> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nexusapi.link/llms.txt
> Use this file to discover all available pages before exploring further.

# Self-check before contacting support

> Collect the minimum evidence needed to diagnose a NexusAPI request failure.

Most setup mistakes and transient failures can be narrowed down in a few minutes.

## 1. Preserve the evidence

Before closing the client or retrying, save:

* Complete error text
* Request ID
* Time and time zone
* Model ID
* Client and version
* Request endpoint, such as `/v1/responses`

Do not share full API keys, passwords, private data, or complete business content.

## 2. Identify the category

| Symptom                | Check first                                                                       |
| ---------------------- | --------------------------------------------------------------------------------- |
| 401                    | API key, `Bearer` prefix, and request address                                     |
| 403                    | Token group, model restriction, IP allowlist, account status, or upstream refusal |
| 404                    | Base URL and accidental duplicate or missing `/v1`                                |
| 400                    | Model, endpoint, parameters, or session state                                     |
| 429                    | Concurrency and request rate                                                      |
| 500 / 502 / 503        | Wait 10–30 seconds, then retry only once                                          |
| 504 / 524              | Request size, wait time, and service status                                       |
| `500, not implemented` | Do not retry; check whether `/v1/responses` was used with a Claude model          |
| `no available channel` | Model ID, token group, and current model availability                             |
| Image timeout          | Do not resubmit the same image task; keep its Request ID                          |

See [API error guide](/en/faq/common-errors) for details.

## 3. Run one minimal test

1. Confirm console credit and API-key status.
2. Copy the model ID again from the Model Marketplace.
3. Start a new session and send a short text request.
4. Retry only once and save the new Request ID.
5. Check whether the matching request appears in the usage log.

If a new short session works, the original conversation state or request content is likely involved. If it still fails, submit the evidence.

## What to send support

```text theme={"system"}
Time (with time zone):
Request ID:
Client and version:
Model:
Endpoint:
Full error text:
Intermittent or persistent:
Tested with a new session:
```

If an API key is necessary for identification, provide no more than its last six characters.