Skip to content

Errors and Operations โ€‹

Updated

What should I do when a 429 or rate limit error occurs? โ€‹

Reduce the number of concurrent requests and the query frequency, and follow the retry conditions indicated by the server. Use a request queue and exponential backoff so that calls are not sent all at once. Check the limits for generation requests and status queries separately.

Why is the generation time not consistent? โ€‹

The completion time varies depending on the model, output specifications, input conditions, and the service queue status. Record the request submission time and the task completion time separately, and use status queries instead of assuming completion after a fixed wait.

Are failed requests also billed? โ€‹

Billing criteria vary by product, model, processing stage, and contract terms. Do not assume that all failed requests are free or that all submitted requests are billed. Based on the task identifier and usage records, check the billing documentation or contact your account representative.

What should I validate when changing the model version? โ€‹

Check the supported resolutions, aspect ratios, and durations, input limits, additional parameters, the output file, and the error handling flow. Run a representative request to completion and check the result file before adopting it. Do not judge compatibility with the previous version based only on successful request submission.

Built with VitePress ยท Deployed on Tencent EdgeOne Pages