API error reference
What each error actually means and what to do about it — including which ones are worth retrying.
Genace API errors
Returned by the Genace gateway. Facts come from the gateway source, not from a generated schema dump.
| Status | Code | What it means |
|---|---|---|
| 400 | INVALID_JSON | Request body is not valid JSON |
| 400 | INVALID_PARAM | Invalid params |
| 400 | UNKNOWN_MODEL | Unknown model |
| 401 | INVALID_API_KEY | API key invalid or revoked |
| 401 | MISSING_API_KEY | Authorization header must be: Bearer <api_key> |
| 402 | NOT_ENOUGH_CREDITS | Not enough credits |
| 404 | JOB_NOT_FOUND | Job not found |
| 429 | CONCURRENCY_LIMIT_EXCEEDED | Concurrency limit reached |
| 429 | RATE_LIMIT_EXCEEDED | api rate limit exceeded |
| 502 | UPSTREAM_SUBMISSION_UNCERTAIN | UPSTREAM_SUBMISSION_UNCERTAIN |
| — | upstream-fail | UPSTREAM_FAIL |
Upstream provider errors
Errors from the model APIs you might be calling directly. Every error text is quoted verbatim from the provider's docs or a public issue, with the source linked.
- fal.ai 3 errors
Queue, auth, balance and validation errors from the fal.ai model API and its JS / Python clients.
- Google Veo, Imagen & Gemini image 31 errors
Safety-filter, quota, region and parameter errors from Veo, Imagen and Gemini image generation, on both the Gemini API and Vertex AI.
- OpenAI image generation 5 errors
Moderation, verification, billing and request-shape errors from the OpenAI Images API (gpt-image models).
- Replicate 1 error
Prediction failures, E-codes, billing and rate limits from the Replicate API and its clients.