Usage & retries

Each successful reaction decision counts as one request, including "none". Failed calls and empty reaction lists don’t count. The free tier includes 1,000 requests total, with no credit card required. The Developer plan includes 100,000 requests for $100 per billing month, with no automatic overage charges.

Send an Idempotency-Key header (1–128 printable ASCII characters, without spaces) for each message. Reusing it with the same input returns the saved result without another charge. If a request is still running, wait before retrying. Use a new key for a new message. There are no automatic inference retries.

X-Request-Id identifies the request, X-Usage-Remaining reports the allowance available when it was reserved, and X-Idempotent-Replay: true identifies a saved result.