Errors

Radium returns standard HTTP status codes. Every error response includes a JSON body with the error type and message.

Error table

HTTP status	Error type	Meaning	Safe to retry	What to do
400	`invalid_request_error`	Malformed request	No	Check request shape and parameters
401	`authentication_error`	Invalid or missing API key	No	Verify key and header format
403	`permission_error`	Key valid but lacks permission	No	Check dashboard permissions
404	`not_found_error`	Endpoint or model not found	No	Check base URL and model string
413	`request_too_large`	Request body exceeds limit	No	Reduce message size or batch
422	`unprocessable_entity`	Valid JSON but semantically invalid	No	Check parameter values
429	`rate_limit_error`	Quota exceeded	Yes	Back off exponentially, honor Retry-After
500	`internal_error`	Server error	Yes	Retry with backoff, then contact support
503	`service_unavailable`	Temporary outage	Yes	Retry with backoff

{
  "error": {
    "message": "Invalid model string",
    "type": "invalid_request_error",
    "param": "model",
    "code": "model_not_found"
  }
}

{
  "type": "error",
  "error": {
    "type": "invalid_request_error",
    "message": "Invalid model string"
  }
}

Use exponential backoff starting at 9 seconds. Honor the Retry-After header when present. Jitter each retry by up to 50% to avoid thundering herd.

Idempotency keys are not yet supported. Each retry may generate a new completion.