feat(taste): phase 2b — taste_overlap candidate arm (#796)

2a re-ranks the existing pool by TasteMatch; this ensures taste-relevant tracks
ARE in the pool. Adds a 6th arm to LoadRadioCandidatesV2: in-library tracks by
the user's top positively-weighted taste-profile artists ($10 K, weight > 0,
deterministic weight-DESC,id order so it doesn't reintroduce same-day
nondeterminism). Pool-inclusion only (sim_score 0) — TasteMatch already scores
the fit. Empty for cold-start users (no profile).

- CandidateSourceLimits.TasteOverlap; default 20 (radio), 80 for For-You via
  systemForYouSourceLimits.
- You-might-like deliberately sets TasteOverlap=0: it surfaces NOT-actively-
  engaged artists, so flooding its pool with top-taste (mostly already-played)
  artists would just feed the read-time dedup.
- Test: positive-weight artist's track enters via the arm; negative-weight one
  is excluded (weight > 0). Existing pool tests unaffected (no profile seeded).

Deferred within 2b: profile-seeded For-You — marginal given the arm + TasteMatch
already inject taste broadly (top-played seed ≈ top-taste artist).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

internal/db/dbq/recommendation.sql.go

@@ -737,6 +737,17 @@ likes_overlap AS (
     ORDER BY random()
     LIMIT $8
 ),
+taste_overlap AS (
+    SELECT t.id AS track_id, 0.0::float8 AS sim_score
+    FROM taste_profile_artists tpa
+    JOIN tracks t ON t.artist_id = tpa.artist_id
+    WHERE tpa.user_id = $1
+      AND tpa.weight > 0
+      AND t.id NOT IN (SELECT id FROM excluded_ids)
+      AND t.id <> $2
+    ORDER BY tpa.weight DESC, t.id
+    LIMIT $10
+),
 random_fill AS (
     SELECT t.id AS track_id, 0.0::float8 AS sim_score
     FROM tracks t
@@ -747,6 +758,7 @@ random_fill AS (
           UNION SELECT track_id FROM similar_artists
           UNION SELECT track_id FROM tag_overlap
           UNION SELECT track_id FROM likes_overlap
+          UNION SELECT track_id FROM taste_overlap
       )
     ORDER BY random()
     LIMIT $9
@@ -763,6 +775,7 @@ FROM (
     UNION ALL SELECT track_id, sim_score FROM similar_artists
     UNION ALL SELECT track_id, sim_score FROM tag_overlap
     UNION ALL SELECT track_id, sim_score FROM likes_overlap
+    UNION ALL SELECT track_id, sim_score FROM taste_overlap
     UNION ALL SELECT track_id, sim_score FROM random_fill
 ) u
 JOIN tracks t ON t.id = u.track_id
@@ -790,6 +803,7 @@ type LoadRadioCandidatesV2Params struct {
 	Limit_3 int32
 	Limit_4 int32
 	Limit_5 int32
+	Limit_6 int32
 }
 
 type LoadRadioCandidatesV2Row struct {
@@ -801,11 +815,15 @@ type LoadRadioCandidatesV2Row struct {
 	SimilarityScore interface{}
 }
 
-// M4c: similarity-driven candidate pool. 5-way UNION:
+// M4c: similarity-driven candidate pool. 6-way UNION:
 //
 //	$1 user_id, $2 seed_track_id, $3 recently_played_hours,
 //	$4 exclude (uuid[]), $5 lb_similar K, $6 similar_artists K,
-//	$7 tag_overlap K, $8 likes_overlap K, $9 random_fill K.
+//	$7 tag_overlap K, $8 likes_overlap K, $9 random_fill K,
+//	$10 taste_overlap K (#796 phase 2b — tracks by the user's top
+//	positively-weighted taste-profile artists, so taste-relevant tracks
+//	enter the pool even when the similarity/random arms miss them; scored
+//	in Go via TasteMatch, so sim_score here is 0 pool-inclusion).
 //
 // Returns same shape as LoadRadioCandidates plus similarity_score column.
 func (q *Queries) LoadRadioCandidatesV2(ctx context.Context, arg LoadRadioCandidatesV2Params) ([]LoadRadioCandidatesV2Row, error) {
@@ -819,6 +837,7 @@ func (q *Queries) LoadRadioCandidatesV2(ctx context.Context, arg LoadRadioCandid
 		arg.Limit_3,
 		arg.Limit_4,
 		arg.Limit_5,
+		arg.Limit_6,
 	)
 	if err != nil {
 		return nil, err

internal/db/queries/recommendation.sql

@@ -33,10 +33,14 @@ WHERE t.id <> $2
   );
 
 -- name: LoadRadioCandidatesV2 :many
--- M4c: similarity-driven candidate pool. 5-way UNION:
+-- M4c: similarity-driven candidate pool. 6-way UNION:
 --   $1 user_id, $2 seed_track_id, $3 recently_played_hours,
 --   $4 exclude (uuid[]), $5 lb_similar K, $6 similar_artists K,
---   $7 tag_overlap K, $8 likes_overlap K, $9 random_fill K.
+--   $7 tag_overlap K, $8 likes_overlap K, $9 random_fill K,
+--   $10 taste_overlap K (#796 phase 2b — tracks by the user's top
+--   positively-weighted taste-profile artists, so taste-relevant tracks
+--   enter the pool even when the similarity/random arms miss them; scored
+--   in Go via TasteMatch, so sim_score here is 0 pool-inclusion).
 -- Returns same shape as LoadRadioCandidates plus similarity_score column.
 
 WITH
@@ -109,6 +113,17 @@ likes_overlap AS (
     ORDER BY random()
     LIMIT $8
 ),
+taste_overlap AS (
+    SELECT t.id AS track_id, 0.0::float8 AS sim_score
+    FROM taste_profile_artists tpa
+    JOIN tracks t ON t.artist_id = tpa.artist_id
+    WHERE tpa.user_id = $1
+      AND tpa.weight > 0
+      AND t.id NOT IN (SELECT id FROM excluded_ids)
+      AND t.id <> $2
+    ORDER BY tpa.weight DESC, t.id
+    LIMIT $10
+),
 random_fill AS (
     SELECT t.id AS track_id, 0.0::float8 AS sim_score
     FROM tracks t
@@ -119,6 +134,7 @@ random_fill AS (
           UNION SELECT track_id FROM similar_artists
           UNION SELECT track_id FROM tag_overlap
           UNION SELECT track_id FROM likes_overlap
+          UNION SELECT track_id FROM taste_overlap
       )
     ORDER BY random()
     LIMIT $9
@@ -135,6 +151,7 @@ FROM (
     UNION ALL SELECT track_id, sim_score FROM similar_artists
     UNION ALL SELECT track_id, sim_score FROM tag_overlap
     UNION ALL SELECT track_id, sim_score FROM likes_overlap
+    UNION ALL SELECT track_id, sim_score FROM taste_overlap
     UNION ALL SELECT track_id, sim_score FROM random_fill
 ) u
 JOIN tracks t ON t.id = u.track_id

internal/playlists/system.go

@@ -329,6 +329,10 @@ func systemForYouSourceLimits() recommendation.CandidateSourceLimits {
 		TagOverlap:    60,
 		LikesOverlap:  40,
 		RandomFill:    150,
+		// For-You / You-might-like are the taste-driven surfaces, so pull a
+		// deep slice of the user's top taste-profile artists into the pool
+		// (#796 phase 2b). Empty for cold-start users (no profile yet).
+		TasteOverlap: 80,
 	}
 }
 

internal/playlists/you_might_like.go

@@ -89,9 +89,15 @@ func buildYouMightLike(
 	}
 
 	zeroVec := recommendation.SessionVector{Seed: true}
+	// You-might-like surfaces in-library artists the user does NOT actively
+	// engage with, so it deliberately skips the taste_overlap arm — that arm
+	// pulls top-taste (mostly already-played) artists, which would crowd the
+	// pool with entities the read-time dedup then strips. For-You/radio keep it.
+	ymlLimits := systemForYouSourceLimits()
+	ymlLimits.TasteOverlap = 0
 	cands, err := recommendation.LoadCandidatesFromSimilarity(
 		ctx, q, userID, seed, 1, zeroVec,
-		[]pgtype.UUID{seed}, systemForYouSourceLimits(),
+		[]pgtype.UUID{seed}, ymlLimits,
 	)
 	if err != nil {
 		logger.Warn("you-might-like: candidate load failed; skipping",

internal/recommendation/candidates.go

@@ -72,6 +72,10 @@ type CandidateSourceLimits struct {
 	TagOverlap    int
 	LikesOverlap  int
 	RandomFill    int
+	// TasteOverlap (#796 phase 2b): tracks by the user's top positively-
+	// weighted taste-profile artists. 0 disables the arm (e.g. cold-start
+	// users have an empty profile, so it contributes nothing anyway).
+	TasteOverlap int
 }
 
 // DefaultCandidateSourceLimits returns the v1 hardcoded constants per spec.
@@ -82,6 +86,7 @@ func DefaultCandidateSourceLimits() CandidateSourceLimits {
 		TagOverlap:    20,
 		LikesOverlap:  20,
 		RandomFill:    30,
+		TasteOverlap:  20,
 	}
 }
 
@@ -113,6 +118,7 @@ func LoadCandidatesFromSimilarity(
 		Limit_3: int32(limits.TagOverlap),
 		Limit_4: int32(limits.LikesOverlap),
 		Limit_5: int32(limits.RandomFill),
+		Limit_6: int32(limits.TasteOverlap),
 	})
 	if err != nil {
 		return nil, err

internal/recommendation/candidates_v2_test.go

@@ -262,6 +262,61 @@ func TestLoadCandidatesFromSimilarity_DedupTakesMaxScore(t *testing.T) {
 	}
 }
 
+// TestLoadCandidatesFromSimilarity_TasteOverlapArm (#796 phase 2b): a track by
+// a positively-weighted taste-profile artist enters the pool via taste_overlap
+// even with every other arm disabled; a negatively-weighted artist's track does
+// not (the WHERE weight > 0 filter).
+func TestLoadCandidatesFromSimilarity_TasteOverlapArm(t *testing.T) {
+	f := newFixture(t, 2) // seed + 1 other, both by the fixture artist
+	seed := f.tracks[0]
+	target := f.tracks[1]
+	ctx := context.Background()
+
+	// Fixture artist gets a positive taste weight.
+	if _, err := f.pool.Exec(ctx,
+		`INSERT INTO taste_profile_artists (user_id, artist_id, weight) VALUES ($1, $2, 5.0)`,
+		f.user, seed.ArtistID); err != nil {
+		t.Fatalf("insert taste (positive): %v", err)
+	}
+
+	// A second artist with a NEGATIVE weight — its track must be excluded.
+	negArtist, _ := f.q.UpsertArtist(ctx, dbq.UpsertArtistParams{Name: "NegArtist", SortName: "NegArtist"})
+	negAlbum, _ := f.q.UpsertAlbum(ctx, dbq.UpsertAlbumParams{Title: "NegAlbum", SortTitle: "NegAlbum", ArtistID: negArtist.ID})
+	negTrack, _ := f.q.UpsertTrack(ctx, dbq.UpsertTrackParams{
+		Title: "NegTrack", AlbumID: negAlbum.ID, ArtistID: negArtist.ID,
+		FilePath: "/tmp/neg.flac", DurationMs: 180_000,
+	})
+	if _, err := f.pool.Exec(ctx,
+		`INSERT INTO taste_profile_artists (user_id, artist_id, weight) VALUES ($1, $2, -3.0)`,
+		f.user, negArtist.ID); err != nil {
+		t.Fatalf("insert taste (negative): %v", err)
+	}
+
+	// Only the taste_overlap arm is enabled.
+	limits := CandidateSourceLimits{TasteOverlap: 10}
+	got, err := LoadCandidatesFromSimilarity(
+		ctx, f.q, f.user, seed.ID, 1, SessionVector{Seed: true}, nil, limits,
+	)
+	if err != nil {
+		t.Fatalf("load: %v", err)
+	}
+	var sawTarget, sawNeg bool
+	for _, c := range got {
+		switch c.Track.ID {
+		case target.ID:
+			sawTarget = true
+		case negTrack.ID:
+			sawNeg = true
+		}
+	}
+	if !sawTarget {
+		t.Error("positive-taste-artist track missing (taste_overlap arm didn't contribute)")
+	}
+	if sawNeg {
+		t.Error("negative-taste-artist track present (weight > 0 filter failed)")
+	}
+}
+
 func TestLoadCandidatesFromSimilarity_EmptyLibrary_NoError(t *testing.T) {
 	f := newFixture(t, 1) // just the seed
 	seed := f.tracks[0]