Update app.py

app.py

@@ -2,6 +2,7 @@ import os
 import shutil
 from fastapi.responses import FileResponse
 import asyncio
+import json
 from fastapi import FastAPI, UploadFile, File, Form, HTTPException, BackgroundTasks
 from fastapi.responses import StreamingResponse
 from pydantic import BaseModel
@@ -65,46 +66,53 @@ async def upload_document(
         raise HTTPException(status_code=500, detail=str(e))
 
 
-import json
+
 
 # ... (keep existing imports) ...
 
 @app.post("/chat")
 async def chat_endpoint(request: ChatRequest):
     """
-    Chat endpoint that returns a STREAMING response in JSON-SSE format.
+    Robust Streaming Endpoint that logs events to console.
     """
-    
     config = {"configurable": {"thread_id": request.thread_id}}
     
     inputs = {
         "query": request.query,
         "RAG": request.use_rag,
         "web_search": request.use_web,
-        "model_name": request.model_name, # Ensure you passed this if you added multi-llm support
+        "model_name": request.model_name,
         "context": [],
         "metadata": [],
         "web_context": "",
     }
 
     async def event_generator():
-        print(f"--- Starting stream for {request.thread_id} ---") # Log to HF console
+        print(f"--- 🚀 Starting stream for {request.thread_id} ---")
         
+        # Use 'v2' if you are on the latest LangGraph, but 'v1' is safer for compatibility
         async for event in rag_app.astream_events(inputs, config=config, version="v1"):
-            kind = event["event"]
             
-            # Check for LLM token events
-            if kind == "on_chat_model_stream":
-                content = event["data"]["chunk"].content
+            # [DEBUG] Print the event type to your Hugging Face Logs
+            # This will show us if the events are firing but named differently
+            event_type = event.get("event")
+            
+            # Logic: We don't care about the event name. 
+            # We only care: "Does this event have a chunk with text?"
+            data = event.get("data", {})
+            chunk = data.get("chunk")
+            
+            # Check if chunk exists and has .content attribute (standard LangChain message chunk)
+            if chunk and hasattr(chunk, "content") and chunk.content:
+                content = chunk.content
                 
-                if content:
-                    # 1. Wrap content in a JSON object (Safer than raw text)
-                    chunk_data = json.dumps({"content": content})
-                    
-                    # 2. Yield the SSE frame
-                    yield f"data: {chunk_data}\n\n"
-
-        # 3. Send a [DONE] signal so the frontend knows to stop
+                # Filter out empty strings or weird artifacts
+                if content.strip() != "":
+                    # JSON encode the content
+                    chunk_json = json.dumps({"content": content})
+                    yield f"data: {chunk_json}\n\n"
+
+        # End of stream
         yield "data: [DONE]\n\n"
 
     return StreamingResponse(
@@ -114,7 +122,7 @@ async def chat_endpoint(request: ChatRequest):
             "Cache-Control": "no-cache",
             "Connection": "keep-alive",
             "Content-Type": "text/event-stream",
-            "X-Accel-Buffering": "no", # CRITICAL: Disables HF/Nginx buffering
+            "X-Accel-Buffering": "no",
         },
     )