Rate limits and budgets

The per-key limiter uses fixed 60-second windows. Its shared cache is eventually consistent, so counts are best-effort rather than exact. A configured limit of zero or less disables this limiter.

Keys default to 60 requests per minute in legacy routing. Catalog routing enforces only operator-configured custom key limits; the default is not an enforced catalog-wide cap. Read x-ratelimit-limit, x-ratelimit-remaining and x-ratelimit-reset when present. On HTTP 429, wait for the Retry-After seconds before retrying.

Daily and per-call budgets are not available in Console. Wallet shows prepaid credit and Usage shows recorded charges. Holds reduce available credit while calls settle. Do not treat the displayed balance as an exact concurrent spending cap.

Wallet · Usage · Billing