Milestone 3: Navidrome (OpenSubsonic) playback provider

Replace the directory-scan queue with a real ingestion pipeline:
provider -> fetcher -> cache -> ready queue, driven by a background
prefetch thread.

- subsonic.py: minimal OpenSubsonic client (salted-token auth,
  getPlaylists/getPlaylist, raw streaming download).
- providers/navidrome.py: pick tracks from a playlist (by name or id),
  with anti-repeat and periodic playlist reload.
- fetchers/subsonic.py: atomic download into the shared cache.
- db.py: SQLite state — append-only play history (anti-repeat + stats)
  and cache_files LRU retention (keep the N most recently played).
- queue.py: prefetch buffer + retention on play; graceful degradation
  to the stream's local-cache fallback when no source is configured.
- api.py: GET /next now carries real title/artist metadata.
- Config via .env (Navidrome credentials), persistent state/ volume,
  httpx dependency.

Verified end-to-end against a live Navidrome: playlist resolved,
tracks downloaded and broadcast, retention and history correct.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

.env.example

@@ -0,0 +1,15 @@
+# radieo — configuration locale. Copier en `.env` et remplir.
+# docker compose lit automatiquement `.env` pour ces variables.
+
+# --- Source Navidrome / OpenSubsonic ---
+# URL de base de ton serveur (sans /rest). Laisser les champs vides désactive
+# la source : le stream joue alors uniquement les fichiers déjà dans cache/.
+RADIEO_NAVIDROME_URL=https://navidrome.example.org
+RADIEO_NAVIDROME_USER=monuser
+RADIEO_NAVIDROME_PASSWORD=monmotdepasse
+# Nom OU identifiant de la playlist à diffuser.
+RADIEO_NAVIDROME_PLAYLIST=Radio
+
+# --- Rétention du cache (optionnel) ---
+# Nombre de morceaux joués conservés sur disque avant éviction (LRU).
+RADIEO_RETENTION_KEEP=20

.gitignore

@@ -2,12 +2,16 @@
 /cache/*
 !/cache/.gitkeep
 
-# État de l'ingestion (jalons suivants)
+# Secrets locaux
+.env
+
+# État de l'ingestion (base SQLite persistante)
+/state/
 *.db
 *.db-journal
 *.db-wal
 
-# Python (jalons suivants)
+# Python
 __pycache__/
 *.pyc
 .venv/

README.md

@@ -46,29 +46,43 @@ The stream is MP3 at 192 kbps. Multiple clients can listen at the same time.
 New files dropped into `cache/` are picked up automatically (the playlist is
 reloaded when the directory changes).
 
+## Configuration
+
+Copy `.env.example` to `.env` and fill in your Navidrome details:
+
+```sh
+cp .env.example .env
+# edit .env: RADIEO_NAVIDROME_URL / USER / PASSWORD / PLAYLIST
+```
+
+If the Navidrome variables are left empty, the source is simply disabled and
+the stream plays whatever is already in `cache/` (the milestone-1/2 behaviour).
+
 ## Current status
 
-**Milestone 2 — ingestion daemon: done.**
+**Milestone 3 — Navidrome provider: done.**
 
-- `ingest` (Python) container exposes `GET /next`, returning the next track as
+- `ingest` pulls tracks from an OpenSubsonic playlist (Navidrome), downloading
-  an annotated Liquidsoap URI (or an empty body when nothing is ready).
+  them into the shared cache ahead of playback (prefetch buffer).
-- `stream` (Liquidsoap v2.4.5) pulls from `ingest` via a `request.dynamic`
+- Play history and LRU retention are tracked in a SQLite database under
-  source, and falls back to the local `cache/` directory when the daemon has
+  `state/`: only the N most recently played files are kept on disk
-  nothing to offer.
+  (`RADIEO_RETENTION_KEEP`, default 20); anti-repeat avoids replaying a track
-- HTTP stream served at `http://localhost:8000/radio.mp3` (MP3, 192 kbps).
+  seen among the last plays.
-- Continuous output guaranteed: silence rather than a crash when everything is
+- `GET /next` returns the next track as an annotated Liquidsoap URI with real
-  empty (`mksafe`).
+  title/artist metadata (or an empty body when nothing is ready).
-- Multiple simultaneous listeners supported.
+- `stream` (Liquidsoap v2.4.5) pulls via `request.dynamic` and falls back to the
+  local `cache/` directory; `mksafe` guarantees silence rather than a crash.
+- HTTP stream served at `http://localhost:8000/radio.mp3` (MP3, 192 kbps),
+  multiple simultaneous listeners supported.
 
-At this stage the daemon just cycles through the files already in `cache/`; the
+The yt-dlp and ListenBrainz sources come next.
-download providers (Navidrome, yt-dlp, ListenBrainz) come next.
 
 ## Roadmap
 
 1. ✅ **Broadcasting skeleton** — Liquidsoap serving the cache directory.
 2. ✅ **Ingestion daemon** — Python daemon exposing `GET /next`; Liquidsoap
    switches to a `request.dynamic` source with the cache as fallback.
-3. **Navidrome provider** — play from an OpenSubsonic playlist, with caching,
+3. ✅ **Navidrome provider** — play from an OpenSubsonic playlist, with caching,
    LRU retention and play history.
 4. **yt-dlp provider** — fetch tracks from a maintained URL/artist list; weighted
    mixing between sources.

docker-compose.yml

@@ -3,10 +3,19 @@ services:
     build: ./ingest
     image: radieo-ingest
     volumes:
-      - ./cache:/cache          # volume partagé avec le stream
+      - ./cache:/cache          # volume partagé avec le stream (rw : téléchargements)
+      - ./state:/state          # état persistant (SQLite) hors du cache éphémère
     environment:
       - RADIEO_CACHE_DIR=/cache
+      - RADIEO_STATE_DIR=/state
       - RADIEO_HTTP_PORT=8080
+      # Source Navidrome / OpenSubsonic (voir .env / .env.example).
+      # Laisser vide désactive la source : le stream joue alors son cache local.
+      - RADIEO_NAVIDROME_URL=${RADIEO_NAVIDROME_URL:-}
+      - RADIEO_NAVIDROME_USER=${RADIEO_NAVIDROME_USER:-}
+      - RADIEO_NAVIDROME_PASSWORD=${RADIEO_NAVIDROME_PASSWORD:-}
+      - RADIEO_NAVIDROME_PLAYLIST=${RADIEO_NAVIDROME_PLAYLIST:-}
+      - RADIEO_RETENTION_KEEP=${RADIEO_RETENTION_KEEP:-20}
     restart: unless-stopped
 
   stream:

ingest/Dockerfile

@@ -2,8 +2,9 @@ FROM python:3.12-slim
 
 WORKDIR /app
 
-# Milestone 2 uses the standard library only; third-party dependencies
+COPY requirements.txt ./
-# (httpx, yt-dlp, feedparser…) will be added in later milestones.
+RUN pip install --no-cache-dir -r requirements.txt
+
 COPY radieo/ ./radieo/
 
 ENV PYTHONUNBUFFERED=1

ingest/radieo/__main__.py

@@ -1,34 +1,76 @@
-"""Entry point: start the HTTP API serving the track queue."""
+"""Entry point: wire providers/fetchers into the queue and serve GET /next."""
 
 import logging
 
 from . import config
 from .api import IngestServer
+from .db import Database
 from .queue import TrackQueue
 
+log = logging.getLogger("radieo")
+
+
+class _NullProvider:
+    """Fallback provider when no source is configured: never yields a track."""
+
+    def next(self):
+        return None
+
+
+def _build_pipeline(db: Database):
+    """Return (provider, fetcher). Falls back to a no-op provider when the
+    Navidrome source is not configured, so the daemon still runs and the
+    stream plays its local cache."""
+    if not config.NAVIDROME_ENABLED:
+        log.warning(
+            "Navidrome not configured (RADIEO_NAVIDROME_*): no source active, "
+            "the stream will play its local cache fallback."
+        )
+        return _NullProvider(), None
+
+    from .fetchers.subsonic import SubsonicFetcher
+    from .providers.navidrome import NavidromeProvider
+    from .subsonic import SubsonicClient
+
+    client = SubsonicClient(
+        config.NAVIDROME_URL, config.NAVIDROME_USER, config.NAVIDROME_PASSWORD
+    )
+    provider = NavidromeProvider(client, config.NAVIDROME_PLAYLIST, db)
+    fetcher = SubsonicFetcher(client, config.CACHE_DIR)
+    log.info("Navidrome source enabled (playlist=%r)", config.NAVIDROME_PLAYLIST)
+    return provider, fetcher
+
 
 def main() -> None:
     logging.basicConfig(
         level=logging.INFO,
         format="%(asctime)s %(levelname)s %(name)s: %(message)s",
     )
-    log = logging.getLogger("radieo")
 
     config.CACHE_DIR.mkdir(parents=True, exist_ok=True)
+    config.STATE_DIR.mkdir(parents=True, exist_ok=True)
 
-    server = IngestServer((config.HTTP_HOST, config.HTTP_PORT), TrackQueue())
+    db = Database(config.STATE_DB)
+    provider, fetcher = _build_pipeline(db)
+    queue = TrackQueue(provider, fetcher, db)
+    queue.start()
+
+    server = IngestServer((config.HTTP_HOST, config.HTTP_PORT), queue)
     log.info(
-        "ingest listening on %s:%d (cache=%s)",
+        "ingest listening on %s:%d (cache=%s, state=%s)",
         config.HTTP_HOST,
         config.HTTP_PORT,
         config.CACHE_DIR,
+        config.STATE_DB,
     )
     try:
         server.serve_forever()
     except KeyboardInterrupt:
         pass
     finally:
+        queue.stop()
         server.server_close()
+        db.close()
 
 
 if __name__ == "__main__":

ingest/radieo/api.py

@@ -10,24 +10,22 @@ import logging
 from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
 from pathlib import Path
 
+from .models import Track
 from .queue import TrackQueue
 
 log = logging.getLogger("radieo.api")
 
 
-def annotate_uri(path: Path) -> str:
+def annotate_uri(path: Path, track: Track) -> str:
-    """Build an annotated Liquidsoap request URI for a cache file.
+    """Build an annotated Liquidsoap request URI for a cache file."""
-
-    Metadata is minimal for now (title derived from the filename); real
-    metadata will come from the providers in later milestones.
-    """
 
     def esc(value: str) -> str:
         return value.replace("\\", "\\\\").replace('"', '\\"')
 
-    title = esc(path.stem)
+    return (
-    artist = esc("radieo")
+        f'annotate:title="{esc(track.title)}",artist="{esc(track.artist)}"'
-    return f'annotate:title="{title}",artist="{artist}":{path}'
+        f":{path}"
+    )
 
 
 class IngestServer(ThreadingHTTPServer):
@@ -48,13 +46,14 @@ class _Handler(BaseHTTPRequestHandler):
             self._text(404, "not found\n")
 
     def _serve_next(self):
-        track = self.server.queue.pop_next()
+        result = self.server.queue.pop_next()
-        if track is None:
+        if result is None:
             # Empty body: tells Liquidsoap to use its fallback for now.
             self._text(200, "")
             return
-        log.info("next -> %s", track.name)
+        path, track = result
-        self._text(200, annotate_uri(track) + "\n")
+        log.info("next -> %s", track)
+        self._text(200, annotate_uri(path, track) + "\n")
 
     def _text(self, code: int, body: str):
         data = body.encode("utf-8")

ingest/radieo/config.py

@@ -11,8 +11,33 @@ from pathlib import Path
 # be valid inside *that* container, so both mount the cache at the same path.
 CACHE_DIR = Path(os.environ.get("RADIEO_CACHE_DIR", "/cache"))
 
+# Persistent state (SQLite). Kept out of the ephemeral cache directory.
+STATE_DIR = Path(os.environ.get("RADIEO_STATE_DIR", "/state"))
+STATE_DB = STATE_DIR / "radieo.db"
+
 HTTP_HOST = os.environ.get("RADIEO_HTTP_HOST", "0.0.0.0")
 HTTP_PORT = int(os.environ.get("RADIEO_HTTP_PORT", "8080"))
 
-# File extensions considered playable when scanning the cache.
+# --- Prefetching / retention ---
-AUDIO_EXTENSIONS = {".mp3", ".flac", ".ogg", ".opus", ".m4a", ".aac", ".wav"}
+# How many downloaded tracks to keep ready ahead of playback.
+PREFETCH = int(os.environ.get("RADIEO_PREFETCH", "3"))
+# Seconds between prefetch-loop wake-ups.
+PREFETCH_INTERVAL = float(os.environ.get("RADIEO_PREFETCH_INTERVAL", "2.0"))
+# Keep the N most recently played files on disk; evict older ones (LRU).
+RETENTION_KEEP = int(os.environ.get("RADIEO_RETENTION_KEEP", "20"))
+# Do not replay a track seen among the last N plays, when avoidable.
+ANTIREPEAT_WINDOW = int(os.environ.get("RADIEO_ANTIREPEAT_WINDOW", "50"))
+
+# --- Navidrome / OpenSubsonic source ---
+# Left empty means the provider is disabled (the stream then plays its own
+# local-cache fallback). Credentials are expected to come from a .env file.
+NAVIDROME_URL = os.environ.get("RADIEO_NAVIDROME_URL", "").strip()
+NAVIDROME_USER = os.environ.get("RADIEO_NAVIDROME_USER", "").strip()
+NAVIDROME_PASSWORD = os.environ.get("RADIEO_NAVIDROME_PASSWORD", "")
+NAVIDROME_PLAYLIST = os.environ.get("RADIEO_NAVIDROME_PLAYLIST", "").strip()
+# How often to reload the playlist contents, in seconds.
+PLAYLIST_REFRESH = float(os.environ.get("RADIEO_PLAYLIST_REFRESH", "300"))
+
+NAVIDROME_ENABLED = bool(
+    NAVIDROME_URL and NAVIDROME_USER and NAVIDROME_PASSWORD and NAVIDROME_PLAYLIST
+)

ingest/radieo/db.py

@@ -0,0 +1,109 @@
+"""SQLite state: play history (anti-repeat + stats) and cache-file retention.
+
+Two concerns, two tables:
+
+- ``history`` is append-only. It drives anti-repeat (recently played track
+  keys) and survives cache eviction, so a track can stay "recently played" even
+  after its file is deleted.
+- ``cache_files`` tracks downloaded files so we can keep only the N most
+  recently *played* ones (LRU retention). Files not yet played are never
+  evicted.
+"""
+
+import sqlite3
+import threading
+import time
+from pathlib import Path
+
+from .models import Track
+
+_SCHEMA = """
+CREATE TABLE IF NOT EXISTS history (
+    id        INTEGER PRIMARY KEY,
+    track_key TEXT NOT NULL,
+    artist    TEXT,
+    title     TEXT,
+    origin    TEXT,
+    played_at REAL NOT NULL
+);
+CREATE INDEX IF NOT EXISTS idx_history_played_at ON history(played_at);
+
+CREATE TABLE IF NOT EXISTS cache_files (
+    path      TEXT PRIMARY KEY,
+    track_key TEXT,
+    played_at REAL           -- NULL until the file has been played
+);
+"""
+
+
+class Database:
+    def __init__(self, path: Path):
+        path.parent.mkdir(parents=True, exist_ok=True)
+        self._lock = threading.Lock()
+        self._conn = sqlite3.connect(
+            path, check_same_thread=False, isolation_level=None
+        )
+        self._conn.row_factory = sqlite3.Row
+        self._conn.executescript(_SCHEMA)
+
+    # --- anti-repeat / history -------------------------------------------
+
+    def recent_keys(self, limit: int) -> set[str]:
+        with self._lock:
+            rows = self._conn.execute(
+                "SELECT track_key FROM history ORDER BY played_at DESC LIMIT ?",
+                (limit,),
+            ).fetchall()
+        return {r["track_key"] for r in rows}
+
+    def record_play(self, track: Track) -> None:
+        with self._lock:
+            self._conn.execute(
+                "INSERT INTO history (track_key, artist, title, origin, played_at)"
+                " VALUES (?, ?, ?, ?, ?)",
+                (track.key, track.artist, track.title, track.origin, time.time()),
+            )
+
+    # --- cache-file retention --------------------------------------------
+
+    def register_download(self, path: str, track_key: str) -> None:
+        with self._lock:
+            self._conn.execute(
+                "INSERT OR REPLACE INTO cache_files (path, track_key, played_at)"
+                " VALUES (?, ?, NULL)",
+                (path, track_key),
+            )
+
+    def mark_played(self, path: str) -> None:
+        with self._lock:
+            self._conn.execute(
+                "UPDATE cache_files SET played_at = ? WHERE path = ?",
+                (time.time(), path),
+            )
+
+    def evict(self, keep: int) -> list[str]:
+        """Return (and forget) played files beyond the ``keep`` most recent.
+
+        The caller is responsible for deleting the returned files from disk.
+        Never touches files that have not been played yet.
+        """
+        with self._lock:
+            rows = self._conn.execute(
+                "SELECT path FROM cache_files WHERE played_at IS NOT NULL"
+                " AND path NOT IN ("
+                "   SELECT path FROM cache_files WHERE played_at IS NOT NULL"
+                "   ORDER BY played_at DESC LIMIT ?"
+                ")",
+                (keep,),
+            ).fetchall()
+            paths = [r["path"] for r in rows]
+            if paths:
+                self._conn.executemany(
+                    "DELETE FROM cache_files WHERE path = ?",
+                    [(p,) for p in paths],
+                )
+        return paths
+
+    def close(self) -> None:
+        with self._lock:
+            self._conn.close()

ingest/radieo/fetchers/__init__.py

@@ -0,0 +1 @@
+"""Fetchers turn a ``Track`` into a local file in the cache directory."""

ingest/radieo/fetchers/subsonic.py

@@ -0,0 +1,46 @@
+"""SubsonicFetcher: download a ``subsonic`` track into the cache directory.
+
+Files are downloaded to a hidden, non-audio temporary name and then atomically
+renamed into place, so the stream container never sees a partial file through
+its shared read-only cache mount.
+"""
+
+import logging
+import re
+from pathlib import Path
+from uuid import uuid4
+
+from ..models import Track
+from ..subsonic import SubsonicClient
+
+log = logging.getLogger("radieo.fetcher.subsonic")
+
+_SAFE = re.compile(r"[^A-Za-z0-9._-]")
+
+
+class SubsonicFetcher:
+    backend = "subsonic"
+
+    def __init__(self, client: SubsonicClient, cache_dir: Path):
+        self._client = client
+        self._cache_dir = cache_dir
+
+    def fetch(self, track: Track) -> Path:
+        stem = f"subsonic-{_SAFE.sub('_', track.locator)}"
+        # Reuse a still-cached copy rather than downloading again.
+        for existing in self._cache_dir.glob(f"{stem}.*"):
+            if not existing.name.endswith(".part"):
+                return existing
+
+        tmp = self._cache_dir / f".{uuid4().hex}.part"
+        try:
+            ext = self._client.download(
+                track.locator, tmp, hint_ext=track.source_ext
+            )
+            dest = self._cache_dir / f"{stem}{ext}"
+            tmp.replace(dest)
+        except BaseException:
+            tmp.unlink(missing_ok=True)
+            raise
+        log.info("downloaded %s -> %s", track, dest.name)
+        return dest

ingest/radieo/models.py

@@ -0,0 +1,34 @@
+"""Shared data model.
+
+A ``Track`` is the uniform object every provider emits: a *resolved* reference
+(which backend can download it, and where) plus display metadata. Fetchers turn
+it into a local file; the queue and the state database use ``key`` for
+de-duplication and anti-repeat.
+"""
+
+from dataclasses import dataclass
+
+
+@dataclass(frozen=True)
+class Track:
+    backend: str  # which fetcher handles it: "subsonic" | "ytdlp"
+    locator: str  # backend-specific: Subsonic song id, or a media URL
+    artist: str
+    title: str
+    origin: str  # provider that produced it, e.g. "navidrome"
+    mbid: str | None = None  # filled by the Canonicalizer (milestone 5)
+    source_ext: str | None = None  # filename hint, e.g. "mp3", "flac"
+
+    @property
+    def key(self) -> str:
+        """Stable identity for de-duplication and anti-repeat.
+
+        Until the Canonicalizer (milestone 5) fills ``mbid``, we key on the
+        backend locator, which is unique within a source.
+        """
+        if self.mbid:
+            return f"mbid:{self.mbid}"
+        return f"{self.backend}:{self.locator}"
+
+    def __str__(self) -> str:
+        return f"{self.artist} — {self.title} [{self.origin}]"

ingest/radieo/providers/__init__.py

@@ -0,0 +1 @@
+"""Providers decide *what* to play, emitting resolved ``Track`` objects."""

ingest/radieo/providers/navidrome.py

@@ -0,0 +1,68 @@
+"""NavidromeProvider: picks tracks from an OpenSubsonic playlist.
+
+Emits ``subsonic`` tracks (locator = song id). The playlist is cached in
+memory and refreshed periodically. Anti-repeat is applied by filtering out
+tracks whose key is among the recently played ones; if that empties the pool
+(short playlist), the filter is dropped so playback never stalls.
+"""
+
+import logging
+import random
+import time
+
+import httpx
+
+from .. import config
+from ..db import Database
+from ..models import Track
+from ..subsonic import SubsonicClient, SubsonicError
+
+log = logging.getLogger("radieo.provider.navidrome")
+
+
+class NavidromeProvider:
+    name = "navidrome"
+
+    def __init__(self, client: SubsonicClient, playlist_ref: str, db: Database):
+        self._client = client
+        self._playlist_ref = playlist_ref
+        self._db = db
+        self._playlist_id: str | None = None
+        self._songs: list[dict] = []
+        self._loaded_at = 0.0
+
+    def _ensure_songs(self) -> None:
+        now = time.time()
+        if self._songs and now - self._loaded_at < config.PLAYLIST_REFRESH:
+            return
+        if self._playlist_id is None:
+            self._playlist_id = self._client.resolve_playlist_id(
+                self._playlist_ref
+            )
+        songs = self._client.get_playlist_songs(self._playlist_id)
+        self._songs = songs
+        self._loaded_at = now
+        log.info("loaded %d songs from playlist %r", len(songs), self._playlist_ref)
+
+    def next(self) -> Track | None:
+        try:
+            self._ensure_songs()
+        except (SubsonicError, httpx.HTTPError, OSError) as exc:
+            log.warning("could not load playlist: %s", exc)
+            return None
+        if not self._songs:
+            return None
+
+        recent = self._db.recent_keys(config.ANTIREPEAT_WINDOW)
+        candidates = [
+            s for s in self._songs if f"subsonic:{s['id']}" not in recent
+        ] or self._songs
+        song = random.choice(candidates)
+        return Track(
+            backend="subsonic",
+            locator=str(song["id"]),
+            artist=song.get("artist", "Unknown artist"),
+            title=song.get("title", str(song["id"])),
+            origin=self.name,
+            source_ext=song.get("suffix"),
+        )

ingest/radieo/queue.py

@@ -1,50 +1,89 @@
-"""Queue of tracks ready to be served to the stream layer.
+"""Prefetching pipeline feeding ready-to-play tracks to the stream layer.
 
-Milestone 2: a track is simply an audio file already present in the cache
+A background thread keeps a small buffer of downloaded tracks: it asks the
-directory. When the queue drains, it is refilled with a fresh shuffle of the
+provider what to play next, has the matching fetcher download it into the
-available files. Later milestones will replace the refill logic with providers
+cache, and enqueues the resulting file. ``pop_next`` hands the oldest ready
-that download tracks (Navidrome, yt-dlp, …) before enqueuing them.
+track to the HTTP API, records the play and runs LRU retention.
+
+If the provider has nothing (e.g. Navidrome not configured, or unreachable),
+the buffer simply stays empty and ``pop_next`` returns ``None`` — the stream
+then plays its own local-cache fallback.
 """
 
-import random
+import logging
 import threading
 from collections import deque
 from pathlib import Path
 
 from . import config
+from .db import Database
+from .models import Track
+
+log = logging.getLogger("radieo.queue")
 
 
 class TrackQueue:
-    def __init__(self):
+    def __init__(self, provider, fetcher, db: Database):
+        self._provider = provider
+        self._fetcher = fetcher
+        self._db = db
         self._lock = threading.Lock()
-        self._upcoming: deque[Path] = deque()
+        self._ready: deque[tuple[Path, Track]] = deque()
-        self._last_served: Path | None = None
+        self._stop = threading.Event()
-
+        self._thread = threading.Thread(
-    def _available_files(self) -> list[Path]:
+            target=self._run, name="prefetch", daemon=True
-        if not config.CACHE_DIR.is_dir():
-            return []
-        return sorted(
-            p
-            for p in config.CACHE_DIR.iterdir()
-            if p.is_file() and p.suffix.lower() in config.AUDIO_EXTENSIONS
         )
 
-    def _refill_locked(self) -> None:
+    def start(self) -> None:
-        files = self._available_files()
+        self._thread.start()
-        if not files:
-            return
-        # Avoid replaying the last served track back-to-back when we can.
-        pool = [f for f in files if f != self._last_served] or files
-        random.shuffle(pool)
-        self._upcoming.extend(pool)
 
-    def pop_next(self) -> Path | None:
+    def stop(self) -> None:
-        """Return the next track to play, or None if the cache is empty."""
+        self._stop.set()
+
+    # --- background prefetching ------------------------------------------
+
+    def _run(self) -> None:
+        while not self._stop.is_set():
+            try:
+                self._prefetch()
+            except Exception:  # never let the loop die
+                log.exception("prefetch loop error")
+            self._stop.wait(config.PREFETCH_INTERVAL)
+
+    def _prefetch(self) -> None:
         with self._lock:
-            if not self._upcoming:
+            missing = config.PREFETCH - len(self._ready)
-                self._refill_locked()
+        for _ in range(max(0, missing)):
-            if not self._upcoming:
+            if self._stop.is_set():
+                return
+            track = self._provider.next()
+            if track is None:
+                return  # nothing to fetch right now
+            try:
+                path = self._fetcher.fetch(track)
+            except Exception:
+                log.exception("fetch failed for %s", track)
+                continue
+            self._db.register_download(str(path), track.key)
+            with self._lock:
+                self._ready.append((path, track))
+
+    # --- serving ----------------------------------------------------------
+
+    def pop_next(self) -> tuple[Path, Track] | None:
+        with self._lock:
+            if not self._ready:
                 return None
-            track = self._upcoming.popleft()
+            path, track = self._ready.popleft()
-            self._last_served = track
+        self._db.mark_played(str(path))
-            return track
+        self._db.record_play(track)
+        self._evict()
+        return path, track
+
+    def _evict(self) -> None:
+        for path in self._db.evict(config.RETENTION_KEEP):
+            try:
+                Path(path).unlink(missing_ok=True)
+                log.info("evicted %s", Path(path).name)
+            except OSError:
+                log.exception("could not evict %s", path)

ingest/radieo/subsonic.py

@@ -0,0 +1,115 @@
+"""Minimal OpenSubsonic client (enough for Navidrome playback).
+
+Uses salted-token authentication (``t = md5(password + salt)``), the scheme
+recommended by the Subsonic API since 1.13.0 and supported by Navidrome.
+"""
+
+import hashlib
+import logging
+import secrets
+from pathlib import Path
+
+import httpx
+
+log = logging.getLogger("radieo.subsonic")
+
+# Advertised API version and client name.
+_API_VERSION = "1.16.1"
+_CLIENT = "radieo"
+
+# Content-Type -> file extension, used to name downloaded files.
+_CTYPE_EXT = {
+    "audio/mpeg": ".mp3",
+    "audio/mp3": ".mp3",
+    "audio/flac": ".flac",
+    "audio/x-flac": ".flac",
+    "audio/ogg": ".ogg",
+    "application/ogg": ".ogg",
+    "audio/opus": ".opus",
+    "audio/mp4": ".m4a",
+    "audio/x-m4a": ".m4a",
+    "audio/aac": ".aac",
+    "audio/wav": ".wav",
+    "audio/x-wav": ".wav",
+}
+
+
+class SubsonicError(Exception):
+    pass
+
+
+class SubsonicClient:
+    def __init__(self, base_url: str, user: str, password: str):
+        self._base = base_url.rstrip("/")
+        self._user = user
+        self._password = password
+        self._http = httpx.Client(timeout=30.0, follow_redirects=True)
+
+    def _auth_params(self) -> dict[str, str]:
+        salt = secrets.token_hex(8)
+        token = hashlib.md5((self._password + salt).encode()).hexdigest()
+        return {
+            "u": self._user,
+            "t": token,
+            "s": salt,
+            "v": _API_VERSION,
+            "c": _CLIENT,
+            "f": "json",
+        }
+
+    def _get_json(self, view: str, **params) -> dict:
+        url = f"{self._base}/rest/{view}"
+        resp = self._http.get(url, params={**self._auth_params(), **params})
+        resp.raise_for_status()
+        body = resp.json()["subsonic-response"]
+        if body.get("status") != "ok":
+            err = body.get("error", {})
+            raise SubsonicError(
+                f"{view}: {err.get('code')} {err.get('message')}"
+            )
+        return body
+
+    def ping(self) -> None:
+        self._get_json("ping")
+
+    def resolve_playlist_id(self, name_or_id: str) -> str:
+        """Accept either a playlist id or a playlist name."""
+        body = self._get_json("getPlaylists")
+        playlists = body.get("playlists", {}).get("playlist", [])
+        for pl in playlists:
+            if pl.get("id") == name_or_id or pl.get("name") == name_or_id:
+                return pl["id"]
+        raise SubsonicError(f"playlist not found: {name_or_id!r}")
+
+    def get_playlist_songs(self, playlist_id: str) -> list[dict]:
+        body = self._get_json("getPlaylist", id=playlist_id)
+        return body.get("playlist", {}).get("entry", [])
+
+    def download(self, song_id: str, dest: Path, hint_ext: str | None = None) -> str:
+        """Download a song to ``dest``; return the file extension used.
+
+        ``format=raw`` asks Navidrome for the original file (no transcoding),
+        keeping quality and letting Liquidsoap decode it.
+        """
+        params = {**self._auth_params(), "id": song_id, "format": "raw"}
+        with self._http.stream(
+            "GET", f"{self._base}/rest/stream", params=params
+        ) as resp:
+            resp.raise_for_status()
+            ctype = resp.headers.get("content-type", "").split(";")[0].strip()
+            if ctype.startswith(("application/json", "text/xml")):
+                raise SubsonicError(
+                    f"stream {song_id}: error response {resp.read()[:200]!r}"
+                )
+            ext = _CTYPE_EXT.get(ctype)
+            if ext is None and hint_ext:
+                ext = "." + hint_ext.lstrip(".")
+            if ext is None:
+                ext = ".mp3"
+            with open(dest, "wb") as fh:
+                for chunk in resp.iter_bytes(chunk_size=65536):
+                    fh.write(chunk)
+        return ext
+
+    def close(self) -> None:
+        self._http.close()

ingest/requirements.txt

@@ -0,0 +1 @@
+httpx>=0.27