Sevak

Our Project aims to serve as a guardian against scam calls for people and it does so by analysing conversations in real time and also providing users with intelligent “trap-setting” replies in parallel to “test” the scammer and boost the confidence of being a scam call.

About

This is a college-level project made for learning and demonstration purposes. It allows users to:

• Leverages real-time interaction instead of just blocking known numbers.
• detect new, evolving scam tactics that keyword filters and number blacklists can’t.
• provides users with intelligent “trap-setting” replies.
• Scales easily as a service—browser extension, mobile app, or telco integration.
• Use a simple and modern UI with React and TailwindCSS

Currently, only local-host is supported. Working on deploying it without losing accuracy.

You’ll need Chrome browser to get the best experience.

Here’s a preview of the dApp:

Tech Stack

Frontend

• React.js
• Vite
• TailwindCSS

Backend

• Flask
• Flask-CORS
• Python
• Torch for deep learning
• Transformers for NLP models
• Requests for HTTP requests

Pre-requisites

Initialize a .env file with hugging face token as:

HF_TOKEN = hf_xYbODZ******

Backend Setup

1. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

2. Install required packages:

    pip install -r requirements.txt

Installation

npm install

Local Development

1. To run the backend:

   python Server.py

2. To run the frontend:

   npm run dev

Notes

• You need to configure .env with your Hugging Face token.
• For testing, use Chrome browser for audio processing and integration with your frontend.