Skip to main content
Send an Idempotency-Key header (8–200 printable ASCII characters) to make a retry return the first result instead of running again.

Inference

On non-streaming inference, a replay returns the stored response without calling the model or charging again. Stored responses are encrypted and deleted after 24 hours; this is the only time Antara keeps response content. Streams can’t be replayed. A key reused after a streamed request finished gets 409 idempotency_replay_unavailable; use a new key for a new request.