acon96/home-llm

What it solves

Home LLM allows users to control their Home Assistant smart home using a Large Language Model (LLM) that runs entirely locally. This eliminates the need for cloud subscriptions and ensures that sensitive home data remains private and under the user's control.

How it works

The project consists of two primary components: a Home Assistant custom integration and a set of specialized AI models. The integration acts as a bridge, connecting the smart home's devices to an LLM backend (such as Llama.cpp, Ollama, or OpenAI-compatible servers). The "Home" models are small, efficient LLMs (under 5B parameters) specifically fine-tuned to translate natural language requests into actionable Home Assistant service calls.

Who it’s for

Smart home enthusiasts using Home Assistant who want AI-powered voice or chat control without relying on external cloud providers, particularly those running on modest hardware like a Raspberry Pi.

Highlights

• Privacy-First: Fully local execution ensures data never leaves the local network.
• Low Hardware Requirements: Optimized models can run on low-power devices without a GPU.
• Flexible Backends: Supports various local engines (Llama.cpp, Ollama) and external APIs (OpenAI, Anthropic).
• Broad Device Support: Controls lights, switches, fans, locks, climate, vacuums, and more.
• AI Task Automation: Can be used to generate structured data and dynamic content for home automations.
• Multilingual: Built-in support for English, German, French, Spanish, and Polish.