feat: show RAG query results as citations

backend/apps/rag/utils.py

@@ -320,11 +320,19 @@ def rag_messages(
         extracted_collections.extend(collection)
 
     context_string = ""
+    citations = []
     for context in relevant_contexts:
         try:
             if "documents" in context:
                 items = [item for item in context["documents"][0] if item is not None]
                 context_string += "\n\n".join(items)
+                if "metadatas" in context:
+                    citations.append(
+                        {
+                            "document": context["documents"][0],
+                            "metadata": context["metadatas"][0],
+                        }
+                    )
         except Exception as e:
             log.exception(e)
     context_string = context_string.strip()
@@ -355,7 +363,7 @@ def rag_messages(
 
     messages[last_user_message_idx] = new_user_message
 
-    return messages
+    return messages, citations
 
 
 def get_model_path(model: str, update_model: bool = False):

backend/main.py

@@ -15,7 +15,7 @@ from fastapi.middleware.wsgi import WSGIMiddleware
 from fastapi.middleware.cors import CORSMiddleware
 from starlette.exceptions import HTTPException as StarletteHTTPException
 from starlette.middleware.base import BaseHTTPMiddleware
-
+from starlette.responses import StreamingResponse
 
 from apps.ollama.main import app as ollama_app
 from apps.openai.main import app as openai_app
@@ -102,6 +102,8 @@ origins = ["*"]
 
 class RAGMiddleware(BaseHTTPMiddleware):
     async def dispatch(self, request: Request, call_next):
+        return_citations = False
+
         if request.method == "POST" and (
             "/api/chat" in request.url.path or "/chat/completions" in request.url.path
         ):
@@ -114,11 +116,15 @@ class RAGMiddleware(BaseHTTPMiddleware):
             # Parse string to JSON
             data = json.loads(body_str) if body_str else {}
 
+            return_citations = data.get("citations", False)
+            if "citations" in data:
+                del data["citations"]
+
             # Example: Add a new key-value pair or modify existing ones
             # data["modified"] = True  # Example modification
             if "docs" in data:
                 data = {**data}
-                data["messages"] = rag_messages(
+                data["messages"], citations = rag_messages(
                     docs=data["docs"],
                     messages=data["messages"],
                     template=rag_app.state.RAG_TEMPLATE,
@@ -130,7 +136,9 @@ class RAGMiddleware(BaseHTTPMiddleware):
                 )
                 del data["docs"]
 
-                log.debug(f"data['messages']: {data['messages']}")
+                log.debug(
+                    f"data['messages']: {data['messages']}, citations: {citations}"
+                )
 
             modified_body_bytes = json.dumps(data).encode("utf-8")
 
@@ -148,11 +156,36 @@ class RAGMiddleware(BaseHTTPMiddleware):
             ]
 
         response = await call_next(request)
+
+        if return_citations:
+            # Inject the citations into the response
+            if isinstance(response, StreamingResponse):
+                # If it's a streaming response, inject it as SSE event or NDJSON line
+                content_type = response.headers.get("Content-Type")
+                if "text/event-stream" in content_type:
+                    return StreamingResponse(
+                        self.openai_stream_wrapper(response.body_iterator, citations),
+                    )
+                if "application/x-ndjson" in content_type:
+                    return StreamingResponse(
+                        self.ollama_stream_wrapper(response.body_iterator, citations),
+                    )
+
         return response
 
     async def _receive(self, body: bytes):
         return {"type": "http.request", "body": body, "more_body": False}
 
+    async def openai_stream_wrapper(self, original_generator, citations):
+        yield f"data: {json.dumps({'citations': citations})}\n\n"
+        async for data in original_generator:
+            yield data
+
+    async def ollama_stream_wrapper(self, original_generator, citations):
+        yield f"{json.dumps({'citations': citations})}\n"
+        async for data in original_generator:
+            yield data
+
 
 app.add_middleware(RAGMiddleware)