Home Generative Agent
A self-hosted Home Assistant AI agent controls devices, creates automations and analyzes cameras with local Ollama models or cloud providers.

Home Generative Agent is a Home Assistant integration for people who want to control their smart home through conversation and get alerts about security issues. It can use fully local models, so home data doesn't have to leave your network.
- Device control and history: Control entities, check their status and ask about past events or energy consumption using Home Assistant's recorded history.
- Automation creation: Describe an automation in plain language and the agent adds it to Home Assistant's automation editor after checking entity IDs and service calls; an optional Critical Action PIN gates actions such as unlocking doors or opening garages.
- Camera analysis: Ask about camera images or trigger analysis after motion events, with support for Axis, Reolink, UniFi Protect, Ring through ring-mqtt and compatible Home Assistant cameras.
- Security alerts and audits: Sentinel uses fixed rules to detect open entries, unlocked locks and other issues, sends phone alerts and audits accounts, tokens, exposed services and connected devices; the LLM advises but doesn't execute Sentinel actions.
- Memory and voice: Remember preferences across conversations and support optional speech input and output through local Speaches servers or OpenAI's audio APIs.
The integration is open source under MIT and requires Home Assistant and PostgreSQL with pgvector. Model choices include Ollama, OpenAI-compatible servers, OpenAI, Gemini and Anthropic. Local model and speech services can keep those requests at home; cloud providers receive requests for the capabilities you assign to them. Face recognition and alerts about unknown people require a separate face-service, with face recognition disabled by default.