Update app.py

app.py

@@ -42,16 +42,13 @@ except (ImportError, RuntimeError, OSError):
     from llama_cpp import Llama
 
 import spaces
-from typing import List, Literal, Optional, Any
-from pydantic import BaseModel, Field
-from fastapi import HTTPException, Header
+from typing import List
 import gradio as gr
 
 # =============================================================
 # 2. Load GGUF Model
 # =============================================================
 MODEL_PATH = "Qwen3.5-0.8B.Q4_K_M.gguf"
-API_KEY = os.getenv("API_KEY", "commitcon-secret-2026")
 
 print(f"Loading {MODEL_PATH} via llama-cpp-python ...")
 llm = Llama(
@@ -69,7 +66,6 @@ SYSTEM_PROMPT = (
     "earnings, and low traffic delays."
 )
 
-@spaces.GPU
 def run_inference(prompt: str, history: List[dict] = None) -> str:
     messages = [{"role": "system", "content": SYSTEM_PROMPT}]
     if history:
@@ -77,60 +73,36 @@ def run_inference(prompt: str, history: List[dict] = None) -> str:
             messages.append({"role": m["role"], "content": m["content"]})
     messages.append({"role": "user", "content": prompt})
     response = llm.create_chat_completion(
-        messages=messages, temperature=0.7, max_tokens=512
+        messages=messages, temperature=0.7, max_tokens=256
     )
     return response["choices"][0]["message"]["content"]
 
 # =============================================================
-# 3. Gradio Interface (Satisfies ZeroGPU startup check)
+# 3. Gradio Interface (duration=15 ensures ZeroGPU quota is preserved)
 # =============================================================
-@spaces.GPU
+@spaces.GPU(duration=15)
 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})
+        for item in history:
+            if isinstance(item, (list, tuple)) and len(item) == 2:
+                u, b = item
+                hist_formatted.append({"role": "user", "content": str(u)})
+                if b:
+                    hist_formatted.append({"role": "assistant", "content": str(b)})
+            elif isinstance(item, dict):
+                hist_formatted.append(item)
     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"
+    description="Fine-tuned Qwen 0.8B Q4_K_M — llama-cpp-python inference",
+    api_name="chat"
 )
 
 # =============================================================
-# 4. Custom API Schemas (Matching CommitCon/src/phase5_api.py)
-# =============================================================
-class HistoryMessage(BaseModel):
-    role: Literal["user", "assistant"]
-    content: str = Field(min_length=1, max_length=4000)
-
-class ChatRequest(BaseModel):
-    driver_id: Optional[str] = "DRV_DEFAULT"
-    driver_prompt: str = Field(min_length=1, max_length=2000)
-    recommendation_data: Optional[dict[str, Any]] = None
-    current_zone_id: Optional[str] = None
-    history: List[HistoryMessage] = Field(default_factory=list)
-
-def chat_endpoint(payload: ChatRequest, x_api_key: Optional[str] = Header(None, alias="X-API-Key")):
-    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)
-    return {
-        "status": "success",
-        "answer": answer,
-        "model": "project-driver-qwen-gguf",
-        "zone_context": payload.current_zone_id
-    }
-
-# Register the POST /chat route directly into Gradio's underlying FastAPI router
-demo.app.add_api_route("/chat", chat_endpoint, methods=["POST"])
-
-# =============================================================
-# 5. Native ZeroGPU Launch
+# 4. Native ZeroGPU Launch
 # =============================================================
 if __name__ == "__main__":
     demo.launch()