# Open WebUI

Connect a self-hosted [Open WebUI](https://openwebui.com/) instance to Aurora
through an OpenAI-compatible connection. Requests in this setup go from the
Open WebUI server to Aurora.

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

## Install and prepare

Follow the [Open WebUI quick start](https://docs.openwebui.com/getting-started/quick-start)
to install the application and create an administrator account. Obtain an
[Aurora token](https://docs.aur.lu/docs/inference/getting-started). The server or container running Open WebUI
must be able to reach the selected Aurora endpoint over HTTPS.

## Add the connection

1. Sign in as an administrator and open **Settings → Admin → Connections**
   (called **Admin Panel → Settings → Connections** in some versions).
2. Find **Manage OpenAI API Connections** and click **Add Connection**.
3. Set **Connection Type** to **External** and set **URL** to:

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

4. Enter your Aurora token as **API Key**. Use bearer/API-key authentication if
   the dialog offers an authentication selector.
5. Leave **Provider** at **Default** and use **Chat Completions**, not Responses.
6. Leave **Model IDs (Filter)** empty for automatic discovery, or add
   `deepseek/deepseek-v4-flash` to limit the visible models.
7. Leave **Prefix ID** empty for this initial setup and save.

The URL includes `/v1`; Open WebUI adds endpoint paths. The Aurora token is an
upstream credential, not an Open WebUI API key. Users allowed to use this
connection consume inference billed to the Aurora account behind that token.

## Start a chat

Open a new chat, select `deepseek/deepseek-v4-flash`, and send:

```text
Explain the difference between a model and a coding agent in three sentences.
```

Expect a streamed answer. Test plain chat first, before adding tools or files.
Review task-model settings for titles, tags, and follow-up suggestions if the
instance already uses other providers. Embeddings for retrieval, speech, and
image generation are configured independently of this connection.

## Troubleshooting

- **No models:** Check the server's network access and any model filter or user
  access restrictions. Discover the ID with Aurora's public catalog.
- **Model list loads but chat fails:** Verify the token with an authenticated
  request; Aurora model discovery itself does not require authentication.
- **Browser works but container fails:** Test network/DNS/proxy configuration
  from the Open WebUI server environment.
- **Streaming buffered:** Check the reverse proxy in front of Open WebUI for
  response buffering and streaming timeouts.
- **Environment changes do not apply:** Existing connections can be stored in
  Open WebUI's database. Update the saved connection in the Admin settings.

See [shared troubleshooting](https://docs.aur.lu/docs/inference/integrations#troubleshooting) and
[Open WebUI's compatible-provider guide](https://docs.openwebui.com/getting-started/quick-start/connect-a-provider/starting-with-openai-compatible).