feat(taste): household co-play similarity — #1533
test-go / test (push) Successful in 29s
test-go / integration (push) Successful in 4m48s

Milestone #160 Opt 5. A collaborative candidate arm: tracks by artists
co-played across the instance with the seed's artist.

Minstrel is a single shared-library, multi-user server (no per-user
library ACL — verified: no owner/share/group model), so the "household"
is the whole instance's user set; the rule #47 scoping is satisfied by
the shared-library boundary. Single-user servers produce no edges.

- No migration: source='user_cooccurrence' was pre-whitelisted in the
  0009 similarity CHECK from day one.
- internal/db/queries/coplay.sql: Delete + Insert artist co-play edges.
  Score = Jaccard of the two artists' distinct-player sets (controls for
  globally-popular artists); >= 2 co-players AND Jaccard >= floor kept
  (the floor also self-limits hub artists). Completed plays, 365d window.
- internal/coplay: periodic worker (6h) that atomic-replaces the
  user_cooccurrence edge set from play_events — pure local SQL, no
  external calls. Wired in main.go alongside the similarity worker.
- LoadRadioCandidatesV2: new coplay_artists arm (source='user_cooccurrence',
  seed-artist based, 0.5 damp like similar_artists) + $11 limit;
  CandidateSourceLimits.UserCoplay (default 20, For-You 40).
- Integration tests: perfect-overlap Jaccard=1.0 edge + single-user
  empty-set gate.

Device axis and AcousticBrainz (Opt 4) are separately tracked; this
closes the milestone-#160 sequential options.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-07-14 10:07:19 -04:00
parent 65dd132b3d
commit 199fec2058
9 changed files with 441 additions and 2 deletions
+78
View File
@@ -0,0 +1,78 @@
// Code generated by sqlc. DO NOT EDIT.
// versions:
// sqlc v1.31.1
// source: coplay.sql
package dbq
import (
"context"
)
const deleteArtistCoplayEdges = `-- name: DeleteArtistCoplayEdges :exec
DELETE FROM artist_similarity WHERE source = 'user_cooccurrence'
`
// Household co-play similarity (#1533, milestone #160 Opt 5). Minstrel is a
// single shared-library, multi-user server (no per-user library ACL), so the
// "household" is the whole instance's user set. These queries recompute
// artistartist co-occurrence edges from play_events and store them in
// artist_similarity under the pre-provisioned source = 'user_cooccurrence'
// (whitelisted in the 0009 CHECK from day one). A periodic worker DELETEs then
// re-INSERTs so edges that fall below threshold disappear. Single-user servers
// produce no edges (the >= 2 co-player gate), so the arm is empty there.
func (q *Queries) DeleteArtistCoplayEdges(ctx context.Context) error {
_, err := q.db.Exec(ctx, deleteArtistCoplayEdges)
return err
}
const insertArtistCoplayEdges = `-- name: InsertArtistCoplayEdges :exec
WITH user_artists AS (
SELECT DISTINCT pe.user_id, t.artist_id
FROM play_events pe
JOIN tracks t ON t.id = pe.track_id
WHERE pe.was_skipped = false
AND pe.started_at > now() - interval '365 days'
),
artist_players AS (
SELECT artist_id, count(*)::float8 AS players
FROM user_artists
GROUP BY artist_id
),
pairs AS (
SELECT ua.artist_id AS a_id,
ub.artist_id AS b_id,
count(*)::float8 AS coplayers
FROM user_artists ua
JOIN user_artists ub
ON ua.user_id = ub.user_id AND ua.artist_id <> ub.artist_id
GROUP BY ua.artist_id, ub.artist_id
HAVING count(*) >= 2
),
scored AS (
SELECT p.a_id, p.b_id,
p.coplayers / (pa.players + pb.players - p.coplayers) AS score
FROM pairs p
JOIN artist_players pa ON pa.artist_id = p.a_id
JOIN artist_players pb ON pb.artist_id = p.b_id
)
INSERT INTO artist_similarity (artist_a_id, artist_b_id, score, source, fetched_at)
SELECT a_id, b_id, score, 'user_cooccurrence', now()
FROM scored
WHERE score >= $1
ON CONFLICT (artist_a_id, artist_b_id, source)
DO UPDATE SET score = EXCLUDED.score, fetched_at = EXCLUDED.fetched_at
`
// Two artists are "co-played" when the same users play both. score is the
// Jaccard of their distinct-player sets — coplayers / (playersA + playersB
// coplayers) in (0,1] — which controls for globally-popular artists (an artist
// everyone plays would otherwise co-occur with everything). Only pairs with
// >= 2 distinct co-players AND Jaccard >= $1 (a floor that both prunes weak
// edges and self-limits hub artists, whose large denominators drag their
// Jaccard down) are kept. Completed plays only, 365-day window.
func (q *Queries) InsertArtistCoplayEdges(ctx context.Context, score float64) error {
_, err := q.db.Exec(ctx, insertArtistCoplayEdges, score)
return err
}