# Errors

Errors come back with an HTTP status and a JSON body in the same shape as OpenAI's, so the OpenAI SDKs raise their usual exceptions.

```json
{"error": {"message": "Your account has no credit left. Add credit at https://www.continuumlabs.ai/console", "type": "insufficient_quota", "code": "insufficient_quota"}}
```

## Every error

| Status | `code` | What it means | What to do |
|---|---|---|---|
| 400 | `invalid_request_error` | The body is not valid JSON. | Check the request body. |
| 401 | `invalid_api_key` | The key is missing, malformed, deleted or unknown. | Send `Authorization: Bearer cl-...`. Create a new key in the [console](/console) if needed. |
| 401 | `key_disabled` | The key has been switched off in the console. | Ask your account owner to switch it back on. |
| 402 | `insufficient_quota` | Your account has no credit left. | Add credit. Requests work again straight away. |
| 402 | `allowance_reached` | The key's or the workspace's [allowance](/docs/allowances) for this period is used up. The message says which. | Wait for the reset, or raise the allowance in the console. |
| 403 | `permission_error` | A management key was used to call a model, or an ordinary key to manage. | Use the right kind of key. |
| 404 | `model_not_found` | The model ID is not one we offer. The message lists the IDs that are. | Use an ID from [models](/docs/models). |
| 404 | `not_found` | The endpoint or the ID does not exist. | Check the path against these docs. |
| 413 | `request_too_large` | The request is over 4 MB. | Send less text, or split the work across requests. |
| 429 | `rate_limited` | The model is busy. | Wait and retry, with backoff. |
| 502 | `upstream_error` | The model returned an error. | Retry. If it keeps happening, tell us. |
| 502 | `upstream_unavailable` | The model could not be reached. | Retry after a few seconds. |
| 503 | `service_unavailable` | The service is not open, or is down for maintenance. | Check our status, then retry later. |

If the model rejects a request because of something in it, such as a parameter it does not support, you get its original status with a short explanation in `message`.

## Retrying safely

Retry `429`, `502` and `503` with exponential backoff: wait about one second, then two, then four, and stop after a few tries. Do not retry `400`, `401`, `402`, `404` or `413`, because they will fail the same way.

The OpenAI SDKs retry some errors for you. You can set how many times:

```python
client = OpenAI(base_url="https://api.continuumlabs.ai/v1", api_key="cl-your-key", max_retries=3)
```

Failed requests are not charged.