Quick start

Install, configure with the wizard, and talk to Watari in about 5 minutes.

Requires Python 3.11+ and uv. For a multi-device deployment you also need Tailscale on every device.

1 · Clone & install

git clone https://github.com/iamvazghen/OpenWatari openwatari
cd openwatari

# Install the base + the stack you want (also installs the `jarvis-setup` CLI):
uv sync --extra edge --extra cloud-voice --extra brain --extra channels --extra identity --extra dev
#   for a 100% local/offline voice stack: drop cloud-voice, add local-voice

Note. uv sync prunes extras you don't list — install the full set you intend to use in one command.

2 · Run the setup wizard

uv run jarvis-setup

A terminal UI (Rich; falls back to plain text) walks you through the decisions and writes a ready .env. It never prints a secret back, backs up an existing .env, and auto-generates your brain auth token and all eight protocol passwords.

Step	What it sets
Identity	display name + wake phrase
Voice	cloud (ElevenLabs + Deepgram) or local (Piper/Kokoro + Whisper)
Brain	LLM backend: free proxy, OpenAI, or local Ollama
Knowledge	path to your Obsidian/Markdown notes (L3 memory)
Deployment	`single` (loopback) or `vps` (binds 0.0.0.0 + token)
Security	generates strong protocol passwords
Integrations	optional: Telegram, Tavily, Google, Notion, ntfy
Behaviour	proactivity on/off; fleet stays off by default

3 · Verify

uv run python bench/run_all_tests.py     # the single gate — should be all green
uv run python bench/efficiency_report.py # hot-path latency vs targets

Publish gate. Keep your repo private until the suite is green and the efficiency report meets its targets.

4 · Talk to it

# Local voice loop on this machine:
uv run python -m jarvis.edge.assistant

# …or the shared 24/7 brain (for phone/glasses, over your tailnet):
uv run python -m jarvis.brain.server

5 · Finish setup

TODO-NOW.md in the repo is the full deployment checklist: Tailscale on every device, voice enrollment, the recurring-reminder ticker, Google/Notion/Telegram logins, and a device-by-device acceptance test (TTFW, barge-in, Siri voice, music room, proactive voice, shared memory). Then make it yours by editing personality/jarvis.md.