What is JARVIS?
RAWA DEV JARVIS is an open-source voice agent for Windows. You speak naturally; JARVIS replies in spoken audio and can trigger real actions on your machine — open apps, search YouTube, check the weather, run workflows, and more.
Unlike a chat window you type into, JARVIS is built for hands-free use: push-to-talk or continuous conversation, low latency, and a personality tuned for helpful, concise spoken answers.
What can it do?
- Real-time voice — Gemini Live audio in both directions
- YouTube — search and play music or videos on command
- Open apps & files — launch programs and paths on Windows
- Weather & web — quick answers without leaving the desk
- Tool calling — ~30 structured functions the model can invoke
- Secure auth — email signup with OTP; daily voice quota per user
- No user API keys — the server proxies Gemini; keys stay on the backend
How it works
JARVIS splits cleanly into two parts:
- Desktop app (open source) — captures microphone audio, plays responses, executes tool calls locally on Windows, and talks to the hosted API over WebSocket.
- Voice API on rawandios.dev — handles accounts, quotas, and proxies Gemini Live so the client never sees a Gemini key.
Typical flow:
register → verify email → login → /v1/live/start → WebSocket /v1/live
Open source
The desktop client is published as an open-source archive. Download the zip, explore the code, fork it, or adapt it for your own assistant stack.
The hosted voice API remains on rawandios.dev so end users get auth, rate limits, and a managed Gemini connection out of the box.