# Chat in Workspace

Source: [https://docs.aventail.co.uk/docs/workspace/getting-started](https://docs.aventail.co.uk/docs/workspace/getting-started)

Chat with the model you already set up through Link. Keep it serving and open Workspace on the same Mac.

### Chat with your Link model

Workspace v0.2.0 MVP · Locai Link (Legacy Aventail) 1.3.2 · Gemma 4 (X-Small).

1.  **Connect your node**
    
    Open Workspace on the same Mac as Link. Sign in with the same account; skip the separate WebGPU download to use your Link model.
    
    [Screenshot: Model and conversation — close-up from Workspace chat.](https://docs.aventail.co.uk/img/guides/context/workspace.png)
    
2.  **Send a message**
    
    Send a short prompt and wait for a complete reply. The generated wording can vary.
    
    [Screenshot: Workspace conversation containing a completed reply from Gemma 4 (X-Small) through Link.](https://docs.aventail.co.uk/img/workspace/onboarding/first-response.jpg)
    
3.  **Try a document**
    
    Once chat works, attach a small supported document and ask a question about it.
    
    -   Attach a document: PDF, DOCX, TXT or MD
    -   Ask a focused question: Start with a small file
    -   Inspect the answer: Check it against the document

**Read the chat and document steps**

Use Workspace to chat with the model you set up through Link.

*Tested chat: Workspace v0.2.0 MVP · Link 1.3.2 Alpha · Gemma 4 (X-Small).*

## Connect your model

1.  Complete [setup and serving](https://docs.aventail.co.uk/docs/getting-started/quickstartguide#deploy). Keep Link running and open [Workspace](https://workspace.locai.co.uk/) **on the same Mac**.
2.  Sign in with the same Aventail account. If the getting-started guide offers a WebGPU model download, select **Skip setup** to use your existing Link model.
3.  Open **Settings** using the cog in the top-right corner. Under **Processing → Inference mode**, select **Your Locai node**, then close Settings.
4.  Open the model picker at the top of the conversation. Select **Your node**, check the device, and choose **Gemma 4 (X-Small)**. The header should show **live :8100**, or your chosen serving port.

**This device** runs a separate browser model through WebGPU. It needs its own download; it does not use the model installed by Link.

## Send a message

Enter a prompt such as “Reply with one short sentence confirming that you can answer a local test request.” Select **Send** and wait for the response.

A reply in the conversation confirms that chat is working. Generated wording can vary. Use **New chat** for another conversation.

## Work with a document

Select **Attach a document**, choose a PDF, DOCX, TXT or MD file, and ask a question about it. For size guidance, use the information button beside the attachment control or see [Document and connection errors](https://docs.aventail.co.uk/docs/workspace/troubleshooting-document-too-large).

## If your model is missing

-   Check Link is running and the model is serving in Control.
-   Confirm **Your node** is selected; **This device** lists browser models.
-   Refresh Workspace after starting serving if the picker has not updated.

To stop the model, turn off its switch under **Serving** in Control. To use it in another application, follow [Connect your software](https://docs.aventail.co.uk/docs/getting-started/run-first-interface).