Noiseblend

Hi @abadesi! I'm a fan of Discover weekly myself! ✨ But I can't rely on it for specific activities like reading or working out for example. Yes, I know that there are quite a lot of playlists for those activities, but I wanted something more personal. That's what Noiseblend gives you, always-fresh playlists generated using the same technologies that Discover Weekly uses, based on what you like to listen on that specific occasion. That, and an easy way to customize those recommendations based on attributes such as Danceability 💃 or Acousticness 🎸. Of course, this wouldn't have been possible without the powerful API that Spotify provides. We're not reinventing another AI, just borrowing theirs 😄

@abadesi @alin23 Love the UI! When you say "borrowing theirs" for the AI, are you doing any playlist manipulation on your end? Or are you using a Spotify Endpoint that crafts a playlist based off of some parameter?

@kevinguebert We're mostly using the Spotify `/recommendations` endpoint for the playlist generation. We're also doing some basic playlist manipulation like a custom ordering algorithm and some fiddling with the seed parameters for `/recommendations`. But no machine learning stuff is done on our part. This would require a huge database with user behavior data and I'm trying to keep as little data as possible about the users.

@alin23 Awesome work! Thanks for sharing!

@reymarr_zamora_hijara Hi! We are experiencing some problems with our database and we'll be back as soon as possible. We are really sorry for this!

@reymarr_zamora_hijara I'm still working on fixing all the errors. There's a database connection limit that breaks the service from time to time. I'm hoping on fixing it today.

@alin23 Aight. Standing by...

@reymarr_zamora_hijara the database problem and other small issues have been resolved. You can use the app now 😇 Though there are still some issues for users where Spotify just launched and most of the recommendations come as not playable for their country ¯\_(ツ)_/¯ Still, if you experience any error/issue, please tell us how it happened because sometimes your words can be much more helpful than an error traceback. We simply cannot predict all the use cases from the start.