Browse a large library without rendering everything

After a scan you have thousands of rows and one screen. This page covers finding tracks, reading their metadata, previewing audio, and turning what you find into seeds or a temporary set.

Panel 2, library and listening, is the middle column. Its heading carries a badge reading tracks plus the library total. Control names below are given in English, and UI language carries the on-screen string for each one.

Direct API equivalent

The browser uses the same current endpoints shown here. For scripting, start the backend on 127.0.0.1 and call them directly:

Invoke-RestMethod -Uri 'http://127.0.0.1:8765/api/library/summary'
Invoke-RestMethod -Uri 'http://127.0.0.1:8765/api/tracks?limit=25'

Use GET /api/tracks/{track_id} for full metadata and GET /media/{track_id} for preview audio. The current response shapes and filter query parameters are in the API reference.

Loading behavior

The library panel uses fixed server-side pages of up to 200 tracks. Prev, Next, and the page-number field request one /api/tracks page at a time. All returned rows render in one scrollable list. There is no second row-window paginator.

The pagination group also shows three read-only status chips: the current page over the page count, the track range on this page, and the filtered total in parentheses. Each reads ... while a page is loading.

Database changes and newer loads invalidate older responses. Duplicate rows are reconciled by their stable catalog_uuid and track_uuid identity. A response whose catalog identity does not match the selected database is rejected with Library response catalog identity does not match the selected database.

Search and filters

The search box carries the placeholder path, title, artist, genre and matches artist, title, album, path, and MAEST genres. Beside it a two-button group selects the matching mode:

• LIKE runs a substring match, so it finds partial text inside a token.
• FTS runs a token search. It is faster on broad text queries and does not match arbitrary substrings inside one token.

To the right of the pagination group sit four controls, left to right:

1. A heart icon filters to liked tracks only. Its tooltip names the count, as in "show liked tracks only, 312 available". It is disabled while no track is liked.
2. A waveform icon titled Show only tracks with the stored MAEST syncopated-rhythm flag filters to tracks whose stored MAEST analysis carries the syncopated-rhythm flag.
3. A shuffle icon toggles random playback order.
4. An up/down arrow icon toggles the display order of the loaded rows.

Shuffle and sort direction act on the rows already loaded on the current page. Neither reorders the server query or reaches other pages. Both are disabled while fewer than two rows are loaded.

Promoted classifiers add minimum-score filters to the same library query. They are configured in the CLASS tab.

The API caps each request at 1..500. The UI keeps rows light and opens full metadata only on demand.

Maintenance actions in panel 1

Five icon buttons in panel 1, database and analysis, act on the library you are browsing:

Icon	Control	What it does
Circular arrows	Refresh tags	rereads Mutagen tags for existing tracks without touching paths or analysis rows
Save	Save genres	writes stored MAEST genres into audio files, the one backend audio write
Shield with a check	Validate the database	starts a read-only database validation job
Two overlapping squares	Find and review duplicates	opens the Audio Dedup reviewer
Trash	Clear the database	deletes every SQLite row after a confirmation, leaving audio files in place

Validate the database is read-only. It opens the file read-only, streams findings into the log, and reports a running notice naming what was checked, such as validation finished with 41235 checked, 3 warnings, and 0 errors. Only one validation job runs at a time. The same check is available from the CLI as dj-sim validate-database, which exits with code 2 when it finds errors.

All five buttons are disabled while the library is empty or a stage is running. Save genres also needs at least one stored MAEST genre.

Metadata dialog

Open track details with the tag icon on a row, titled Tags and genres. The dialog separates:

• MAEST genres, with a syncopated-rhythm indicator when the flag is stored,
• Track Details, with file name, path, and size,
• Tags, the Mutagen fields,
• Data, the technical audio facts,
• SONARA features, grouped into tempo, tonal, loudness, structure, spectral, timbral, perceptual, mood, aggression, and vocalness,
• Classifier scores,
• Scan analyses details, a collapsible block with scan, SONARA, embedding, and classifier timestamps.

Each SONARA feature value carries an inline # comment in a fixed description column. BPM and key candidate lists use the full value width and omit that comment. A value that has not been computed renders as -, and an empty group says so in Russian, with a note that SONARA features have not been calculated yet.

The header carries a copy button for the file path, a copy button for the file name, an Open containing folder action, and a delete action titled Delete from the database. The delete action removes the track and its catalog-owned SQLite data after a confirmation and leaves the audio file on disk.

Keep those sources separate when judging a track. A MAEST label and a file tag are not the same evidence.

Preview

The /media/{track_id} endpoint streams the local file when the browser can play it. AIFF, FLAC, DSD/DSF, WMA, APE, WV, M4B/M4R, TAK, TTA, and browser-unsafe WAV files are decoded and encoded in process with TorchCodec and the configured shared FFmpeg libraries to a temporary WAV for streaming, then the temporary file is deleted. The source audio file is not rewritten, and no ffmpeg.exe process is started.

If the file is missing, preview returns an error instead of hiding the problem.

The row playing right now replaces its score or index area with a seek slider and a M:SS / M:SS readout.

Continuous playback on the loaded page

When a library preview finishes, the player starts the next visible track in the displayed order of the currently loaded page. Playback stops after the last visible track. It does not advance to another page.

With shuffle enabled, the next track is chosen at random from the current page. Shuffle never selects the track that just finished.

Likes

The like button writes a local SQLite row with catalog_uuid and track_uuid. It does not edit audio tags. Its tooltip toggles between Like and Remove the like. Likes can be used for browsing and filtering.

Seeds and the current set

From each visible result row you can:

• add the track as a seed with the magnifier icon titled Seed,
• add or remove the track from the current set with the plus or minus icon, titled Add to the set and Remove from the set,
• start or stop preview,
• open metadata,
• toggle liked state.

Seeds feed the SIMILARITY, SONARA, and LAB tabs. Selected seeds appear as removable chips above the tab strip in panel 3, search and listening. The current set is the temporary playlist inside the collapsible Set and export block. It is not written to disk until you export it or save it as a Rhythm Lab collection.

Add visible tracks

The plus button at the right of the control group appends only the tracks already loaded on the current library page to the current set, in the displayed order. Tracks already in the set are skipped. A run that adds nothing reports that every track on this page is already in the set. It does not make a separate library or database request, load other pages, validate track data, or recalculate analysis.