fix: add fallback to free models when Claude credits exhausted in /api/ask

server.py

@@ -114,6 +114,16 @@ async def ask_jarvis(req: Request):
     except Exception as e:
         return {"error": str(e)}
 
+    # Fallback: if primary backend returned an error, retry with free models
+    if full_response.startswith("Error:") or full_response.startswith("\nError:"):
+        if backend in ("auto", "claude"):
+            full_response = ""
+            try:
+                async for chunk in stream_response(messages, system, "openrouter", params):
+                    full_response += chunk
+            except Exception:
+                pass
+
     # Apply STM
     cleaned = apply_stm(full_response, aggressive=True)