This is a submission for the Gemma 4 Challenge: Build with Gemma 4
🎙️ AI Debate Arena — Multi-Agent Critical Thinking Powered by Gemma 4
What I Built
Most AI tools generate a single response.
But real thinking happens when ideas are challenged.
So I built AI Debate Arena — a multi-agent AI system where multiple Gemma 4 powered agents debate opposite sides of any topic across several rounds, while a judge AI evaluates logic, persuasion, rebuttals, and clarity to declare the winner.
This project transforms passive AI answers into active reasoning battles that help users think critically, compare viewpoints, and make better decisions.
Demo
This is a CLI-based project
Code
💻 https://github.com/Sripadh-Sujith/debate-arena
How I Used Gemma 4
I used Gemma 4 via OpenRouter as the reasoning engine behind every agent in the system.
Model Choice
I chose Gemma 4 27B Dense because this project requires:
- Strong multi-turn reasoning
- Consistent roleplay between agents
- High quality rebuttals
- Good memory across rounds
- Balanced speed + intelligence
Gemma 4 Powers Four Agents
- 🎙️ Moderator — manages flow and rules
- 🟢 Pro — argues in favor n- 🔴 Con — argues against
- ⚖️ Judge — evaluates and decides winner
Why Gemma 4 Was Perfect
Gemma 4 excels at structured reasoning and contextual understanding, making it ideal for multi-agent workflows.
Instead of using separate models, one Gemma 4 model can convincingly perform multiple roles through prompt engineering.
Tech Stack
- Gemma 4 (OpenRouter)
- LangGraph
- LangChain
- Python
- Rich
Future Improvements
- Web UI
- Voice debates
- Audience voting
- Persona battles
- Export transcripts
- Debate Yourself mode
Closing Thoughts
I believe the future of AI is not one assistant giving one answer, but multiple intelligent agents reasoning together.
Built by Sripadh Sujith
Top comments (0)