Update app.py

app.py

@@ -8,23 +8,25 @@ import uvicorn
 
 app = FastAPI()
 
-# 1. Start Ollama in the background
+# FIX 1: Increase Context Window to 16k to stop the "Truncation" error
+# This allows the model to 'read' more of your project at once.
+os.environ["OLLAMA_NUM_CTX"] = "16384"
+
+# Start Ollama in the background
 subprocess.Popen(["ollama", "serve"])
 
 @app.api_route("/{path:path}", methods=["GET", "POST", "PUT", "DELETE"])
 async def proxy_to_ollama(request: Request, path: str):
-    # This forwards EVERYTHING to the local Ollama engine
+    # Reverse Proxy: Forwards Claude CLI requests to the internal Ollama engine
     url = httpx.URL(f"http://127.0.0.1:11434/{path}", query=request.url.query.encode("utf-8"))
     
     async with httpx.AsyncClient(timeout=None) as client:
-        # Build the request to forward
         req = client.build_request(
             request.method,
             url,
             headers=request.headers.raw,
             content=request.stream(),
         )
-        # Send it and stream the response back (important for coding speed)
         resp = await client.send(req, stream=True)
         return StreamingResponse(
             resp.aiter_raw(),
@@ -32,11 +34,18 @@ async def proxy_to_ollama(request: Request, path: str):
             headers=dict(resp.headers)
         )
 
+@app.get("/")
+def health_check():
+    return {"status": "running", "model": "qwen2.5-coder:3b", "context": "16k"}
+
 if __name__ == "__main__":
-    # 2. Wait for Ollama to wake up and pull the model
+    # Wait for Ollama to wake up
     time.sleep(5)
-    subprocess.run(["ollama", "pull", "qwen2.5-coder:7b"])
     
-    # 3. Start the web server on the port Hugging Face expects (7860)
-    print("Bridge is active on port 7860. Forwarding to Ollama on 11434...")
-    uvicorn.run(app, host="0.0.0.0", port=7860)
+    # FIX 2: Switch to the 3B model
+    # It provides a 'snappy' response on CPU compared to the 14-minute 7B delay.
+    print("Pulling Qwen 2.5 Coder 3B...")
+    subprocess.run(["ollama", "pull", "qwen2.5-coder:3b"])
+    
+    # Start the web server for Hugging Face
+    uvicorn.run(app, host="0.0.0.0", port=7860)