Rate Limits

Analysis uses a fixed-window quota per authenticated API token, shared across /text, /email, /audio, and WebSocket audio messages. Defaults: 10 requests per 60 seconds. Operators can configure MAX_REQUEST_LIMIT and RATE_LIMIT_WINDOW.

HTTP quota failures return 429 and a conservative Retry-After window in seconds:

{"detail":"Rate limit exceeded"}

Redis 7+ transactions keep concurrent requests within the quota. Later requests do not extend the window. Quotas reset on Redis restart; Redis outages fail closed with 503. Clients sharing a token share its quota, so bots/gateways must add per-user limits.