We're relying on endpoints for this agent and they can be paused and sometimes the replicas that
are being used can be deprecated.  

Go to https://endpoints.huggingface.co/{my user name}}/endpoints/

Depending on the model size / image can make cold start 5–10+ minutes; for demos, consider min_replicas: 1 or disabling scale-to-zero (cost ↑).

Auth: ensure HF_TOKEN has access; 401/403 will also look like “not ready”.

Correct URL (includes https:// and no extra path).