Experimental — This integration is a proof of concept and has not been heavily tested.
A Home Assistant custom integration that adds a conversation agent and AI task entity powered by GitHub Models.
- Conversation agent — Chat with AI models, control Home Assistant via tool calling
- AI task agent — Structured data generation with attachment/vision support
- Model catalog — Automatically fetches available models from the GitHub Models API
- Recommended settings — One-click setup with sensible defaults, or customize model parameters
- Add this repository as a custom repository in HACS
- Search for "GitHub Models" and install it
- Restart Home Assistant
- Copy the
custom_components/github_models_conversationfolder into yourcustom_componentsdirectory - Restart Home Assistant
- Go to Settings > Devices & Services > Add Integration
- Search for GitHub Models
- Enter your GitHub personal access token
- Fine-grained tokens (recommended) — more secure, requires the
models:readpermission - Classic tokens — work without any additional scope
- Fine-grained tokens (recommended) — more secure, requires the
- A default conversation agent and AI task agent are created automatically
- Go to Settings > Voice assistants and select GitHub Models Conversation as the conversation agent
You can add more conversation agents or AI task agents from the integration page.
| Option | Description | Default |
|---|---|---|
| Instructions | System prompt for the AI (supports HA templates) | Default HA instructions |
| Control Home Assistant | LLM APIs to expose (tool calling) | Assist |
| Recommended | Use recommended model settings | Enabled |
| Option | Description | Default |
|---|---|---|
| Model | The AI model to use | openai/gpt-4o-mini |
| Temperature | Controls randomness (0-2) | 1.0 |
| Top P | Alternative to temperature (0-1) | 1.0 |
| Max Tokens | Maximum tokens in the response | 1024 |
Same advanced model settings as conversation agents. No system prompt or LLM API options.
GitHub Models comes with a free tier — no payment required. Any GitHub account can use it with a personal access token.
Rate limits for the free tier (Copilot Free):
| Tier | Requests/min | Requests/day | Tokens/request | Concurrent |
|---|---|---|---|---|
| Low models | 15 | 150 | 8k in / 4k out | 5 |
| High models | 10 | 50 | 8k in / 4k out | 2 |
| Embedding models | 15 | 150 | 64k | 5 |
These limits are subject to change. See the GitHub Models rate limits documentation for the latest details.