Skip to content

KENZY

Kenzy is a voice assistant for your whole home that runs on your own hardware. You talk to it naturally — "hey Kenzie, turn off the lights downstairs, I'm ready for bed" — and it understands you the way a person would, because it's powered by a real language model (one you choose: a cloud provider or a model running locally). Your rooms share one assistant, it can control your smart home through Home Assistant, and it can recognize who is talking.

What you can do with it

  • Ask it things, naturally — no memorized phrases. "What time is it?", "Is it going to rain tomorrow?", "What's in the news?"
  • Control your smart home by voice — Kenzy connects to Home Assistant and understands your devices by the names you actually use ("the lamp by the chair").
  • Talk to the whole house — "tell everyone dinner's ready" speaks it in every room; "call the living room" starts a live two-way intercom (the other room has to say "yes" first).
  • Get answers that know who's asking — enroll your voice and Kenzy can tell family members apart, and can require a recognized voice for sensitive actions like unlocking a door.
  • Manage it all from a web page — the built-in dashboard shows every room and service, and is where you name rooms, tune microphones, and update everything with a click.

What it looks like in your home

One computer runs the "brain" (the server and the speech/language services). Each room gets a small, cheap device — a Raspberry Pi with a USB speakerphone — that listens for the wake word and plays the replies. You can also start with everything on a single computer to try it out.

Kenzy is a self-hosted project for people comfortable running Linux on their own machine. You don't need to be a programmer — the installer is one command and day-to-day management happens in the dashboard — but you should be comfortable opening a terminal and editing a text file. If that's you, you'll be talking to Kenzy in under an hour.

→ Start here: Getting Started

Under the hood (you don't need this to get started)

Kenzy is built as six small services that you can run together on one machine or spread across several:

Service Command Port Role
node kenzy-node — Room device: wake word, audio capture, playback
server kenzy-server 8765 The hub: connects rooms and runs the pipeline
stt kenzy-stt 8767 Speech-to-text (local faster-whisper, or OpenAI cloud)
tts kenzy-tts 8769 Text-to-speech (OpenAI, or local Kokoro)
llm kenzy-llm 8766 The language model + skills (any LiteLLM provider)
speaker kenzy-speaker 8768 Voice identification (runs locally)

Voice identification always runs on your hardware, and speech recognition does by default. Every stage is your choice — run the speech, language, and voice services fully local, use a cloud provider (the easiest start, and the lightest on your hardware), or mix and match per service. See Architecture for how it all fits together.