feat(taste): household co-play similarity — #1533

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>
2026-07-14 10:07:19 -04:00
parent 65dd132b3d
commit 199fec2058
9 changed files with 441 additions and 2 deletions
@@ -0,0 +1,86 @@
+// Package coplay builds household co-play similarity edges (#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.
+// A periodic worker recomputes artist–artist co-occurrence from play_events
+// entirely in SQL (no external API) and stores the edges in artist_similarity
+// under source='user_cooccurrence' — the pre-provisioned co-occurrence source
+// from the 0009 schema — which the radio/mix candidate pool consumes as a
+// collaborative arm. Single-user servers produce no edges (the >= 2 co-player
+// gate), so the feature is a graceful no-op there.
+package coplay
+
+import (
+	"context"
+	"log/slog"
+	"time"
+
+	"github.com/jackc/pgx/v5/pgxpool"
+
+	"git.fabledsword.com/bvandeusen/minstrel/internal/db/dbq"
+)
+
+// minJaccard is the co-play edge floor: prune weak pairs and self-limit hub
+// artists, whose large combined-audience denominators pull their Jaccard down.
+// 0.1 keeps only pairs whose co-player overlap is a real fraction of their
+// combined audience.
+const minJaccard = 0.1
+
+// Worker periodically rebuilds the household co-play edge set. Pure local SQL,
+// so there are no rate limits — but the pair self-join is roughly
+// O(users × artists_per_user²), so it ticks slowly: co-listening shifts over
+// days, not minutes, and a frequent rebuild would burn CPU for no fresher
+// signal.
+type Worker struct {
+	pool   *pgxpool.Pool
+	logger *slog.Logger
+	tick   time.Duration
+}
+
+// NewWorker constructs a worker with the production default 6h tick.
+func NewWorker(pool *pgxpool.Pool, logger *slog.Logger) *Worker {
+	return &Worker{pool: pool, logger: logger, tick: 6 * time.Hour}
+}
+
+// Run rebuilds once at startup (so edges exist without waiting a full tick),
+// then every w.tick until ctx is cancelled. Runs in its own goroutine, so a
+// slow initial rebuild never blocks boot.
+func (w *Worker) Run(ctx context.Context) {
+	if err := w.rebuild(ctx); err != nil {
+		w.logger.Error("coplay: initial rebuild failed", "err", err)
+	}
+	t := time.NewTicker(w.tick)
+	defer t.Stop()
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-t.C:
+			if err := w.rebuild(ctx); err != nil {
+				w.logger.Error("coplay: rebuild failed", "err", err)
+			}
+		}
+	}
+}
+
+// rebuild atomic-replaces the user_cooccurrence edge set inside one
+// transaction — delete the old edges, recompute from current play history — so
+// the candidate pool never observes a half-built set.
+func (w *Worker) rebuild(ctx context.Context) error {
+	tx, err := w.pool.Begin(ctx)
+	if err != nil {
+		return err
+	}
+	defer func() { _ = tx.Rollback(ctx) }()
+	q := dbq.New(tx)
+	if err := q.DeleteArtistCoplayEdges(ctx); err != nil {
+		return err
+	}
+	if err := q.InsertArtistCoplayEdges(ctx, minJaccard); err != nil {
+		return err
+	}
+	if err := tx.Commit(ctx); err != nil {
+		return err
+	}
+	w.logger.Debug("coplay: edges rebuilt")
+	return nil
+}
@@ -0,0 +1,162 @@
+package coplay
+
+import (
+	"context"
+	"io"
+	"log/slog"
+	"os"
+	"testing"
+	"time"
+
+	"github.com/jackc/pgx/v5/pgtype"
+	"github.com/jackc/pgx/v5/pgxpool"
+
+	"git.fabledsword.com/bvandeusen/minstrel/internal/db"
+	"git.fabledsword.com/bvandeusen/minstrel/internal/db/dbq"
+	"git.fabledsword.com/bvandeusen/minstrel/internal/dbtest"
+)
+
+func testPool(t *testing.T) *pgxpool.Pool {
+	t.Helper()
+	if testing.Short() {
+		t.Skip("skipping in -short mode")
+	}
+	dsn := os.Getenv("MINSTREL_TEST_DATABASE_URL")
+	if dsn == "" {
+		t.Skip("MINSTREL_TEST_DATABASE_URL not set")
+	}
+	if err := db.Migrate(dsn, slog.New(slog.NewTextHandler(io.Discard, nil))); err != nil {
+		t.Fatalf("migrate: %v", err)
+	}
+	pool, err := pgxpool.New(context.Background(), dsn)
+	if err != nil {
+		t.Fatalf("pool: %v", err)
+	}
+	t.Cleanup(pool.Close)
+	dbtest.ResetDB(t, pool)
+	return pool
+}
+
+func discardLogger() *slog.Logger { return slog.New(slog.NewTextHandler(io.Discard, nil)) }
+
+func seedUser(t *testing.T, q *dbq.Queries, name string) pgtype.UUID {
+	t.Helper()
+	u, err := q.CreateUser(context.Background(), dbq.CreateUserParams{
+		Username: dbtest.TestUserPrefix + name, PasswordHash: "x", ApiToken: name + "-tok", IsAdmin: false,
+	})
+	if err != nil {
+		t.Fatalf("seed user %s: %v", name, err)
+	}
+	return u.ID
+}
+
+// seedArtistTrack creates an artist + album + one track, returning both ids.
+func seedArtistTrack(t *testing.T, q *dbq.Queries, name string) (pgtype.UUID, pgtype.UUID) {
+	t.Helper()
+	ctx := context.Background()
+	a, err := q.UpsertArtist(ctx, dbq.UpsertArtistParams{Name: name, SortName: name})
+	if err != nil {
+		t.Fatalf("seed artist: %v", err)
+	}
+	al, err := q.UpsertAlbum(ctx, dbq.UpsertAlbumParams{Title: name, SortTitle: name, ArtistID: a.ID})
+	if err != nil {
+		t.Fatalf("seed album: %v", err)
+	}
+	tr, err := q.UpsertTrack(ctx, dbq.UpsertTrackParams{
+		Title: name, AlbumID: al.ID, ArtistID: a.ID,
+		DurationMs: 1000, FilePath: "/tmp/coplay-" + name + ".mp3",
+		FileSize: 1, FileFormat: "mp3",
+	})
+	if err != nil {
+		t.Fatalf("seed track: %v", err)
+	}
+	return a.ID, tr.ID
+}
+
+func play(t *testing.T, pool *pgxpool.Pool, userID, trackID pgtype.UUID) {
+	t.Helper()
+	ctx := context.Background()
+	at := time.Now()
+	var sessionID pgtype.UUID
+	if err := pool.QueryRow(ctx,
+		`INSERT INTO play_sessions (user_id, started_at, last_event_at, client_id)
+		 VALUES ($1, $2, $2, 'coplay-test') RETURNING id`,
+		userID, at).Scan(&sessionID); err != nil {
+		t.Fatalf("insert play_session: %v", err)
+	}
+	if _, err := pool.Exec(ctx,
+		`INSERT INTO play_events (user_id, track_id, session_id, started_at, was_skipped)
+		 VALUES ($1, $2, $3, $4, false)`,
+		userID, trackID, sessionID, at); err != nil {
+		t.Fatalf("insert play_event: %v", err)
+	}
+}
+
+func coplayEdgeCount(t *testing.T, pool *pgxpool.Pool) int {
+	t.Helper()
+	var n int
+	if err := pool.QueryRow(context.Background(),
+		`SELECT count(*) FROM artist_similarity WHERE source='user_cooccurrence'`).Scan(&n); err != nil {
+		t.Fatalf("count edges: %v", err)
+	}
+	return n
+}
+
+// TestCoplayWorker_BuildsEdges: two users who both play two artists produce a
+// perfect-overlap co-play edge (Jaccard 1.0) in each direction.
+func TestCoplayWorker_BuildsEdges(t *testing.T) {
+	pool := testPool(t)
+	ctx := context.Background()
+	q := dbq.New(pool)
+
+	u1 := seedUser(t, q, "cop1")
+	u2 := seedUser(t, q, "cop2")
+	aA, tA := seedArtistTrack(t, q, "ArtA")
+	aB, tB := seedArtistTrack(t, q, "ArtB")
+
+	for _, u := range []pgtype.UUID{u1, u2} {
+		play(t, pool, u, tA)
+		play(t, pool, u, tB)
+	}
+
+	if err := NewWorker(pool, discardLogger()).rebuild(ctx); err != nil {
+		t.Fatalf("rebuild: %v", err)
+	}
+
+	var score float64
+	if err := pool.QueryRow(ctx,
+		`SELECT score FROM artist_similarity
+		 WHERE artist_a_id=$1 AND artist_b_id=$2 AND source='user_cooccurrence'`,
+		aA, aB).Scan(&score); err != nil {
+		t.Fatalf("expected co-play edge ArtA→ArtB: %v", err)
+	}
+	// Jaccard of perfectly-overlapping player sets (both users play both):
+	// coplayers=2, players each=2 → 2/(2+2-2) = 1.0.
+	if score != 1.0 {
+		t.Errorf("edge score = %v, want 1.0 (perfect overlap)", score)
+	}
+	if got := coplayEdgeCount(t, pool); got != 2 {
+		t.Errorf("edge count = %d, want 2 (A→B and B→A)", got)
+	}
+}
+
+// TestCoplayWorker_SingleUserNoEdges: one user's own co-listening yields no
+// edges — the >= 2 distinct-co-player gate keeps single-user servers empty.
+func TestCoplayWorker_SingleUserNoEdges(t *testing.T) {
+	pool := testPool(t)
+	ctx := context.Background()
+	q := dbq.New(pool)
+
+	u := seedUser(t, q, "solo")
+	_, tA := seedArtistTrack(t, q, "SoloA")
+	_, tB := seedArtistTrack(t, q, "SoloB")
+	play(t, pool, u, tA)
+	play(t, pool, u, tB)
+
+	if err := NewWorker(pool, discardLogger()).rebuild(ctx); err != nil {
+		t.Fatalf("rebuild: %v", err)
+	}
+	if got := coplayEdgeCount(t, pool); got != 0 {
+		t.Errorf("single-user edge count = %d, want 0 (>= 2 co-player gate)", got)
+	}
+}