Errors and reliability FAQ
Links on this page to features marked Preview describe designs that are not available yet.
Which errors should I retry?
Types rate_limit_error, service_unavailable_error, api_error and timeout_error, with back-off, honouring Retry-After. Do not retry other 4xx errors unchanged. Full list: Error code index.
What should I send to support?
The request_id from the error body or the x-syntic-request-id header, the time, and the endpoint. See Request IDs.
A stream stopped halfway. Can I resume it?
No. Retry the whole request. For agent sessions, reconnect to the session stream with Last-Event-ID; session events can be replayed.
How do I avoid duplicate creates when retrying?
Send an Idempotency-Key on POST requests to preview endpoints; a retry within 24 hours returns the original result.
Why did my typo in the model name not fail?
Today an unknown name falls back to the default model. The stricter 404 model_not_found behaviour is on the way; keep model names in configuration and test them.
Was this page helpful?