Errors

OpenAI-compatible error envelope with Infery-specific codes.

Envelope

All errors follow OpenAI’s format:

{
  "error": {
    "message": "Human-readable description",
    "type": "invalid_request_error",
    "code": "file_too_large",
    "param": "file"
  }
}

The body is always valid JSON. param is included when the error relates to a specific field.

Types

type	HTTP	Meaning
invalid_request_error	400 / 413	Bad input, quota exceeded, unsupported format
invalid_request_error with code: model_not_found	404	Model slug doesn’t exist or isn’t available on your plan
auth_error	401	Missing or invalid API key
rate_limit_error	429	Per-key or per-IP rate limit hit
quota_exceeded	402	Out of credits or monthly budget reached
server_error	5xx	Upstream provider failure (often retried by fallback chain)

Common codes

code	Meaning	Fix
missing_api_key	No Authorization header	Add Bearer token
invalid_api_key	Unknown or revoked key	Create a new key
expired_api_key	Key has passed its expiry	Rotate
model_not_found	Slug doesn’t exist	GET /v1/models to list available
model_not_supported	Plan doesn’t include this model	Upgrade or pick another
file_too_large	Exceeds plan or provider limit	Upload smaller
file_count_quota_exceeded	Too many stored files on plan	Delete old ones or upgrade
storage_quota_exceeded	Storage bytes cap	Same
unsupported_file_type	MIME not allowed	Check supported list
mime_mismatch	File content doesn’t match declared MIME	Usually an upload bug in client
file_not_found	file_id doesn’t exist in workspace	Check id, ensure same workspace
captcha_required / captcha_failed	Contact form protection	Complete Turnstile challenge
rate_limit_exceeded	429	Back off, retry with jitter
insufficient_credits	Balance too low	Top up or wait for monthly renewal
audio_not_supported_by_model	e.g. Claude + audio	Transcribe first

Error IDs

Every error response includes x-request-id. Include it when reporting issues — our support can look up the full request/response in seconds.

Retrying

Retryable (with exponential backoff + jitter):

• 429 rate_limit_exceeded
• 502, 503, 504

Not retryable:

• 4xx other than 429 — they’re your client’s fault; fix the request
• 402 insufficient_credits — top up first

Our fallback chains already retry server-side for you if configured — see Fallbacks.