JARVIS

Local Gemini Live desktop assistant with a PyQt6 interface, voice interaction, detachable panels, and optional browser, file, screen, and messaging tools.

JARVIS also includes a dedicated presentation studio that creates, edits, redesigns, and extends editable widescreen .pptx decks from documents, data, images, audio, and video, with optional PDF export. See the usage guide for examples.

Requirements

You need Python 3.11 or newer installed to set up and run JARVIS. Confirm your Python version before continuing:

python --version

Quick start (Windows, macOS, Linux)

In Terminal, run:

git clone https://github.com/MAL19INDUSTRIES/JARVIS-OS-V.2.git
cd JARVIS-OS-V.2
python scripts/setup_jarvis.py

On Windows, you can double-click scripts/setup_jarvis.bat instead.

Open .env, add your GEMINI_API_KEY, then launch JARVIS:

jarvis

You only need to run setup once. Activate .venv when opening a new terminal, then type jarvis.

JARVIS's core UI, Gemini connection, presentations, research, files, and CLI are cross-platform. Some computer-control, email, media, and browser integrations depend on permissions and available applications on each operating system.

Hosted web application

The repository also contains a multi-user FastAPI service and a Next.js web client. Hosted sessions use Postgres for user-scoped memory and configuration, Redis for request quotas, encrypted per-user Gemini keys, and Gemini Live over an authenticated WebSocket. The desktop launcher continues to use its local stores and full local action inventory.

Start the complete local web stack with Docker:

docker compose up --build

Then open http://localhost:3000. To run each service directly:

# API
cp .env.example .env
alembic upgrade head
uvicorn api.server:app --reload

# Web client
cd web
cp .env.example .env.local
npm install
npm run dev

Production templates are included for Fly.io (fly.toml), Render (render.yaml), and Vercel (web/vercel.json). Configure DATABASE_URL, REDIS_URL, JWT_SECRET, JARVIS_ENCRYPTION_KEY, and CORS_ORIGINS on the API host. Configure NEXT_PUBLIC_API_URL and NEXT_PUBLIC_WS_URL on Vercel. The deployment workflow runs manually after the Fly and Vercel repository secrets have been added.

Manual setup

python3 -m venv .venv
source .venv/bin/activate
python3 -m pip install --upgrade pip
python3 -m pip install -r requirements.txt
cp .env.example .env
./scripts/install_jarvis_cli.sh
jarvis

Set GEMINI_API_KEY in .env before launch. Optional settings such as voice and local API keys are documented in .env.example.

Launch with jarvis

The CLI launcher is included in this repository. After cloning and completing the one-time setup, install it for your user with:

./scripts/install_jarvis_cli.sh

Open a new terminal (or reload your shell profile), then start JARVIS with:

jarvis

Before packaging or releasing the desktop app, run the side-effect-safe capability audit:

jarvis --self-test

The audit exercises voice/tool contracts, messaging routing and approval boundaries, a local browser interaction, isolated file operations, vision, agent recovery, and memory. It never sends a real message or performs a live desktop mutation. Results that still require a person, account, or physical device are labeled LIVE CHECK REQUIRED, and a JSON report is written under .qa-artifacts/.

Alternatively, from an activated virtual environment, python3 -m pip install -e . installs the same jarvis command through the standard Python package entry point.

Documentation

• Usage guide
• Tutorial
• Complete QA and bug-audit guide
• Contribution notes

Configuration files

Template files are included for local setup:

• .env.example
• config/api_keys.example.json
• config/layout_settings.example.json
• config/ui_settings.example.json
• memory/long_term.example.json
• memory/task_history.example.json

Publishing checklist

• Keep .env and local secret files out of git.
• Do not commit memory/long_term.json or config/api_keys.json.
• Run python3 -m py_compile main.py ui.py before tagging a release.

License

MIT License, see LICENSE.