feat(recommendations): collect demographics and advance age cohorts

services/recommender/pipu_rec/wave/store.py

@@ -161,6 +161,9 @@ def load_profile(conn: psycopg.Connection[Any], user_id: str) -> Profile:
     ).fetchone()
     if row and row[3] and len(scores) < ONBOARDING_UNTIL:
         apply_onboarding(conn, scores, row[3])
+        apply_demographics(conn, scores, row[3], user_id)
+        for tid in disliked:
+            scores.pop(tid, None)
     exploration = float(row[0]) if row else 0.15
     taste_cf = parse_vector(row[1]) if row else None
     bandit = {str(k): [float(x) for x in v] for k, v in (row[2] or {}).items()} if row else {}
@@ -212,6 +215,35 @@ def apply_onboarding(conn: psycopg.Connection[Any], scores: dict[str, float], ch
         scores.setdefault(tid, v)
 
 
+def apply_demographics(
+    conn: psycopg.Connection[Any], scores: dict[str, float], choice: dict[str, Any], user_id: str
+) -> None:
+    """A weak cold-start prior from actual likes of peers, without assumed musical stereotypes."""
+    gender, age = choice.get("gender"), choice.get("age_group")
+    if not gender and not age:
+        return
+    rows = conn.execute(
+        """
+        WITH own AS (
+            SELECT onboarding FROM user_profiles_rec WHERE user_id = %s
+        ), peers AS (
+            SELECT p.user_id FROM user_profiles_rec p JOIN users u ON u.id = p.user_id, own
+            WHERE p.user_id <> %s AND NOT u.is_blocked
+                AND (%s::text IS NULL OR p.onboarding->>'gender' = %s)
+                AND (%s::text IS NULL OR rec_age_group(p.onboarding) = rec_age_group(own.onboarding))
+        )
+        SELECT r.track_id::text, count(DISTINCT r.user_id) AS fans
+        FROM track_reactions r JOIN peers p ON p.user_id = r.user_id
+        WHERE r.reaction = 'like'
+        GROUP BY r.track_id HAVING count(DISTINCT r.user_id) >= 2
+        ORDER BY fans DESC, r.track_id LIMIT %s
+        """,
+        (user_id, user_id, gender, gender, age, ONBOARDING_LIMIT),
+    )
+    for tid, fans in rows:
+        scores.setdefault(tid, min(0.25, 0.05 * fans))
+
+
 def seed_tracks(conn: psycopg.Connection[Any], seed_type: str, seed_id: str) -> list[str]:
     """Треки, от которых запущена волна: трек, альбом или исполнитель."""
     queries = {