opensource music reccomendation / playlist, similar to spotify radio / YT music mix?

Any recommendations for this?

Initially, i was thinking that LLMs probably not the right thing for this (assuming your source data is all listening metrics), HOWEVER, if you combine
a) user listening data; AND
b) user comments / text data / reccs/ reviews / forum posts / social media mentions etc
and put taht ALL inside the LLM, it might work. Like your ultimate LLM DJ that is intune with not just data, but the zeitgeist as well.

anyway, I've did the obligatory search and seems like nothing really worthy comes up.

Apart from last.fm / various APIs which are heavily limited, there's also this
https://www.reddit.com/r/navidrome/comments/1eoc0cz/generating_weekly_recommendations_playlists_for/

but it seems pretty janky and not exacltly what I'm thinking of.

Is this obscure / rare because BULK user listening data is not really public (ie all hidden behind spotify / youtube / soundhound / shazam walled gardens?)

The ask:
Put in a song / list of songs, and it generates playlist based on that.

So far, spotify's reccs are best for me, i can do endless listening and enjoy most of their suggestions.