LLM failed to respond. Invalid API key error from '{llm.config.model_provider}'."
-        else:
-            error_msg = "An unexpected error occurred while processing your request. Please try again later."
-
-        yield StreamingError(error=error_msg)
+        yield StreamingError(error=client_error_msg, stack_trace=stack_trace)
         db_session.rollback()
         return
 
     # Post-LLM answer processing
     try:
-        db_citations = None
+        message_specific_citations: MessageSpecificCitations | None = None
         if reference_db_search_docs:
-            db_citations = translate_citations(
+            message_specific_citations = _translate_citations(
                 citations_list=answer.citations,
                 db_docs=reference_db_search_docs,
             )
+            yield AllCitations(citations=answer.citations)
 
         # Saving Gen AI answer and responding with message info
         tool_name_to_tool_id: dict[str, int] = {}
@@ -734,6 +790,7 @@ def stream_chat_message_objects(
                 tool_name_to_tool_id[tool.name] = tool_id
 
         gen_ai_response_message = partial_response(
+            reserved_message_id=reserved_message_id,
             message=answer.llm_answer,
             rephrased_query=(
                 qa_docs_response.rephrased_query if qa_docs_response else None
@@ -741,19 +798,25 @@ def stream_chat_message_objects(
             reference_docs=reference_db_search_docs,
             files=ai_message_files,
             token_count=len(llm_tokenizer_encode_func(answer.llm_answer)),
-            citations=db_citations,
+            citations=message_specific_citations.citation_map
+            if message_specific_citations
+            else None,
             error=None,
-            tool_calls=[
-                ToolCall(
-                    tool_id=tool_name_to_tool_id[tool_result.tool_name],
-                    tool_name=tool_result.tool_name,
-                    tool_arguments=tool_result.tool_args,
-                    tool_result=tool_result.tool_result,
-                )
-            ]
-            if tool_result
-            else [],
+            tool_calls=(
+                [
+                    ToolCall(
+                        tool_id=tool_name_to_tool_id[tool_result.tool_name],
+                        tool_name=tool_result.tool_name,
+                        tool_arguments=tool_result.tool_args,
+                        tool_result=tool_result.tool_result,
+                    )
+                ]
+                if tool_result
+                else []
+            ),
         )
+
+        logger.debug("Committing messages")
         db_session.commit()  # actually save user / assistant message
 
         msg_detail_response = translate_db_message_to_chat_message_detail(
@@ -762,7 +825,8 @@ def stream_chat_message_objects(
 
         yield msg_detail_response
     except Exception as e:
-        logger.exception(e)
+        error_msg = str(e)
+        logger.exception(error_msg)
 
         # Frontend will erase whatever answer and show this instead
         yield StreamingError(error="Failed to parse LLM output")
@@ -774,6 +838,7 @@ def stream_chat_message(
     user: User | None,
     use_existing_user_message: bool = False,
     litellm_additional_headers: dict[str, str] | None = None,
+    is_connected: Callable[[], bool] | None = None,
 ) -> Iterator[str]:
     with get_session_context_manager() as db_session:
         objects = stream_chat_message_objects(
@@ -782,6 +847,7 @@ def stream_chat_message(
             db_session=db_session,
             use_existing_user_message=use_existing_user_message,
             litellm_additional_headers=litellm_additional_headers,
+            is_connected=is_connected,
         )
         for obj in objects:
-            yield get_json_line(obj.dict())
+            yield get_json_line(obj.model_dump())

backend/danswer/chat/tools.py

@@ -1,4 +1,4 @@
-from typing import TypedDict
+from typing_extensions import TypedDict  # noreorder
 
 from pydantic import BaseModel
 

backend/danswer/configs/app_configs.py

@@ -93,6 +93,14 @@ SMTP_USER = os.environ.get("SMTP_USER", "your-email@gmail.com")
 SMTP_PASS = os.environ.get("SMTP_PASS", "your-gmail-password")
 EMAIL_FROM = os.environ.get("EMAIL_FROM") or SMTP_USER
 
+# If set, Danswer will listen to the `expires_at` returned by the identity
+# provider (e.g. Okta, Google, etc.) and force the user to re-authenticate
+# after this time has elapsed. Disabled since by default many auth providers
+# have very short expiry times (e.g. 1 hour) which provide a poor user experience
+TRACK_EXTERNAL_IDP_EXPIRY = (
+    os.environ.get("TRACK_EXTERNAL_IDP_EXPIRY", "").lower() == "true"
+)
+
 
 #####
 # DB Configs
@@ -118,6 +126,7 @@ try:
 except ValueError:
     INDEX_BATCH_SIZE = 16
 
+
 # Below are intended to match the env variables names used by the official postgres docker image
 # https://hub.docker.com/_/postgres
 POSTGRES_USER = os.environ.get("POSTGRES_USER") or "postgres"
@@ -141,6 +150,27 @@ try:
 except ValueError:
     POSTGRES_POOL_RECYCLE = POSTGRES_POOL_RECYCLE_DEFAULT
 
+REDIS_SSL = os.getenv("REDIS_SSL", "").lower() == "true"
+REDIS_HOST = os.environ.get("REDIS_HOST") or "localhost"
+REDIS_PORT = int(os.environ.get("REDIS_PORT", 6379))
+REDIS_PASSWORD = os.environ.get("REDIS_PASSWORD") or ""
+
+# Used for general redis things
+REDIS_DB_NUMBER = int(os.environ.get("REDIS_DB_NUMBER", 0))
+
+# Used by celery as broker and backend
+REDIS_DB_NUMBER_CELERY_RESULT_BACKEND = int(
+    os.environ.get("REDIS_DB_NUMBER_CELERY_RESULT_BACKEND", 14)
+)
+REDIS_DB_NUMBER_CELERY = int(os.environ.get("REDIS_DB_NUMBER_CELERY", 15))  # broker
+
+# https://docs.celeryq.dev/en/stable/userguide/configuration.html#redis-backend-settings
+# should be one of "required", "optional", or "none"
+REDIS_SSL_CERT_REQS = os.getenv("REDIS_SSL_CERT_REQS", "none")
+REDIS_SSL_CA_CERTS = os.getenv("REDIS_SSL_CA_CERTS", "")
+
+CELERY_RESULT_EXPIRES = int(os.environ.get("CELERY_RESULT_EXPIRES", 86400))  # seconds
+
 #####
 # Connector Configs
 #####
@@ -192,8 +222,8 @@ CONFLUENCE_CONNECTOR_LABELS_TO_SKIP = [
 ]
 
 # Avoid to get archived pages
-CONFLUENCE_CONNECTOR_INDEX_ONLY_ACTIVE_PAGES = (
-    os.environ.get("CONFLUENCE_CONNECTOR_INDEX_ONLY_ACTIVE_PAGES", "").lower() == "true"
+CONFLUENCE_CONNECTOR_INDEX_ARCHIVED_PAGES = (
+    os.environ.get("CONFLUENCE_CONNECTOR_INDEX_ARCHIVED_PAGES", "").lower() == "true"
 )
 
 # Save pages labels as Danswer metadata tags
@@ -204,7 +234,12 @@ CONFLUENCE_CONNECTOR_SKIP_LABEL_INDEXING = (
 
 # Attachments exceeding this size will not be retrieved (in bytes)
 CONFLUENCE_CONNECTOR_ATTACHMENT_SIZE_THRESHOLD = int(
-    os.environ.get("CONFLUENCE_CONNECTOR_ATTACHMENT_SIZE_THRESHOLD", 50 * 1024 * 1024)
+    os.environ.get("CONFLUENCE_CONNECTOR_ATTACHMENT_SIZE_THRESHOLD", 10 * 1024 * 1024)
+)
+# Attachments with more chars than this will not be indexed. This is to prevent extremely
+# large files from freezing indexing. 200,000 is ~100 google doc pages.
+CONFLUENCE_CONNECTOR_ATTACHMENT_CHAR_COUNT_THRESHOLD = int(
+    os.environ.get("CONFLUENCE_CONNECTOR_ATTACHMENT_CHAR_COUNT_THRESHOLD", 200_000)
 )
 
 JIRA_CONNECTOR_LABELS_TO_SKIP = [
@@ -276,6 +311,10 @@ ENABLE_MULTIPASS_INDEXING = (
 # Slightly larger since the sentence aware split is a max cutoff so most minichunks will be under MINI_CHUNK_SIZE
 # tokens. But we need it to be at least as big as 1/4th chunk size to avoid having a tiny mini-chunk at the end
 MINI_CHUNK_SIZE = 150
+
+# This is the number of regular chunks per large chunk
+LARGE_CHUNK_RATIO = 4
+
 # Include the document level metadata in each chunk. If the metadata is too long, then it is thrown out
 # We don't want the metadata to overwhelm the actual contents of the chunk
 SKIP_METADATA_IN_CHUNK = os.environ.get("SKIP_METADATA_IN_CHUNK", "").lower() == "true"
@@ -287,6 +326,14 @@ INDEXING_SIZE_WARNING_THRESHOLD = int(
     os.environ.get("INDEXING_SIZE_WARNING_THRESHOLD", 100 * 1024 * 1024)
 )
 
+# during indexing, will log verbose memory diff stats every x batches and at the end.
+# 0 disables this behavior and is the default.
+INDEXING_TRACER_INTERVAL = int(os.environ.get("INDEXING_TRACER_INTERVAL", 0))
+
+# During an indexing attempt, specifies the number of batches which are allowed to
+# exception without aborting the attempt.
+INDEXING_EXCEPTION_LIMIT = int(os.environ.get("INDEXING_EXCEPTION_LIMIT", 0))
+
 #####
 # Miscellaneous
 #####
@@ -313,6 +360,10 @@ LOG_VESPA_TIMING_INFORMATION = (
     os.environ.get("LOG_VESPA_TIMING_INFORMATION", "").lower() == "true"
 )
 LOG_ENDPOINT_LATENCY = os.environ.get("LOG_ENDPOINT_LATENCY", "").lower() == "true"
+LOG_POSTGRES_LATENCY = os.environ.get("LOG_POSTGRES_LATENCY", "").lower() == "true"
+LOG_POSTGRES_CONN_COUNTS = (
+    os.environ.get("LOG_POSTGRES_CONN_COUNTS", "").lower() == "true"
+)
 # Anonymous usage telemetry
 DISABLE_TELEMETRY = os.environ.get("DISABLE_TELEMETRY", "").lower() == "true"
 

backend/danswer/configs/chat_configs.py

@@ -31,8 +31,9 @@ FAVOR_RECENT_DECAY_MULTIPLIER = 2.0
 DISABLE_LLM_QUERY_ANSWERABILITY = QA_PROMPT_OVERRIDE == "weak"
 # For the highest matching base size chunk, how many chunks above and below do we pull in by default
 # Note this is not in any of the deployment configs yet
-CONTEXT_CHUNKS_ABOVE = int(os.environ.get("CONTEXT_CHUNKS_ABOVE") or 0)
-CONTEXT_CHUNKS_BELOW = int(os.environ.get("CONTEXT_CHUNKS_BELOW") or 0)
+# Currently only applies to search flow not chat
+CONTEXT_CHUNKS_ABOVE = int(os.environ.get("CONTEXT_CHUNKS_ABOVE") or 1)
+CONTEXT_CHUNKS_BELOW = int(os.environ.get("CONTEXT_CHUNKS_BELOW") or 1)
 # Whether the LLM should be used to decide if a search would help given the chat history
 DISABLE_LLM_CHOOSE_SEARCH = (
     os.environ.get("DISABLE_LLM_CHOOSE_SEARCH", "").lower() == "true"
@@ -44,7 +45,7 @@ DISABLE_LLM_QUERY_REPHRASE = (
 QUOTE_ALLOWED_ERROR_PERCENT = 0.05
 QA_TIMEOUT = int(os.environ.get("QA_TIMEOUT") or "60")  # 60 seconds
 # Weighting factor between Vector and Keyword Search, 1 for completely vector search
-HYBRID_ALPHA = max(0, min(1, float(os.environ.get("HYBRID_ALPHA") or 0.62)))
+HYBRID_ALPHA = max(0, min(1, float(os.environ.get("HYBRID_ALPHA") or 0.5)))
 HYBRID_ALPHA_KEYWORD = max(
     0, min(1, float(os.environ.get("HYBRID_ALPHA_KEYWORD") or 0.4))
 )
@@ -53,7 +54,7 @@ HYBRID_ALPHA_KEYWORD = max(
 # Content. This is to avoid cases where the Content is very relevant but it may not be clear
 # if the title is separated out. Title is most of a "boost" than a separate field.
 TITLE_CONTENT_RATIO = max(
-    0, min(1, float(os.environ.get("TITLE_CONTENT_RATIO") or 0.20))
+    0, min(1, float(os.environ.get("TITLE_CONTENT_RATIO") or 0.10))
 )
 
 # A list of languages passed to the LLM to rephase the query
@@ -82,8 +83,15 @@ DISABLE_LLM_DOC_RELEVANCE = (
 # Stops streaming answers back to the UI if this pattern is seen:
 STOP_STREAM_PAT = os.environ.get("STOP_STREAM_PAT") or None
 
-# The backend logic for this being True isn't fully supported yet
-HARD_DELETE_CHATS = False
+# Set this to "true" to hard delete chats
+# This will make chats unviewable by admins after a user deletes them
+# As opposed to soft deleting them, which just hides them from non-admin users
+HARD_DELETE_CHATS = os.environ.get("HARD_DELETE_CHATS", "").lower() == "true"
 
 # Internet Search
 BING_API_KEY = os.environ.get("BING_API_KEY") or None
+
+# Enable in-house model for detecting connector-based filtering in queries
+ENABLE_CONNECTOR_CLASSIFIER = os.environ.get("ENABLE_CONNECTOR_CLASSIFIER", False)
+
+VESPA_SEARCHER_THREADS = int(os.environ.get("VESPA_SEARCHER_THREADS") or 2)

backend/danswer/configs/constants.py

@@ -1,26 +1,7 @@
+from enum import auto
 from enum import Enum
 
-DOCUMENT_ID = "document_id"
-CHUNK_ID = "chunk_id"
-BLURB = "blurb"
-CONTENT = "content"
 SOURCE_TYPE = "source_type"
-SOURCE_LINKS = "source_links"
-SOURCE_LINK = "link"
-SEMANTIC_IDENTIFIER = "semantic_identifier"
-TITLE = "title"
-SKIP_TITLE_EMBEDDING = "skip_title"
-SECTION_CONTINUATION = "section_continuation"
-EMBEDDINGS = "embeddings"
-TITLE_EMBEDDING = "title_embedding"
-ALLOWED_USERS = "allowed_users"
-ACCESS_CONTROL_LIST = "access_control_list"
-DOCUMENT_SETS = "document_sets"
-TIME_FILTER = "time_filter"
-METADATA = "metadata"
-METADATA_LIST = "metadata_list"
-METADATA_SUFFIX = "metadata_suffix"
-MATCH_HIGHLIGHTS = "match_highlights"
 # stored in the `metadata` of a chunk. Used to signify that this chunk should
 # not be used for QA. For example, Google Drive file types which can't be parsed
 # are still useful as a search result but not for QA.
@@ -28,20 +9,9 @@ IGNORE_FOR_QA = "ignore_for_qa"
 # NOTE: deprecated, only used for porting key from old system
 GEN_AI_API_KEY_STORAGE_KEY = "genai_api_key"
 PUBLIC_DOC_PAT = "PUBLIC"
-PUBLIC_DOCUMENT_SET = "__PUBLIC"
-QUOTE = "quote"
-BOOST = "boost"
-DOC_UPDATED_AT = "doc_updated_at"  # Indexed as seconds since epoch
-PRIMARY_OWNERS = "primary_owners"
-SECONDARY_OWNERS = "secondary_owners"
-RECENCY_BIAS = "recency_bias"
-HIDDEN = "hidden"
-SCORE = "score"
 ID_SEPARATOR = ":;:"
 DEFAULT_BOOST = 0
 SESSION_KEY = "session"
-QUERY_EVENT_ID = "query_event_id"
-LLM_CHUNKS = "llm_chunks"
 
 # For chunking/processing chunks
 RETURN_SEPARATOR = "\n\r\n"
@@ -87,9 +57,12 @@ KV_SLACK_BOT_TOKENS_CONFIG_KEY = "slack_bot_tokens_config_key"
 KV_GEN_AI_KEY_CHECK_TIME = "genai_api_key_last_check_time"
 KV_SETTINGS_KEY = "danswer_settings"
 KV_CUSTOMER_UUID_KEY = "customer_uuid"
+KV_INSTANCE_DOMAIN_KEY = "instance_domain"
 KV_ENTERPRISE_SETTINGS_KEY = "danswer_enterprise_settings"
 KV_CUSTOM_ANALYTICS_SCRIPT_KEY = "__custom_analytics_script__"
 
+CELERY_VESPA_SYNC_BEAT_LOCK_TIMEOUT = 60
+
 
 class DocumentSource(str, Enum):
     # Special case, document passed in via Danswer APIs without specifying a source type
@@ -126,6 +99,7 @@ class DocumentSource(str, Enum):
     CLICKUP = "clickup"
     MEDIAWIKI = "mediawiki"
     WIKIPEDIA = "wikipedia"
+    ASANA = "asana"
     S3 = "s3"
     R2 = "r2"
     GOOGLE_CLOUD_STORAGE = "google_cloud_storage"
@@ -160,6 +134,12 @@ class AuthType(str, Enum):
     SAML = "saml"
 
 
+class SessionType(str, Enum):
+    CHAT = "Chat"
+    SEARCH = "Search"
+    SLACK = "Slack"
+
+
 class QAFeedbackType(str, Enum):
     LIKE = "like"  # User likes the answer, used for metrics
     DISLIKE = "dislike"  # User dislikes the answer, used for metrics
@@ -192,3 +172,29 @@ class FileOrigin(str, Enum):
     CONNECTOR = "connector"
     GENERATED_REPORT = "generated_report"
     OTHER = "other"
+
+
+class PostgresAdvisoryLocks(Enum):
+    KOMBU_MESSAGE_CLEANUP_LOCK_ID = auto()
+
+
+class DanswerCeleryQueues:
+    VESPA_DOCSET_SYNC_GENERATOR = "vespa_docset_sync_generator"
+    VESPA_USERGROUP_SYNC_GENERATOR = "vespa_usergroup_sync_generator"
+    VESPA_METADATA_SYNC = "vespa_metadata_sync"
+    CONNECTOR_DELETION = "connector_deletion"
+
+
+class DanswerRedisLocks:
+    CHECK_VESPA_SYNC_BEAT_LOCK = "da_lock:check_vespa_sync_beat"
+    MONITOR_VESPA_SYNC_BEAT_LOCK = "da_lock:monitor_vespa_sync_beat"
+    CHECK_CONNECTOR_DELETION_BEAT_LOCK = "da_lock:check_connector_deletion_beat"
+    MONITOR_CONNECTOR_DELETION_BEAT_LOCK = "da_lock:monitor_connector_deletion_beat"
+
+
+class DanswerCeleryPriority(int, Enum):
+    HIGHEST = 0
+    HIGH = auto()
+    MEDIUM = auto()
+    LOW = auto()
+    LOWEST = auto()

backend/danswer/configs/danswerbot_configs.py

@@ -73,3 +73,15 @@ DANSWER_BOT_FEEDBACK_REMINDER = int(
 DANSWER_BOT_REPHRASE_MESSAGE = (
     os.environ.get("DANSWER_BOT_REPHRASE_MESSAGE", "").lower() == "true"
 )
+
+# DANSWER_BOT_RESPONSE_LIMIT_PER_TIME_PERIOD is the number of
+# responses DanswerBot can send in a given time period.
+# Set to 0 to disable the limit.
+DANSWER_BOT_RESPONSE_LIMIT_PER_TIME_PERIOD = int(
+    os.environ.get("DANSWER_BOT_RESPONSE_LIMIT_PER_TIME_PERIOD", "5000")
+)
+# DANSWER_BOT_RESPONSE_LIMIT_TIME_PERIOD_SECONDS is the number
+# of seconds until the response limit is reset.
+DANSWER_BOT_RESPONSE_LIMIT_TIME_PERIOD_SECONDS = int(
+    os.environ.get("DANSWER_BOT_RESPONSE_LIMIT_TIME_PERIOD_SECONDS", "86400")
+)

backend/danswer/configs/model_configs.py

@@ -19,6 +19,8 @@ DOCUMENT_ENCODER_MODEL = (
 # If the below is changed, Vespa deployment must also be changed
 DOC_EMBEDDING_DIM = int(os.environ.get("DOC_EMBEDDING_DIM") or 768)
 # Model should be chosen with 512 context size, ideally don't change this
+# If multipass_indexing is enabled, the max context size would be set to
+# DOC_EMBEDDING_CONTEXT_SIZE * LARGE_CHUNK_RATIO
 DOC_EMBEDDING_CONTEXT_SIZE = 512
 NORMALIZE_EMBEDDINGS = (
     os.environ.get("NORMALIZE_EMBEDDINGS") or "true"
@@ -37,9 +39,13 @@ SIM_SCORE_RANGE_HIGH = float(os.environ.get("SIM_SCORE_RANGE_HIGH") or 1.0)
 ASYM_QUERY_PREFIX = os.environ.get("ASYM_QUERY_PREFIX", "search_query: ")
 ASYM_PASSAGE_PREFIX = os.environ.get("ASYM_PASSAGE_PREFIX", "search_document: ")
 # Purely an optimization, memory limitation consideration
-BATCH_SIZE_ENCODE_CHUNKS = 8
+
+# User's set embedding batch size overrides the default encoding batch sizes
+EMBEDDING_BATCH_SIZE = int(os.environ.get("EMBEDDING_BATCH_SIZE") or 0) or None
+
+BATCH_SIZE_ENCODE_CHUNKS = EMBEDDING_BATCH_SIZE or 8
 # don't send over too many chunks at once, as sending too many could cause timeouts
-BATCH_SIZE_ENCODE_CHUNKS_FOR_API_EMBEDDING_SERVICES = 512
+BATCH_SIZE_ENCODE_CHUNKS_FOR_API_EMBEDDING_SERVICES = EMBEDDING_BATCH_SIZE or 512
 # For score display purposes, only way is to know the expected ranges
 CROSS_ENCODER_RANGE_MAX = 1
 CROSS_ENCODER_RANGE_MIN = 0
@@ -49,37 +55,23 @@ CROSS_ENCODER_RANGE_MIN = 0
 # Generative AI Model Configs
 #####
 
-# If changing GEN_AI_MODEL_PROVIDER or GEN_AI_MODEL_VERSION from the default,
-# be sure to use one that is LiteLLM compatible:
-# https://litellm.vercel.app/docs/providers/azure#completion---using-env-variables
-# The provider is the prefix before / in the model argument
-
-# Additionally Danswer supports GPT4All and custom request library based models
-# Set GEN_AI_MODEL_PROVIDER to "custom" to use the custom requests approach
-# Set GEN_AI_MODEL_PROVIDER to "gpt4all" to use gpt4all models running locally
-GEN_AI_MODEL_PROVIDER = os.environ.get("GEN_AI_MODEL_PROVIDER") or "openai"
-# If using Azure, it's the engine name, for example: Danswer
+# NOTE: the 3 below should only be used for dev.
+GEN_AI_API_KEY = os.environ.get("GEN_AI_API_KEY")
 GEN_AI_MODEL_VERSION = os.environ.get("GEN_AI_MODEL_VERSION")
-
-# For secondary flows like extracting filters or deciding if a chunk is useful, we don't need
-# as powerful of a model as say GPT-4 so we can use an alternative that is faster and cheaper
 FAST_GEN_AI_MODEL_VERSION = os.environ.get("FAST_GEN_AI_MODEL_VERSION")
 
-# If the Generative AI model requires an API key for access, otherwise can leave blank
-GEN_AI_API_KEY = (
-    os.environ.get("GEN_AI_API_KEY", os.environ.get("OPENAI_API_KEY")) or None
-)
-
-# API Base, such as (for Azure): https://danswer.openai.azure.com/
-GEN_AI_API_ENDPOINT = os.environ.get("GEN_AI_API_ENDPOINT") or None
-# API Version, such as (for Azure): 2023-09-15-preview
-GEN_AI_API_VERSION = os.environ.get("GEN_AI_API_VERSION") or None
-# LiteLLM custom_llm_provider
-GEN_AI_LLM_PROVIDER_TYPE = os.environ.get("GEN_AI_LLM_PROVIDER_TYPE") or None
 # Override the auto-detection of LLM max context length
 GEN_AI_MAX_TOKENS = int(os.environ.get("GEN_AI_MAX_TOKENS") or 0) or None
+
 # Set this to be enough for an answer + quotes. Also used for Chat
-GEN_AI_MAX_OUTPUT_TOKENS = int(os.environ.get("GEN_AI_MAX_OUTPUT_TOKENS") or 1024)
+# This is the minimum token context we will leave for the LLM to generate an answer
+GEN_AI_NUM_RESERVED_OUTPUT_TOKENS = int(
+    os.environ.get("GEN_AI_NUM_RESERVED_OUTPUT_TOKENS") or 1024
+)
+
+# Typically, GenAI models nowadays are at least 4K tokens
+GEN_AI_MODEL_FALLBACK_MAX_TOKENS = 4096
+
 # Number of tokens from chat history to include at maximum
 # 3000 should be enough context regardless of use, no need to include as much as possible
 # as this drives up the cost unnecessarily

backend/danswer/connectors/README.md

@@ -59,6 +59,8 @@ if __name__ == "__main__":
     latest_docs = test_connector.poll_source(one_day_ago, current)
 ```
 
+> Note: Be sure to set PYTHONPATH to danswer/backend before running the above main.
+
 
 ### Additional Required Changes:
 #### Backend Changes
@@ -68,17 +70,16 @@ if __name__ == "__main__":
 [here](https://github.com/danswer-ai/danswer/blob/main/backend/danswer/connectors/factory.py#L33)
 
 #### Frontend Changes
-- Create the new connector directory and admin page under `danswer/web/src/app/admin/connectors/`
-- Create the new icon, type, source, and filter changes
-(refer to existing [PR](https://github.com/danswer-ai/danswer/pull/139))
+- Add the new Connector definition to the `SOURCE_METADATA_MAP` [here](https://github.com/danswer-ai/danswer/blob/main/web/src/lib/sources.ts#L59).
+- Add the definition for the new Form to the `connectorConfigs` object [here](https://github.com/danswer-ai/danswer/blob/main/web/src/lib/connectors/connectors.ts#L79). 
 
 #### Docs Changes
 Create the new connector page (with guiding images!) with how to get the connector credentials and how to set up the
-connector in Danswer. Then create a Pull Request in https://github.com/danswer-ai/danswer-docs
-
+connector in Danswer. Then create a Pull Request in https://github.com/danswer-ai/danswer-docs.
 
 ### Before opening PR
 1. Be sure to fully test changes end to end with setting up the connector and updating the index with new docs from the
-new connector.
-2. Be sure to run the linting/formatting, refer to the formatting and linting section in
+new connector. To make it easier to review, please attach a video showing the successful creation of the connector via the UI (starting from the `Add Connector` page).
+2. Add a folder + tests under `backend/tests/daily/connectors` director. For an example, checkout the [test for Confluence](https://github.com/danswer-ai/danswer/blob/main/backend/tests/daily/connectors/confluence/test_confluence_basic.py). In the PR description, include a guide on how to setup the new source to pass the test. Before merging, we will re-create the environment and make sure the test(s) pass.  
+3. Be sure to run the linting/formatting, refer to the formatting and linting section in
 [CONTRIBUTING.md](https://github.com/danswer-ai/danswer/blob/main/CONTRIBUTING.md#formatting-and-linting)

backend/danswer/connectors/asana/asana_api.py

@@ -0,0 +1,233 @@
+import time
+from collections.abc import Iterator
+from datetime import datetime
+from typing import Dict
+
+import asana  # type: ignore
+
+from danswer.utils.logger import setup_logger
+
+logger = setup_logger()
+
+
+# https://github.com/Asana/python-asana/tree/master?tab=readme-ov-file#documentation-for-api-endpoints
+class AsanaTask:
+    def __init__(
+        self,
+        id: str,
+        title: str,
+        text: str,
+        link: str,
+        last_modified: datetime,
+        project_gid: str,
+        project_name: str,
+    ) -> None:
+        self.id = id
+        self.title = title
+        self.text = text
+        self.link = link
+        self.last_modified = last_modified
+        self.project_gid = project_gid
+        self.project_name = project_name
+
+    def __str__(self) -> str:
+        return f"ID: {self.id}\nTitle: {self.title}\nLast modified: {self.last_modified}\nText: {self.text}"
+
+
+class AsanaAPI:
+    def __init__(
+        self, api_token: str, workspace_gid: str, team_gid: str | None
+    ) -> None:
+        self._user = None  # type: ignore
+        self.workspace_gid = workspace_gid
+        self.team_gid = team_gid
+
+        self.configuration = asana.Configuration()
+        self.api_client = asana.ApiClient(self.configuration)
+        self.tasks_api = asana.TasksApi(self.api_client)
+        self.stories_api = asana.StoriesApi(self.api_client)
+        self.users_api = asana.UsersApi(self.api_client)
+        self.project_api = asana.ProjectsApi(self.api_client)
+        self.workspaces_api = asana.WorkspacesApi(self.api_client)
+
+        self.api_error_count = 0
+        self.configuration.access_token = api_token
+        self.task_count = 0
+
+    def get_tasks(
+        self, project_gids: list[str] | None, start_date: str
+    ) -> Iterator[AsanaTask]:
+        """Get all tasks from the projects with the given gids that were modified since the given date.
+        If project_gids is None, get all tasks from all projects in the workspace."""
+        logger.info("Starting to fetch Asana projects")
+        projects = self.project_api.get_projects(
+            opts={
+                "workspace": self.workspace_gid,
+                "opt_fields": "gid,name,archived,modified_at",
+            }
+        )
+        start_seconds = int(time.mktime(datetime.now().timetuple()))
+        projects_list = []
+        project_count = 0
+        for project_info in projects:
+            project_gid = project_info["gid"]
+            if project_gids is None or project_gid in project_gids:
+                projects_list.append(project_gid)
+            else:
+                logger.debug(
+                    f"Skipping project: {project_gid} - not in accepted project_gids"
+                )
+            project_count += 1
+            if project_count % 100 == 0:
+                logger.info(f"Processed {project_count} projects")
+
+        logger.info(f"Found {len(projects_list)} projects to process")
+        for project_gid in projects_list:
+            for task in self._get_tasks_for_project(
+                project_gid, start_date, start_seconds
+            ):
+                yield task
+        logger.info(f"Completed fetching {self.task_count} tasks from Asana")
+        if self.api_error_count > 0:
+            logger.warning(
+                f"Encountered {self.api_error_count} API errors during task fetching"
+            )
+
+    def _get_tasks_for_project(
+        self, project_gid: str, start_date: str, start_seconds: int
+    ) -> Iterator[AsanaTask]:
+        project = self.project_api.get_project(project_gid, opts={})
+        if project["archived"]:
+            logger.info(f"Skipping archived project: {project['name']} ({project_gid})")
+            return []
+        if not project["team"] or not project["team"]["gid"]:
+            logger.info(
+                f"Skipping project without a team: {project['name']} ({project_gid})"
+            )
+            return []
+        if project["privacy_setting"] == "private":
+            if self.team_gid and project["team"]["gid"] != self.team_gid:
+                logger.info(
+                    f"Skipping private project not in configured team: {project['name']} ({project_gid})"
+                )
+                return []
+            else:
+                logger.info(
+                    f"Processing private project in configured team: {project['name']} ({project_gid})"
+                )
+
+        simple_start_date = start_date.split(".")[0].split("+")[0]
+        logger.info(
+            f"Fetching tasks modified since {simple_start_date} for project: {project['name']} ({project_gid})"
+        )
+
+        opts = {
+            "opt_fields": "name,memberships,memberships.project,completed_at,completed_by,created_at,"
+            "created_by,custom_fields,dependencies,due_at,due_on,external,html_notes,liked,likes,"
+            "modified_at,notes,num_hearts,parent,projects,resource_subtype,resource_type,start_on,"
+            "workspace,permalink_url",
+            "modified_since": start_date,
+        }
+        tasks_from_api = self.tasks_api.get_tasks_for_project(project_gid, opts)
+        for data in tasks_from_api:
+            self.task_count += 1
+            if self.task_count % 10 == 0:
+                end_seconds = time.mktime(datetime.now().timetuple())
+                runtime_seconds = end_seconds - start_seconds
+                if runtime_seconds > 0:
+                    logger.info(
+                        f"Processed {self.task_count} tasks in {runtime_seconds:.0f} seconds "
+                        f"({self.task_count / runtime_seconds:.2f} tasks/second)"
+                    )
+
+            logger.debug(f"Processing Asana task: {data['name']}")
+
+            text = self._construct_task_text(data)
+
+            try:
+                text += self._fetch_and_add_comments(data["gid"])
+
+                last_modified_date = self.format_date(data["modified_at"])
+                text += f"Last modified: {last_modified_date}\n"
+
+                task = AsanaTask(
+                    id=data["gid"],
+                    title=data["name"],
+                    text=text,
+                    link=data["permalink_url"],
+                    last_modified=datetime.fromisoformat(data["modified_at"]),
+                    project_gid=project_gid,
+                    project_name=project["name"],
+                )
+                yield task
+            except Exception:
+                logger.error(
+                    f"Error processing task {data['gid']} in project {project_gid}",
+                    exc_info=True,
+                )
+                self.api_error_count += 1
+
+    def _construct_task_text(self, data: Dict) -> str:
+        text = f"{data['name']}\n\n"
+
+        if data["notes"]:
+            text += f"{data['notes']}\n\n"
+
+        if data["created_by"] and data["created_by"]["gid"]:
+            creator = self.get_user(data["created_by"]["gid"])["name"]
+            created_date = self.format_date(data["created_at"])
+            text += f"Created by: {creator} on {created_date}\n"
+
+        if data["due_on"]:
+            due_date = self.format_date(data["due_on"])
+            text += f"Due date: {due_date}\n"
+
+        if data["completed_at"]:
+            completed_date = self.format_date(data["completed_at"])
+            text += f"Completed on: {completed_date}\n"
+
+        text += "\n"
+        return text
+
+    def _fetch_and_add_comments(self, task_gid: str) -> str:
+        text = ""
+        stories_opts: Dict[str, str] = {}
+        story_start = time.time()
+        stories = self.stories_api.get_stories_for_task(task_gid, stories_opts)
+
+        story_count = 0
+        comment_count = 0
+
+        for story in stories:
+            story_count += 1
+            if story["resource_subtype"] == "comment_added":
+                comment = self.stories_api.get_story(
+                    story["gid"], opts={"opt_fields": "text,created_by,created_at"}
+                )
+                commenter = self.get_user(comment["created_by"]["gid"])["name"]
+                text += f"Comment by {commenter}: {comment['text']}\n\n"
+                comment_count += 1
+
+        story_duration = time.time() - story_start
+        logger.debug(
+            f"Processed {story_count} stories (including {comment_count} comments) in {story_duration:.2f} seconds"
+        )
+
+        return text
+
+    def get_user(self, user_gid: str) -> Dict:
+        if self._user is not None:
+            return self._user
+        self._user = self.users_api.get_user(user_gid, {"opt_fields": "name,email"})
+
+        if not self._user:
+            logger.warning(f"Unable to fetch user information for user_gid: {user_gid}")
+            return {"name": "Unknown"}
+        return self._user
+
+    def format_date(self, date_str: str) -> str:
+        date = datetime.fromisoformat(date_str)
+        return time.strftime("%Y-%m-%d", date.timetuple())
+
+    def get_time(self) -> str:
+        return time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())

backend/danswer/connectors/asana/connector.py

@@ -0,0 +1,120 @@
+import datetime
+from typing import Any
+
+from danswer.configs.app_configs import CONTINUE_ON_CONNECTOR_FAILURE
+from danswer.configs.app_configs import INDEX_BATCH_SIZE
+from danswer.configs.constants import DocumentSource
+from danswer.connectors.asana import asana_api
+from danswer.connectors.interfaces import GenerateDocumentsOutput
+from danswer.connectors.interfaces import LoadConnector
+from danswer.connectors.interfaces import PollConnector
+from danswer.connectors.interfaces import SecondsSinceUnixEpoch
+from danswer.connectors.models import Document
+from danswer.connectors.models import Section
+from danswer.utils.logger import setup_logger
+
+logger = setup_logger()
+
+
+class AsanaConnector(LoadConnector, PollConnector):
+    def __init__(
+        self,
+        asana_workspace_id: str,
+        asana_project_ids: str | None = None,
+        asana_team_id: str | None = None,
+        batch_size: int = INDEX_BATCH_SIZE,
+        continue_on_failure: bool = CONTINUE_ON_CONNECTOR_FAILURE,
+    ) -> None:
+        self.workspace_id = asana_workspace_id
+        self.project_ids_to_index: list[str] | None = (
+            asana_project_ids.split(",") if asana_project_ids is not None else None
+        )
+        self.asana_team_id = asana_team_id
+        self.batch_size = batch_size
+        self.continue_on_failure = continue_on_failure
+        logger.info(
+            f"AsanaConnector initialized with workspace_id: {asana_workspace_id}"
+        )
+
+    def load_credentials(self, credentials: dict[str, Any]) -> dict[str, Any] | None:
+        self.api_token = credentials["asana_api_token_secret"]
+        self.asana_client = asana_api.AsanaAPI(
+            api_token=self.api_token,
+            workspace_gid=self.workspace_id,
+            team_gid=self.asana_team_id,
+        )
+        logger.info("Asana credentials loaded and API client initialized")
+        return None
+
+    def poll_source(
+        self, start: SecondsSinceUnixEpoch, end: SecondsSinceUnixEpoch | None
+    ) -> GenerateDocumentsOutput:
+        start_time = datetime.datetime.fromtimestamp(start).isoformat()
+        logger.info(f"Starting Asana poll from {start_time}")
+        asana = asana_api.AsanaAPI(
+            api_token=self.api_token,
+            workspace_gid=self.workspace_id,
+            team_gid=self.asana_team_id,
+        )
+        docs_batch: list[Document] = []
+        tasks = asana.get_tasks(self.project_ids_to_index, start_time)
+
+        for task in tasks:
+            doc = self._message_to_doc(task)
+            docs_batch.append(doc)
+
+            if len(docs_batch) >= self.batch_size:
+                logger.info(f"Yielding batch of {len(docs_batch)} documents")
+                yield docs_batch
+                docs_batch = []
+
+        if docs_batch:
+            logger.info(f"Yielding final batch of {len(docs_batch)} documents")
+            yield docs_batch
+
+        logger.info("Asana poll completed")
+
+    def load_from_state(self) -> GenerateDocumentsOutput:
+        logger.notice("Starting full index of all Asana tasks")
+        return self.poll_source(start=0, end=None)
+
+    def _message_to_doc(self, task: asana_api.AsanaTask) -> Document:
+        logger.debug(f"Converting Asana task {task.id} to Document")
+        return Document(
+            id=task.id,
+            sections=[Section(link=task.link, text=task.text)],
+            doc_updated_at=task.last_modified,
+            source=DocumentSource.ASANA,
+            semantic_identifier=task.title,
+            metadata={
+                "group": task.project_gid,
+                "project": task.project_name,
+            },
+        )
+
+
+if __name__ == "__main__":
+    import time
+    import os
+
+    logger.notice("Starting Asana connector test")
+    connector = AsanaConnector(
+        os.environ["WORKSPACE_ID"],
+        os.environ["PROJECT_IDS"],
+        os.environ["TEAM_ID"],
+    )
+    connector.load_credentials(
+        {
+            "asana_api_token_secret": os.environ["API_TOKEN"],
+        }
+    )
+    logger.info("Loading all documents from Asana")
+    all_docs = connector.load_from_state()
+    current = time.time()
+    one_day_ago = current - 24 * 60 * 60  # 1 day
+    logger.info("Polling for documents updated in the last 24 hours")
+    latest_docs = connector.poll_source(one_day_ago, current)
+    for docs in latest_docs:
+        for doc in docs:
+            print(doc.id)
+    logger.notice("Asana connector test completed")

backend/danswer/connectors/blob/connector.py

@@ -56,7 +56,7 @@ class BlobStorageConnector(LoadConnector, PollConnector):
         Raises ValueError for unsupported bucket types.
         """
 
-        logger.info(
+        logger.debug(
             f"Loading credentials for {self.bucket_name} or type {self.bucket_type}"
         )
 
@@ -169,7 +169,7 @@ class BlobStorageConnector(LoadConnector, PollConnector):
         end: datetime,
     ) -> GenerateDocumentsOutput:
         if self.s3_client is None:
-            raise ConnectorMissingCredentialError("Blog storage")
+            raise ConnectorMissingCredentialError("Blob storage")
 
         paginator = self.s3_client.get_paginator("list_objects_v2")
         pages = paginator.paginate(Bucket=self.bucket_name, Prefix=self.prefix)
@@ -220,7 +220,7 @@ class BlobStorageConnector(LoadConnector, PollConnector):
             yield batch
 
     def load_from_state(self) -> GenerateDocumentsOutput:
-        logger.info("Loading blob objects")
+        logger.debug("Loading blob objects")
         return self._yield_blob_objects(
             start=datetime(1970, 1, 1, tzinfo=timezone.utc),
             end=datetime.now(timezone.utc),
@@ -230,7 +230,7 @@ class BlobStorageConnector(LoadConnector, PollConnector):
         self, start: SecondsSinceUnixEpoch, end: SecondsSinceUnixEpoch
     ) -> GenerateDocumentsOutput:
         if self.s3_client is None:
-            raise ConnectorMissingCredentialError("Blog storage")
+            raise ConnectorMissingCredentialError("Blob storage")
 
         start_datetime = datetime.fromtimestamp(start, tz=timezone.utc)
         end_datetime = datetime.fromtimestamp(end, tz=timezone.utc)

backend/danswer/connectors/confluence/connector.py

@@ -7,14 +7,16 @@ from datetime import timezone
 from functools import lru_cache
 from typing import Any
 from typing import cast
-from urllib.parse import urlparse
 
 import bs4
 from atlassian import Confluence  # type:ignore
 from requests import HTTPError
 
+from danswer.configs.app_configs import (
+    CONFLUENCE_CONNECTOR_ATTACHMENT_CHAR_COUNT_THRESHOLD,
+)
 from danswer.configs.app_configs import CONFLUENCE_CONNECTOR_ATTACHMENT_SIZE_THRESHOLD
-from danswer.configs.app_configs import CONFLUENCE_CONNECTOR_INDEX_ONLY_ACTIVE_PAGES
+from danswer.configs.app_configs import CONFLUENCE_CONNECTOR_INDEX_ARCHIVED_PAGES
 from danswer.configs.app_configs import CONFLUENCE_CONNECTOR_LABELS_TO_SKIP
 from danswer.configs.app_configs import CONFLUENCE_CONNECTOR_SKIP_LABEL_INDEXING
 from danswer.configs.app_configs import CONTINUE_ON_CONNECTOR_FAILURE
@@ -42,77 +44,12 @@ logger = setup_logger()
 # 2. Segment into Sections for more accurate linking, can split by headers but make sure no text/ordering is lost
 
 
-def _extract_confluence_keys_from_cloud_url(wiki_url: str) -> tuple[str, str, str]:
-    """Sample
-    URL w/ page: https://danswer.atlassian.net/wiki/spaces/1234abcd/pages/5678efgh/overview
-    URL w/o page: https://danswer.atlassian.net/wiki/spaces/ASAM/overview
-
-    wiki_base is https://danswer.atlassian.net/wiki
-    space is 1234abcd
-    page_id is 5678efgh
-    """
-    parsed_url = urlparse(wiki_url)
-    wiki_base = (
-        parsed_url.scheme
-        + "://"
-        + parsed_url.netloc
-        + parsed_url.path.split("/spaces")[0]
-    )
-
-    path_parts = parsed_url.path.split("/")
-    space = path_parts[3]
-
-    page_id = path_parts[5] if len(path_parts) > 5 else ""
-    return wiki_base, space, page_id
-
-
-def _extract_confluence_keys_from_datacenter_url(wiki_url: str) -> tuple[str, str, str]:
-    """Sample
-    URL w/ page https://danswer.ai/confluence/display/1234abcd/pages/5678efgh/overview
-    URL w/o page https://danswer.ai/confluence/display/1234abcd/overview
-    wiki_base is https://danswer.ai/confluence
-    space is 1234abcd
-    page_id is 5678efgh
-    """
-    # /display/ is always right before the space and at the end of the base print()
-    DISPLAY = "/display/"
-    PAGE = "/pages/"
-
-    parsed_url = urlparse(wiki_url)
-    wiki_base = (
-        parsed_url.scheme
-        + "://"
-        + parsed_url.netloc
-        + parsed_url.path.split(DISPLAY)[0]
-    )
-    space = DISPLAY.join(parsed_url.path.split(DISPLAY)[1:]).split("/")[0]
-    page_id = ""
-    if (content := parsed_url.path.split(PAGE)) and len(content) > 1:
-        page_id = content[1]
-    return wiki_base, space, page_id
-
-
-def extract_confluence_keys_from_url(wiki_url: str) -> tuple[str, str, str, bool]:
-    is_confluence_cloud = (
-        ".atlassian.net/wiki/spaces/" in wiki_url
-        or ".jira.com/wiki/spaces/" in wiki_url
-    )
-
-    try:
-        if is_confluence_cloud:
-            wiki_base, space, page_id = _extract_confluence_keys_from_cloud_url(
-                wiki_url
-            )
-        else:
-            wiki_base, space, page_id = _extract_confluence_keys_from_datacenter_url(
-                wiki_url
-            )
-    except Exception as e:
-        error_msg = f"Not a valid Confluence Wiki Link, unable to extract wiki base, space, and page id. Exception: {e}"
-        logger.error(error_msg)
-        raise ValueError(error_msg)
-
-    return wiki_base, space, page_id, is_confluence_cloud
+NO_PERMISSIONS_TO_VIEW_ATTACHMENTS_ERROR_STR = (
+    "User not permitted to view attachments on content"
+)
+NO_PARENT_OR_NO_PERMISSIONS_ERROR_STR = (
+    "No parent or not permitted to view content with id"
+)
 
 
 @lru_cache()
@@ -200,19 +137,38 @@ def _comment_dfs(
         comments_str += "\nComment:\n" + parse_html_page(
             comment_html, confluence_client
         )
-        child_comment_pages = get_page_child_by_type(
-            comment_page["id"],
-            type="comment",
-            start=None,
-            limit=None,
-            expand="body.storage.value",
-        )
-        comments_str = _comment_dfs(
-            comments_str, child_comment_pages, confluence_client
-        )
+        try:
+            child_comment_pages = get_page_child_by_type(
+                comment_page["id"],
+                type="comment",
+                start=None,
+                limit=None,
+                expand="body.storage.value",
+            )
+            comments_str = _comment_dfs(
+                comments_str, child_comment_pages, confluence_client
+            )
+        except HTTPError as e:
+            # not the cleanest, but I'm not aware of a nicer way to check the error
+            if NO_PARENT_OR_NO_PERMISSIONS_ERROR_STR not in str(e):
+                raise
+
     return comments_str
 
 
+def _datetime_from_string(datetime_string: str) -> datetime:
+    datetime_object = datetime.fromisoformat(datetime_string)
+
+    if datetime_object.tzinfo is None:
+        # If no timezone info, assume it is UTC
+        datetime_object = datetime_object.replace(tzinfo=timezone.utc)
+    else:
+        # If not in UTC, translate it
+        datetime_object = datetime_object.astimezone(timezone.utc)
+
+    return datetime_object
+
+
 class RecursiveIndexer:
     def __init__(
         self,
@@ -342,7 +298,10 @@ class RecursiveIndexer:
 class ConfluenceConnector(LoadConnector, PollConnector):
     def __init__(
         self,
-        wiki_page_url: str,
+        wiki_base: str,
+        space: str,
+        is_cloud: bool,
+        page_id: str = "",
         index_recursively: bool = True,
         batch_size: int = INDEX_BATCH_SIZE,
         continue_on_failure: bool = CONTINUE_ON_CONNECTOR_FAILURE,
@@ -356,15 +315,15 @@ class ConfluenceConnector(LoadConnector, PollConnector):
         self.labels_to_skip = set(labels_to_skip)
         self.recursive_indexer: RecursiveIndexer | None = None
         self.index_recursively = index_recursively
-        (
-            self.wiki_base,
-            self.space,
-            self.page_id,
-            self.is_cloud,
-        ) = extract_confluence_keys_from_url(wiki_page_url)
+
+        # Remove trailing slash from wiki_base if present
+        self.wiki_base = wiki_base.rstrip("/")
+        self.space = space
+        self.page_id = page_id
+
+        self.is_cloud = is_cloud
 
         self.space_level_scan = False
-
         self.confluence_client: Confluence | None = None
 
         if self.page_id is None or self.page_id == "":
@@ -384,7 +343,6 @@ class ConfluenceConnector(LoadConnector, PollConnector):
             username=username if self.is_cloud else None,
             password=access_token if self.is_cloud else None,
             token=access_token if not self.is_cloud else None,
-            cloud=self.is_cloud,
         )
         return None
 
@@ -403,9 +361,7 @@ class ConfluenceConnector(LoadConnector, PollConnector):
                     start=start_ind,
                     limit=batch_size,
                     status=(
-                        "current"
-                        if CONFLUENCE_CONNECTOR_INDEX_ONLY_ACTIVE_PAGES
-                        else None
+                        None if CONFLUENCE_CONNECTOR_INDEX_ARCHIVED_PAGES else "current"
                     ),
                     expand="body.storage.value,version",
                 )
@@ -426,9 +382,9 @@ class ConfluenceConnector(LoadConnector, PollConnector):
                                 start=start_ind + i,
                                 limit=1,
                                 status=(
-                                    "current"
-                                    if CONFLUENCE_CONNECTOR_INDEX_ONLY_ACTIVE_PAGES
-                                    else None
+                                    None
+                                    if CONFLUENCE_CONNECTOR_INDEX_ARCHIVED_PAGES
+                                    else "current"
                                 ),
                                 expand="body.storage.value,version",
                             )
@@ -535,145 +491,249 @@ class ConfluenceConnector(LoadConnector, PollConnector):
             logger.exception("Ran into exception when fetching labels from Confluence")
             return []
 
+    @classmethod
+    def _attachment_to_download_link(
+        cls, confluence_client: Confluence, attachment: dict[str, Any]
+    ) -> str:
+        return confluence_client.url + attachment["_links"]["download"]
+
+    @classmethod
+    def _attachment_to_content(
+        cls,
+        confluence_client: Confluence,
+        attachment: dict[str, Any],
+    ) -> str | None:
+        """If it returns None, assume that we should skip this attachment."""
+        if attachment["metadata"]["mediaType"] in [
+            "image/jpeg",
+            "image/png",
+            "image/gif",
+            "image/svg+xml",
+            "video/mp4",
+            "video/quicktime",
+        ]:
+            return None
+
+        download_link = cls._attachment_to_download_link(confluence_client, attachment)
+
+        attachment_size = attachment["extensions"]["fileSize"]
+        if attachment_size > CONFLUENCE_CONNECTOR_ATTACHMENT_SIZE_THRESHOLD:
+            logger.warning(
+                f"Skipping {download_link} due to size. "
+                f"size={attachment_size} "
+                f"threshold={CONFLUENCE_CONNECTOR_ATTACHMENT_SIZE_THRESHOLD}"
+            )
+            return None
+
+        response = confluence_client._session.get(download_link)
+        if response.status_code != 200:
+            logger.warning(
+                f"Failed to fetch {download_link} with invalid status code {response.status_code}"
+            )
+            return None
+
+        extracted_text = extract_file_text(
+            attachment["title"], io.BytesIO(response.content), False
+        )
+        if len(extracted_text) > CONFLUENCE_CONNECTOR_ATTACHMENT_CHAR_COUNT_THRESHOLD:
+            logger.warning(
+                f"Skipping {download_link} due to char count. "
+                f"char count={len(extracted_text)} "
+                f"threshold={CONFLUENCE_CONNECTOR_ATTACHMENT_CHAR_COUNT_THRESHOLD}"
+            )
+            return None
+
+        return extracted_text
+
     def _fetch_attachments(
         self, confluence_client: Confluence, page_id: str, files_in_used: list[str]
-    ) -> str:
+    ) -> tuple[str, list[dict[str, Any]]]:
+        unused_attachments: list = []
+
         get_attachments_from_content = make_confluence_call_handle_rate_limit(
             confluence_client.get_attachments_from_content
         )
         files_attachment_content: list = []
 
         try:
+            expand = "history.lastUpdated,metadata.labels"
             attachments_container = get_attachments_from_content(
-                page_id, start=0, limit=500
+                page_id, start=0, limit=500, expand=expand
             )
             for attachment in attachments_container["results"]:
-                if attachment["metadata"]["mediaType"] in [
-                    "image/jpeg",
-                    "image/png",
-                    "image/gif",
-                    "image/svg+xml",
-                    "video/mp4",
-                    "video/quicktime",
-                ]:
-                    continue
-
                 if attachment["title"] not in files_in_used:
+                    unused_attachments.append(attachment)
                     continue
 
-                download_link = confluence_client.url + attachment["_links"]["download"]
-
-                attachment_size = attachment["extensions"]["fileSize"]
-                if attachment_size > CONFLUENCE_CONNECTOR_ATTACHMENT_SIZE_THRESHOLD:
-                    logger.warning(
-                        f"Skipping {download_link} due to size. "
-                        f"size={attachment_size} "
-                        f"threshold={CONFLUENCE_CONNECTOR_ATTACHMENT_SIZE_THRESHOLD}"
-                    )
-                    continue
-
-                download_link = confluence_client.url + attachment["_links"]["download"]
-                response = confluence_client._session.get(download_link)
-
-                if response.status_code == 200:
-                    extract = extract_file_text(
-                        attachment["title"], io.BytesIO(response.content), False
-                    )
-                    files_attachment_content.append(extract)
+                attachment_content = self._attachment_to_content(
+                    confluence_client, attachment
+                )
+                if attachment_content:
+                    files_attachment_content.append(attachment_content)
 
         except Exception as e:
+            if isinstance(
+                e, HTTPError
+            ) and NO_PERMISSIONS_TO_VIEW_ATTACHMENTS_ERROR_STR in str(e):
+                logger.warning(
+                    f"User does not have access to attachments on page '{page_id}'"
+                )
+                return "", []
+
             if not self.continue_on_failure:
                 raise e
             logger.exception(
                 f"Ran into exception when fetching attachments from Confluence: {e}"
             )
 
-        return "\n".join(files_attachment_content)
+        return "\n".join(files_attachment_content), unused_attachments
 
     def _get_doc_batch(
         self, start_ind: int, time_filter: Callable[[datetime], bool] | None = None
-    ) -> tuple[list[Document], int]:
+    ) -> tuple[list[Document], list[dict[str, Any]], int]:
         doc_batch: list[Document] = []
+        unused_attachments: list[dict[str, Any]] = []
 
         if self.confluence_client is None:
             raise ConnectorMissingCredentialError("Confluence")
         batch = self._fetch_pages(self.confluence_client, start_ind)
 
         for page in batch:
-            last_modified_str = page["version"]["when"]
+            last_modified = _datetime_from_string(page["version"]["when"])
             author = cast(str | None, page["version"].get("by", {}).get("email"))
-            last_modified = datetime.fromisoformat(last_modified_str)
 
-            if last_modified.tzinfo is None:
-                # If no timezone info, assume it is UTC
-                last_modified = last_modified.replace(tzinfo=timezone.utc)
-            else:
-                # If not in UTC, translate it
-                last_modified = last_modified.astimezone(timezone.utc)
+            if time_filter and not time_filter(last_modified):
+                continue
 
-            if time_filter is None or time_filter(last_modified):
-                page_id = page["id"]
+            page_id = page["id"]
 
-                if self.labels_to_skip or not CONFLUENCE_CONNECTOR_SKIP_LABEL_INDEXING:
-                    page_labels = self._fetch_labels(self.confluence_client, page_id)
+            if self.labels_to_skip or not CONFLUENCE_CONNECTOR_SKIP_LABEL_INDEXING:
+                page_labels = self._fetch_labels(self.confluence_client, page_id)
 
-                # check disallowed labels
-                if self.labels_to_skip:
-                    label_intersection = self.labels_to_skip.intersection(page_labels)
-                    if label_intersection:
-                        logger.info(
-                            f"Page with ID '{page_id}' has a label which has been "
-                            f"designated as disallowed: {label_intersection}. Skipping."
-                        )
-
-                        continue
-
-                page_html = (
-                    page["body"]
-                    .get("storage", page["body"].get("view", {}))
-                    .get("value")
-                )
-                page_url = self.wiki_base + page["_links"]["webui"]
-                if not page_html:
-                    logger.debug("Page is empty, skipping: %s", page_url)
-                    continue
-                page_text = parse_html_page(page_html, self.confluence_client)
-
-                files_in_used = get_used_attachments(page_html, self.confluence_client)
-                attachment_text = self._fetch_attachments(
-                    self.confluence_client, page_id, files_in_used
-                )
-                page_text += attachment_text
-                comments_text = self._fetch_comments(self.confluence_client, page_id)
-                page_text += comments_text
-                doc_metadata: dict[str, str | list[str]] = {
-                    "Wiki Space Name": self.space
-                }
-                if not CONFLUENCE_CONNECTOR_SKIP_LABEL_INDEXING and page_labels:
-                    doc_metadata["labels"] = page_labels
-
-                doc_batch.append(
-                    Document(
-                        id=page_url,
-                        sections=[Section(link=page_url, text=page_text)],
-                        source=DocumentSource.CONFLUENCE,
-                        semantic_identifier=page["title"],
-                        doc_updated_at=last_modified,
-                        primary_owners=(
-                            [BasicExpertInfo(email=author)] if author else None
-                        ),
-                        metadata=doc_metadata,
+            # check disallowed labels
+            if self.labels_to_skip:
+                label_intersection = self.labels_to_skip.intersection(page_labels)
+                if label_intersection:
+                    logger.info(
+                        f"Page with ID '{page_id}' has a label which has been "
+                        f"designated as disallowed: {label_intersection}. Skipping."
                     )
+
+                    continue
+
+            page_html = (
+                page["body"].get("storage", page["body"].get("view", {})).get("value")
+            )
+            page_url = self.wiki_base + page["_links"]["webui"]
+            if not page_html:
+                logger.debug("Page is empty, skipping: %s", page_url)
+                continue
+            page_text = parse_html_page(page_html, self.confluence_client)
+
+            files_in_used = get_used_attachments(page_html, self.confluence_client)
+            attachment_text, unused_page_attachments = self._fetch_attachments(
+                self.confluence_client, page_id, files_in_used
+            )
+            unused_attachments.extend(unused_page_attachments)
+
+            page_text += attachment_text
+            comments_text = self._fetch_comments(self.confluence_client, page_id)
+            page_text += comments_text
+            doc_metadata: dict[str, str | list[str]] = {"Wiki Space Name": self.space}
+            if not CONFLUENCE_CONNECTOR_SKIP_LABEL_INDEXING and page_labels:
+                doc_metadata["labels"] = page_labels
+
+            doc_batch.append(
+                Document(
+                    id=page_url,
+                    sections=[Section(link=page_url, text=page_text)],
+                    source=DocumentSource.CONFLUENCE,
+                    semantic_identifier=page["title"],
+                    doc_updated_at=last_modified,
+                    primary_owners=(
+                        [BasicExpertInfo(email=author)] if author else None
+                    ),
+                    metadata=doc_metadata,
                 )
-        return doc_batch, len(batch)
+            )
+        return (
+            doc_batch,
+            unused_attachments,
+            len(batch),
+        )
+
+    def _get_attachment_batch(
+        self,
+        start_ind: int,
+        attachments: list[dict[str, Any]],
+        time_filter: Callable[[datetime], bool] | None = None,
+    ) -> tuple[list[Document], int]:
+        doc_batch: list[Document] = []
+
+        if self.confluence_client is None:
+            raise ConnectorMissingCredentialError("Confluence")
+
+        end_ind = min(start_ind + self.batch_size, len(attachments))
+
+        for attachment in attachments[start_ind:end_ind]:
+            last_updated = _datetime_from_string(
+                attachment["history"]["lastUpdated"]["when"]
+            )
+
+            if time_filter and not time_filter(last_updated):
+                continue
+
+            attachment_url = self._attachment_to_download_link(
+                self.confluence_client, attachment
+            )
+            attachment_content = self._attachment_to_content(
+                self.confluence_client, attachment
+            )
+            if attachment_content is None:
+                continue
+
+            creator_email = attachment["history"]["createdBy"].get("email")
+
+            comment = attachment["metadata"].get("comment", "")
+            doc_metadata: dict[str, str | list[str]] = {"comment": comment}
+
+            attachment_labels: list[str] = []
+            if not CONFLUENCE_CONNECTOR_SKIP_LABEL_INDEXING:
+                for label in attachment["metadata"]["labels"]["results"]:
+                    attachment_labels.append(label["name"])
+
+            doc_metadata["labels"] = attachment_labels
+
+            doc_batch.append(
+                Document(
+                    id=attachment_url,
+                    sections=[Section(link=attachment_url, text=attachment_content)],
+                    source=DocumentSource.CONFLUENCE,
+                    semantic_identifier=attachment["title"],
+                    doc_updated_at=last_updated,
+                    primary_owners=(
+                        [BasicExpertInfo(email=creator_email)]
+                        if creator_email
+                        else None
+                    ),
+                    metadata=doc_metadata,
+                )
+            )
+
+        return doc_batch, end_ind - start_ind
 
     def load_from_state(self) -> GenerateDocumentsOutput:
+        unused_attachments = []
+
         if self.confluence_client is None:
             raise ConnectorMissingCredentialError("Confluence")
 
         start_ind = 0
         while True:
-            doc_batch, num_pages = self._get_doc_batch(start_ind)
+            doc_batch, unused_attachments_batch, num_pages = self._get_doc_batch(
+                start_ind
+            )
+            unused_attachments.extend(unused_attachments_batch)
             start_ind += num_pages
             if doc_batch:
                 yield doc_batch
@@ -681,9 +741,23 @@ class ConfluenceConnector(LoadConnector, PollConnector):
             if num_pages < self.batch_size:
                 break
 
+        start_ind = 0
+        while True:
+            attachment_batch, num_attachments = self._get_attachment_batch(
+                start_ind, unused_attachments
+            )
+            start_ind += num_attachments
+            if attachment_batch:
+                yield attachment_batch
+
+            if num_attachments < self.batch_size:
+                break
+
     def poll_source(
         self, start: SecondsSinceUnixEpoch, end: SecondsSinceUnixEpoch
     ) -> GenerateDocumentsOutput:
+        unused_attachments = []
+
         if self.confluence_client is None:
             raise ConnectorMissingCredentialError("Confluence")
 
@@ -692,9 +766,11 @@ class ConfluenceConnector(LoadConnector, PollConnector):
 
         start_ind = 0
         while True:
-            doc_batch, num_pages = self._get_doc_batch(
+            doc_batch, unused_attachments_batch, num_pages = self._get_doc_batch(
                 start_ind, time_filter=lambda t: start_time <= t <= end_time
             )
+            unused_attachments.extend(unused_attachments_batch)
+
             start_ind += num_pages
             if doc_batch:
                 yield doc_batch
@@ -702,9 +778,29 @@ class ConfluenceConnector(LoadConnector, PollConnector):
             if num_pages < self.batch_size:
                 break
 
+        start_ind = 0
+        while True:
+            attachment_batch, num_attachments = self._get_attachment_batch(
+                start_ind,
+                unused_attachments,
+                time_filter=lambda t: start_time <= t <= end_time,
+            )
+            start_ind += num_attachments
+            if attachment_batch:
+                yield attachment_batch
+
+            if num_attachments < self.batch_size:
+                break
+
 
 if __name__ == "__main__":
-    connector = ConfluenceConnector(os.environ["CONFLUENCE_TEST_SPACE_URL"])
+    connector = ConfluenceConnector(
+        wiki_base=os.environ["CONFLUENCE_TEST_SPACE_URL"],
+        space=os.environ["CONFLUENCE_TEST_SPACE"],
+        is_cloud=os.environ.get("CONFLUENCE_IS_CLOUD", "true").lower() == "true",
+        page_id=os.environ.get("CONFLUENCE_TEST_PAGE_ID", ""),
+        index_recursively=True,
+    )
     connector.load_credentials(
         {
             "confluence_username": os.environ["CONFLUENCE_USER_NAME"],

backend/danswer/connectors/confluence/rate_limit_handler.py

@@ -23,25 +23,33 @@ class ConfluenceRateLimitError(Exception):
 
 def make_confluence_call_handle_rate_limit(confluence_call: F) -> F:
     def wrapped_call(*args: list[Any], **kwargs: Any) -> Any:
+        max_retries = 5
         starting_delay = 5
         backoff = 2
         max_delay = 600
 
-        for attempt in range(10):
+        for attempt in range(max_retries):
             try:
                 return confluence_call(*args, **kwargs)
             except HTTPError as e:
+                # Check if the response or headers are None to avoid potential AttributeError
+                if e.response is None or e.response.headers is None:
+                    logger.warning("HTTPError with `None` as response or as headers")
+                    raise e
+
+                retry_after_header = e.response.headers.get("Retry-After")
                 if (
                     e.response.status_code == 429
                     or RATE_LIMIT_MESSAGE_LOWERCASE in e.response.text.lower()
                 ):
                     retry_after = None
-                    try:
-                        retry_after = int(e.response.headers.get("Retry-After"))
-                    except (ValueError, TypeError):
-                        pass
+                    if retry_after_header is not None:
+                        try:
+                            retry_after = int(retry_after_header)
+                        except ValueError:
+                            pass
 
-                    if retry_after:
+                    if retry_after is not None:
                         logger.warning(
                             f"Rate limit hit. Retrying after {retry_after} seconds..."
                         )
@@ -55,5 +63,14 @@ def make_confluence_call_handle_rate_limit(confluence_call: F) -> F:
                 else:
                     # re-raise, let caller handle
                     raise
+            except AttributeError as e:
+                # Some error within the Confluence library, unclear why it fails.
+                # Users reported it to be intermittent, so just retry
+                logger.warning(f"Confluence Internal Error, retrying... {e}")
+                delay = min(starting_delay * (backoff**attempt), max_delay)
+                time.sleep(delay)
+
+                if attempt == max_retries - 1:
+                    raise e
 
     return cast(F, wrapped_call)

backend/danswer/connectors/connector_runner.py

@@ -0,0 +1,70 @@
+import sys
+from datetime import datetime
+
+from danswer.connectors.interfaces import BaseConnector
+from danswer.connectors.interfaces import GenerateDocumentsOutput
+from danswer.connectors.interfaces import LoadConnector
+from danswer.connectors.interfaces import PollConnector
+from danswer.utils.logger import setup_logger
+
+
+logger = setup_logger()
+
+
+TimeRange = tuple[datetime, datetime]
+
+
+class ConnectorRunner:
+    def __init__(
+        self,
+        connector: BaseConnector,
+        time_range: TimeRange | None = None,
+        fail_loudly: bool = False,
+    ):
+        self.connector = connector
+
+        if isinstance(self.connector, PollConnector):
+            if time_range is None:
+                raise ValueError("time_range is required for PollConnector")
+
+            self.doc_batch_generator = self.connector.poll_source(
+                time_range[0].timestamp(), time_range[1].timestamp()
+            )
+
+        elif isinstance(self.connector, LoadConnector):
+            if time_range and fail_loudly:
+                raise ValueError(
+                    "time_range specified, but passed in connector is not a PollConnector"
+                )
+
+            self.doc_batch_generator = self.connector.load_from_state()
+
+        else:
+            raise ValueError(f"Invalid connector. type: {type(self.connector)}")
+
+    def run(self) -> GenerateDocumentsOutput:
+        """Adds additional exception logging to the connector."""
+        try:
+            yield from self.doc_batch_generator
+        except Exception:
+            exc_type, _, exc_traceback = sys.exc_info()
+
+            # Traverse the traceback to find the last frame where the exception was raised
+            tb = exc_traceback
+            if tb is None:
+                logger.error("No traceback found for exception")
+                raise
+
+            while tb.tb_next:
+                tb = tb.tb_next  # Move to the next frame in the traceback
+
+            # Get the local variables from the frame where the exception occurred
+            local_vars = tb.tb_frame.f_locals
+            local_vars_str = "\n".join(
+                f"{key}: {value}" for key, value in local_vars.items()
+            )
+            logger.error(
+                f"Error in connector. type: {exc_type};\n"
+                f"local_vars below -> \n{local_vars_str}"
+            )
+            raise

backend/danswer/connectors/cross_connector_utils/rate_limit_wrapper.py

@@ -56,7 +56,7 @@ class _RateLimitDecorator:
             sleep_cnt = 0
             while len(self.call_history) == self.max_calls:
                 sleep_time = self.sleep_time * (self.sleep_backoff**sleep_cnt)
-                logger.info(
+                logger.notice(
                     f"Rate limit exceeded for function {func.__name__}. "
                     f"Waiting {sleep_time} seconds before retrying."
                 )

backend/danswer/connectors/danswer_jira/connector.py

@@ -45,10 +45,15 @@ def extract_jira_project(url: str) -> tuple[str, str]:
     return jira_base, jira_project
 
 
-def extract_text_from_content(content: dict) -> str:
+def extract_text_from_adf(adf: dict | None) -> str:
+    """Extracts plain text from Atlassian Document Format:
+    https://developer.atlassian.com/cloud/jira/platform/apis/document/structure/
+
+    WARNING: This function is incomplete and will e.g. skip lists!
+    """
     texts = []
-    if "content" in content:
-        for block in content["content"]:
+    if adf is not None and "content" in adf:
+        for block in adf["content"]:
             if "content" in block:
                 for item in block["content"]:
                     if item["type"] == "text":
@@ -72,18 +77,15 @@ def _get_comment_strs(
     comment_strs = []
     for comment in jira.fields.comment.comments:
         try:
-            if hasattr(comment, "body"):
-                body_text = extract_text_from_content(comment.raw["body"])
-            elif hasattr(comment, "raw"):
-                body = comment.raw.get("body", "No body content available")
-                body_text = (
-                    extract_text_from_content(body) if isinstance(body, dict) else body
-                )
-            else:
-                body_text = "No body attribute found"
+            body_text = (
+                comment.body
+                if JIRA_API_VERSION == "2"
+                else extract_text_from_adf(comment.raw["body"])
+            )
 
             if (
                 hasattr(comment, "author")
+                and hasattr(comment.author, "emailAddress")
                 and comment.author.emailAddress in comment_email_blacklist
             ):
                 continue  # Skip adding comment if author's email is in blacklist
@@ -126,11 +128,14 @@ def fetch_jira_issues_batch(
             )
             continue
 
+        description = (
+            jira.fields.description
+            if JIRA_API_VERSION == "2"
+            else extract_text_from_adf(jira.raw["fields"]["description"])
+        )
         comments = _get_comment_strs(jira, comment_email_blacklist)
-        semantic_rep = (
-            f"{jira.fields.description}\n"
-            if jira.fields.description
-            else "" + "\n".join([f"Comment: {comment}" for comment in comments])
+        semantic_rep = f"{description}\n" + "\n".join(
+            [f"Comment: {comment}" for comment in comments if comment]
         )
 
         page_url = f"{jira_client.client_info()}/browse/{jira.key}"

backend/danswer/connectors/factory.py

@@ -4,6 +4,7 @@ from typing import Type
 from sqlalchemy.orm import Session
 
 from danswer.configs.constants import DocumentSource
+from danswer.connectors.asana.connector import AsanaConnector
 from danswer.connectors.axero.connector import AxeroConnector
 from danswer.connectors.blob.connector import BlobStorageConnector
 from danswer.connectors.bookstack.connector import BookstackConnector
@@ -91,6 +92,7 @@ def identify_connector_class(
         DocumentSource.CLICKUP: ClickupConnector,
         DocumentSource.MEDIAWIKI: MediaWikiConnector,
         DocumentSource.WIKIPEDIA: WikipediaConnector,
+        DocumentSource.ASANA: AsanaConnector,
         DocumentSource.S3: BlobStorageConnector,
         DocumentSource.R2: BlobStorageConnector,
         DocumentSource.GOOGLE_CLOUD_STORAGE: BlobStorageConnector,
@@ -124,11 +126,11 @@ def identify_connector_class(
 
 
 def instantiate_connector(
+    db_session: Session,
     source: DocumentSource,
     input_type: InputType,
     connector_specific_config: dict[str, Any],
     credential: Credential,
-    db_session: Session,
 ) -> BaseConnector:
     connector_class = identify_connector_class(source, input_type)
     connector = connector_class(**connector_specific_config)

backend/danswer/connectors/file/connector.py

@@ -23,7 +23,7 @@ from danswer.file_processing.extract_file_text import extract_file_text
 from danswer.file_processing.extract_file_text import get_file_ext
 from danswer.file_processing.extract_file_text import is_text_file_extension
 from danswer.file_processing.extract_file_text import load_files_from_zip
-from danswer.file_processing.extract_file_text import pdf_to_text
+from danswer.file_processing.extract_file_text import read_pdf_file
 from danswer.file_processing.extract_file_text import read_text_file
 from danswer.file_store.file_store import get_default_file_store
 from danswer.utils.logger import setup_logger
@@ -75,7 +75,7 @@ def _process_file(
 
     # Using the PDF reader function directly to pass in password cleanly
     elif extension == ".pdf":
-        file_content_raw = pdf_to_text(file=file, pdf_pass=pdf_pass)
+        file_content_raw, file_metadata = read_pdf_file(file=file, pdf_pass=pdf_pass)
 
     else:
         file_content_raw = extract_file_text(

backend/danswer/connectors/github/connector.py

@@ -38,7 +38,7 @@ def _sleep_after_rate_limit_exception(github_client: Github) -> None:
         tzinfo=timezone.utc
     ) - datetime.now(tz=timezone.utc)
     sleep_time += timedelta(minutes=1)  # add an extra minute just to be safe
-    logger.info(f"Ran into Github rate-limit. Sleeping {sleep_time.seconds} seconds.")
+    logger.notice(f"Ran into Github rate-limit. Sleeping {sleep_time.seconds} seconds.")
     time.sleep(sleep_time.seconds)
 
 

backend/danswer/connectors/gmail/connector_auth.py

@@ -50,7 +50,7 @@ def get_gmail_creds_for_authorized_user(
         try:
             creds.refresh(Request())
             if creds.valid:
-                logger.info("Refreshed Gmail tokens.")
+                logger.notice("Refreshed Gmail tokens.")
                 return creds
         except Exception as e:
             logger.exception(f"Failed to refresh gmail access token due to: {e}")
@@ -125,7 +125,7 @@ def update_gmail_credential_access_tokens(
 ) -> OAuthCredentials | None:
     app_credentials = get_google_app_gmail_cred()
     flow = InstalledAppFlow.from_client_config(
-        app_credentials.dict(),
+        app_credentials.model_dump(),
         scopes=SCOPES,
         redirect_uri=_build_frontend_gmail_redirect(),
     )

backend/danswer/connectors/gong/connector.py

@@ -81,10 +81,10 @@ class GongConnector(LoadConnector, PollConnector):
 
         for workspace in workspace_list:
             if workspace:
-                logger.info(f"Updating workspace: {workspace}")
+                logger.info(f"Updating Gong workspace: {workspace}")
                 workspace_id = workspace_map.get(workspace)
                 if not workspace_id:
-                    logger.error(f"Invalid workspace: {workspace}")
+                    logger.error(f"Invalid Gong workspace: {workspace}")
                     if not self.continue_on_fail:
                         raise ValueError(f"Invalid workspace: {workspace}")
                     continue

backend/danswer/connectors/google_drive/connector.py

@@ -6,7 +6,6 @@ from datetime import timezone
 from enum import Enum
 from itertools import chain
 from typing import Any
-from typing import cast
 
 from google.oauth2.credentials import Credentials as OAuthCredentials  # type: ignore
 from google.oauth2.service_account import Credentials as ServiceAccountCredentials  # type: ignore
@@ -21,19 +20,13 @@ from danswer.configs.app_configs import INDEX_BATCH_SIZE
 from danswer.configs.constants import DocumentSource
 from danswer.configs.constants import IGNORE_FOR_QA
 from danswer.connectors.cross_connector_utils.retry_wrapper import retry_builder
-from danswer.connectors.google_drive.connector_auth import (
-    get_google_drive_creds_for_authorized_user,
-)
-from danswer.connectors.google_drive.connector_auth import (
-    get_google_drive_creds_for_service_account,
-)
+from danswer.connectors.google_drive.connector_auth import get_google_drive_creds
 from danswer.connectors.google_drive.constants import (
     DB_CREDENTIALS_DICT_DELEGATED_USER_KEY,
 )
 from danswer.connectors.google_drive.constants import (
     DB_CREDENTIALS_DICT_SERVICE_ACCOUNT_KEY,
 )
-from danswer.connectors.google_drive.constants import DB_CREDENTIALS_DICT_TOKEN_KEY
 from danswer.connectors.interfaces import GenerateDocumentsOutput
 from danswer.connectors.interfaces import LoadConnector
 from danswer.connectors.interfaces import PollConnector
@@ -41,8 +34,8 @@ from danswer.connectors.interfaces import SecondsSinceUnixEpoch
 from danswer.connectors.models import Document
 from danswer.connectors.models import Section
 from danswer.file_processing.extract_file_text import docx_to_text
-from danswer.file_processing.extract_file_text import pdf_to_text
 from danswer.file_processing.extract_file_text import pptx_to_text
+from danswer.file_processing.extract_file_text import read_pdf_file
 from danswer.utils.batching import batch_generator
 from danswer.utils.logger import setup_logger
 
@@ -62,6 +55,8 @@ class GDriveMimeType(str, Enum):
     POWERPOINT = (
         "application/vnd.openxmlformats-officedocument.presentationml.presentation"
     )
+    PLAIN_TEXT = "text/plain"
+    MARKDOWN = "text/markdown"
 
 
 GoogleDriveFileType = dict[str, Any]
@@ -267,7 +262,7 @@ def get_all_files_batched(
     yield from batch_generator(
         items=found_files,
         batch_size=batch_size,
-        pre_batch_yield=lambda batch_files: logger.info(
+        pre_batch_yield=lambda batch_files: logger.debug(
             f"Parseable Documents in batch: {[file['name'] for file in batch_files]}"
         ),
     )
@@ -316,25 +311,29 @@ def extract_text(file: dict[str, str], service: discovery.Resource) -> str:
         GDriveMimeType.PPT.value,
         GDriveMimeType.SPREADSHEET.value,
     ]:
-        export_mime_type = "text/plain"
-        if mime_type == GDriveMimeType.SPREADSHEET.value:
-            export_mime_type = "text/csv"
-        elif mime_type == GDriveMimeType.PPT.value:
-            export_mime_type = "text/plain"
-
-        response = (
+        export_mime_type = (
+            "text/plain"
+            if mime_type != GDriveMimeType.SPREADSHEET.value
+            else "text/csv"
+        )
+        return (
             service.files()
             .export(fileId=file["id"], mimeType=export_mime_type)
             .execute()
+            .decode("utf-8")
         )
-        return response.decode("utf-8")
-
+    elif mime_type in [
+        GDriveMimeType.PLAIN_TEXT.value,
+        GDriveMimeType.MARKDOWN.value,
+    ]:
+        return service.files().get_media(fileId=file["id"]).execute().decode("utf-8")
     elif mime_type == GDriveMimeType.WORD_DOC.value:
         response = service.files().get_media(fileId=file["id"]).execute()
         return docx_to_text(file=io.BytesIO(response))
     elif mime_type == GDriveMimeType.PDF.value:
         response = service.files().get_media(fileId=file["id"]).execute()
-        return pdf_to_text(file=io.BytesIO(response))
+        text, _ = read_pdf_file(file=io.BytesIO(response))
+        return text
     elif mime_type == GDriveMimeType.POWERPOINT.value:
         response = service.files().get_media(fileId=file["id"]).execute()
         return pptx_to_text(file=io.BytesIO(response))
@@ -401,42 +400,7 @@ class GoogleDriveConnector(LoadConnector, PollConnector):
         (2) A credential which holds a service account key JSON file, which
         can then be used to impersonate any user in the workspace.
         """
-        creds: OAuthCredentials | ServiceAccountCredentials | None = None
-        new_creds_dict = None
-        if DB_CREDENTIALS_DICT_TOKEN_KEY in credentials:
-            access_token_json_str = cast(
-                str, credentials[DB_CREDENTIALS_DICT_TOKEN_KEY]
-            )
-            creds = get_google_drive_creds_for_authorized_user(
-                token_json_str=access_token_json_str
-            )
-
-            # tell caller to update token stored in DB if it has changed
-            # (e.g. the token has been refreshed)
-            new_creds_json_str = creds.to_json() if creds else ""
-            if new_creds_json_str != access_token_json_str:
-                new_creds_dict = {DB_CREDENTIALS_DICT_TOKEN_KEY: new_creds_json_str}
-
-        if DB_CREDENTIALS_DICT_SERVICE_ACCOUNT_KEY in credentials:
-            service_account_key_json_str = credentials[
-                DB_CREDENTIALS_DICT_SERVICE_ACCOUNT_KEY
-            ]
-            creds = get_google_drive_creds_for_service_account(
-                service_account_key_json_str=service_account_key_json_str
-            )
-
-            # "Impersonate" a user if one is specified
-            delegated_user_email = cast(
-                str | None, credentials.get(DB_CREDENTIALS_DICT_DELEGATED_USER_KEY)
-            )
-            if delegated_user_email:
-                creds = creds.with_subject(delegated_user_email) if creds else None  # type: ignore
-
-        if creds is None:
-            raise PermissionError(
-                "Unable to access Google Drive - unknown credential structure."
-            )
-
+        creds, new_creds_dict = get_google_drive_creds(credentials)
         self.creds = creds
         return new_creds_dict
 
@@ -503,6 +467,7 @@ class GoogleDriveConnector(LoadConnector, PollConnector):
                                 file["modifiedTime"]
                             ).astimezone(timezone.utc),
                             metadata={} if text_contents else {IGNORE_FOR_QA: "True"},
+                            additional_info=file.get("id"),
                         )
                     )
                 except Exception as e:

backend/danswer/connectors/google_drive/connector_auth.py

@@ -10,11 +10,13 @@ from google.oauth2.service_account import Credentials as ServiceAccountCredentia
 from google_auth_oauthlib.flow import InstalledAppFlow  # type: ignore
 from sqlalchemy.orm import Session
 
+from danswer.configs.app_configs import ENTERPRISE_EDITION_ENABLED
 from danswer.configs.app_configs import WEB_DOMAIN
 from danswer.configs.constants import DocumentSource
 from danswer.configs.constants import KV_CRED_KEY
 from danswer.configs.constants import KV_GOOGLE_DRIVE_CRED_KEY
 from danswer.configs.constants import KV_GOOGLE_DRIVE_SERVICE_ACCOUNT_KEY
+from danswer.connectors.google_drive.constants import BASE_SCOPES
 from danswer.connectors.google_drive.constants import (
     DB_CREDENTIALS_DICT_DELEGATED_USER_KEY,
 )
@@ -22,7 +24,8 @@ from danswer.connectors.google_drive.constants import (
     DB_CREDENTIALS_DICT_SERVICE_ACCOUNT_KEY,
 )
 from danswer.connectors.google_drive.constants import DB_CREDENTIALS_DICT_TOKEN_KEY
-from danswer.connectors.google_drive.constants import SCOPES
+from danswer.connectors.google_drive.constants import FETCH_GROUPS_SCOPES
+from danswer.connectors.google_drive.constants import FETCH_PERMISSIONS_SCOPES
 from danswer.db.credentials import update_credential_json
 from danswer.db.models import User
 from danswer.dynamic_configs.factory import get_dynamic_config_store
@@ -34,15 +37,25 @@ from danswer.utils.logger import setup_logger
 logger = setup_logger()
 
 
+def build_gdrive_scopes() -> list[str]:
+    base_scopes: list[str] = BASE_SCOPES
+    permissions_scopes: list[str] = FETCH_PERMISSIONS_SCOPES
+    groups_scopes: list[str] = FETCH_GROUPS_SCOPES
+
+    if ENTERPRISE_EDITION_ENABLED:
+        return base_scopes + permissions_scopes + groups_scopes
+    return base_scopes + permissions_scopes
+
+
 def _build_frontend_google_drive_redirect() -> str:
     return f"{WEB_DOMAIN}/admin/connectors/google-drive/auth/callback"
 
 
 def get_google_drive_creds_for_authorized_user(
-    token_json_str: str,
+    token_json_str: str, scopes: list[str] = build_gdrive_scopes()
 ) -> OAuthCredentials | None:
     creds_json = json.loads(token_json_str)
-    creds = OAuthCredentials.from_authorized_user_info(creds_json, SCOPES)
+    creds = OAuthCredentials.from_authorized_user_info(creds_json, scopes)
     if creds.valid:
         return creds
 
@@ -50,7 +63,7 @@ def get_google_drive_creds_for_authorized_user(
         try:
             creds.refresh(Request())
             if creds.valid:
-                logger.info("Refreshed Google Drive tokens.")
+                logger.notice("Refreshed Google Drive tokens.")
                 return creds
         except Exception as e:
             logger.exception(f"Failed to refresh google drive access token due to: {e}")
@@ -59,18 +72,67 @@ def get_google_drive_creds_for_authorized_user(
     return None
 
 
-def get_google_drive_creds_for_service_account(
-    service_account_key_json_str: str,
+def _get_google_drive_creds_for_service_account(
+    service_account_key_json_str: str, scopes: list[str] = build_gdrive_scopes()
 ) -> ServiceAccountCredentials | None:
     service_account_key = json.loads(service_account_key_json_str)
     creds = ServiceAccountCredentials.from_service_account_info(
-        service_account_key, scopes=SCOPES
+        service_account_key, scopes=scopes
     )
     if not creds.valid or not creds.expired:
         creds.refresh(Request())
     return creds if creds.valid else None
 
 
+def get_google_drive_creds(
+    credentials: dict[str, str], scopes: list[str] = build_gdrive_scopes()
+) -> tuple[ServiceAccountCredentials | OAuthCredentials, dict[str, str] | None]:
+    oauth_creds = None
+    service_creds = None
+    new_creds_dict = None
+    if DB_CREDENTIALS_DICT_TOKEN_KEY in credentials:
+        access_token_json_str = cast(str, credentials[DB_CREDENTIALS_DICT_TOKEN_KEY])
+        oauth_creds = get_google_drive_creds_for_authorized_user(
+            token_json_str=access_token_json_str, scopes=scopes
+        )
+
+        # tell caller to update token stored in DB if it has changed
+        # (e.g. the token has been refreshed)
+        new_creds_json_str = oauth_creds.to_json() if oauth_creds else ""
+        if new_creds_json_str != access_token_json_str:
+            new_creds_dict = {DB_CREDENTIALS_DICT_TOKEN_KEY: new_creds_json_str}
+
+    elif DB_CREDENTIALS_DICT_SERVICE_ACCOUNT_KEY in credentials:
+        service_account_key_json_str = credentials[
+            DB_CREDENTIALS_DICT_SERVICE_ACCOUNT_KEY
+        ]
+        service_creds = _get_google_drive_creds_for_service_account(
+            service_account_key_json_str=service_account_key_json_str,
+            scopes=scopes,
+        )
+
+        # "Impersonate" a user if one is specified
+        delegated_user_email = cast(
+            str | None, credentials.get(DB_CREDENTIALS_DICT_DELEGATED_USER_KEY)
+        )
+        if delegated_user_email:
+            service_creds = (
+                service_creds.with_subject(delegated_user_email)
+                if service_creds
+                else None
+            )
+
+    creds: ServiceAccountCredentials | OAuthCredentials | None = (
+        oauth_creds or service_creds
+    )
+    if creds is None:
+        raise PermissionError(
+            "Unable to access Google Drive - unknown credential structure."
+        )
+
+    return creds, new_creds_dict
+
+
 def verify_csrf(credential_id: int, state: str) -> None:
     csrf = get_dynamic_config_store().load(KV_CRED_KEY.format(str(credential_id)))
     if csrf != state:
@@ -84,7 +146,7 @@ def get_auth_url(credential_id: int) -> str:
     credential_json = json.loads(creds_str)
     flow = InstalledAppFlow.from_client_config(
         credential_json,
-        scopes=SCOPES,
+        scopes=build_gdrive_scopes(),
         redirect_uri=_build_frontend_google_drive_redirect(),
     )
     auth_url, _ = flow.authorization_url(prompt="consent")
@@ -106,8 +168,8 @@ def update_credential_access_tokens(
 ) -> OAuthCredentials | None:
     app_credentials = get_google_app_cred()
     flow = InstalledAppFlow.from_client_config(
-        app_credentials.dict(),
-        scopes=SCOPES,
+        app_credentials.model_dump(),
+        scopes=build_gdrive_scopes(),
         redirect_uri=_build_frontend_google_drive_redirect(),
     )
     flow.fetch_token(code=auth_code)

backend/danswer/connectors/google_drive/constants.py

@@ -1,7 +1,7 @@
 DB_CREDENTIALS_DICT_TOKEN_KEY = "google_drive_tokens"
 DB_CREDENTIALS_DICT_SERVICE_ACCOUNT_KEY = "google_drive_service_account_key"
 DB_CREDENTIALS_DICT_DELEGATED_USER_KEY = "google_drive_delegated_user"
-SCOPES = [
-    "https://www.googleapis.com/auth/drive.readonly",
-    "https://www.googleapis.com/auth/drive.metadata.readonly",
-]
+
+BASE_SCOPES = ["https://www.googleapis.com/auth/drive.readonly"]
+FETCH_PERMISSIONS_SCOPES = ["https://www.googleapis.com/auth/drive.metadata.readonly"]
+FETCH_GROUPS_SCOPES = ["https://www.googleapis.com/auth/cloud-identity.groups.readonly"]

backend/danswer/connectors/guru/connector.py

@@ -103,6 +103,10 @@ class GuruConnector(LoadConnector, PollConnector):
                     # In UI it's called Folders
                     metadata_dict["folders"] = boards
 
+                collection = card.get("collection", {})
+                if collection:
+                    metadata_dict["collection_name"] = collection.get("name", "")
+
                 owner = card.get("owner", {})
                 author = None
                 if owner:

backend/danswer/connectors/models.py

@@ -113,6 +113,9 @@ class DocumentBase(BaseModel):
     # The default title is semantic_identifier though unless otherwise specified
     title: str | None = None
     from_ingestion_api: bool = False
+    # Anything else that may be useful that is specific to this particular connector type that other
+    # parts of the code may need. If you're unsure, this can be left as None
+    additional_info: Any = None
 
     def get_title_for_document_index(
         self,
@@ -166,6 +169,36 @@ class Document(DocumentBase):
         )
 
 
+class DocumentErrorSummary(BaseModel):
+    id: str
+    semantic_id: str
+    section_link: str | None
+
+    @classmethod
+    def from_document(cls, doc: Document) -> "DocumentErrorSummary":
+        section_link = doc.sections[0].link if len(doc.sections) > 0 else None
+        return cls(
+            id=doc.id, semantic_id=doc.semantic_identifier, section_link=section_link
+        )
+
+    @classmethod
+    def from_dict(cls, data: dict) -> "DocumentErrorSummary":
+        return cls(
+            id=str(data.get("id")),
+            semantic_id=str(data.get("semantic_id")),
+            section_link=str(data.get("section_link")),
+        )
+
+    def to_dict(self) -> dict[str, str | None]:
+        return {
+            "id": self.id,
+            "semantic_id": self.semantic_id,
+            "section_link": self.section_link,
+        }
+
+
 class IndexAttemptMetadata(BaseModel):
+    batch_num: int | None = None
+    num_exceptions: int = 0
     connector_id: int
     credential_id: int

backend/danswer/connectors/notion/connector.py

@@ -237,6 +237,14 @@ class NotionConnector(LoadConnector, PollConnector):
                     )
                     continue
 
+                if result_type == "external_object_instance_page":
+                    logger.warning(
+                        f"Skipping 'external_object_instance_page' ('{result_block_id}') for base block '{base_block_id}': "
+                        f"Notion API does not currently support reading external blocks (as of 24/07/03) "
+                        f"(discussion: https://github.com/danswer-ai/danswer/issues/1761)"
+                    )
+                    continue
+
                 cur_result_text_arr = []
                 if "rich_text" in result_obj:
                     for rich_text in result_obj["rich_text"]:

backend/danswer/connectors/productboard/connector.py

@@ -98,6 +98,15 @@ class ProductboardConnector(PollConnector):
             owner = self._get_owner_email(feature)
             experts = [BasicExpertInfo(email=owner)] if owner else None
 
+            metadata: dict[str, str | list[str]] = {}
+            entity_type = feature.get("type", "feature")
+            if entity_type:
+                metadata["entity_type"] = str(entity_type)
+
+            status = feature.get("status", {}).get("name")
+            if status:
+                metadata["status"] = str(status)
+
             yield Document(
                 id=feature["id"],
                 sections=[
@@ -110,10 +119,7 @@ class ProductboardConnector(PollConnector):
                 source=DocumentSource.PRODUCTBOARD,
                 doc_updated_at=time_str_to_utc(feature["updatedAt"]),
                 primary_owners=experts,
-                metadata={
-                    "entity_type": feature["type"],
-                    "status": feature["status"]["name"],
-                },
+                metadata=metadata,
             )
 
     def _get_components(self) -> Generator[Document, None, None]:
@@ -174,6 +180,12 @@ class ProductboardConnector(PollConnector):
             owner = self._get_owner_email(objective)
             experts = [BasicExpertInfo(email=owner)] if owner else None
 
+            metadata: dict[str, str | list[str]] = {
+                "entity_type": "objective",
+            }
+            if objective.get("state"):
+                metadata["state"] = str(objective["state"])
+
             yield Document(
                 id=objective["id"],
                 sections=[
@@ -186,10 +198,7 @@ class ProductboardConnector(PollConnector):
                 source=DocumentSource.PRODUCTBOARD,
                 doc_updated_at=time_str_to_utc(objective["updatedAt"]),
                 primary_owners=experts,
-                metadata={
-                    "entity_type": "release",
-                    "state": objective["state"],
-                },
+                metadata=metadata,
             )
 
     def _is_updated_at_out_of_time_range(

backend/danswer/connectors/sharepoint/connector.py

@@ -25,7 +25,6 @@ from danswer.connectors.models import Section
 from danswer.file_processing.extract_file_text import extract_file_text
 from danswer.utils.logger import setup_logger
 
-
 logger = setup_logger()
 
 
@@ -137,7 +136,7 @@ class SharepointConnector(LoadConnector, PollConnector):
                     .execute_query()
                 ]
         else:
-            sites = self.graph_client.sites.get().execute_query()
+            sites = self.graph_client.sites.get_all().execute_query()
             self.site_data = [
                 SiteData(url=None, folder=None, sites=sites, driveitems=[])
             ]

backend/danswer/connectors/slack/connector.py

@@ -29,6 +29,7 @@ from danswer.connectors.slack.utils import make_slack_api_rate_limited
 from danswer.connectors.slack.utils import SlackTextCleaner
 from danswer.utils.logger import setup_logger
 
+
 logger = setup_logger()
 
 

backend/danswer/connectors/web/connector.py

@@ -1,6 +1,8 @@
 import io
 import ipaddress
 import socket
+from datetime import datetime
+from datetime import timezone
 from enum import Enum
 from typing import Any
 from typing import cast
@@ -27,7 +29,7 @@ from danswer.connectors.interfaces import GenerateDocumentsOutput
 from danswer.connectors.interfaces import LoadConnector
 from danswer.connectors.models import Document
 from danswer.connectors.models import Section
-from danswer.file_processing.extract_file_text import pdf_to_text
+from danswer.file_processing.extract_file_text import read_pdf_file
 from danswer.file_processing.html_utils import web_html_cleanup
 from danswer.utils.logger import setup_logger
 from danswer.utils.sitemap import list_pages_for_site
@@ -84,6 +86,20 @@ def check_internet_connection(url: str) -> None:
     try:
         response = requests.get(url, timeout=3)
         response.raise_for_status()
+    except requests.exceptions.HTTPError as e:
+        # Extract status code from the response, defaulting to -1 if response is None
+        status_code = e.response.status_code if e.response is not None else -1
+        error_msg = {
+            400: "Bad Request",
+            401: "Unauthorized",
+            403: "Forbidden",
+            404: "Not Found",
+            500: "Internal Server Error",
+            502: "Bad Gateway",
+            503: "Service Unavailable",
+            504: "Gateway Timeout",
+        }.get(status_code, "HTTP Error")
+        raise Exception(f"{error_msg} ({status_code}) for {url} - {e}")
     except requests.exceptions.SSLError as e:
         cause = (
             e.args[0].reason
@@ -91,8 +107,8 @@ def check_internet_connection(url: str) -> None:
             else e.args
         )
         raise Exception(f"SSL error {str(cause)}")
-    except (requests.RequestException, ValueError):
-        raise Exception(f"Unable to reach {url} - check your internet connection")
+    except (requests.RequestException, ValueError) as e:
+        raise Exception(f"Unable to reach {url} - check your internet connection: {e}")
 
 
 def is_valid_url(url: str) -> bool:
@@ -189,6 +205,15 @@ def _read_urls_file(location: str) -> list[str]:
     return urls
 
 
+def _get_datetime_from_last_modified_header(last_modified: str) -> datetime | None:
+    try:
+        return datetime.strptime(last_modified, "%a, %d %b %Y %H:%M:%S %Z").replace(
+            tzinfo=timezone.utc
+        )
+    except (ValueError, TypeError):
+        return None
+
+
 class WebConnector(LoadConnector):
     def __init__(
         self,
@@ -271,7 +296,10 @@ class WebConnector(LoadConnector):
                 if current_url.split(".")[-1] == "pdf":
                     # PDF files are not checked for links
                     response = requests.get(current_url)
-                    page_text = pdf_to_text(file=io.BytesIO(response.content))
+                    page_text, metadata = read_pdf_file(
+                        file=io.BytesIO(response.content)
+                    )
+                    last_modified = response.headers.get("Last-Modified")
 
                     doc_batch.append(
                         Document(
@@ -279,13 +307,23 @@ class WebConnector(LoadConnector):
                             sections=[Section(link=current_url, text=page_text)],
                             source=DocumentSource.WEB,
                             semantic_identifier=current_url.split("/")[-1],
-                            metadata={},
+                            metadata=metadata,
+                            doc_updated_at=_get_datetime_from_last_modified_header(
+                                last_modified
+                            )
+                            if last_modified
+                            else None,
                         )
                     )
                     continue
 
                 page = context.new_page()
                 page_response = page.goto(current_url)
+                last_modified = (
+                    page_response.header_value("Last-Modified")
+                    if page_response
+                    else None
+                )
                 final_page = page.url
                 if final_page != current_url:
                     logger.info(f"Redirected to {final_page}")
@@ -321,6 +359,11 @@ class WebConnector(LoadConnector):
                         source=DocumentSource.WEB,
                         semantic_identifier=parsed_html.title or current_url,
                         metadata={},
+                        doc_updated_at=_get_datetime_from_last_modified_header(
+                            last_modified
+                        )
+                        if last_modified
+                        else None,
                     )
                 )
 

backend/danswer/connectors/zendesk/connector.py

@@ -3,6 +3,7 @@ from typing import Any
 import requests
 from retry import retry
 from zenpy import Zenpy  # type: ignore
+from zenpy.lib.api_objects import Ticket  # type: ignore
 from zenpy.lib.api_objects.help_centre_objects import Article  # type: ignore
 
 from danswer.configs.app_configs import INDEX_BATCH_SIZE
@@ -59,10 +60,15 @@ class ZendeskClientNotSetUpError(PermissionError):
 
 
 class ZendeskConnector(LoadConnector, PollConnector):
-    def __init__(self, batch_size: int = INDEX_BATCH_SIZE) -> None:
+    def __init__(
+        self,
+        batch_size: int = INDEX_BATCH_SIZE,
+        content_type: str = "articles",
+    ) -> None:
         self.batch_size = batch_size
         self.zendesk_client: Zenpy | None = None
         self.content_tags: dict[str, str] = {}
+        self.content_type = content_type
 
     @retry(tries=3, delay=2, backoff=2)
     def _set_content_tags(
@@ -122,16 +128,86 @@ class ZendeskConnector(LoadConnector, PollConnector):
     def load_from_state(self) -> GenerateDocumentsOutput:
         return self.poll_source(None, None)
 
+    def _ticket_to_document(self, ticket: Ticket) -> Document:
+        if self.zendesk_client is None:
+            raise ZendeskClientNotSetUpError()
+
+        owner = None
+        if ticket.requester and ticket.requester.name and ticket.requester.email:
+            owner = [
+                BasicExpertInfo(
+                    display_name=ticket.requester.name, email=ticket.requester.email
+                )
+            ]
+        update_time = time_str_to_utc(ticket.updated_at) if ticket.updated_at else None
+
+        metadata: dict[str, str | list[str]] = {}
+        if ticket.status is not None:
+            metadata["status"] = ticket.status
+        if ticket.priority is not None:
+            metadata["priority"] = ticket.priority
+        if ticket.tags:
+            metadata["tags"] = ticket.tags
+        if ticket.type is not None:
+            metadata["ticket_type"] = ticket.type
+
+        # Fetch comments for the ticket
+        comments = self.zendesk_client.tickets.comments(ticket=ticket)
+
+        # Combine all comments into a single text
+        comments_text = "\n\n".join(
+            [
+                f"Comment{f' by {comment.author.name}' if comment.author and comment.author.name else ''}"
+                f"{f' at {comment.created_at}' if comment.created_at else ''}:\n{comment.body}"
+                for comment in comments
+                if comment.body
+            ]
+        )
+
+        # Combine ticket description and comments
+        description = (
+            ticket.description
+            if hasattr(ticket, "description") and ticket.description
+            else ""
+        )
+        full_text = f"Ticket Description:\n{description}\n\nComments:\n{comments_text}"
+
+        # Extract subdomain from ticket.url
+        subdomain = ticket.url.split("//")[1].split(".zendesk.com")[0]
+
+        # Build the html url for the ticket
+        ticket_url = f"https://{subdomain}.zendesk.com/agent/tickets/{ticket.id}"
+
+        return Document(
+            id=f"zendesk_ticket_{ticket.id}",
+            sections=[Section(link=ticket_url, text=full_text)],
+            source=DocumentSource.ZENDESK,
+            semantic_identifier=f"Ticket #{ticket.id}: {ticket.subject or 'No Subject'}",
+            doc_updated_at=update_time,
+            primary_owners=owner,
+            metadata=metadata,
+        )
+
     def poll_source(
         self, start: SecondsSinceUnixEpoch | None, end: SecondsSinceUnixEpoch | None
     ) -> GenerateDocumentsOutput:
         if self.zendesk_client is None:
             raise ZendeskClientNotSetUpError()
 
+        if self.content_type == "articles":
+            yield from self._poll_articles(start)
+        elif self.content_type == "tickets":
+            yield from self._poll_tickets(start)
+        else:
+            raise ValueError(f"Unsupported content_type: {self.content_type}")
+
+    def _poll_articles(
+        self, start: SecondsSinceUnixEpoch | None
+    ) -> GenerateDocumentsOutput:
         articles = (
-            self.zendesk_client.help_center.articles(cursor_pagination=True)
+            self.zendesk_client.help_center.articles(cursor_pagination=True)  # type: ignore
             if start is None
-            else self.zendesk_client.help_center.articles.incremental(
+            else self.zendesk_client.help_center.articles.incremental(  # type: ignore
                 start_time=int(start)
             )
         )
@@ -155,9 +231,43 @@ class ZendeskConnector(LoadConnector, PollConnector):
         if doc_batch:
             yield doc_batch
 
+    def _poll_tickets(
+        self, start: SecondsSinceUnixEpoch | None
+    ) -> GenerateDocumentsOutput:
+        if self.zendesk_client is None:
+            raise ZendeskClientNotSetUpError()
+
+        ticket_generator = self.zendesk_client.tickets.incremental(start_time=start)
+
+        while True:
+            doc_batch = []
+            for _ in range(self.batch_size):
+                try:
+                    ticket = next(ticket_generator)
+
+                    # Check if the ticket status is deleted and skip it if so
+                    if ticket.status == "deleted":
+                        continue
+
+                    doc_batch.append(self._ticket_to_document(ticket))
+
+                    if len(doc_batch) >= self.batch_size:
+                        yield doc_batch
+                        doc_batch.clear()
+
+                except StopIteration:
+                    # No more tickets to process
+                    if doc_batch:
+                        yield doc_batch
+                    return
+
+            if doc_batch:
+                yield doc_batch
+
 
 if __name__ == "__main__":
     import os
+
     import time
 
     connector = ZendeskConnector()

backend/danswer/connectors/zulip/schemas.py

@@ -3,6 +3,7 @@ from typing import List
 from typing import Optional
 
 from pydantic import BaseModel
+from pydantic import Field
 
 
 class Message(BaseModel):
@@ -18,11 +19,11 @@ class Message(BaseModel):
     sender_realm_str: str
     subject: str
     topic_links: Optional[List[Any]] = None
-    last_edit_timestamp: Optional[int] = None
-    edit_history: Any
+    last_edit_timestamp: Optional[int]
+    edit_history: Any = None
     reactions: List[Any]
     submessages: List[Any]
-    flags: List[str] = []
+    flags: List[str] = Field(default_factory=list)
     display_recipient: Optional[str] = None
     type: Optional[str] = None
     stream_id: int
@@ -39,4 +40,4 @@ class GetMessagesResponse(BaseModel):
     found_newest: Optional[bool] = None
     history_limited: Optional[bool] = None
     anchor: Optional[str] = None
-    messages: List[Message] = []
+    messages: List[Message] = Field(default_factory=list)

backend/danswer/danswerbot/slack/blocks.py

@@ -25,7 +25,6 @@ from danswer.danswerbot.slack.constants import DISLIKE_BLOCK_ACTION_ID
 from danswer.danswerbot.slack.constants import FEEDBACK_DOC_BUTTON_BLOCK_ACTION_ID
 from danswer.danswerbot.slack.constants import FOLLOWUP_BUTTON_ACTION_ID
 from danswer.danswerbot.slack.constants import FOLLOWUP_BUTTON_RESOLVED_ACTION_ID
-from danswer.danswerbot.slack.constants import GENERATE_ANSWER_BUTTON_ACTION_ID
 from danswer.danswerbot.slack.constants import IMMEDIATE_RESOLVED_BUTTON_ACTION_ID
 from danswer.danswerbot.slack.constants import LIKE_BLOCK_ACTION_ID
 from danswer.danswerbot.slack.icons import source_to_github_img_link
@@ -360,22 +359,6 @@ def build_quotes_block(
     return [SectionBlock(text="*Relevant Snippets*\n" + "\n".join(quote_lines))]
 
 
-def build_standard_answer_blocks(
-    answer_message: str,
-) -> list[Block]:
-    generate_button_block = ButtonElement(
-        action_id=GENERATE_ANSWER_BUTTON_ACTION_ID,
-        text="Generate Full Answer",
-    )
-    answer_block = SectionBlock(text=answer_message)
-    return [
-        answer_block,
-        ActionsBlock(
-            elements=[generate_button_block],
-        ),
-    ]
-
-
 def build_qa_response_blocks(
     message_id: int | None,
     answer: str | None,