fix starter message editing

* trying out a fix

* add ability to manually run model tests

---------

Co-authored-by: Richard Kuo (Danswer) <rkuo@onyx.app>

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

@@ -145,7 +145,7 @@ jobs:
         run: |
           cd deployment/docker_compose
           docker compose -f docker-compose.multitenant-dev.yml -p onyx-stack down -v
-      
+
       # NOTE: Use pre-ping/null pool to reduce flakiness due to dropped connections
       - name: Start Docker containers
         run: |
@@ -157,6 +157,7 @@ jobs:
           REQUIRE_EMAIL_VERIFICATION=false \
           DISABLE_TELEMETRY=true \
           IMAGE_TAG=test \
+          INTEGRATION_TESTS_MODE=true \
           docker compose -f docker-compose.dev.yml -p onyx-stack up -d
         id: start_docker
 
@@ -199,7 +200,7 @@ jobs:
           cd backend/tests/integration/mock_services
           docker compose -f docker-compose.mock-it-services.yml \
             -p mock-it-services-stack up -d
-      
+
       # NOTE: Use pre-ping/null to reduce flakiness due to dropped connections
       - name: Run Standard Integration Tests
         run: |

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

@@ -1,10 +1,16 @@
-name: Connector Tests
+name: Model Server Tests
 
 on:
   schedule:
     # This cron expression runs the job daily at 16:00 UTC (9am PT)
     - cron: "0 16 * * *"
-
+  workflow_dispatch:
+    inputs:
+      branch:
+        description: 'Branch to run the workflow on'
+        required: false
+        default: 'main'
+        
 env:
   # Bedrock
   AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
@@ -26,6 +32,23 @@ jobs:
       - name: Checkout code
         uses: actions/checkout@v4
 
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKER_USERNAME }}
+          password: ${{ secrets.DOCKER_TOKEN }}
+
+      # tag every docker image with "test" so that we can spin up the correct set
+      # of images during testing
+
+      # We don't need to build the Web Docker image since it's not yet used
+      # in the integration tests. We have a separate action to verify that it builds
+      # successfully.
+      - name: Pull Model Server Docker image
+        run: |
+          docker pull onyxdotapp/onyx-model-server:latest
+          docker tag onyxdotapp/onyx-model-server:latest onyxdotapp/onyx-model-server:test
+          
       - name: Set up Python
         uses: actions/setup-python@v5
         with:
@@ -41,6 +64,49 @@ jobs:
           pip install --retries 5 --timeout 30 -r backend/requirements/default.txt
           pip install --retries 5 --timeout 30 -r backend/requirements/dev.txt
 
+      - name: Start Docker containers
+        run: |
+          cd deployment/docker_compose
+          ENABLE_PAID_ENTERPRISE_EDITION_FEATURES=true \
+          AUTH_TYPE=basic \
+          REQUIRE_EMAIL_VERIFICATION=false \
+          DISABLE_TELEMETRY=true \
+          IMAGE_TAG=test \
+          docker compose -f docker-compose.dev.yml -p onyx-stack up -d indexing_model_server
+        id: start_docker
+
+      - name: Wait for service to be ready
+        run: |
+          echo "Starting wait-for-service script..."
+
+          start_time=$(date +%s)
+          timeout=300  # 5 minutes in seconds
+
+          while true; do
+            current_time=$(date +%s)
+            elapsed_time=$((current_time - start_time))
+            
+            if [ $elapsed_time -ge $timeout ]; then
+              echo "Timeout reached. Service did not become ready in 5 minutes."
+              exit 1
+            fi
+            
+            # Use curl with error handling to ignore specific exit code 56
+            response=$(curl -s -o /dev/null -w "%{http_code}" http://localhost:9000/api/health || echo "curl_error")
+            
+            if [ "$response" = "200" ]; then
+              echo "Service is ready!"
+              break
+            elif [ "$response" = "curl_error" ]; then
+              echo "Curl encountered an error, possibly exit code 56. Continuing to retry..."
+            else
+              echo "Service not ready yet (HTTP status $response). Retrying in 5 seconds..."
+            fi
+            
+            sleep 5
+          done
+          echo "Finished waiting for service."
+          
       - name: Run Tests
         shell: script -q -e -c "bash --noprofile --norc -eo pipefail {0}"
         run: |
@@ -56,3 +122,10 @@ jobs:
             -H 'Content-type: application/json' \
             --data '{"text":"Scheduled Model Tests failed! Check the run at: https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }}"}' \
             $SLACK_WEBHOOK
+            
+      - name: Stop Docker containers
+        if: always()
+        run: |
+          cd deployment/docker_compose
+          docker compose -f docker-compose.dev.yml -p onyx-stack down -v
+          

README.md

@@ -26,12 +26,12 @@
 
 <strong>[Onyx](https://www.onyx.app/)</strong> (formerly Danswer) is the AI platform connected to your company's docs, apps, and people.
 Onyx provides a feature rich Chat interface and plugs into any LLM of your choice.
-There are over 40 supported connectors such as Google Drive, Slack, Confluence, Salesforce, etc. which keep knowledge and permissions up to date.
-Create custom AI agents with unique prompts, knowledge, and actions the agents can take.
+Keep knowledge and access controls sync-ed across over 40 connectors like Google Drive, Slack, Confluence, Salesforce, etc.
+Create custom AI agents with unique prompts, knowledge, and actions that the agents can take.
 Onyx can be deployed securely anywhere and for any scale - on a laptop, on-premise, or to cloud.
 
 
-<h3>Feature Showcase</h3>
+<h3>Feature Highlights</h3>
 
 **Deep research over your team's knowledge:**
 
@@ -63,22 +63,21 @@ We also have built-in support for high-availability/scalable deployment on Kuber
 References [here](https://github.com/onyx-dot-app/onyx/tree/main/deployment).
 
 
+## 🔍 Other Notable Benefits of Onyx
+- Custom deep learning models for indexing and inference time, only through Onyx + learning from user feedback.
+- Flexible security features like SSO (OIDC/SAML/OAuth2), RBAC, encryption of credentials, etc.
+- Knowledge curation features like document-sets, query history, usage analytics, etc.
+- Scalable deployment options tested up to many tens of thousands users and hundreds of millions of documents.
+
+
 ## 🚧 Roadmap
-- Extensions to the Chrome Plugin
-- Latest methods in information retrieval (StructRAG, LightGraphRAG, etc.)
+- New methods in information retrieval (StructRAG, LightGraphRAG, etc.)
 - Personalized Search
 - Organizational understanding and ability to locate and suggest experts from your team.
 - Code Search
 - SQL and Structured Query Language
 
 
-## 🔍 Other Notable Benefits of Onyx
-- Custom deep learning models only through Onyx + learn from user feedback.
-- Flexible security features like SSO (OIDC/SAML/OAuth2), RBAC, encryption of credentials, etc.
-- Knowledge curation features like document-sets, query history, usage analytics, etc.
-- Scalable deployment options tested up to many tens of thousands users and hundreds of millions of documents.
-
-
 ## 🔌 Connectors
 Keep knowledge and access up to sync across 40+ connectors:
 

backend/alembic/versions/f13db29f3101_add_composite_index_for_last_modified_.py

@@ -0,0 +1,27 @@
+"""Add composite index for last_modified and last_synced to document
+
+Revision ID: f13db29f3101
+Revises: b388730a2899
+Create Date: 2025-02-18 22:48:11.511389
+
+"""
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision = "f13db29f3101"
+down_revision = "acaab4ef4507"
+branch_labels: str | None = None
+depends_on: str | None = None
+
+
+def upgrade() -> None:
+    op.create_index(
+        "ix_document_sync_status",
+        "document",
+        ["last_modified", "last_synced"],
+        unique=False,
+    )
+
+
+def downgrade() -> None:
+    op.drop_index("ix_document_sync_status", table_name="document")

backend/model_server/encoders.py

@@ -98,12 +98,17 @@ class CloudEmbedding:
             return final_embeddings
         except Exception as e:
             error_string = (
-                f"Error embedding text with OpenAI: {str(e)} \n"
-                f"Model: {model} \n"
-                f"Provider: {self.provider} \n"
-                f"Texts: {texts}"
+                f"Exception embedding text with OpenAI - {type(e)}: "
+                f"Model: {model} "
+                f"Provider: {self.provider} "
+                f"Exception: {e}"
             )
             logger.error(error_string)
+
+            # only log text when it's not an authentication error.
+            if not isinstance(e, openai.AuthenticationError):
+                logger.debug(f"Exception texts: {texts}")
+
             raise RuntimeError(error_string)
 
     async def _embed_cohere(

backend/onyx/auth/email_utils.py

@@ -10,6 +10,7 @@ from onyx.configs.app_configs import SMTP_PORT
 from onyx.configs.app_configs import SMTP_SERVER
 from onyx.configs.app_configs import SMTP_USER
 from onyx.configs.app_configs import WEB_DOMAIN
+from onyx.configs.constants import AuthType
 from onyx.configs.constants import TENANT_ID_COOKIE_NAME
 from onyx.db.models import User
 
@@ -187,23 +188,51 @@ def send_subscription_cancellation_email(user_email: str) -> None:
     send_email(user_email, subject, html_content, text_content)
 
 
-def send_user_email_invite(user_email: str, current_user: User) -> None:
+def send_user_email_invite(
+    user_email: str, current_user: User, auth_type: AuthType
+) -> None:
     subject = "Invitation to Join Onyx Organization"
     heading = "You've Been Invited!"
-    message = (
-        f"<p>You have been invited by {current_user.email} to join an organization on Onyx.</p>"
-        "<p>To join the organization, please click the button below to set a password "
-        "or login with Google and complete your registration.</p>"
-    )
+
+    # the exact action taken by the user, and thus the message, depends on the auth type
+    message = f"<p>You have been invited by {current_user.email} to join an organization on Onyx.</p>"
+    if auth_type == AuthType.CLOUD:
+        message += (
+            "<p>To join the organization, please click the button below to set a password "
+            "or login with Google and complete your registration.</p>"
+        )
+    elif auth_type == AuthType.BASIC:
+        message += (
+            "<p>To join the organization, please click the button below to set a password "
+            "and complete your registration.</p>"
+        )
+    elif auth_type == AuthType.GOOGLE_OAUTH:
+        message += (
+            "<p>To join the organization, please click the button below to login with Google "
+            "and complete your registration.</p>"
+        )
+    elif auth_type == AuthType.OIDC or auth_type == AuthType.SAML:
+        message += (
+            "<p>To join the organization, please click the button below to"
+            " complete your registration.</p>"
+        )
+    else:
+        raise ValueError(f"Invalid auth type: {auth_type}")
+
     cta_text = "Join Organization"
     cta_link = f"{WEB_DOMAIN}/auth/signup?email={user_email}"
     html_content = build_html_email(heading, message, cta_text, cta_link)
+
+    # text content is the fallback for clients that don't support HTML
+    # not as critical, so not having special cases for each auth type
     text_content = (
         f"You have been invited by {current_user.email} to join an organization on Onyx.\n"
         "To join the organization, please visit the following link:\n"
         f"{WEB_DOMAIN}/auth/signup?email={user_email}\n"
-        "You'll be asked to set a password or login with Google to complete your registration."
     )
+    if auth_type == AuthType.CLOUD:
+        text_content += "You'll be asked to set a password or login with Google to complete your registration."
+
     send_email(user_email, subject, html_content, text_content)
 
 

backend/onyx/background/celery/apps/app_base.py

@@ -140,7 +140,7 @@ def on_task_postrun(
         f"{f'for tenant_id={tenant_id}' if tenant_id else ''}"
     )
 
-    r = get_redis_client()
+    r = get_redis_client(tenant_id=tenant_id)
 
     if task_id.startswith(RedisConnectorCredentialPair.PREFIX):
         r.srem(RedisConnectorCredentialPair.get_taskset_key(), task_id)

backend/onyx/background/celery/tasks/external_group_syncing/tasks.py

@@ -361,6 +361,7 @@ def connector_external_group_sync_generator_task(
             cc_pair = get_connector_credential_pair_from_id(
                 db_session=db_session,
                 cc_pair_id=cc_pair_id,
+                eager_load_credential=True,
             )
             if cc_pair is None:
                 raise ValueError(

backend/onyx/background/indexing/run_indexing.py

@@ -15,6 +15,7 @@ from onyx.background.indexing.memory_tracer import MemoryTracer
 from onyx.configs.app_configs import INDEX_BATCH_SIZE
 from onyx.configs.app_configs import INDEXING_SIZE_WARNING_THRESHOLD
 from onyx.configs.app_configs import INDEXING_TRACER_INTERVAL
+from onyx.configs.app_configs import INTEGRATION_TESTS_MODE
 from onyx.configs.app_configs import LEAVE_CONNECTOR_ACTIVE_ON_INITIALIZATION_FAILURE
 from onyx.configs.app_configs import POLL_CONNECTOR_OFFSET
 from onyx.configs.constants import DocumentSource
@@ -89,8 +90,8 @@ def _get_connector_runner(
         )
 
         # validate the connector settings
-
-        runnable_connector.validate_connector_settings()
+        if not INTEGRATION_TESTS_MODE:
+            runnable_connector.validate_connector_settings()
 
     except Exception as e:
         logger.exception(f"Unable to instantiate connector due to {e}")

backend/onyx/chat/chat_utils.py

@@ -190,7 +190,8 @@ def create_chat_chain(
             and previous_message.message_type == MessageType.ASSISTANT
             and mainline_messages
         ):
-            mainline_messages[-1] = current_message
+            if current_message.refined_answer_improvement:
+                mainline_messages[-1] = current_message
         else:
             mainline_messages.append(current_message)
 

backend/onyx/chat/models.py

@@ -142,6 +142,15 @@ class MessageResponseIDInfo(BaseModel):
     reserved_assistant_message_id: int
 
 
+class AgentMessageIDInfo(BaseModel):
+    level: int
+    message_id: int
+
+
+class AgenticMessageResponseIDInfo(BaseModel):
+    agentic_message_ids: list[AgentMessageIDInfo]
+
+
 class StreamingError(BaseModel):
     error: str
     stack_trace: str | None = None

backend/onyx/chat/process_message.py

@@ -11,6 +11,8 @@ from onyx.agents.agent_search.orchestration.nodes.call_tool import ToolCallExcep
 from onyx.chat.answer import Answer
 from onyx.chat.chat_utils import create_chat_chain
 from onyx.chat.chat_utils import create_temporary_persona
+from onyx.chat.models import AgenticMessageResponseIDInfo
+from onyx.chat.models import AgentMessageIDInfo
 from onyx.chat.models import AgentSearchPacket
 from onyx.chat.models import AllCitations
 from onyx.chat.models import AnswerPostInfo
@@ -308,6 +310,7 @@ ChatPacket = (
     | CustomToolResponse
     | MessageSpecificCitations
     | MessageResponseIDInfo
+    | AgenticMessageResponseIDInfo
     | StreamStopInfo
     | AgentSearchPacket
 )
@@ -1035,6 +1038,7 @@ def stream_chat_message_objects(
         next_level = 1
         prev_message = gen_ai_response_message
         agent_answers = answer.llm_answer_by_level()
+        agentic_message_ids = []
         while next_level in agent_answers:
             next_answer = agent_answers[next_level]
             info = info_by_subq[
@@ -1059,17 +1063,18 @@ def stream_chat_message_objects(
                 refined_answer_improvement=refined_answer_improvement,
                 is_agentic=True,
             )
+            agentic_message_ids.append(
+                AgentMessageIDInfo(level=next_level, message_id=next_answer_message.id)
+            )
             next_level += 1
             prev_message = next_answer_message
 
         logger.debug("Committing messages")
         db_session.commit()  # actually save user / assistant message
 
-        msg_detail_response = translate_db_message_to_chat_message_detail(
-            gen_ai_response_message
-        )
+        yield AgenticMessageResponseIDInfo(agentic_message_ids=agentic_message_ids)
 
-        yield msg_detail_response
+        yield translate_db_message_to_chat_message_detail(gen_ai_response_message)
     except Exception as e:
         error_msg = str(e)
         logger.exception(error_msg)

backend/onyx/configs/app_configs.py

@@ -158,7 +158,7 @@ POSTGRES_USER = os.environ.get("POSTGRES_USER") or "postgres"
 POSTGRES_PASSWORD = urllib.parse.quote_plus(
     os.environ.get("POSTGRES_PASSWORD") or "password"
 )
-POSTGRES_HOST = os.environ.get("POSTGRES_HOST") or "localhost"
+POSTGRES_HOST = os.environ.get("POSTGRES_HOST") or "127.0.0.1"
 POSTGRES_PORT = os.environ.get("POSTGRES_PORT") or "5432"
 POSTGRES_DB = os.environ.get("POSTGRES_DB") or "postgres"
 AWS_REGION_NAME = os.environ.get("AWS_REGION_NAME") or "us-east-2"
@@ -626,6 +626,8 @@ POD_NAMESPACE = os.environ.get("POD_NAMESPACE")
 
 DEV_MODE = os.environ.get("DEV_MODE", "").lower() == "true"
 
+INTEGRATION_TESTS_MODE = os.environ.get("INTEGRATION_TESTS_MODE", "").lower() == "true"
+
 MOCK_CONNECTOR_FILE_PATH = os.environ.get("MOCK_CONNECTOR_FILE_PATH")
 
 TEST_ENV = os.environ.get("TEST_ENV", "").lower() == "true"

backend/onyx/connectors/factory.py

@@ -3,6 +3,7 @@ from typing import Type
 
 from sqlalchemy.orm import Session
 
+from onyx.configs.app_configs import INTEGRATION_TESTS_MODE
 from onyx.configs.constants import DocumentSource
 from onyx.configs.constants import DocumentSourceRequiringTenantContext
 from onyx.connectors.airtable.airtable_connector import AirtableConnector
@@ -187,6 +188,9 @@ def validate_ccpair_for_user(
     user: User | None,
     tenant_id: str | None,
 ) -> None:
+    if INTEGRATION_TESTS_MODE:
+        return
+
     # Validate the connector settings
     connector = fetch_connector_by_id(connector_id, db_session)
     credential = fetch_credential_by_id_for_user(
@@ -195,11 +199,19 @@ def validate_ccpair_for_user(
         db_session,
         get_editable=False,
     )
-    if not credential:
-        raise ValueError("Credential not found")
+
     if not connector:
         raise ValueError("Connector not found")
 
+    if (
+        connector.source == DocumentSource.INGESTION_API
+        or connector.source == DocumentSource.MOCK_CONNECTOR
+    ):
+        return
+
+    if not credential:
+        raise ValueError("Credential not found")
+
     try:
         runnable_connector = instantiate_connector(
             db_session=db_session,

backend/onyx/connectors/gitbook/connector.py

@@ -229,16 +229,20 @@ class GitbookConnector(LoadConnector, PollConnector):
 
         try:
             content = self.client.get(f"/spaces/{self.space_id}/content")
-            pages = content.get("pages", [])
-
+            pages: list[dict[str, Any]] = content.get("pages", [])
             current_batch: list[Document] = []
-            for page in pages:
-                updated_at = datetime.fromisoformat(page["updatedAt"])
 
+            while pages:
+                page = pages.pop(0)
+
+                updated_at_raw = page.get("updatedAt")
+                if updated_at_raw is None:
+                    # if updatedAt is not present, that means the page has never been edited
+                    continue
+
+                updated_at = datetime.fromisoformat(updated_at_raw)
                 if start and updated_at < start:
-                    if current_batch:
-                        yield current_batch
-                    return
+                    continue
                 if end and updated_at > end:
                     continue
 
@@ -250,6 +254,8 @@ class GitbookConnector(LoadConnector, PollConnector):
                     yield current_batch
                     current_batch = []
 
+                pages.extend(page.get("pages", []))
+
             if current_batch:
                 yield current_batch
 

backend/onyx/connectors/google_drive/connector.py

@@ -220,7 +220,14 @@ class GoogleDriveConnector(LoadConnector, PollConnector, SlimConnector):
         return self._creds
 
     def load_credentials(self, credentials: dict[str, Any]) -> dict[str, str] | None:
-        self._primary_admin_email = credentials[DB_CREDENTIALS_PRIMARY_ADMIN_KEY]
+        try:
+            self._primary_admin_email = credentials[DB_CREDENTIALS_PRIMARY_ADMIN_KEY]
+        except KeyError:
+            raise ValueError(
+                "Primary admin email missing, "
+                "should not call this property "
+                "before calling load_credentials"
+            )
 
         self._creds, new_creds_dict = get_google_creds(
             credentials=credentials,

backend/onyx/db/connector_credential_pair.py

@@ -194,9 +194,14 @@ def get_connector_credential_pair_from_id_for_user(
 def get_connector_credential_pair_from_id(
     db_session: Session,
     cc_pair_id: int,
+    eager_load_credential: bool = False,
 ) -> ConnectorCredentialPair | None:
     stmt = select(ConnectorCredentialPair).distinct()
     stmt = stmt.where(ConnectorCredentialPair.id == cc_pair_id)
+
+    if eager_load_credential:
+        stmt = stmt.options(joinedload(ConnectorCredentialPair.credential))
+
     result = db_session.execute(stmt)
     return result.scalar_one_or_none()
 

backend/onyx/db/document.py

@@ -60,9 +60,8 @@ def count_documents_by_needs_sync(session: Session) -> int:
     This function executes the query and returns the count of
     documents matching the criteria."""
 
-    count = (
-        session.query(func.count(DbDocument.id.distinct()))
-        .select_from(DbDocument)
+    return (
+        session.query(DbDocument.id)
         .join(
             DocumentByConnectorCredentialPair,
             DbDocument.id == DocumentByConnectorCredentialPair.id,
@@ -73,63 +72,53 @@ def count_documents_by_needs_sync(session: Session) -> int:
                 DbDocument.last_synced.is_(None),
             )
         )
-        .scalar()
+        .count()
     )
 
-    return count
-
 
 def construct_document_select_for_connector_credential_pair_by_needs_sync(
     connector_id: int, credential_id: int
 ) -> Select:
-    initial_doc_ids_stmt = select(DocumentByConnectorCredentialPair.id).where(
-        and_(
-            DocumentByConnectorCredentialPair.connector_id == connector_id,
-            DocumentByConnectorCredentialPair.credential_id == credential_id,
-        )
-    )
-
-    stmt = (
+    return (
         select(DbDocument)
-        .where(
-            DbDocument.id.in_(initial_doc_ids_stmt),
-            or_(
-                DbDocument.last_modified
-                > DbDocument.last_synced,  # last_modified is newer than last_synced
-                DbDocument.last_synced.is_(None),  # never synced
-            ),
+        .join(
+            DocumentByConnectorCredentialPair,
+            DbDocument.id == DocumentByConnectorCredentialPair.id,
+        )
+        .where(
+            and_(
+                DocumentByConnectorCredentialPair.connector_id == connector_id,
+                DocumentByConnectorCredentialPair.credential_id == credential_id,
+                or_(
+                    DbDocument.last_modified > DbDocument.last_synced,
+                    DbDocument.last_synced.is_(None),
+                ),
+            )
         )
-        .distinct()
     )
 
-    return stmt
-
 
 def construct_document_id_select_for_connector_credential_pair_by_needs_sync(
     connector_id: int, credential_id: int
 ) -> Select:
-    initial_doc_ids_stmt = select(DocumentByConnectorCredentialPair.id).where(
-        and_(
-            DocumentByConnectorCredentialPair.connector_id == connector_id,
-            DocumentByConnectorCredentialPair.credential_id == credential_id,
-        )
-    )
-
-    stmt = (
+    return (
         select(DbDocument.id)
-        .where(
-            DbDocument.id.in_(initial_doc_ids_stmt),
-            or_(
-                DbDocument.last_modified
-                > DbDocument.last_synced,  # last_modified is newer than last_synced
-                DbDocument.last_synced.is_(None),  # never synced
-            ),
+        .join(
+            DocumentByConnectorCredentialPair,
+            DbDocument.id == DocumentByConnectorCredentialPair.id,
+        )
+        .where(
+            and_(
+                DocumentByConnectorCredentialPair.connector_id == connector_id,
+                DocumentByConnectorCredentialPair.credential_id == credential_id,
+                or_(
+                    DbDocument.last_modified > DbDocument.last_synced,
+                    DbDocument.last_synced.is_(None),
+                ),
+            )
         )
-        .distinct()
     )
 
-    return stmt
-
 
 def get_all_documents_needing_vespa_sync_for_cc_pair(
     db_session: Session, cc_pair_id: int

backend/onyx/db/models.py

@@ -570,6 +570,14 @@ class Document(Base):
         back_populates="documents",
     )
 
+    __table_args__ = (
+        Index(
+            "ix_document_sync_status",
+            last_modified,
+            last_synced,
+        ),
+    )
+
 
 class Tag(Base):
     __tablename__ = "tag"

backend/onyx/llm/models.py

@@ -23,6 +23,7 @@ class PreviousMessage(BaseModel):
     message_type: MessageType
     files: list[InMemoryChatFile]
     tool_call: ToolCallFinalResult | None
+    refined_answer_improvement: bool | None
 
     @classmethod
     def from_chat_message(
@@ -47,6 +48,7 @@ class PreviousMessage(BaseModel):
             )
             if chat_message.tool_call
             else None,
+            refined_answer_improvement=chat_message.refined_answer_improvement,
         )
 
     def to_langchain_msg(self) -> BaseMessage:

backend/onyx/server/manage/users.py

@@ -311,19 +311,23 @@ def bulk_invite_users(
     all_emails = list(set(new_invited_emails) | set(initial_invited_users))
     number_of_invited_users = write_invited_users(all_emails)
 
+    # send out email invitations if enabled
+    if ENABLE_EMAIL_INVITES:
+        try:
+            for email in new_invited_emails:
+                send_user_email_invite(email, current_user, AUTH_TYPE)
+        except Exception as e:
+            logger.error(f"Error sending email invite to invited users: {e}")
+
     if not MULTI_TENANT:
         return number_of_invited_users
+
+    # for billing purposes, write to the control plane about the number of new users
     try:
         logger.info("Registering tenant users")
         fetch_ee_implementation_or_noop(
             "onyx.server.tenants.billing", "register_tenant_users", None
         )(tenant_id, get_total_users_count(db_session))
-        if ENABLE_EMAIL_INVITES:
-            try:
-                for email in new_invited_emails:
-                    send_user_email_invite(email, current_user)
-            except Exception as e:
-                logger.error(f"Error sending email invite to invited users: {e}")
 
         return number_of_invited_users
     except Exception as e:

backend/onyx/server/settings/models.py

@@ -45,7 +45,7 @@ class Settings(BaseModel):
     gpu_enabled: bool | None = None
     application_status: ApplicationStatus = ApplicationStatus.ACTIVE
     anonymous_user_enabled: bool | None = None
-    pro_search_disabled: bool | None = None
+    pro_search_enabled: bool | None = None
 
     temperature_override_enabled: bool = False
     auto_scroll: bool = False

backend/requirements/ee.txt

@@ -1,3 +1,4 @@
 cohere==5.6.1
 posthog==3.7.4
 python3-saml==1.15.0
+xmlsec==1.3.14

backend/scripts/debugging/onyx_redis.py

@@ -3,6 +3,7 @@ import json
 import logging
 import sys
 import time
+from enum import Enum
 from logging import getLogger
 from typing import cast
 from uuid import UUID
@@ -20,10 +21,13 @@ from onyx.configs.app_configs import REDIS_PORT
 from onyx.configs.app_configs import REDIS_SSL
 from onyx.db.engine import get_session_with_tenant
 from onyx.db.users import get_user_by_email
+from onyx.redis.redis_connector import RedisConnector
+from onyx.redis.redis_connector_index import RedisConnectorIndex
 from onyx.redis.redis_pool import RedisPool
 from shared_configs.configs import MULTI_TENANT
 from shared_configs.configs import POSTGRES_DEFAULT_SCHEMA
 from shared_configs.contextvars import CURRENT_TENANT_ID_CONTEXTVAR
+from shared_configs.contextvars import get_current_tenant_id
 
 # Tool to run helpful operations on Redis in production
 # This is targeted for internal usage and may not have all the necessary parameters
@@ -42,6 +46,19 @@ SCAN_ITER_COUNT = 10000
 BATCH_DEFAULT = 1000
 
 
+class OnyxRedisCommand(Enum):
+    purge_connectorsync_taskset = "purge_connectorsync_taskset"
+    purge_documentset_taskset = "purge_documentset_taskset"
+    purge_usergroup_taskset = "purge_usergroup_taskset"
+    purge_locks_blocking_deletion = "purge_locks_blocking_deletion"
+    purge_vespa_syncing = "purge_vespa_syncing"
+    get_user_token = "get_user_token"
+    delete_user_token = "delete_user_token"
+
+    def __str__(self) -> str:
+        return self.value
+
+
 def get_user_id(user_email: str) -> tuple[UUID, str]:
     tenant_id = (
         get_tenant_id_for_email(user_email) if MULTI_TENANT else POSTGRES_DEFAULT_SCHEMA
@@ -55,50 +72,79 @@ def get_user_id(user_email: str) -> tuple[UUID, str]:
 
 
 def onyx_redis(
-    command: str,
+    command: OnyxRedisCommand,
     batch: int,
     dry_run: bool,
+    ssl: bool,
     host: str,
     port: int,
     db: int,
     password: str | None,
     user_email: str | None = None,
+    cc_pair_id: int | None = None,
 ) -> int:
+    # this is global and not tenant aware
     pool = RedisPool.create_pool(
         host=host,
         port=port,
         db=db,
         password=password if password else "",
-        ssl=REDIS_SSL,
+        ssl=ssl,
         ssl_cert_reqs="optional",
         ssl_ca_certs=None,
     )
 
     r = Redis(connection_pool=pool)
 
+    logger.info("Redis ping starting. This may hang if your settings are incorrect.")
+
     try:
         r.ping()
     except:
         logger.exception("Redis ping exceptioned")
         raise
 
-    if command == "purge_connectorsync_taskset":
+    logger.info("Redis ping succeeded.")
+
+    if command == OnyxRedisCommand.purge_connectorsync_taskset:
         """Purge connector tasksets. Used when the tasks represented in the tasksets
         have been purged."""
         return purge_by_match_and_type(
             "*connectorsync_taskset*", "set", batch, dry_run, r
         )
-    elif command == "purge_documentset_taskset":
+    elif command == OnyxRedisCommand.purge_documentset_taskset:
         return purge_by_match_and_type(
             "*documentset_taskset*", "set", batch, dry_run, r
         )
-    elif command == "purge_usergroup_taskset":
+    elif command == OnyxRedisCommand.purge_usergroup_taskset:
         return purge_by_match_and_type("*usergroup_taskset*", "set", batch, dry_run, r)
-    elif command == "purge_vespa_syncing":
+    elif command == OnyxRedisCommand.purge_locks_blocking_deletion:
+        if cc_pair_id is None:
+            logger.error("You must specify --cc-pair with purge_deletion_locks")
+            return 1
+
+        tenant_id = get_current_tenant_id()
+        logger.info(f"Purging locks associated with deleting cc_pair={cc_pair_id}.")
+        redis_connector = RedisConnector(tenant_id, cc_pair_id)
+
+        match_pattern = f"{tenant_id}:{RedisConnectorIndex.FENCE_PREFIX}_{cc_pair_id}/*"
+        purge_by_match_and_type(match_pattern, "string", batch, dry_run, r)
+
+        redis_delete_if_exists_helper(
+            f"{tenant_id}:{redis_connector.prune.fence_key}", dry_run, r
+        )
+        redis_delete_if_exists_helper(
+            f"{tenant_id}:{redis_connector.permissions.fence_key}", dry_run, r
+        )
+        redis_delete_if_exists_helper(
+            f"{tenant_id}:{redis_connector.external_group_sync.fence_key}", dry_run, r
+        )
+        return 0
+    elif command == OnyxRedisCommand.purge_vespa_syncing:
         return purge_by_match_and_type(
             "*connectorsync:vespa_syncing*", "string", batch, dry_run, r
         )
-    elif command == "get_user_token":
+    elif command == OnyxRedisCommand.get_user_token:
         if not user_email:
             logger.error("You must specify --user-email with get_user_token")
             return 1
@@ -109,7 +155,7 @@ def onyx_redis(
         else:
             print(f"No token found for user {user_email}")
             return 2
-    elif command == "delete_user_token":
+    elif command == OnyxRedisCommand.delete_user_token:
         if not user_email:
             logger.error("You must specify --user-email with delete_user_token")
             return 1
@@ -131,6 +177,25 @@ def flush_batch_delete(batch_keys: list[bytes], r: Redis) -> None:
         pipe.execute()
 
 
+def redis_delete_if_exists_helper(key: str, dry_run: bool, r: Redis) -> bool:
+    """Returns True if the key was found, False if not.
+    This function exists for logging purposes as the delete operation itself
+    doesn't really need to check the existence of the key.
+    """
+
+    if not r.exists(key):
+        logger.info(f"Did not find {key}.")
+        return False
+
+    if dry_run:
+        logger.info(f"(DRY-RUN) Deleting {key}.")
+    else:
+        logger.info(f"Deleting {key}.")
+        r.delete(key)
+
+    return True
+
+
 def purge_by_match_and_type(
     match_pattern: str, match_type: str, batch_size: int, dry_run: bool, r: Redis
 ) -> int:
@@ -138,6 +203,12 @@ def purge_by_match_and_type(
     match_type: https://redis.io/docs/latest/commands/type/
     """
 
+    logger.info(
+        f"purge_by_match_and_type start: "
+        f"match_pattern={match_pattern} "
+        f"match_type={match_type}"
+    )
+
     # cursor = "0"
     # while cursor != 0:
     #     cursor, data = self.scan(
@@ -164,13 +235,15 @@ def purge_by_match_and_type(
         logger.info(f"Deleting item {count}: {key_str}")
 
         batch_keys.append(key)
+
+        # flush if batch size has been reached
         if len(batch_keys) >= batch_size:
             flush_batch_delete(batch_keys, r)
             batch_keys.clear()
 
-    if len(batch_keys) >= batch_size:
-        flush_batch_delete(batch_keys, r)
-        batch_keys.clear()
+    # final flush
+    flush_batch_delete(batch_keys, r)
+    batch_keys.clear()
 
     logger.info(f"Deleted {count} matches.")
 
@@ -279,7 +352,21 @@ def delete_user_token_from_redis(
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description="Onyx Redis Manager")
-    parser.add_argument("--command", type=str, help="Operation to run", required=True)
+    parser.add_argument(
+        "--command",
+        type=OnyxRedisCommand,
+        help="The command to run",
+        choices=list(OnyxRedisCommand),
+        required=True,
+    )
+
+    parser.add_argument(
+        "--ssl",
+        type=bool,
+        default=REDIS_SSL,
+        help="Use SSL when connecting to Redis. Usually True for prod and False for local testing",
+        required=False,
+    )
 
     parser.add_argument(
         "--host",
@@ -342,6 +429,13 @@ if __name__ == "__main__":
         required=False,
     )
 
+    parser.add_argument(
+        "--cc-pair",
+        type=int,
+        help="A connector credential pair id. Used with the purge_deletion_locks command.",
+        required=False,
+    )
+
     args = parser.parse_args()
 
     if args.tenant_id:
@@ -368,10 +462,12 @@ if __name__ == "__main__":
         command=args.command,
         batch=args.batch,
         dry_run=args.dry_run,
+        ssl=args.ssl,
         host=args.host,
         port=args.port,
         db=args.db,
         password=args.password,
         user_email=args.user_email,
+        cc_pair_id=args.cc_pair,
     )
     sys.exit(exitcode)

backend/tests/integration/common_utils/constants.py

@@ -3,7 +3,7 @@ import os
 ADMIN_USER_NAME = "admin_user"
 
 API_SERVER_PROTOCOL = os.getenv("API_SERVER_PROTOCOL") or "http"
-API_SERVER_HOST = os.getenv("API_SERVER_HOST") or "localhost"
+API_SERVER_HOST = os.getenv("API_SERVER_HOST") or "127.0.0.1"
 API_SERVER_PORT = os.getenv("API_SERVER_PORT") or "8080"
 API_SERVER_URL = f"{API_SERVER_PROTOCOL}://{API_SERVER_HOST}:{API_SERVER_PORT}"
 MAX_DELAY = 45

backend/tests/integration/common_utils/managers/connector.py

@@ -30,8 +30,10 @@ class ConnectorManager:
             name=name,
             source=source,
             input_type=input_type,
-            connector_specific_config=connector_specific_config
-            or {"file_locations": []},
+            connector_specific_config=(
+                connector_specific_config
+                or ({"file_locations": []} if source == DocumentSource.FILE else {})
+            ),
             access_type=access_type,
             groups=groups or [],
         )

backend/tests/integration/common_utils/managers/user.py

@@ -88,8 +88,6 @@ class UserManager:
         if not session_cookie:
             raise Exception("Failed to login")
 
-        print(f"Logged in as {test_user.email}")
-
         # Set cookies in the headers
         test_user.headers["Cookie"] = f"fastapiusersauth={session_cookie}; "
         test_user.cookies = {"fastapiusersauth": session_cookie}

deployment/data/nginx/app.conf.template

@@ -4,6 +4,24 @@ log_format custom_main '$remote_addr - $remote_user [$time_local] "$request" '
                 '"$http_user_agent" "$http_x_forwarded_for" '
                 'rt=$request_time';
 
+# Map X-Forwarded-Proto or fallback to $scheme
+map $http_x_forwarded_proto $forwarded_proto {
+    default $http_x_forwarded_proto;
+    ""      $scheme;
+}
+
+# Map X-Forwarded-Host or fallback to $host
+map $http_x_forwarded_host $forwarded_host {
+    default $http_x_forwarded_host;
+    ""      $host;
+}
+
+# Map X-Forwarded-Port or fallback to server port
+map $http_x_forwarded_port $forwarded_port {
+    default $http_x_forwarded_port;
+    ""      $server_port;
+}
+
 upstream api_server {
     # fail_timeout=0 means we always retry an upstream even if it failed
     # to return a good HTTP response
@@ -21,8 +39,7 @@ upstream web_server {
 }
 
 server {
-    listen 80;
-    server_name ${DOMAIN};
+    listen 80 default_server;
 
     client_max_body_size 5G;    # Maximum upload size
 
@@ -36,8 +53,9 @@ server {
         # misc headers
         proxy_set_header X-Real-IP $remote_addr;
         proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-        proxy_set_header X-Forwarded-Proto $scheme;
-        proxy_set_header X-Forwarded-Host $host; 
+        proxy_set_header X-Forwarded-Proto $forwarded_proto;
+        proxy_set_header X-Forwarded-Host $forwarded_host; 
+        proxy_set_header X-Forwarded-Port $forwarded_port;
         proxy_set_header Host $host;
 
         # need to use 1.1 to support chunked transfers
@@ -54,8 +72,9 @@ server {
         # misc headers
         proxy_set_header X-Real-IP $remote_addr;
         proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-        proxy_set_header X-Forwarded-Proto $scheme;
-        proxy_set_header X-Forwarded-Host $host; 
+        proxy_set_header X-Forwarded-Proto $forwarded_proto;
+        proxy_set_header X-Forwarded-Host $forwarded_host; 
+        proxy_set_header X-Forwarded-Port $forwarded_port;
         proxy_set_header Host $host;
 
         proxy_http_version 1.1;
@@ -72,14 +91,25 @@ server {
 }
 
 server {
-    listen 443 ssl;
-    server_name ${DOMAIN};
+    listen 443 ssl default_server;
 
     client_max_body_size 5G;    # Maximum upload size
     
     location / {
+        # misc headers
+        proxy_set_header X-Real-IP $remote_addr;
+        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+        # don't use forwarded schema, host, or port here - this is the entry point
+        proxy_set_header X-Forwarded-Proto $scheme;
+        proxy_set_header X-Forwarded-Host $host; 
+        proxy_set_header X-Forwarded-Port $server_port;
+        proxy_set_header Host $host;
+
         proxy_http_version 1.1;
         proxy_buffering off;
+        # we don't want nginx trying to do something clever with
+        # redirects, we set the Host: header above already.
+        proxy_redirect off;
         proxy_pass http://localhost:80;
     }
 

deployment/data/nginx/app.conf.template.dev

@@ -21,8 +21,7 @@ upstream web_server {
 }
 
 server {
-    listen 80;
-    server_name ${DOMAIN};
+    listen 80 default_server;
 
     client_max_body_size 5G;    # Maximum upload size    
 
@@ -37,7 +36,8 @@ server {
         proxy_set_header X-Real-IP $remote_addr;
         proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
         proxy_set_header X-Forwarded-Proto $scheme;
-        proxy_set_header X-Forwarded-Host $host; 
+        proxy_set_header X-Forwarded-Host $host;
+        proxy_set_header X-Forwarded-Port $server_port;
         proxy_set_header Host $host;
 
         # need to use 1.1 to support chunked transfers
@@ -55,7 +55,8 @@ server {
         proxy_set_header X-Real-IP $remote_addr;
         proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
         proxy_set_header X-Forwarded-Proto $scheme;
-        proxy_set_header X-Forwarded-Host $host; 
+        proxy_set_header X-Forwarded-Host $host;
+        proxy_set_header X-Forwarded-Port $server_port;
         proxy_set_header Host $host;
 
         proxy_http_version 1.1;

deployment/data/nginx/app.conf.template.no-letsencrypt

@@ -4,6 +4,24 @@ log_format custom_main '$remote_addr - $remote_user [$time_local] "$request" '
                 '"$http_user_agent" "$http_x_forwarded_for" '
                 'rt=$request_time';
 
+# Map X-Forwarded-Proto or fallback to $scheme
+map $http_x_forwarded_proto $forwarded_proto {
+    default $http_x_forwarded_proto;
+    ""      $scheme;
+}
+
+# Map X-Forwarded-Host or fallback to $host
+map $http_x_forwarded_host $forwarded_host {
+    default $http_x_forwarded_host;
+    ""      $host;
+}
+
+# Map X-Forwarded-Port or fallback to server port
+map $http_x_forwarded_port $forwarded_port {
+    default $http_x_forwarded_port;
+    ""      $server_port;
+}
+
 upstream api_server {
     # fail_timeout=0 means we always retry an upstream even if it failed
     # to return a good HTTP response
@@ -21,8 +39,7 @@ upstream web_server {
 }
 
 server {
-    listen 80;
-    server_name ${DOMAIN};
+    listen 80 default_server;
 
     client_max_body_size 5G;    # Maximum upload size
 
@@ -36,8 +53,9 @@ server {
         # misc headers
         proxy_set_header X-Real-IP $remote_addr;
         proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-        proxy_set_header X-Forwarded-Proto $scheme;
-        proxy_set_header X-Forwarded-Host $host; 
+        proxy_set_header X-Forwarded-Proto $forwarded_proto;
+        proxy_set_header X-Forwarded-Host $forwarded_host; 
+        proxy_set_header X-Forwarded-Port $forwarded_port;
         proxy_set_header Host $host;
 
         # need to use 1.1 to support chunked transfers
@@ -54,8 +72,9 @@ server {
         # misc headers
         proxy_set_header X-Real-IP $remote_addr;
         proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-        proxy_set_header X-Forwarded-Proto $scheme;
-        proxy_set_header X-Forwarded-Host $host; 
+        proxy_set_header X-Forwarded-Proto $forwarded_proto;
+        proxy_set_header X-Forwarded-Host $forwarded_host; 
+        proxy_set_header X-Forwarded-Port $forwarded_port;
         proxy_set_header Host $host;
 
         proxy_http_version 1.1;
@@ -68,14 +87,25 @@ server {
 }
 
 server {
-    listen 443 ssl;
-    server_name ${DOMAIN};
+    listen 443 ssl default_server;
 
     client_max_body_size 5G;    # Maximum upload size
     
     location / {
+        # misc headers
+        proxy_set_header X-Real-IP $remote_addr;
+        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+        # don't use forwarded schema, host, or port here - this is the entry point
+        proxy_set_header X-Forwarded-Proto $scheme;
+        proxy_set_header X-Forwarded-Host $host; 
+        proxy_set_header X-Forwarded-Port $server_port;
+        proxy_set_header Host $host;
+
         proxy_http_version 1.1;
         proxy_buffering off;
+        # we don't want nginx trying to do something clever with
+        # redirects, we set the Host: header above already.
+        proxy_redirect off;
         proxy_pass http://localhost:80;
     }
 

deployment/docker_compose/docker-compose.dev.yml

@@ -36,6 +36,7 @@ services:
       - OPENID_CONFIG_URL=${OPENID_CONFIG_URL:-}
       - TRACK_EXTERNAL_IDP_EXPIRY=${TRACK_EXTERNAL_IDP_EXPIRY:-}
       - CORS_ALLOWED_ORIGIN=${CORS_ALLOWED_ORIGIN:-}
+      - INTEGRATION_TESTS_MODE=${INTEGRATION_TESTS_MODE:-}
       # Gen AI Settings
       - GEN_AI_MAX_TOKENS=${GEN_AI_MAX_TOKENS:-}
       - QA_TIMEOUT=${QA_TIMEOUT:-}

web/src/app/admin/assistants/AssistantEditor.tsx

@@ -113,7 +113,6 @@ export function AssistantEditor({
   documentSets,
   user,
   defaultPublic,
-  redirectType,
   llmProviders,
   tools,
   shouldAddAssistantToUserPreferences,
@@ -124,7 +123,6 @@ export function AssistantEditor({
   documentSets: DocumentSet[];
   user: User | null;
   defaultPublic: boolean;
-  redirectType: SuccessfulPersonaUpdateRedirectType;
   llmProviders: FullLLMProvider[];
   tools: ToolSnapshot[];
   shouldAddAssistantToUserPreferences?: boolean;
@@ -502,7 +500,7 @@ export function AssistantEditor({
             )
             .map((message: { message: string; name?: string }) => ({
               message: message.message,
-              name: message.name || message.message,
+              name: message.message,
             }));
 
           // don't set groups if marked as public

web/src/app/admin/settings/SettingsForm.tsx

@@ -240,11 +240,11 @@ export function SettingsForm() {
       />
 
       <Checkbox
-        label="Pro Search Disabled"
-        sublabel="If set, users will not be able to use Pro Search."
-        checked={settings.pro_search_disabled ?? false}
+        label="Agent Search"
+        sublabel="If set, users will be able to use Agent Search."
+        checked={settings.pro_search_enabled ?? true}
         onChange={(e) =>
-          handleToggleSettingsField("pro_search_disabled", e.target.checked)
+          handleToggleSettingsField("pro_search_enabled", e.target.checked)
         }
       />
 

web/src/app/admin/settings/interfaces.ts

@@ -10,7 +10,7 @@ export interface Settings {
   notifications: Notification[];
   needs_reindexing: boolean;
   gpu_enabled: boolean;
-  pro_search_disabled: boolean | null;
+  pro_search_enabled: boolean | null;
   application_status: ApplicationStatus;
   auto_scroll: boolean;
   temperature_override_enabled: boolean;

web/src/app/assistants/edit/[id]/page.tsx

@@ -21,11 +21,7 @@ export default async function Page(props: { params: Promise<{ id: string }> }) {
         <div className="px-32">
           <div className="mx-auto container">
             <CardSection className="!border-none !bg-transparent !ring-none">
-              <AssistantEditor
-                {...values}
-                defaultPublic={false}
-                redirectType={SuccessfulPersonaUpdateRedirectType.CHAT}
-              />
+              <AssistantEditor {...values} defaultPublic={false} />
             </CardSection>
           </div>
         </div>

web/src/app/assistants/new/page.tsx

@@ -26,7 +26,6 @@ export default async function Page() {
               <AssistantEditor
                 {...values}
                 defaultPublic={false}
-                redirectType={SuccessfulPersonaUpdateRedirectType.CHAT}
                 shouldAddAssistantToUserPreferences={true}
               />
             </CardSection>

web/src/app/chat/ChatPage.tsx

@@ -23,6 +23,7 @@ import {
   SubQuestionDetail,
   constructSubQuestions,
   DocumentsResponse,
+  AgenticMessageResponseIDInfo,
 } from "./interfaces";
 
 import Prism from "prismjs";
@@ -46,6 +47,7 @@ import {
   removeMessage,
   sendMessage,
   setMessageAsLatest,
+  updateLlmOverrideForChatSession,
   updateParentChildren,
   uploadFilesForChat,
   useScrollonStream,
@@ -64,7 +66,7 @@ import {
 import { usePopup } from "@/components/admin/connectors/Popup";
 import { SEARCH_PARAM_NAMES, shouldSubmitOnLoad } from "./searchParams";
 import { useDocumentSelection } from "./useDocumentSelection";
-import { LlmOverride, useFilters, useLlmOverride } from "@/lib/hooks";
+import { LlmDescriptor, useFilters, useLlmManager } from "@/lib/hooks";
 import { ChatState, FeedbackType, RegenerationState } from "./types";
 import { DocumentResults } from "./documentSidebar/DocumentResults";
 import { OnyxInitializingLoader } from "@/components/OnyxInitializingLoader";
@@ -88,7 +90,11 @@ import {
 import { buildFilters } from "@/lib/search/utils";
 import { SettingsContext } from "@/components/settings/SettingsProvider";
 import Dropzone from "react-dropzone";
-import { checkLLMSupportsImageInput, getFinalLLM } from "@/lib/llm/utils";
+import {
+  checkLLMSupportsImageInput,
+  getFinalLLM,
+  structureValue,
+} from "@/lib/llm/utils";
 import { ChatInputBar } from "./input/ChatInputBar";
 import { useChatContext } from "@/components/context/ChatContext";
 import { v4 as uuidv4 } from "uuid";
@@ -355,7 +361,7 @@ export function ChatPage({
     ]
   );
 
-  const llmOverrideManager = useLlmOverride(
+  const llmManager = useLlmManager(
     llmProviders,
     selectedChatSession,
     liveAssistant
@@ -1137,7 +1143,7 @@ export function ChatPage({
     forceSearch,
     isSeededChat,
     alternativeAssistantOverride = null,
-    modelOverRide,
+    modelOverride,
     regenerationRequest,
     overrideFileDescriptors,
   }: {
@@ -1147,7 +1153,7 @@ export function ChatPage({
     forceSearch?: boolean;
     isSeededChat?: boolean;
     alternativeAssistantOverride?: Persona | null;
-    modelOverRide?: LlmOverride;
+    modelOverride?: LlmDescriptor;
     regenerationRequest?: RegenerationRequest | null;
     overrideFileDescriptors?: FileDescriptor[];
   } = {}) => {
@@ -1190,6 +1196,22 @@ export function ChatPage({
       currChatSessionId = chatSessionIdRef.current as string;
     }
     frozenSessionId = currChatSessionId;
+    // update the selected model for the chat session if one is specified so that
+    // it persists across page reloads. Do not `await` here so that the message
+    // request can continue and this will just happen in the background.
+    // NOTE: only set the model override for the chat session once we send a
+    // message with it. If the user switches models and then starts a new
+    // chat session, it is unexpected for that model to be used when they
+    // return to this session the next day.
+    let finalLLM = modelOverride || llmManager.currentLlm;
+    updateLlmOverrideForChatSession(
+      currChatSessionId,
+      structureValue(
+        finalLLM.name || "",
+        finalLLM.provider || "",
+        finalLLM.modelName || ""
+      )
+    );
 
     updateStatesWithNewSessionId(currChatSessionId);
 
@@ -1249,11 +1271,14 @@ export function ChatPage({
         : null) ||
       (messageMap.size === 1 ? Array.from(messageMap.values())[0] : null);
 
-    const currentAssistantId = alternativeAssistantOverride
-      ? alternativeAssistantOverride.id
-      : alternativeAssistant
-        ? alternativeAssistant.id
-        : liveAssistant.id;
+    let currentAssistantId;
+    if (alternativeAssistantOverride) {
+      currentAssistantId = alternativeAssistantOverride.id;
+    } else if (alternativeAssistant) {
+      currentAssistantId = alternativeAssistant.id;
+    } else {
+      currentAssistantId = liveAssistant.id;
+    }
 
     resetInputBar();
     let messageUpdates: Message[] | null = null;
@@ -1280,6 +1305,8 @@ export function ChatPage({
     let toolCall: ToolCallMetadata | null = null;
     let isImprovement: boolean | undefined = undefined;
     let isStreamingQuestions = true;
+    let includeAgentic = false;
+    let secondLevelMessageId: number | null = null;
 
     let initialFetchDetails: null | {
       user_message_id: number;
@@ -1323,20 +1350,18 @@ export function ChatPage({
         forceSearch,
         regenerate: regenerationRequest !== undefined,
         modelProvider:
-          modelOverRide?.name ||
-          llmOverrideManager.llmOverride.name ||
-          undefined,
+          modelOverride?.name || llmManager.currentLlm.name || undefined,
         modelVersion:
-          modelOverRide?.modelName ||
-          llmOverrideManager.llmOverride.modelName ||
+          modelOverride?.modelName ||
+          llmManager.currentLlm.modelName ||
           searchParams.get(SEARCH_PARAM_NAMES.MODEL_VERSION) ||
           undefined,
-        temperature: llmOverrideManager.temperature || undefined,
+        temperature: llmManager.temperature || undefined,
         systemPromptOverride:
           searchParams.get(SEARCH_PARAM_NAMES.SYSTEM_PROMPT) || undefined,
         useExistingUserMessage: isSeededChat,
         useLanggraph:
-          !settings?.settings.pro_search_disabled &&
+          settings?.settings.pro_search_enabled &&
           proSearchEnabled &&
           retrievalEnabled,
       });
@@ -1417,6 +1442,17 @@ export function ChatPage({
             resetRegenerationState();
           } else {
             const { user_message_id, frozenMessageMap } = initialFetchDetails;
+            if (Object.hasOwn(packet, "agentic_message_ids")) {
+              const agenticMessageIds = (packet as AgenticMessageResponseIDInfo)
+                .agentic_message_ids;
+              const level1MessageId = agenticMessageIds.find(
+                (item) => item.level === 1
+              )?.message_id;
+              if (level1MessageId) {
+                secondLevelMessageId = level1MessageId;
+                includeAgentic = true;
+              }
+            }
 
             setChatState((prevState) => {
               if (prevState.get(chatSessionIdRef.current!) === "loading") {
@@ -1568,7 +1604,10 @@ export function ChatPage({
                   };
                 }
               );
-            } else if (Object.hasOwn(packet, "error")) {
+            } else if (
+              Object.hasOwn(packet, "error") &&
+              (packet as any).error != null
+            ) {
               if (
                 sub_questions.length > 0 &&
                 sub_questions
@@ -1580,8 +1619,8 @@ export function ChatPage({
                 setAgenticGenerating(false);
                 setAlternativeGeneratingAssistant(null);
                 setSubmittedMessage("");
-                return;
-                // throw new Error((packet as StreamingError).error);
+
+                throw new Error((packet as StreamingError).error);
               } else {
                 error = (packet as StreamingError).error;
                 stackTrace = (packet as StreamingError).stack_trace;
@@ -1664,6 +1703,19 @@ export function ChatPage({
                 second_level_generating: second_level_generating,
                 agentic_docs: agenticDocs,
               },
+              ...(includeAgentic
+                ? [
+                    {
+                      messageId: secondLevelMessageId!,
+                      message: second_level_answer,
+                      type: "assistant" as const,
+                      files: [],
+                      toolCall: null,
+                      parentMessageId:
+                        initialFetchDetails.assistant_message_id!,
+                    },
+                  ]
+                : []),
             ]);
           }
         }
@@ -1772,7 +1824,7 @@ export function ChatPage({
     const [_, llmModel] = getFinalLLM(
       llmProviders,
       liveAssistant,
-      llmOverrideManager.llmOverride
+      llmManager.currentLlm
     );
     const llmAcceptsImages = checkLLMSupportsImageInput(llmModel);
 
@@ -2091,7 +2143,7 @@ export function ChatPage({
   }, [searchParams, router]);
 
   useEffect(() => {
-    llmOverrideManager.updateImageFilesPresent(imageFileInMessageHistory);
+    llmManager.updateImageFilesPresent(imageFileInMessageHistory);
   }, [imageFileInMessageHistory]);
 
   const pathname = usePathname();
@@ -2145,9 +2197,9 @@ export function ChatPage({
 
   function createRegenerator(regenerationRequest: RegenerationRequest) {
     // Returns new function that only needs `modelOverRide` to be specified when called
-    return async function (modelOverRide: LlmOverride) {
+    return async function (modelOverride: LlmDescriptor) {
       return await onSubmit({
-        modelOverRide,
+        modelOverride,
         messageIdToResend: regenerationRequest.parentMessage.messageId,
         regenerationRequest,
         forceSearch: regenerationRequest.forceSearch,
@@ -2228,9 +2280,7 @@ export function ChatPage({
       {(settingsToggled || userSettingsToggled) && (
         <UserSettingsModal
           setPopup={setPopup}
-          setLlmOverride={(newOverride) =>
-            llmOverrideManager.updateLLMOverride(newOverride)
-          }
+          setCurrentLlm={(newLlm) => llmManager.updateCurrentLlm(newLlm)}
           defaultModel={user?.preferences.default_model!}
           llmProviders={llmProviders}
           onClose={() => {
@@ -2294,7 +2344,7 @@ export function ChatPage({
         <ShareChatSessionModal
           assistantId={liveAssistant?.id}
           message={message}
-          modelOverride={llmOverrideManager.llmOverride}
+          modelOverride={llmManager.currentLlm}
           chatSessionId={sharedChatSession.id}
           existingSharedStatus={sharedChatSession.shared_status}
           onClose={() => setSharedChatSession(null)}
@@ -2312,7 +2362,7 @@ export function ChatPage({
         <ShareChatSessionModal
           message={message}
           assistantId={liveAssistant?.id}
-          modelOverride={llmOverrideManager.llmOverride}
+          modelOverride={llmManager.currentLlm}
           chatSessionId={chatSessionIdRef.current}
           existingSharedStatus={chatSessionSharedStatus}
           onClose={() => setSharingModalVisible(false)}
@@ -2692,6 +2742,11 @@ export function ChatPage({
                                     ? messageHistory[i + 1]?.documents
                                     : undefined;
 
+                                const nextMessage =
+                                  messageHistory[i + 1]?.type === "assistant"
+                                    ? messageHistory[i + 1]
+                                    : undefined;
+
                                 return (
                                   <div
                                     className="text-text"
@@ -2720,7 +2775,10 @@ export function ChatPage({
                                             selectedMessageForDocDisplay ==
                                               secondLevelMessage?.messageId)
                                         }
-                                        isImprovement={message.isImprovement}
+                                        isImprovement={
+                                          message.isImprovement ||
+                                          nextMessage?.isImprovement
+                                        }
                                         secondLevelGenerating={
                                           (message.second_level_generating &&
                                             currentSessionChatState !==
@@ -3020,7 +3078,7 @@ export function ChatPage({
                                               messageId: message.messageId,
                                               parentMessage: parentMessage!,
                                               forceSearch: true,
-                                            })(llmOverrideManager.llmOverride);
+                                            })(llmManager.currentLlm);
                                           } else {
                                             setPopup({
                                               type: "error",
@@ -3165,7 +3223,7 @@ export function ChatPage({
                               availableDocumentSets={documentSets}
                               availableTags={tags}
                               filterManager={filterManager}
-                              llmOverrideManager={llmOverrideManager}
+                              llmManager={llmManager}
                               removeDocs={() => {
                                 clearSelectedDocuments();
                               }}

web/src/app/chat/RegenerateOption.tsx

@@ -1,8 +1,8 @@
 import { useChatContext } from "@/components/context/ChatContext";
 import {
   getDisplayNameForModel,
-  LlmOverride,
-  useLlmOverride,
+  LlmDescriptor,
+  useLlmManager,
 } from "@/lib/hooks";
 import { StringOrNumberOption } from "@/components/Dropdown";
 
@@ -106,13 +106,13 @@ export default function RegenerateOption({
   onDropdownVisibleChange,
 }: {
   selectedAssistant: Persona;
-  regenerate: (modelOverRide: LlmOverride) => Promise<void>;
+  regenerate: (modelOverRide: LlmDescriptor) => Promise<void>;
   overriddenModel?: string;
   onHoverChange: (isHovered: boolean) => void;
   onDropdownVisibleChange: (isVisible: boolean) => void;
 }) {
   const { llmProviders } = useChatContext();
-  const llmOverrideManager = useLlmOverride(llmProviders);
+  const llmManager = useLlmManager(llmProviders);
 
   const [_, llmName] = getFinalLLM(llmProviders, selectedAssistant, null);
 
@@ -148,7 +148,7 @@ export default function RegenerateOption({
   );
 
   const currentModelName =
-    llmOverrideManager?.llmOverride.modelName ||
+    llmManager?.currentLlm.modelName ||
     (selectedAssistant
       ? selectedAssistant.llm_model_version_override || llmName
       : llmName);

web/src/app/chat/input/ChatInputBar.tsx

@@ -6,7 +6,7 @@ import { Persona } from "@/app/admin/assistants/interfaces";
 import LLMPopover from "./LLMPopover";
 import { InputPrompt } from "@/app/chat/interfaces";
 
-import { FilterManager, LlmOverrideManager } from "@/lib/hooks";
+import { FilterManager, LlmManager } from "@/lib/hooks";
 import { useChatContext } from "@/components/context/ChatContext";
 import { ChatFileType, FileDescriptor } from "../interfaces";
 import {
@@ -180,7 +180,7 @@ interface ChatInputBarProps {
   setMessage: (message: string) => void;
   stopGenerating: () => void;
   onSubmit: () => void;
-  llmOverrideManager: LlmOverrideManager;
+  llmManager: LlmManager;
   chatState: ChatState;
   alternativeAssistant: Persona | null;
   // assistants
@@ -225,7 +225,7 @@ export function ChatInputBar({
   availableSources,
   availableDocumentSets,
   availableTags,
-  llmOverrideManager,
+  llmManager,
   proSearchEnabled,
   setProSearchEnabled,
 }: ChatInputBarProps) {
@@ -781,7 +781,7 @@ export function ChatInputBar({
 
                 <LLMPopover
                   llmProviders={llmProviders}
-                  llmOverrideManager={llmOverrideManager}
+                  llmManager={llmManager}
                   requiresImageGeneration={false}
                   currentAssistant={selectedAssistant}
                 />
@@ -805,13 +805,12 @@ export function ChatInputBar({
                 )}
               </div>
               <div className="flex items-center my-auto">
-                {retrievalEnabled &&
-                  !settings?.settings.pro_search_disabled && (
-                    <AgenticToggle
-                      proSearchEnabled={proSearchEnabled}
-                      setProSearchEnabled={setProSearchEnabled}
-                    />
-                  )}
+                {retrievalEnabled && settings?.settings.pro_search_enabled && (
+                  <AgenticToggle
+                    proSearchEnabled={proSearchEnabled}
+                    setProSearchEnabled={setProSearchEnabled}
+                  />
+                )}
                 <button
                   id="onyx-chat-input-send-button"
                   className={`cursor-pointer ${

web/src/app/chat/input/LLMPopover.tsx

@@ -16,7 +16,7 @@ import {
   LLMProviderDescriptor,
 } from "@/app/admin/configuration/llm/interfaces";
 import { Persona } from "@/app/admin/assistants/interfaces";
-import { LlmOverrideManager } from "@/lib/hooks";
+import { LlmManager } from "@/lib/hooks";
 
 import {
   Tooltip,
@@ -31,21 +31,19 @@ import { useUser } from "@/components/user/UserProvider";
 
 interface LLMPopoverProps {
   llmProviders: LLMProviderDescriptor[];
-  llmOverrideManager: LlmOverrideManager;
+  llmManager: LlmManager;
   requiresImageGeneration?: boolean;
   currentAssistant?: Persona;
 }
 
 export default function LLMPopover({
   llmProviders,
-  llmOverrideManager,
+  llmManager,
   requiresImageGeneration,
   currentAssistant,
 }: LLMPopoverProps) {
   const [isOpen, setIsOpen] = useState(false);
   const { user } = useUser();
-  const { llmOverride, updateLLMOverride } = llmOverrideManager;
-  const currentLlm = llmOverride.modelName;
 
   const llmOptionsByProvider: {
     [provider: string]: {
@@ -93,19 +91,19 @@ export default function LLMPopover({
     : null;
 
   const [localTemperature, setLocalTemperature] = useState(
-    llmOverrideManager.temperature ?? 0.5
+    llmManager.temperature ?? 0.5
   );
 
   useEffect(() => {
-    setLocalTemperature(llmOverrideManager.temperature ?? 0.5);
-  }, [llmOverrideManager.temperature]);
+    setLocalTemperature(llmManager.temperature ?? 0.5);
+  }, [llmManager.temperature]);
 
   const handleTemperatureChange = (value: number[]) => {
     setLocalTemperature(value[0]);
   };
 
   const handleTemperatureChangeComplete = (value: number[]) => {
-    llmOverrideManager.updateTemperature(value[0]);
+    llmManager.updateTemperature(value[0]);
   };
 
   return (
@@ -120,15 +118,15 @@ export default function LLMPopover({
             toggle
             flexPriority="stiff"
             name={getDisplayNameForModel(
-              llmOverrideManager?.llmOverride.modelName ||
+              llmManager?.currentLlm.modelName ||
                 defaultModelDisplayName ||
                 "Models"
             )}
             Icon={getProviderIcon(
-              llmOverrideManager?.llmOverride.provider ||
+              llmManager?.currentLlm.provider ||
                 defaultProvider?.provider ||
                 "anthropic",
-              llmOverrideManager?.llmOverride.modelName ||
+              llmManager?.currentLlm.modelName ||
                 defaultProvider?.default_model_name ||
                 "claude-3-5-sonnet-20240620"
             )}
@@ -147,12 +145,12 @@ export default function LLMPopover({
                 <button
                   key={index}
                   className={`w-full flex items-center gap-x-2 px-3 py-2 text-sm text-left hover:bg-background-100 dark:hover:bg-neutral-800 transition-colors duration-150 ${
-                    currentLlm === name
+                    llmManager.currentLlm.modelName === name
                       ? "bg-background-100 dark:bg-neutral-900 text-text"
                       : "text-text-darker"
                   }`}
                   onClick={() => {
-                    updateLLMOverride(destructureValue(value));
+                    llmManager.updateCurrentLlm(destructureValue(value));
                     setIsOpen(false);
                   }}
                 >
@@ -172,7 +170,7 @@ export default function LLMPopover({
                       );
                     }
                   })()}
-                  {llmOverrideManager.imageFilesPresent &&
+                  {llmManager.imageFilesPresent &&
                     !checkLLMSupportsImageInput(name) && (
                       <TooltipProvider>
                         <Tooltip delayDuration={0}>
@@ -199,7 +197,7 @@ export default function LLMPopover({
             <div className="w-full px-3 py-2">
               <Slider
                 value={[localTemperature]}
-                max={llmOverrideManager.maxTemperature}
+                max={llmManager.maxTemperature}
                 min={0}
                 step={0.01}
                 onValueChange={handleTemperatureChange}

web/src/app/chat/interfaces.ts

@@ -155,6 +155,15 @@ export interface MessageResponseIDInfo {
   reserved_assistant_message_id: number;
 }
 
+export interface AgentMessageIDInfo {
+  level: number;
+  message_id: number;
+}
+
+export interface AgenticMessageResponseIDInfo {
+  agentic_message_ids: AgentMessageIDInfo[];
+}
+
 export interface DocumentsResponse {
   top_documents: OnyxDocument[];
   rephrased_query: string | null;

web/src/app/chat/lib.tsx

@@ -25,6 +25,7 @@ import {
   RetrievalType,
   StreamingError,
   ToolCallMetadata,
+  AgenticMessageResponseIDInfo,
 } from "./interfaces";
 import { Persona } from "../admin/assistants/interfaces";
 import { ReadonlyURLSearchParams } from "next/navigation";
@@ -64,7 +65,7 @@ export function getChatRetentionInfo(
   };
 }
 
-export async function updateModelOverrideForChatSession(
+export async function updateLlmOverrideForChatSession(
   chatSessionId: string,
   newAlternateModel: string
 ) {
@@ -154,7 +155,8 @@ export type PacketType =
   | AgentAnswerPiece
   | SubQuestionPiece
   | ExtendedToolResponse
-  | RefinedAnswerImprovement;
+  | RefinedAnswerImprovement
+  | AgenticMessageResponseIDInfo;
 
 export async function* sendMessage({
   regenerate,
@@ -234,7 +236,7 @@ export async function* sendMessage({
           }
         : null,
     use_existing_user_message: useExistingUserMessage,
-    use_agentic_search: useLanggraph,
+    use_agentic_search: useLanggraph ?? false,
   });
 
   const response = await fetch(`/api/chat/send-message`, {

web/src/app/chat/message/AgenticMessage.tsx

@@ -44,7 +44,7 @@ import { ValidSources } from "@/lib/types";
 import { useMouseTracking } from "./hooks";
 import { SettingsContext } from "@/components/settings/SettingsProvider";
 import RegenerateOption from "../RegenerateOption";
-import { LlmOverride } from "@/lib/hooks";
+import { LlmDescriptor } from "@/lib/hooks";
 import { ContinueGenerating } from "./ContinueMessage";
 import { MemoizedAnchor, MemoizedParagraph } from "./MemoizedTextComponents";
 import { extractCodeText, preprocessLaTeX } from "./codeUtils";
@@ -117,7 +117,7 @@ export const AgenticMessage = ({
   isComplete?: boolean;
   handleFeedback?: (feedbackType: FeedbackType) => void;
   overriddenModel?: string;
-  regenerate?: (modelOverRide: LlmOverride) => Promise<void>;
+  regenerate?: (modelOverRide: LlmDescriptor) => Promise<void>;
   setPresentingDocument?: (document: OnyxDocument) => void;
   toggleDocDisplay?: (agentic: boolean) => void;
   error?: string | null;

web/src/app/chat/message/Messages.tsx

@@ -58,7 +58,7 @@ import { useMouseTracking } from "./hooks";
 import { SettingsContext } from "@/components/settings/SettingsProvider";
 import GeneratingImageDisplay from "../tools/GeneratingImageDisplay";
 import RegenerateOption from "../RegenerateOption";
-import { LlmOverride } from "@/lib/hooks";
+import { LlmDescriptor } from "@/lib/hooks";
 import { ContinueGenerating } from "./ContinueMessage";
 import { MemoizedAnchor, MemoizedParagraph } from "./MemoizedTextComponents";
 import { extractCodeText, preprocessLaTeX } from "./codeUtils";
@@ -213,7 +213,7 @@ export const AIMessage = ({
   handleForceSearch?: () => void;
   retrievalDisabled?: boolean;
   overriddenModel?: string;
-  regenerate?: (modelOverRide: LlmOverride) => Promise<void>;
+  regenerate?: (modelOverRide: LlmDescriptor) => Promise<void>;
   setPresentingDocument: (document: OnyxDocument) => void;
   removePadding?: boolean;
 }) => {

web/src/app/chat/modal/ShareChatSessionModal.tsx

@@ -11,7 +11,7 @@ import { CopyButton } from "@/components/CopyButton";
 import { SEARCH_PARAM_NAMES } from "../searchParams";
 import { usePopup } from "@/components/admin/connectors/Popup";
 import { structureValue } from "@/lib/llm/utils";
-import { LlmOverride } from "@/lib/hooks";
+import { LlmDescriptor } from "@/lib/hooks";
 import { Separator } from "@/components/ui/separator";
 import { AdvancedOptionsToggle } from "@/components/AdvancedOptionsToggle";
 
@@ -38,7 +38,7 @@ async function generateShareLink(chatSessionId: string) {
 async function generateSeedLink(
   message?: string,
   assistantId?: number,
-  modelOverride?: LlmOverride
+  modelOverride?: LlmDescriptor
 ) {
   const baseUrl = `${window.location.protocol}//${window.location.host}`;
   const model = modelOverride
@@ -92,7 +92,7 @@ export function ShareChatSessionModal({
   onClose: () => void;
   message?: string;
   assistantId?: number;
-  modelOverride?: LlmOverride;
+  modelOverride?: LlmDescriptor;
 }) {
   const [shareLink, setShareLink] = useState<string>(
     existingSharedStatus === ChatSessionSharedStatus.Public

web/src/app/chat/modal/UserSettingsModal.tsx

@@ -1,6 +1,6 @@
 import { useContext, useEffect, useRef, useState } from "react";
 import { Modal } from "@/components/Modal";
-import { getDisplayNameForModel, LlmOverride } from "@/lib/hooks";
+import { getDisplayNameForModel, LlmDescriptor } from "@/lib/hooks";
 import { LLMProviderDescriptor } from "@/app/admin/configuration/llm/interfaces";
 
 import { destructureValue, structureValue } from "@/lib/llm/utils";
@@ -31,12 +31,12 @@ export function UserSettingsModal({
   setPopup,
   llmProviders,
   onClose,
-  setLlmOverride,
+  setCurrentLlm,
   defaultModel,
 }: {
   setPopup: (popupSpec: PopupSpec | null) => void;
   llmProviders: LLMProviderDescriptor[];
-  setLlmOverride?: (newOverride: LlmOverride) => void;
+  setCurrentLlm?: (newLlm: LlmDescriptor) => void;
   onClose: () => void;
   defaultModel: string | null;
 }) {
@@ -127,18 +127,14 @@ export function UserSettingsModal({
     );
   });
 
-  const llmOptions = Object.entries(llmOptionsByProvider).flatMap(
-    ([provider, options]) => [...options]
-  );
-
   const router = useRouter();
   const handleChangedefaultModel = async (defaultModel: string | null) => {
     try {
       const response = await setUserDefaultModel(defaultModel);
 
       if (response.ok) {
-        if (defaultModel && setLlmOverride) {
-          setLlmOverride(destructureValue(defaultModel));
+        if (defaultModel && setCurrentLlm) {
+          setCurrentLlm(destructureValue(defaultModel));
         }
         setPopup({
           message: "Default model updated successfully",

web/src/app/layout.tsx

@@ -21,11 +21,9 @@ import { fetchAssistantData } from "@/lib/chat/fetchAssistantdata";
 import { AppProvider } from "@/components/context/AppProvider";
 import { PHProvider } from "./providers";
 import { getCurrentUserSS } from "@/lib/userSS";
-import CardSection from "@/components/admin/CardSection";
 import { Suspense } from "react";
 import PostHogPageView from "./PostHogPageView";
 import Script from "next/script";
-import { LogoType } from "@/components/logo/Logo";
 import { Hanken_Grotesk } from "next/font/google";
 import { WebVitals } from "./web-vitals";
 import { ThemeProvider } from "next-themes";

web/src/components/settings/lib.ts

@@ -51,7 +51,7 @@ export async function fetchSettingsSS(): Promise<CombinedSettings | null> {
           notifications: [],
           needs_reindexing: false,
           anonymous_user_enabled: false,
-          pro_search_disabled: false,
+          pro_search_enabled: true,
           temperature_override_enabled: true,
         };
       } else {
@@ -95,8 +95,8 @@ export async function fetchSettingsSS(): Promise<CombinedSettings | null> {
       }
     }
 
-    if (enterpriseSettings && settings.pro_search_disabled == null) {
-      settings.pro_search_disabled = true;
+    if (settings.pro_search_enabled == null) {
+      settings.pro_search_enabled = true;
     }
 
     const webVersion = getWebVersion();

web/src/lib/hooks.ts

@@ -360,18 +360,18 @@ export const useUsers = ({ includeApiKeys }: UseUsersParams) => {
   };
 };
 
-export interface LlmOverride {
+export interface LlmDescriptor {
   name: string;
   provider: string;
   modelName: string;
 }
 
-export interface LlmOverrideManager {
-  llmOverride: LlmOverride;
-  updateLLMOverride: (newOverride: LlmOverride) => void;
+export interface LlmManager {
+  currentLlm: LlmDescriptor;
+  updateCurrentLlm: (newOverride: LlmDescriptor) => void;
   temperature: number;
   updateTemperature: (temperature: number) => void;
-  updateModelOverrideForChatSession: (chatSession?: ChatSession) => void;
+  updateModelOverrideBasedOnChatSession: (chatSession?: ChatSession) => void;
   imageFilesPresent: boolean;
   updateImageFilesPresent: (present: boolean) => void;
   liveAssistant: Persona | null;
@@ -400,7 +400,7 @@ Thus, the input should be
 
 Changes take place as
 - liveAssistant or currentChatSession changes (and the associated model override is set)
-- (uploadLLMOverride) User explicitly setting a model override (and we explicitly override and set the userSpecifiedOverride which we'll use in place of the user preferences unless overridden by an assistant)
+- (updateCurrentLlm) User explicitly setting a model override (and we explicitly override and set the userSpecifiedOverride which we'll use in place of the user preferences unless overridden by an assistant)
 
 If we have a live assistant, we should use that model override
 
@@ -419,55 +419,78 @@ This approach ensures that user preferences are maintained for existing chats wh
 providing appropriate defaults for new conversations based on the available tools.
 */
 
-export function useLlmOverride(
+export function useLlmManager(
   llmProviders: LLMProviderDescriptor[],
   currentChatSession?: ChatSession,
   liveAssistant?: Persona
-): LlmOverrideManager {
+): LlmManager {
   const { user } = useUser();
 
+  const [userHasManuallyOverriddenLLM, setUserHasManuallyOverriddenLLM] =
+    useState(false);
   const [chatSession, setChatSession] = useState<ChatSession | null>(null);
+  const [currentLlm, setCurrentLlm] = useState<LlmDescriptor>({
+    name: "",
+    provider: "",
+    modelName: "",
+  });
 
-  const llmOverrideUpdate = () => {
-    if (liveAssistant?.llm_model_version_override) {
-      setLlmOverride(
-        getValidLlmOverride(liveAssistant.llm_model_version_override)
-      );
-    } else if (currentChatSession?.current_alternate_model) {
-      setLlmOverride(
-        getValidLlmOverride(currentChatSession.current_alternate_model)
-      );
-    } else if (user?.preferences?.default_model) {
-      setLlmOverride(getValidLlmOverride(user.preferences.default_model));
-      return;
-    } else {
-      const defaultProvider = llmProviders.find(
-        (provider) => provider.is_default_provider
-      );
+  const llmUpdate = () => {
+    /* Should be called when the live assistant or current chat session changes */
 
-      if (defaultProvider) {
-        setLlmOverride({
-          name: defaultProvider.name,
-          provider: defaultProvider.provider,
-          modelName: defaultProvider.default_model_name,
-        });
+    // separate function so we can `return` to break out
+    const _llmUpdate = () => {
+      // if the user has overridden in this session and just switched to a brand
+      // new session, use their manually specified model
+      if (userHasManuallyOverriddenLLM && !currentChatSession) {
+        return;
       }
-    }
+
+      if (currentChatSession?.current_alternate_model) {
+        setCurrentLlm(
+          getValidLlmDescriptor(currentChatSession.current_alternate_model)
+        );
+      } else if (liveAssistant?.llm_model_version_override) {
+        setCurrentLlm(
+          getValidLlmDescriptor(liveAssistant.llm_model_version_override)
+        );
+      } else if (userHasManuallyOverriddenLLM) {
+        // if the user has an override and there's nothing special about the
+        // current chat session, use the override
+        return;
+      } else if (user?.preferences?.default_model) {
+        setCurrentLlm(getValidLlmDescriptor(user.preferences.default_model));
+      } else {
+        const defaultProvider = llmProviders.find(
+          (provider) => provider.is_default_provider
+        );
+
+        if (defaultProvider) {
+          setCurrentLlm({
+            name: defaultProvider.name,
+            provider: defaultProvider.provider,
+            modelName: defaultProvider.default_model_name,
+          });
+        }
+      }
+    };
+
+    _llmUpdate();
     setChatSession(currentChatSession || null);
   };
 
-  const getValidLlmOverride = (
-    overrideModel: string | null | undefined
-  ): LlmOverride => {
-    if (overrideModel) {
-      const model = destructureValue(overrideModel);
+  const getValidLlmDescriptor = (
+    modelName: string | null | undefined
+  ): LlmDescriptor => {
+    if (modelName) {
+      const model = destructureValue(modelName);
       if (!(model.modelName && model.modelName.length > 0)) {
         const provider = llmProviders.find((p) =>
-          p.model_names.includes(overrideModel)
+          p.model_names.includes(modelName)
         );
         if (provider) {
           return {
-            modelName: overrideModel,
+            modelName: modelName,
             name: provider.name,
             provider: provider.provider,
           };
@@ -491,38 +514,32 @@ export function useLlmOverride(
     setImageFilesPresent(present);
   };
 
-  const [llmOverride, setLlmOverride] = useState<LlmOverride>({
-    name: "",
-    provider: "",
-    modelName: "",
-  });
-
-  // Manually set the override
-  const updateLLMOverride = (newOverride: LlmOverride) => {
+  // Manually set the LLM
+  const updateCurrentLlm = (newLlm: LlmDescriptor) => {
     const provider =
-      newOverride.provider ||
-      findProviderForModel(llmProviders, newOverride.modelName);
+      newLlm.provider || findProviderForModel(llmProviders, newLlm.modelName);
     const structuredValue = structureValue(
-      newOverride.name,
+      newLlm.name,
       provider,
-      newOverride.modelName
+      newLlm.modelName
     );
-    setLlmOverride(getValidLlmOverride(structuredValue));
+    setCurrentLlm(getValidLlmDescriptor(structuredValue));
+    setUserHasManuallyOverriddenLLM(true);
   };
 
-  const updateModelOverrideForChatSession = (chatSession?: ChatSession) => {
+  const updateModelOverrideBasedOnChatSession = (chatSession?: ChatSession) => {
     if (chatSession && chatSession.current_alternate_model?.length > 0) {
-      setLlmOverride(getValidLlmOverride(chatSession.current_alternate_model));
+      setCurrentLlm(getValidLlmDescriptor(chatSession.current_alternate_model));
     }
   };
 
   const [temperature, setTemperature] = useState<number>(() => {
-    llmOverrideUpdate();
+    llmUpdate();
 
     if (currentChatSession?.current_temperature_override != null) {
       return Math.min(
         currentChatSession.current_temperature_override,
-        isAnthropic(llmOverride.provider, llmOverride.modelName) ? 1.0 : 2.0
+        isAnthropic(currentLlm.provider, currentLlm.modelName) ? 1.0 : 2.0
       );
     } else if (
       liveAssistant?.tools.some((tool) => tool.name === SEARCH_TOOL_ID)
@@ -533,22 +550,23 @@ export function useLlmOverride(
   });
 
   const maxTemperature = useMemo(() => {
-    return isAnthropic(llmOverride.provider, llmOverride.modelName) ? 1.0 : 2.0;
-  }, [llmOverride]);
+    return isAnthropic(currentLlm.provider, currentLlm.modelName) ? 1.0 : 2.0;
+  }, [currentLlm]);
 
   useEffect(() => {
-    if (isAnthropic(llmOverride.provider, llmOverride.modelName)) {
+    if (isAnthropic(currentLlm.provider, currentLlm.modelName)) {
       const newTemperature = Math.min(temperature, 1.0);
       setTemperature(newTemperature);
       if (chatSession?.id) {
         updateTemperatureOverrideForChatSession(chatSession.id, newTemperature);
       }
     }
-  }, [llmOverride]);
+  }, [currentLlm]);
 
   useEffect(() => {
+    llmUpdate();
+
     if (!chatSession && currentChatSession) {
-      setChatSession(currentChatSession || null);
       if (temperature) {
         updateTemperatureOverrideForChatSession(
           currentChatSession.id,
@@ -570,7 +588,7 @@ export function useLlmOverride(
   }, [liveAssistant, currentChatSession]);
 
   const updateTemperature = (temperature: number) => {
-    if (isAnthropic(llmOverride.provider, llmOverride.modelName)) {
+    if (isAnthropic(currentLlm.provider, currentLlm.modelName)) {
       setTemperature((prevTemp) => Math.min(temperature, 1.0));
     } else {
       setTemperature(temperature);
@@ -581,9 +599,9 @@ export function useLlmOverride(
   };
 
   return {
-    updateModelOverrideForChatSession,
-    llmOverride,
-    updateLLMOverride,
+    updateModelOverrideBasedOnChatSession,
+    currentLlm,
+    updateCurrentLlm,
     temperature,
     updateTemperature,
     imageFilesPresent,

web/src/lib/llm/utils.ts

@@ -1,11 +1,11 @@
 import { Persona } from "@/app/admin/assistants/interfaces";
 import { LLMProviderDescriptor } from "@/app/admin/configuration/llm/interfaces";
-import { LlmOverride } from "@/lib/hooks";
+import { LlmDescriptor } from "@/lib/hooks";
 
 export function getFinalLLM(
   llmProviders: LLMProviderDescriptor[],
   persona: Persona | null,
-  llmOverride: LlmOverride | null
+  currentLlm: LlmDescriptor | null
 ): [string, string] {
   const defaultProvider = llmProviders.find(
     (llmProvider) => llmProvider.is_default_provider
@@ -26,9 +26,9 @@ export function getFinalLLM(
     model = persona.llm_model_version_override || model;
   }
 
-  if (llmOverride) {
-    provider = llmOverride.provider || provider;
-    model = llmOverride.modelName || model;
+  if (currentLlm) {
+    provider = currentLlm.provider || provider;
+    model = currentLlm.modelName || model;
   }
 
   return [provider, model];
@@ -37,7 +37,7 @@ export function getFinalLLM(
 export function getLLMProviderOverrideForPersona(
   liveAssistant: Persona,
   llmProviders: LLMProviderDescriptor[]
-): LlmOverride | null {
+): LlmDescriptor | null {
   const overrideProvider = liveAssistant.llm_model_provider_override;
   const overrideModel = liveAssistant.llm_model_version_override;
 
@@ -135,7 +135,7 @@ export const structureValue = (
   return `${name}__${provider}__${modelName}`;
 };
 
-export const destructureValue = (value: string): LlmOverride => {
+export const destructureValue = (value: string): LlmDescriptor => {
   const [displayName, provider, modelName] = value.split("__");
   return {
     name: displayName,

web/src/lib/users/UserSettings.tsx

@@ -1,5 +1,3 @@
-import { LlmOverride } from "../hooks";
-
 export async function setUserDefaultModel(
   model: string | null
 ): Promise<Response> {