# taraassistant-public

> Self-hosted AI that learns home patterns and generates Home Assistant automations in YAML for approval, running locally with optional LLM providers.

- **Category:** Home Automation
- **Licence:** Apache-2.0
- **Latest version:** 1.0.0
- **Open source:** oss
- **Languages:** Typescript, Python
- **Platforms:** Docker, Docker Compose, Windows, macOS
- **Tags:** Automation, Local First, LLM, Home Automation, Home Assistant, Yaml, AI, Self Hosted, Smart Home, Pattern detection

## Description

Tara Assistant is a self-hosted service that observes your Home Assistant events to detect recurring patterns in your smart home usage. When it finds consistent routines — like lights turning on at sunset or doors locking at night — it drafts Home Assistant automations in plain YAML and presents them for your approval before applying them. It runs locally as a Docker container, stores data on your own volume, and encrypts credentials. Optional LLM providers (OpenAI, Anthropic, Google, Ollama) can handle complex natural language commands, while simple intents execute directly via Home Assistant services. A safety layer classifies actions and requests confirmation for sensitive operations like locking doors. Home Assistant remains the control plane; Tara only suggests automations you can inspect, edit, version, and delete. The project is early-stage with conservative pattern detection and a browser-based setup wizard.

## Links

- [Homepage](https://github.com/TaraHome/taraassistant-public)
- [Docker](ghcr.io/tarahome/taraassistant)
- [GitHub Repository](https://github.com/TaraHome/taraassistant-public)
- [GitHub Issues](https://github.com/TaraHome/taraassistant-public/issues)
- [GitHub Discussions](https://github.com/TaraHome/taraassistant-public/discussions)

---
Source: https://selfhost.directory/project/taraassistant-public