advaitpaliwal Managed Model Endpoints
Use this skill to make a model endpoint usable from Feynman.
Workflow:
- Define the endpoint purpose, model family, input/output schema, auth method, hardware need, and expected latency.
- Record whether the endpoint is local, remote HTTP, Modal-backed, SSH-backed, or a custom connector.
- Add only non-secret endpoint metadata to settings. Store secret references as environment variable names or credential refs.
- Implement start/stop/health/inference checks when Feynman owns the endpoint lifecycle.
- Run a tiny inference smoke and save request/response shape without leaking secrets.
Expose endpoints as research infrastructure, not as permanent claims that a model is installed when health has not passed.