Update main.py

main.py

@@ -18,7 +18,7 @@ API_KEY = "sk-XzS5hhsa3vpIcRLz3prQirBQXOx2hPydPzSpzdRcE1YddnNm"
 BACKEND_MODEL = "pixtral-large-latest"
 
 # Load model -> system prompt mappings
-with open("model_map.json", "r") as f:
+with open("model_map.json", "r", encoding="utf-8") as f:
     MODEL_PROMPTS = json.load(f)
 
 # Request schema
@@ -39,17 +39,12 @@ class ChatRequest(BaseModel):
 
 # Construct payload with enforced system prompt
 def build_payload(chat: ChatRequest):
-    # Use internal system prompt
     system_prompt = MODEL_PROMPTS.get(chat.model, "You are a helpful assistant.")
-
-    # Remove any user-provided system messages
+    # Strip user system messages
     filtered_messages = [msg for msg in chat.messages if msg.role != "system"]
-
-    # Insert enforced system prompt
     payload_messages = [{"role": "system", "content": system_prompt}] + [
         {"role": msg.role, "content": msg.content} for msg in filtered_messages
     ]
-
     return {
         "model": BACKEND_MODEL,
         "messages": payload_messages,
@@ -62,7 +57,7 @@ def build_payload(chat: ChatRequest):
         "frequency_penalty": chat.frequency_penalty
     }
 
-# Streaming response handler
+# Properly streamed UTF-8 chunks with model rewrite
 def stream_generator(requested_model: str, payload: dict, headers: dict):
     with requests.post(API_URL, headers=headers, json=payload, stream=True) as r:
         for line in r.iter_lines(decode_unicode=True):
@@ -77,13 +72,13 @@ def stream_generator(requested_model: str, payload: dict, headers: dict):
                     json_obj = json.loads(content)
                     if json_obj.get("model") == BACKEND_MODEL:
                         json_obj["model"] = requested_model
-                    yield f"data: {json.dumps(json_obj)}\n\n"
+                    yield f"data: {json.dumps(json_obj, ensure_ascii=False)}\n\n"
                 except json.JSONDecodeError:
                     logger.warning("Invalid JSON in stream chunk: %s", content)
             else:
                 logger.debug("Non-data stream line skipped: %s", line)
 
-# Proxy endpoint
+# Main endpoint
 @app.post("/v1/chat/completions")
 async def proxy_chat(request: Request):
     try:
@@ -99,14 +94,15 @@ async def proxy_chat(request: Request):
         if chat_request.stream:
             return StreamingResponse(
                 stream_generator(chat_request.model, payload, headers),
-                media_type="text/event-stream"
+                media_type="text/event-stream; charset=utf-8",
+                headers={"Content-Type": "text/event-stream; charset=utf-8"}
             )
         else:
             response = requests.post(API_URL, headers=headers, json=payload)
             data = response.json()
             if "model" in data and data["model"] == BACKEND_MODEL:
                 data["model"] = chat_request.model
-            return JSONResponse(content=data)
+            return JSONResponse(content=data, media_type="application/json; charset=utf-8")
 
     except Exception as e:
         logger.error("Error in /v1/chat/completions: %s", str(e))