Update api/utils.py

api/utils.py

@@ -63,6 +63,12 @@ def message_to_dict(message, model_prefix: Optional[str] = None):
     else:
         return {"role": message.role, "content": message.content}
 
+def strip_model_prefix(content: str, model_prefix: str) -> str:
+    """Remove the model prefix from the response content if present."""
+    if content.startswith(model_prefix):
+        return content[len(model_prefix):].strip()
+    return content
+
 async def process_streaming_response(request: ChatRequest):
     agent_mode = AGENT_MODE.get(request.model, {})
     trending_agent_mode = TRENDING_AGENT_MODE.get(request.model, {})
@@ -112,7 +118,9 @@ async def process_streaming_response(request: ChatRequest):
                         content = line
                         if content.startswith("$@$v=undefined-rv1$@$"):
                             content = content[21:]
-                        yield f"data: {json.dumps(create_chat_completion_data(content, request.model, timestamp))}\n\n"
+                        # Strip the model prefix from the response content
+                        cleaned_content = strip_model_prefix(content, model_prefix)
+                        yield f"data: {json.dumps(create_chat_completion_data(cleaned_content, request.model, timestamp))}\n\n"
 
                 yield f"data: {json.dumps(create_chat_completion_data('', request.model, timestamp, 'stop'))}\n\n"
                 yield "data: [DONE]\n\n"
@@ -173,6 +181,9 @@ async def process_non_streaming_response(request: ChatRequest):
     if full_response.startswith("$@$v=undefined-rv1$@$"):
         full_response = full_response[21:]
 
+    # Strip the model prefix from the full response
+    cleaned_full_response = strip_model_prefix(full_response, model_prefix)
+
     return {
         "id": f"chatcmpl-{uuid.uuid4()}",
         "object": "chat.completion",
@@ -181,7 +192,7 @@ async def process_non_streaming_response(request: ChatRequest):
         "choices": [
             {
                 "index": 0,
-                "message": {"role": "assistant", "content": full_response},
+                "message": {"role": "assistant", "content": cleaned_full_response},
                 "finish_reason": "stop",
             }
         ],