fixing

app.py

@@ -44,7 +44,7 @@ except (ImportError, RuntimeError, OSError):
 import spaces
 from typing import List, Literal, Optional
 from pydantic import BaseModel, Field
-from fastapi import FastAPI, HTTPException, Header
+from fastapi import HTTPException, Header, Request
 import gradio as gr
 
 # =============================================================
@@ -69,11 +69,14 @@ SYSTEM_PROMPT = (
     "earnings, and low traffic delays."
 )
 
+# =============================================================
+# 3. Core ZeroGPU Inference function
+# =============================================================
 @spaces.GPU
-def run_inference(prompt: str, history: List[dict] = None) -> str:
+def generate_response(prompt: str, history_messages: list = None) -> str:
     messages = [{"role": "system", "content": SYSTEM_PROMPT}]
-    if history:
-        for m in history:
+    if history_messages:
+        for m in history_messages:
             messages.append({"role": m["role"], "content": m["content"]})
     messages.append({"role": "user", "content": prompt})
     response = llm.create_chat_completion(
@@ -82,10 +85,27 @@ def run_inference(prompt: str, history: List[dict] = None) -> str:
     return response["choices"][0]["message"]["content"]
 
 # =============================================================
-# 3. Create FastAPI app & define POST /chat endpoint
+# 4. Gradio Chat UI with @spaces.GPU
 # =============================================================
-app = FastAPI(title="CommitCon Sherpa AI Copilot")
+@spaces.GPU
+def gradio_chat(user_message, history):
+    hist_formatted = []
+    if history:
+        for user_msg, bot_msg in history:
+            hist_formatted.append({"role": "user", "content": user_msg})
+            if bot_msg:
+                hist_formatted.append({"role": "assistant", "content": bot_msg})
+    return generate_response(user_message, hist_formatted)
 
+demo = gr.ChatInterface(
+    fn=gradio_chat,
+    title="🧭 CommitCon Sherpa Driver Copilot",
+    description="Fine-tuned Qwen 0.8B Q4_K_M — llama-cpp-python inference"
+)
+
+# =============================================================
+# 5. Attach POST /chat to demo.app using Gradio's native router
+# =============================================================
 class HistoryMessage(BaseModel):
     role: Literal["user", "assistant"]
     content: str = Field(min_length=1, max_length=4000)
@@ -96,12 +116,12 @@ class ChatRequest(BaseModel):
     current_zone_id: Optional[str] = None
     history: List[HistoryMessage] = Field(default_factory=list)
 
-@app.post("/chat")
-def chat_endpoint(payload: ChatRequest, x_api_key: Optional[str] = Header(None)):
+@demo.app.post("/chat")
+def chat_endpoint(payload: ChatRequest, request: Request, x_api_key: Optional[str] = Header(None)):
     if API_KEY and x_api_key != API_KEY:
         raise HTTPException(status_code=401, detail="Unauthorized: Invalid API Key")
     formatted_history = [{"role": m.role, "content": m.content} for m in payload.history]
-    answer = run_inference(payload.driver_prompt, formatted_history)
+    answer = generate_response(payload.driver_prompt, formatted_history)
     return {
         "status": "success",
         "answer": answer,
@@ -110,30 +130,7 @@ def chat_endpoint(payload: ChatRequest, x_api_key: Optional[str] = Header(None))
     }
 
 # =============================================================
-# 4. Gradio Chat UI
-# =============================================================
-def gradio_chat(user_message, history):
-    hist_formatted = []
-    if history:
-        for user_msg, bot_msg in history:
-            hist_formatted.append({"role": "user", "content": user_msg})
-            if bot_msg:
-                hist_formatted.append({"role": "assistant", "content": bot_msg})
-    return run_inference(user_message, hist_formatted)
-
-demo = gr.ChatInterface(
-    fn=gradio_chat,
-    title="🧭 CommitCon Sherpa Driver Copilot",
-    description="Fine-tuned Qwen 0.8B Q4_K_M — llama-cpp-python inference"
-)
-
-# Mount Gradio UI on the FastAPI app
-app = gr.mount_gradio_app(app, demo, path="/")
-
-# =============================================================
-# 5. Uvicorn Runner on Environment PORT
+# 6. Launch with demo.launch() (DO NOT USE UVICORN DIRECTLY ON ZEROGPU)
 # =============================================================
 if __name__ == "__main__":
-    import uvicorn
-    port = int(os.environ.get("PORT", 7860))
-    uvicorn.run(app, host="0.0.0.0", port=port)
+    demo.launch()