Sachet Open Source — GitHub Repository & Architecture

Repository Structure

The codebase is organized into modular components designed for reproducibility and easy extension:

datagen/

Scenario grid generator creating 17 scam categories, 14 hard negatives, 4 victim personas, and 20 psychological risk signals.

api/

FastAPI backend running MuRIL transformer scoring, regex signal parsing, ephemeral RAM session management, and Whisper STT.

eval/

Evaluation benchmarks comparing fine-tuned MuRIL against XLM-RoBERTa and Gemini baseline models.

client/

Lightweight, zero-eavesdropping dark civic interface with real-time radial risk meter, audio recording, and evidence export.

Local Quickstart Guide

Clone the repository and set up your local development environment:

1. Clone & Set Up Virtual Environment

git clone https://github.com/rizzesh/Sachet.git
cd Sachet

# Create and activate virtual environment
python -m venv venv
.\venv\Scripts\Activate.ps1  # On Windows PowerShell
# source venv/bin/activate   # On Linux/macOS

# Install dependencies
pip install -r requirements.txt
        

2. Launch the FastAPI Service

uvicorn api.main:app --host 127.0.0.1 --port 8000 --reload

3. Launch the Web Client

# Serve the client on port 5500
cd client
python -m http.server 5500
# Open http://localhost:5500 in your browser
        