Errors

/v1 errors look like:

{
  "error": {
    "code": "invalid_request",
    "message": "Request could not be accepted.",
    "details": {}
  }
}

Handle the API from error.code. The message is for a person. details carries extra fields when the code defines them, such as required and available on insufficient_credits.

HTTP	Code
401	`authentication_required`, `invalid_api_key`
400	`invalid_request`, `idempotency_key_required`, `invalid_idempotency_key`, `invalid_cursor`, `quick_pulse_unavailable`
402	`insufficient_credits`, `subscription_required`
404	`run_not_found`, `model_response_not_found`
409	`idempotency_conflict`, `concurrency_limit_reached`, `result_not_ready`, `responses_not_ready`, `research_failed`
429	`rate_limit_exceeded`
500	`internal_error`

quick_pulse_unavailable rejects a retired panel id. It does not reserve credits. The public panel ids are wide, consensus, and seneca. Another workspace’s run is 404 run_not_found, not the report. A 500 internal_error body does not include a traceback or a secret.

A lost create response is not an error to paper over with a new idempotency key. Repeat the same POST with the same key, prompt, and panel. See Create a run.

Getting started

Research

Operations