{Tech:Europe} Paris Hackathon

Yongkang Zou · 2025-09-28 · hackathon · Music & Audio AI

Turning commits into diss tracks

Need an elegant way to call out a colleague who doesn't contribute much code but still takes the credit? Write them a diss track.

At the Europe Paris AI Hackathon, I teamed up with Salim Boujaddi and Mounir Lbath to build exactly that. We created GitBeat, an interactive platform that converts your GitHub repository into music.

The concept is simple. You paste a GitHub URL. Our service reads through your code, analyzes the commit history, and parses the README to understand the project. Then it generates a unique song that sings about the actual work, the inspiration, and the team dynamics behind the repository.

GitBeat workflow diagram showing GitHub MCP, Dust.tt, Suno AI, and Supabase

The core workflow from GitHub URL to generated beat

To make this happen, we built a pipeline to process repository data and feed it into a music generation model. We used a Dust.tt Analysis Agent connected to a GitHub MCP (Model Context Protocol). This agent extracts the core themes from your repository and passes that context to the Suno AI API to generate the audio track.

GitBeat system architecture diagram

Frontend and backend architecture built around Next.js and Supabase

The application layer is built around a Next.js frontend with Tailwind UI. We split the backend logic into distinct API routes. Repository analysis is handled by the Dust Agent, music generation routes through Suno AI, and Supabase manages the state for our song library and voting system.

You can try the live demo at gitbeat.vercel.app or check out the source code on GitHub. Generate a track for your own repo, share it, and upvote your favorites on the leaderboard.

Photos

The GitBeat team at Europe Paris AI Hackathon

GitBeat interface showing the leaderboard