55c7497ca7cdf33d34f79ea011f6b5de9a19993f
3
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
|
55c7497ca7 |
feat: Minstrel stops acquiring music and drops Discover
release / go (push) Successful in 1m54s
release / web (push) Failing after 5s
release / integration (push) Failing after 3m44s
release / govulncheck (push) Successful in 20s
release / Attach APK to the Release (tag releases only) (push) Canceled after 0s
release / Build + push container image (push) Canceled after 0s
release / Verify release artifacts (tag releases only) (push) Canceled after 0s
release / Build signed APK (push) Canceled after 19s
release / android (push) Canceled after 2m37s
Minstrel is now a viewer of the library it is pointed at. Finding, downloading and filing music moves out of it. Removed from the server: - the request flow (requests, approval, auto-approve) and the wanted list, follows, release profiles and release blocks; - searching indexers, the download clients, the importer, the album lifecycle and missing-file re-acquisition; - the catalogue's discography, search and art (the MusicBrainz release and track-list cache stays, for the duplicate resolver); - Discover: its suggestions of artists not in the library, snoozes, the candidate-artist tag enrichment and its tuning scope; - writes to acquisition tables from moving the library folder and from merging duplicates, the want-list join on Suspect sources, the sealing of acquisition credentials, and the release-group lookup only re-acquisition used. Settling a wrong-slot duplicate no longer sets the release aside: the copies go and their slots show as missing until a good copy is in the library. Removed from the web: Admin > Acquisition and Requests, /requests, /discover, Follow on the artist page, the auto-approve toggle, and the Collection's wanted-list tabs. The Collection is now the library's health: an overview, Reports, Duplicates, Suspect sources and Missing files. The README and docs no longer describe getting music. The tables stay for now; a later migration drops them. The request notification kinds and the Android screens are still to go. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
|
d191a28217 |
docs: server comments say what the code does, not how it came to be (#6277)
Comment blocks in cmd/minstrel and 30 smaller internal packages, rewritten
to rule 32's standard with each citation read in context: task, milestone
and rule ids go, and so do the stories they were attached to ("used to",
"Before #N", measurements from one install, quotes, pointers to Fabled
Librarian's records and to docs/superpowers files that are not in the
repo). Where a story wrapped a real constraint, the constraint stays as a
present-tense fact.
Comments that had gone stale are corrected:
- taste.Config: the tuned values are overlaid by internal/recsettings;
it is no longer "a later follow-up".
- StorageConfig: the data dir holds art, caches and the stream secret,
not only playlist covers.
- the Subsonic album list and getUser comments describe what they do
today rather than a milestone that would change it.
Checked: the non-comment token stream of every changed file matches
HEAD, and gofmt -l is empty.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
||
|
|
f367eeaa9d |
fix(recommendation): Songs-like gets its own profile so it stops wandering
test-web / test (push) Successful in 1m7s
test-go / test (push) Successful in 1m31s
test-go / integration (push) Failing after 4m21s
release / Build signed APK (releases and dev) (push) Successful in 5m11s
release / Build + push container image (push) Successful in 1m52s
release / Verify release artifacts (tag releases only) (push) Skipped
Operator, 2026-09-10: "when I play it I'm expecting to get a consistent
sound and style from the experience... I was getting a seeming wide variety
of music from each one when I was hoping to stay in a certain neighborhood."
Songs-like shared the `daily_mix` weight profile with For-You, and that
sharing WAS the bug. The two surfaces want opposite things: For-You answers
"what will they enjoy today" and is supposed to roam; Songs-like answers
"what sounds like THIS". Under one profile the broad answer wins.
The arithmetic, from the shared weights:
unrelated track, liked, not played recently → 1.0 + 2.0 + 1.0 = 4.0
PERFECT similarity match, not liked → 1.0 + 1.5 = 2.5
Liking something outranked sounding like the seed, because LikeBoost (2.0)
exceeded SimilarityWeight's whole range (1.5) and TasteWeight (1.5, and
seed-INDEPENDENT) matched it outright. Under the new profile the same pair
scores 5.00 vs 2.00.
Two levers, because either alone leaves the other's failure intact:
POOL. Songs-like now takes its own CandidateSourceLimits. The default gave
~29% of candidates a sim_score of literally zero — `taste_overlap` and
`random_fill` are both `0.0::float8` in recommendation.sql, seed-independent
by construction. Same total pool size; composition shifts to arms that
measure distance from the seed, LBSimilar doubled.
WEIGHTS. A third profile beside radio and daily_mix, DB-backed and live per
rule 25, with the property that similarity's range exceeds the combined
range of every seed-independent differentiator — so a closer match cannot
be beaten on likes, freshness and taste alone, while tracks within ~0.39
similarity of each other still get ordered by what the user likes.
Rule 131 changed the pool design mid-way and for the better. Zeroing the
two seed-independent arms was the first instinct and is exactly the
vanish-or-nothing shape that rule forbids: a seed with thin ListenBrainz
coverage would yield a short mix or none. They are the tier-3 FLOOR — cut
hard, never removed — and the weights keep them at the bottom of the
ranking rather than out of the pool. "A few tracks further from the seed
than we'd like" beats "no playlist".
Caught while wiring it: switching only pickTopN's final Score would have
been nearly INERT. scoreAndSortCandidates does the selection sort, and the
caller caps and truncates in that order — so the playlist would still have
been chosen by daily_mix and merely relabelled with songs_like numbers. It
now takes the profile as a parameter, and each surface passes its own.
Also corrects the daily_mix card's blurb, which claimed Songs-like as one
of its surfaces and no longer is.
Guards pin behaviour rather than the numbers, since numbers get retuned:
that similarity beats an unrelated liked track, that daily_mix still
DOESN'T (or the split buys nothing), that the tier-3 floor is non-zero,
and that the UI card shows its own values rather than falling back. Each
falsified against its named regression first.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SQ31KQpYbStyK5y58UmPLH
|