#metabrainz

oh yes. thanks for that correction.

rabbitmq runs on trile.

anyway, just throwing out some random ideas. if the data transfer continues to be a sticking point it might be worth revisiting, but it sounds like it's a bunch extra work that we don't want to poke right at the moment

iliekcomputers: no need for a discussion tomorrow, thanks :)

I *think* that if we improve the logging on the request consumer to identify each client correctly, we can understand things better.

In general I am happy with how things are working -- that isn't to say that we don't have a pile of improvements to make.

also, remember that we're just now putting things into production on this front. still things to be ironed out.

pristine___: your jobs ran.

shivam-kapila: how are you? how is the PR? should we have meeting or do you instead want to work on the PR?

ruaok: seen. They are within the range. No negatives.

ok, cool. lets undo the debug statements then.

also, I asked for 200 recs from the API, but only got 100. Is that intentional?

Hmmm. Will do.

Nope

100 is the max we can retrieve from API

You can use offset.

To get the remaining 100

ruaok: ^^

ok, I'll paginate my shit then...

:p

As per our discussion a few weeks ago, our next step is to generate similar artists using CF. But I was thinking of playing with the current data and get a playable list before moving to that. What do you think?

I am also interested in the algo to get a playlist of 25 or something from these 200 recs for a user.

ruaok: sorry. I was away for dinner

Are you free now

shivam-kapila: here yes. how is that PR?

Multiple PRs

pristine___: hang on a sec

1 is almost ready

2 on it

3rd tomorrow

shivam-kapila: then I think you should focus on those PRs and keep moving.

I will post the screenshots in the PRs itself

Hanging.

1 thing I had in mind was that I am using Spotify to fetch cover arts

reosarevok: You may have missed https://tickets.metabrainz.org/browse/MBS-10193...

MBS-10193: Update iTunes/Apple Music URL cleanup

See also MBS-9375

Can it lead to rate limit problems

MBS-9375: Remove autoselect from iTunes/Apple Music release URLs https://tickets.metabrainz.org/browse/MBS-9375

ruaok: ^^

shivam-kapila: I am not sure, we'd have to check the docs.

!m outsidecontext

You're doing good work, outsidecontext!

Ok I will do that

I'd say don;'t worry about that right this second. get the PR in and note that as a potential problem.

especially about improved responsiveness :)

outsidecontext: grüße aus dem Scharzwald!

I figured out the multiple component thing and figured that I will pass an extra prop to render the design for mobile devices

ruaok: Grüße zurück :) I'd been to the Schwarzwald this year also for holiday. Was great, a lot of walking without meeting too many people

I guess further updates through demos from my side for which you have been waiting for more than 2 weeks 😬

Anything else you want to discuss?

nächste woche am 16-17. habe ich ein tag frei. ich bin noch nie in Konstanz gewesen...

shivam-kapila: not really. just work on those PRs, we're so close.

Yes we are finally

Ok I am onto thrm thanks

pristine___: if you want, and I would love that, is for you to start working with troi.

Mr_Monkey: iliekcomputers: how should I define my custom styles LESS or CSS in React

I've got the MSB mapping generating periodically now and have a couple of minor things left to finish off.

ruaok: leider bin ich ja nicht mehr in Konstanz

oh. wo nu?

ruaok: hmmm. Sounds good. So I am gonna spend few days on it.

want me to walk you through the example?

Why not :)

ok

the whole system is build on the concept of Elements.

and Element can take a number of initializers (literally args to __init__) and be connected to 0 or more other elements.

let's take the recommondation element: https://github.com/metabrainz/troi-recommendati...

__init__ takes the user_name and type of the recommendations

inputs are none -- it is a data source.

outputs is a list of Recordings.

Hmm. With you.

entities are defined here: https://github.com/metabrainz/troi-recommendati...

so the rec fetcher uses iliekcomputers' pylistenbrainz to fetch data and then to convert it into a list of Recordings. that's it.

that is invoked by line 26 here: https://github.com/metabrainz/troi-recommendati...

now, if you recall, the recs only giving recording_mbid and rating, nothing more.

Cool. So how do we choose the 25 tracks? The top 25, as per rating, no?

> now, if you recall, the recs only giving recording_mbid and rating, nothing more.

the output from the recs are then pushed to the next step: https://github.com/metabrainz/troi-recommendati...

Yes

the next step takes the MBIDs and fetches all the relevant data from musicbrainz.

yes, top 25 per rating -- no intelligence there.

the dump element here: https://github.com/metabrainz/troi-recommendati...

all it does is debug things, but its not used right now.

Hmmm. So what should be the next step? Making the playlist playable or an algo for choosing tracks for playlists.

instead the last element take the recording list and makes a "playlist" from it and then launches it to the brainzplayer.

I guess the former.

> instead the last element take the recording list and makes a "playlist" from it and then launches it to the brainzplayer.

Ooooo

the list is "playable" its just, not very smart.

ruaok: bei Mainz

ruaok: okay. So I understand the general flow.

outsidecontext: ahhhh, ok.

pristine___: great.

we've basically done all of the hard work -- and now we can be creative.

to make a playlist you'll need to add... something else.

there are endless possibilities.

Like

one suggestion is to make a playlist generator for jogging.

start kinda slow. pick up the pace, then wind down and chill out.

you could give as input: warm up: 8 minutes. ramp up: 4 minutes, steady: 10 minutes. cool down 10 minutes.

each of these sections could have BPM associated with it and then you could find tracks in the recommended tracks that fit the profile.

BPM?

one thought I had is this: (might be of interest to alastairp too) What if we had a personal music collection that we submitted to AB.

beats per minute

then we could download all of the acoustic data from AB and create our own Annoy indexes.

yep, perfect usecase for unifying the brainzes

then a user could pick a seed track and have troi generate a similar track playlist based on that seed song, based only on what is available in the local collection.

MusicIP's tool used to do that really well. I miss that!

oh hey boss!

ruaok: I worry about saying it out loud, but it kind of feels like we need a local collection management app :(

I wonder how much of this you could do with a picard plugin

beets seems pretty well suited to do something like that. amarok too, probably

in terms of keeping a database of mbids of your local tracks

alastairp: https://github.com/mayhem/musicpod I've been working on one.

it turns your music collection into an API with search.

ruaok I guess we should make a plan so that we are not lost in the endless possibilities :p I will generate a playlist for myself tomorrow and get back to you for further discussion.

well, problem solved then. just finish that in all of the copious spare time that you have, and we can use that

next step: AB submission and brainzplayer support.

alastairp: that's the plan!

so, see you in 10 years then? :D

pristine___: good plan. there are so many possibilities it is easy to get lost!

alastairp: far less than that!

picard does so much of the work, a chunk is done.

nn!

I'm curious to see if I can generate annoy indexes for a private collection.

nn pristine___

hi?

It’s Lazy Monday!

I haven’t received any mailed in reviews, so…

People up for reviews tonight: zas, yvanzo, alastairp, reosarevok, Mr_Monkey, ruaok, bitmap, Freso, diru1100, prabal, jmp_music, CatQuest, shivam-kapila, ishaanshah, _lucifer – anyone else, let me know ASAP!

zas: Go!

Hey

apart usual stuff (maintenance, supervision, minor fixes, fighting back abusers, whitelisting), we released Picard 2.4 :)

zas: u know any public musicbrainz mirror servers i didnt know? https://wiki.musicbrainz.org/User:Colbydray/Mus...