#metabrainz

aerozol: i am not sure i understand the context. who are the creators in question? but adding those links sound good in any case.

O’Moin

Prototype with the nivo circle packing library (and aerozol's data :p ). I reckon with a bit more work this could look pretty good: https://usercontent.irccloud-cdn.com/file/JG4eN...

Not quite as nice as your mockup, but…

Or we show fewer artists: https://usercontent.irccloud-cdn.com/file/nXIuC...

The library produces an SVG in the end, so we're not far from the other mockups

moin!

nice start on those monkey! but I think the problem is that there is a gulf of effort between "working" and "looks aesthetically pleasing".

Indeed

It looks all right, but I don't think I can make it look a lot better than this https://usercontent.irccloud-cdn.com/file/N0Zir...

Better font style, choosable colors, perhaps

If anyone wants to play: https://codesandbox.io/s/nivo-artist-bubbles-sd...

my pr is leet https://usercontent.irccloud-cdn.com/file/NKrJb...

daily jams are down to 9 tracks now, lol.

will fix that once i am done fixing my bootloader yet again lol

oy, not the evil bootloader again.

mayhem: https://open.spotify.com/playlist/4PtK8B2EqKEh5...

[troi-recommendation-playground] 14amCap1712 opened pull request #75 (03main…more-tracls): Add more tracks to daily jams if needed https://github.com/metabrainz/troi-recommendati...

not the most efficient approach but works... . should probably add some memoization.

mayhem: https://labs.api.listenbrainz.org/similar-recor...

so much awesome stuff to take in!

I hadn't gotten far on the improved playlist before you hit me with the similar recordings.

these similarities are pretty good -- best we've had, I think.

what is the data source? listens? not MHLD, right?

yes listens. and only using 7 days of listens with old mb metadata cache dumps

only 7 days???

do you have something like 30 days kicking around?

yeah, i wanted to test the whole pipeline once first so only did a small batch.

this is pretty promising -- whatever might've changed.

right now things get off into the weed pretty quickly, but the top hits are quite good.

intend to request a 90-180 days version tomorrow

great.

huh, 2 hits for these tracks.

will also request a new dump of mb metadata cache table before that batch.

can you please add the export to spotify button to the instant playlists?

sure will do

brainzplayer just doesn't work for me. :(

thanks!

what else changed in the alg? these are pretty promising for a score of TWO!

I think we can kick generated playlists up a notch once we have this data for all listens.

nothing changed in the algorithm afaik. i just cherry picked commits from aug and put it up on spark.

i think the results were already good when we re did it in august but never had a ui to explore it so didnt notice well

the first iteration we did in feb was the buggy one iirc.

hey, go past us. :)

this is quite eggciting!

indeed

i had to deviate from the schema we discussed earlier because it was actually harder to work with.

yeah, that was still a concern.

just mbid0, mbid1, metadata worked fine. queries are simple.

it feels odd that old skool format is pretty good

k

wow, I'm digging the music I've got with a score of 2.

the case when thing did the trick

so only 1 mbid column to join amything to

noice

when we meet in person, I'd love to have you give alastair and myself a crash course on all the postgres oddities you've learned in the last few months. ;)

hehe lol. sure

I guess with this data we can create an "artist radio" feature. we should have enough data to accomplish that.

is it wrong of me to want to chain datasethoster queries together into a data pipeline?

ACR -> similarities -> upload to spotify

lucifer: https://labs.api.listenbrainz.org/similar-recor...

that is a guaranteed to not be in MB uuid. :)

[troi-recommendation-playground] 14mayhem opened pull request #76 (03main…top-recording-similar-recordings): Create a playlist from the similar recordings of your top recordings https://github.com/metabrainz/troi-recommendati...

that was a fun hour to waste. :)

if anyone wants a playlist to groove to, shout!

zas: https://listenbrainz.org/playlist/778cc309-95e0...

monkey: https://listenbrainz.org/playlist/6ba22fb2-9fe6...

lucifer: https://listenbrainz.org/user/troi-bot/playlists/ -- many users's daily jams stopped updating on 10-18. :(

lucifer: https://listenbrainz.org/playlist/93eb2478-ae5c...

CatQuest: https://listenbrainz.org/playlist/b1c27b4f-d995...

ok, enough for tonight. thanks for attending my monologue this evening! <3

we're really building some cool shit, if I do say so myself. :)