# Chatbox AI

Use [Chatbox AI](https://chatboxai.app/) with your Aurora token through a custom
OpenAI API Compatible provider. This guide uses the desktop application's
bring-your-own-key settings. The fastest path is to import a ready-made provider
configuration from the clipboard and then fetch the model list from Aurora; the
manual steps below cover older versions and mobile builds.

> Available Inference endpoint: `https://ai.aur.lu/v1` (Aurora).

## Install and prepare

Download Chatbox from its [official site](https://chatboxai.app/en/guide/getting-started/download).
Create an [Aurora API token](https://docs.aur.lu/docs/inference/getting-started). The token goes into the
custom provider's settings, not the Chatbox AI subscription/license field.

## Import Aurora from the clipboard

Chatbox desktop **1.15.1 or later** can create a provider from a JSON
configuration on the clipboard. Mobile builds do not show this option; use the
[manual steps](https://docs.aur.lu/docs/inference/integrations/chatbox#add-aurora-manually) instead.

1. Copy the configuration below. The API Host is filled in from the endpoint
   selected above.

```json
{
  "id": "aurora",
  "name": "Aurora",
  "type": "openai",
  "urls": { "website": "https://docs.aur.lu/" },
  "settings": {
    "apiHost": "https://ai.aur.lu/v1",
    "apiPath": "/chat/completions"
  }
}
```

2. Open **Settings → Model Provider → Add** and select **Import from clipboard**.
3. Review the **Import Provider Configuration** dialog. API Host should be the
   selected Aurora endpoint and API Path `/chat/completions`. Click **Save**.
4. Open the new **Aurora** provider and paste your Aurora token into
   **API Key**. The import dialog shows fields read-only, so the key is entered
   after import.
5. Continue with [Add models](https://docs.aur.lu/docs/inference/integrations/chatbox#add-models).

Keep `apiPath` in the configuration. Chatbox defaults it to `/v1/chat/completions`,
and the API Host already includes `/v1`, so omitting it sends requests to
`/v1/v1/chat/completions`.

Importing again with the same `id` overwrites the existing Aurora provider. This
is a quick way to switch endpoints, but the API Key must be entered again unless
you add an `"apiKey"` field to the JSON before copying it.

## Add Aurora manually

Use these steps on mobile, on Chatbox versions before 1.15.1, or if the import
dialog reports an invalid configuration.

1. Open **Settings → Model Provider → Add**.
2. Select **Add Custom Provider**.
3. Name it `Aurora` and select **OpenAI API Compatible**.
4. Set **API Host** to:

```text
https://ai.aur.lu/v1
```

5. Set **API Path** to `/chat/completions`.
6. Paste your Aurora token into **API Key**.

The API Host already includes `/v1`. Chatbox combines the host and API path;
do not put `/v1/chat/completions` in API Path or repeat the path in API Host.

## Add models

### Fetch from Aurora (recommended)

Under **Models**, click **Fetch**. The list shows every model in the Aurora
catalog; add the ones you want to use. Chatbox fills in each model's ID, display
name, and context window from the catalog. It might not correctly infer other
model capabilities or **Max output tokens**. Verify these settings against the
[Aurora catalog and model information](https://docs.aur.lu/docs/inference/models-and-pricing).

Fetch does not update a model that is already in the list. To refresh its
metadata, delete the entry and fetch again.

### Add a model manually

If Fetch is unavailable, pick a model from the
[Aurora catalog and model information](https://docs.aur.lu/docs/inference/models-and-pricing), then click
**New** under **Models** and enter:

| Field                   | Value                                                          |
| ----------------------- | -------------------------------------------------------------- |
| Model ID                | The model's complete catalog ID, including its author prefix   |
| Display name / nickname | Any label, for example the catalog name followed by `(Aurora)` |
| Model type              | **Chat**                                                       |
| Context window          | The model's context length from the catalog                    |
| Max output tokens       | An output budget within the model's limit                      |

Enable capabilities such as tool use or vision only after checking model support.

## Test the model

Run **Check** on an Aurora model, save, and select it in a **new chat**. Send:

```text
Say hello in one sentence, then list three things you can help me write.
```

Expect an answer in the chat pane. Chatbox supports streamed chat responses;
if the answer arrives all at once, check its streaming/network compatibility
settings and any proxy between the application and Aurora.

## Troubleshooting

- **"Invalid provider configuration format" on import:** Copy the whole JSON
  block, including both outer braces, and undo any edits that broke the JSON.
  Versions before 1.15.1 cannot import; use the manual steps.
- **"Failed to read from clipboard":** The clipboard must hold the configuration
  when you click the option. Copy the block again and retry.
- **"Failed to fetch models":** Fetch requests `<API Host>/models`. Check that
  API Host is the selected Aurora endpoint and that API Key holds your token,
  then retry. Add the model manually if the problem persists.
- **`404`:** Check API Host and API Path separately. The completed URL must end
  in exactly `/v1/chat/completions`. For an imported provider, API Path must be
  `/chat/completions`.
- **Check succeeds but another model answers:** Select the saved Aurora model
  in the conversation, not a built-in Chatbox AI model.
- **Model not found:** Use the model's complete catalog ID, including its author
  prefix, not the nickname. Fetch again if the catalog has changed.
- **Attachments or Work Mode fail:** Plain chat is the first connection test.
  Images require a compatible model; knowledge-base embeddings, speech, and
  other Work Mode services may need separate configuration.

See [shared troubleshooting](https://docs.aur.lu/docs/inference/integrations#troubleshooting) and Chatbox's
[custom-provider examples](https://chatboxai.app/en/guide/byok/other-platforms)
and [provider import format](https://docs.chatboxai.app/en/guides/providers/import-config).