KTT2 Song Finder.
Finds song mentions in KTT2 Music conversations and turns reviewed Spotify matches into private playlists.
- Role
- Creator & full-stack engineer
- When
- 2026
- Built with
- React, TypeScript, Express, SQLite, Spotify API, OpenAI API
KTT2 Song Finder is a local web app for turning music recommendations in KTT2 threads into a private Spotify playlist. A scan reads recent conversations and presents candidate tracks for review.
- Scans recent music threads
- Rules first, optional AI for informal mentions
- Review before Spotify playlist export
How it works.
Read KTT2
A paced crawler reads recent KTT2 Music threads and stores resumable scans in SQLite.
Extract songs
Spotify links and artist–title patterns come first, with an optional OpenAI fallback.
Review & export
Uncertain Spotify matches wait for review before export to a private playlist.
The challenge
Recommendations in discussion threads are easy to miss and hard to collect. Song names appear in different formats, and a fuzzy Spotify result should not silently become a playlist track.
What I did
I built a crawler for recent threads and replies, removed quoted text, extracted explicit Spotify links and artist-title patterns, and used an optional OpenAI fallback for less structured mentions. Clear matches can be accepted automatically; uncertain results remain reviewable before export.
Under the hood
A React and TypeScript frontend talks to an Express API. SQLite stores scans, posts, mentions, and resumable jobs. The worker handles collection, extraction, and Spotify matching in stages; Spotify OAuth is used to create a private playlist after review. The app runs locally and needs user-provided Spotify credentials.
Where it landed
The local workflow can scan conversations, preserve partial results when a stage fails, and turn selected matches into a private playlist. It identifies itself to the source site, limits request frequency, and stops on access challenges.