Spotify Vibe Analyzer

Two things Spotify doesn't give you:

1. Playlist mood: paste a public playlist or album and get its mood profile (radar chart, mood label, tempo, happiest/saddest/most energetic tracks).
2. Similar vibes: paste a track you love and get songs with the same listening feel, matched purely on audio features (energy, valence, danceability, acousticness, tempo…), not on what other people listen to.

A dependency-free Python API on Vercel serverless functions, plus a single-page demo.

How it works

Spotify's Web API no longer exposes audio features (Nov 2024) or playlist items for apps you don't own (Feb 2026 Development Mode changes), so:

• Track lists and metadata come from Spotify's public embed pages. No credentials needed.
• Audio features come from ReccoBeats, looked up by Spotify track ID. If a track isn't in their database, its 30-second preview is sent to ReccoBeats' audio analysis instead.
• Similar vibes is content-based: ReccoBeats recommendations are used only as a candidate pool. Every candidate is re-ranked by weighted distance between audio features (valence and energy weigh most; tempo is compared octave-aware, so 70 and 140 BPM count as close). A second round seeds from the best matches to move the pool closer to the original vibe.

API

Full reference: docs/API.md.

Both endpoints are GET, return JSON, send CORS headers, and are cached at Vercel's CDN for 6 hours. Errors return { "error": "..." } with status 400/404/422/502.

GET /api/analyze?url=<playlist or album link>

{
  "playlist": { "id": "…", "type": "playlist", "name": "Today's Top Hits", "owner": "Spotify", "image": "https://…", "url": "https://open.spotify.com/playlist/…" },
  "trackCount": 50,
  "analyzedCount": 24,
  "mood": "Happy & Energetic",
  "features": [{ "key": "danceability", "label": "Danceable", "value": 0.635 }, "…"],
  "tempo": 125,
  "loudness": -7.142,
  "highlights": { "happiest": { "id": "…", "title": "…", "artist": "…", "value": 0.96 }, "saddest": {}, "mostEnergetic": {}, "mostDanceable": {} },
  "tracks": [{ "id": "…", "title": "…", "artist": "…", "danceability": 0.5, "energy": 0.5, "…": 0 }]
}

GET /api/similar?url=<track link>&limit=10 (limit 1–25)

{
  "seed": { "id": "…", "title": "Knights of Cydonia", "artist": "Muse", "image": "https://…", "url": "…", "features": { "energy": 0.96, "…": 0 } },
  "labels": { "energy": "Energy Level", "…": "…" },
  "tracks": [{ "id": "…", "title": "Drain You", "artist": "Nirvana", "url": "…", "similarity": 96, "features": { "…": 0 } }]
}

Deploy

npm i -g vercel
vercel          # preview
vercel --prod   # production

Optional environment variable:

Name	Example	Default
ALLOWED_ORIGINS	https://naufaldi.vercel.app,http://localhost:3000	* (any origin)

Limitations

• The embed page lists at most ~100 tracks per playlist.
• ReccoBeats doesn't know every track, so analyzedCount can be lower than trackCount. Very new releases are the most likely to be missing.
• Audio features describe feel, not genre. A rock song can match a pop song with the same energy, mood and tempo.
• Both data sources are unofficial and could change without notice.

Development

python test_api.py   # offline logic checks
vercel dev           # run the API + demo locally