Errors
Errors return JSON with error and message. Optional: limit, reset_at, retry_after_seconds.
| HTTP | Meaning |
|---|---|
401 | Missing or invalid API key |
413 | Payload / token cap exceeded |
422 | JSON or schema validation failed |
429 | Account RPM, daily or concurrency limit, or global capacity — see limit and Retry-After |
503 | inference_unavailable / inference_not_ready — GPU temporarily unavailable; not charged |
504 | inference_timeout — exceeded the 60 s deadline; not charged |
5xx | Server error — retry with backoff |
Unavailable example
{
"error": "inference_unavailable",
"message": "GPU worker is temporarily unavailable."
}
Rate limit example
{ "error": "rate_limited", "limit": "rpm", "reset_at": "…" }
Full table: project file api-errors-v1.md.