Reliability & Control
Trace requests, review failures, and handle upstream instability without losing cost visibility.
Trace every request
Use the request ID, model, status, latency, usage, charge, and error type shown in the console to investigate failures and reconcile billing. When reporting a problem, include the request ID and timestamp so support can locate the request efficiently.
Failed calls and billing review
If a request fails, check its status, error type, usage, and charge in the console before retrying. If the billing record appears inconsistent with the result, contact support with the request ID. Any balance adjustment is reviewed under the service terms and the available request record.
Availability snapshots
Availability monitoring is useful for visible health signals, but it is not the same as a universal guarantee. Some model types, especially image and embedding models, may not be part of every probe set.