T.O.R.C.H

Thinking, Observing, Reasoning, Creating & Handling

Your personal AI agent that gives you complete control over your PC and online life through voice and text commands.

---

Features

• Full Desktop Control — Find files, send emails, post on social media, control the browser
• Hey TORCH — Siri-style floating overlay responds to voice even when minimized
• Agent Execution — Multi-step task execution with real-time progress streaming
• Human-in-the-Loop — Approval required for irreversible actions (email, delete, post)
• Screen Watcher — 24/7 screen observer that learns what you're doing
• Memory & Habits — Learns your patterns and predicts your needs
• 100% Local — Your data stays on your machine. No telemetry.

Tech Stack

Layer	Technology
Desktop Shell	Electron
Frontend	React 18, TypeScript, Tailwind CSS, Framer Motion, Zustand
Backend	Python 3.11+, FastAPI, WebSockets
AI Brain	Gemini 2.5 Flash (free tier)
Voice	OpenAI Whisper (local, offline)
Automation	PyAutoGUI, Playwright
Memory	SQLite + ChromaDB

Getting Started

Prerequisites

• Node.js 18+
• Python 3.11+
• npm or yarn

Frontend Setup

cd TORCH
npm install
npm run dev

Backend Setup

cd backend
pip install -r requirements.txt
python main.py

Configuration

1. Copy .env.example to .env
2. Add your Gemini API key from aistudio.google.com
3. (Optional) Add Gmail credentials for email features

Architecture

Frontend (Electron + React) ←→ WebSocket ←→ Backend (FastAPI + Python)
                                               ↓
                                          Gemini 2.5 Flash
                                               ↓
                                     Agent Tools (files, email, browser, etc.)

License

MIT