JukeBox

JukeBox is a music recommendation web app built around a pre train collaborative filtering model trained with with LightFM on Lastfm-VADS dataset using React and shadcn components for a modern and smooth frontend.

This app supports discovering, playing audio, viewing lyrics and album art for over 40,000+ tracks. It uses different recommendation strategies to adapt to different users interactions (cold/warm starts)

What’s inside

• Recommender: LightFM-based pure collabaritive filtering model
• Backend: Flask REST API + SQLite for storage
• Frontend: React UI (with shadcn/ui)
• Integrations: Genius (lyrics/metadata), YouTube (streaming via youtube-dl)

Getting started

1) Backend

pip install -r requirements.txt
python -m app.api

2) Frontend

# from frontend directory 
npm install
npm run dev

To do

• Integrate genius and youtube services
• Reliable playlist/queue system for autoplay
• Sentimental info based mood visualizer
• Updating the search system to a non-linear index based search