Oracle of Music

Find the shortest chain of collaborations linking any two music artists, like the Oracle of Bacon (Six Degrees of Kevin Bacon), but for music.

Live at oracleofmusic.app: currently covers the top 5,000 Spotify artists and everyone they've collaborated with.

Built with: Python · Flask · SQLAlchemy · PostgreSQL · Docker · Gunicorn · Nginx · vanilla JS/HTML/CSS · MusicBrainz · Spotify Web API

How it works

flowchart LR
    A[MusicBrainz dump] --> B[(PostgreSQL)]
    B --> C[Adjacency list]
    C --> D[BFS]
    D --> E[Spotify Web API]
    E --> F[HTML/CSS Path]

• Self-hosted data: a PostgreSQL database built from the full MusicBrainz data dump, running in Docker.
• Clean collaborations only: tracks are filtered to official albums, singles and EPs, excluding bootlegs and unofficial releases.
• Fast search: collaborations are worked out once and cached in the database, then searched in memory with breadth-first search, so each Path is the shortest one possible.
• Lazy Spotify data: artist pictures, album art and Spotify links load after the Path is shown, one request per item.
• Typeahead search: accent-insensitive artist suggestions that also match alternative names.
• Accessible Path display: keyboard-focusable, respects reduced-motion settings, and switches to a vertical layout on narrow screens.

Running locally

Prerequisites

• Python 3.13+
• Docker
• About 40 GB of free disk space for the database
• A Spotify developer app (client ID and secret)

1. Build the database

Follow deploy_database/instructions.txt to start PostgreSQL in Docker and load the MusicBrainz data dump into it.

2. Configure the app

Create a .env file in the repo root:

Variable	Description
DATABASE_URL	SQLAlchemy URL for the MusicBrainz database, e.g. postgresql+psycopg://postgres:<password>@localhost:5432/musicbrainz
SPOTIFY_CLIENT_ID	Spotify app client ID
SPOTIFY_CLIENT_SECRET	Spotify app client secret
ARTIST_NAMES_FILE_PATH	Text file with one artist name per line, most important first
ARTISTS_TO_LOAD	How many names to read from that file. The app finds Paths between these artists and everyone they've collaborated with. The live site uses the top 5,000 Spotify artists.

3. Install and run

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

cd app
gunicorn app:app        # or: flask --app app run

The first start builds the collaboration graph and caches it in the database. Later starts reuse the cache.

Tests

The tests don't need a database or Spotify credentials:

pytest

Acknowledgements

• MusicBrainz for the artist and track data
• Spotify Web API for artist pictures, album art and links
• mbdata, which the database build scripts are adapted from

License

MIT