diff --git a/core/radio/__init__.py b/core/radio/__init__.py
new file mode 100644
index 00000000..eb4f86d5
--- /dev/null
+++ b/core/radio/__init__.py
@@ -0,0 +1,22 @@
+"""Radio / auto-play recommendation logic.
+
+Pure, DB-agnostic helpers that decide *what* radio should play. The SQL
+execution stays in ``database.music_database.get_radio_tracks``; this package
+owns the decisions (tag parsing, tier caps, dedup/collection, LIKE-condition
+building) so they're unit-testable without a live DB — the seam Phase 2's
+smarter ranking will plug into.
+"""
+
+from core.radio.selection import (
+    RadioCollector,
+    build_like_conditions,
+    parse_tags,
+    same_artist_cap,
+)
+
+__all__ = [
+    "RadioCollector",
+    "build_like_conditions",
+    "parse_tags",
+    "same_artist_cap",
+]
diff --git a/core/radio/selection.py b/core/radio/selection.py
new file mode 100644
index 00000000..17d008d1
--- /dev/null
+++ b/core/radio/selection.py
@@ -0,0 +1,139 @@
+"""Pure radio-selection decisions, lifted out of the DB layer.
+
+``database.music_database.get_radio_tracks`` used to inline all of this between
+``cursor.execute`` calls, so the algorithm couldn't be tested without a live DB
+(which also happens to throw in the dev sandbox). These helpers carry the same
+behavior as before — they're a faithful extraction, not a rewrite — but as
+plain functions they're unit-testable and give Phase 2 (smart ranking) a clean
+place to evolve the logic.
+
+Nothing here touches sqlite; callers pass already-fetched rows (as dicts) and
+get back decisions.
+"""
+
+from __future__ import annotations
+
+import json
+from typing import Any, Dict, Iterable, List, Optional, Sequence, Tuple
+
+
+def parse_tags(raw_val: Any) -> List[str]:
+    """Parse a genre/mood/style field into a list of tags.
+
+    The field may be a JSON array (canonical) or a legacy comma-separated
+    string. Mirrors the inline ``_parse_tags`` the DB method used.
+    """
+    if not raw_val:
+        return []
+    try:
+        parsed = json.loads(raw_val)
+        return parsed if isinstance(parsed, list) else [str(parsed)]
+    except (json.JSONDecodeError, ValueError, TypeError):
+        return [t.strip() for t in str(raw_val).split(",") if t.strip()]
+
+
+def same_artist_cap(limit: int) -> int:
+    """How many same-artist tracks tier 1 may contribute.
+
+    Capped so radio doesn't become an all-one-artist playlist: 30% of the
+    limit, floored at 5 (matches the original ``max(5, limit * 3 // 10)``).
+    """
+    return max(5, limit * 3 // 10)
+
+
+def merge_tags(*tag_groups: Iterable[str]) -> List[str]:
+    """Concatenate tag lists, dedupe, preserve first-seen order.
+
+    Mirrors ``list(dict.fromkeys(a + b))`` used for genre/mood/style merges.
+    """
+    merged: List[str] = []
+    for group in tag_groups:
+        for tag in group:
+            merged.append(tag)
+    return list(dict.fromkeys(merged))
+
+
+def build_like_conditions(
+    tags: Sequence[str], columns: Sequence[str]
+) -> Tuple[str, List[str]]:
+    """Build an OR-of-LIKEs SQL fragment + params for matching ``tags``
+    against each of ``columns``.
+
+    Returns ``(sql_fragment, params)`` where the fragment is
+    ``"col1 LIKE ? OR col1 LIKE ? OR col2 LIKE ? ..."`` (one LIKE per
+    column per tag) and params are the ``%tag%`` wildcards in matching
+    order. Returns ``("", [])`` when there are no tags or no columns, so
+    callers can skip the tier cleanly.
+
+    This reproduces the original per-tier condition building, which paired
+    every tag against album-level and artist-level columns.
+    """
+    if not tags or not columns:
+        return "", []
+    conditions: List[str] = []
+    params: List[str] = []
+    # Group by column (all tags for column A, then all tags for column B) to
+    # match the original ordering: it emitted every ``al.<f> LIKE ?`` then
+    # every ``ar.<f> LIKE ?``, with params being ``[%tag%...] * 2``.
+    for col in columns:
+        for tag in tags:
+            conditions.append(f"{col} LIKE ?")
+            params.append(f"%{tag}%")
+    return " OR ".join(conditions), params
+
+
+class RadioCollector:
+    """Accumulates radio candidates across tiers with dedup + cap logic.
+
+    Replaces the inline ``collected`` list + ``seen_ids`` set + ``_collect``
+    closure the DB method used. Construct with the overall ``limit`` and the
+    set of IDs to exclude up front (seed track + caller-supplied), then feed
+    each tier's fetched rows through :meth:`collect`.
+    """
+
+    def __init__(self, limit: int, exclude_ids: Optional[Iterable[Any]] = None):
+        self.limit = limit
+        self._collected: List[Dict[str, Any]] = []
+        # seen_ids seeds with the exclude set so excluded tracks never collect
+        # AND so the placeholders/values used in WHERE ... NOT IN stay in sync.
+        self._seen: set[str] = {str(e) for e in (exclude_ids or [])}
+
+    @property
+    def tracks(self) -> List[Dict[str, Any]]:
+        return self._collected
+
+    @property
+    def filled(self) -> bool:
+        """True once we've reached the overall limit."""
+        return len(self._collected) >= self.limit
+
+    def exclude_placeholders(self) -> str:
+        """SQL ``?,?,...`` placeholder string sized to the current seen set."""
+        return ",".join("?" * len(self._seen))
+
+    def exclude_values(self) -> List[str]:
+        """Param values for the placeholders above (current seen set)."""
+        return list(self._seen)
+
+    def remaining(self) -> int:
+        """How many more tracks are needed to hit the limit."""
+        return max(0, self.limit - len(self._collected))
+
+    def collect(self, rows: Iterable[Dict[str, Any]], cap: Optional[int] = None) -> bool:
+        """Append ``rows`` (dict-like) to the result, skipping already-seen IDs.
+
+        ``cap`` bounds how many THIS call may add (on top of what's already
+        collected); ``None`` means bounded only by the overall limit. Returns
+        True once the overall limit is reached. Mirrors the original
+        ``_collect`` closure exactly.
+        """
+        target = min(self.limit, len(self._collected) + cap) if cap else self.limit
+        for row in rows:
+            r = dict(row)
+            rid = str(r["id"])
+            if rid not in self._seen:
+                self._seen.add(rid)
+                self._collected.append(r)
+                if len(self._collected) >= target:
+                    return True
+        return self.filled
diff --git a/database/music_database.py b/database/music_database.py
index 1394719f..46f87d88 100644
--- a/database/music_database.py
+++ b/database/music_database.py
@@ -12796,19 +12796,24 @@ class MusicDatabase:
                 seed = dict(seed)
                 artist_name = seed['artist_name']
 
-                # Build the set of IDs to exclude (seed + caller-supplied)
-                excluded = {str(track_id)}
-                if exclude_ids:
-                    excluded.update(str(eid) for eid in exclude_ids)
-
-                collected: list[dict] = []
-                seen_ids: set[str] = set(excluded)
-
-                def _exclude_placeholders():
-                    return ','.join('?' * len(seen_ids))
+                # Selection decisions (dedup, caps, tag parsing, condition
+                # building) live in core.radio.selection so they're unit-
+                # testable without a live DB. The cursor work stays here.
+                from core.radio.selection import (
+                    RadioCollector,
+                    build_like_conditions,
+                    merge_tags,
+                    parse_tags,
+                    same_artist_cap,
+                )
 
-                def _exclude_values():
-                    return list(seen_ids)
+                # Seed + caller-supplied IDs to exclude (seeds the collector's
+                # seen-set so excluded tracks never collect and the NOT IN
+                # placeholders/values stay in sync).
+                exclude_seed = [str(track_id)]
+                if exclude_ids:
+                    exclude_seed.extend(str(eid) for eid in exclude_ids)
+                collector = RadioCollector(limit, exclude_ids=exclude_seed)
 
                 _track_select = """
                     SELECT t.id, t.title, t.track_number, t.duration,
@@ -12824,98 +12829,71 @@ class MusicDatabase:
                 # Only return tracks that have actual files on disk
                 _file_filter = "t.file_path IS NOT NULL AND t.file_path != ''"
 
-                def _collect(rows, cap=None):
-                    """Append rows to collected. Stop at cap or limit."""
-                    target = min(limit, (len(collected) + cap)) if cap else limit
-                    for row in rows:
-                        r = dict(row)
-                        rid = str(r['id'])
-                        if rid not in seen_ids:
-                            seen_ids.add(rid)
-                            collected.append(r)
-                            if len(collected) >= target:
-                                return True
-                    return len(collected) >= limit
-
-                def _parse_tags(raw_val):
-                    """Parse a JSON array or comma-separated string into a list."""
-                    if not raw_val:
-                        return []
-                    try:
-                        parsed = json.loads(raw_val)
-                        return parsed if isinstance(parsed, list) else [str(parsed)]
-                    except (json.JSONDecodeError, ValueError):
-                        return [t.strip() for t in raw_val.split(',') if t.strip()]
-
                 # --- 1. Same artist, different albums (capped at 30% of limit) ---
-                same_artist_cap = max(5, limit * 3 // 10)
+                artist_cap = same_artist_cap(limit)
                 cursor.execute(f"""
                     {_track_select}
-                    WHERE {_file_filter} AND ar.name = ? AND t.album_id != ? AND t.id NOT IN ({_exclude_placeholders()})
+                    WHERE {_file_filter} AND ar.name = ? AND t.album_id != ? AND t.id NOT IN ({collector.exclude_placeholders()})
                     ORDER BY RANDOM()
                     LIMIT ?
-                """, [artist_name, seed['album_id']] + _exclude_values() + [same_artist_cap])
-                _collect(cursor.fetchall(), cap=same_artist_cap)
+                """, [artist_name, seed['album_id']] + collector.exclude_values() + [artist_cap])
+                collector.collect(cursor.fetchall(), cap=artist_cap)
 
-                if len(collected) >= limit:
-                    return {'success': True, 'tracks': collected}
+                if collector.filled:
+                    return {'success': True, 'tracks': collector.tracks}
 
                 # --- 2. Same genre (album genres + artist genres, other artists) ---
-                genre_list = _parse_tags(seed.get('album_genres'))
-                artist_genre_list = _parse_tags(seed.get('artist_genres'))
-                all_genres = list(dict.fromkeys(genre_list + artist_genre_list))  # dedupe, preserve order
-
-                if all_genres:
-                    genre_conditions = ' OR '.join(
-                        ['al.genres LIKE ?' for _ in all_genres] +
-                        ['ar.genres LIKE ?' for _ in all_genres]
-                    )
-                    genre_params = [f'%{g}%' for g in all_genres] * 2
+                all_genres = merge_tags(
+                    parse_tags(seed.get('album_genres')),
+                    parse_tags(seed.get('artist_genres')),
+                )
+                genre_conditions, genre_params = build_like_conditions(
+                    all_genres, ('al.genres', 'ar.genres')
+                )
+                if genre_conditions:
                     cursor.execute(f"""
                         {_track_select}
                         WHERE {_file_filter} AND ({genre_conditions})
                           AND ar.name != ?
-                          AND t.id NOT IN ({_exclude_placeholders()})
+                          AND t.id NOT IN ({collector.exclude_placeholders()})
                         ORDER BY RANDOM()
                         LIMIT ?
-                    """, genre_params + [artist_name] + _exclude_values() + [limit - len(collected)])
-                    if _collect(cursor.fetchall()):
-                        return {'success': True, 'tracks': collected}
+                    """, genre_params + [artist_name] + collector.exclude_values() + [collector.remaining()])
+                    if collector.collect(cursor.fetchall()):
+                        return {'success': True, 'tracks': collector.tracks}
 
                 # --- 3. Same mood / style (album + artist level) ---
                 for field_name in ('mood', 'style'):
-                    album_tags = _parse_tags(seed.get(f'album_{field_name}'))
-                    artist_tags = _parse_tags(seed.get(f'artist_{field_name}'))
-                    all_tags = list(dict.fromkeys(album_tags + artist_tags))
-
-                    if all_tags:
-                        tag_conditions = ' OR '.join(
-                            [f'al.{field_name} LIKE ?' for _ in all_tags] +
-                            [f'ar.{field_name} LIKE ?' for _ in all_tags]
-                        )
-                        tag_params = [f'%{t}%' for t in all_tags] * 2
+                    all_tags = merge_tags(
+                        parse_tags(seed.get(f'album_{field_name}')),
+                        parse_tags(seed.get(f'artist_{field_name}')),
+                    )
+                    tag_conditions, tag_params = build_like_conditions(
+                        all_tags, (f'al.{field_name}', f'ar.{field_name}')
+                    )
+                    if tag_conditions:
                         cursor.execute(f"""
                             {_track_select}
                             WHERE {_file_filter} AND ({tag_conditions})
                               AND ar.name != ?
-                              AND t.id NOT IN ({_exclude_placeholders()})
+                              AND t.id NOT IN ({collector.exclude_placeholders()})
                             ORDER BY RANDOM()
                             LIMIT ?
-                        """, tag_params + [artist_name] + _exclude_values() + [limit - len(collected)])
-                        if _collect(cursor.fetchall()):
-                            return {'success': True, 'tracks': collected}
+                        """, tag_params + [artist_name] + collector.exclude_values() + [collector.remaining()])
+                        if collector.collect(cursor.fetchall()):
+                            return {'success': True, 'tracks': collector.tracks}
 
                 # --- 4. Random library tracks ---
-                if len(collected) < limit:
+                if not collector.filled:
                     cursor.execute(f"""
                         {_track_select}
-                        WHERE {_file_filter} AND t.id NOT IN ({_exclude_placeholders()})
+                        WHERE {_file_filter} AND t.id NOT IN ({collector.exclude_placeholders()})
                         ORDER BY RANDOM()
                         LIMIT ?
-                    """, _exclude_values() + [limit - len(collected)])
-                    _collect(cursor.fetchall())
+                    """, collector.exclude_values() + [collector.remaining()])
+                    collector.collect(cursor.fetchall())
 
-                return {'success': True, 'tracks': collected}
+                return {'success': True, 'tracks': collector.tracks}
 
         except Exception as e:
             logger.error(f"Error getting radio tracks for track {track_id}: {e}")
diff --git a/revamp_plan.md b/revamp_plan.md
new file mode 100644
index 00000000..6aa952a1
--- /dev/null
+++ b/revamp_plan.md
@@ -0,0 +1,37 @@
+# Stream / Player / Radio Revamp — Plan
+
+Goal: bring the audio stream + media-player + radio system to Spotify/Apple-level polish and feature set. Target stack: **plain JS** (`webui/static/media-player.js`), not the React migration. Intended architecture direction: **multi-listener** (final call deferred to Phase 3; Phases 0–2 stay compatible either way).
+
+Rule for every phase: kettui standard — importable/testable logic, seam-level + differential tests, break nothing, ship one reviewable phase at a time.
+
+---
+
+## Phase 0 — Make it provable (foundation, no user-visible change)
+
+- [ ] **0a. Extract radio selection logic into testable `core/radio/`.** The algorithm (tier orchestration, cap math, dedup, tag parsing, SQL-condition building) is currently tangled with `cursor.execute` inside `database/music_database.py:get_radio_tracks` (~12756) — untestable without a live DB. Pull the pure decisions into `core/radio/selection.py`; the DB method keeps SQL execution but delegates the decisions. Differential-test: same inputs → same output as today.
+- [ ] **0b. Centralize frontend player state.** ~10 scattered `np*` globals in `media-player.js` → one `PlayerState` object. Seam for every later frontend phase. No behavior change.
+
+## Phase 1 — Polish / feel (frontend)
+
+- [ ] Persistent queue across refresh (localStorage first; server-side in P3)
+- [ ] Drag-to-reorder queue; duration + art per queue item
+- [ ] Seek tooltip (hover timestamp); smoother progress
+- [ ] Crossfade via dual-`<audio>` swap (honest approximation of gapless — true gapless impossible w/ single element)
+- [ ] Full Media Session API (lockscreen / hardware transport keys)
+- [ ] Keyboard shortcut overlay + fuller bindings
+
+## Phase 2 — Smart radio (backend algorithm)
+
+- [ ] Replace `ORDER BY RANDOM()` with real seeding: play-count + recency weighting, genre-adjacency, recently-played memory. Slots into the Phase-0a pure module → fully unit-testable (seed → expected ordering). Both radio buttons benefit (shared function).
+
+## Phase 3 — Architecture (deepest, riskiest — listener decision lands here)
+
+- [ ] Per-session (or multi-tenant) stream state — replaces the single global `stream_state` + 1-worker executor + single `Stream/` staging file (`web_server.py:747`).
+- [ ] Server-side persistent queue (resume across devices/refresh).
+- [ ] Final multi-listener vs single-listener scope decided here, with real usage in hand.
+
+---
+
+## Order of execution
+
+0a (radio extraction) → 2 (smart radio) first: highest *visible* upgrade, backend-only, cleanest to prove, zero playback risk. Then 0b → 1 (polish). Then 3 (architecture) last.
diff --git a/tests/radio/__init__.py b/tests/radio/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/tests/radio/test_get_radio_tracks_db.py b/tests/radio/test_get_radio_tracks_db.py
new file mode 100644
index 00000000..afc9ef7e
--- /dev/null
+++ b/tests/radio/test_get_radio_tracks_db.py
@@ -0,0 +1,222 @@
+"""End-to-end behavioral pin for MusicDatabase.get_radio_tracks.
+
+Phase 0a extracted the radio SELECTION logic into core.radio.selection but the
+DB method still owns the SQL. These tests drive the REAL get_radio_tracks
+against an in-memory sqlite to prove the refactor preserved behavior — the
+4-tier fallback (same-artist cap → genre → mood/style → random), dedup, and
+exclude handling all still work through the extracted helpers.
+
+Reuses the in-memory MusicDatabase harness pattern from
+tests/test_reorganize_db_methods.py.
+"""
+
+import sqlite3
+import sys
+import types
+
+import pytest
+
+
+# ── stubs (same shape used elsewhere in the suite) ────────────────────────
+if "spotipy" not in sys.modules:
+    spotipy = types.ModuleType("spotipy")
+    spotipy.Spotify = object
+    oauth2 = types.ModuleType("spotipy.oauth2")
+    oauth2.SpotifyOAuth = object
+    oauth2.SpotifyClientCredentials = object
+    spotipy.oauth2 = oauth2
+    sys.modules["spotipy"] = spotipy
+    sys.modules["spotipy.oauth2"] = oauth2
+
+if "config.settings" not in sys.modules:
+    config_pkg = types.ModuleType("config")
+    settings_mod = types.ModuleType("config.settings")
+
+    class _DummyConfigManager:
+        def get(self, key, default=None):
+            return default
+
+        def get_active_media_server(self):
+            return "primary"
+
+    settings_mod.config_manager = _DummyConfigManager()
+    config_pkg.settings = settings_mod
+    sys.modules["config"] = config_pkg
+    sys.modules["config.settings"] = settings_mod
+
+
+from database.music_database import MusicDatabase  # noqa: E402
+
+
+class _InMemoryDB(MusicDatabase):
+    def __init__(self):
+        self._conn = sqlite3.connect(":memory:")
+        self._conn.row_factory = sqlite3.Row
+
+    def _get_connection(self):
+        return _NonClosingConn(self._conn)
+
+
+class _NonClosingConn:
+    def __init__(self, real):
+        self._real = real
+
+    def cursor(self):
+        return self._real.cursor()
+
+    def commit(self):
+        return self._real.commit()
+
+    def close(self):
+        pass
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, *args):
+        pass
+
+
+def _schema(db):
+    cur = db._conn.cursor()
+    cur.execute("""
+        CREATE TABLE artists (
+            id TEXT PRIMARY KEY, name TEXT,
+            genres TEXT, mood TEXT, style TEXT, thumb_url TEXT
+        )
+    """)
+    cur.execute("""
+        CREATE TABLE albums (
+            id TEXT PRIMARY KEY, artist_id TEXT, title TEXT,
+            genres TEXT, mood TEXT, style TEXT, thumb_url TEXT
+        )
+    """)
+    cur.execute("""
+        CREATE TABLE tracks (
+            id TEXT PRIMARY KEY, album_id TEXT, artist_id TEXT,
+            title TEXT, track_number INTEGER, duration INTEGER,
+            file_path TEXT, bitrate INTEGER
+        )
+    """)
+    db._conn.commit()
+
+
+def _add_artist(db, aid, name, genres="", mood="", style=""):
+    db._conn.execute(
+        "INSERT INTO artists (id, name, genres, mood, style, thumb_url) VALUES (?,?,?,?,?,?)",
+        (aid, name, genres, mood, style, ""),
+    )
+
+
+def _add_album(db, alid, aid, title, genres="", mood="", style=""):
+    db._conn.execute(
+        "INSERT INTO albums (id, artist_id, title, genres, mood, style, thumb_url) VALUES (?,?,?,?,?,?,?)",
+        (alid, aid, title, genres, mood, style, ""),
+    )
+
+
+def _add_track(db, tid, alid, aid, title, file_path="/m/x.flac"):
+    db._conn.execute(
+        "INSERT INTO tracks (id, album_id, artist_id, title, track_number, duration, file_path, bitrate) "
+        "VALUES (?,?,?,?,?,?,?,?)",
+        (tid, alid, aid, title, 1, 200, file_path, 1000),
+    )
+
+
+@pytest.fixture
+def db():
+    d = _InMemoryDB()
+    _schema(d)
+    return d
+
+
+def test_missing_seed_track_returns_failure(db):
+    res = db.get_radio_tracks("nope", limit=10)
+    assert res["success"] is False
+
+
+def test_tier1_same_artist_other_albums(db):
+    _add_artist(db, "ar1", "Artist One")
+    _add_album(db, "al1", "ar1", "Album A")
+    _add_album(db, "al2", "ar1", "Album B")
+    _add_track(db, "seed", "al1", "ar1", "Seed")
+    _add_track(db, "t2", "al2", "ar1", "Other Album Track")
+    db._conn.commit()
+
+    res = db.get_radio_tracks("seed", limit=10)
+    assert res["success"] is True
+    ids = [t["id"] for t in res["tracks"]]
+    assert "t2" in ids
+    assert "seed" not in ids          # seed always excluded
+
+
+def test_excludes_caller_supplied_ids(db):
+    _add_artist(db, "ar1", "Artist One")
+    _add_album(db, "al1", "ar1", "Album A")
+    _add_album(db, "al2", "ar1", "Album B")
+    _add_track(db, "seed", "al1", "ar1", "Seed")
+    _add_track(db, "t2", "al2", "ar1", "T2")
+    _add_track(db, "t3", "al2", "ar1", "T3")
+    db._conn.commit()
+
+    res = db.get_radio_tracks("seed", limit=10, exclude_ids=["t2"])
+    ids = [t["id"] for t in res["tracks"]]
+    assert "t2" not in ids
+    assert "t3" in ids
+
+
+def test_tier2_genre_match_other_artists(db):
+    # No same-artist alternatives; falls to genre tier.
+    _add_artist(db, "ar1", "Seed Artist", genres='["shoegaze"]')
+    _add_artist(db, "ar2", "Other Artist", genres='["shoegaze"]')
+    _add_album(db, "al1", "ar1", "Seed Album", genres='["shoegaze"]')
+    _add_album(db, "al2", "ar2", "Other Album", genres='["shoegaze"]')
+    _add_track(db, "seed", "al1", "ar1", "Seed")
+    _add_track(db, "g1", "al2", "ar2", "Genre Match")
+    db._conn.commit()
+
+    res = db.get_radio_tracks("seed", limit=10)
+    ids = [t["id"] for t in res["tracks"]]
+    assert "g1" in ids
+
+
+def test_tier4_random_fallback_fills_when_no_metadata_match(db):
+    # Seed has no genre/mood/style and no same-artist alts → random tier.
+    _add_artist(db, "ar1", "Seed Artist")
+    _add_artist(db, "ar2", "Unrelated")
+    _add_album(db, "al1", "ar1", "Seed Album")
+    _add_album(db, "al2", "ar2", "Unrelated Album")
+    _add_track(db, "seed", "al1", "ar1", "Seed")
+    _add_track(db, "r1", "al2", "ar2", "Random One")
+    db._conn.commit()
+
+    res = db.get_radio_tracks("seed", limit=10)
+    ids = [t["id"] for t in res["tracks"]]
+    assert "r1" in ids                # filled from random tier
+
+
+def test_only_returns_tracks_with_files(db):
+    _add_artist(db, "ar1", "Artist One")
+    _add_album(db, "al1", "ar1", "Album A")
+    _add_album(db, "al2", "ar1", "Album B")
+    _add_track(db, "seed", "al1", "ar1", "Seed")
+    _add_track(db, "nofile", "al2", "ar1", "No File", file_path="")
+    db._conn.commit()
+
+    res = db.get_radio_tracks("seed", limit=10)
+    ids = [t["id"] for t in res["tracks"]]
+    assert "nofile" not in ids        # file_path filter still enforced
+
+
+def test_no_duplicate_ids_across_tiers(db):
+    # A track that qualifies for both same-artist AND genre must appear once.
+    _add_artist(db, "ar1", "Artist One", genres='["pop"]')
+    _add_album(db, "al1", "ar1", "Album A", genres='["pop"]')
+    _add_album(db, "al2", "ar1", "Album B", genres='["pop"]')
+    _add_track(db, "seed", "al1", "ar1", "Seed")
+    _add_track(db, "dup", "al2", "ar1", "Could Match Twice")
+    db._conn.commit()
+
+    res = db.get_radio_tracks("seed", limit=10)
+    ids = [t["id"] for t in res["tracks"]]
+    assert ids.count("dup") == 1
diff --git a/tests/radio/test_selection.py b/tests/radio/test_selection.py
new file mode 100644
index 00000000..9a9564c5
--- /dev/null
+++ b/tests/radio/test_selection.py
@@ -0,0 +1,139 @@
+"""Tests for the extracted radio-selection logic (Phase 0a of the player revamp).
+
+These pin the behavior that used to be inline + untestable inside
+``database.music_database.get_radio_tracks``. They lock current behavior so
+Phase 2 (smart ranking) can evolve it against a green baseline.
+"""
+
+from __future__ import annotations
+
+from core.radio.selection import (
+    RadioCollector,
+    build_like_conditions,
+    merge_tags,
+    parse_tags,
+    same_artist_cap,
+)
+
+
+class TestParseTags:
+    def test_json_array(self):
+        assert parse_tags('["rock", "indie"]') == ["rock", "indie"]
+
+    def test_comma_separated_legacy(self):
+        assert parse_tags("rock, indie, folk") == ["rock", "indie", "folk"]
+
+    def test_comma_separated_strips_whitespace_and_blanks(self):
+        assert parse_tags("rock,  , indie ,") == ["rock", "indie"]
+
+    def test_empty_and_none(self):
+        assert parse_tags("") == []
+        assert parse_tags(None) == []
+
+    def test_non_list_json_scalar_wrapped(self):
+        # A bare JSON scalar (e.g. a quoted string) becomes a single-item list.
+        assert parse_tags('"rock"') == ["rock"]
+
+    def test_garbage_falls_back_to_split(self):
+        assert parse_tags("not json at all") == ["not json at all"]
+
+
+class TestSameArtistCap:
+    def test_thirty_percent(self):
+        assert same_artist_cap(50) == 15   # 50 * 3 // 10
+        assert same_artist_cap(20) == 6
+
+    def test_floored_at_five(self):
+        assert same_artist_cap(10) == 5     # 3, floored to 5
+        assert same_artist_cap(1) == 5
+
+
+class TestMergeTags:
+    def test_dedupes_preserving_order(self):
+        assert merge_tags(["rock", "indie"], ["indie", "folk"]) == ["rock", "indie", "folk"]
+
+    def test_empty_groups(self):
+        assert merge_tags([], []) == []
+
+
+class TestBuildLikeConditions:
+    def test_single_tag_two_columns(self):
+        sql, params = build_like_conditions(["rock"], ("al.genres", "ar.genres"))
+        assert sql == "al.genres LIKE ? OR ar.genres LIKE ?"
+        assert params == ["%rock%", "%rock%"]
+
+    def test_grouping_matches_original_order(self):
+        # Original emitted all album-col LIKEs, then all artist-col LIKEs;
+        # params were [%t%...] * 2. Reproduce that ordering exactly.
+        sql, params = build_like_conditions(["rock", "indie"], ("al.genres", "ar.genres"))
+        assert sql == (
+            "al.genres LIKE ? OR al.genres LIKE ? OR "
+            "ar.genres LIKE ? OR ar.genres LIKE ?"
+        )
+        assert params == ["%rock%", "%indie%", "%rock%", "%indie%"]
+
+    def test_no_tags_returns_empty(self):
+        assert build_like_conditions([], ("al.genres",)) == ("", [])
+
+    def test_no_columns_returns_empty(self):
+        assert build_like_conditions(["rock"], ()) == ("", [])
+
+
+class TestRadioCollector:
+    def _rows(self, *ids):
+        return [{"id": i, "title": f"t{i}"} for i in ids]
+
+    def test_collects_and_dedupes(self):
+        c = RadioCollector(limit=10)
+        c.collect(self._rows(1, 2, 2, 3))   # dup 2 ignored
+        assert [t["id"] for t in c.tracks] == [1, 2, 3]
+
+    def test_excludes_seed_and_caller_ids(self):
+        c = RadioCollector(limit=10, exclude_ids=["1", "2"])
+        c.collect(self._rows(1, 2, 3, 4))
+        assert [t["id"] for t in c.tracks] == [3, 4]
+
+    def test_exclude_ids_coerced_to_str(self):
+        # Caller may pass ints; seen-set stores strings.
+        c = RadioCollector(limit=10, exclude_ids=[1])
+        c.collect(self._rows(1, 2))
+        assert [t["id"] for t in c.tracks] == [2]
+
+    def test_cap_bounds_a_single_tier(self):
+        c = RadioCollector(limit=10)
+        c.collect(self._rows(1, 2, 3, 4, 5), cap=2)   # only 2 from this tier
+        assert [t["id"] for t in c.tracks] == [1, 2]
+        assert not c.filled
+        assert c.remaining() == 8
+
+    def test_filled_at_limit(self):
+        c = RadioCollector(limit=3)
+        ret = c.collect(self._rows(1, 2, 3, 4))
+        assert ret is True
+        assert c.filled
+        assert len(c.tracks) == 3
+        assert c.remaining() == 0
+
+    def test_capped_collect_returns_true_at_cap_target(self):
+        # Faithful to the original _collect: it returns True once the
+        # cap-bounded target is hit, even below the overall limit. The DB
+        # method IGNORES tier 1's capped return and checks .filled instead, so
+        # this never causes early exit — but the contract must match exactly.
+        c = RadioCollector(limit=5)
+        assert c.collect(self._rows(1, 2), cap=2) is True   # hit cap target (2)
+        assert not c.filled                                  # but not at limit (5)
+
+    def test_uncapped_collect_returns_true_only_at_limit(self):
+        c = RadioCollector(limit=5)
+        assert c.collect(self._rows(1, 2)) is False          # below limit
+        assert c.collect(self._rows(3, 4, 5)) is True         # now at limit
+
+    def test_exclude_placeholders_and_values_track_seen_set(self):
+        c = RadioCollector(limit=10, exclude_ids=["a", "b"])
+        assert c.exclude_placeholders() == "?,?"
+        assert set(c.exclude_values()) == {"a", "b"}
+        # After collecting, already-collected IDs join the NOT-IN set so the
+        # next tier's SQL won't re-pull them.
+        c.collect(self._rows("c"))
+        assert c.exclude_placeholders() == "?,?,?"
+        assert set(c.exclude_values()) == {"a", "b", "c"}