# AMD Lemonade (https://www.librechat.ai/docs/configuration/librechat_yaml/ai_endpoints/lemonade)

AMD Lemonade runs models locally through an OpenAI-compatible API. Follow the [Lemonade documentation](https://lemonade-server.ai/docs/) to install and start the server, then download a chat model.

## Configuration

Lemonade requires no API key; the placeholder below satisfies the OpenAI client. Add the endpoint under `endpoints.custom` in your `librechat.yaml`:

```yaml filename="librechat.yaml"
    - name: 'lemonade'
      apiKey: 'lemonade'
      baseURL: 'http://localhost:13305/v1'
      models:
        default: ['Qwen3-0.6B-GGUF']
        fetch: true
      titleConvo: true
      titleModel: 'current_model'
      modelDisplayLabel: 'AMD Lemonade'
      iconURL: 'lemonade'
```

## Notes

- Replace `Qwen3-0.6B-GGUF` with a locally installed chat model. With `fetch: true`, LibreChat retrieves the model list from Lemonade. `titleModel: 'current_model'` reuses the conversation's model for titles.
- For Docker, replace `localhost` with `host.docker.internal` or a reachable Lemonade hostname.
- For user-provided base URLs, add the local server's exact `host:port` (e.g. `host.docker.internal:13305`) to [`endpoints.allowedAddresses`](/docs/configuration/librechat_yaml/object_structure/config#endpoints).
- Built-in branding recognizes `lemonade`, `AMD Lemonade`, and `Lemonade Server`. Keep `iconURL: 'lemonade'` to preserve the logo when renaming the endpoint.