music

Personal music rating and playlist generation system that combines iTunes listening history with Pitchfork reviews.

What it does

• Computes a personal engagement score per album based on your play/skip behavior
• Matches your library against ~29k Pitchfork reviews
• Surfaces insights: your top albums, critic picks you've ignored, hidden gems, albums worth revisiting
• Generates mood-based playlists using k-means clustering on genre, listening behavior, and review text sentiment
• Exports playlists as M3U files for import into Music.app

Setup

1. Export your iTunes library: Music.app → File → Library → Export Library → save to data/library/
2. Run the Pitchfork scraper to collect reviews: python scraper.py
3. Analyze audio features: python audio_features.py
4. Generate playlists: python playlists.py
5. Launch the web app: streamlit run app.py

Commands

python playlists.py                     # generate playlists (k=9 default)
python playlists.py --k 6 --size 30    # custom k and playlist size
python playlists.py --optimize         # find optimal k via silhouette score
python export_playlist.py               # list available playlists
python export_playlist.py "Alternative" # export one playlist as M3U
python export_playlist.py --all         # export all playlists
python audio_features.py               # analyze audio features (run once)
python scraper.py                       # scrape new Pitchfork reviews
streamlit run app.py                    # web UI

Data

All data lives in data/ (gitignored):

• data/library/ — iTunes XML exports (drop new exports here, latest is auto-selected)
• data/reviews/ — Pitchfork review CSVs
• data/playlists/ — generated playlist CSVs
• data/m3u/ — exported M3U files for Music.app