Music API Blog

Most Spotify audio_features replacements stop at raw BPM and key. FreqBlog rebuilt the dead endpoints, then added a set-builder — transition scoring, next-track ranking and Camelot-wheel setlist ordering — and exposed it all over a remote MCP server an LLM can call directly.

Enterprise auto-taggers hand you a flat label and a bill. Here's a self-serve, no-upload alternative: a ~30-line Python walkthrough using /tag, where every tag carries a confidence + provenance so you know which to trust — an honest tagging API on open-data analysis.

Spotify removed GET /v1/recommendations and related-artists too — not just audio_features. A practical Python walkthrough rebuilding both on the FreqBlog Music API: seed-track recommendations and a derived artist graph, genre-aware, no Spotify account.

Spotify removed audio_features — but you can still build a real harmonic set planner. A practical Python walkthrough using /next-track, /transition, /setlist and /export/rekordbox: beat-matched, key-compatible set flow in ~50 lines.

A track that's in every major music database can still fail to match. The three failure modes behind it — punctuation and Unicode drift, search-index blind spots, and naive first-result matching — and how to fix each.

You swapped the host and the response shape works unchanged — but your danceability and acousticness thresholds are firing wrong. Honest field-by-field guide with real catalog distribution data and recommended threshold adjustments.

Eighteen months after Spotify deprecated audio_features and audio_analysis, there's still no official replacement. Honest comparison of the alternatives, with migration code.

Most music-analysis APIs want a 30-second audio clip. If all you have is the track name, you need a catalog API. Here's how name-based lookup actually works, when it fails, and how to architect around the gaps.

AB shut down February 2022 but published the entire 7.5M-track dataset before going dark. What you can still get from the frozen dump, what the live alternatives offer, and how to tier them.

Three professional DJ platforms produce three different keys for the same track. Why, what to do about it, and how Camelot saves you anyway.

50 lines of Python = a working harmonic-mixing matcher. Rules, code, edge cases, and why the wheel is robust to algorithm errors.

Why beat-tracking algorithms return 85 BPM for Blinding Lights (which is 171), and a 30-line heuristic to detect and correct it.