This Home Assistant integration automatically scans your entities, devices, and existing automations to suggest new YAML automations using various AI providers. It targets Home Assistant users who want to build more sophisticated smart home routines but struggle with the complexity of YAML or identifying potential automation opportunities. The benefit is faster creation of context-aware automations, reducing manual effort and unlocking more smart home potential.

How It Works

The integration takes a snapshot of your Home Assistant data (entities, devices, areas, automations), embeds it into a system prompt, and sends it to a chosen AI provider (OpenAI, Anthropic, Ollama, etc.). The AI generates automation suggestions, including human-readable descriptions and ready-to-paste YAML code, which are then surfaced as notifications and sensor attributes within Home Assistant. This approach leverages LLMs to overcome "YAML writer's block" and identify synergistic device interactions.

Quick Start & Requirements

• Install: Recommended via HACS (Home Assistant Community Store).
• Prerequisites: Home Assistant installation. API keys for cloud AI providers if used.
• Setup: Minimal, primarily configuration of AI provider credentials.
• Docs: HACS, Home Assistant

Highlighted Details

• Supports multiple AI back-ends: OpenAI, Anthropic, Google, Groq, LocalAI, Ollama, Mistral, Perplexity.
• Suggestions are delivered with human-readable descriptions and directly usable YAML blocks.
• Includes example automations for new-entity detection and weekly reviews.
• Service ai_automation_suggester.generate_suggestions allows manual triggering with configurable parameters.

Maintenance & Community

• Actively maintained with recent fixes for deprecation warnings (v1.3.1).
• Open to Issues, PRs, and translations. Contribution support via coffee donations.

Licensing & Compatibility

• MIT License.
• Compatible with Home Assistant and commercial use. Not affiliated with Home Assistant or AI providers.

Limitations & Caveats

Suggestions are AI-generated and require user review before implementation; the integration does not automatically execute automations. The quality of suggestions depends on the AI model used and the completeness of the Home Assistant snapshot.