lirify/Overview
connecting api L
Indexer control room

Feed the lyric graph.

Scrape playlists, watch the two-stage queue, and inspect the similarity matrices that power Lirify's search and recommendations. Everything is accumulative — playlists add to the live index.

Listen
en + es only Fetching index state…
—terms
—tokens
—in queue
—Songs indexed
—Inverted terms
—Lyric tokens
—Pending jobs
Ingest

Extract & index

A YouTube video or playlist URL. Extraction is read-only; indexing commits eligible lyrics.
idle
Commands run over gRPC to the python-backend.
Idle 0%
No indexing job running—

Live events

0
Queue

Playlist intake

One URL per line. Jobs are persisted, retried with backoff, and committed batch by batch.
0 jobs
JobSourceStatusStageAddedDetail
No jobs yet.
Metrics

What the index knows

Committed state only — build stats, language mix, provider chain and the weighted similarity model.
— × —

Languages

Lyrics backends

Similarity weights

Allowed languages

Data files

Lyrics provider fallback chain

Top terms by document frequency

Matrices

The similarity graph

A centrality-ranked preview. The full matrices stay on the API — the dashboard only pulls a small read-only slice.
Loading a centrality-ranked matrix preview…
Weak
Strongunified
Diagnostics

Cache & failed songs

Negative results are retried in the background; this is the queue behind that worker.
0Lyrics cached
0Failed songs
0Max attempts
0sRetry interval
No failed songs. Everything the providers could resolve is in the index.

Committed songs

—

Build summary

—