JARVIS — Advanced AI Virtual Assistant

JARVIS is a futuristic, highly capable virtual assistant built with Electron and Python. It features a stunning "Stark Industries" HUD interface and leverages a multi-LLM architecture to provide real-time intelligence, system control, and automation.

🚀 System Architecture

1. Frontend (Electron HUD)

• Visuals: A premium, glassmorphism-inspired "Iron Man" HUD with real-time audio visualizers (Arc Reactor), system stat monitors, and cinematic animations.

• Speech Stack:

  • STT (Speech-to-Text): Supports Vosk (Local/Offline), Groq Whisper, and Sarvam.

  • TTS (Text-to-Speech): Integrated with Sarvam (Bulbul v3), Groq Orpheus, and native Web Speech API.

• Interaction: Features a "Clap to Wake" cinematic sequence and always-on voice listening.

2. Backend (Python MCP Server)

• Model Context Protocol (MCP): A dedicated Python server (mcp_server.py) provides JARVIS with "hands" to interact with the OS.

• Capabilities:

  • macOS Control: Application launching, volume control, screen locking, screenshots, and system info.

  • Web Intelligence: Web scraping for summarization, news fetching via RSS, and advanced Chrome control via AppleScript.

  • Workspace Automation: One-command setup for 'Coding', 'Research', 'Relax', and 'Web Dev' modes.

3. Intelligence Layer

• Groq (Llama 3.3/3.1): Used for sub-500ms intent detection and tool routing.

• Gemini 3 Flash: The primary conversational brain, providing high-intelligence responses with minimal latency.

• OpenRouter (Gemma 4): Fallback engine and advanced reasoning specialist.

Related MCP server: macos-control-mcp

🛠️ Setup & Installation

Prerequisites

• macOS (Optimized for Mac; some features may not work on Windows).

• Node.js (v18+)

• Python 3.10+

1. Clone & Install Dependencies

# Install JS dependencies
npm install

# Install Python dependencies
pip install fastmcp psutil feedparser requests beautifulsoup4

2. Environment Configuration

Create a .env file in the root directory and add your API keys:

GROQ_API_KEY=gsk_...
GEMINI_API_KEY=AIza...
OPENROUTER_API_KEY=sk-or-...
SARVAM_API_KEY=your_sarvam_key

3. Run the App

npm start

Note: On first run, it will download the Vosk model (~40MB) if not present in the models/ folder.

📦 Exporting & Distribution

To package the application into a standalone macOS .app or .dmg file:

npm run build

The build output will be located in the dist/ folder.

🔧 Core Features

• "Hey JARVIS": Start speaking anytime to interact.

• "Clap to Wake": A loud clap wakes JARVIS up with a cinematic intro sequence.

• Workspace Modes: Say "Setup coding workspace" to automatically open Terminal, VS Code, and relevant browser tabs.

• System Telemetry: Real-time monitoring of CPU, RAM, and Network on the HUD.

• Memory System: JARVIS remembers facts you tell it about yourself (name, profession, etc.) across sessions.

👨‍💻 Created By

Akshat Singh — Tech Creator & Developer. Designed to bring the future to the present.