Errors & limits

Check the HTTP status code before reading a response as a successful completion. Website ingress errors have an error string; upstream errors may use the provider's own error object.

Website ingress errors

Status	Meaning	What to do
401	Missing, invalid, expired or revoked API key	Check the bearer header and key status in your workspace.
404	The requested path or method is not supported	Check the endpoint contract and HTTP method.
503	Upstream connection error	Check the error details. Use bounded exponential backoff for transient connection failures.

Example website ingress error:

{
  "error": "A valid active API key is required."
}

Upstream errors

Model validation, rate limits, spending-limit errors and service failures are returned by the gateway. Their exact request limits and response formats depend on the selected model and account configuration.

For a rate-limit response, respect Retry-After when provided. Do not retry an invalid key, an invalid model identifier or an exhausted spending limit without resolving the cause.

Key and request limits

• At most 20 active keys per account.
• Newly created keys expire after 30 days.
• Allowed models and the spending limit are determined by the invitation.
• The public ingress allows only the explicitly supported /v1 routes, not arbitrary upstream administration endpoints.

Additional model-specific request size, video format and context-length limits will be published alongside the connected inference release.