Troubleshoot from a request record

Start with a request ID, then check configuration, access, quotas, and upstream responses.

Concept & preparation guideUpdated Oct 21, 2026

Identify the same request

Keep the request time, request ID, model, and returned status. Query the same time window and use the request ID to connect application-side and gateway-side information.

  • request_id: an identifier for correlating a request.
  • model / status: the requested model and request status.
  • duration / usage: elapsed time and consumption; field definitions depend on the release.

Check each stage of the request

  • Address and protocol: did the request reach the correct endpoint with appropriate parameters?
  • Authentication and access: is the key valid and is the model allowed?
  • Quotas and rate limits: did the request reach a configured limit?
  • Channel and upstream: is the connection configured correctly and what status did the upstream return?

Keep reproducible details

Record the reproduction time, client and deployment versions, request ID, redacted error information, and diagnostic steps. Remove real keys, provider credentials, and business content before sharing.

Request status and metadata do not imply complete distributed tracing or request-body replay.

Match capabilities and procedures to your deployment version. Read the documentation scope

BOS API · TYPEX

Sales consultation is unavailable

Sales reception could not be initialized. Please try again later. Documentation and free self-hosting remain available.

Read the docsSelf-host for free