nit

* update assistant logic

* quick nit

* k

* fix "featured" logic

* Small tweaks

* k

---------

Co-authored-by: Weves <chrisweaver101@gmail.com>

.github/pull_request_template.md

@@ -1,11 +1,14 @@
 ## Description
+
 [Provide a brief description of the changes in this PR]
 
-
 ## How Has This Been Tested?
+
 [Describe the tests you ran to verify your changes]
 
-
 ## Backporting (check the box to trigger backport action)
+
 Note: You have to check that the action passes, otherwise resolve the conflicts manually and tag the patches.
+
 - [ ] This PR should be backported (make sure to check that the backport attempt succeeds)
+- [ ] [Optional] Override Linear Check

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

@@ -67,6 +67,8 @@ jobs:
             NEXT_PUBLIC_SENTRY_DSN=${{ secrets.SENTRY_DSN }}
             NEXT_PUBLIC_GTM_ENABLED=true
             NEXT_PUBLIC_FORGOT_PASSWORD_ENABLED=true
+            NEXT_PUBLIC_INCLUDE_ERROR_POPUP_SUPPORT_LINK=true
+            NODE_OPTIONS=--max-old-space-size=8192
           # needed due to weird interactions with the builds for different platforms
           no-cache: true
           labels: ${{ steps.meta.outputs.labels }}

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

@@ -4,6 +4,9 @@ on:
   push:
     tags:
       - "*"
+    paths:
+      - 'backend/model_server/**'
+      - 'backend/Dockerfile.model_server'
 
 env:
   REGISTRY_IMAGE: ${{ contains(github.ref_name, 'cloud') && 'onyxdotapp/onyx-model-server-cloud' || 'onyxdotapp/onyx-model-server' }}

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

@@ -60,6 +60,8 @@ jobs:
           push: true
           build-args: |
             ONYX_VERSION=${{ github.ref_name }}
+            NODE_OPTIONS=--max-old-space-size=8192
+
           # needed due to weird interactions with the builds for different platforms
           no-cache: true
           labels: ${{ steps.meta.outputs.labels }}

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

@@ -8,6 +8,8 @@ on: push
 env:
   OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
   SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}
+  GEN_AI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+  MOCK_LLM_RESPONSE: true
 
 jobs:
   playwright-tests:

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

@@ -21,10 +21,10 @@ jobs:
     - name: Set up Helm
       uses: azure/setup-helm@v4.2.0
       with:
-        version: v3.14.4
+        version: v3.17.0
       
     - name: Set up chart-testing
-      uses: helm/chart-testing-action@v2.6.1
+      uses: helm/chart-testing-action@v2.7.0
 
     # even though we specify chart-dirs in ct.yaml, it isn't used by ct for the list-changed command...
     - name: Run chart-testing (list-changed)
@@ -37,22 +37,6 @@ jobs:
           echo "changed=true" >> "$GITHUB_OUTPUT"
         fi
 
-#     rkuo: I don't think we need python?
-#     - name: Set up Python
-#       uses: actions/setup-python@v5
-#       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
-
     # lint all charts if any changes were detected
     - name: Run chart-testing (lint)
       if: steps.list-changed.outputs.changed == 'true'
@@ -62,7 +46,7 @@ jobs:
 
     - name: Create kind cluster
       if: steps.list-changed.outputs.changed == 'true'
-      uses: helm/kind-action@v1.10.0
+      uses: helm/kind-action@v1.12.0
 
     - name: Run chart-testing (install)
       if: steps.list-changed.outputs.changed == 'true'

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

@@ -94,16 +94,19 @@ jobs:
           cd deployment/docker_compose
           ENABLE_PAID_ENTERPRISE_EDITION_FEATURES=true \
           MULTI_TENANT=true \
-          AUTH_TYPE=basic \
+          AUTH_TYPE=cloud \
           REQUIRE_EMAIL_VERIFICATION=false \
           DISABLE_TELEMETRY=true \
           IMAGE_TAG=test \
-          docker compose -f docker-compose.dev.yml -p danswer-stack up -d
+          DEV_MODE=true \
+          docker compose -f docker-compose.multitenant-dev.yml -p danswer-stack up -d
         id: start_docker_multi_tenant
 
       # In practice, `cloud` Auth type would require OAUTH credentials to be set.
       - name: Run Multi-Tenant Integration Tests
         run: |
+          echo "Waiting for 3 minutes to ensure API server is ready..."
+          sleep 180
           echo "Running integration tests..."
           docker run --rm --network danswer-stack_default \
             --name test-runner \
@@ -119,6 +122,10 @@ jobs:
             -e TEST_WEB_HOSTNAME=test-runner \
             -e AUTH_TYPE=cloud \
             -e MULTI_TENANT=true \
+            -e REQUIRE_EMAIL_VERIFICATION=false \
+            -e DISABLE_TELEMETRY=true \
+            -e IMAGE_TAG=test \
+            -e DEV_MODE=true \
             onyxdotapp/onyx-integration:test \
             /app/tests/integration/multitenant_tests
         continue-on-error: true
@@ -126,17 +133,17 @@ jobs:
 
       - name: Check multi-tenant test results
         run: |
-          if [ ${{ steps.run_tests.outcome }} == 'failure' ]; then
-            echo "Integration tests failed. Exiting with error."
+          if [ ${{ steps.run_multitenant_tests.outcome }} == 'failure' ]; then
+            echo "Multi-tenant integration tests failed. Exiting with error."
             exit 1
           else
-            echo "All integration tests passed successfully."
+            echo "All multi-tenant integration tests passed successfully."
           fi
 
       - name: Stop multi-tenant Docker containers
         run: |
           cd deployment/docker_compose
-          docker compose -f docker-compose.dev.yml -p danswer-stack down -v
+          docker compose -f docker-compose.multitenant-dev.yml -p danswer-stack down -v
 
       - name: Start Docker containers
         run: |
@@ -216,27 +223,30 @@ jobs:
             echo "All integration tests passed successfully."
           fi
 
-      # save before stopping the containers so the logs can be captured
-      - name: Save Docker logs
-        if: success() || failure()
+      # ------------------------------------------------------------
+      # Always gather logs BEFORE "down":
+      - name: Dump API server logs
+        if: always()
         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
+          docker compose -f docker-compose.dev.yml -p danswer-stack logs --no-color api_server > $GITHUB_WORKSPACE/api_server.log || true
 
-      - name: Stop Docker containers
+      - name: Dump all-container logs (optional)
+        if: always()
         run: |
           cd deployment/docker_compose
-          docker compose -f docker-compose.dev.yml -p danswer-stack down -v
+          docker compose -f docker-compose.dev.yml -p danswer-stack logs --no-color > $GITHUB_WORKSPACE/docker-compose.log || true
 
       - name: Upload logs
-        if: success() || failure()
+        if: always()
         uses: actions/upload-artifact@v4
         with:
-          name: docker-logs
+          name: docker-all-logs
           path: ${{ github.workspace }}/docker-compose.log
+      # ------------------------------------------------------------
 
       - name: Stop Docker containers
+        if: always()
         run: |
           cd deployment/docker_compose
           docker compose -f docker-compose.dev.yml -p danswer-stack down -v

.github/workflows/pr-linear-check.yml

@@ -0,0 +1,29 @@
+name: Ensure PR references Linear
+
+on:
+  pull_request:
+    types: [opened, edited, reopened, synchronize]
+
+jobs:
+  linear-check:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Check PR body for Linear link or override
+        env:
+          PR_BODY: ${{ github.event.pull_request.body }}
+        run: |
+          # Looking for "https://linear.app" in the body
+          if echo "$PR_BODY" | grep -qE "https://linear\.app"; then
+            echo "Found a Linear link. Check passed."
+            exit 0
+          fi
+
+          # Looking for a checked override: "[x] Override Linear Check"
+          if echo "$PR_BODY" | grep -q "\[x\].*Override Linear Check"; then
+            echo "Override box is checked. Check passed."
+            exit 0
+          fi
+
+          # Otherwise, fail the run
+          echo "No Linear link or override found in the PR description."
+          exit 1

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

@@ -39,6 +39,12 @@ env:
   AIRTABLE_TEST_TABLE_ID: ${{ secrets.AIRTABLE_TEST_TABLE_ID }}
   AIRTABLE_TEST_TABLE_NAME: ${{ secrets.AIRTABLE_TEST_TABLE_NAME }}
   AIRTABLE_ACCESS_TOKEN: ${{ secrets.AIRTABLE_ACCESS_TOKEN }}
+  # Sharepoint
+  SHAREPOINT_CLIENT_ID: ${{ secrets.SHAREPOINT_CLIENT_ID }}
+  SHAREPOINT_CLIENT_SECRET: ${{ secrets.SHAREPOINT_CLIENT_SECRET }}
+  SHAREPOINT_CLIENT_DIRECTORY_ID: ${{ secrets.SHAREPOINT_CLIENT_DIRECTORY_ID }}
+  SHAREPOINT_SITE: ${{ secrets.SHAREPOINT_SITE }}
+
 jobs:
   connectors-check:
     # See https://runs-on.com/runners/linux/

.vscode/env_template.txt

@@ -29,6 +29,7 @@ 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>
+OPENAI_API_KEY=<REPLACE THIS>
 # 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

.vscode/launch.template.jsonc

@@ -28,6 +28,7 @@
 		  		"Celery heavy", 
 		  		"Celery indexing", 
 		  		"Celery beat",
+                "Celery monitoring",
             ],
 			"presentation": {
 				 "group": "1",
@@ -51,7 +52,8 @@
 		  		"Celery light", 
 		  		"Celery heavy", 
 		  		"Celery indexing", 
-		  		"Celery beat"
+		  		"Celery beat",
+                "Celery monitoring",
 		  	],
 			"presentation": {
 				 "group": "1",
@@ -269,6 +271,31 @@
 			 },
             "consoleTitle": "Celery indexing Console"
         },
+        {
+            "name": "Celery monitoring",
+            "type": "debugpy",
+            "request": "launch",
+            "module": "celery",
+            "cwd": "${workspaceFolder}/backend",
+            "envFile": "${workspaceFolder}/.vscode/.env",
+            "env": {},
+            "args": [
+                "-A",
+                "onyx.background.celery.versioned_apps.monitoring",
+                "worker",
+                "--pool=solo",
+                "--concurrency=1",
+                "--prefetch-multiplier=1",
+                "--loglevel=INFO",
+                "--hostname=monitoring@%n",
+                "-Q",
+                "monitoring",
+            ],
+            "presentation": {
+				 "group": "2",
+			 },
+            "consoleTitle": "Celery monitoring Console"
+        },
         {
             "name": "Celery beat",
             "type": "debugpy",

CONTRIBUTING_VSCODE.md

@@ -17,9 +17,10 @@ Before starting, make sure the Docker Daemon is running.
 1. Open the Debug view in VSCode (Cmd+Shift+D on macOS)
 2. From the dropdown at the top, select "Clear and Restart External Volumes and Containers" and press the green play button
 3. From the dropdown at the top, select "Run All Onyx Services" and press the green play button
-4. Now, you can navigate to onyx in your browser (default is http://localhost:3000) and start using the app
-5. You can set breakpoints by clicking to the left of line numbers to help debug while the app is running
-6. Use the debug toolbar to step through code, inspect variables, etc.
+4. CD into web, run "npm i" followed by npm run dev.
+5. Now, you can navigate to onyx in your browser (default is http://localhost:3000) and start using the app
+6. You can set breakpoints by clicking to the left of line numbers to help debug while the app is running
+7. Use the debug toolbar to step through code, inspect variables, etc.
 
 ## Features
 

README.md

@@ -119,12 +119,12 @@ There are two editions of Onyx:
   - Whitelabeling
   - API key authentication
   - Encryption of secrets
-  - Any many more! Checkout [our website](https://www.onyx.app/) for the latest.
+  - And many more! Checkout [our website](https://www.onyx.app/) for the latest.
 
 To try the Onyx Enterprise Edition:
 
 1. Checkout our [Cloud product](https://cloud.onyx.app/signup).
-2. For self-hosting, contact us at [founders@onyx.app](mailto:founders@onyx.app) or book a call with us on our [Cal](https://cal.com/team/danswer/founders).
+2. For self-hosting, contact us at [founders@onyx.app](mailto:founders@onyx.app) or book a call with us on our [Cal](https://cal.com/team/onyx/founders).
 
 ## 💡 Contributing
 
@@ -133,3 +133,4 @@ Looking to contribute? Please check out the [Contribution Guide](CONTRIBUTING.md
 ## ⭐Star History
 
 [![Star History Chart](https://api.star-history.com/svg?repos=onyx-dot-app/onyx&type=Date)](https://star-history.com/#onyx-dot-app/onyx&Date)
+

backend/Dockerfile

@@ -9,8 +9,10 @@ founders@onyx.app for more information. Please visit https://github.com/onyx-dot
 
 # Default ONYX_VERSION, typically overriden during builds by GitHub Actions.
 ARG ONYX_VERSION=0.8-dev
+# DO_NOT_TRACK is used to disable telemetry for Unstructured
 ENV ONYX_VERSION=${ONYX_VERSION} \
-    DANSWER_RUNNING_IN_DOCKER="true"
+    DANSWER_RUNNING_IN_DOCKER="true" \
+    DO_NOT_TRACK="true"
 
 
 RUN echo "ONYX_VERSION: ${ONYX_VERSION}"
@@ -99,7 +101,8 @@ COPY ./alembic_tenants /app/alembic_tenants
 COPY ./alembic.ini /app/alembic.ini
 COPY supervisord.conf /usr/etc/supervisord.conf
 
-# Escape hatch
+# Escape hatch scripts
+COPY ./scripts/debugging /app/scripts/debugging
 COPY ./scripts/force_delete_connector_by_id.py /app/scripts/force_delete_connector_by_id.py
 
 # Put logo in assets

backend/alembic/versions/027381bce97c_add_shortcut_option_for_users.py

@@ -20,7 +20,7 @@ def upgrade() -> None:
     op.add_column(
         "user",
         sa.Column(
-            "shortcut_enabled", sa.Boolean(), nullable=False, server_default="true"
+            "shortcut_enabled", sa.Boolean(), nullable=False, server_default="false"
         ),
     )
 

backend/alembic/versions/0f7ff6d75b57_add_index_to_index_attempt_time_created.py

@@ -0,0 +1,36 @@
+"""add index to index_attempt.time_created
+
+Revision ID: 0f7ff6d75b57
+Revises: 369644546676
+Create Date: 2025-01-10 14:01:14.067144
+
+"""
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision = "0f7ff6d75b57"
+down_revision = "fec3db967bf7"
+branch_labels: None = None
+depends_on: None = None
+
+
+def upgrade() -> None:
+    op.create_index(
+        op.f("ix_index_attempt_status"),
+        "index_attempt",
+        ["status"],
+        unique=False,
+    )
+
+    op.create_index(
+        op.f("ix_index_attempt_time_created"),
+        "index_attempt",
+        ["time_created"],
+        unique=False,
+    )
+
+
+def downgrade() -> None:
+    op.drop_index(op.f("ix_index_attempt_time_created"), table_name="index_attempt")
+
+    op.drop_index(op.f("ix_index_attempt_status"), table_name="index_attempt")

backend/alembic/versions/1adf5ea20d2b_agent_doc_result_col.py

@@ -1,29 +0,0 @@
-"""agent_doc_result_col
-
-Revision ID: 1adf5ea20d2b
-Revises: e9cf2bd7baed
-Create Date: 2025-01-05 13:14:58.344316
-
-"""
-from alembic import op
-import sqlalchemy as sa
-from sqlalchemy.dialects import postgresql
-
-# revision identifiers, used by Alembic.
-revision = "1adf5ea20d2b"
-down_revision = "e9cf2bd7baed"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # Add the new column with JSONB type
-    op.add_column(
-        "sub_question",
-        sa.Column("sub_question_doc_results", postgresql.JSONB(), nullable=True),
-    )
-
-
-def downgrade() -> None:
-    # Drop the column
-    op.drop_column("sub_question", "sub_question_doc_results")

backend/alembic/versions/2cdeff6d8c93_set_built_in_to_default.py

@@ -0,0 +1,32 @@
+"""set built in to default
+
+Revision ID: 2cdeff6d8c93
+Revises: f5437cc136c5
+Create Date: 2025-02-11 14:57:51.308775
+
+"""
+from alembic import op
+
+
+# revision identifiers, used by Alembic.
+revision = "2cdeff6d8c93"
+down_revision = "f5437cc136c5"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # Prior to this migration / point in the codebase history,
+    # built in personas were implicitly treated as default personas (with no option to change this)
+    # This migration makes that explicit
+    op.execute(
+        """
+        UPDATE persona
+        SET is_default_persona = TRUE
+        WHERE builtin_persona = TRUE
+    """
+    )
+
+
+def downgrade() -> None:
+    pass

backend/alembic/versions/2f80c6a2550f_add_chat_session_specific_temperature_.py

@@ -0,0 +1,36 @@
+"""add chat session specific temperature override
+
+Revision ID: 2f80c6a2550f
+Revises: 33ea50e88f24
+Create Date: 2025-01-31 10:30:27.289646
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = "2f80c6a2550f"
+down_revision = "33ea50e88f24"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "chat_session", sa.Column("temperature_override", sa.Float(), nullable=True)
+    )
+    op.add_column(
+        "user",
+        sa.Column(
+            "temperature_override_enabled",
+            sa.Boolean(),
+            nullable=False,
+            server_default=sa.false(),
+        ),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("chat_session", "temperature_override")
+    op.drop_column("user", "temperature_override_enabled")

backend/alembic/versions/33ea50e88f24_foreign_key_input_prompts.py

@@ -0,0 +1,80 @@
+"""foreign key input prompts
+
+Revision ID: 33ea50e88f24
+Revises: a6df6b88ef81
+Create Date: 2025-01-29 10:54:22.141765
+
+"""
+from alembic import op
+
+
+# revision identifiers, used by Alembic.
+revision = "33ea50e88f24"
+down_revision = "a6df6b88ef81"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # Safely drop constraints if exists
+    op.execute(
+        """
+        ALTER TABLE inputprompt__user
+        DROP CONSTRAINT IF EXISTS inputprompt__user_input_prompt_id_fkey
+        """
+    )
+    op.execute(
+        """
+        ALTER TABLE inputprompt__user
+        DROP CONSTRAINT IF EXISTS inputprompt__user_user_id_fkey
+        """
+    )
+
+    # Recreate with ON DELETE CASCADE
+    op.create_foreign_key(
+        "inputprompt__user_input_prompt_id_fkey",
+        "inputprompt__user",
+        "inputprompt",
+        ["input_prompt_id"],
+        ["id"],
+        ondelete="CASCADE",
+    )
+
+    op.create_foreign_key(
+        "inputprompt__user_user_id_fkey",
+        "inputprompt__user",
+        "user",
+        ["user_id"],
+        ["id"],
+        ondelete="CASCADE",
+    )
+
+
+def downgrade() -> None:
+    # Drop the new FKs with ondelete
+    op.drop_constraint(
+        "inputprompt__user_input_prompt_id_fkey",
+        "inputprompt__user",
+        type_="foreignkey",
+    )
+    op.drop_constraint(
+        "inputprompt__user_user_id_fkey",
+        "inputprompt__user",
+        type_="foreignkey",
+    )
+
+    # Recreate them without cascading
+    op.create_foreign_key(
+        "inputprompt__user_input_prompt_id_fkey",
+        "inputprompt__user",
+        "inputprompt",
+        ["input_prompt_id"],
+        ["id"],
+    )
+    op.create_foreign_key(
+        "inputprompt__user_user_id_fkey",
+        "inputprompt__user",
+        "user",
+        ["user_id"],
+        ["id"],
+    )

backend/alembic/versions/3c6531f32351_add_back_input_prompts.py

@@ -41,6 +41,7 @@ def upgrade() -> None:
         sa.Column(
             "user_id", fastapi_users_db_sqlalchemy.generics.GUID(), nullable=False
         ),
+        sa.Column("disabled", sa.Boolean(), nullable=False, default=False),
         sa.ForeignKeyConstraint(
             ["input_prompt_id"],
             ["inputprompt.id"],

backend/alembic/versions/47e5bef3a1d7_add_persona_categories.py

@@ -40,6 +40,6 @@ def upgrade() -> None:
 
 
 def downgrade() -> None:
-    op.drop_constraint("fk_persona_category", "persona", type_="foreignkey")
+    op.drop_constraint("persona_category_id_fkey", "persona", type_="foreignkey")
     op.drop_column("persona", "category_id")
     op.drop_table("persona_category")

backend/alembic/versions/4d58345da04a_lowercase_user_emails.py

@@ -0,0 +1,37 @@
+"""lowercase_user_emails
+
+Revision ID: 4d58345da04a
+Revises: f1ca58b2f2ec
+Create Date: 2025-01-29 07:48:46.784041
+
+"""
+from alembic import op
+from sqlalchemy.sql import text
+
+
+# revision identifiers, used by Alembic.
+revision = "4d58345da04a"
+down_revision = "f1ca58b2f2ec"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # Get database connection
+    connection = op.get_bind()
+
+    # Update all user emails to lowercase
+    connection.execute(
+        text(
+            """
+            UPDATE "user"
+            SET email = LOWER(email)
+            WHERE email != LOWER(email)
+            """
+        )
+    )
+
+
+def downgrade() -> None:
+    # Cannot restore original case of emails
+    pass

backend/alembic/versions/4ee1287bd26a_add_multiple_slack_bot_support.py

@@ -5,7 +5,6 @@ Revises: 47e5bef3a1d7
 Create Date: 2024-11-06 13:15:53.302644
 
 """
-import logging
 from typing import cast
 from alembic import op
 import sqlalchemy as sa
@@ -20,13 +19,8 @@ down_revision = "47e5bef3a1d7"
 branch_labels: None = None
 depends_on: None = None
 
-# Configure logging
-logger = logging.getLogger("alembic.runtime.migration")
-logger.setLevel(logging.INFO)
-
 
 def upgrade() -> None:
-    logger.info(f"{revision}: create_table: slack_bot")
     # Create new slack_bot table
     op.create_table(
         "slack_bot",
@@ -63,7 +57,6 @@ def upgrade() -> None:
     )
 
     # Handle existing Slack bot tokens first
-    logger.info(f"{revision}: Checking for existing Slack bot.")
     bot_token = None
     app_token = None
     first_row_id = None
@@ -71,15 +64,12 @@ def upgrade() -> None:
     try:
         tokens = cast(dict, get_kv_store().load("slack_bot_tokens_config_key"))
     except Exception:
-        logger.warning("No existing Slack bot tokens found.")
         tokens = {}
 
     bot_token = tokens.get("bot_token")
     app_token = tokens.get("app_token")
 
     if bot_token and app_token:
-        logger.info(f"{revision}: Found bot and app tokens.")
-
         session = Session(bind=op.get_bind())
         new_slack_bot = SlackBot(
             name="Slack Bot (Migrated)",
@@ -170,10 +160,9 @@ def upgrade() -> None:
     # Clean up old tokens if they existed
     try:
         if bot_token and app_token:
-            logger.info(f"{revision}: Removing old bot and app tokens.")
             get_kv_store().delete("slack_bot_tokens_config_key")
     except Exception:
-        logger.warning("tried to delete tokens in dynamic config but failed")
+        pass
     # Rename the table
     op.rename_table(
         "slack_bot_config__standard_answer_category",
@@ -190,8 +179,6 @@ def upgrade() -> None:
     # Drop the table with CASCADE to handle dependent objects
     op.execute("DROP TABLE slack_bot_config CASCADE")
 
-    logger.info(f"{revision}: Migration complete.")
-
 
 def downgrade() -> None:
     # Recreate the old slack_bot_config table
@@ -273,7 +260,7 @@ def downgrade() -> None:
             }
             get_kv_store().store("slack_bot_tokens_config_key", tokens)
     except Exception:
-        logger.warning("Failed to save tokens back to KV store")
+        pass
 
     # Drop the new tables in reverse order
     op.drop_table("slack_channel_config")

backend/alembic/versions/6fc7886d665d_make_categories_labels_and_many_to_many.py

@@ -32,6 +32,7 @@ def upgrade() -> None:
         sa.ForeignKeyConstraint(
             ["persona_label_id"],
             ["persona_label.id"],
+            ondelete="CASCADE",
         ),
         sa.PrimaryKeyConstraint("persona_id", "persona_label_id"),
     )

backend/alembic/versions/925b58bd75b6_agent_metric_col_rename__s.py

@@ -1,35 +0,0 @@
-"""agent_metric_col_rename__s
-
-Revision ID: 925b58bd75b6
-Revises: 9787be927e58
-Create Date: 2025-01-06 11:20:26.752441
-
-"""
-from alembic import op
-
-
-# revision identifiers, used by Alembic.
-revision = "925b58bd75b6"
-down_revision = "9787be927e58"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # Rename columns using PostgreSQL syntax
-    op.alter_column(
-        "agent__search_metrics", "base_duration_s", new_column_name="base_duration__s"
-    )
-    op.alter_column(
-        "agent__search_metrics", "full_duration_s", new_column_name="full_duration__s"
-    )
-
-
-def downgrade() -> None:
-    # Revert the column renames
-    op.alter_column(
-        "agent__search_metrics", "base_duration__s", new_column_name="base_duration_s"
-    )
-    op.alter_column(
-        "agent__search_metrics", "full_duration__s", new_column_name="full_duration_s"
-    )

backend/alembic/versions/9787be927e58_agent_metric_table_renames__agent__.py

@@ -1,25 +0,0 @@
-"""agent_metric_table_renames__agent__
-
-Revision ID: 9787be927e58
-Revises: bceb76d618ec
-Create Date: 2025-01-06 11:01:44.210160
-
-"""
-from alembic import op
-
-
-# revision identifiers, used by Alembic.
-revision = "9787be927e58"
-down_revision = "bceb76d618ec"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # Rename table from agent_search_metrics to agent__search_metrics
-    op.rename_table("agent_search_metrics", "agent__search_metrics")
-
-
-def downgrade() -> None:
-    # Rename table back from agent__search_metrics to agent_search_metrics
-    op.rename_table("agent__search_metrics", "agent_search_metrics")

backend/alembic/versions/97dbb53fa8c8_add_syncrecord.py

@@ -0,0 +1,72 @@
+"""Add SyncRecord
+
+Revision ID: 97dbb53fa8c8
+Revises: 369644546676
+Create Date: 2025-01-11 19:39:50.426302
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "97dbb53fa8c8"
+down_revision = "be2ab2aa50ee"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "sync_record",
+        sa.Column("id", sa.Integer(), nullable=False),
+        sa.Column("entity_id", sa.Integer(), nullable=False),
+        sa.Column(
+            "sync_type",
+            sa.Enum(
+                "DOCUMENT_SET",
+                "USER_GROUP",
+                "CONNECTOR_DELETION",
+                name="synctype",
+                native_enum=False,
+                length=40,
+            ),
+            nullable=False,
+        ),
+        sa.Column(
+            "sync_status",
+            sa.Enum(
+                "IN_PROGRESS",
+                "SUCCESS",
+                "FAILED",
+                "CANCELED",
+                name="syncstatus",
+                native_enum=False,
+                length=40,
+            ),
+            nullable=False,
+        ),
+        sa.Column("num_docs_synced", sa.Integer(), nullable=False),
+        sa.Column("sync_start_time", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("sync_end_time", sa.DateTime(timezone=True), nullable=True),
+        sa.PrimaryKeyConstraint("id"),
+    )
+
+    # Add index for fetch_latest_sync_record query
+    op.create_index(
+        "ix_sync_record_entity_id_sync_type_sync_start_time",
+        "sync_record",
+        ["entity_id", "sync_type", "sync_start_time"],
+    )
+
+    # Add index for cleanup_sync_records query
+    op.create_index(
+        "ix_sync_record_entity_id_sync_type_sync_status",
+        "sync_record",
+        ["entity_id", "sync_type", "sync_status"],
+    )
+
+
+def downgrade() -> None:
+    op.drop_index("ix_sync_record_entity_id_sync_type_sync_status")
+    op.drop_index("ix_sync_record_entity_id_sync_type_sync_start_time")
+    op.drop_table("sync_record")

backend/alembic/versions/98a5008d8711_agent_tracking.py

@@ -1,24 +1,25 @@
 """agent_tracking
 
 Revision ID: 98a5008d8711
-Revises: 027381bce97c
-Create Date: 2025-01-04 14:41:52.732238
+Revises: 2f80c6a2550f
+Create Date: 2025-01-29 17:00:00.000001
 
 """
 from alembic import op
 import sqlalchemy as sa
 from sqlalchemy.dialects import postgresql
+from sqlalchemy.dialects.postgresql import UUID
 
 # revision identifiers, used by Alembic.
 revision = "98a5008d8711"
-down_revision = "027381bce97c"
+down_revision = "2f80c6a2550f"
 branch_labels = None
 depends_on = None
 
 
 def upgrade() -> None:
     op.create_table(
-        "agent_search_metrics",
+        "agent__search_metrics",
         sa.Column("id", sa.Integer(), nullable=False),
         sa.Column("user_id", postgresql.UUID(as_uuid=True), nullable=True),
         sa.Column("persona_id", sa.Integer(), nullable=True),
@@ -37,6 +38,70 @@ def upgrade() -> None:
         sa.PrimaryKeyConstraint("id"),
     )
 
+    # Create sub_question table
+    op.create_table(
+        "agent__sub_question",
+        sa.Column("id", sa.Integer, primary_key=True),
+        sa.Column("primary_question_id", sa.Integer, sa.ForeignKey("chat_message.id")),
+        sa.Column(
+            "chat_session_id", UUID(as_uuid=True), sa.ForeignKey("chat_session.id")
+        ),
+        sa.Column("sub_question", sa.Text),
+        sa.Column(
+            "time_created", sa.DateTime(timezone=True), server_default=sa.func.now()
+        ),
+        sa.Column("sub_answer", sa.Text),
+        sa.Column("sub_question_doc_results", postgresql.JSONB(), nullable=True),
+        sa.Column("level", sa.Integer(), nullable=False),
+        sa.Column("level_question_num", sa.Integer(), nullable=False),
+    )
+
+    # Create sub_query table
+    op.create_table(
+        "agent__sub_query",
+        sa.Column("id", sa.Integer, primary_key=True),
+        sa.Column(
+            "parent_question_id", sa.Integer, sa.ForeignKey("agent__sub_question.id")
+        ),
+        sa.Column(
+            "chat_session_id", UUID(as_uuid=True), sa.ForeignKey("chat_session.id")
+        ),
+        sa.Column("sub_query", sa.Text),
+        sa.Column(
+            "time_created", sa.DateTime(timezone=True), server_default=sa.func.now()
+        ),
+    )
+
+    # Create sub_query__search_doc association table
+    op.create_table(
+        "agent__sub_query__search_doc",
+        sa.Column(
+            "sub_query_id",
+            sa.Integer,
+            sa.ForeignKey("agent__sub_query.id"),
+            primary_key=True,
+        ),
+        sa.Column(
+            "search_doc_id",
+            sa.Integer,
+            sa.ForeignKey("search_doc.id"),
+            primary_key=True,
+        ),
+    )
+
+    op.add_column(
+        "chat_message",
+        sa.Column(
+            "refined_answer_improvement",
+            sa.Boolean(),
+            nullable=True,
+        ),
+    )
+
 
 def downgrade() -> None:
-    op.drop_table("agent_search_metrics")
+    op.drop_column("chat_message", "refined_answer_improvement")
+    op.drop_table("agent__sub_query__search_doc")
+    op.drop_table("agent__sub_query")
+    op.drop_table("agent__sub_question")
+    op.drop_table("agent__search_metrics")

backend/alembic/versions/a6df6b88ef81_remove_recent_assistants.py

@@ -0,0 +1,29 @@
+"""remove recent assistants
+
+Revision ID: a6df6b88ef81
+Revises: 4d58345da04a
+Create Date: 2025-01-29 10:25:52.790407
+
+"""
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision = "a6df6b88ef81"
+down_revision = "4d58345da04a"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.drop_column("user", "recent_assistants")
+
+
+def downgrade() -> None:
+    op.add_column(
+        "user",
+        sa.Column(
+            "recent_assistants", postgresql.JSONB(), server_default="[]", nullable=False
+        ),
+    )

backend/alembic/versions/aeda5f2df4f6_add_pinned_assistants.py

@@ -1,7 +1,7 @@
 """add pinned assistants
 
 Revision ID: aeda5f2df4f6
-Revises: 369644546676
+Revises: c5eae4a75a1b
 Create Date: 2025-01-09 16:04:10.770636
 
 """
@@ -11,7 +11,7 @@ from sqlalchemy.dialects import postgresql
 
 # revision identifiers, used by Alembic.
 revision = "aeda5f2df4f6"
-down_revision = "369644546676"
+down_revision = "c5eae4a75a1b"
 branch_labels = None
 depends_on = None
 

backend/alembic/versions/bceb76d618ec_agent_table_renames__agent__.py

@@ -1,84 +0,0 @@
-"""agent_table_renames__agent__
-
-Revision ID: bceb76d618ec
-Revises: c0132518a25b
-Create Date: 2025-01-06 10:50:48.109285
-
-"""
-from alembic import op
-
-
-# revision identifiers, used by Alembic.
-revision = "bceb76d618ec"
-down_revision = "c0132518a25b"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    op.drop_constraint(
-        "sub_query__search_doc_sub_query_id_fkey",
-        "sub_query__search_doc",
-        type_="foreignkey",
-    )
-    op.drop_constraint(
-        "sub_query__search_doc_search_doc_id_fkey",
-        "sub_query__search_doc",
-        type_="foreignkey",
-    )
-    # Rename tables
-    op.rename_table("sub_query", "agent__sub_query")
-    op.rename_table("sub_question", "agent__sub_question")
-    op.rename_table("sub_query__search_doc", "agent__sub_query__search_doc")
-
-    # Update both foreign key constraints for agent__sub_query__search_doc
-
-    # Create new foreign keys with updated names
-    op.create_foreign_key(
-        "agent__sub_query__search_doc_sub_query_id_fkey",
-        "agent__sub_query__search_doc",
-        "agent__sub_query",
-        ["sub_query_id"],
-        ["id"],
-    )
-    op.create_foreign_key(
-        "agent__sub_query__search_doc_search_doc_id_fkey",
-        "agent__sub_query__search_doc",
-        "search_doc",  # This table name doesn't change
-        ["search_doc_id"],
-        ["id"],
-    )
-
-
-def downgrade() -> None:
-    # Update foreign key constraints for sub_query__search_doc
-    op.drop_constraint(
-        "agent__sub_query__search_doc_sub_query_id_fkey",
-        "agent__sub_query__search_doc",
-        type_="foreignkey",
-    )
-    op.drop_constraint(
-        "agent__sub_query__search_doc_search_doc_id_fkey",
-        "agent__sub_query__search_doc",
-        type_="foreignkey",
-    )
-
-    # Rename tables back
-    op.rename_table("agent__sub_query__search_doc", "sub_query__search_doc")
-    op.rename_table("agent__sub_question", "sub_question")
-    op.rename_table("agent__sub_query", "sub_query")
-
-    op.create_foreign_key(
-        "sub_query__search_doc_sub_query_id_fkey",
-        "sub_query__search_doc",
-        "sub_query",
-        ["sub_query_id"],
-        ["id"],
-    )
-    op.create_foreign_key(
-        "sub_query__search_doc_search_doc_id_fkey",
-        "sub_query__search_doc",
-        "search_doc",  # This table name doesn't change
-        ["search_doc_id"],
-        ["id"],
-    )

backend/alembic/versions/be2ab2aa50ee_fix_capitalization.py

@@ -0,0 +1,38 @@
+"""fix_capitalization
+
+Revision ID: be2ab2aa50ee
+Revises: 369644546676
+Create Date: 2025-01-10 13:13:26.228960
+
+"""
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision = "be2ab2aa50ee"
+down_revision = "369644546676"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.execute(
+        """
+        UPDATE document
+        SET
+            external_user_group_ids = ARRAY(
+                SELECT LOWER(unnest(external_user_group_ids))
+            ),
+            last_modified = NOW()
+        WHERE
+            external_user_group_ids IS NOT NULL
+            AND external_user_group_ids::text[] <> ARRAY(
+                SELECT LOWER(unnest(external_user_group_ids))
+            )::text[]
+    """
+    )
+
+
+def downgrade() -> None:
+    # No way to cleanly persist the bad state through an upgrade/downgrade
+    # cycle, so we just pass
+    pass

backend/alembic/versions/c0132518a25b_agent_table_changes_rename_level.py

@@ -1,40 +0,0 @@
-"""agent_table_changes_rename_level
-
-Revision ID: c0132518a25b
-Revises: 1adf5ea20d2b
-Create Date: 2025-01-05 16:38:37.660152
-
-"""
-from alembic import op
-import sqlalchemy as sa
-
-
-# revision identifiers, used by Alembic.
-revision = "c0132518a25b"
-down_revision = "1adf5ea20d2b"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # Add level and level_question_nr columns with NOT NULL constraint
-    op.add_column(
-        "sub_question",
-        sa.Column("level", sa.Integer(), nullable=False, server_default="0"),
-    )
-    op.add_column(
-        "sub_question",
-        sa.Column(
-            "level_question_nr", sa.Integer(), nullable=False, server_default="0"
-        ),
-    )
-
-    # Remove the server_default after the columns are created
-    op.alter_column("sub_question", "level", server_default=None)
-    op.alter_column("sub_question", "level_question_nr", server_default=None)
-
-
-def downgrade() -> None:
-    # Remove the columns
-    op.drop_column("sub_question", "level_question_nr")
-    op.drop_column("sub_question", "level")

backend/alembic/versions/c5eae4a75a1b_add_chat_message__standard_answer_table.py

@@ -0,0 +1,36 @@
+"""Add chat_message__standard_answer table
+
+Revision ID: c5eae4a75a1b
+Revises: 0f7ff6d75b57
+Create Date: 2025-01-15 14:08:49.688998
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "c5eae4a75a1b"
+down_revision = "0f7ff6d75b57"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "chat_message__standard_answer",
+        sa.Column("chat_message_id", sa.Integer(), nullable=False),
+        sa.Column("standard_answer_id", sa.Integer(), nullable=False),
+        sa.ForeignKeyConstraint(
+            ["chat_message_id"],
+            ["chat_message.id"],
+        ),
+        sa.ForeignKeyConstraint(
+            ["standard_answer_id"],
+            ["standard_answer.id"],
+        ),
+        sa.PrimaryKeyConstraint("chat_message_id", "standard_answer_id"),
+    )
+
+
+def downgrade() -> None:
+    op.drop_table("chat_message__standard_answer")

backend/alembic/versions/c7bf5721733e_add_has_been_indexed_to_.py

@@ -0,0 +1,48 @@
+"""Add has_been_indexed to DocumentByConnectorCredentialPair
+
+Revision ID: c7bf5721733e
+Revises: fec3db967bf7
+Create Date: 2025-01-13 12:39:05.831693
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "c7bf5721733e"
+down_revision = "027381bce97c"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # assume all existing rows have been indexed, no better approach
+    op.add_column(
+        "document_by_connector_credential_pair",
+        sa.Column("has_been_indexed", sa.Boolean(), nullable=True),
+    )
+    op.execute(
+        "UPDATE document_by_connector_credential_pair SET has_been_indexed = TRUE"
+    )
+    op.alter_column(
+        "document_by_connector_credential_pair",
+        "has_been_indexed",
+        nullable=False,
+    )
+
+    # Add index to optimize get_document_counts_for_cc_pairs query pattern
+    op.create_index(
+        "idx_document_cc_pair_counts",
+        "document_by_connector_credential_pair",
+        ["connector_id", "credential_id", "has_been_indexed"],
+        unique=False,
+    )
+
+
+def downgrade() -> None:
+    # Remove the index first before removing the column
+    op.drop_index(
+        "idx_document_cc_pair_counts",
+        table_name="document_by_connector_credential_pair",
+    )
+    op.drop_column("document_by_connector_credential_pair", "has_been_indexed")

backend/alembic/versions/e9cf2bd7baed_create_pro_search_persistence_tables.py

@@ -1,68 +0,0 @@
-"""create pro search persistence tables
-
-Revision ID: e9cf2bd7baed
-Revises: 98a5008d8711
-Create Date: 2025-01-02 17:55:56.544246
-
-"""
-from alembic import op
-import sqlalchemy as sa
-from sqlalchemy.dialects.postgresql import UUID
-
-
-# revision identifiers, used by Alembic.
-revision = "e9cf2bd7baed"
-down_revision = "98a5008d8711"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # Create sub_question table
-    op.create_table(
-        "sub_question",
-        sa.Column("id", sa.Integer, primary_key=True),
-        sa.Column("primary_question_id", sa.Integer, sa.ForeignKey("chat_message.id")),
-        sa.Column(
-            "chat_session_id", UUID(as_uuid=True), sa.ForeignKey("chat_session.id")
-        ),
-        sa.Column("sub_question", sa.Text),
-        sa.Column(
-            "time_created", sa.DateTime(timezone=True), server_default=sa.func.now()
-        ),
-        sa.Column("sub_answer", sa.Text),
-    )
-
-    # Create sub_query table
-    op.create_table(
-        "sub_query",
-        sa.Column("id", sa.Integer, primary_key=True),
-        sa.Column("parent_question_id", sa.Integer, sa.ForeignKey("sub_question.id")),
-        sa.Column(
-            "chat_session_id", UUID(as_uuid=True), sa.ForeignKey("chat_session.id")
-        ),
-        sa.Column("sub_query", sa.Text),
-        sa.Column(
-            "time_created", sa.DateTime(timezone=True), server_default=sa.func.now()
-        ),
-    )
-
-    # Create sub_query__search_doc association table
-    op.create_table(
-        "sub_query__search_doc",
-        sa.Column(
-            "sub_query_id", sa.Integer, sa.ForeignKey("sub_query.id"), primary_key=True
-        ),
-        sa.Column(
-            "search_doc_id",
-            sa.Integer,
-            sa.ForeignKey("search_doc.id"),
-            primary_key=True,
-        ),
-    )
-
-
-def downgrade() -> None:
-    op.drop_table("sub_query__search_doc")
-    op.drop_table("sub_query")
-    op.drop_table("sub_question")

backend/alembic/versions/eaa3b5593925_add_default_slack_channel_config.py

@@ -0,0 +1,80 @@
+"""add default slack channel config
+
+Revision ID: eaa3b5593925
+Revises: 98a5008d8711
+Create Date: 2025-02-03 18:07:56.552526
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = "eaa3b5593925"
+down_revision = "98a5008d8711"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    # Add is_default column
+    op.add_column(
+        "slack_channel_config",
+        sa.Column("is_default", sa.Boolean(), nullable=False, server_default="false"),
+    )
+
+    op.create_index(
+        "ix_slack_channel_config_slack_bot_id_default",
+        "slack_channel_config",
+        ["slack_bot_id", "is_default"],
+        unique=True,
+        postgresql_where=sa.text("is_default IS TRUE"),
+    )
+
+    # Create default channel configs for existing slack bots without one
+    conn = op.get_bind()
+    slack_bots = conn.execute(sa.text("SELECT id FROM slack_bot")).fetchall()
+
+    for slack_bot in slack_bots:
+        slack_bot_id = slack_bot[0]
+        existing_default = conn.execute(
+            sa.text(
+                "SELECT id FROM slack_channel_config WHERE slack_bot_id = :bot_id AND is_default = TRUE"
+            ),
+            {"bot_id": slack_bot_id},
+        ).fetchone()
+
+        if not existing_default:
+            conn.execute(
+                sa.text(
+                    """
+                    INSERT INTO slack_channel_config (
+                        slack_bot_id, persona_id, channel_config, enable_auto_filters, is_default
+                    ) VALUES (
+                        :bot_id, NULL,
+                        '{"channel_name": null, '
+                        '"respond_member_group_list": [], '
+                        '"answer_filters": [], '
+                        '"follow_up_tags": [], '
+                        '"respond_tag_only": true}',
+                        FALSE, TRUE
+                    )
+                """
+                ),
+                {"bot_id": slack_bot_id},
+            )
+
+
+def downgrade() -> None:
+    # Delete default slack channel configs
+    conn = op.get_bind()
+    conn.execute(sa.text("DELETE FROM slack_channel_config WHERE is_default = TRUE"))
+
+    # Remove index
+    op.drop_index(
+        "ix_slack_channel_config_slack_bot_id_default",
+        table_name="slack_channel_config",
+    )
+
+    # Remove is_default column
+    op.drop_column("slack_channel_config", "is_default")

backend/alembic/versions/f1ca58b2f2ec_add_passthrough_auth_to_tool.py

@@ -0,0 +1,33 @@
+"""add passthrough auth to tool
+
+Revision ID: f1ca58b2f2ec
+Revises: c7bf5721733e
+Create Date: 2024-03-19
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = "f1ca58b2f2ec"
+down_revision: Union[str, None] = "c7bf5721733e"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    # Add passthrough_auth column to tool table with default value of False
+    op.add_column(
+        "tool",
+        sa.Column(
+            "passthrough_auth", sa.Boolean(), nullable=False, server_default=sa.false()
+        ),
+    )
+
+
+def downgrade() -> None:
+    # Remove passthrough_auth column from tool table
+    op.drop_column("tool", "passthrough_auth")

backend/alembic/versions/f5437cc136c5_delete_non_search_assistants.py

@@ -0,0 +1,53 @@
+"""delete non-search assistants
+
+Revision ID: f5437cc136c5
+Revises: eaa3b5593925
+Create Date: 2025-02-04 16:17:15.677256
+
+"""
+from alembic import op
+
+
+# revision identifiers, used by Alembic.
+revision = "f5437cc136c5"
+down_revision = "eaa3b5593925"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    pass
+
+
+def downgrade() -> None:
+    # Fix: split the statements into multiple op.execute() calls
+    op.execute(
+        """
+        WITH personas_without_search AS (
+            SELECT p.id
+            FROM persona p
+            LEFT JOIN persona__tool pt ON p.id = pt.persona_id
+            LEFT JOIN tool t ON pt.tool_id = t.id
+            GROUP BY p.id
+            HAVING COUNT(CASE WHEN t.in_code_tool_id = 'run_search' THEN 1 END) = 0
+        )
+        UPDATE slack_channel_config
+        SET persona_id = NULL
+        WHERE is_default = TRUE AND persona_id IN (SELECT id FROM personas_without_search)
+        """
+    )
+
+    op.execute(
+        """
+        WITH personas_without_search AS (
+            SELECT p.id
+            FROM persona p
+            LEFT JOIN persona__tool pt ON p.id = pt.persona_id
+            LEFT JOIN tool t ON pt.tool_id = t.id
+            GROUP BY p.id
+            HAVING COUNT(CASE WHEN t.in_code_tool_id = 'run_search' THEN 1 END) = 0
+        )
+        DELETE FROM slack_channel_config
+        WHERE is_default = FALSE AND persona_id IN (SELECT id FROM personas_without_search)
+        """
+    )

backend/alembic/versions/fec3db967bf7_add_time_updated_to_usergroup_and_.py

@@ -0,0 +1,41 @@
+"""Add time_updated to UserGroup and DocumentSet
+
+Revision ID: fec3db967bf7
+Revises: 97dbb53fa8c8
+Create Date: 2025-01-12 15:49:02.289100
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision = "fec3db967bf7"
+down_revision = "97dbb53fa8c8"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "document_set",
+        sa.Column(
+            "time_last_modified_by_user",
+            sa.DateTime(timezone=True),
+            nullable=False,
+            server_default=sa.func.now(),
+        ),
+    )
+    op.add_column(
+        "user_group",
+        sa.Column(
+            "time_last_modified_by_user",
+            sa.DateTime(timezone=True),
+            nullable=False,
+            server_default=sa.func.now(),
+        ),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("user_group", "time_last_modified_by_user")
+    op.drop_column("document_set", "time_last_modified_by_user")

backend/chatt.txt

@@ -1,536 +0,0 @@
-"{\"user_message_id\": 475, \"reserved_assistant_message_id\": 476}\n"
-"{\"sub_question\": \"What\", \"level\": 0, \"level_question_nr\": 1}\n"
-"{\"sub_query\": \"ony\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_question\": \" is\", \"level\": 0, \"level_question_nr\": 1}\n"
-"{\"sub_query\": \"x\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_question\": \" On\", \"level\": 0, \"level_question_nr\": 1}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_question\": \"yx\", \"level\": 0, \"level_question_nr\": 1}\n"
-"{\"sub_question\": \" \", \"level\": 0, \"level_question_nr\": 1}\n"
-"{\"sub_query\": \"1\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_question\": \"1\", \"level\": 0, \"level_question_nr\": 1}\n"
-"{\"sub_query\": \" features\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_query\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_question\": \"?\", \"level\": 0, \"level_question_nr\": 1}\n"
-"{\"sub_question\": \"  \", \"level\": 0, \"level_question_nr\": 1}\n"
-"{\"sub_question\": \"\", \"level\": 0, \"level_question_nr\": 2}\n"
-"{\"sub_question\": \"What\", \"level\": 0, \"level_question_nr\": 2}\n"
-"{\"sub_question\": \" is\", \"level\": 0, \"level_question_nr\": 2}\n"
-"{\"sub_query\": \" specifications\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_question\": \" On\", \"level\": 0, \"level_question_nr\": 2}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_question\": \"yx\", \"level\": 0, \"level_question_nr\": 2}\n"
-"{\"sub_question\": \" \", \"level\": 0, \"level_question_nr\": 2}\n"
-"{\"sub_question\": \"2\", \"level\": 0, \"level_question_nr\": 2}\n"
-"{\"sub_query\": \"ony\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_question\": \"?\", \"level\": 0, \"level_question_nr\": 2}\n"
-"{\"sub_query\": \"x\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_question\": \"  \", \"level\": 0, \"level_question_nr\": 2}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_question\": \"\", \"level\": 0, \"level_question_nr\": 3}\n"
-"{\"sub_query\": \"2\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_question\": \"What\", \"level\": 0, \"level_question_nr\": 3}\n"
-"{\"sub_query\": \" applications\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_question\": \" is\", \"level\": 0, \"level_question_nr\": 3}\n"
-"{\"sub_query\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_question\": \" On\", \"level\": 0, \"level_question_nr\": 3}\n"
-"{\"sub_query\": \" use\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_question\": \"yx\", \"level\": 0, \"level_question_nr\": 3}\n"
-"{\"sub_query\": \" cases\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_question\": \" \", \"level\": 0, \"level_question_nr\": 3}\n"
-"{\"sub_question\": \"3\", \"level\": 0, \"level_question_nr\": 3}\n"
-"{\"sub_question\": \"?\", \"level\": 0, \"level_question_nr\": 3}\n"
-"{\"sub_question\": \"  \", \"level\": 0, \"level_question_nr\": 3}\n"
-"{\"sub_question\": \"\", \"level\": 0, \"level_question_nr\": 4}\n"
-"{\"sub_question\": \"What\", \"level\": 0, \"level_question_nr\": 4}\n"
-"{\"sub_question\": \" is\", \"level\": 0, \"level_question_nr\": 4}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_question\": \" On\", \"level\": 0, \"level_question_nr\": 4}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_question\": \"yx\", \"level\": 0, \"level_question_nr\": 4}\n"
-"{\"sub_question\": \" \", \"level\": 0, \"level_question_nr\": 4}\n"
-"{\"sub_query\": \"ony\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_question\": \"4\", \"level\": 0, \"level_question_nr\": 4}\n"
-"{\"sub_question\": \"?\", \"level\": 0, \"level_question_nr\": 4}\n"
-"{\"sub_query\": \"x\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_question\": \"  \", \"level\": 0, \"level_question_nr\": 4}\n"
-"{\"sub_question\": \"\", \"level\": 0, \"level_question_nr\": 4}\n"
-"{\"sub_query\": \"3\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \"4\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" comparison\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" differences\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 0}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 0}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 0}\n"
-"{\"sub_query\": \"4\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 0}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 0}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 0}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_query\": \"1\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 0}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 0}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 0}\n"
-"{\"sub_query\": \" product\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_query\": \"3\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 0}\n"
-"{\"sub_query\": \" information\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_query\": \" software\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 0}\n"
-"{\"sub_query\": \" features\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 0}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 0}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 0}\n"
-"{\"sub_query\": \" software\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 0}\n"
-"{\"sub_query\": \"2\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 0}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 1}\n"
-"{\"sub_query\": \" features\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 0}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 1}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 0}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 1}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 1}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_query\": \" software\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 0}\n"
-"{\"sub_query\": \" features\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 0}\n"
-"{\"sub_query\": \"4\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 1}\n"
-"{\"sub_query\": \" applications\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 1}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 0}\n"
-"{\"sub_query\": \"1\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 1}\n"
-"{\"sub_query\": \" features\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 1}\n"
-"{\"sub_query\": \" in\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 1}\n"
-"{\"sub_query\": \" industry\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 1}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 0}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 1}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 1}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 1}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 1}\n"
-"{\"sub_query\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 1}\n"
-"{\"sub_query\": \" specifications\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 1}\n"
-"{\"sub_query\": \"2\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 1}\n"
-"{\"sub_query\": \" applications\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 1}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 1}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 1}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 2}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 2}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \"3\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 1}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" applications\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 1}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 2}\n"
-"{\"sub_query\": \" in\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 1}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 2}\n"
-"{\"sub_query\": \" industry\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 1}\n"
-"{\"sub_query\": \"1\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" applications\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 1}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 2}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 2}\n"
-"{\"sub_query\": \"4\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 2}\n"
-"{\"sub_query\": \" comparison\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 2}\n"
-"{\"sub_query\": \" with\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 2}\n"
-"{\"sub_query\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" previous\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 2}\n"
-"{\"sub_query\": \" use\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" versions\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 2}\n"
-"{\"sub_query\": \" in\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 1}\n"
-"{\"sub_query\": \" industry\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 1}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 2}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 3, \"query_id\": 2}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 2}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 1}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 2}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 2}\n"
-"{\"sub_query\": \"On\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 2}\n"
-"{\"sub_query\": \"yx\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 2}\n"
-"{\"sub_query\": \" \", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 2}\n"
-"{\"sub_query\": \"3\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 2}\n"
-"{\"sub_query\": \" comparison\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 2}\n"
-"{\"sub_query\": \"2\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 2}\n"
-"{\"sub_query\": \" comparison\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 2}\n"
-"{\"sub_query\": \" with\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 2}\n"
-"{\"sub_query\": \" cases\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" previous\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 2}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 0, \"query_id\": 2}\n"
-"{\"sub_query\": \" with\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 2}\n"
-"{\"sub_query\": \" other\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 2}\n"
-"{\"sub_query\": \" software\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 2}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 1, \"query_id\": 2}\n"
-"{\"sub_query\": \" versions\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 2}\n"
-"{\"sub_query\": \"  \", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 2}\n"
-"{\"sub_query\": \"\", \"level\": 0, \"level_question_nr\": 2, \"query_id\": 2}\n"
-"{\"top_documents\": [], \"rephrased_query\": \"What is Onyx 4?\", \"predicted_flow\": \"question-answer\", \"predicted_search\": \"keyword\", \"applied_source_filters\": null, \"applied_time_cutoff\": null, \"recency_bias_multiplier\": 0.5}\n"
-"{\"llm_selected_doc_indices\": []}\n"
-"{\"final_context_docs\": []}\n"
-"{\"answer_piece\": \"I\", \"level\": 0, \"level_question_nr\": 3, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" don't\", \"level\": 0, \"level_question_nr\": 3, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"On\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" know\", \"level\": 0, \"level_question_nr\": 3, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"yx\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 3, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" \", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"1\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" formerly\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" known\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" as\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" D\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"answer\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" is\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" an\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" AI\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" Assistant\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" that\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" connects\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" to\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" a\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" company's\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" documents\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" applications\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" personnel\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" It\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" provides\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" a\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" chat\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" interface\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" can\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" integrate\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" with\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" any\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" large\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" language\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" model\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" (\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"LL\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"M\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \")\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"top_documents\": [], \"rephrased_query\": \"What is Onyx 2?\", \"predicted_flow\": \"question-answer\", \"predicted_search\": \"keyword\", \"applied_source_filters\": null, \"applied_time_cutoff\": null, \"recency_bias_multiplier\": 0.5}\n"
-"{\"llm_selected_doc_indices\": []}\n"
-"{\"final_context_docs\": []}\n"
-"{\"answer_piece\": \" of\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" choice\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" On\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"yx\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" is\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" designed\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" to\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" be\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" modular\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" easily\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" extens\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"ible\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" allowing\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" for\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" deployment\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" on\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" various\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" platforms\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" including\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" laptops\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" on\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"-prem\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"ise\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" or\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" cloud\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" environments\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" It\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" ensures\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" that\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" user\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" data\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" chats\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" remain\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"I\", \"level\": 0, \"level_question_nr\": 1, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" under\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" don't\", \"level\": 0, \"level_question_nr\": 1, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" the\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" user's\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" know\", \"level\": 0, \"level_question_nr\": 1, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 1, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" control\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" as\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" the\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" deployment\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" is\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" owned\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" by\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" the\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" user\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" On\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"yx\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" is\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" MIT\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" licensed\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" comes\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" ready\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" for\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" production\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" use\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" featuring\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" user\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" authentication\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" role\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" management\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" chat\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" persistence\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" a\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" user\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" interface\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" for\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" configuring\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" AI\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" Assist\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"ants\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" their\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" prompts\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" Additionally\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" On\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"yx\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" serves\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" as\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" a\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" unified\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" search\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" tool\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" across\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" common\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" workplace\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" applications\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" like\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" Slack\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" Google\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" Drive\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" Con\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"fluence\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" enabling\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" it\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" to\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" act\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" as\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" a\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" subject\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" matter\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" expert\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" for\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" teams\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" by\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" combining\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" L\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"LM\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"s\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" with\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" team\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \"-specific\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" knowledge\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" [[1]]()\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"top_documents\": [], \"rephrased_query\": \"What is Onyx 3?\", \"predicted_flow\": \"question-answer\", \"predicted_search\": \"keyword\", \"applied_source_filters\": null, \"applied_time_cutoff\": null, \"recency_bias_multiplier\": 0.5}\n"
-"{\"llm_selected_doc_indices\": []}\n"
-"{\"final_context_docs\": []}\n"
-"{\"answer_piece\": \"I\", \"level\": 0, \"level_question_nr\": 2, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" don't\", \"level\": 0, \"level_question_nr\": 2, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \" know\", \"level\": 0, \"level_question_nr\": 2, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 2, \"answer_type\": \"agent_sub_answer\"}\n"
-"{\"top_documents\": [{\"document_id\": \"https://docs.onyx.app/introduction\", \"chunk_ind\": 0, \"semantic_identifier\": \"Introduction - Onyx Documentation\", \"link\": \"https://docs.onyx.app/introduction\", \"blurb\": \"Onyx Documentation home page\\nSearch...\\nNavigation\\nWelcome to Onyx\\nIntroduction\\nWelcome to Onyx\\nIntroduction\\nOnyx Overview\\n\\nWhat is Onyx\\nOnyx (Formerly Danswer) is the AI Assistant connected to your companys docs, apps, and people. Onyx provides a Chat interface and plugs into any LLM of your choice. Onyx can be deployed anywhere and for any scale - on a laptop, on-premise, or to cloud. Since you own the deployment, your user data and chats are fully in your own control. Onyx is MIT licensed and designed to be modular and easily extensible.\", \"source_type\": \"web\", \"boost\": 0, \"hidden\": false, \"metadata\": {}, \"score\": 0.6275177643886491, \"is_relevant\": null, \"relevance_explanation\": null, \"match_highlights\": [\"\", \"such as A customer wants feature X, is this already supported? or Wheres the pull request for feature Y?\\n<hi>Onyx</hi> can also be plugged into existing tools like Slack to get answers and AI chats directly in Slack.\\n\\nDemo\\n\\nMain <hi>Features</hi> \\n- Chat UI with the ability to select documents to chat with.\\n- Create custom AI Assistants\", \"\"], \"updated_at\": null, \"primary_owners\": null, \"secondary_owners\": null, \"is_internet\": false, \"db_doc_id\": 35923}], \"rephrased_query\": \"what is onyx 1, 2, 3, 4\", \"predicted_flow\": \"question-answer\", \"predicted_search\": \"keyword\", \"applied_source_filters\": null, \"applied_time_cutoff\": null, \"recency_bias_multiplier\": 0.5}\n"
-"{\"llm_selected_doc_indices\": []}\n"
-"{\"final_context_docs\": [{\"document_id\": \"https://docs.onyx.app/introduction\", \"content\": \"Onyx Documentation home page\\nSearch...\\nNavigation\\nWelcome to Onyx\\nIntroduction\\nWelcome to Onyx\\nIntroduction\\nOnyx Overview\\n\\nWhat is Onyx\\nOnyx (Formerly Danswer) is the AI Assistant connected to your companys docs, apps, and people. Onyx provides a Chat interface and plugs into any LLM of your choice. Onyx can be deployed anywhere and for any scale - on a laptop, on-premise, or to cloud. Since you own the deployment, your user data and chats are fully in your own control. Onyx is MIT licensed and designed to be modular and easily extensible. The system also comes fully ready for production usage with user authentication, role management (admin/basic users), chat persistence, and a UI for configuring Personas (AI Assistants) and their Prompts.\\nOnyx also serves as a Unified Search across all common workplace tools such as Slack, Google Drive, Confluence, etc. By combining LLMs and team specific knowledge, Onyx becomes a subject matter expert for the team. Its like ChatGPT if it had access to your teams unique knowledge! It enables questions such as A customer wants feature X, is this already supported? or Wheres the pull request for feature Y?\\nOnyx can also be plugged into existing tools like Slack to get answers and AI chats directly in Slack.\\n\\nDemo\\n\\nMain Features \\n- Chat UI with the ability to select documents to chat with.\\n- Create custom AI Assistants with different prompts and backing knowledge sets.\\n- Connect Onyx with LLM of your choice (self-host for a fully airgapped solution).\\n- Document Search + AI Answers for natural language queries.\\n- Connectors to all common workplace tools like Google Drive, Confluence, Slack, etc.\\n- Slack integration to get answers and search results directly in Slack.\\n\\nUpcoming\\n- Chat/Prompt sharing with specific teammates and user groups.\\n- Multi-modal model support, chat with images, video etc.\\n- Choosing between LLMs and parameters during chat session.\\n- Tool calling and agent configurations options.\\n- Organizational understanding and ability to locate and suggest experts from your team.\\n\\nOther Noteable Benefits of Onyx\\n- User Authentication with document level access management.\\n- Best in class Hybrid Search across all sources (BM-25 + prefix aware embedding models).\\n- Admin Dashboard to configure connectors, document-sets, access, etc.\\n- Custom deep learning models + learn from user feedback.\\n- Easy deployment and ability to host Onyx anywhere of your choosing.\\nQuickstart\", \"blurb\": \"Onyx Documentation home page\\nSearch...\\nNavigation\\nWelcome to Onyx\\nIntroduction\\nWelcome to Onyx\\nIntroduction\\nOnyx Overview\\n\\nWhat is Onyx\\nOnyx (Formerly Danswer) is the AI Assistant connected to your companys docs, apps, and people. Onyx provides a Chat interface and plugs into any LLM of your choice. Onyx can be deployed anywhere and for any scale - on a laptop, on-premise, or to cloud. Since you own the deployment, your user data and chats are fully in your own control. Onyx is MIT licensed and designed to be modular and easily extensible.\", \"semantic_identifier\": \"Introduction - Onyx Documentation\", \"source_type\": \"web\", \"metadata\": {}, \"updated_at\": null, \"link\": \"https://docs.onyx.app/introduction\", \"source_links\": {\"0\": \"https://docs.onyx.app/introduction\"}, \"match_highlights\": [\"\", \"such as A customer wants feature X, is this already supported? or Wheres the pull request for feature Y?\\n<hi>Onyx</hi> can also be plugged into existing tools like Slack to get answers and AI chats directly in Slack.\\n\\nDemo\\n\\nMain <hi>Features</hi> \\n- Chat UI with the ability to select documents to chat with.\\n- Create custom AI Assistants\", \"\"]}]}\n"
-"{\"answer_piece\": \"I\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" cannot\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" reliably\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" answer\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" the\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" question\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" about\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" On\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"yx\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" \", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"2\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" \", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"3\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" \", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"4\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" as\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" the\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" provided\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" information\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" only\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" describes\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" On\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"yx\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" \", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"1\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" which\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" is\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" an\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" AI\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" Assistant\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" formerly\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" known\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" as\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" D\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"answer\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" On\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"yx\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" \", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"1\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" connects\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" to\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" a\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" company's\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" documents\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" applications\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" personnel\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" providing\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" a\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" chat\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" interface\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" integration\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" with\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" any\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" large\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" language\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" model\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" (\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"LL\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"M\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \")\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" of\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" choice\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" It\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" is\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" designed\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" to\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" be\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" modular\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" easily\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" extens\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"ible\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" can\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" be\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" deployed\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" on\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" various\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" platforms\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" while\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" ensuring\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" user\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" data\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" control\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" It\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" also\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" serves\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" as\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" a\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" unified\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" search\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" tool\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" across\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" common\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" workplace\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" applications\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" like\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" Slack\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" Google\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" Drive\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" and\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" Con\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"fluence\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" acting\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" as\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" a\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" subject\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" matter\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" expert\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" for\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" teams\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" [[1]]()\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"{{1}}\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"There\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" is\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" no\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" information\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" available\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" regarding\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" On\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"yx\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" \", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"2\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" \", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"3\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" or\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" \", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \"4\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \",\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" so\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" I\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" cannot\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" provide\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" details\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" about\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \" them\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"answer_piece\": \".\", \"level\": 0, \"level_question_nr\": 0, \"answer_type\": \"agent_level_answer\"}\n"
-"{\"citations\": []}\n"
-"{\"message_id\": 476, \"parent_message\": 475, \"latest_child_message\": null, \"message\": \"I cannot reliably answer the question about Onyx 2, 3, and 4, as the provided information only describes Onyx 1, which is an AI Assistant formerly known as Danswer. Onyx 1 connects to a company's documents, applications, and personnel, providing a chat interface and integration with any large language model (LLM) of choice. It is designed to be modular, easily extensible, and can be deployed on various platforms while ensuring user data control. It also serves as a unified search tool across common workplace applications like Slack, Google Drive, and Confluence, acting as a subject matter expert for teams [[1]](){{1}}There is no information available regarding Onyx 2, 3, or 4, so I cannot provide details about them.\", \"rephrased_query\": \"what is onyx 1, 2, 3, 4\", \"context_docs\": {\"top_documents\": [{\"document_id\": \"https://docs.onyx.app/introduction\", \"chunk_ind\": 0, \"semantic_identifier\": \"Introduction - Onyx Documentation\", \"link\": \"https://docs.onyx.app/introduction\", \"blurb\": \"Onyx Documentation home page\\nSearch...\\nNavigation\\nWelcome to Onyx\\nIntroduction\\nWelcome to Onyx\\nIntroduction\\nOnyx Overview\\n\\nWhat is Onyx\\nOnyx (Formerly Danswer) is the AI Assistant connected to your companys docs, apps, and people. Onyx provides a Chat interface and plugs into any LLM of your choice. Onyx can be deployed anywhere and for any scale - on a laptop, on-premise, or to cloud. Since you own the deployment, your user data and chats are fully in your own control. Onyx is MIT licensed and designed to be modular and easily extensible.\", \"source_type\": \"web\", \"boost\": 0, \"hidden\": false, \"metadata\": {}, \"score\": 0.6275177643886491, \"is_relevant\": null, \"relevance_explanation\": null, \"match_highlights\": [\"\", \"such as A customer wants feature X, is this already supported? or Wheres the pull request for feature Y?\\n<hi>Onyx</hi> can also be plugged into existing tools like Slack to get answers and AI chats directly in Slack.\\n\\nDemo\\n\\nMain <hi>Features</hi> \\n- Chat UI with the ability to select documents to chat with.\\n- Create custom AI Assistants\", \"\"], \"updated_at\": null, \"primary_owners\": null, \"secondary_owners\": null, \"is_internet\": false, \"db_doc_id\": 35923}]}, \"message_type\": \"assistant\", \"time_sent\": \"2025-01-12T05:37:18.318251+00:00\", \"overridden_model\": \"gpt-4o\", \"alternate_assistant_id\": 0, \"chat_session_id\": \"40f91916-7419-48d1-9681-5882b0869d88\", \"citations\": {}, \"sub_questions\": [], \"files\": [], \"tool_call\": null}\n"

backend/ee/onyx/background/celery/apps/primary.py

@@ -32,6 +32,7 @@ def perform_ttl_management_task(
 
 @celery_app.task(
     name="check_ttl_management_task",
+    ignore_result=True,
     soft_time_limit=JOB_TIMEOUT,
 )
 def check_ttl_management_task(*, tenant_id: str | None) -> None:
@@ -56,6 +57,7 @@ def check_ttl_management_task(*, tenant_id: str | None) -> None:
 
 @celery_app.task(
     name="autogenerate_usage_report_task",
+    ignore_result=True,
     soft_time_limit=JOB_TIMEOUT,
 )
 def autogenerate_usage_report_task(*, tenant_id: str | None) -> None:

backend/ee/onyx/background/celery/tasks/beat_schedule.py

@@ -1,24 +1,80 @@
 from datetime import timedelta
 from typing import Any
 
+from onyx.background.celery.tasks.beat_schedule import BEAT_EXPIRES_DEFAULT
 from onyx.background.celery.tasks.beat_schedule import (
-    tasks_to_schedule as base_tasks_to_schedule,
+    beat_system_tasks as base_beat_system_tasks,
 )
+from onyx.background.celery.tasks.beat_schedule import (
+    beat_task_templates as base_beat_task_templates,
+)
+from onyx.background.celery.tasks.beat_schedule import generate_cloud_tasks
+from onyx.background.celery.tasks.beat_schedule import (
+    get_tasks_to_schedule as base_get_tasks_to_schedule,
+)
+from onyx.configs.constants import OnyxCeleryPriority
 from onyx.configs.constants import OnyxCeleryTask
+from shared_configs.configs import MULTI_TENANT
 
-ee_tasks_to_schedule = [
-    {
-        "name": "autogenerate_usage_report",
-        "task": OnyxCeleryTask.AUTOGENERATE_USAGE_REPORT_TASK,
-        "schedule": timedelta(days=30),  # TODO: change this to config flag
-    },
-    {
-        "name": "check-ttl-management",
-        "task": OnyxCeleryTask.CHECK_TTL_MANAGEMENT_TASK,
-        "schedule": timedelta(hours=1),
-    },
-]
+ee_beat_system_tasks: list[dict] = []
+
+ee_beat_task_templates: list[dict] = []
+ee_beat_task_templates.extend(
+    [
+        {
+            "name": "autogenerate-usage-report",
+            "task": OnyxCeleryTask.AUTOGENERATE_USAGE_REPORT_TASK,
+            "schedule": timedelta(days=30),
+            "options": {
+                "priority": OnyxCeleryPriority.MEDIUM,
+                "expires": BEAT_EXPIRES_DEFAULT,
+            },
+        },
+        {
+            "name": "check-ttl-management",
+            "task": OnyxCeleryTask.CHECK_TTL_MANAGEMENT_TASK,
+            "schedule": timedelta(hours=1),
+            "options": {
+                "priority": OnyxCeleryPriority.MEDIUM,
+                "expires": BEAT_EXPIRES_DEFAULT,
+            },
+        },
+    ]
+)
+
+ee_tasks_to_schedule: list[dict] = []
+
+if not MULTI_TENANT:
+    ee_tasks_to_schedule = [
+        {
+            "name": "autogenerate-usage-report",
+            "task": OnyxCeleryTask.AUTOGENERATE_USAGE_REPORT_TASK,
+            "schedule": timedelta(days=30),  # TODO: change this to config flag
+            "options": {
+                "priority": OnyxCeleryPriority.MEDIUM,
+                "expires": BEAT_EXPIRES_DEFAULT,
+            },
+        },
+        {
+            "name": "check-ttl-management",
+            "task": OnyxCeleryTask.CHECK_TTL_MANAGEMENT_TASK,
+            "schedule": timedelta(hours=1),
+            "options": {
+                "priority": OnyxCeleryPriority.MEDIUM,
+                "expires": BEAT_EXPIRES_DEFAULT,
+            },
+        },
+    ]
+
+
+def get_cloud_tasks_to_schedule(beat_multiplier: float) -> list[dict[str, Any]]:
+    beat_system_tasks = ee_beat_system_tasks + base_beat_system_tasks
+    beat_task_templates = ee_beat_task_templates + base_beat_task_templates
+    cloud_tasks = generate_cloud_tasks(
+        beat_system_tasks, beat_task_templates, beat_multiplier
+    )
+    return cloud_tasks
 
 
 def get_tasks_to_schedule() -> list[dict[str, Any]]:
-    return ee_tasks_to_schedule + base_tasks_to_schedule
+    return ee_tasks_to_schedule + base_get_tasks_to_schedule()

backend/ee/onyx/background/celery/tasks/vespa/tasks.py

@@ -8,6 +8,9 @@ from ee.onyx.db.user_group import fetch_user_group
 from ee.onyx.db.user_group import mark_user_group_as_synced
 from ee.onyx.db.user_group import prepare_user_group_for_deletion
 from onyx.background.celery.apps.app_base import task_logger
+from onyx.db.enums import SyncStatus
+from onyx.db.enums import SyncType
+from onyx.db.sync_record import update_sync_record_status
 from onyx.redis.redis_usergroup import RedisUserGroup
 from onyx.utils.logger import setup_logger
 
@@ -43,24 +46,59 @@ def monitor_usergroup_taskset(
         f"User group sync progress: usergroup_id={usergroup_id} remaining={count} initial={initial_count}"
     )
     if count > 0:
+        update_sync_record_status(
+            db_session=db_session,
+            entity_id=usergroup_id,
+            sync_type=SyncType.USER_GROUP,
+            sync_status=SyncStatus.IN_PROGRESS,
+            num_docs_synced=count,
+        )
         return
 
     user_group = fetch_user_group(db_session=db_session, user_group_id=usergroup_id)
     if user_group:
         usergroup_name = user_group.name
-        if user_group.is_up_for_deletion:
-            # this prepare should have been run when the deletion was scheduled,
-            # but run it again to be sure we're ready to go
-            mark_user_group_as_synced(db_session, user_group)
-            prepare_user_group_for_deletion(db_session, usergroup_id)
-            delete_user_group(db_session=db_session, user_group=user_group)
-            task_logger.info(
-                f"Deleted usergroup: name={usergroup_name} id={usergroup_id}"
-            )
-        else:
-            mark_user_group_as_synced(db_session=db_session, user_group=user_group)
-            task_logger.info(
-                f"Synced usergroup. name={usergroup_name} id={usergroup_id}"
+        try:
+            if user_group.is_up_for_deletion:
+                # this prepare should have been run when the deletion was scheduled,
+                # but run it again to be sure we're ready to go
+                mark_user_group_as_synced(db_session, user_group)
+                prepare_user_group_for_deletion(db_session, usergroup_id)
+                delete_user_group(db_session=db_session, user_group=user_group)
+
+                update_sync_record_status(
+                    db_session=db_session,
+                    entity_id=usergroup_id,
+                    sync_type=SyncType.USER_GROUP,
+                    sync_status=SyncStatus.SUCCESS,
+                    num_docs_synced=initial_count,
+                )
+
+                task_logger.info(
+                    f"Deleted usergroup: name={usergroup_name} id={usergroup_id}"
+                )
+            else:
+                mark_user_group_as_synced(db_session=db_session, user_group=user_group)
+
+                update_sync_record_status(
+                    db_session=db_session,
+                    entity_id=usergroup_id,
+                    sync_type=SyncType.USER_GROUP,
+                    sync_status=SyncStatus.SUCCESS,
+                    num_docs_synced=initial_count,
+                )
+
+                task_logger.info(
+                    f"Synced usergroup. name={usergroup_name} id={usergroup_id}"
+                )
+        except Exception as e:
+            update_sync_record_status(
+                db_session=db_session,
+                entity_id=usergroup_id,
+                sync_type=SyncType.USER_GROUP,
+                sync_status=SyncStatus.FAILED,
+                num_docs_synced=initial_count,
             )
+            raise e
 
     rug.reset()

backend/ee/onyx/configs/app_configs.py

@@ -4,6 +4,20 @@ import os
 # Applicable for OIDC Auth
 OPENID_CONFIG_URL = os.environ.get("OPENID_CONFIG_URL", "")
 
+# Applicable for OIDC Auth, allows you to override the scopes that
+# are requested from the OIDC provider. Currently used when passing
+# over access tokens to tool calls and the tool needs more scopes
+OIDC_SCOPE_OVERRIDE: list[str] | None = None
+_OIDC_SCOPE_OVERRIDE = os.environ.get("OIDC_SCOPE_OVERRIDE")
+
+if _OIDC_SCOPE_OVERRIDE:
+    try:
+        OIDC_SCOPE_OVERRIDE = [
+            scope.strip() for scope in _OIDC_SCOPE_OVERRIDE.split(",")
+        ]
+    except Exception:
+        pass
+
 # Applicable for SAML Auth
 SAML_CONF_DIR = os.environ.get("SAML_CONF_DIR") or "/app/ee/onyx/configs/saml_config"
 

backend/ee/onyx/db/document.py

@@ -5,7 +5,7 @@ from sqlalchemy import select
 from sqlalchemy.orm import Session
 
 from onyx.access.models import ExternalAccess
-from onyx.access.utils import prefix_group_w_source
+from onyx.access.utils import build_ext_group_name_for_onyx
 from onyx.configs.constants import DocumentSource
 from onyx.db.models import Document as DbDocument
 
@@ -25,7 +25,7 @@ def upsert_document_external_perms__no_commit(
     ).first()
 
     prefixed_external_groups = [
-        prefix_group_w_source(
+        build_ext_group_name_for_onyx(
             ext_group_name=group_id,
             source=source_type,
         )
@@ -66,7 +66,7 @@ def upsert_document_external_perms(
     ).first()
 
     prefixed_external_groups: set[str] = {
-        prefix_group_w_source(
+        build_ext_group_name_for_onyx(
             ext_group_name=group_id,
             source=source_type,
         )

backend/ee/onyx/db/external_perm.py

@@ -6,8 +6,9 @@ from sqlalchemy import delete
 from sqlalchemy import select
 from sqlalchemy.orm import Session
 
-from onyx.access.utils import prefix_group_w_source
+from onyx.access.utils import build_ext_group_name_for_onyx
 from onyx.configs.constants import DocumentSource
+from onyx.db.models import User
 from onyx.db.models import User__ExternalUserGroupId
 from onyx.db.users import batch_add_ext_perm_user_if_not_exists
 from onyx.db.users import get_user_by_email
@@ -61,8 +62,10 @@ def replace_user__ext_group_for_cc_pair(
             all_group_member_emails.add(user_email)
 
     # batch add users if they don't exist and get their ids
-    all_group_members = batch_add_ext_perm_user_if_not_exists(
-        db_session=db_session, emails=list(all_group_member_emails)
+    all_group_members: list[User] = batch_add_ext_perm_user_if_not_exists(
+        db_session=db_session,
+        # NOTE: this function handles case sensitivity for emails
+        emails=list(all_group_member_emails),
     )
 
     delete_user__ext_group_for_cc_pair__no_commit(
@@ -84,12 +87,14 @@ def replace_user__ext_group_for_cc_pair(
                     f" with email {user_email} not found"
                 )
                 continue
+            external_group_id = build_ext_group_name_for_onyx(
+                ext_group_name=external_group.id,
+                source=source,
+            )
             new_external_permissions.append(
                 User__ExternalUserGroupId(
                     user_id=user_id,
-                    external_user_group_id=prefix_group_w_source(
-                        external_group.id, source
-                    ),
+                    external_user_group_id=external_group_id,
                     cc_pair_id=cc_pair_id,
                 )
             )

backend/ee/onyx/db/persona.py

@@ -2,8 +2,11 @@ from uuid import UUID
 
 from sqlalchemy.orm import Session
 
+from onyx.configs.constants import NotificationType
 from onyx.db.models import Persona__User
 from onyx.db.models import Persona__UserGroup
+from onyx.db.notification import create_notification
+from onyx.server.features.persona.models import PersonaSharedNotificationData
 
 
 def make_persona_private(
@@ -12,6 +15,9 @@ def make_persona_private(
     group_ids: list[int] | None,
     db_session: Session,
 ) -> None:
+    """NOTE(rkuo): This function batches all updates into a single commit. If we don't
+    dedupe the inputs, the commit will exception."""
+
     db_session.query(Persona__User).filter(
         Persona__User.persona_id == persona_id
     ).delete(synchronize_session="fetch")
@@ -20,11 +26,22 @@ def make_persona_private(
     ).delete(synchronize_session="fetch")
 
     if user_ids:
-        for user_uuid in user_ids:
-            db_session.add(Persona__User(persona_id=persona_id, user_id=user_uuid))
+        user_ids_set = set(user_ids)
+        for user_id in user_ids_set:
+            db_session.add(Persona__User(persona_id=persona_id, user_id=user_id))
+
+            create_notification(
+                user_id=user_id,
+                notif_type=NotificationType.PERSONA_SHARED,
+                db_session=db_session,
+                additional_data=PersonaSharedNotificationData(
+                    persona_id=persona_id,
+                ).model_dump(),
+            )
 
     if group_ids:
-        for group_id in group_ids:
+        group_ids_set = set(group_ids)
+        for group_id in group_ids_set:
             db_session.add(
                 Persona__UserGroup(persona_id=persona_id, user_group_id=group_id)
             )

backend/ee/onyx/db/query_history.py

@@ -1,27 +1,138 @@
-import datetime
-from typing import Literal
+from collections.abc import Sequence
+from datetime import datetime
 
 from sqlalchemy import asc
 from sqlalchemy import BinaryExpression
 from sqlalchemy import ColumnElement
 from sqlalchemy import desc
+from sqlalchemy import distinct
 from sqlalchemy.orm import contains_eager
 from sqlalchemy.orm import joinedload
 from sqlalchemy.orm import Session
+from sqlalchemy.sql import case
+from sqlalchemy.sql import func
+from sqlalchemy.sql import select
+from sqlalchemy.sql.expression import literal
 from sqlalchemy.sql.expression import UnaryExpression
 
+from onyx.configs.constants import QAFeedbackType
 from onyx.db.models import ChatMessage
+from onyx.db.models import ChatMessageFeedback
 from onyx.db.models import ChatSession
 
-SortByOptions = Literal["time_sent"]
+
+def _build_filter_conditions(
+    start_time: datetime | None,
+    end_time: datetime | None,
+    feedback_filter: QAFeedbackType | None,
+) -> list[ColumnElement]:
+    """
+    Helper function to build all filter conditions for chat sessions.
+    Filters by start and end time, feedback type, and any sessions without messages.
+    start_time: Date from which to filter
+    end_time: Date to which to filter
+    feedback_filter: Feedback type to filter by
+    Returns: List of filter conditions
+    """
+    conditions = []
+
+    if start_time is not None:
+        conditions.append(ChatSession.time_created >= start_time)
+    if end_time is not None:
+        conditions.append(ChatSession.time_created <= end_time)
+
+    if feedback_filter is not None:
+        feedback_subq = (
+            select(ChatMessage.chat_session_id)
+            .join(ChatMessageFeedback)
+            .group_by(ChatMessage.chat_session_id)
+            .having(
+                case(
+                    (
+                        case(
+                            {literal(feedback_filter == QAFeedbackType.LIKE): True},
+                            else_=False,
+                        ),
+                        func.bool_and(ChatMessageFeedback.is_positive),
+                    ),
+                    (
+                        case(
+                            {literal(feedback_filter == QAFeedbackType.DISLIKE): True},
+                            else_=False,
+                        ),
+                        func.bool_and(func.not_(ChatMessageFeedback.is_positive)),
+                    ),
+                    else_=func.bool_or(ChatMessageFeedback.is_positive)
+                    & func.bool_or(func.not_(ChatMessageFeedback.is_positive)),
+                )
+            )
+        )
+        conditions.append(ChatSession.id.in_(feedback_subq))
+
+    return conditions
+
+
+def get_total_filtered_chat_sessions_count(
+    db_session: Session,
+    start_time: datetime | None,
+    end_time: datetime | None,
+    feedback_filter: QAFeedbackType | None,
+) -> int:
+    conditions = _build_filter_conditions(start_time, end_time, feedback_filter)
+    stmt = (
+        select(func.count(distinct(ChatSession.id)))
+        .select_from(ChatSession)
+        .filter(*conditions)
+    )
+    return db_session.scalar(stmt) or 0
+
+
+def get_page_of_chat_sessions(
+    start_time: datetime | None,
+    end_time: datetime | None,
+    db_session: Session,
+    page_num: int,
+    page_size: int,
+    feedback_filter: QAFeedbackType | None = None,
+) -> Sequence[ChatSession]:
+    conditions = _build_filter_conditions(start_time, end_time, feedback_filter)
+
+    subquery = (
+        select(ChatSession.id)
+        .filter(*conditions)
+        .order_by(desc(ChatSession.time_created), ChatSession.id)
+        .limit(page_size)
+        .offset(page_num * page_size)
+        .subquery()
+    )
+
+    stmt = (
+        select(ChatSession)
+        .join(subquery, ChatSession.id == subquery.c.id)
+        .outerjoin(ChatMessage, ChatSession.id == ChatMessage.chat_session_id)
+        .options(
+            joinedload(ChatSession.user),
+            joinedload(ChatSession.persona),
+            contains_eager(ChatSession.messages).joinedload(
+                ChatMessage.chat_message_feedbacks
+            ),
+        )
+        .order_by(
+            desc(ChatSession.time_created),
+            ChatSession.id,
+            asc(ChatMessage.id),  # Ensure chronological message order
+        )
+    )
+
+    return db_session.scalars(stmt).unique().all()
 
 
 def fetch_chat_sessions_eagerly_by_time(
-    start: datetime.datetime,
-    end: datetime.datetime,
+    start: datetime,
+    end: datetime,
     db_session: Session,
     limit: int | None = 500,
-    initial_time: datetime.datetime | None = None,
+    initial_time: datetime | None = None,
 ) -> list[ChatSession]:
     time_order: UnaryExpression = desc(ChatSession.time_created)
     message_order: UnaryExpression = asc(ChatMessage.id)

backend/ee/onyx/db/token_limit.py

@@ -111,10 +111,10 @@ def insert_user_group_token_rate_limit(
     return token_limit
 
 
-def fetch_user_group_token_rate_limits(
+def fetch_user_group_token_rate_limits_for_user(
     db_session: Session,
     group_id: int,
-    user: User | None = None,
+    user: User | None,
     enabled_only: bool = False,
     ordered: bool = True,
     get_editable: bool = True,

backend/ee/onyx/db/user_group.py

@@ -218,14 +218,14 @@ def fetch_user_groups_for_user(
     return db_session.scalars(stmt).all()
 
 
-def construct_document_select_by_usergroup(
+def construct_document_id_select_by_usergroup(
     user_group_id: int,
 ) -> Select:
     """This returns a statement that should be executed using
     .yield_per() to minimize overhead. The primary consumers of this function
     are background processing task generators."""
     stmt = (
-        select(Document)
+        select(Document.id)
         .join(
             DocumentByConnectorCredentialPair,
             Document.id == DocumentByConnectorCredentialPair.id,
@@ -374,7 +374,9 @@ def _add_user_group__cc_pair_relationships__no_commit(
 
 
 def insert_user_group(db_session: Session, user_group: UserGroupCreate) -> UserGroup:
-    db_user_group = UserGroup(name=user_group.name)
+    db_user_group = UserGroup(
+        name=user_group.name, time_last_modified_by_user=func.now()
+    )
     db_session.add(db_user_group)
     db_session.flush()  # give the group an ID
 
@@ -630,6 +632,10 @@ def update_user_group(
         select(User).where(User.id.in_(removed_user_ids))  # type: ignore
     ).unique()
     _validate_curator_status__no_commit(db_session, list(removed_users))
+
+    # update "time_updated" to now
+    db_user_group.time_last_modified_by_user = func.now()
+
     db_session.commit()
     return db_user_group
 
@@ -699,7 +705,10 @@ def delete_user_group_cc_pair_relationship__no_commit(
     connector_credential_pair_id matches the given cc_pair_id.
 
     Should be used very carefully (only for connectors that are being deleted)."""
-    cc_pair = get_connector_credential_pair_from_id(cc_pair_id, db_session)
+    cc_pair = get_connector_credential_pair_from_id(
+        db_session=db_session,
+        cc_pair_id=cc_pair_id,
+    )
     if not cc_pair:
         raise ValueError(f"Connector Credential Pair '{cc_pair_id}' does not exist")
 

backend/ee/onyx/external_permissions/confluence/doc_sync.py

@@ -13,6 +13,7 @@ from onyx.connectors.confluence.onyx_confluence import OnyxConfluence
 from onyx.connectors.confluence.utils import get_user_email_from_username__server
 from onyx.connectors.models import SlimDocument
 from onyx.db.models import ConnectorCredentialPair
+from onyx.indexing.indexing_heartbeat import IndexingHeartbeatInterface
 from onyx.utils.logger import setup_logger
 
 logger = setup_logger()
@@ -257,6 +258,7 @@ def _fetch_all_page_restrictions(
     slim_docs: list[SlimDocument],
     space_permissions_by_space_key: dict[str, ExternalAccess],
     is_cloud: bool,
+    callback: IndexingHeartbeatInterface | None,
 ) -> list[DocExternalAccess]:
     """
     For all pages, if a page has restrictions, then use those restrictions.
@@ -265,6 +267,12 @@ def _fetch_all_page_restrictions(
     document_restrictions: list[DocExternalAccess] = []
 
     for slim_doc in slim_docs:
+        if callback:
+            if callback.should_stop():
+                raise RuntimeError("confluence_doc_sync: Stop signal detected")
+
+            callback.progress("confluence_doc_sync:fetch_all_page_restrictions", 1)
+
         if slim_doc.perm_sync_data is None:
             raise ValueError(
                 f"No permission sync data found for document {slim_doc.id}"
@@ -334,7 +342,7 @@ def _fetch_all_page_restrictions(
 
 
 def confluence_doc_sync(
-    cc_pair: ConnectorCredentialPair,
+    cc_pair: ConnectorCredentialPair, callback: IndexingHeartbeatInterface | None
 ) -> list[DocExternalAccess]:
     """
     Adds the external permissions to the documents in postgres
@@ -357,8 +365,16 @@ def confluence_doc_sync(
 
     slim_docs = []
     logger.debug("Fetching all slim documents from confluence")
-    for doc_batch in confluence_connector.retrieve_all_slim_documents():
+    for doc_batch in confluence_connector.retrieve_all_slim_documents(
+        callback=callback
+    ):
         logger.debug(f"Got {len(doc_batch)} slim documents from confluence")
+        if callback:
+            if callback.should_stop():
+                raise RuntimeError("confluence_doc_sync: Stop signal detected")
+
+            callback.progress("confluence_doc_sync", 1)
+
         slim_docs.extend(doc_batch)
 
     logger.debug("Fetching all page restrictions for space")
@@ -367,4 +383,5 @@ def confluence_doc_sync(
         slim_docs=slim_docs,
         space_permissions_by_space_key=space_permissions_by_space_key,
         is_cloud=is_cloud,
+        callback=callback,
     )

backend/ee/onyx/external_permissions/confluence/group_sync.py

@@ -14,6 +14,8 @@ def _build_group_member_email_map(
 ) -> dict[str, set[str]]:
     group_member_emails: dict[str, set[str]] = {}
     for user_result in confluence_client.paginated_cql_user_retrieval():
+        logger.debug(f"Processing groups for user: {user_result}")
+
         user = user_result.get("user", {})
         if not user:
             logger.warning(f"user result missing user field: {user_result}")
@@ -33,10 +35,17 @@ def _build_group_member_email_map(
             logger.warning(f"user result missing email field: {user_result}")
             continue
 
+        all_users_groups: set[str] = set()
         for group in confluence_client.paginated_groups_by_user_retrieval(user):
             # group name uniqueness is enforced by Confluence, so we can use it as a group ID
             group_id = group["name"]
             group_member_emails.setdefault(group_id, set()).add(email)
+            all_users_groups.add(group_id)
+
+        if not group_member_emails:
+            logger.warning(f"No groups found for user with email: {email}")
+        else:
+            logger.debug(f"Found groups {all_users_groups} for user with email {email}")
 
     return group_member_emails
 

backend/ee/onyx/external_permissions/gmail/doc_sync.py

@@ -6,6 +6,7 @@ from onyx.access.models import ExternalAccess
 from onyx.connectors.gmail.connector import GmailConnector
 from onyx.connectors.interfaces import GenerateSlimDocumentOutput
 from onyx.db.models import ConnectorCredentialPair
+from onyx.indexing.indexing_heartbeat import IndexingHeartbeatInterface
 from onyx.utils.logger import setup_logger
 
 logger = setup_logger()
@@ -14,6 +15,7 @@ logger = setup_logger()
 def _get_slim_doc_generator(
     cc_pair: ConnectorCredentialPair,
     gmail_connector: GmailConnector,
+    callback: IndexingHeartbeatInterface | None = None,
 ) -> GenerateSlimDocumentOutput:
     current_time = datetime.now(timezone.utc)
     start_time = (
@@ -23,12 +25,14 @@ def _get_slim_doc_generator(
     )
 
     return gmail_connector.retrieve_all_slim_documents(
-        start=start_time, end=current_time.timestamp()
+        start=start_time,
+        end=current_time.timestamp(),
+        callback=callback,
     )
 
 
 def gmail_doc_sync(
-    cc_pair: ConnectorCredentialPair,
+    cc_pair: ConnectorCredentialPair, callback: IndexingHeartbeatInterface | None
 ) -> list[DocExternalAccess]:
     """
     Adds the external permissions to the documents in postgres
@@ -39,11 +43,19 @@ def gmail_doc_sync(
     gmail_connector = GmailConnector(**cc_pair.connector.connector_specific_config)
     gmail_connector.load_credentials(cc_pair.credential.credential_json)
 
-    slim_doc_generator = _get_slim_doc_generator(cc_pair, gmail_connector)
+    slim_doc_generator = _get_slim_doc_generator(
+        cc_pair, gmail_connector, callback=callback
+    )
 
     document_external_access: list[DocExternalAccess] = []
     for slim_doc_batch in slim_doc_generator:
         for slim_doc in slim_doc_batch:
+            if callback:
+                if callback.should_stop():
+                    raise RuntimeError("gmail_doc_sync: Stop signal detected")
+
+                callback.progress("gmail_doc_sync", 1)
+
             if slim_doc.perm_sync_data is None:
                 logger.warning(f"No permissions found for document {slim_doc.id}")
                 continue

backend/ee/onyx/external_permissions/google_drive/doc_sync.py

@@ -10,6 +10,7 @@ from onyx.connectors.google_utils.resources import get_drive_service
 from onyx.connectors.interfaces import GenerateSlimDocumentOutput
 from onyx.connectors.models import SlimDocument
 from onyx.db.models import ConnectorCredentialPair
+from onyx.indexing.indexing_heartbeat import IndexingHeartbeatInterface
 from onyx.utils.logger import setup_logger
 
 logger = setup_logger()
@@ -20,6 +21,7 @@ _PERMISSION_ID_PERMISSION_MAP: dict[str, dict[str, Any]] = {}
 def _get_slim_doc_generator(
     cc_pair: ConnectorCredentialPair,
     google_drive_connector: GoogleDriveConnector,
+    callback: IndexingHeartbeatInterface | None = None,
 ) -> GenerateSlimDocumentOutput:
     current_time = datetime.now(timezone.utc)
     start_time = (
@@ -29,7 +31,9 @@ def _get_slim_doc_generator(
     )
 
     return google_drive_connector.retrieve_all_slim_documents(
-        start=start_time, end=current_time.timestamp()
+        start=start_time,
+        end=current_time.timestamp(),
+        callback=callback,
     )
 
 
@@ -42,24 +46,22 @@ def _fetch_permissions_for_permission_ids(
     if not permission_info or not doc_id:
         return []
 
-    # Check cache first for all permission IDs
     permissions = [
         _PERMISSION_ID_PERMISSION_MAP[pid]
         for pid in permission_ids
         if pid in _PERMISSION_ID_PERMISSION_MAP
     ]
 
-    # If we found all permissions in cache, return them
     if len(permissions) == len(permission_ids):
         return permissions
 
     owner_email = permission_info.get("owner_email")
+
     drive_service = get_drive_service(
         creds=google_drive_connector.creds,
         user_email=(owner_email or google_drive_connector.primary_admin_email),
     )
 
-    # Otherwise, fetch all permissions and update cache
     fetched_permissions = execute_paginated_retrieval(
         retrieval_function=drive_service.permissions().list,
         list_key="permissions",
@@ -69,7 +71,6 @@ def _fetch_permissions_for_permission_ids(
     )
 
     permissions_for_doc_id = []
-    # Update cache and return all permissions
     for permission in fetched_permissions:
         permissions_for_doc_id.append(permission)
         _PERMISSION_ID_PERMISSION_MAP[permission["id"]] = permission
@@ -120,15 +121,18 @@ def _get_permissions_from_slim_doc(
         elif permission_type == "anyone":
             public = True
 
+    drive_id = permission_info.get("drive_id")
+    group_ids = group_emails | ({drive_id} if drive_id is not None else set())
+
     return ExternalAccess(
         external_user_emails=user_emails,
-        external_user_group_ids=group_emails,
+        external_user_group_ids=group_ids,
         is_public=public,
     )
 
 
 def gdrive_doc_sync(
-    cc_pair: ConnectorCredentialPair,
+    cc_pair: ConnectorCredentialPair, callback: IndexingHeartbeatInterface | None
 ) -> list[DocExternalAccess]:
     """
     Adds the external permissions to the documents in postgres
@@ -146,6 +150,12 @@ def gdrive_doc_sync(
     document_external_accesses = []
     for slim_doc_batch in slim_doc_generator:
         for slim_doc in slim_doc_batch:
+            if callback:
+                if callback.should_stop():
+                    raise RuntimeError("gdrive_doc_sync: Stop signal detected")
+
+                callback.progress("gdrive_doc_sync", 1)
+
             ext_access = _get_permissions_from_slim_doc(
                 google_drive_connector=google_drive_connector,
                 slim_doc=slim_doc,

backend/ee/onyx/external_permissions/google_drive/group_sync.py

@@ -1,16 +1,127 @@
 from ee.onyx.db.external_perm import ExternalUserGroup
 from onyx.connectors.google_drive.connector import GoogleDriveConnector
 from onyx.connectors.google_utils.google_utils import execute_paginated_retrieval
+from onyx.connectors.google_utils.resources import AdminService
 from onyx.connectors.google_utils.resources import get_admin_service
+from onyx.connectors.google_utils.resources import get_drive_service
 from onyx.db.models import ConnectorCredentialPair
 from onyx.utils.logger import setup_logger
 
 logger = setup_logger()
 
 
+def _get_drive_members(
+    google_drive_connector: GoogleDriveConnector,
+) -> dict[str, tuple[set[str], set[str]]]:
+    """
+    This builds a map of drive ids to their members (group and user emails).
+    E.g. {
+        "drive_id_1": ({"group_email_1"}, {"user_email_1", "user_email_2"}),
+        "drive_id_2": ({"group_email_3"}, {"user_email_3"}),
+    }
+    """
+    drive_ids = google_drive_connector.get_all_drive_ids()
+
+    drive_id_to_members_map: dict[str, tuple[set[str], set[str]]] = {}
+    drive_service = get_drive_service(
+        google_drive_connector.creds,
+        google_drive_connector.primary_admin_email,
+    )
+
+    for drive_id in drive_ids:
+        group_emails: set[str] = set()
+        user_emails: set[str] = set()
+        for permission in execute_paginated_retrieval(
+            drive_service.permissions().list,
+            list_key="permissions",
+            fileId=drive_id,
+            fields="permissions(emailAddress, type)",
+            supportsAllDrives=True,
+        ):
+            if permission["type"] == "group":
+                group_emails.add(permission["emailAddress"])
+            elif permission["type"] == "user":
+                user_emails.add(permission["emailAddress"])
+        drive_id_to_members_map[drive_id] = (group_emails, user_emails)
+    return drive_id_to_members_map
+
+
+def _get_all_groups(
+    admin_service: AdminService,
+    google_domain: str,
+) -> set[str]:
+    """
+    This gets all the group emails.
+    """
+    group_emails: set[str] = set()
+    for group in execute_paginated_retrieval(
+        admin_service.groups().list,
+        list_key="groups",
+        domain=google_domain,
+        fields="groups(email)",
+    ):
+        group_emails.add(group["email"])
+    return group_emails
+
+
+def _map_group_email_to_member_emails(
+    admin_service: AdminService,
+    group_emails: set[str],
+) -> dict[str, set[str]]:
+    """
+    This maps group emails to their member emails.
+    """
+    group_to_member_map: dict[str, set[str]] = {}
+    for group_email in group_emails:
+        group_member_emails: set[str] = set()
+        for member in execute_paginated_retrieval(
+            admin_service.members().list,
+            list_key="members",
+            groupKey=group_email,
+            fields="members(email)",
+        ):
+            group_member_emails.add(member["email"])
+
+        group_to_member_map[group_email] = group_member_emails
+    return group_to_member_map
+
+
+def _build_onyx_groups(
+    drive_id_to_members_map: dict[str, tuple[set[str], set[str]]],
+    group_email_to_member_emails_map: dict[str, set[str]],
+) -> list[ExternalUserGroup]:
+    onyx_groups: list[ExternalUserGroup] = []
+
+    # Convert all drive member definitions to onyx groups
+    # This is because having drive level access means you have
+    # irrevocable access to all the files in the drive.
+    for drive_id, (group_emails, user_emails) in drive_id_to_members_map.items():
+        all_member_emails: set[str] = user_emails
+        for group_email in group_emails:
+            all_member_emails.update(group_email_to_member_emails_map[group_email])
+        onyx_groups.append(
+            ExternalUserGroup(
+                id=drive_id,
+                user_emails=list(all_member_emails),
+            )
+        )
+
+    # Convert all group member definitions to onyx groups
+    for group_email, member_emails in group_email_to_member_emails_map.items():
+        onyx_groups.append(
+            ExternalUserGroup(
+                id=group_email,
+                user_emails=list(member_emails),
+            )
+        )
+
+    return onyx_groups
+
+
 def gdrive_group_sync(
     cc_pair: ConnectorCredentialPair,
 ) -> list[ExternalUserGroup]:
+    # Initialize connector and build credential/service objects
     google_drive_connector = GoogleDriveConnector(
         **cc_pair.connector.connector_specific_config
     )
@@ -19,34 +130,23 @@ def gdrive_group_sync(
         google_drive_connector.creds, google_drive_connector.primary_admin_email
     )
 
-    onyx_groups: list[ExternalUserGroup] = []
-    for group in execute_paginated_retrieval(
-        admin_service.groups().list,
-        list_key="groups",
-        domain=google_drive_connector.google_domain,
-        fields="groups(email)",
-    ):
-        # The id is the group email
-        group_email = group["email"]
+    # Get all drive members
+    drive_id_to_members_map = _get_drive_members(google_drive_connector)
 
-        # Gather group member emails
-        group_member_emails: list[str] = []
-        for member in execute_paginated_retrieval(
-            admin_service.members().list,
-            list_key="members",
-            groupKey=group_email,
-            fields="members(email)",
-        ):
-            group_member_emails.append(member["email"])
+    # Get all group emails
+    all_group_emails = _get_all_groups(
+        admin_service, google_drive_connector.google_domain
+    )
 
-        if not group_member_emails:
-            continue
+    # Map group emails to their members
+    group_email_to_member_emails_map = _map_group_email_to_member_emails(
+        admin_service, all_group_emails
+    )
 
-        onyx_groups.append(
-            ExternalUserGroup(
-                id=group_email,
-                user_emails=list(group_member_emails),
-            )
-        )
+    # Convert the maps to onyx groups
+    onyx_groups = _build_onyx_groups(
+        drive_id_to_members_map=drive_id_to_members_map,
+        group_email_to_member_emails_map=group_email_to_member_emails_map,
+    )
 
     return onyx_groups

backend/ee/onyx/external_permissions/salesforce/utils.py

@@ -161,7 +161,10 @@ def _get_salesforce_client_for_doc_id(db_session: Session, doc_id: str) -> Sales
 
     cc_pair_id = _DOC_ID_TO_CC_PAIR_ID_MAP[doc_id]
     if cc_pair_id not in _CC_PAIR_ID_SALESFORCE_CLIENT_MAP:
-        cc_pair = get_connector_credential_pair_from_id(cc_pair_id, db_session)
+        cc_pair = get_connector_credential_pair_from_id(
+            db_session=db_session,
+            cc_pair_id=cc_pair_id,
+        )
         if cc_pair is None:
             raise ValueError(f"CC pair {cc_pair_id} not found")
         credential_json = cc_pair.credential.credential_json

backend/ee/onyx/external_permissions/slack/doc_sync.py

@@ -7,6 +7,7 @@ from onyx.connectors.slack.connector import get_channels
 from onyx.connectors.slack.connector import make_paginated_slack_api_call_w_retries
 from onyx.connectors.slack.connector import SlackPollConnector
 from onyx.db.models import ConnectorCredentialPair
+from onyx.indexing.indexing_heartbeat import IndexingHeartbeatInterface
 from onyx.utils.logger import setup_logger
 
 
@@ -14,12 +15,12 @@ logger = setup_logger()
 
 
 def _get_slack_document_ids_and_channels(
-    cc_pair: ConnectorCredentialPair,
+    cc_pair: ConnectorCredentialPair, callback: IndexingHeartbeatInterface | None
 ) -> dict[str, list[str]]:
     slack_connector = SlackPollConnector(**cc_pair.connector.connector_specific_config)
     slack_connector.load_credentials(cc_pair.credential.credential_json)
 
-    slim_doc_generator = slack_connector.retrieve_all_slim_documents()
+    slim_doc_generator = slack_connector.retrieve_all_slim_documents(callback=callback)
 
     channel_doc_map: dict[str, list[str]] = {}
     for doc_metadata_batch in slim_doc_generator:
@@ -31,6 +32,14 @@ def _get_slack_document_ids_and_channels(
                 channel_doc_map[channel_id] = []
             channel_doc_map[channel_id].append(doc_metadata.id)
 
+        if callback:
+            if callback.should_stop():
+                raise RuntimeError(
+                    "_get_slack_document_ids_and_channels: Stop signal detected"
+                )
+
+            callback.progress("_get_slack_document_ids_and_channels", 1)
+
     return channel_doc_map
 
 
@@ -114,7 +123,7 @@ def _fetch_channel_permissions(
 
 
 def slack_doc_sync(
-    cc_pair: ConnectorCredentialPair,
+    cc_pair: ConnectorCredentialPair, callback: IndexingHeartbeatInterface | None
 ) -> list[DocExternalAccess]:
     """
     Adds the external permissions to the documents in postgres
@@ -127,7 +136,7 @@ def slack_doc_sync(
     )
     user_id_to_email_map = fetch_user_id_to_email_map(slack_client)
     channel_doc_map = _get_slack_document_ids_and_channels(
-        cc_pair=cc_pair,
+        cc_pair=cc_pair, callback=callback
     )
     workspace_permissions = _fetch_workspace_permissions(
         user_id_to_email_map=user_id_to_email_map,

backend/ee/onyx/external_permissions/sync_params.py

@@ -15,11 +15,13 @@ from ee.onyx.external_permissions.slack.doc_sync import slack_doc_sync
 from onyx.access.models import DocExternalAccess
 from onyx.configs.constants import DocumentSource
 from onyx.db.models import ConnectorCredentialPair
+from onyx.indexing.indexing_heartbeat import IndexingHeartbeatInterface
 
 # Defining the input/output types for the sync functions
 DocSyncFuncType = Callable[
     [
         ConnectorCredentialPair,
+        IndexingHeartbeatInterface | None,
     ],
     list[DocExternalAccess],
 ]

backend/ee/onyx/main.py

@@ -1,7 +1,9 @@
 from fastapi import FastAPI
 from httpx_oauth.clients.google import GoogleOAuth2
+from httpx_oauth.clients.openid import BASE_SCOPES
 from httpx_oauth.clients.openid import OpenID
 
+from ee.onyx.configs.app_configs import OIDC_SCOPE_OVERRIDE
 from ee.onyx.configs.app_configs import OPENID_CONFIG_URL
 from ee.onyx.server.analytics.api import router as analytics_router
 from ee.onyx.server.auth_check import check_ee_router_auth
@@ -88,7 +90,13 @@ def get_application() -> FastAPI:
         include_auth_router_with_prefix(
             application,
             create_onyx_oauth_router(
-                OpenID(OAUTH_CLIENT_ID, OAUTH_CLIENT_SECRET, OPENID_CONFIG_URL),
+                OpenID(
+                    OAUTH_CLIENT_ID,
+                    OAUTH_CLIENT_SECRET,
+                    OPENID_CONFIG_URL,
+                    # BASE_SCOPES is the same as not setting this
+                    base_scopes=OIDC_SCOPE_OVERRIDE or BASE_SCOPES,
+                ),
                 auth_backend,
                 USER_AUTH_SECRET,
                 associate_by_email=True,

backend/ee/onyx/onyxbot/slack/handlers/handle_standard_answers.py

@@ -80,7 +80,7 @@ def oneoff_standard_answers(
 def _handle_standard_answers(
     message_info: SlackMessageInfo,
     receiver_ids: list[str] | None,
-    slack_channel_config: SlackChannelConfig | None,
+    slack_channel_config: SlackChannelConfig,
     prompt: Prompt | None,
     logger: OnyxLoggingAdapter,
     client: WebClient,
@@ -94,13 +94,10 @@ def _handle_standard_answers(
     Returns True if standard answers are found to match the user's message and therefore,
     we still need to respond to the users.
     """
-    # if no channel config, then no standard answers are configured
-    if not slack_channel_config:
-        return False
 
     slack_thread_id = message_info.thread_to_respond
     configured_standard_answer_categories = (
-        slack_channel_config.standard_answer_categories if slack_channel_config else []
+        slack_channel_config.standard_answer_categories
     )
     configured_standard_answers = set(
         [
@@ -150,9 +147,9 @@ def _handle_standard_answers(
             db_session=db_session,
             description="",
             user_id=None,
-            persona_id=slack_channel_config.persona.id
-            if slack_channel_config.persona
-            else 0,
+            persona_id=(
+                slack_channel_config.persona.id if slack_channel_config.persona else 0
+            ),
             onyxbot_flow=True,
             slack_thread_id=slack_thread_id,
         )
@@ -182,7 +179,7 @@ def _handle_standard_answers(
             formatted_answers.append(formatted_answer)
         answer_message = "\n\n".join(formatted_answers)
 
-        _ = create_new_chat_message(
+        chat_message = create_new_chat_message(
             chat_session_id=chat_session.id,
             parent_message=new_user_message,
             prompt_id=prompt.id if prompt else None,
@@ -191,8 +188,13 @@ def _handle_standard_answers(
             message_type=MessageType.ASSISTANT,
             error=None,
             db_session=db_session,
-            commit=True,
+            commit=False,
         )
+        # attach the standard answers to the chat message
+        chat_message.standard_answers = [
+            standard_answer for standard_answer, _ in matching_standard_answers
+        ]
+        db_session.commit()
 
         update_emote_react(
             emoji=DANSWER_REACT_EMOJI,

backend/ee/onyx/server/analytics/api.py

@@ -228,8 +228,6 @@ def get_assistant_stats(
         datetime.datetime.utcnow() - datetime.timedelta(days=_DEFAULT_LOOKBACK_DAYS)
     )
     end = end or datetime.datetime.utcnow()
-    print("current user")
-    print(user)
 
     if not user_can_view_assistant_stats(db_session, user, assistant_id):
         raise HTTPException(

backend/ee/onyx/server/middleware/tenant_tracking.py

@@ -10,6 +10,7 @@ from fastapi import Response
 from ee.onyx.auth.users import decode_anonymous_user_jwt_token
 from ee.onyx.configs.app_configs import ANONYMOUS_USER_COOKIE_NAME
 from onyx.auth.api_key import extract_tenant_from_api_key_header
+from onyx.configs.constants import TENANT_ID_COOKIE_NAME
 from onyx.db.engine import is_valid_schema_name
 from onyx.redis.redis_pool import retrieve_auth_token_data_from_redis
 from shared_configs.configs import MULTI_TENANT
@@ -43,6 +44,7 @@ async def _get_tenant_id_from_request(
     Attempt to extract tenant_id from:
     1) The API key header
     2) The Redis-based token (stored in Cookie: fastapiusersauth)
+    3)  Reset token cookie
     Fallback: POSTGRES_DEFAULT_SCHEMA
     """
     # Check for API key
@@ -62,6 +64,7 @@ async def _get_tenant_id_from_request(
 
     try:
         # Look up token data in Redis
+
         token_data = await retrieve_auth_token_data_from_redis(request)
 
         if not token_data:
@@ -85,8 +88,18 @@ async def _get_tenant_id_from_request(
         if not is_valid_schema_name(tenant_id):
             raise HTTPException(status_code=400, detail="Invalid tenant ID format")
 
-        return tenant_id
-
     except Exception as e:
         logger.error(f"Unexpected error in _get_tenant_id_from_request: {str(e)}")
         raise HTTPException(status_code=500, detail="Internal server error")
+
+    finally:
+        if tenant_id:
+            return tenant_id
+
+        # As a final step, check for explicit tenant_id cookie
+        tenant_id_cookie = request.cookies.get(TENANT_ID_COOKIE_NAME)
+        if tenant_id_cookie and is_valid_schema_name(tenant_id_cookie):
+            return tenant_id_cookie
+
+        # If we've reached this point, return the default schema
+        return POSTGRES_DEFAULT_SCHEMA

backend/ee/onyx/server/oauth.py

@@ -286,6 +286,7 @@ def prepare_authorization_request(
     oauth_state = (
         base64.urlsafe_b64encode(oauth_uuid.bytes).rstrip(b"=").decode("utf-8")
     )
+    session: str
 
     if connector == DocumentSource.SLACK:
         oauth_url = SlackOAuth.generate_oauth_url(oauth_state)
@@ -554,6 +555,7 @@ def handle_google_drive_oauth_callback(
         )
 
     session_json = session_json_bytes.decode("utf-8")
+    session: GoogleDriveOAuth.OAuthSession
     try:
         session = GoogleDriveOAuth.parse_session(session_json)
 

backend/ee/onyx/server/query_and_chat/models.py

@@ -125,10 +125,10 @@ class OneShotQARequest(ChunkContext):
     # will also disable Thread-based Rewording if specified
     query_override: str | None = None
 
-    # If True, skips generative an AI response to the search query
+    # If True, skips generating an AI response to the search query
     skip_gen_ai_answer_generation: bool = False
 
-    # If True, uses pro search instead of basic search
+    # If True, uses agentic search instead of basic search
     use_agentic_search: bool = False
 
     @model_validator(mode="after")

backend/ee/onyx/server/query_and_chat/query_backend.py

@@ -197,6 +197,7 @@ def get_answer_stream(
         rerank_settings=query_request.rerank_settings,
         db_session=db_session,
         use_agentic_search=query_request.use_agentic_search,
+        skip_gen_ai_answer_generation=query_request.skip_gen_ai_answer_generation,
     )
 
     packets = stream_chat_message_objects(

backend/ee/onyx/server/query_history/api.py

@@ -1,19 +1,23 @@
 import csv
 import io
 from datetime import datetime
-from datetime import timedelta
 from datetime import timezone
-from typing import Literal
 from uuid import UUID
 
 from fastapi import APIRouter
 from fastapi import Depends
 from fastapi import HTTPException
+from fastapi import Query
 from fastapi.responses import StreamingResponse
-from pydantic import BaseModel
 from sqlalchemy.orm import Session
 
 from ee.onyx.db.query_history import fetch_chat_sessions_eagerly_by_time
+from ee.onyx.db.query_history import get_page_of_chat_sessions
+from ee.onyx.db.query_history import get_total_filtered_chat_sessions_count
+from ee.onyx.server.query_history.models import ChatSessionMinimal
+from ee.onyx.server.query_history.models import ChatSessionSnapshot
+from ee.onyx.server.query_history.models import MessageSnapshot
+from ee.onyx.server.query_history.models import QuestionAnswerPairSnapshot
 from onyx.auth.users import current_admin_user
 from onyx.auth.users import get_display_email
 from onyx.chat.chat_utils import create_chat_chain
@@ -23,257 +27,15 @@ from onyx.configs.constants import SessionType
 from onyx.db.chat import get_chat_session_by_id
 from onyx.db.chat import get_chat_sessions_by_user
 from onyx.db.engine import get_session
-from onyx.db.models import ChatMessage
 from onyx.db.models import ChatSession
 from onyx.db.models import User
+from onyx.server.documents.models import PaginatedReturn
 from onyx.server.query_and_chat.models import ChatSessionDetails
 from onyx.server.query_and_chat.models import ChatSessionsResponse
 
 router = APIRouter()
 
 
-class AbridgedSearchDoc(BaseModel):
-    """A subset of the info present in `SearchDoc`"""
-
-    document_id: str
-    semantic_identifier: str
-    link: str | None
-
-
-class MessageSnapshot(BaseModel):
-    message: str
-    message_type: MessageType
-    documents: list[AbridgedSearchDoc]
-    feedback_type: QAFeedbackType | None
-    feedback_text: str | None
-    time_created: datetime
-
-    @classmethod
-    def build(cls, message: ChatMessage) -> "MessageSnapshot":
-        latest_messages_feedback_obj = (
-            message.chat_message_feedbacks[-1]
-            if len(message.chat_message_feedbacks) > 0
-            else None
-        )
-        feedback_type = (
-            (
-                QAFeedbackType.LIKE
-                if latest_messages_feedback_obj.is_positive
-                else QAFeedbackType.DISLIKE
-            )
-            if latest_messages_feedback_obj
-            else None
-        )
-        feedback_text = (
-            latest_messages_feedback_obj.feedback_text
-            if latest_messages_feedback_obj
-            else None
-        )
-        return cls(
-            message=message.message,
-            message_type=message.message_type,
-            documents=[
-                AbridgedSearchDoc(
-                    document_id=document.document_id,
-                    semantic_identifier=document.semantic_id,
-                    link=document.link,
-                )
-                for document in message.search_docs
-            ],
-            feedback_type=feedback_type,
-            feedback_text=feedback_text,
-            time_created=message.time_sent,
-        )
-
-
-class ChatSessionMinimal(BaseModel):
-    id: UUID
-    user_email: str
-    name: str | None
-    first_user_message: str
-    first_ai_message: str
-    assistant_id: int | None
-    assistant_name: str | None
-    time_created: datetime
-    feedback_type: QAFeedbackType | Literal["mixed"] | None
-    flow_type: SessionType
-    conversation_length: int
-
-
-class ChatSessionSnapshot(BaseModel):
-    id: UUID
-    user_email: str
-    name: str | None
-    messages: list[MessageSnapshot]
-    assistant_id: int | None
-    assistant_name: str | None
-    time_created: datetime
-    flow_type: SessionType
-
-
-class QuestionAnswerPairSnapshot(BaseModel):
-    chat_session_id: UUID
-    # 1-indexed message number in the chat_session
-    # e.g. the first message pair in the chat_session is 1, the second is 2, etc.
-    message_pair_num: int
-    user_message: str
-    ai_response: str
-    retrieved_documents: list[AbridgedSearchDoc]
-    feedback_type: QAFeedbackType | None
-    feedback_text: str | None
-    persona_name: str | None
-    user_email: str
-    time_created: datetime
-    flow_type: SessionType
-
-    @classmethod
-    def from_chat_session_snapshot(
-        cls,
-        chat_session_snapshot: ChatSessionSnapshot,
-    ) -> list["QuestionAnswerPairSnapshot"]:
-        message_pairs: list[tuple[MessageSnapshot, MessageSnapshot]] = []
-        for ind in range(1, len(chat_session_snapshot.messages), 2):
-            message_pairs.append(
-                (
-                    chat_session_snapshot.messages[ind - 1],
-                    chat_session_snapshot.messages[ind],
-                )
-            )
-
-        return [
-            cls(
-                chat_session_id=chat_session_snapshot.id,
-                message_pair_num=ind + 1,
-                user_message=user_message.message,
-                ai_response=ai_message.message,
-                retrieved_documents=ai_message.documents,
-                feedback_type=ai_message.feedback_type,
-                feedback_text=ai_message.feedback_text,
-                persona_name=chat_session_snapshot.assistant_name,
-                user_email=get_display_email(chat_session_snapshot.user_email),
-                time_created=user_message.time_created,
-                flow_type=chat_session_snapshot.flow_type,
-            )
-            for ind, (user_message, ai_message) in enumerate(message_pairs)
-        ]
-
-    def to_json(self) -> dict[str, str | None]:
-        return {
-            "chat_session_id": str(self.chat_session_id),
-            "message_pair_num": str(self.message_pair_num),
-            "user_message": self.user_message,
-            "ai_response": self.ai_response,
-            "retrieved_documents": "|".join(
-                [
-                    doc.link or doc.semantic_identifier
-                    for doc in self.retrieved_documents
-                ]
-            ),
-            "feedback_type": self.feedback_type.value if self.feedback_type else "",
-            "feedback_text": self.feedback_text or "",
-            "persona_name": self.persona_name,
-            "user_email": self.user_email,
-            "time_created": str(self.time_created),
-            "flow_type": self.flow_type,
-        }
-
-
-def determine_flow_type(chat_session: ChatSession) -> SessionType:
-    return SessionType.SLACK if chat_session.onyxbot_flow else SessionType.CHAT
-
-
-def fetch_and_process_chat_session_history_minimal(
-    db_session: Session,
-    start: datetime,
-    end: datetime,
-    feedback_filter: QAFeedbackType | None = None,
-    limit: int | None = 500,
-) -> list[ChatSessionMinimal]:
-    chat_sessions = fetch_chat_sessions_eagerly_by_time(
-        start=start, end=end, db_session=db_session, limit=limit
-    )
-
-    minimal_sessions = []
-    for chat_session in chat_sessions:
-        if not chat_session.messages:
-            continue
-
-        first_user_message = next(
-            (
-                message.message
-                for message in chat_session.messages
-                if message.message_type == MessageType.USER
-            ),
-            "",
-        )
-        first_ai_message = next(
-            (
-                message.message
-                for message in chat_session.messages
-                if message.message_type == MessageType.ASSISTANT
-            ),
-            "",
-        )
-
-        has_positive_feedback = any(
-            feedback.is_positive
-            for message in chat_session.messages
-            for feedback in message.chat_message_feedbacks
-        )
-
-        has_negative_feedback = any(
-            not feedback.is_positive
-            for message in chat_session.messages
-            for feedback in message.chat_message_feedbacks
-        )
-
-        feedback_type: QAFeedbackType | Literal["mixed"] | None = (
-            "mixed"
-            if has_positive_feedback and has_negative_feedback
-            else QAFeedbackType.LIKE
-            if has_positive_feedback
-            else QAFeedbackType.DISLIKE
-            if has_negative_feedback
-            else None
-        )
-
-        if feedback_filter:
-            if feedback_filter == QAFeedbackType.LIKE and not has_positive_feedback:
-                continue
-            if feedback_filter == QAFeedbackType.DISLIKE and not has_negative_feedback:
-                continue
-
-        flow_type = determine_flow_type(chat_session)
-
-        minimal_sessions.append(
-            ChatSessionMinimal(
-                id=chat_session.id,
-                user_email=get_display_email(
-                    chat_session.user.email if chat_session.user else None
-                ),
-                name=chat_session.description,
-                first_user_message=first_user_message,
-                first_ai_message=first_ai_message,
-                assistant_id=chat_session.persona_id,
-                assistant_name=(
-                    chat_session.persona.name if chat_session.persona else None
-                ),
-                time_created=chat_session.time_created,
-                feedback_type=feedback_type,
-                flow_type=flow_type,
-                conversation_length=len(
-                    [
-                        m
-                        for m in chat_session.messages
-                        if m.message_type != MessageType.SYSTEM
-                    ]
-                ),
-            )
-        )
-
-    return minimal_sessions
-
-
 def fetch_and_process_chat_session_history(
     db_session: Session,
     start: datetime,
@@ -319,7 +81,7 @@ def snapshot_from_chat_session(
     except RuntimeError:
         return None
 
-    flow_type = determine_flow_type(chat_session)
+    flow_type = SessionType.SLACK if chat_session.onyxbot_flow else SessionType.CHAT
 
     return ChatSessionSnapshot(
         id=chat_session.id,
@@ -371,22 +133,38 @@ def get_user_chat_sessions(
 
 @router.get("/admin/chat-session-history")
 def get_chat_session_history(
+    page_num: int = Query(0, ge=0),
+    page_size: int = Query(10, ge=1),
     feedback_type: QAFeedbackType | None = None,
-    start: datetime | None = None,
-    end: datetime | None = None,
+    start_time: datetime | None = None,
+    end_time: datetime | None = None,
     _: User | None = Depends(current_admin_user),
     db_session: Session = Depends(get_session),
-) -> list[ChatSessionMinimal]:
-    return fetch_and_process_chat_session_history_minimal(
+) -> PaginatedReturn[ChatSessionMinimal]:
+    page_of_chat_sessions = get_page_of_chat_sessions(
+        page_num=page_num,
+        page_size=page_size,
         db_session=db_session,
-        start=start
-        or (
-            datetime.now(tz=timezone.utc) - timedelta(days=30)
-        ),  # default is 30d lookback
-        end=end or datetime.now(tz=timezone.utc),
+        start_time=start_time,
+        end_time=end_time,
         feedback_filter=feedback_type,
     )
 
+    total_filtered_chat_sessions_count = get_total_filtered_chat_sessions_count(
+        db_session=db_session,
+        start_time=start_time,
+        end_time=end_time,
+        feedback_filter=feedback_type,
+    )
+
+    return PaginatedReturn(
+        items=[
+            ChatSessionMinimal.from_chat_session(chat_session)
+            for chat_session in page_of_chat_sessions
+        ],
+        total_items=total_filtered_chat_sessions_count,
+    )
+
 
 @router.get("/admin/chat-session-history/{chat_session_id}")
 def get_chat_session_admin(

backend/ee/onyx/server/query_history/models.py

@@ -0,0 +1,218 @@
+from datetime import datetime
+from uuid import UUID
+
+from pydantic import BaseModel
+
+from onyx.auth.users import get_display_email
+from onyx.configs.constants import MessageType
+from onyx.configs.constants import QAFeedbackType
+from onyx.configs.constants import SessionType
+from onyx.db.models import ChatMessage
+from onyx.db.models import ChatSession
+
+
+class AbridgedSearchDoc(BaseModel):
+    """A subset of the info present in `SearchDoc`"""
+
+    document_id: str
+    semantic_identifier: str
+    link: str | None
+
+
+class MessageSnapshot(BaseModel):
+    id: int
+    message: str
+    message_type: MessageType
+    documents: list[AbridgedSearchDoc]
+    feedback_type: QAFeedbackType | None
+    feedback_text: str | None
+    time_created: datetime
+
+    @classmethod
+    def build(cls, message: ChatMessage) -> "MessageSnapshot":
+        latest_messages_feedback_obj = (
+            message.chat_message_feedbacks[-1]
+            if len(message.chat_message_feedbacks) > 0
+            else None
+        )
+        feedback_type = (
+            (
+                QAFeedbackType.LIKE
+                if latest_messages_feedback_obj.is_positive
+                else QAFeedbackType.DISLIKE
+            )
+            if latest_messages_feedback_obj
+            else None
+        )
+        feedback_text = (
+            latest_messages_feedback_obj.feedback_text
+            if latest_messages_feedback_obj
+            else None
+        )
+        return cls(
+            id=message.id,
+            message=message.message,
+            message_type=message.message_type,
+            documents=[
+                AbridgedSearchDoc(
+                    document_id=document.document_id,
+                    semantic_identifier=document.semantic_id,
+                    link=document.link,
+                )
+                for document in message.search_docs
+            ],
+            feedback_type=feedback_type,
+            feedback_text=feedback_text,
+            time_created=message.time_sent,
+        )
+
+
+class ChatSessionMinimal(BaseModel):
+    id: UUID
+    user_email: str
+    name: str | None
+    first_user_message: str
+    first_ai_message: str
+    assistant_id: int | None
+    assistant_name: str | None
+    time_created: datetime
+    feedback_type: QAFeedbackType | None
+    flow_type: SessionType
+    conversation_length: int
+
+    @classmethod
+    def from_chat_session(cls, chat_session: ChatSession) -> "ChatSessionMinimal":
+        first_user_message = next(
+            (
+                message.message
+                for message in chat_session.messages
+                if message.message_type == MessageType.USER
+            ),
+            "",
+        )
+        first_ai_message = next(
+            (
+                message.message
+                for message in chat_session.messages
+                if message.message_type == MessageType.ASSISTANT
+            ),
+            "",
+        )
+
+        list_of_message_feedbacks = [
+            feedback.is_positive
+            for message in chat_session.messages
+            for feedback in message.chat_message_feedbacks
+        ]
+        session_feedback_type = None
+        if list_of_message_feedbacks:
+            if all(list_of_message_feedbacks):
+                session_feedback_type = QAFeedbackType.LIKE
+            elif not any(list_of_message_feedbacks):
+                session_feedback_type = QAFeedbackType.DISLIKE
+            else:
+                session_feedback_type = QAFeedbackType.MIXED
+
+        return cls(
+            id=chat_session.id,
+            user_email=get_display_email(
+                chat_session.user.email if chat_session.user else None
+            ),
+            name=chat_session.description,
+            first_user_message=first_user_message,
+            first_ai_message=first_ai_message,
+            assistant_id=chat_session.persona_id,
+            assistant_name=(
+                chat_session.persona.name if chat_session.persona else None
+            ),
+            time_created=chat_session.time_created,
+            feedback_type=session_feedback_type,
+            flow_type=SessionType.SLACK
+            if chat_session.onyxbot_flow
+            else SessionType.CHAT,
+            conversation_length=len(
+                [
+                    message
+                    for message in chat_session.messages
+                    if message.message_type != MessageType.SYSTEM
+                ]
+            ),
+        )
+
+
+class ChatSessionSnapshot(BaseModel):
+    id: UUID
+    user_email: str
+    name: str | None
+    messages: list[MessageSnapshot]
+    assistant_id: int | None
+    assistant_name: str | None
+    time_created: datetime
+    flow_type: SessionType
+
+
+class QuestionAnswerPairSnapshot(BaseModel):
+    chat_session_id: UUID
+    # 1-indexed message number in the chat_session
+    # e.g. the first message pair in the chat_session is 1, the second is 2, etc.
+    message_pair_num: int
+    user_message: str
+    ai_response: str
+    retrieved_documents: list[AbridgedSearchDoc]
+    feedback_type: QAFeedbackType | None
+    feedback_text: str | None
+    persona_name: str | None
+    user_email: str
+    time_created: datetime
+    flow_type: SessionType
+
+    @classmethod
+    def from_chat_session_snapshot(
+        cls,
+        chat_session_snapshot: ChatSessionSnapshot,
+    ) -> list["QuestionAnswerPairSnapshot"]:
+        message_pairs: list[tuple[MessageSnapshot, MessageSnapshot]] = []
+        for ind in range(1, len(chat_session_snapshot.messages), 2):
+            message_pairs.append(
+                (
+                    chat_session_snapshot.messages[ind - 1],
+                    chat_session_snapshot.messages[ind],
+                )
+            )
+
+        return [
+            cls(
+                chat_session_id=chat_session_snapshot.id,
+                message_pair_num=ind + 1,
+                user_message=user_message.message,
+                ai_response=ai_message.message,
+                retrieved_documents=ai_message.documents,
+                feedback_type=ai_message.feedback_type,
+                feedback_text=ai_message.feedback_text,
+                persona_name=chat_session_snapshot.assistant_name,
+                user_email=get_display_email(chat_session_snapshot.user_email),
+                time_created=user_message.time_created,
+                flow_type=chat_session_snapshot.flow_type,
+            )
+            for ind, (user_message, ai_message) in enumerate(message_pairs)
+        ]
+
+    def to_json(self) -> dict[str, str | None]:
+        return {
+            "chat_session_id": str(self.chat_session_id),
+            "message_pair_num": str(self.message_pair_num),
+            "user_message": self.user_message,
+            "ai_response": self.ai_response,
+            "retrieved_documents": "|".join(
+                [
+                    doc.link or doc.semantic_identifier
+                    for doc in self.retrieved_documents
+                ]
+            ),
+            "feedback_type": self.feedback_type.value if self.feedback_type else "",
+            "feedback_text": self.feedback_text or "",
+            "persona_name": self.persona_name,
+            "user_email": self.user_email,
+            "time_created": str(self.time_created),
+            "flow_type": self.flow_type,
+        }

backend/ee/onyx/server/seeding.py

@@ -24,7 +24,7 @@ from onyx.db.llm import update_default_provider
 from onyx.db.llm import upsert_llm_provider
 from onyx.db.models import Tool
 from onyx.db.persona import upsert_persona
-from onyx.server.features.persona.models import CreatePersonaRequest
+from onyx.server.features.persona.models import PersonaUpsertRequest
 from onyx.server.manage.llm.models import LLMProviderUpsertRequest
 from onyx.server.settings.models import Settings
 from onyx.server.settings.store import store_settings as store_base_settings
@@ -57,7 +57,7 @@ class SeedConfiguration(BaseModel):
     llms: list[LLMProviderUpsertRequest] | None = None
     admin_user_emails: list[str] | None = None
     seeded_logo_path: str | None = None
-    personas: list[CreatePersonaRequest] | None = None
+    personas: list[PersonaUpsertRequest] | None = None
     settings: Settings | None = None
     enterprise_settings: EnterpriseSettings | None = None
 
@@ -128,7 +128,7 @@ def _seed_llms(
         )
 
 
-def _seed_personas(db_session: Session, personas: list[CreatePersonaRequest]) -> None:
+def _seed_personas(db_session: Session, personas: list[PersonaUpsertRequest]) -> None:
     if personas:
         logger.notice("Seeding Personas")
         for persona in personas:

backend/ee/onyx/server/tenants/api.py

@@ -34,6 +34,7 @@ from onyx.auth.users import get_redis_strategy
 from onyx.auth.users import optional_user
 from onyx.auth.users import User
 from onyx.configs.app_configs import WEB_DOMAIN
+from onyx.configs.constants import FASTAPI_USERS_AUTH_COOKIE_NAME
 from onyx.db.auth import get_user_count
 from onyx.db.engine import get_current_tenant_id
 from onyx.db.engine import get_session
@@ -111,6 +112,7 @@ async def login_as_anonymous_user(
     token = generate_anonymous_user_jwt_token(tenant_id)
 
     response = Response()
+    response.delete_cookie(FASTAPI_USERS_AUTH_COOKIE_NAME)
     response.set_cookie(
         key=ANONYMOUS_USER_COOKIE_NAME,
         value=token,

backend/ee/onyx/server/tenants/provisioning.py

@@ -24,6 +24,7 @@ from ee.onyx.server.tenants.user_mapping import get_tenant_id_for_email
 from ee.onyx.server.tenants.user_mapping import user_owns_a_tenant
 from onyx.auth.users import exceptions
 from onyx.configs.app_configs import CONTROL_PLANE_API_BASE_URL
+from onyx.configs.app_configs import DEV_MODE
 from onyx.configs.constants import MilestoneRecordType
 from onyx.db.engine import get_session_with_tenant
 from onyx.db.engine import get_sqlalchemy_engine
@@ -85,7 +86,8 @@ async def create_tenant(email: str, referral_source: str | None = None) -> str:
         # Provision tenant on data plane
         await provision_tenant(tenant_id, email)
         # Notify control plane
-        await notify_control_plane(tenant_id, email, referral_source)
+        if not DEV_MODE:
+            await notify_control_plane(tenant_id, email, referral_source)
     except Exception as e:
         logger.error(f"Tenant provisioning failed: {e}")
         await rollback_tenant_provisioning(tenant_id)

backend/ee/onyx/server/token_rate_limits/api.py

@@ -5,7 +5,7 @@ from fastapi import Depends
 from sqlalchemy.orm import Session
 
 from ee.onyx.db.token_limit import fetch_all_user_group_token_rate_limits_by_group
-from ee.onyx.db.token_limit import fetch_user_group_token_rate_limits
+from ee.onyx.db.token_limit import fetch_user_group_token_rate_limits_for_user
 from ee.onyx.db.token_limit import insert_user_group_token_rate_limit
 from onyx.auth.users import current_admin_user
 from onyx.auth.users import current_curator_or_admin_user
@@ -51,8 +51,10 @@ def get_group_token_limit_settings(
 ) -> list[TokenRateLimitDisplay]:
     return [
         TokenRateLimitDisplay.from_db(token_rate_limit)
-        for token_rate_limit in fetch_user_group_token_rate_limits(
-            db_session, group_id, user
+        for token_rate_limit in fetch_user_group_token_rate_limits_for_user(
+            db_session=db_session,
+            group_id=group_id,
+            user=user,
         )
     ]
 

backend/ee/onyx/server/user_group/models.py

@@ -58,6 +58,7 @@ class UserGroup(BaseModel):
                     credential=CredentialSnapshot.from_credential_db_model(
                         cc_pair_relationship.cc_pair.credential
                     ),
+                    access_type=cc_pair_relationship.cc_pair.access_type,
                 )
                 for cc_pair_relationship in user_group_model.cc_pair_relationships
                 if cc_pair_relationship.is_current

backend/model_server/constants.py

@@ -28,3 +28,9 @@ class EmbeddingModelTextType:
     @staticmethod
     def get_type(provider: EmbeddingProvider, text_type: EmbedTextType) -> str:
         return EmbeddingModelTextType.PROVIDER_TEXT_TYPE_MAP[provider][text_type]
+
+
+class GPUStatus:
+    CUDA = "cuda"
+    MAC_MPS = "mps"
+    NONE = "none"

backend/model_server/encoders.py

@@ -12,6 +12,7 @@ import voyageai  # type: ignore
 from cohere import AsyncClient as CohereAsyncClient
 from fastapi import APIRouter
 from fastapi import HTTPException
+from fastapi import Request
 from google.oauth2 import service_account  # type: ignore
 from litellm import aembedding
 from litellm.exceptions import RateLimitError
@@ -320,6 +321,7 @@ async def embed_text(
     prefix: str | None,
     api_url: str | None,
     api_version: str | None,
+    gpu_type: str = "UNKNOWN",
 ) -> list[Embedding]:
     if not all(texts):
         logger.error("Empty strings provided for embedding")
@@ -373,8 +375,11 @@ async def embed_text(
 
         elapsed = time.monotonic() - start
         logger.info(
-            f"Successfully embedded {len(texts)} texts with {total_chars} total characters "
-            f"with provider {provider_type} in {elapsed:.2f}"
+            f"event=embedding_provider "
+            f"texts={len(texts)} "
+            f"chars={total_chars} "
+            f"provider={provider_type} "
+            f"elapsed={elapsed:.2f}"
         )
     elif model_name is not None:
         logger.info(
@@ -403,6 +408,14 @@ async def embed_text(
             f"Successfully embedded {len(texts)} texts with {total_chars} total characters "
             f"with local model {model_name} in {elapsed:.2f}"
         )
+        logger.info(
+            f"event=embedding_model "
+            f"texts={len(texts)} "
+            f"chars={total_chars} "
+            f"model={model_name} "
+            f"gpu={gpu_type} "
+            f"elapsed={elapsed:.2f}"
+        )
     else:
         logger.error("Neither model name nor provider specified for embedding")
         raise ValueError(
@@ -455,8 +468,15 @@ async def litellm_rerank(
 
 
 @router.post("/bi-encoder-embed")
-async def process_embed_request(
+async def route_bi_encoder_embed(
+    request: Request,
     embed_request: EmbedRequest,
+) -> EmbedResponse:
+    return await process_embed_request(embed_request, request.app.state.gpu_type)
+
+
+async def process_embed_request(
+    embed_request: EmbedRequest, gpu_type: str = "UNKNOWN"
 ) -> EmbedResponse:
     if not embed_request.texts:
         raise HTTPException(status_code=400, detail="No texts to be embedded")
@@ -484,6 +504,7 @@ async def process_embed_request(
             api_url=embed_request.api_url,
             api_version=embed_request.api_version,
             prefix=prefix,
+            gpu_type=gpu_type,
         )
         return EmbedResponse(embeddings=embeddings)
     except RateLimitError as e:

backend/model_server/main.py

@@ -16,6 +16,7 @@ from model_server.custom_models import router as custom_models_router
 from model_server.custom_models import warm_up_intent_model
 from model_server.encoders import router as encoders_router
 from model_server.management_endpoints import router as management_router
+from model_server.utils import get_gpu_type
 from onyx import __version__
 from onyx.utils.logger import setup_logger
 from shared_configs.configs import INDEXING_ONLY
@@ -58,12 +59,10 @@ def _move_files_recursively(source: Path, dest: Path, overwrite: bool = False) -
 
 @asynccontextmanager
 async def lifespan(app: FastAPI) -> AsyncGenerator:
-    if torch.cuda.is_available():
-        logger.notice("CUDA GPU is available")
-    elif torch.backends.mps.is_available():
-        logger.notice("Mac MPS is available")
-    else:
-        logger.notice("GPU is not available, using CPU")
+    gpu_type = get_gpu_type()
+    logger.notice(f"Torch GPU Detection: gpu_type={gpu_type}")
+
+    app.state.gpu_type = gpu_type
 
     if TEMP_HF_CACHE_PATH.is_dir():
         logger.notice("Moving contents of temp_huggingface to huggingface cache.")

backend/model_server/management_endpoints.py

@@ -1,7 +1,9 @@
-import torch
 from fastapi import APIRouter
 from fastapi import Response
 
+from model_server.constants import GPUStatus
+from model_server.utils import get_gpu_type
+
 router = APIRouter(prefix="/api")
 
 
@@ -11,10 +13,7 @@ async def healthcheck() -> Response:
 
 
 @router.get("/gpu-status")
-async def gpu_status() -> dict[str, bool | str]:
-    if torch.cuda.is_available():
-        return {"gpu_available": True, "type": "cuda"}
-    elif torch.backends.mps.is_available():
-        return {"gpu_available": True, "type": "mps"}
-    else:
-        return {"gpu_available": False, "type": "none"}
+async def route_gpu_status() -> dict[str, bool | str]:
+    gpu_type = get_gpu_type()
+    gpu_available = gpu_type != GPUStatus.NONE
+    return {"gpu_available": gpu_available, "type": gpu_type}

backend/model_server/utils.py

@@ -8,6 +8,9 @@ from typing import Any
 from typing import cast
 from typing import TypeVar
 
+import torch
+
+from model_server.constants import GPUStatus
 from onyx.utils.logger import setup_logger
 
 logger = setup_logger()
@@ -58,3 +61,12 @@ def simple_log_function_time(
             return cast(F, wrapped_sync_func)
 
     return decorator
+
+
+def get_gpu_type() -> str:
+    if torch.cuda.is_available():
+        return GPUStatus.CUDA
+    if torch.backends.mps.is_available():
+        return GPUStatus.MAC_MPS
+
+    return GPUStatus.NONE

backend/onyx/access/utils.py

@@ -19,6 +19,9 @@ def prefix_external_group(ext_group_name: str) -> str:
     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}"
+def build_ext_group_name_for_onyx(ext_group_name: str, source: DocumentSource) -> str:
+    """
+    External groups may collide across sources, every source needs its own prefix.
+    NOTE: the name is lowercased to handle case sensitivity for group names
+    """
+    return f"{source.value}_{ext_group_name}".lower()

backend/onyx/agent_search/basic/graph_builder.py

@@ -1,103 +0,0 @@
-from typing import cast
-
-from langchain_core.callbacks.manager import dispatch_custom_event
-from langchain_core.runnables.config import RunnableConfig
-from langgraph.graph import END
-from langgraph.graph import START
-from langgraph.graph import StateGraph
-
-from onyx.agent_search.basic.states import BasicInput
-from onyx.agent_search.basic.states import BasicOutput
-from onyx.agent_search.basic.states import BasicState
-from onyx.agent_search.basic.states import BasicStateUpdate
-from onyx.agent_search.models import ProSearchConfig
-from onyx.chat.stream_processing.utils import (
-    map_document_id_order,
-)
-from onyx.tools.tool_implementations.search.search_tool import SearchTool
-
-
-def basic_graph_builder() -> StateGraph:
-    graph = StateGraph(
-        state_schema=BasicState,
-        input=BasicInput,
-        output=BasicOutput,
-    )
-
-    ### Add nodes ###
-
-    graph.add_node(
-        node="get_response",
-        action=get_response,
-    )
-
-    ### Add edges ###
-
-    graph.add_edge(start_key=START, end_key="get_response")
-
-    graph.add_conditional_edges("get_response", should_continue, ["get_response", END])
-    graph.add_edge(
-        start_key="get_response",
-        end_key=END,
-    )
-
-    return graph
-
-
-def should_continue(state: BasicState) -> str:
-    return (
-        END if state["last_llm_call"] is None or state["calls"] > 1 else "get_response"
-    )
-
-
-def get_response(state: BasicState, config: RunnableConfig) -> BasicStateUpdate:
-    pro_search_config = cast(ProSearchConfig, config["metadata"]["config"])
-    llm = pro_search_config.primary_llm
-    current_llm_call = state["last_llm_call"]
-    if current_llm_call is None:
-        raise ValueError("last_llm_call is None")
-    structured_response_format = pro_search_config.structured_response_format
-    response_handler_manager = state["response_handler_manager"]
-    # DEBUG: good breakpoint
-    stream = llm.stream(
-        # For tool calling LLMs, we want to insert the task prompt as part of this flow, this is because the LLM
-        # may choose to not call any tools and just generate the answer, in which case the task prompt is needed.
-        prompt=current_llm_call.prompt_builder.build(),
-        tools=[tool.tool_definition() for tool in current_llm_call.tools] or None,
-        tool_choice=(
-            "required"
-            if current_llm_call.tools and current_llm_call.force_use_tool.force_use
-            else None
-        ),
-        structured_response_format=structured_response_format,
-    )
-
-    for response in response_handler_manager.handle_llm_response(stream):
-        dispatch_custom_event(
-            "basic_response",
-            response,
-        )
-
-    next_call = response_handler_manager.next_llm_call(current_llm_call)
-    if next_call is not None:
-        final_search_results, displayed_search_results = SearchTool.get_search_result(
-            next_call
-        ) or ([], [])
-    else:
-        final_search_results, displayed_search_results = [], []
-
-    response_handler_manager.answer_handler.update(
-        (
-            final_search_results,
-            map_document_id_order(final_search_results),
-            map_document_id_order(displayed_search_results),
-        )
-    )
-    return BasicStateUpdate(
-        last_llm_call=next_call,
-        calls=state["calls"] + 1,
-    )
-
-
-if __name__ == "__main__":
-    pass

backend/onyx/agent_search/basic/states.py

@@ -1,38 +0,0 @@
-from typing import TypedDict
-
-from onyx.chat.llm_response_handler import LLMResponseHandlerManager
-from onyx.chat.prompt_builder.build import LLMCall
-
-## Update States
-
-
-## Graph Input State
-
-
-class BasicInput(TypedDict):
-    base_question: str
-    last_llm_call: LLMCall | None
-    response_handler_manager: LLMResponseHandlerManager
-    calls: int
-
-
-## Graph Output State
-
-
-class BasicOutput(TypedDict):
-    pass
-
-
-class BasicStateUpdate(TypedDict):
-    last_llm_call: LLMCall | None
-    calls: int
-
-
-## Graph State
-
-
-class BasicState(
-    BasicInput,
-    BasicOutput,
-):
-    pass

backend/onyx/agent_search/db_operations.py

@@ -1,66 +0,0 @@
-from uuid import UUID
-
-from sqlalchemy.orm import Session
-
-from onyx.db.models import AgentSubQuery
-from onyx.db.models import AgentSubQuestion
-
-
-def create_sub_question(
-    db_session: Session,
-    chat_session_id: UUID,
-    primary_message_id: int,
-    sub_question: str,
-    sub_answer: str,
-) -> AgentSubQuestion:
-    """Create a new sub-question record in the database."""
-    sub_q = AgentSubQuestion(
-        chat_session_id=chat_session_id,
-        primary_question_id=primary_message_id,
-        sub_question=sub_question,
-        sub_answer=sub_answer,
-    )
-    db_session.add(sub_q)
-    db_session.flush()
-    return sub_q
-
-
-def create_sub_query(
-    db_session: Session,
-    chat_session_id: UUID,
-    parent_question_id: int,
-    sub_query: str,
-) -> AgentSubQuery:
-    """Create a new sub-query record in the database."""
-    sub_q = AgentSubQuery(
-        chat_session_id=chat_session_id,
-        parent_question_id=parent_question_id,
-        sub_query=sub_query,
-    )
-    db_session.add(sub_q)
-    db_session.flush()
-    return sub_q
-
-
-def get_sub_questions_for_message(
-    db_session: Session,
-    primary_message_id: int,
-) -> list[AgentSubQuestion]:
-    """Get all sub-questions for a given primary message."""
-    return (
-        db_session.query(AgentSubQuestion)
-        .filter(AgentSubQuestion.primary_question_id == primary_message_id)
-        .all()
-    )
-
-
-def get_sub_queries_for_question(
-    db_session: Session,
-    sub_question_id: int,
-) -> list[AgentSubQuery]:
-    """Get all sub-queries for a given sub-question."""
-    return (
-        db_session.query(AgentSubQuery)
-        .filter(AgentSubQuery.parent_question_id == sub_question_id)
-        .all()
-    )

backend/onyx/agent_search/models.py

@@ -1,57 +0,0 @@
-from dataclasses import dataclass
-from uuid import UUID
-
-from pydantic import BaseModel
-from sqlalchemy.orm import Session
-
-from onyx.context.search.models import SearchRequest
-from onyx.llm.interfaces import LLM
-from onyx.llm.models import PreviousMessage
-from onyx.tools.tool_implementations.search.search_tool import SearchTool
-
-
-@dataclass
-class ProSearchConfig:
-    """
-    Configuration for the Pro Search feature.
-    """
-
-    # The search request that was used to generate the Pro Search
-    search_request: SearchRequest
-
-    primary_llm: LLM
-    fast_llm: LLM
-    search_tool: SearchTool
-    use_agentic_search: bool = False
-
-    # For persisting agent search data
-    chat_session_id: UUID | None = None
-
-    # The message ID of the user message that triggered the Pro Search
-    message_id: int | None = None
-
-    # Whether to persistence data for the Pro Search (turned off for testing)
-    use_persistence: bool = True
-
-    # The database session for the Pro Search
-    db_session: Session | None = None
-
-    # Whether to perform initial search to inform decomposition
-    perform_initial_search_path_decision: bool = False
-
-    # Whether to perform initial search to inform decomposition
-    perform_initial_search_decomposition: bool = False
-
-    # Whether to allow creation of refinement questions (and entity extraction, etc.)
-    allow_refinement: bool = False
-
-    # Message history for the current chat session
-    message_history: list[PreviousMessage] | None = None
-
-    structured_response_format: dict | None = None
-
-
-class AgentDocumentCitations(BaseModel):
-    document_id: str
-    document_title: str
-    link: str

backend/onyx/agent_search/pro_search_a/answer_initial_sub_question/edges.py

@@ -1,26 +0,0 @@
-from collections.abc import Hashable
-
-from langgraph.types import Send
-
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionInput,
-)
-from onyx.agent_search.pro_search_a.expanded_retrieval.states import (
-    ExpandedRetrievalInput,
-)
-from onyx.utils.logger import setup_logger
-
-logger = setup_logger()
-
-
-def send_to_expanded_retrieval(state: AnswerQuestionInput) -> Send | Hashable:
-    logger.debug("sending to expanded retrieval via edge")
-
-    return Send(
-        "initial_sub_question_expanded_retrieval",
-        ExpandedRetrievalInput(
-            question=state["question"],
-            base_search=False,
-            sub_question_id=state["question_id"],
-        ),
-    )

backend/onyx/agent_search/pro_search_a/answer_initial_sub_question/graph_builder.py

@@ -1,125 +0,0 @@
-from langgraph.graph import END
-from langgraph.graph import START
-from langgraph.graph import StateGraph
-
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.edges import (
-    send_to_expanded_retrieval,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.nodes.answer_check import (
-    answer_check,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.nodes.answer_generation import (
-    answer_generation,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.nodes.format_answer import (
-    format_answer,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.nodes.ingest_retrieval import (
-    ingest_retrieval,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionInput,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionOutput,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionState,
-)
-from onyx.agent_search.pro_search_a.expanded_retrieval.graph_builder import (
-    expanded_retrieval_graph_builder,
-)
-from onyx.agent_search.shared_graph_utils.utils import get_test_config
-from onyx.utils.logger import setup_logger
-
-logger = setup_logger()
-
-
-def answer_query_graph_builder() -> StateGraph:
-    graph = StateGraph(
-        state_schema=AnswerQuestionState,
-        input=AnswerQuestionInput,
-        output=AnswerQuestionOutput,
-    )
-
-    ### Add nodes ###
-
-    expanded_retrieval = expanded_retrieval_graph_builder().compile()
-    graph.add_node(
-        node="initial_sub_question_expanded_retrieval",
-        action=expanded_retrieval,
-    )
-    graph.add_node(
-        node="answer_check",
-        action=answer_check,
-    )
-    graph.add_node(
-        node="answer_generation",
-        action=answer_generation,
-    )
-    graph.add_node(
-        node="format_answer",
-        action=format_answer,
-    )
-    graph.add_node(
-        node="ingest_retrieval",
-        action=ingest_retrieval,
-    )
-
-    ### Add edges ###
-
-    graph.add_conditional_edges(
-        source=START,
-        path=send_to_expanded_retrieval,
-        path_map=["initial_sub_question_expanded_retrieval"],
-    )
-    graph.add_edge(
-        start_key="initial_sub_question_expanded_retrieval",
-        end_key="ingest_retrieval",
-    )
-    graph.add_edge(
-        start_key="ingest_retrieval",
-        end_key="answer_generation",
-    )
-    graph.add_edge(
-        start_key="answer_generation",
-        end_key="answer_check",
-    )
-    graph.add_edge(
-        start_key="answer_check",
-        end_key="format_answer",
-    )
-    graph.add_edge(
-        start_key="format_answer",
-        end_key=END,
-    )
-
-    return graph
-
-
-if __name__ == "__main__":
-    from onyx.db.engine import get_session_context_manager
-    from onyx.llm.factory import get_default_llms
-    from onyx.context.search.models import SearchRequest
-
-    graph = answer_query_graph_builder()
-    compiled_graph = graph.compile()
-    primary_llm, fast_llm = get_default_llms()
-    search_request = SearchRequest(
-        query="what can you do with onyx or danswer?",
-    )
-    with get_session_context_manager() as db_session:
-        pro_search_config, search_tool = get_test_config(
-            db_session, primary_llm, fast_llm, search_request
-        )
-        inputs = AnswerQuestionInput(
-            question="what can you do with onyx?",
-            question_id="0_0",
-        )
-        for thing in compiled_graph.stream(
-            input=inputs,
-            config={"configurable": {"config": pro_search_config}},
-            # debug=True,
-            # subgraphs=True,
-        ):
-            logger.debug(thing)

backend/onyx/agent_search/pro_search_a/answer_initial_sub_question/models.py

@@ -1,8 +0,0 @@
-from pydantic import BaseModel
-
-
-### Models ###
-
-
-class AnswerRetrievalStats(BaseModel):
-    answer_retrieval_stats: dict[str, float | int]

backend/onyx/agent_search/pro_search_a/answer_initial_sub_question/nodes/answer_check.py

@@ -1,45 +0,0 @@
-from typing import cast
-
-from langchain_core.messages import HumanMessage
-from langchain_core.messages import merge_message_runs
-from langchain_core.runnables.config import RunnableConfig
-
-from onyx.agent_search.models import ProSearchConfig
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionState,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    QACheckUpdate,
-)
-from onyx.agent_search.shared_graph_utils.prompts import SUB_CHECK_NO
-from onyx.agent_search.shared_graph_utils.prompts import SUB_CHECK_PROMPT
-from onyx.agent_search.shared_graph_utils.prompts import UNKNOWN_ANSWER
-
-
-def answer_check(state: AnswerQuestionState, config: RunnableConfig) -> QACheckUpdate:
-    if state["answer"] == UNKNOWN_ANSWER:
-        return QACheckUpdate(
-            answer_quality=SUB_CHECK_NO,
-        )
-    msg = [
-        HumanMessage(
-            content=SUB_CHECK_PROMPT.format(
-                question=state["question"],
-                base_answer=state["answer"],
-            )
-        )
-    ]
-
-    pro_search_config = cast(ProSearchConfig, config["metadata"]["config"])
-    fast_llm = pro_search_config.fast_llm
-    response = list(
-        fast_llm.stream(
-            prompt=msg,
-        )
-    )
-
-    quality_str = merge_message_runs(response, chunk_separator="")[0].content
-
-    return QACheckUpdate(
-        answer_quality=quality_str,
-    )

backend/onyx/agent_search/pro_search_a/answer_initial_sub_question/nodes/answer_generation.py

@@ -1,106 +0,0 @@
-import datetime
-from typing import Any
-from typing import cast
-
-from langchain_core.callbacks.manager import dispatch_custom_event
-from langchain_core.messages import merge_message_runs
-from langchain_core.runnables.config import RunnableConfig
-
-from onyx.agent_search.models import ProSearchConfig
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionState,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    QAGenerationUpdate,
-)
-from onyx.agent_search.shared_graph_utils.agent_prompt_ops import (
-    build_sub_question_answer_prompt,
-)
-from onyx.agent_search.shared_graph_utils.prompts import ASSISTANT_SYSTEM_PROMPT_DEFAULT
-from onyx.agent_search.shared_graph_utils.prompts import ASSISTANT_SYSTEM_PROMPT_PERSONA
-from onyx.agent_search.shared_graph_utils.prompts import UNKNOWN_ANSWER
-from onyx.agent_search.shared_graph_utils.utils import get_persona_prompt
-from onyx.agent_search.shared_graph_utils.utils import parse_question_id
-from onyx.chat.models import AgentAnswerPiece
-from onyx.chat.models import StreamStopInfo
-from onyx.chat.models import StreamStopReason
-from onyx.utils.logger import setup_logger
-
-logger = setup_logger()
-
-
-def answer_generation(
-    state: AnswerQuestionState, config: RunnableConfig
-) -> QAGenerationUpdate:
-    now_start = datetime.datetime.now()
-    logger.debug(f"--------{now_start}--------START ANSWER GENERATION---")
-
-    pro_search_config = cast(ProSearchConfig, config["metadata"]["config"])
-    question = state["question"]
-    docs = state["documents"]
-    level, question_nr = parse_question_id(state["question_id"])
-    persona_prompt = get_persona_prompt(pro_search_config.search_request.persona)
-
-    if len(docs) == 0:
-        answer_str = UNKNOWN_ANSWER
-        dispatch_custom_event(
-            "sub_answers",
-            AgentAnswerPiece(
-                answer_piece=answer_str,
-                level=level,
-                level_question_nr=question_nr,
-                answer_type="agent_sub_answer",
-            ),
-        )
-    else:
-        if len(persona_prompt) > 0:
-            persona_specification = ASSISTANT_SYSTEM_PROMPT_DEFAULT
-        else:
-            persona_specification = ASSISTANT_SYSTEM_PROMPT_PERSONA.format(
-                persona_prompt=persona_prompt
-            )
-
-        logger.debug(f"Number of verified retrieval docs: {len(docs)}")
-
-        fast_llm = pro_search_config.fast_llm
-        msg = build_sub_question_answer_prompt(
-            question=question,
-            original_question=pro_search_config.search_request.query,
-            docs=docs,
-            persona_specification=persona_specification,
-            config=fast_llm.config,
-        )
-
-        response: list[str | list[str | dict[str, Any]]] = []
-        for message in fast_llm.stream(
-            prompt=msg,
-        ):
-            # TODO: in principle, the answer here COULD contain images, but we don't support that yet
-            content = message.content
-            if not isinstance(content, str):
-                raise ValueError(
-                    f"Expected content to be a string, but got {type(content)}"
-                )
-            dispatch_custom_event(
-                "sub_answers",
-                AgentAnswerPiece(
-                    answer_piece=content,
-                    level=level,
-                    level_question_nr=question_nr,
-                    answer_type="agent_sub_answer",
-                ),
-            )
-            response.append(content)
-
-        answer_str = merge_message_runs(response, chunk_separator="")[0].content
-
-    stop_event = StreamStopInfo(
-        stop_reason=StreamStopReason.FINISHED,
-        level=level,
-        level_question_nr=question_nr,
-    )
-    dispatch_custom_event("sub_answer_finished", stop_event)
-
-    return QAGenerationUpdate(
-        answer=answer_str,
-    )

backend/onyx/agent_search/pro_search_a/answer_initial_sub_question/nodes/format_answer.py

@@ -1,25 +0,0 @@
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionOutput,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionState,
-)
-from onyx.agent_search.shared_graph_utils.models import (
-    QuestionAnswerResults,
-)
-
-
-def format_answer(state: AnswerQuestionState) -> AnswerQuestionOutput:
-    return AnswerQuestionOutput(
-        answer_results=[
-            QuestionAnswerResults(
-                question=state["question"],
-                question_id=state["question_id"],
-                quality=state.get("answer_quality", "No"),
-                answer=state["answer"],
-                expanded_retrieval_results=state["expanded_retrieval_results"],
-                documents=state["documents"],
-                sub_question_retrieval_stats=state["sub_question_retrieval_stats"],
-            )
-        ],
-    )

backend/onyx/agent_search/pro_search_a/answer_initial_sub_question/nodes/ingest_retrieval.py

@@ -1,23 +0,0 @@
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    RetrievalIngestionUpdate,
-)
-from onyx.agent_search.pro_search_a.expanded_retrieval.states import (
-    ExpandedRetrievalOutput,
-)
-from onyx.agent_search.shared_graph_utils.models import AgentChunkStats
-
-
-def ingest_retrieval(state: ExpandedRetrievalOutput) -> RetrievalIngestionUpdate:
-    sub_question_retrieval_stats = state[
-        "expanded_retrieval_result"
-    ].sub_question_retrieval_stats
-    if sub_question_retrieval_stats is None:
-        sub_question_retrieval_stats = [AgentChunkStats()]
-
-    return RetrievalIngestionUpdate(
-        expanded_retrieval_results=state[
-            "expanded_retrieval_result"
-        ].expanded_queries_results,
-        documents=state["expanded_retrieval_result"].all_documents,
-        sub_question_retrieval_stats=sub_question_retrieval_stats,
-    )

backend/onyx/agent_search/pro_search_a/answer_initial_sub_question/states.py

@@ -1,63 +0,0 @@
-from operator import add
-from typing import Annotated
-from typing import TypedDict
-
-from onyx.agent_search.core_state import SubgraphCoreState
-from onyx.agent_search.shared_graph_utils.models import AgentChunkStats
-from onyx.agent_search.shared_graph_utils.models import QueryResult
-from onyx.agent_search.shared_graph_utils.models import (
-    QuestionAnswerResults,
-)
-from onyx.agent_search.shared_graph_utils.operators import dedup_inference_sections
-from onyx.context.search.models import InferenceSection
-
-
-## Update States
-class QACheckUpdate(TypedDict):
-    answer_quality: str
-
-
-class QAGenerationUpdate(TypedDict):
-    answer: str
-    # answer_stat: AnswerStats
-
-
-class RetrievalIngestionUpdate(TypedDict):
-    expanded_retrieval_results: list[QueryResult]
-    documents: Annotated[list[InferenceSection], dedup_inference_sections]
-    sub_question_retrieval_stats: AgentChunkStats
-
-
-## Graph Input State
-
-
-class AnswerQuestionInput(SubgraphCoreState):
-    question: str
-    question_id: str  # 0_0 is original question, everything else is <level>_<question_num>.
-    # level 0 is original question and first decomposition, level 1 is follow up, etc
-    # question_num is a unique number per original question per level.
-
-
-## Graph State
-
-
-class AnswerQuestionState(
-    AnswerQuestionInput,
-    QAGenerationUpdate,
-    QACheckUpdate,
-    RetrievalIngestionUpdate,
-):
-    pass
-
-
-## Graph Output State
-
-
-class AnswerQuestionOutput(TypedDict):
-    """
-    This is a list of results even though each call of this subgraph only returns one result.
-    This is because if we parallelize the answer query subgraph, there will be multiple
-      results in a list so the add operator is used to add them together.
-    """
-
-    answer_results: Annotated[list[QuestionAnswerResults], add]

backend/onyx/agent_search/pro_search_a/answer_refinement_sub_question/edges.py

@@ -1,26 +0,0 @@
-from collections.abc import Hashable
-
-from langgraph.types import Send
-
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionInput,
-)
-from onyx.agent_search.pro_search_a.expanded_retrieval.states import (
-    ExpandedRetrievalInput,
-)
-from onyx.utils.logger import setup_logger
-
-logger = setup_logger()
-
-
-def send_to_expanded_refined_retrieval(state: AnswerQuestionInput) -> Send | Hashable:
-    logger.debug("sending to expanded retrieval for follow up question via edge")
-
-    return Send(
-        "refined_sub_question_expanded_retrieval",
-        ExpandedRetrievalInput(
-            question=state["question"],
-            sub_question_id=state["question_id"],
-            base_search=False,
-        ),
-    )

backend/onyx/agent_search/pro_search_a/answer_refinement_sub_question/graph_builder.py

@@ -1,122 +0,0 @@
-from langgraph.graph import END
-from langgraph.graph import START
-from langgraph.graph import StateGraph
-
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.nodes.answer_check import (
-    answer_check,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.nodes.answer_generation import (
-    answer_generation,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.nodes.format_answer import (
-    format_answer,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.nodes.ingest_retrieval import (
-    ingest_retrieval,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionInput,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionOutput,
-)
-from onyx.agent_search.pro_search_a.answer_initial_sub_question.states import (
-    AnswerQuestionState,
-)
-from onyx.agent_search.pro_search_a.answer_refinement_sub_question.edges import (
-    send_to_expanded_refined_retrieval,
-)
-from onyx.agent_search.pro_search_a.expanded_retrieval.graph_builder import (
-    expanded_retrieval_graph_builder,
-)
-from onyx.utils.logger import setup_logger
-
-logger = setup_logger()
-
-
-def answer_refined_query_graph_builder() -> StateGraph:
-    graph = StateGraph(
-        state_schema=AnswerQuestionState,
-        input=AnswerQuestionInput,
-        output=AnswerQuestionOutput,
-    )
-
-    ### Add nodes ###
-
-    expanded_retrieval = expanded_retrieval_graph_builder().compile()
-    graph.add_node(
-        node="refined_sub_question_expanded_retrieval",
-        action=expanded_retrieval,
-    )
-    graph.add_node(
-        node="refined_sub_answer_check",
-        action=answer_check,
-    )
-    graph.add_node(
-        node="refined_sub_answer_generation",
-        action=answer_generation,
-    )
-    graph.add_node(
-        node="format_refined_sub_answer",
-        action=format_answer,
-    )
-    graph.add_node(
-        node="ingest_refined_retrieval",
-        action=ingest_retrieval,
-    )
-
-    ### Add edges ###
-
-    graph.add_conditional_edges(
-        source=START,
-        path=send_to_expanded_refined_retrieval,
-        path_map=["refined_sub_question_expanded_retrieval"],
-    )
-    graph.add_edge(
-        start_key="refined_sub_question_expanded_retrieval",
-        end_key="ingest_refined_retrieval",
-    )
-    graph.add_edge(
-        start_key="ingest_refined_retrieval",
-        end_key="refined_sub_answer_generation",
-    )
-    graph.add_edge(
-        start_key="refined_sub_answer_generation",
-        end_key="refined_sub_answer_check",
-    )
-    graph.add_edge(
-        start_key="refined_sub_answer_check",
-        end_key="format_refined_sub_answer",
-    )
-    graph.add_edge(
-        start_key="format_refined_sub_answer",
-        end_key=END,
-    )
-
-    return graph
-
-
-if __name__ == "__main__":
-    from onyx.db.engine import get_session_context_manager
-    from onyx.llm.factory import get_default_llms
-    from onyx.context.search.models import SearchRequest
-
-    graph = answer_refined_query_graph_builder()
-    compiled_graph = graph.compile()
-    primary_llm, fast_llm = get_default_llms()
-    search_request = SearchRequest(
-        query="what can you do with onyx or danswer?",
-    )
-    with get_session_context_manager() as db_session:
-        inputs = AnswerQuestionInput(
-            question="what can you do with onyx?",
-            question_id="0_0",
-        )
-        for thing in compiled_graph.stream(
-            input=inputs,
-            # debug=True,
-            # subgraphs=True,
-        ):
-            logger.debug(thing)
-        # output = compiled_graph.invoke(inputs)
-        #  logger.debug(output)

backend/onyx/agent_search/pro_search_a/answer_refinement_sub_question/models.py

@@ -1,19 +0,0 @@
-from pydantic import BaseModel
-
-from onyx.agent_search.shared_graph_utils.models import AgentChunkStats
-from onyx.context.search.models import InferenceSection
-
-### Models ###
-
-
-class AnswerRetrievalStats(BaseModel):
-    answer_retrieval_stats: dict[str, float | int]
-
-
-class QuestionAnswerResults(BaseModel):
-    question: str
-    answer: str
-    quality: str
-    # expanded_retrieval_results: list[QueryResult]
-    documents: list[InferenceSection]
-    sub_question_retrieval_stats: AgentChunkStats

backend/onyx/agent_search/pro_search_a/base_raw_search/graph_builder.py

@@ -1,70 +0,0 @@
-from langgraph.graph import END
-from langgraph.graph import START
-from langgraph.graph import StateGraph
-
-from onyx.agent_search.pro_search_a.base_raw_search.nodes.format_raw_search_results import (
-    format_raw_search_results,
-)
-from onyx.agent_search.pro_search_a.base_raw_search.nodes.generate_raw_search_data import (
-    generate_raw_search_data,
-)
-from onyx.agent_search.pro_search_a.base_raw_search.states import BaseRawSearchInput
-from onyx.agent_search.pro_search_a.base_raw_search.states import BaseRawSearchOutput
-from onyx.agent_search.pro_search_a.base_raw_search.states import BaseRawSearchState
-from onyx.agent_search.pro_search_a.expanded_retrieval.graph_builder import (
-    expanded_retrieval_graph_builder,
-)
-
-
-def base_raw_search_graph_builder() -> StateGraph:
-    graph = StateGraph(
-        state_schema=BaseRawSearchState,
-        input=BaseRawSearchInput,
-        output=BaseRawSearchOutput,
-    )
-
-    ### Add nodes ###
-
-    graph.add_node(
-        node="generate_raw_search_data",
-        action=generate_raw_search_data,
-    )
-
-    expanded_retrieval = expanded_retrieval_graph_builder().compile()
-    graph.add_node(
-        node="expanded_retrieval_base_search",
-        action=expanded_retrieval,
-    )
-    graph.add_node(
-        node="format_raw_search_results",
-        action=format_raw_search_results,
-    )
-
-    ### Add edges ###
-
-    graph.add_edge(start_key=START, end_key="generate_raw_search_data")
-
-    graph.add_edge(
-        start_key="generate_raw_search_data",
-        end_key="expanded_retrieval_base_search",
-    )
-    graph.add_edge(
-        start_key="expanded_retrieval_base_search",
-        end_key="format_raw_search_results",
-    )
-
-    # graph.add_edge(
-    #     start_key="expanded_retrieval_base_search",
-    #     end_key=END,
-    # )
-
-    graph.add_edge(
-        start_key="format_raw_search_results",
-        end_key=END,
-    )
-
-    return graph
-
-
-if __name__ == "__main__":
-    pass

backend/onyx/agent_search/pro_search_a/base_raw_search/models.py

@@ -1,20 +0,0 @@
-from pydantic import BaseModel
-
-from onyx.agent_search.shared_graph_utils.models import AgentChunkStats
-from onyx.agent_search.shared_graph_utils.models import QueryResult
-from onyx.context.search.models import InferenceSection
-
-### Models ###
-
-
-class AnswerRetrievalStats(BaseModel):
-    answer_retrieval_stats: dict[str, float | int]
-
-
-class QuestionAnswerResults(BaseModel):
-    question: str
-    answer: str
-    quality: str
-    expanded_retrieval_results: list[QueryResult]
-    documents: list[InferenceSection]
-    sub_question_retrieval_stats: list[AgentChunkStats]

backend/onyx/agent_search/pro_search_a/base_raw_search/nodes/format_raw_search_results.py

@@ -1,16 +0,0 @@
-from onyx.agent_search.pro_search_a.base_raw_search.states import BaseRawSearchOutput
-from onyx.agent_search.pro_search_a.expanded_retrieval.states import (
-    ExpandedRetrievalOutput,
-)
-from onyx.utils.logger import setup_logger
-
-logger = setup_logger()
-
-
-def format_raw_search_results(state: ExpandedRetrievalOutput) -> BaseRawSearchOutput:
-    logger.debug("format_raw_search_results")
-    return BaseRawSearchOutput(
-        base_expanded_retrieval_result=state["expanded_retrieval_result"],
-        # base_retrieval_results=[state["expanded_retrieval_result"]],
-        # base_search_documents=[],
-    )

backend/onyx/agent_search/pro_search_a/base_raw_search/nodes/generate_raw_search_data.py

@@ -1,24 +0,0 @@
-from typing import cast
-
-from langchain_core.runnables.config import RunnableConfig
-
-from onyx.agent_search.core_state import CoreState
-from onyx.agent_search.models import ProSearchConfig
-from onyx.agent_search.pro_search_a.expanded_retrieval.states import (
-    ExpandedRetrievalInput,
-)
-from onyx.utils.logger import setup_logger
-
-logger = setup_logger()
-
-
-def generate_raw_search_data(
-    state: CoreState, config: RunnableConfig
-) -> ExpandedRetrievalInput:
-    logger.debug("generate_raw_search_data")
-    pro_search_config = cast(ProSearchConfig, config["metadata"]["config"])
-    return ExpandedRetrievalInput(
-        question=pro_search_config.search_request.query,
-        base_search=True,
-        sub_question_id=None,  # This graph is always and only used for the original question
-    )