Projects: Teleri

I am an avid music listener. My digital music library is over 500 albums and 7.000 songs, and I’ve heard everything at least twice. Despite having more than 400 hours of music in my collection, I am always searching for more, and often add over 20 new albums to it each month—call me lazy, but that’s a lot of work to do by hand. Teleri, named after the best singers of Middle-earth, is a command line tool that scrapes new album releases and automatically adds the ones I am likely to enjoy to my Apple Music Library.

Then, Teleri gathers information about my library (artist and album names, song genres, etc), and then scrapes music aggregation websites for new album releases (currently only Metacritic, and Pitchork), and using information gathered from my Apple Music library either adds the new album to my library or doesn’t. This decision process is made up of a few simple rules: do I currently have any other albums by this artist? Is this a genre I like a lot? Does this album comes well recommended?

The first rule is very clear, but the next two are a bit more interesting. In order to know if I “like a genre a lot”, I make use of Apple Music’s “love” feature, where you mark songs as loved so their algorithm can recommend similar songs and artists to you. I extract all the songs I have “loved” and get their respective genres. If a specific genre is responsible than more than the 30% of all songs, I can confidently say I “like this genre a lot”. To answer “Is this album well recommended”, on the other hand, I make use of the ratings (if present), on the aggregation websites, and a “recommended” file I manually curate. If the album has a rating over 7/10, or is on the recommended list, it’s added to the library.