Skip to main content

Checklist

Use this public checklist to diagnose client-visible issues before contacting support. It avoids internal origin details and administrator-only runbooks.

When you contact support, include the timestamp, public URL or API path, HTTP status, request ID if available, and a redacted client log. Do not send raw API keys or full Authorization headers.

DNS Problems​

Symptoms:

  • The domain cannot be resolved.
  • curl reports Could not resolve host.
  • Only one network or region is affected.

Checks:

DNS checks
dig docs.codexpertise.com
dig api.codexpertise.com

If DNS fails, retry from another network, clear local DNS cache if you control the client machine, and record the resolver or region where it failed.

TLS Problems​

Symptoms:

  • The browser shows a certificate warning.
  • curl reports a certificate, handshake, or protocol error.
  • Corporate proxies work for other sites but fail for Codexpertise.

Checks:

TLS checks
curl -Iv https://docs.codexpertise.com/en/
curl -Iv https://api.codexpertise.com/

If TLS fails only behind a proxy, ask the network administrator whether TLS inspection, blocked ciphers, or custom trust stores are involved.

CloudFront 5xx​

Symptoms:

  • You receive a 500, 502, 503, or 504 response from a public Codexpertise domain.
  • Retrying from another region gives a different result.

Collect the timestamp, URL, status code, and any public request ID in the response headers. Do not include internal hostnames or secrets in the report. The operations team will check the private routing path.

Nginx 502​

Symptoms:

  • The public site returns 502 Bad Gateway.
  • A specific path fails while other Codexpertise pages still load.

From the client side, confirm the public URL and whether the same failure happens on /en/, /zh/, and the exact API path. Administrator-level Nginx checks live outside the public user documentation.

sub2api Upstream Error​

Symptoms:

  • The API returns a 5xx Upstream Error.
  • The relay responds, but the provider-side request fails.

Retry with backoff and a smaller request. If the issue persists, contact support with the timestamp, API path, HTTP status, and request ID if available. Do not send your API key.

TODO: verify against sub2api before documenting final upstream error codes and request ID fields.

Authorization Header Missing​

Symptoms:

  • API requests return 401 Unauthorized.
  • The same key works in one client but fails in another.

Checks:

Header pattern
Authorization: Bearer YOUR_API_KEY

Common causes include environment variables not being loaded, proxy rules stripping headers, misspelled header names, copied whitespace, or using a key that has been revoked.

Streaming Interrupted​

Symptoms:

  • Output starts and then stops.
  • The client reports a broken pipe, timeout, cancelled request, or incomplete event stream.

Retry only when safe, test a non-streaming request, reduce response length, and check whether your client, network, or proxy has a short timeout. If this happens consistently, include the timestamp and client timeout settings when contacting support.

Slow Response​

Symptoms:

  • Requests eventually succeed but take longer than expected.
  • Coding-agent tasks wait for a long response before continuing.

Try a smaller prompt, reduce concurrency, test from another network, and compare streaming versus non-streaming behavior. Slow upstream generation can be normal for large requests, but repeated delays should be reported with timestamps and request paths.