Errors and rate limits

Every error the API returns is a Problem object served as application/problem+json. Errors explains how to read it, what to do for each status from 400 to 500, which requests to retry and what to quote to support; the page also carries the catalogue generated from the contract.

Rate limits then describes the operational burst protection: two token buckets, per key and per workspace, the noryo-ratelimit-limit and noryo-ratelimit-remaining headers, the 429 response with Retry-After, what consumes a token, and how to retry with the same Idempotency-Key.