# Use Genesis models in WorkBuddy

To use a Genesis model in WorkBuddy, add a custom model with its URL, API key, and model ID, then select it for a task. This tutorial connects the **WorkBuddy desktop app** to **Kimi K3 (`kimi-k3`)** and checks a conversation.

Before starting, install and sign in to [WorkBuddy](https://www.workbuddy.cn/) and prepare an MOI account with access to Genesis.

## 1. Create an API token in MOI

1. Click your avatar in MOI and select **Access Credentials**.

   ```{figure} images/genesis-access-credentials-entry.png
   :figclass: mo-shot
   :alt: Open Access Credentials from the MOI avatar menu
   ```

2. Choose **Personal access tokens**, then **Create token**. Use a name such as `genesis-workbuddy`, select an expiration, and create it.
3. Copy and save the full token. This is the **API key** you will enter later.

You can also use an existing valid personal access token.

## 2. Copy the model ID and connection URL from Genesis

1. In Genesis, open **Models**, search for `kimi-k3`, and open Kimi K3's details.
2. Click **Copy Model ID**. This example uses `kimi-k3`; keep the lowercase letters and hyphen.

   ```{figure} images/workbuddy-kimi-k3-model-details.png
   :figclass: mo-shot
   :alt: Kimi K3 model ID and API request example in Genesis
   ```

3. Open **Usage** and copy the Base URL under **Connection information**. This example uses:

   ```text
   https://token.moi.matrixorigin.cn/v1
   ```

4. Under **API usage**, check the **Chat Completions** example for this model. The complete request URL is `https://token.moi.matrixorigin.cn/v1/chat/completions`.

You now have the API key, Base URL, and model ID. WorkBuddy's custom integration uses Chat Completions. Enter these values as shown below.

## 3. Add a custom model in WorkBuddy

1. Open WorkBuddy, click your avatar, and go to **System settings → Models**.
2. Under **Custom models**, click **Add model** and choose **Custom** as the provider.
3. Fill in:

   | Field | Value |
   | --- | --- |
   | Provider | **Custom** |
   | Endpoint | `https://token.moi.matrixorigin.cn/v1` |
   | API Key | The full token from step 1 |
   | Model Name | `kimi-k3` |
   | Advanced Settings | Select **Tool Calling**; leave **Image Input**, **Reasoning**, and **Custom Protocol** cleared |

4. Click **Test connection**. After **Connection succeeded** appears, click **Save**.

   ```{figure} images/workbuddy-add-kimi-k3-model.png
   :figclass: mo-shot
   :alt: Genesis Kimi K3 settings and a successful connection test in WorkBuddy
   ```

WorkBuddy appends `/chat/completions` to the Endpoint, so do not enter the full request URL here.

Labels may vary by version. The conversation model picker also links to custom model settings. See [WorkBuddy's model configuration guide](https://www.workbuddy.ai/docs/workbuddy/From-Beginner-to-Expert-Guide/Function-Description/Model-Configuration).

## 4. Start a task with Kimi K3

1. Create a new task and open the model picker in the input area.
2. Under **Custom models**, choose the `kimi-k3` entry you added. If a built-in Kimi model is also listed, choose your custom entry.

   ```{figure} images/workbuddy-select-kimi-k3-model.png
   :figclass: mo-shot
   :alt: Selecting the custom kimi-k3 model in WorkBuddy
   ```

3. Send:

   ```text
   Turn these project updates into three short weekly-report bullets:
   The sign-in page design is complete. Data import integration is complete,
   but large file uploads occasionally time out. Next week we will investigate
   the timeouts and need the backend team to provide request logs.
   ```

4. Send the prompt and confirm that WorkBuddy returns the weekly report.

   ```{figure} images/workbuddy-kimi-k3-response.png
   :figclass: mo-shot
   :alt: WorkBuddy returning an English weekly report with the custom kimi-k3 model

   Kimi K3 returns the formatted weekly report in WorkBuddy.
   ```

5. In Genesis, open **Logs** and confirm that the request used `kimi-k3`.

You have connected WorkBuddy to Genesis and completed a conversation with Kimi K3.

For later tasks, select this custom model in the picker. To use a built-in WorkBuddy model, select its built-in entry instead.
