Update main.py

main.py

@@ -17,7 +17,7 @@ API_URL = "https://api.typegpt.net/v1/chat/completions"
 API_KEY = "sk-XzS5hhsa3vpIcRLz3prQirBQXOx2hPydPzSpzdRcE1YddnNm"
 BACKEND_MODEL = "pixtral-large-latest"
 
-# Load model-system-prompt mappings
+# Load model -> system prompt mappings
 with open("model_map.json", "r") as f:
     MODEL_PROMPTS = json.load(f)
 
@@ -37,24 +37,32 @@ class ChatRequest(BaseModel):
     presence_penalty: Optional[float] = 0.0
     frequency_penalty: Optional[float] = 0.0
 
-# Build payload to send to actual backend API
+# 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
+    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,
         "stream": chat.stream,
         "temperature": chat.temperature,
         "top_p": chat.top_p,
         "n": chat.n,
         "stop": chat.stop,
         "presence_penalty": chat.presence_penalty,
-        "frequency_penalty": chat.frequency_penalty,
-        "messages": [{"role": "system", "content": system_prompt}] + [
-            {"role": msg.role, "content": msg.content} for msg in chat.messages
-        ]
+        "frequency_penalty": chat.frequency_penalty
     }
 
-# Properly forward streaming data and replace model
+# Streaming response handler
 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):
@@ -75,13 +83,14 @@ def stream_generator(requested_model: str, payload: dict, headers: dict):
             else:
                 logger.debug("Non-data stream line skipped: %s", line)
 
-# Main endpoint
+# Proxy endpoint
 @app.post("/v1/chat/completions")
 async def proxy_chat(request: Request):
     try:
         body = await request.json()
         chat_request = ChatRequest(**body)
         payload = build_payload(chat_request)
+
         headers = {
             "Authorization": f"Bearer {API_KEY}",
             "Content-Type": "application/json"
@@ -100,5 +109,5 @@ async def proxy_chat(request: Request):
             return JSONResponse(content=data)
 
     except Exception as e:
-        logger.error("Error in proxy_chat: %s", str(e))
+        logger.error("Error in /v1/chat/completions: %s", str(e))
         return JSONResponse(content={"error": "Internal server error."}, status_code=500)