Sunnify

Spotify has no "download my playlist" path, and the usual scripts break once a playlist passes a few hundred tracks or hits a rate limit. Sunnify handles the full pipeline from URL to tagged file.

The hard part is the data layer. Instead of an official API, it parses the __NEXT_DATA__ JSON blob from Spotify's public embed pages, using layered path resolution and recursive fallbacks to survive the structural drift Spotify's A/B tests introduce. Embed pages cap near 100 tracks, so larger playlists fall back to Spotify's internal spclient API with a short-lived anonymous token to recover canonical order.

Matching is where correctness lives. Each track is scored against YouTube candidates on normalized title, artist, and a duration tolerance, so it grabs the right recording instead of a remix or live cut, with multi-client and query-widening fallbacks when a search comes back empty. Around that: a 4-worker thread pool, 429-aware exponential backoff, an append-only JSONL manifest for resume across sessions, and cooperative cancellation.

The suite runs 160+ pytest cases on every push across a 15-job matrix (3 OSes x Python 3.9-3.13), and releases ship SLSA Build Level 3 provenance, CycloneDX SBOMs, and Sigstore signatures.