Errors

Every error response uses the OpenAI-compatible envelope:

{
  "error": {
    "code": "insufficient_credits",
    "message": "Insufficient credits to complete this request. Add credits or upgrade your plan.",
    "type": "insufficient_quota",
    "request_id": "..."
  }
}

The code field is the stable contract — match on it in your client. The message is human-readable and may evolve. Always include the request_id when contacting support.

Stable error codes

code	HTTP	meaning
`invalid_api_key`	401	The Authorization header was missing or the key didn't match a row.
`api_key_revoked`	401	The key was revoked. Create a new one in /api-keys.
`api_key_paused`	401	The key is paused. Resume it in /api-keys.
`insufficient_credits`	402	Wallet balance minus locked credits would go negative.
`model_not_permitted`	403	This key is restricted and the requested model is not on its allow-list.
`endpoint_not_permitted`	403	This key cannot call this endpoint.
`org_suspended`	403	The organisation owning this key is no longer active.
`model_not_found`	404	No model with this slug, or it is currently disabled.
`rate_limit_exceeded`	429	Per-key requests-per-minute or tokens-per-minute exceeded.
`concurrency_limit_exceeded`	429	Too many in-flight requests on this key.
`invalid_request`	400	Bad request body, missing required field, or schema violation.
`context_length_exceeded`	400	Input + max_tokens exceeds the model's context window.
`provider_error`	502	Upstream provider returned a non-recoverable error.
`provider_timeout`	504	Upstream provider didn't respond in time.
`provider_unavailable`	503	No provider is configured for this model right now.
`gateway_error`	500	An error inside our gateway. Quote x-infra-request-id when reporting.
`internal_error`	500	Catch-all for unexpected failures. Quote x-infra-request-id.