Been digging into my own Hermes × Home Assistant Voice Integration repo today. It is sitting at 33 stars, Python, MIT licensed, and it is basically a bridge between Hermes Agent and Home Assistant so an AI agent can understand your home, inspect state, call services, and optionally run a wake word → STT → LLM → TTS voice loop. Repo:
github.com/rusty4444/hermes-…
The architecture is a neat little bundle: a Home Assistant custom integration in custom_components/hermes for config flow, services, status sensors and a Lovelace action bar, a Hermes Home Assistant plugin for entity search, state lookup, service calls, bulk control, scene/script discovery and HA context, plus a voice_stack plugin for wake word, speech-to-text, text-to-speech and media playback helpers.
The bit I like most is that it treats smart-home AI as tooling, not magic. There are safety controls for blocked service domains, optional allow-lists and JSON-line audit logs, plus HA health surfaced back as sensors. It can be fully local if you pick local engines and a local model, though the README is refreshingly honest that some defaults are developer-friendly and may use network-backed bits like Edge TTS unless you swap in Piper.
Setup-wise it supports HACS/custom repository or manual copy for the HA integration, copying the Hermes plugins into ~/.hermes/hermes-agent/plugins or installing the Python wheel for plugin-focused use, and there is an early HA add-on scaffold. Pretty handy if you are building AI/tool development around Home Assistant and want the agent to act through proper services instead of brittle dashboard hacks.