PRIVATE BY DESIGN

Your Home's Local AI Brain.

A local AI smart home assistant that manages your home, remembers your preferences and understands nuanced requests.

See Features GitHub
JARVIS
CPU
GPU
Hide Chat
Turn on the studio lights.
Studio lights enabled. Scene set to 'Creative'.
| Direct Interface...
IDLE

How It Works

You talk to it, it reasons, executes and replies.

Listen

Audio capture and speech-to-text run locally on your Mac.

Details

Audio & STT

Uses OpenAI Whisper locally to transcribe audio in near real-time.

Think

Jarvis understands context, memory, and intent.

Details

Agent Routing

Intent Agent classifies the request. Action Agent produces strict JSON for Home Assistant. Response Agent explains outcomes. Memory provides retrieval (RAG) when needed.

Act

Lights flip. Music plays. Scenes change. All via Home Assistant.

Details

Home Assistant Control

Sends authenticated HTTP calls to the Home Assistant API to trigger services, update entities, and read states in real time.

Private by Default

Jarvis is local by default. No telemetry. No background data collection. Cloud models are optional and only used when you explicitly enable a provider.

Local Mode

Recommended

Runs fully on your Mac. Works offline. Your home context stays on your network.

  • Ollama
  • Open Source LLM's
  • 100% Private
Details

Ollama & Offline Models

Jarvis uses Ollama as the primary engine for local intelligence. You can download and run state-of-the-art models like Llama, Mistral, or Qwen with zero internet connection once pulled.

Cloud Mode

Optional

Optionally use cloud models for heavy reasoning. When enabled, requests are sent to the selected provider under their data policy.

  • OpenAI
  • Gemini
  • OpenCode
Details

Cloud Providers

OpenCode offers a variety of powerful models that are free to use without any account creation. For enterprise-grade models from Gemini or OpenAI, you can easily connect your own API keys in the system settings.

Who this is for

Jarvis is for people who want to use cutting-edge AI without outsourcing their private life.

For those who want to integrate AI into daily routines while keeping personal data out of centralized data centers.

For people who appreciate automation, but have outgrown brittle, opaque systems like Alexa and similar assistants.

Built for users who want systems that are capable, private, and actually understand context.

Get Started

  1. Clone the repository from GitHub.
  2. Open your terminal and navigate to the project folder.
  3. Run the installer command:
$ ./scripts/start.sh

This script configures the environment and launches the reference system locally.