Update api/utils.py

api/utils.py

@@ -141,6 +141,7 @@ async def process_streaming_response(request: ChatRequest):
         "imageGenerationMode": False,  # Added this line
     }
 
+    response_chunks = []
     async with httpx.AsyncClient() as client:
         try:
             async with client.stream(
@@ -166,15 +167,26 @@ async def process_streaming_response(request: ChatRequest):
                             if not content:
                                 continue  # Skip if content is empty after removal
                         cleaned_content = strip_model_prefix(content, model_prefix)
-                        
-                        # Append advertisement text
-                        if ADVERTISEMENT_TEXT:
-                            cleaned_content += "\n\n" + ADVERTISEMENT_TEXT
+                        response_chunks.append(create_chat_completion_data(cleaned_content, request.model, timestamp))
+                
+                # At the very end, add the advertisement text once
+                if ADVERTISEMENT_TEXT:
+                    # If there are chunks already, update the last one with the advertisement text.
+                    if response_chunks:
+                        last_chunk = response_chunks[-1]
+                        last_chunk["choices"][0]["delta"]["content"] += "\n\n" + ADVERTISEMENT_TEXT
+                    else:
+                        # If no chunks are returned, just create an empty response with the ad.
+                        response_chunks.append(create_chat_completion_data(ADVERTISEMENT_TEXT, request.model, timestamp))
 
-                        yield f"data: {json.dumps(create_chat_completion_data(cleaned_content, request.model, timestamp))}\n\n"
+                # Finalize the response
+                response_chunks.append(create_chat_completion_data('', request.model, timestamp, 'stop'))
+                response_chunks.append({"data": "[DONE]\n\n"})
+                
+                # Yield each chunk as part of the stream response
+                for chunk in response_chunks:
+                    yield f"data: {json.dumps(chunk)}\n\n"
 
-                yield f"data: {json.dumps(create_chat_completion_data('', request.model, timestamp, 'stop'))}\n\n"
-                yield "data: [DONE]\n\n"
         except httpx.HTTPStatusError as e:
             logger.error(f"HTTP error occurred for Request ID {request_id}: {e}")
             raise HTTPException(status_code=e.response.status_code, detail=str(e))
@@ -283,7 +295,7 @@ async def process_non_streaming_response(request: ChatRequest):
 
     cleaned_full_response = strip_model_prefix(full_response, model_prefix)
 
-    # Append the advertisement text if available
+    # Append the advertisement text only once at the end
     if ADVERTISEMENT_TEXT:
         cleaned_full_response += "\n\n" + ADVERTISEMENT_TEXT