# Open WebUI

Source: [https://docs.aventail.co.uk/docs/integration/openwebui](https://docs.aventail.co.uk/docs/integration/openwebui)

You need administrator access to Open WebUI and a Link endpoint reachable from its backend.

### Connect Open WebUI to Link

Screenshots from the original guide. Use your own model ID and port.

1.  **Open the admin panel**
    
    Open the profile menu and choose Admin Panel. You need administrator access to manage connections.
    
    [Screenshot: Profile menu and Admin Panel — close-up from Open WebUI chat page.](https://docs.aventail.co.uk/img/open-webuistep_5_.png)
    
2.  **Manage OpenAI connections**
    
    Open Settings → Connections, then manage the OpenAI API connections. Add or edit the connection for Link.
    
    [Screenshot: Open WebUI Settings and Connections, with the OpenAI connection controls highlighted.](https://docs.aventail.co.uk/img/open-webuistep_6.png)
    
3.  **Set the endpoint**
    
    Enter a URL reachable from the Open WebUI backend, ending in /v1. Use the written settings below, verify the connection and save.
    
    In Docker, use a host address reachable from the container instead of its own 127.0.0.1.
    
    [Screenshot: Open WebUI Edit Connection dialog showing its URL and authentication fields.](https://docs.aventail.co.uk/img/openstep7.png)
    
4.  **Select your model**
    
    Open a new chat and choose the model returned by your Link connection. The example model name may differ from yours.
    
    [Screenshot: Open WebUI model dropdown with a local model selected.](https://docs.aventail.co.uk/img/open-webuistep_8_.png)
    
5.  **Write your first message**
    
    With your Link model selected, type a short prompt in the message box and choose Send. You can also choose a suggested prompt.
    
    The crop shows the message controls. Keep the Link model selected from the previous step.
    
    [Screenshot: Open WebUI message box, Send arrow and suggested prompts.](https://docs.aventail.co.uk/img/open-webuistep_9.png)
    
6.  **Check a reply**
    
    Send a short prompt and inspect the answer before trying documents, tools or other capabilities.
    
    [Screenshot: Open WebUI chat displaying an answer from a configured model.](https://docs.aventail.co.uk/img/open-webui-setup-step6.png)
    

**Read the connection settings and Docker guidance**

You need a running Open WebUI instance with administrator access and a [tested Link endpoint](https://docs.aventail.co.uk/docs/getting-started/run-first-interface).

## Add a connection

Open the profile menu, then **Admin Panel → Settings → Connections → Manage OpenAI API Connections**, and add a connection.

| Field | Value |
| --- | --- |
| URL | `http://127.0.0.1:8100/v1` when the Open WebUI backend runs directly on the Link machine |
| API key | Leave empty for an unauthenticated endpoint, or use its configured credential |
| Model | Select the exact ID returned by the connection, for example `Gemma 4 (X-Small)` |

Use **Verify Connection**, then save. Verification requests the model list; saving alone does not test connectivity. This follows [Open WebUI's OpenAI-compatible connection guide](https://docs.openwebui.com/getting-started/quick-start/connect-a-provider/starting-with-openai-compatible/).

## If Open WebUI runs in Docker

The backend connects to Link. Inside a container, `127.0.0.1` refers to the container itself.

For Docker Desktop on the Link machine, use `http://host.docker.internal:8100/v1`. Linux Docker may require an explicit host-gateway mapping. For a separate server, use a reachable address with appropriate network access controls.

## Send a message

Open a new chat, select your Aventail model and send a short prompt. Confirm a complete response.

If verification fails, check serving status, port and connectivity **from the backend's environment**. If verification succeeds but chat fails, inspect the error and model ID. Start with text chat before testing uploads, tools or other model capabilities.