Aurelius - Debate Simulator and Mentor

An AI-powered debate simulator with smart evaluation and personality-driven opponents, designed to sharpen critical thinking and argumentation skills.

Comment

Inspiration

As someone who's participated in several debates during school and college, I’ve always enjoyed the challenge of forming strong arguments and thinking on my feet. But one problem I kept running into was that preparing for debates on my own wasn’t enough—there was no one to push back, test my reasoning, or help me improve under pressure.

That’s where the idea for this project came from. After completing my first year in Electronics and Telecommunication Engineering (ExTC), I spent my summer break learning web development and exploring AI tools. I decided to build an AI-powered debate simulator that could act as a personal sparring partner—one that not only responds intelligently but also evaluates your performance and helps you grow over time.

What it does

Unlike basic chatbot-based debate tools, this application focuses on real-time, intelligent interaction and structured performance feedback. Here are the standout features that set it apart:

Gemini-Powered AI Opponents: Uses Google’s Gemini API for dynamic, intelligent responses—no templated or scripted replies.

AI Personalities: Choose between Aggressive, Assertive, and Passive-Aggressive debating styles to simulate diverse opponents.

Smart Evaluation Engine: Automatically scores your debate based on:

Source credibility (e.g., UN, WHO = 3 points)

Articulation and language sophistication

Speaking time and participation balance

Real-Time Feedback: After every debate, receive a clear score breakdown and personalized improvement tips.

Transcript System: A new addition that lets you review the full conversation after each session—great for reflection and learning.

Fully Responsive UI: Works smoothly on desktop and mobile, with a redesigned, intuitive interface.

How we built it

This app was built entirely during the hackathon using the following technologies:

Frontend Framework: React 18 with TypeScript

Styling: Tailwind CSS

AI Integration: Google Gemini API via Google AI Studio

Voice Features (initial version): Web Speech API (speech recognition and text-to-speech)

Build Tool: Vite for fast reload and development builds

Environment Config: .env files for secure API key storage

Challenges we ran into

Challenges Faced Managing Debate Flow Logic: Getting the debate to run in the correct order (user → AI → user) without the AI responding twice was a major challenge. It required fine-tuned state management using React hooks and careful control of transitions between rounds.

State and Side Effects: As a beginner, handling useEffect dependencies and avoiding unwanted re-renders took some learning.

Prompt Structuring: Making the AI give high-quality, contextual arguments consistently took a lot of trial and error with prompt engineering.

UI Redesign: I had to completely overhaul the user interface to improve the flow and make the experience more intuitive and enjoyable.

Accomplishments that we're proud of

uilt an end-to-end AI debate simulator from scratch Using Gemini API, React, and TypeScript, the app went from concept to functional prototype within a limited time.

Successfully implemented structured debate flow Created a clear, round-based debate system with opening statements, rebuttals, and closing arguments.

Designed AI personalities for realistic practice Added Aggressive, Passive-Aggressive, and Assertive modes to simulate different debating styles and train adaptability.

Created a smart evaluation system Built a scoring engine that measures articulation, source credibility, and participation—providing feedback that users can actually learn from.

Implemented a real-time transcript feature Added a feature that logs debates for later review, helping users reflect and analyze their performance.

Improved UX with a complete UI overhaul Redesigned the interface to feel more polished, responsive, and intuitive across devices.

What we learned

React State Management: I gained a solid understanding of how to structure state and control the flow in multi-step interactions.

Prompt Engineering: I learned how small changes in prompt structure can dramatically affect AI responses.

Debugging and Testing: I became better at isolating bugs, testing edge cases, and creating fallback logic for voice and AI errors.

Full Project Lifecycle: From idea to deployment, I learned how to take a project through setup, design, implementation, polish, and presentation.

What's next for Aurelius - Debate Simulator and Mentor

Topic Suggestion Engine Add AI-generated debate topics based on current events, user skill level, or trending themes.

Leaderboard & Ranking System Introduce a competitive layer by ranking users based on debate performance over multiple sessions.

Multiplayer or Peer Debates (Stretch Goal) Enable real-time debates between users with AI as moderator and evaluator.

Advanced AI Personalities Fine-tune personality behavior further—e.g., sarcastic, highly logical, or emotionally charged AI debaters.

Built With

Share this project:

Updates