Add API to Ingest OpenAI Assistant and Migrate to Support Multiple Vector Store IDs

backend/app/alembic/versions/f2589428c1d0_change_vector_store_id_to_vector_store_ids.py

@@ -0,0 +1,58 @@
+"""Change vector_store_id to vector_store_ids in openai_assistant table
+
+Revision ID: f2589428c1d0
+Revises: 3389c67fdcb4
+Create Date: 2025-07-10 11:18:21.223114
+
+"""
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision = "f2589428c1d0"
+down_revision = "3389c67fdcb4"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.add_column(
+        "openai_assistant",
+        sa.Column("vector_store_ids", postgresql.ARRAY(sa.String()), nullable=True),
+    )
+
+    op.execute(
+        """
+        UPDATE openai_assistant
+        SET vector_store_ids = ARRAY[vector_store_id]
+        WHERE vector_store_id IS NOT NULL
+    """
+    )
+
+    op.drop_column("openai_assistant", "vector_store_id")
+    # ### end Alembic commands ###
+
+
+def downgrade():
+    # Add back the single vector_store_id column as nullable for safe data migration
+    op.add_column(
+        "openai_assistant",
+        sa.Column(
+            "vector_store_id",
+            sa.VARCHAR(length=255),
+            autoincrement=False,
+            nullable=True,  # Allow nulls temporarily for safe migration
+        ),
+    )
+
+    op.execute(
+        """
+        UPDATE openai_assistant
+        SET vector_store_id = vector_store_ids[1]
+        WHERE vector_store_ids IS NOT NULL AND array_length(vector_store_ids, 1) > 0
+        """
+    )
+
+    op.drop_column("openai_assistant", "vector_store_ids")

backend/app/api/main.py

@@ -2,6 +2,7 @@
 
 from app.api.routes import (
     api_keys,
+    assistants,
     collections,
     documents,
     login,
@@ -20,6 +21,7 @@
 
 api_router = APIRouter()
 api_router.include_router(api_keys.router)
+api_router.include_router(assistants.router)
 api_router.include_router(collections.router)
 api_router.include_router(credentials.router)
 api_router.include_router(documents.router)

backend/app/api/routes/assistants.py

@@ -0,0 +1,43 @@
+from typing import Annotated
+
+from fastapi import APIRouter, Depends, Path
+from sqlmodel import Session
+
+from app.api.deps import get_db, get_current_user_org_project
+from app.crud import (
+    fetch_assistant_from_openai,
+    sync_assistant,
+)
+from app.models import UserProjectOrg
+from app.utils import APIResponse, get_openai_client
+
+router = APIRouter(prefix="/assistant", tags=["Assistants"])
+
+
+@router.post(
+    "/{assistant_id}/ingest",
+    response_model=APIResponse,
+    status_code=201,
+)
+def ingest_assistant_route(
+    assistant_id: Annotated[str, Path(description="The ID of the assistant to ingest")],
+    session: Session = Depends(get_db),
+    current_user: UserProjectOrg = Depends(get_current_user_org_project),
+):
+    """
+    Ingest an assistant from OpenAI and store it in the platform.
+    """
+
+    client = get_openai_client(
+        session, current_user.organization_id, current_user.project_id
+    )
+
+    openai_assistant = fetch_assistant_from_openai(assistant_id, client)
+    assistant = sync_assistant(
+        session=session,
+        organization_id=current_user.organization_id,
+        project_id=current_user.project_id,
+        openai_assistant=openai_assistant,
+    )
+
+    return APIResponse.success_response(assistant)

backend/app/api/routes/responses.py

@@ -125,11 +125,11 @@ def process_response(
             "input": [{"role": "user", "content": request.question}],
         }
 
-        if assistant.vector_store_id:
+        if assistant.vector_store_ids:
             params["tools"] = [
                 {
                     "type": "file_search",
-                    "vector_store_ids": [assistant.vector_store_id],
+                    "vector_store_ids": assistant.vector_store_ids,
                     "max_num_results": assistant.max_num_results,
                 }
             ]

backend/app/crud/__init__.py

@@ -5,6 +5,7 @@
     update_user,
 )
 from .collection import CollectionCrud
+
 from .document import DocumentCrud
 from .document_collection import DocumentCollectionCrud
 
@@ -43,4 +44,8 @@
 
 from .thread_results import upsert_thread_result, get_thread_result
 
-from .assistants import get_assistant_by_id
+from .assistants import (
+    get_assistant_by_id,
+    fetch_assistant_from_openai,
+    sync_assistant,
+)

backend/app/crud/assistants.py

@@ -1,8 +1,17 @@
-from typing import Optional, List, Tuple
-from sqlmodel import Session, select, and_
+import logging
+
+from typing import Optional
+
+import openai
+from fastapi import HTTPException
+from openai import OpenAI
+from openai.types.beta import Assistant as OpenAIAssistant
+from sqlmodel import Session, and_, select
 
-from app.core.util import now
 from app.models import Assistant
+from app.utils import mask_string
+
+logger = logging.getLogger(__name__)
 
 
 def get_assistant_by_id(
@@ -16,3 +25,89 @@ def get_assistant_by_id(
         )
     )
     return session.exec(statement).first()
+
+
+def fetch_assistant_from_openai(assistant_id: str, client: OpenAI) -> OpenAIAssistant:
+    """
+    Fetch an assistant from OpenAI.
+    Returns OpenAI Assistant model.
+    """
+
+    try:
+        assistant = client.beta.assistants.retrieve(assistant_id=assistant_id)
+        return assistant
+    except openai.NotFoundError as e:
+        logger.error(
+            f"[fetch_assistant_from_openai] Assistant not found: {mask_string(assistant_id)} | {e}"
+        )
+        raise HTTPException(status_code=404, detail="Assistant not found in OpenAI.")
+    except openai.OpenAIError as e:
+        logger.error(
+            f"[fetch_assistant_from_openai] OpenAI API error while retrieving assistant {mask_string(assistant_id)}: {e}"
+        )
+        raise HTTPException(status_code=502, detail=f"OpenAI API error: {e}")
+
+
+def sync_assistant(
+    session: Session,
+    organization_id: int,
+    project_id: int,
+    openai_assistant: OpenAIAssistant,
+) -> Assistant:
+    """
+    Insert an assistant into the database by converting OpenAI Assistant to local Assistant model.
+    """
+    assistant_id = openai_assistant.id
+
+    existing_assistant = get_assistant_by_id(session, assistant_id, organization_id)
+    if existing_assistant:
+        logger.info(
+            f"[sync_assistant] Assistant with ID {mask_string(assistant_id)} already exists in the database."
+        )
+        raise HTTPException(
+            status_code=409,
+            detail=f"Assistant with ID {assistant_id} already exists.",
+        )
+
+    if not openai_assistant.instructions:
+        raise HTTPException(
+            status_code=400,
+            detail="Assistant has no instruction.",
+        )
+
+    vector_store_ids = []
+    if openai_assistant.tool_resources and hasattr(
+        openai_assistant.tool_resources, "file_search"
+    ):
+        file_search = openai_assistant.tool_resources.file_search
+        if file_search and hasattr(file_search, "vector_store_ids"):
+            vector_store_ids = file_search.vector_store_ids or []
+
+    max_num_results = 20
+    for tool in openai_assistant.tools or []:
+        if tool.type == "file_search":
+            file_search = getattr(tool, "file_search", None)
+            if file_search and hasattr(file_search, "max_num_results"):
+                max_num_results = file_search.max_num_results
+            break
+
+    db_assistant = Assistant(
+        assistant_id=openai_assistant.id,
+        name=openai_assistant.name or openai_assistant.id,
+        instructions=openai_assistant.instructions,
+        model=openai_assistant.model,
+        vector_store_ids=vector_store_ids,
+        temperature=openai_assistant.temperature or 0.1,
+        max_num_results=max_num_results,
+        project_id=project_id,
+        organization_id=organization_id,
+    )
+
+    session.add(db_assistant)
+    session.commit()
+    session.refresh(db_assistant)
+
+    logger.info(
+        f"[sync_assistant] Successfully ingested assistant with ID {mask_string(assistant_id)}."
+    )
+    return db_assistant

backend/app/models/assistants.py

@@ -1,6 +1,8 @@
 from datetime import datetime
 from typing import Optional, List
 from sqlmodel import Field, Relationship, SQLModel
+from sqlalchemy import Column, String
+from sqlalchemy.dialects.postgresql import ARRAY
 
 from app.core.util import now
 
@@ -10,7 +12,9 @@ class AssistantBase(SQLModel):
     name: str
     instructions: str
     model: str
-    vector_store_id: str
+    vector_store_ids: List[str] = Field(
+        default_factory=list, sa_column=Column(ARRAY(String))
+    )
     temperature: float = 0.1
     max_num_results: int = 20
     project_id: int = Field(foreign_key="project.id")

backend/app/seed_data/seed_data.json

@@ -57,23 +57,42 @@
     {
       "is_active": true,
       "provider": "openai",
-      "credential": "{\"openai\": {\"api_key\": \"sk-proj-YxK21qI3i5SCxN\"}}",
+      "credential": "{\"api_key\": \"sk-proj-GlificI3i5SCxN\"}",
       "project_name": "Glific",
       "organization_name": "Project Tech4dev",
       "deleted_at": null
+    },
+    {
+      "is_active": true,
+      "provider": "openai",
+      "credential": "{\"api_key\": \"sk-proj-DalgoI3i5SCxN\"}",
+      "project_name": "Dalgo",
+      "organization_name": "Project Tech4dev",
+      "deleted_at": null
     }
   ],
   "assistants": [
     {
-      "assistant_id": "assistant_123",
-      "name": "Test Assistant",
+      "assistant_id": "assistant_glific",
+      "name": "Test Assistant Glific",
       "instructions": "Test instructions",
       "model": "gpt-4o",
-      "vector_store_id": "vs_123",
+      "vector_store_ids": ["vs_glific"],
       "temperature": 0.1,
       "max_num_results": 20,
       "project_name": "Glific",
       "organization_name": "Project Tech4dev"
+    },
+    {
+      "assistant_id": "assistant_dalgo",
+      "name": "Test Assistant Dalgo",
+      "instructions": "Test instructions",
+      "model": "gpt-4o",
+      "vector_store_ids": ["vs_dalgo"],
+      "temperature": 0.1,
+      "max_num_results": 20,
+      "project_name": "Dalgo",
+      "organization_name": "Project Tech4dev"
     }
   ]
 }

backend/app/seed_data/seed_data.py

@@ -57,7 +57,7 @@ class AssistantData(BaseModel):
     name: str
     instructions: str
     model: str
-    vector_store_id: str
+    vector_store_ids: list[str]
     temperature: float
     max_num_results: int
     project_name: str
@@ -261,7 +261,7 @@ def create_assistant(session: Session, assistant_data_raw: dict) -> Assistant:
             name=assistant_data.name,
             instructions=assistant_data.instructions,
             model=assistant_data.model,
-            vector_store_id=assistant_data.vector_store_id,
+            vector_store_ids=assistant_data.vector_store_ids,
             temperature=assistant_data.temperature,
             max_num_results=assistant_data.max_num_results,
             organization_id=organization.id,

backend/app/tests/api/routes/test_assistants.py

@@ -0,0 +1,31 @@
+import pytest
+from fastapi.testclient import TestClient
+from unittest.mock import patch
+from app.tests.utils.openai import mock_openai_assistant
+
+
+@pytest.fixture
+def normal_user_api_key_header():
+    return {"X-API-KEY": "ApiKey Px8y47B6roJHin1lWLkR88eiDrFdXSJRZmFQazzai8j9"}
+
+
+@patch("app.api.routes.assistants.fetch_assistant_from_openai")
+def test_ingest_assistant_success(
+    mock_fetch_assistant,
+    client: TestClient,
+    normal_user_api_key_header: str,
+):
+    """Test successful assistant ingestion from OpenAI."""
+    mock_assistant = mock_openai_assistant()
+
+    mock_fetch_assistant.return_value = mock_assistant
+
+    response = client.post(
+        f"/api/v1/assistant/{mock_assistant.id}/ingest",
+        headers=normal_user_api_key_header,
+    )
+
+    assert response.status_code == 201
+    response_json = response.json()
+    assert response_json["success"] is True
+    assert response_json["data"]["assistant_id"] == mock_assistant.id

backend/app/tests/api/routes/test_responses.py

@@ -49,7 +49,7 @@ def test_responses_endpoint_success(
 
     headers = {"X-API-KEY": original_api_key}
     request_data = {
-        "assistant_id": "assistant_123",
+        "assistant_id": "assistant_glific",
         "question": "What is Glific?",
         "callback_url": "http://example.com/callback",
     }
@@ -80,7 +80,7 @@ def test_responses_endpoint_without_vector_store(
     mock_assistant.model = "gpt-4"
     mock_assistant.instructions = "Test instructions"
     mock_assistant.temperature = 0.1
-    mock_assistant.vector_store_id = None  # No vector store configured
+    mock_assistant.vector_store_ids = []  # No vector store configured
     mock_get_assistant.return_value = mock_assistant
 
     # Setup mock OpenAI client