# Plugins

Custom client features can be distributed as a npm plugin.

- `@agent-smith/feat-git` a command to use a language model to generate a commit message.

Install:

```bash
npm i -g @agent-smith/feat-git
```

Add the plugin to your `config.yml` file

```yml
plugins:
  - "@agent-smith/feat-git"
```

Update your client config by running this command:

```bash
lm conf ~/lm/features/config.yml
```

To use the plugin run the `commit` command in a git repository:

```bash
lm commit
# or
lm commit .
# or
lm commit --msg "The first line of the commit" # the ai will provide the details
```

The default model used is qwencoder. To use another model:

```bash
lm commit -m mymodelname
```

## Create a plugin

Create an npm package with a `package.json`:

```json
{
  "name": "agent-smith-myplugin",
  "version": "0.0.1",
  "description": "Description of the plugin",
  "type": "module"
}
```

Use the standard directory structure for your features in a `dist` directory:

- actions
- commands
- workflows
- tasks

All directories are optional. For the development declare your features as a local folder:

```yml
features:
  - /home/me/lm/agent-smith-myplugin/dist
```

Update your client config by running this command:

```bash
lm conf ~/lm/features/config.yml
```

Once your features completed and the plugin published to npm install it
globally:

```bash
npm i -g agent-smith-myplugin
```

Declare it as a plugin in your cli config:


```yml
plugins:
  - "agent-smith-myplugin"
```

Run the `lm conf` command to update

<a href="javascript:openLink('/plugins/inference')">Next: Inference</a>