Skip to main content

Troubleshooting

{baseUrl} means your platform API root without a trailing / or /v1. Replace it with the actual address when checking configuration.

Error reference​

HTTP statuserror.codeMeaning and action
400INVALID_REQUESTInvalid or missing parameters such as model or prompt. Check the model ID and request format.
401—Missing authorization or an invalid token. Check the full key and the Bearer prefix.
404TASK_NOT_FOUNDThe task does not exist or belongs to a different API key.
429—Rate limit reached or insufficient balance, depending on the platform response. Reduce concurrency and check your balance.
500INTERNAL_ERRORInternal server error. Retry later.
503—No available upstream channel. Try another permitted group or model.

Standard error response:

{ "error": { "code": "...", "message": "..." } }

Frequently encountered issues​

Q1: 401 token.not_provided​

The key was not passed correctly. Check:

  • The Authorization header has Bearer followed by one space and the key.
  • The client key is in the API Key field, not Organization ID.
  • The key has not expired or been deleted.

Q2: 401 even though the key is correct​

Check the credential type. Model endpoints use an sk- model API key. Management endpoints under /api/... use a management access token. They are not interchangeable.

Q3: 404 Invalid URL​

The client may have built a URL such as /v1/v1/chat/completions. Review the URL rules: when the client appends /v1, configure only the root URL.

Q4: model not found or no available channel in the group​

The model ID may be misspelled, or the selected group may not provide it.

  1. Use GET /v1/models to check the exact ID.
  2. Check the token's group in the console, or choose automatic grouping if your deployment supports it.

Q5: Claude Code asks you to sign in or subscribe​

The environment variables may not have taken effect, or ANTHROPIC_BASE_URL may incorrectly include /v1.

echo $ANTHROPIC_BASE_URL

It should show the root URL represented by {baseUrl}. Confirm that ANTHROPIC_AUTH_TOKEN is set for this configuration, and reopen the terminal if necessary.

Q6: Codex CLI reports a wire_api error​

Check the configuration format supported by your Codex version. Use wire_api = "responses" and base_url = "{baseUrl}/v1". The model and upstream must support /v1/responses; do not assume the current version still supports the legacy chat mode.

Q7: Connection succeeds but the model list is empty​

Some clients maintain their own model list. Add model IDs manually, or use Sync Models / Check to retrieve them through /v1/models.

Q8: Slow responses or frequent timeouts​

The selected group's upstream route may be congested. Choose another available group, such as cn1 or ovip2 if offered by your deployment, or reduce max_tokens.

Q9: Streaming stops unexpectedly​

Clients handle SSE differently. Try disabling the client's proxy setting, or use a non-streaming request to isolate the problem.

Q10: No result after submitting an image or video task​

Endpoints that return a task_id are asynchronous and require polling. Synchronous image endpoints return the result directly. If a query returns 404 TASK_NOT_FOUND, confirm that the task ID is correct and that the query uses the same API key as the submission.