Debate Guard 🛡️

Debate Guard is an adversarial AI debate simulator and coaching platform that acts as a live, adversarial opponent. It monitors your logical arguments and analyzes your body language (gaze, posture, shielding, etc.) in real-time, offering actionable feedback to help you become a better debater and speaker.

Features

• Live AI Debate: A fierce, fast-paced Gemini-powered opponent that challenges your logical arguments relentlessly.
• Coach Mode: A supportive persona that weaves in logical feedback and body-language nudges organically.
• Real-Time Body Language Heuristics: Client-side tracking using MediaPipe to monitor your:
  • Gaze (eye contact)
  • Slouching/Posture
  • Defensive Shielding (crossed arms)
  • Yaw Instability (looking side-to-side)
  • Self-Soothing (touching face/neck)
  • Swaying and Tilting
• Live Truth-Checking: The AI automatically uses Google Search to fact-check the statistics and claims you make on the fly.
• Comprehensive Post-Debate Analytics: Generates a highly detailed breakdown of your rhetorical effectiveness, logical fallacies, and physical stage presence.

Tech Stack

• Backend: Python + FastAPI + SQLModel + SQLite
• AI Integration: Google GenAI SDK (Gemini Live API)
• Frontend: React + Vite + TypeScript + TailwindCSS
• Real-Time Streaming: WebSockets (Audio and Video frames)
• Computer Vision: MediaPipe Vision API for Face/Pose landmarks

Setting Up and Running

The project includes a convenient runner script run.sh to handle dependencies and startup.

First-Time Setup

If this is your first time running Debate Guard, or if you need to install missing dependencies, start the project with the setup flag:

./run.sh -setup

This will automatically:

1. Initialize python dependencies (uv sync)
2. Install Node.js frontend dependencies (npm install)
3. Boot up the services

Standard Run

If you already have your dependencies installed, simply run:

./run.sh

If you try to run the project without dependencies installed, the script will gracefully exit and prompt you to use the -setup flag.

Services will be accessible at:

• Frontend App: http://localhost:5173
• Backend API: http://localhost:8000

Environment Variables

Make sure to configure your .env file in the backend/ directory:

GEMINI_API_KEY=your_gemini_api_key_here

Architecture

Below is a high-level representation of the Debate Guard architecture: