Loading documentation…
Loading documentation…
Interpret validation, authentication, rate-limit, and upstream failures.
Use HTTP status and the documented response schema together. Do not parse human-readable messages to drive retry behaviour.
Never expose raw upstream payloads directly to end users.
V2 errors expose a stable code, safe message, and request ID. Capture the code for application logic and the request ID for diagnosis. Treat fields in the API Reference as canonical; do not infer internal provider details from the public message.
Centralize classification in your API client so every worker applies the same policy. The result should answer whether to correct input, stop and alert, wait for rate capacity, retry within a deadline, or show a declared unavailable state.
| Class | Default client action |
|---|---|
| Validation or authentication | Stop; correct the request or credential |
| Product access, credits, or availability | Stop the operation and surface an account or product state |
| Rate limit | Wait according to documented response metadata and coordinate callers |
| Privacy or not found | Preserve the product-specific state; do not retry unchanged |
| Transient upstream or capacity | Retry only if the operation and workflow budget permit |
| Unexpected error | Preserve evidence, fail safely, and avoid blind retry storms |
Continue with Retries & timeouts, Request IDs, and Troubleshooting.