Errors

Errors follow the OpenAI-compatible error envelope:

{
  "error": {
    "message": "insufficient credits",
    "type": "insufficient_quota",
    "param": null,
    "code": "insufficient_quota"
  }
}

Status	Type	Meaning	Retry guidance
400	invalid_request_error	The request body, model, endpoint, parameters, or tool schema is invalid.	Fix the request; do not retry unchanged.
401	invalid_request_error	The API key is missing, malformed, revoked, or invalid.	Check the API key.
402	insufficient_quota	The balance cannot cover the request reservation.	Top up the organization balance.
408	upstream_error	The provider did not respond before the gateway timeout.	Retry with backoff; check usage before repeating long requests.
429	rate_limit_error	The provider or gateway rate limit was reached.	Retry after a delay with exponential backoff.
499	request_canceled	The client disconnected or canceled the request.	Retry only if the operation is safe to repeat.
500	server_error	An unexpected gateway error occurred.	Retry with backoff and include X-Request-ID when contacting support.
501	invalid_request_error	The requested capability or endpoint is not implemented.	Use a supported endpoint or model.
502	upstream_error	The provider rejected the request or returned a provider-side error.	Fix request/model compatibility or retry only when appropriate.
503	upstream_error	The provider is temporarily unavailable.	Retry with exponential backoff.
504	upstream_error	The provider request timed out.	Retry with backoff; check usage before repeating.

Every response carries X-Request-ID. Include it in support requests; the same ID keys your usage history and the provider call.

A provider response after the request was accepted may still consume tokens and affect your balance. Check the Usage & Audit Log before retrying.