Personal Project · Python · Terminal Application
a personal terminal app for music discovery and movie roulette
I was inspired by Justin Li, a CS department head at Occidental College, who uses Python terminal programs for everyday tasks. I wanted to build something similarly practical but personal: something that felt like mine. I started by asking what I actually needed. A way to stop being indecisive about movies, and a way to genuinely find music I had never heard, not just cycle back to artists I already know.
Imports a Letterboxd watchlist CSV, spins a randomized selector, opens the film's page, and optionally removes it from the list.
Five discovery paths: surprise genre, freeform vibe/mood, own playlists, public playlists, and a quiet/obscure instrumental mode. Filters out known songs and familiar artists. Uses Spotify audio features (energy, valence, danceability) to rank results by mood match.
Screen recording walkthrough of movie roulette, song discovery modes, and the adaptive feedback loop.
Stack
Python, no heavy dependencies (stdlib + certifi only)
Auth
Spotify OAuth implemented from scratch
Audio Analysis
Audio feature scoring across 5 dimensions
Discovery Pipeline
Adaptive 4-stage discovery pipeline with fallback logic
Anti-Repeat
Session track window + artist rotation
Preferences
Persistent preference profile that evolves with feedback
Caching
Local JSON caching for liked songs, recent picks, and preferences
Browser
macOS browser integration (Opera same-tab behavior)
Every design decision in this project came from actual use: identifying where things broke, and going back in to fix them.