rekordbox web

experiment shelf

Lab Routes

Quick entry points for family-song coverage and route simulation experiments. These pages are intentionally separated from search results so the experimental surface stays visible but contained.

screenshot text

Tracklist Import

Paste screenshot-derived track text, queue raw rows, and prepare them for Apple Music Search normalization.

raw textApple Searchnormalizer
Open preview ->

family-song

Family Song Coverage

Song-recursive family coverage with identify queue, works/abstract-like checks, and provider matching diagnostics.

recent artistssong0 identifyprovider match
Open preview ->

song0 fast path

Simple Coverage

Song0-only identify with no debug UI, tuned for the fastest useful preview. No fallback to other layers yet when Song0 can't be matched.

song0no debugfast
Open preview ->

artist identity

Artist Identity Lab

Search artist identity first, then branch into rooms for clusters, name strength, membership, identity maps, and relation graphs.

searchidentity roomsrelation maps
Open preview ->

route simulation

Song Route Preview

Uses TopCoverage song0 layers to find related artists, then expands each artist into a small coverage preview.

Takkyu Ishinoroute previewrelated artists
Open preview ->

member route

Scene Route Lab

Resolve a source artist to group members or self, then inspect each seed artist's top three Song0 label profiles.

Daft Punkmembers/selftop 3 songs
Open preview ->

layer genre/style

Song Layer Genre

Run song-recursive from a core_hash + artist_id and read genre/style per layer. All layers, with a mark on the ones that fell back to release-level tags.

core_hashall layersgs fallback
Open preview ->

song0 profile

Song0 Profile Probe

Select a family core hash from an artist coverage list, then fetch Song0 genre, style, label, year, duration, and role artists.

core_hashsong0genre/style/label
Open preview ->

profileの版

Profile Sweep

Counts artists whose families were baked with an older profile version, then re-bakes the top few and shows their profile-version breakdown and taste before and after.

profile_versionsweeplm_artist_taste
Open preview ->

view_daily

View Daily

How often artists and songs are played (60s) or dug (song / artist page), as daily viewer counts over today, 7 or 30 days.

playdigranking
Open preview ->

discogs -> rekordbox

Rekordbox Owned

Checks whether a Discogs-side track can be found back in the user's rekordbox collection, with ground truth so false positives and misses are counted separately.

owned badgeblocking + fuzzyrecall / false positive
Open preview ->

rekordbox -> discogs

Rekordbox Identify

Samples tracks from the user's rekordbox master DB and runs the same identifyDiscogsSong path as the song page, to measure hit rate and latency.

master DBidentify accuracyhit rate
Open preview ->

labelmate

Label Release Probe

Inspect label-centered release blocks: releases, tracks, labels, release artists, track artists, and core credits.

label -> releasetrack creditslabelmate
Open preview ->

master releases

Artist Discography

List an artist's master releases from the Discogs dump (master_artist → master → main_release). Main works only, unofficial pressings excluded, ordered by year.

mastermain onlyno bootlegs
Open preview ->

release coverage

Artist TopRelease Probe

Rank releases where the artist appears as main on family/core titles. First pass for release-centered Dig infrastructure.

Takkyu Ishinomain onlyrelease coverage
Open preview ->

works coverage

Artist Works Coverage

Read non-main family works coverage with representative titles. First probe for writer, producer, remix, and credit-side Dig routes.

TopWorksnon-main creditsrole bits
Open preview ->

lab data policy

Experimental, But Not Careless

Display minimum

Lab pages may expose visible Song facts, but should avoid sending raw release ids, evidence rows, or internal scoring unless debug explicitly needs it.

Derived data boundary

Genre, style, label, and year summaries are product assets. Prefer rounded display summaries over full reconstructable datasets.

Future UI scope

UI experiments can recalculate visible ranges, but long-term artist features and time-series metrics should be computed server-side.

Current lab entrances

/lab/family-coverage

/lab/simple-coverage

/lab/scene-route

/lab/song-route-preview

/lab/artist-graph-preview

/lab/artist-identity

/lab/artist-works-coverage

/lab/artist-discography

/lab/song-layer-genre

/lab/song0-profile

/lab/rekordbox-identify

/lab/rekordbox-owned

/lab/label-release-probe

/lab/artist-top-release