Health check: is Nicotine+ reachable, does the music dir exist, what is configured. Call this first when something fails; the result says how to fix it.

Fetch named songs and albums in one call. Items: "Artist - Title" for a song, "Artist - Album (album)" for a whole album. Resolves them on MusicBrainz, skips what the library already has, and starts a background job that searches Soulseek and queues every match at or above min_confidence; the request itself is the go-ahead, no extra confirmation. Report what was understood, then call status(playlist_id) in a minute or two.

Import a playlist (file path: Spotify export, CSV, M3U, JSPF, XSPF; or a TIDAL / Deezer / YouTube Music share URL; or an existing playlist id), resolve it on MusicBrainz, diff it against the library, and match the missing tracks on Soulseek in the background. Nothing is downloaded unless yes=True, which the user must have asked for. Follow with status(playlist_id) until the job is finished, then queue_downloads.

All playlists, or one playlist: counts by state, the running or last job, download progress. Finished downloads are filed into the library (tags normalised, Artist/Album/NN - Title) as a side effect. The 'next' field says what to do next.

Tracks that need a decision with their best candidates: confidence, quality, why they scored, user queue. doubtful_only hides everything at or above 0.85. status="not_found" lists what Soulseek did not have.

Approve candidates: give track_ids (candidate picks which one, 0 = best), or min_confidence to approve every track in the playlist at or above it. Approving queues nothing; queue_downloads does.

Leave these tracks out of matching and the M3U.

Without yes: the totals (tracks, size, users) of what would be downloaded; show them to the user. With yes=True: queue the transfers in Nicotine+. min_confidence first approves every candidate at or above it.

Write the playlist as M3U8 in its original order from the local files; lists the tracks still missing.

Library clean-up: normalise tags, drop lossy duplicates of FLACs, file everything as Artist/Album/NN - Title. apply=False is a dry run that writes a report; apply=True changes files and must only follow the user's yes to the listed deletions.

Stop the running matching job of a playlist.

Artists and albums in the library that have no bio / wiki text yet, with what a writer needs (tracks, year, MusicBrainz ids). Start here, then wiki_fill; write the rest yourself with wiki_sources + wiki_write.

Give every entry without text its English Wikipedia lead paragraph when MusicBrainz links one (attributed, CC BY-SA), limit entries per call. The rest come back in to_write with MusicBrainz facts and links: write those yourself, from the facts only, and store each with wiki_write. Call again while remaining > 0.

Facts and links for one artist (album=None) or album: MusicBrainz type, dates, area, labels, tags, annotation, outbound links (Wikidata, Discogs, Bandcamp, homepage), and the Wikipedia lead when there is an article.

Store the text for an artist (album=None) or an album: written to a sidecar file beside the music and pushed into the player's cache. Plain text, no markup; one paragraph for an album, two for an artist; only what the sources support. attribution is shown under the text: name the sources and, if you wrote it, yourself (e.g. "Written by Claude from MusicBrainz and Discogs, 2026-09-17"). force replaces existing text.