replace file discovery by udp discovery

src/server.py

@@ -35,7 +35,6 @@ import threading
 
 import stremio
 import transcoder
-from storage import init_service_discovery, get_service_discovery
 import fileserver
 import webdav_client
 
@@ -58,19 +57,24 @@ API_PREFIX = f'/s/{API_KEY}' if API_KEY else ''
 # upstream by hash-lock/OIDC instead of the path-token.
 DASHBOARD_PATHS = frozenset({
     '/', '/index.html', '/configure', '/health',
-    '/api/stats', '/api/library', '/api/services', '/api/languages',
+    '/api/stats', '/api/library', '/api/services', '/api/neighbors', '/api/languages',
+    # The dashboard's shared nav element, loaded as a module from index.html.
+    '/meta-service-menu.js',
     '/transcode/metrics',
 })
 
 # Initialize storage
 storage = stremio.init_storage()
 
-# Initialize service discovery (only if META_CORE_PATH exists)
+# Service discovery is UDP now (meta-discovery v1), so there is no
+# META_CORE_PATH gate: the leader client's node announces this service on the
+# same multicast group it listens on. Being discoverable and discovering are
+# one thing. See docs/project-architecture/service-discovery.md.
 service_discovery = None
-if os.path.exists(META_CORE_PATH):
+try:
-    try:
+    from storage.leader_client import get_leader_client
-        service_discovery = init_service_discovery(base_url=BASE_URL if BASE_URL else None)
+    get_leader_client().start_watching()
-    except Exception as e:
+except Exception as e:
     print(f"[Server] Service discovery init failed: {e}")
 
 # Global Stremio handler
@@ -252,6 +256,13 @@ class Handler(BaseHTTPRequestHandler):
         if path == '/' or path == '/index.html':
             return self.serve_setup_page()
 
+        # The one static asset under www/ that the dashboard loads as a module:
+        # the shared <meta-service-menu> element. There is no generic static
+        # route here (index.html is string-injected on every request), so this
+        # is an explicit allowlist of one rather than an open file server.
+        if path == '/meta-service-menu.js':
+            return self.serve_www_asset('meta-service-menu.js', 'text/javascript')
+
         # Configure page
         if path == '/configure':
             return self.serve_configure_page()
@@ -287,7 +298,7 @@ class Handler(BaseHTTPRequestHandler):
             })
 
         # Discovered services (for inter-service navigation)
-        if path == '/api/services':
+        if path == '/api/services' or path == '/api/neighbors':
             return self.handle_services_api()
 
         # === File API ===
@@ -368,6 +379,22 @@ class Handler(BaseHTTPRequestHandler):
 
         self.send_error(404)
 
+    def serve_www_asset(self, name, content_type):
+        """Serve one named file from www/ verbatim (no template injection)."""
+        for base in (os.path.dirname(__file__), os.path.dirname(os.path.dirname(__file__))):
+            candidate = os.path.join(base, 'www', name)
+            if os.path.exists(candidate):
+                with open(candidate, 'rb') as f:
+                    body = f.read()
+                self.send_response(200)
+                self.send_header('Content-Type', content_type)
+                self.send_header('Content-Length', len(body))
+                self.end_headers()
+                self.wfile.write(body)
+                return
+        self.send_response(404)
+        self.end_headers()
+
     def serve_setup_page(self):
         """Serve the setup/dashboard page."""
         html_path = os.path.join(os.path.dirname(__file__), 'www', 'index.html')
@@ -627,35 +654,35 @@ class Handler(BaseHTTPRequestHandler):
             self.wfile.write(content.encode('utf-8'))
 
     def handle_services_api(self):
-        """Return list of discovered services for dashboard navigation.
+        """Neighbours for the dashboard nav, from this peer's own UDP map.
 
-        Fetches services from meta-core (single source of truth) instead of
+        No hop through meta-core any more: every service keeps its own
-        doing local filesystem-based discovery.
+        neighbour table, so the nav still renders when meta-core is down.
+        `services` is kept as an alias of `neighbors` so the existing
+        dashboards keep working during the UI migration.
         """
         from storage.leader_client import get_leader_client
-        from urllib.request import urlopen
-        from urllib.error import URLError
 
         try:
-            leader_client = get_leader_client()
+            client = get_leader_client()
-            leader_info = leader_client.get_leader_info() if leader_client else None
+            neighbors = client.neighbors() if client else []
-
+            return self.send_json({
-            if leader_info and leader_info.api_url:
+                'current': 'meta-stremio',
-                # Fetch services from meta-core API
+                'enabled': True,
-                url = f"{leader_info.api_url}/services?current=meta-stremio"
+                'count': len(neighbors),
-                with urlopen(url, timeout=5) as response:
+                'neighbors': neighbors,
-                    data = json.loads(response.read().decode())
+                'services': neighbors,
-                    return self.send_json(data)
+                'self': client.self_announce() if client else None,
-        except URLError as e:
+            })
-            print(f"[Server] Error fetching services from meta-core: {e}")
         except Exception as e:
             print(f"[Server] Error in handle_services_api: {e}")
 
-        # Fallback: return empty list if meta-core is unavailable
         return self.send_json({
-            'services': [],
             'current': 'meta-stremio',
+            'enabled': False,
             'count': 0,
+            'neighbors': [],
+            'services': [],
         })
 
     def handle_file(self, cid: str, width: int = None):
@@ -906,11 +933,14 @@ def shutdown_handler(signum, frame):
     """Handle graceful shutdown."""
     print("\n[Server] Shutting down...")
 
-    # Stop service discovery
+    # Stop service discovery. close() emits a final "stopping" announce so
+    # neighbours drop us immediately instead of waiting out the staleness
+    # window.
     try:
-        sd = get_service_discovery()
+        from storage.leader_client import get_leader_client
-        if sd:
+        client = get_leader_client()
-            sd.stop()
+        if client:
+            client.close()
     except Exception as e:
         print(f"[Server] Error stopping service discovery: {e}")
 

src/storage/__init__.py

@@ -13,15 +13,6 @@ from .provider import StorageProvider, VideoMetadata
 from .redis_storage import RedisStorage
 from .leader_storage import LeaderStorage
 from .leader_client import LeaderClient, LeaderLockInfo, get_leader_client
-from .service_registration import (
-    ServiceRegistration,
-    ServiceInfo,
-    get_service_discovery,  # Kept for backward compatibility
-    init_service_discovery  # Kept for backward compatibility
-)
-
-# Backward compatibility alias
-ServiceDiscovery = ServiceRegistration
 
 __all__ = [
     'StorageProvider',
@@ -31,9 +22,4 @@ __all__ = [
     'LeaderClient',
     'LeaderLockInfo',
     'get_leader_client',
-    'ServiceRegistration',
-    'ServiceDiscovery',  # Backward compatibility alias
-    'ServiceInfo',
-    'get_service_discovery',
-    'init_service_discovery',
 ]

src/storage/leader_client.py

@@ -1,40 +1,41 @@
+"""LeaderClient — locates meta-core and exposes its URLs.
+
+Since meta-discovery v1 this is a thin adapter over ``MetaCoreLocator``
+(``meshdisco.py``): meta-core is found by UDP announce instead of by reading
+``/meta-core/locks/kv-leader.info``, so meta-stremio no longer needs the
+``/meta-core`` volume mounted at all.
+
+The public surface is deliberately unchanged so ``LeaderStorage`` did not have
+to move: ``get_leader_info``, ``get_urls``, ``get_api_url``,
+``get_webdav_url*``, ``wait_for_leader``, ``start_watching`` / ``on_change``
+(the watchdog on the lock directory became "an announce arrived carrying a
+different apiUrl", feeding the same callback).
+
+Precedence is the pin: when ``meta_core_url`` (``META_CORE_URL``) is set it
+always wins and the wire is never consulted for core selection. See
+docs/project-architecture/service-discovery.md.
+
+The name "leader" is legacy: meta-core's flock election is vestigial, and
+``redis_url`` has been empty since the api-mediated-access lockdown — every
+read goes over HTTP.
 """
-LeaderClient - Client for reading leader info from meta-core
 
-This replaces the old LeaderDiscovery.py by delegating leader election
-to meta-core (Go sidecar) and just reading the results.
-
-Features:
-- Reads kv-leader.info file for leader URLs
-- Calls meta-core /urls API for current leader URLs
-- Watches for leader changes via file system watcher
-"""
-from __future__ import annotations
-
-import os
 import json
-import threading
+import os
+import time
+import urllib.request
 from dataclasses import dataclass
-from pathlib import Path
+from typing import Callable, List, Optional
-from typing import Optional, Callable
-from urllib.request import urlopen, Request
-from urllib.error import URLError
 
-# Try to import watchdog for file watching (optional)
+from .meshdisco import MetaCoreLocator
-try:
-    from watchdog.observers import Observer
-    from watchdog.events import FileSystemEventHandler
-    WATCHDOG_AVAILABLE = True
-except ImportError:
-    WATCHDOG_AVAILABLE = False
-    Observer = None
-    FileSystemEventHandler = object
 
 
 @dataclass
 class LeaderLockInfo:
-    """Leader lock file content format (kv-leader.info).
+    """meta-core's identity and URLs.
-    Written by meta-core leader, read by other services.
+
+    Historically the parsed kv-leader.info + /urls pair; now assembled from the
+    announce (or from /urls when meta-core is pinned).
     """
     hostname: str
     base_url: str
@@ -53,140 +54,133 @@ class URLsResponse:
     base_url: str
     api_url: str
     redis_url: str
-    webdav_url: str           # External WebDAV URL (via nginx/HTTPS)
+    webdav_url: str
-    webdav_url_internal: str  # Internal WebDAV URL (direct to port 9000)
+    webdav_url_internal: str
     is_leader: bool
 
 
-class LockFileHandler(FileSystemEventHandler):
-    """File system event handler for lock file changes."""
-
-    def __init__(self, callback: Callable[[], None], lock_filename: str):
-        self.callback = callback
-        self.lock_filename = lock_filename
-
-    def on_modified(self, event):
-        if event.src_path.endswith(self.lock_filename):
-            self.callback()
-
-    def on_created(self, event):
-        if event.src_path.endswith(self.lock_filename):
-            self.callback()
-
-    def on_deleted(self, event):
-        if event.src_path.endswith(self.lock_filename):
-            self.callback()
-
-
 class LeaderClient:
-    """
+    """Locates meta-core over UDP and resolves its URL set."""
-    LeaderClient for meta-stremio.
-
-    Reads leader info from kv-leader.info file written by meta-core.
-    """
 
     def __init__(
         self,
-        meta_core_path: str,
+        meta_core_path: Optional[str] = None,
-        meta_core_url: Optional[str] = None
+        meta_core_url: Optional[str] = None,
-    ):
+        service_name: str = "meta-stremio",
-        self.meta_core_path = meta_core_path
+        base_url: Optional[str] = None,
-        self.meta_core_url = meta_core_url
+        version: str = "",
+    ) -> None:
+        # meta_core_path is accepted and ignored — kept so existing callers
+        # construct unchanged. Nothing reads the volume any more.
+        self._meta_core_path = meta_core_path
+        self._meta_core_url = (meta_core_url or "").strip() or None
+        self._leader_info: Optional[LeaderLockInfo] = None
+        self._on_change_callbacks: List[Callable[[], None]] = []
+        self._started = False
 
-        self.info_file_path = os.path.join(meta_core_path, 'locks', 'kv-leader.info')
+        # Only used on the pinned path, where the URLs still come from an HTTP
-        self.leader_info: Optional[LeaderLockInfo] = None
+        # GET /urls. A discovered core carries them in the announce itself.
-
-        self._file_observer = None
-        self._on_change_callbacks: list[Callable[[], None]] = []
-        self._stop_event = threading.Event()
-
-        # URL caching
         self._cached_urls: Optional[URLsResponse] = None
-        self._urls_cache_time: float = 0
+        self._urls_cache_time: float = 0.0
-        self._urls_cache_ttl: float = 5.0  # 5 seconds
+        self._urls_cache_ttl: float = 5.0
 
-    def _get_api_url_from_file(self) -> Optional[str]:
+        self._locator = MetaCoreLocator(
-        """Read API URL from file (plain text format)."""
+            service_name=service_name,
-        try:
+            base_url=base_url,
-            with open(self.info_file_path, 'r') as f:
+            version=version,
-                content = f.read()
+            meta_core_url=self._meta_core_url,
-            return content.strip() or None
+        )
-        except FileNotFoundError:
+        self._locator.on_change(self._on_locator_change)
-            print(f"[LeaderClient] Leader info file not found: {self.info_file_path}")
+
-            return None
+    # -- lifecycle ---------------------------------------------------------
-        except Exception as e:
+
-            print(f"[LeaderClient] Error reading API URL from file: {e}")
+    def _ensure_started(self) -> None:
-            return None
+        if self._started:
+            return
+        self._started = True
+        self._locator.start()
+
+    def _on_locator_change(self) -> None:
+        self._cached_urls = None
+        self._urls_cache_time = 0.0
+        self._notify_change()
+
+    # -- url resolution ----------------------------------------------------
+
+    def _assert_no_redis_url(self, redis_url: Optional[str]) -> None:
+        """Guard against a rollback reintroducing direct Redis exposure.
+
+        api-mediated-access PR D removed redis_url; ALLOW_LEGACY_REDIS_URL=1
+        downgrades this to a warning during a deliberate temporary rollback.
+        """
+        if not redis_url:
+            return
+        msg = (
+            "meta-core still publishes redisUrl; direct Redis access was retired "
+            "by the api-mediated-access lockdown. Verify meta-core version."
+        )
+        if os.environ.get("ALLOW_LEGACY_REDIS_URL") == "1":
+            print(f"[LeaderClient] WARNING: {msg}")
+        else:
+            raise RuntimeError(msg)
 
     def _fetch_urls(self, api_url: str) -> Optional[URLsResponse]:
-        """Fetch URLs from meta-core /urls API with caching."""
-        import time
-
-        # Check cache
         now = time.time()
         if self._cached_urls and (now - self._urls_cache_time) < self._urls_cache_ttl:
             return self._cached_urls
-
         try:
-            url = f"{api_url}/urls"
+            req = urllib.request.Request(
-            req = Request(url, headers={'Accept': 'application/json'})
+                f"{api_url.rstrip('/')}/urls", headers={"Accept": "application/json"}
-            with urlopen(req, timeout=5) as response:
-                if response.status != 200:
-                    print(f"[LeaderClient] Failed to fetch URLs: {response.status}")
-                    return None
-
-                data = json.loads(response.read().decode())
-
-                # api-mediated-access PR D: redisUrl is no longer published by
-                # meta-core. If it shows up, an old build snuck back in —
-                # fail loudly so the rollback can't go unnoticed. Set
-                # ALLOW_LEGACY_REDIS_URL=1 to downgrade to a warning during
-                # a deliberate temporary rollback.
-                if data.get('redisUrl'):
-                    msg = (
-                        "[LeaderClient] meta-core still publishes redisUrl; "
-                        "direct Redis access was retired by the api-mediated-access "
-                        "lockdown. Verify meta-core version."
             )
-                    if os.environ.get("ALLOW_LEGACY_REDIS_URL") == "1":
+            with urllib.request.urlopen(req, timeout=5) as resp:
-                        print("WARNING: " + msg)
+                data = json.loads(resp.read().decode("utf-8"))
-                    else:
+            self._assert_no_redis_url(data.get("redisUrl"))
-                        raise RuntimeError(msg)
             self._cached_urls = URLsResponse(
-                    hostname=data.get('hostname', ''),
+                hostname=data.get("hostname", ""),
-                    base_url=data.get('baseUrl', ''),
+                base_url=data.get("baseUrl", ""),
-                    api_url=data.get('apiUrl', ''),
+                api_url=data.get("apiUrl", ""),
-                    redis_url=data.get('redisUrl', ''),
+                redis_url=data.get("redisUrl", "") or "",
-                    webdav_url=data.get('webdavUrl', ''),
+                webdav_url=data.get("webdavUrl", ""),
-                    webdav_url_internal=data.get('webdavUrlInternal', ''),
+                webdav_url_internal=data.get("webdavUrlInternal", ""),
-                    is_leader=data.get('isLeader', False)
+                is_leader=bool(data.get("isLeader", True)),
             )
             self._urls_cache_time = now
             return self._cached_urls
-
+        except Exception as e:
-        except URLError as e:
             print(f"[LeaderClient] Error calling /urls API: {e}")
             return None
-        except Exception as e:
-            print(f"[LeaderClient] Error fetching URLs: {e}")
-            return None
 
-    def get_leader_info(self) -> Optional[LeaderLockInfo]:
+    def get_urls(self) -> Optional[URLsResponse]:
-        """Read leader info from file and /urls API."""
+        """Resolve meta-core's URLs.
-        try:
+
-            # Read API URL from file (plain text)
+        Straight from the announce when discovered (no HTTP hop needed); over
-            api_url = self._get_api_url_from_file()
+        HTTP when meta-core is pinned.
+        """
+        self._ensure_started()
+
+        wire = self._locator.get_urls()
+        if wire and not self._meta_core_url:
+            return URLsResponse(
+                hostname=wire.get("hostname", ""),
+                base_url=wire.get("baseUrl", ""),
+                api_url=wire.get("apiUrl", ""),
+                redis_url="",
+                webdav_url=wire.get("webdavUrl", ""),
+                webdav_url_internal=wire.get("webdavUrlInternal", ""),
+                is_leader=True,
+            )
+
+        api_url = self._locator.get_api_url()
         if not api_url:
             return None
+        return self._fetch_urls(api_url)
 
-            # Fetch full info from /urls API
+    def get_leader_info(self) -> Optional[LeaderLockInfo]:
-            urls = self._fetch_urls(api_url)
+        try:
+            urls = self.get_urls()
             if not urls:
                 return None
-
+            self._leader_info = LeaderLockInfo(
-            # Convert URLsResponse to LeaderLockInfo
-            import time
-            self.leader_info = LeaderLockInfo(
                 hostname=urls.hostname,
                 base_url=urls.base_url,
                 api_url=urls.api_url,
@@ -194,145 +188,110 @@ class LeaderClient:
                 webdav_url=urls.webdav_url,
                 webdav_url_internal=urls.webdav_url_internal,
                 timestamp=int(time.time() * 1000),
-                pid=0  # Unknown for remote leader
+                pid=0,
             )
-
+            return self._leader_info
-            return self.leader_info
-
         except Exception as e:
-            print(f"[LeaderClient] Error reading leader info: {e}")
+            print(f"[LeaderClient] Failed to read leader info: {e}")
             return None
 
     def get_redis_url(self) -> Optional[str]:
-        """Get Redis URL from leader info."""
         info = self.get_leader_info()
         return info.redis_url if info else None
 
     def get_webdav_url(self) -> Optional[str]:
-        """Get WebDAV URL from leader info (external, via nginx/HTTPS)."""
         info = self.get_leader_info()
         return info.webdav_url if info else None
 
     def get_webdav_url_internal(self) -> Optional[str]:
-        """Get internal WebDAV URL from leader info (direct to port 9000).
+        """Internal WebDAV URL — use for container-to-container access."""
-        Use this for container-to-container communication.
-        """
         info = self.get_leader_info()
         return info.webdav_url_internal if info else None
 
     def get_api_url(self) -> Optional[str]:
-        """Get meta-core API URL from leader info."""
         info = self.get_leader_info()
         return info.api_url if info else None
 
-    def get_urls(self) -> Optional[URLsResponse]:
-        """Call meta-core /urls API to get current URLs."""
-        # First try using configured meta_core_url
-        api_url = self.meta_core_url
-
-        # Fall back to reading from file
-        if not api_url:
-            api_url = self._get_api_url_from_file()
-
-        if not api_url:
-            print("[LeaderClient] No meta-core API URL available")
-            return None
-
-        return self._fetch_urls(api_url)
-
     def wait_for_leader(self, timeout_ms: int = 30000) -> LeaderLockInfo:
-        """Wait for leader info to be available."""
+        """Block until meta-core is reachable.
-        import time
-        start_time = time.time()
-        poll_interval = 0.5  # 500ms
 
-        while (time.time() - start_time) * 1000 < timeout_ms:
+        Probes immediately then every 500ms, matching the previous
+        file-polling loop's timing.
+        """
+        self._ensure_started()
+        timeout_s = None if timeout_ms <= 0 else timeout_ms / 1000.0
+        deadline = None if timeout_s is None else time.time() + timeout_s
+
+        if self._locator.wait_for_core(timeout_s) is None:
+            raise TimeoutError(f"No meta-core found within {timeout_ms}ms")
+
+        while deadline is None or time.time() < deadline:
             info = self.get_leader_info()
             if info:
-                print(f"[LeaderClient] Leader found: {info.hostname} at {info.redis_url}")
+                print(f"[LeaderClient] meta-core found: {info.hostname} at {info.api_url}")
                 return info
+            time.sleep(0.5)
+        raise TimeoutError(f"No meta-core found within {timeout_ms}ms")
 
-            print("[LeaderClient] Waiting for leader...")
+    # -- change notification ----------------------------------------------
-            time.sleep(poll_interval)
-
-        raise TimeoutError(f"No leader found within {timeout_ms}ms")
 
     def start_watching(self) -> None:
-        """Start watching for leader changes."""
+        """Previously a watchdog on the lock directory.
-        if not WATCHDOG_AVAILABLE:
-            print("[LeaderClient] watchdog not available, file watching disabled")
-            return
 
-        try:
+        Discovery is always listening, so this only has to make sure the node
-            lock_dir = os.path.dirname(self.info_file_path)
+        is running.
-
+        """
-            handler = LockFileHandler(
+        self._ensure_started()
-                callback=self._on_file_change,
-                lock_filename='kv-leader.info'
-            )
-
-            self._file_observer = Observer()
-            self._file_observer.schedule(handler, lock_dir, recursive=False)
-            self._file_observer.start()
-            print(f"[LeaderClient] Watching for leader changes in {lock_dir}")
-
-        except Exception as e:
-            print(f"[LeaderClient] Could not start file watcher: {e}")
 
     def stop_watching(self) -> None:
-        """Stop watching for leader changes."""
+        """No-op: the locator keeps listening until close()."""
-        if self._file_observer:
-            self._file_observer.stop()
-            self._file_observer.join(timeout=2.0)
-            self._file_observer = None
-            print("[LeaderClient] Stopped watching for leader changes")
 
-    def _on_file_change(self) -> None:
+    def on_change(self, callback: Callable[[], None]) -> "LeaderClient":
-        """Handle file change event."""
-        print("[LeaderClient] Leader info changed, invalidating cache...")
-        # Invalidate cache to force fresh API call
-        self._cached_urls = None
-        self._urls_cache_time = 0
-
-        self.get_leader_info()
-        self._notify_change()
-
-    def on_change(self, callback: Callable[[], None]) -> 'LeaderClient':
-        """Register callback for leader changes."""
         self._on_change_callbacks.append(callback)
         return self
 
     def _notify_change(self) -> None:
-        """Notify all change callbacks."""
+        print("[LeaderClient] meta-core changed, invalidating cache...")
-        for callback in self._on_change_callbacks:
+        for cb in list(self._on_change_callbacks):
             try:
-                callback()
+                cb()
             except Exception as e:
                 print(f"[LeaderClient] Error in change callback: {e}")
 
     def get_cached_leader_info(self) -> Optional[LeaderLockInfo]:
-        """Get cached leader info (without re-reading file)."""
+        return self._leader_info
-        return self.leader_info
+
+    # -- nav menu ----------------------------------------------------------
+
+    def neighbors(self) -> List[dict]:
+        """Neighbours for the nav menu (one row per service name)."""
+        self._ensure_started()
+        return self._locator.neighbors()
+
+    def self_announce(self) -> dict:
+        """This service's own announce, so the menu can show itself."""
+        self._ensure_started()
+        return self._locator.self_announce()
 
     def close(self) -> None:
-        """Clean up resources."""
-        self._stop_event.set()
-        self.stop_watching()
         self._on_change_callbacks = []
+        self._locator.stop()
+        self._started = False
 
 
-# Singleton instance
 _leader_client: Optional[LeaderClient] = None
 
 
 def get_leader_client() -> Optional[LeaderClient]:
-    """Get or create the LeaderClient singleton.
+    """Process-wide singleton, so one UDP socket serves every caller."""
-
-    Returns None if META_CORE_PATH is not set.
-    """
     global _leader_client
     if _leader_client is None:
-        meta_core_path = os.environ.get('META_CORE_PATH', '/meta-core')
+        _leader_client = LeaderClient(
-        meta_core_url = os.environ.get('META_CORE_URL')
+            meta_core_path=os.environ.get("META_CORE_PATH"),
-        _leader_client = LeaderClient(meta_core_path, meta_core_url)
+            meta_core_url=os.environ.get("META_CORE_URL"),
+            # PUBLIC_URL (a browser-reachable debug-direct port) wins over
+            # BASE_URL (the Caddy URL), same ladder as meta-core. Announcing a
+            # Caddy URL on a stack with no Caddy puts a dead link in every menu.
+            base_url=os.environ.get("PUBLIC_URL") or os.environ.get("BASE_URL") or None,
+        )
     return _leader_client

src/storage/leader_storage.py

@@ -18,7 +18,7 @@ import time
 from typing import List, Optional, Callable
 
 from .provider import StorageProvider, VideoMetadata
-from .leader_client import LeaderClient, LeaderLockInfo
+from .leader_client import LeaderClient, LeaderLockInfo, get_leader_client
 from .meta_consumer import MetaConsumer
 from .meta_core_api_client import MetaCoreApiClient
 
@@ -97,10 +97,11 @@ class LeaderStorage(StorageProvider):
             self._connect_to_redis(self._override_url)
             return
 
-        # Use leader client
+        # Use the process-wide leader client so meta-stremio holds ONE UDP
-        self._leader_client = LeaderClient(
+        # discovery socket, shared with server.py's neighbours endpoint.
-            meta_core_path=self._meta_core_path
+        # META_CORE_URL, when set, pins meta-core and discovery never
-        )
+        # overrides it — see docs/project-architecture/service-discovery.md.
+        self._leader_client = get_leader_client()
 
         # Set up change callback
         self._leader_client.on_change(self._on_leader_change)

src/storage/meshdisco.py

@@ -0,0 +1,437 @@
+"""meta-discovery v1 — UDP multicast service discovery.
+
+Replaces the two file-based mechanisms that required a shared /meta-core
+volume: ``locks/kv-leader.info`` (locating meta-core at boot) and
+``services/<name>-<host>.json`` (the dashboard nav registry). Both answered the
+same question — where is meta-core, and who else is on this network — so one
+announce packet answers both, and discovery scope becomes the docker network
+rather than the mounted volume.
+
+The normative spec is docs/project-architecture/service-discovery.md.
+
+Two details this module must get right, both of which fail silently:
+
+1. **Per-interface fan-out.** A socket that joins with ``INADDR_ANY`` and sends
+   via the default route covers exactly one interface. Containers here are
+   routinely attached to two docker networks, so we enumerate interfaces and
+   join/send on each using a hand-packed ``ip_mreqn``.
+2. **The pin wins.** When ``META_CORE_URL`` is set, the wire is never consulted
+   for core selection — otherwise a container that can see two cores could
+   latch onto the wrong one.
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import socket
+import struct
+import threading
+import time
+from typing import Callable, Dict, List, Optional
+
+PROTOCOL_VERSION = 1
+DEFAULT_GROUP = "239.255.77.1"
+DEFAULT_PORT = 9399
+DEFAULT_INTERVAL = 10.0
+# A neighbour unseen for interval * this is dropped. Replaces the reaper.
+LIVENESS_FACTOR = 3
+READ_BUFFER = 8192
+
+TYPE_DISCOVERY = "discovery"
+TYPE_ANNOUNCE = "announce"
+ROLE_CORE = "core"
+ROLE_SERVICE = "service"
+
+
+def _env_flag(name: str, default: bool = True) -> bool:
+    raw = os.environ.get(name)
+    if raw is None or raw == "":
+        return default
+    return raw.strip().lower() not in ("false", "0", "no", "off")
+
+
+def _interface_indexes() -> List[tuple]:
+    """(name, index, ipv4) for every up, non-loopback interface.
+
+    Falls back to a single unspecified entry when the platform cannot
+    enumerate, so discovery still works on the default route.
+    """
+    out: List[tuple] = []
+    try:
+        for idx, name in socket.if_nameindex():
+            if name == "lo":
+                continue
+            try:
+                # Cheap way to get the interface's IPv4 without netifaces.
+                addr = _ipv4_of(name)
+            except OSError:
+                continue
+            if addr:
+                out.append((name, idx, addr))
+    except (AttributeError, OSError):
+        pass
+    return out
+
+
+def _ipv4_of(ifname: str) -> Optional[str]:
+    """SIOCGIFADDR — avoids a netifaces dependency."""
+    import fcntl
+
+    s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+    try:
+        packed = struct.pack("256s", ifname[:15].encode("utf-8"))
+        return socket.inet_ntoa(fcntl.ioctl(s.fileno(), 0x8915, packed)[20:24])
+    except OSError:
+        return None
+    finally:
+        s.close()
+
+
+def local_ipv4() -> str:
+    """First non-loopback IPv4, else the hostname. Matches the Go helper."""
+    ifaces = _interface_indexes()
+    if ifaces:
+        return ifaces[0][2]
+    return socket.gethostname()
+
+
+class MeshNode:
+    """Announces this service, answers probes, tracks neighbours."""
+
+    def __init__(
+        self,
+        name: str,
+        role: str = ROLE_SERVICE,
+        instance: Optional[str] = None,
+        version: str = "",
+        group: str = DEFAULT_GROUP,
+        port: int = DEFAULT_PORT,
+        interval: float = DEFAULT_INTERVAL,
+        enabled: bool = True,
+        payload: Optional[Callable[[], dict]] = None,
+    ) -> None:
+        self.name = name
+        self.role = role
+        self.instance = instance or socket.gethostname()
+        self.version = version
+        self.group = group
+        self.port = port
+        self.interval = interval
+        self.enabled = enabled
+        self._payload = payload
+
+        self._sock: Optional[socket.socket] = None
+        self._ifaces: List[tuple] = []
+        self._neighbors: Dict[str, dict] = {}
+        self._lock = threading.Lock()
+        self._send_lock = threading.Lock()
+        self._running = False
+        self._threads: List[threading.Thread] = []
+        self._listeners: List[Callable[[dict], None]] = []
+
+    # -- lifecycle ---------------------------------------------------------
+
+    def start(self) -> None:
+        if self._running or not self.enabled:
+            if not self.enabled:
+                print("[meshdisco] Disabled by configuration")
+            return
+
+        sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+        sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
+        try:
+            sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEPORT, 1)
+        except (AttributeError, OSError):
+            pass
+        sock.bind(("0.0.0.0", self.port))
+        # TTL 1: link-local only, never routed.
+        sock.setsockopt(socket.IPPROTO_IP, socket.IP_MULTICAST_TTL, 1)
+
+        self._ifaces = _interface_indexes()
+        group_bin = socket.inet_aton(self.group)
+        joined = 0
+        for ifname, idx, addr in self._ifaces:
+            try:
+                # ip_mreqn: multiaddr, address (INADDR_ANY), ifindex. The
+                # ifindex is what makes this a per-interface join instead of
+                # "whatever the routing table picks".
+                mreq = group_bin + socket.inet_aton("0.0.0.0") + struct.pack("@i", idx)
+                sock.setsockopt(socket.IPPROTO_IP, socket.IP_ADD_MEMBERSHIP, mreq)
+                joined += 1
+            except OSError as e:
+                print(f"[meshdisco] Warning: join {self.group} on {ifname} failed: {e}")
+
+        self._sock = sock
+        self._running = True
+        print(
+            f"[meshdisco] Listening on {self.group}:{self.port} as "
+            f"{self.name}/{self.instance} (joined {joined} of {len(self._ifaces)} interfaces)"
+        )
+
+        for target in (self._read_loop, self._announce_loop):
+            t = threading.Thread(target=target, daemon=True)
+            t.start()
+            self._threads.append(t)
+
+    def stop(self) -> None:
+        if not self._running:
+            return
+        self._running = False
+        # Final announce so neighbours drop us immediately rather than waiting
+        # out the staleness window.
+        try:
+            self._send(self._build_announce("stopping"))
+        except OSError:
+            pass
+        if self._sock:
+            try:
+                self._sock.close()
+            except OSError:
+                pass
+        self._sock = None
+
+    # -- api ---------------------------------------------------------------
+
+    def probe(self) -> None:
+        """Multicast a probe; every listener replies immediately."""
+        self._send({"v": PROTOCOL_VERSION, "type": TYPE_DISCOVERY})
+
+    def neighbors(self) -> List[dict]:
+        cutoff = time.time() - self.interval * LIVENESS_FACTOR
+        with self._lock:
+            for key in [k for k, v in self._neighbors.items() if v["lastSeen"] < cutoff]:
+                del self._neighbors[key]
+            out = list(self._neighbors.values())
+        return sorted(out, key=lambda n: n.get("name") or "")
+
+    def neighbors_by_name(self) -> List[dict]:
+        best: Dict[str, dict] = {}
+        for nb in self.neighbors():
+            cur = best.get(nb.get("name") or "")
+            if cur is None or nb["lastSeen"] > cur["lastSeen"]:
+                best[nb.get("name") or ""] = nb
+        return sorted(best.values(), key=lambda n: n.get("name") or "")
+
+    def cores(self) -> List[dict]:
+        return [n for n in self.neighbors() if n.get("role") == ROLE_CORE and n.get("urls")]
+
+    def self_announce(self) -> dict:
+        msg = self._build_announce("running")
+        msg["addr"] = ""
+        msg["lastSeen"] = time.time()
+        return msg
+
+    def on_announce(self, cb: Callable[[dict], None]) -> None:
+        self._listeners.append(cb)
+
+    # -- internals ---------------------------------------------------------
+
+    def _build_announce(self, status: str) -> dict:
+        msg = {
+            "v": PROTOCOL_VERSION,
+            "type": TYPE_ANNOUNCE,
+            "name": self.name,
+            "instance": self.instance,
+            "role": self.role,
+            "version": self.version,
+            "status": status,
+        }
+        if self._payload:
+            try:
+                p = self._payload() or {}
+            except Exception:
+                p = {}
+            if p.get("baseUrl"):
+                msg["baseUrl"] = p["baseUrl"]
+            if p.get("status") and status == "running":
+                msg["status"] = p["status"]
+            # Only a core may carry a URLs block.
+            if self.role == ROLE_CORE and p.get("urls"):
+                msg["urls"] = p["urls"]
+        return msg
+
+    def _send(self, msg: dict, dest: Optional[tuple] = None) -> None:
+        sock = self._sock
+        if sock is None:
+            return
+        body = json.dumps(msg).encode("utf-8")
+        with self._send_lock:
+            if dest is not None:
+                try:
+                    sock.sendto(body, dest)
+                except OSError as e:
+                    print(f"[meshdisco] Reply to {dest} failed: {e}")
+                return
+            # One write per interface — the default route alone reaches a
+            # single network.
+            for ifname, idx, addr in self._ifaces:
+                try:
+                    sock.setsockopt(
+                        socket.IPPROTO_IP,
+                        socket.IP_MULTICAST_IF,
+                        socket.inet_aton(addr) + socket.inet_aton("0.0.0.0") + struct.pack("@i", idx),
+                    )
+                    sock.sendto(body, (self.group, self.port))
+                except OSError as e:
+                    print(f"[meshdisco] Warning: send on {ifname} failed: {e}")
+
+    def _read_loop(self) -> None:
+        while self._running:
+            sock = self._sock
+            if sock is None:
+                return
+            try:
+                data, addr = sock.recvfrom(READ_BUFFER)
+            except OSError:
+                if not self._running:
+                    return
+                continue
+            try:
+                msg = json.loads(data.decode("utf-8"))
+            except (ValueError, UnicodeDecodeError):
+                continue  # foreign traffic on the group
+            if msg.get("v") != PROTOCOL_VERSION:
+                continue
+            # Drop our own multicast echo.
+            if msg.get("instance") == self.instance and msg.get("name") == self.name:
+                continue
+
+            if msg.get("type") == TYPE_DISCOVERY:
+                self._send(self._build_announce("running"), dest=addr)
+                continue
+            if msg.get("type") != TYPE_ANNOUNCE or not msg.get("name"):
+                continue
+
+            key = f"{msg['name']}|{msg.get('instance', '')}"
+            if msg.get("status") == "stopping":
+                with self._lock:
+                    self._neighbors.pop(key, None)
+                continue
+
+            # Source address comes from the packet, never the payload.
+            msg["addr"] = addr[0]
+            msg["lastSeen"] = time.time()
+            with self._lock:
+                self._neighbors[key] = msg
+            for cb in list(self._listeners):
+                try:
+                    cb(msg)
+                except Exception as e:
+                    print(f"[meshdisco] Listener raised: {e}")
+
+    def _announce_loop(self) -> None:
+        self._send(self._build_announce("running"))
+        self.probe()
+        while self._running:
+            time.sleep(self.interval)
+            if not self._running:
+                return
+            try:
+                self._send(self._build_announce("running"))
+            except OSError:
+                pass
+
+
+class MetaCoreLocator:
+    """Locates meta-core over UDP, replacing the kv-leader.info read."""
+
+    def __init__(
+        self,
+        service_name: str,
+        base_url: Optional[str] = None,
+        version: str = "",
+        meta_core_url: Optional[str] = None,
+        enabled: Optional[bool] = None,
+    ) -> None:
+        self._pin = (meta_core_url or "").strip() or None
+        self._current: Optional[dict] = None
+        self._change_callbacks: List[Callable[[], None]] = []
+
+        self._node = MeshNode(
+            name=service_name,
+            role=ROLE_SERVICE,
+            version=version,
+            enabled=_env_flag("ENABLE_UDP_DISCOVERY", True) if enabled is None else enabled,
+            payload=lambda: {"baseUrl": base_url or f"http://{local_ipv4()}"},
+        )
+        self._node.on_announce(self._on_announce)
+
+    def _on_announce(self, nb: dict) -> None:
+        if nb.get("role") != ROLE_CORE or not nb.get("urls"):
+            return
+        if self._pin:
+            return  # pinned: the wire cannot move us
+
+        prev = self._current
+        if prev is None:
+            self._current = nb["urls"]
+            print(f"[meshdisco] meta-core discovered at {nb['urls']['apiUrl']} ({nb.get('addr')})")
+            for cb in list(self._change_callbacks):
+                try:
+                    cb()
+                except Exception as e:
+                    print(f"[meshdisco] Change callback raised: {e}")
+            return
+        if prev.get("apiUrl") == nb["urls"].get("apiUrl"):
+            self._current = nb["urls"]
+            return
+        # A second, different core. Do not flap — keep the first and say so,
+        # because on a PCS box this means a client container can see a core it
+        # must not use.
+        print(
+            f"[meshdisco] Ignoring a second meta-core at {nb['urls']['apiUrl']} "
+            f"({nb.get('addr')}); staying with {prev.get('apiUrl')}. "
+            f"Set META_CORE_URL to pin this explicitly."
+        )
+
+    def start(self) -> None:
+        if self._pin:
+            print(f"[meshdisco] meta-core pinned to {self._pin}; discovery is advisory")
+        self._node.start()
+
+    def stop(self) -> None:
+        self._node.stop()
+        self._change_callbacks = []
+
+    def get_api_url(self) -> Optional[str]:
+        if self._pin:
+            return self._pin
+        if self._current:
+            return self._current.get("apiUrl")
+        return None
+
+    def get_urls(self) -> Optional[dict]:
+        return self._current
+
+    def wait_for_core(self, timeout: Optional[float] = 30.0) -> Optional[str]:
+        """Block until meta-core is located.
+
+        ``timeout=None`` waits forever, which is meta-stremio's default
+        (LEADER_WAIT_TIMEOUT=0). Re-probes every 500ms rather than waiting out
+        an announce interval.
+        """
+        deadline = None if timeout is None else time.time() + timeout
+        logged = False
+        while deadline is None or time.time() < deadline:
+            url = self.get_api_url()
+            if url:
+                return url
+            if not logged:
+                print("[meshdisco] Waiting for meta-core...")
+                logged = True
+            self._node.probe()
+            time.sleep(0.5)
+        return None
+
+    def on_change(self, cb: Callable[[], None]) -> None:
+        self._change_callbacks.append(cb)
+
+    def neighbors(self) -> List[dict]:
+        return self._node.neighbors_by_name()
+
+    def self_announce(self) -> dict:
+        return self._node.self_announce()
+
+    def probe(self) -> None:
+        self._node.probe()

src/storage/meta_consumer.py

@@ -50,11 +50,19 @@ class MetaConsumer:
     def __init__(
         self,
         api_url: str,
-        cursor_path: str = "/meta-core/cursors/meta-stremio-meta.cursor",
+        cursor_path: Optional[str] = None,
         request_timeout: int = 35,
     ) -> None:
         self._api_url = api_url.rstrip("/")
-        self._cursor_path = cursor_path
+        # The SSE resume cursor used to default under /meta-core, which is not
+        # mounted any more (meta-discovery v1 removed the shared volume). It
+        # belongs with the rest of this service's local state: CACHE_FOLDER_PATH
+        # (/data/cache in the dev stack). Without a writable home the consumer
+        # silently replays from `$` on every restart.
+        self._cursor_path = cursor_path or os.path.join(
+            os.environ.get("CACHE_FOLDER_PATH", "/data/cache"),
+            "meta-stremio-meta.cursor",
+        )
         self._request_timeout = request_timeout
         self._running = False
         self._thread: Optional[threading.Thread] = None

src/storage/service_registration.py

@@ -1,297 +0,0 @@
-"""
-Service Registration for meta-stremio.
-
-Python implementation of the service registration pattern used in meta-sort/meta-fuse.
-Each service registers itself in /meta-core/services/{service-name}-{hostname}.json.
-
-Service discovery is centralized in meta-core - this module only handles registration.
-For service discovery, use the /api/services endpoint (fetches from meta-core).
-
-Features:
-- Service registration with heartbeat
-- Automatic heartbeat loop
-"""
-from __future__ import annotations
-
-import os
-import json
-import socket
-import threading
-import time
-from dataclasses import dataclass, asdict
-from datetime import datetime
-from pathlib import Path
-from typing import List, Optional, Dict, Callable
-from urllib.request import urlopen
-from urllib.error import URLError
-
-
-# Configuration
-META_CORE_PATH = os.environ.get('META_CORE_PATH', '/meta-core')
-SERVICE_NAME = os.environ.get('SERVICE_NAME', 'meta-stremio')
-SERVICE_VERSION = os.environ.get('SERVICE_VERSION', '1.0.0')
-BASE_URL = os.environ.get('BASE_URL', '')
-
-
-@dataclass
-class ServiceInfo:
-    """
-    Service registration info stored in JSON file.
-
-    Simplified format matching TypeScript ServiceInfo interface:
-    - name: Service name (e.g., 'meta-stremio')
-    - hostname: Container/host hostname
-    - baseUrl: Base URL for the service (e.g., 'http://localhost:8182')
-    - status: Current status ('running' | 'stale' | 'stopped')
-    - lastHeartbeat: ISO timestamp of last heartbeat
-    - role: Optional role ('leader' | 'follower') - only used by meta-core
-    """
-    name: str
-    hostname: str
-    baseUrl: str
-    status: str = 'running'  # running | stale | stopped
-    lastHeartbeat: str = ''
-    role: Optional[str] = None  # "leader", "follower", or None (for non-meta-core services)
-
-    def to_dict(self) -> dict:
-        """Convert to dictionary for JSON serialization.
-        Excludes None values to avoid polluting JSON with null fields.
-        """
-        result = asdict(self)
-        # Remove None values (e.g., role should only appear for meta-core)
-        return {k: v for k, v in result.items() if v is not None}
-
-
-class ServiceRegistration:
-    """
-    Service Discovery using shared filesystem.
-
-    Each service registers itself in /meta-core/services/{service-name}-{hostname}.json
-    and can discover other registered services.
-    """
-
-    def __init__(
-        self,
-        meta_core_path: str = None,
-        service_name: str = None,
-        version: str = None,
-        api_url: str = None,
-        base_url: str = None,
-        capabilities: List[str] = None,
-        endpoints: Dict[str, str] = None,
-        heartbeat_interval: float = 30.0,
-        stale_threshold: float = 60.0
-    ):
-        self.meta_core_path = meta_core_path or META_CORE_PATH
-        self.service_name = service_name or SERVICE_NAME
-        # Use base_url if provided, then api_url, then env BASE_URL, then auto-detect
-        self.base_url = base_url or api_url or BASE_URL or self._get_default_api_url()
-        self.heartbeat_interval = heartbeat_interval
-        self.stale_threshold = stale_threshold
-
-        # Keep for backwards compatibility but not used in simplified format
-        self._version = version or SERVICE_VERSION
-        self._capabilities = capabilities or []
-        self._endpoints = endpoints or {}
-
-        self.current_hostname = socket.gethostname()
-        self.services_dir = os.path.join(self.meta_core_path, 'services')
-        # Use hostname-based file naming like TypeScript services
-        self.service_file_path = os.path.join(self.services_dir, f'{self.service_name}-{self.current_hostname}.json')
-
-        self._heartbeat_thread: Optional[threading.Thread] = None
-        self._stop_event = threading.Event()
-        self._is_started = False
-
-    def _get_default_api_url(self) -> str:
-        """Get default API URL using local IP and PORT."""
-        port = os.environ.get('PORT', '7000')
-        ip = self._get_local_ip()
-        return f'http://{ip}:{port}'
-
-    def _get_local_ip(self) -> str:
-        """Get the local machine's IP address."""
-        try:
-            # Create a socket to determine the outbound IP
-            s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
-            s.connect(('8.8.8.8', 80))
-            ip = s.getsockname()[0]
-            s.close()
-            return ip
-        except Exception:
-            return socket.gethostname() or 'localhost'
-
-    def _ensure_services_dir(self) -> None:
-        """Ensure services directory exists."""
-        Path(self.services_dir).mkdir(parents=True, exist_ok=True)
-
-    def _build_service_info(self, status: str = 'running') -> ServiceInfo:
-        """Build service info for this service (simplified format)."""
-        return ServiceInfo(
-            name=self.service_name,
-            hostname=self.current_hostname,
-            baseUrl=self.base_url,
-            status=status,
-            lastHeartbeat=datetime.utcnow().isoformat() + 'Z'
-        )
-
-    def register(self) -> None:
-        """Register this service."""
-        self._ensure_services_dir()
-
-        info = self._build_service_info('starting')
-
-        with open(self.service_file_path, 'w') as f:
-            json.dump(info.to_dict(), f, indent=2)
-
-        print(f'[ServiceRegistration] Registered {self.service_name}')
-
-    def update_status(self, status: str) -> None:
-        """Update service status."""
-        try:
-            with open(self.service_file_path, 'r') as f:
-                info = json.load(f)
-
-            info['status'] = status
-            info['lastHeartbeat'] = datetime.utcnow().isoformat() + 'Z'
-
-            with open(self.service_file_path, 'w') as f:
-                json.dump(info, f, indent=2)
-
-        except Exception as e:
-            print(f'[ServiceRegistration] Failed to update status: {e}')
-
-    def heartbeat(self) -> None:
-        """Send heartbeat (update lastHeartbeat timestamp)."""
-        try:
-            with open(self.service_file_path, 'r') as f:
-                info = json.load(f)
-
-            info['lastHeartbeat'] = datetime.utcnow().isoformat() + 'Z'
-
-            with open(self.service_file_path, 'w') as f:
-                json.dump(info, f, indent=2)
-
-        except FileNotFoundError:
-            # If file was deleted, re-register
-            self.register()
-            self.update_status('running')
-        except Exception as e:
-            print(f'[ServiceRegistration] Heartbeat failed: {e}')
-
-    def _heartbeat_loop(self) -> None:
-        """Background heartbeat loop."""
-        while not self._stop_event.is_set():
-            try:
-                self.heartbeat()
-            except Exception as e:
-                print(f'[ServiceRegistration] Heartbeat error: {e}')
-
-            self._stop_event.wait(self.heartbeat_interval)
-
-    def start_heartbeat(self) -> None:
-        """Start heartbeat loop in background thread."""
-        if self._heartbeat_thread and self._heartbeat_thread.is_alive():
-            return
-
-        self._stop_event.clear()
-        self._heartbeat_thread = threading.Thread(
-            target=self._heartbeat_loop,
-            daemon=True,
-            name='service-heartbeat'
-        )
-        self._heartbeat_thread.start()
-
-    def stop_heartbeat(self) -> None:
-        """Stop heartbeat loop."""
-        self._stop_event.set()
-        if self._heartbeat_thread and self._heartbeat_thread.is_alive():
-            self._heartbeat_thread.join(timeout=2.0)
-
-    def unregister(self) -> None:
-        """Unregister this service (on shutdown)."""
-        self.stop_heartbeat()
-
-        try:
-            self.update_status('stopped')
-            print(f'[ServiceRegistration] Unregistered {self.service_name}')
-        except Exception as e:
-            print(f'[ServiceRegistration] Failed to unregister: {e}')
-
-    # ========================================================================
-    # Lifecycle
-    # ========================================================================
-
-    def start(self) -> None:
-        """Full startup sequence."""
-        if self._is_started:
-            return
-
-        # Try to register, but don't fail if filesystem is read-only
-        # This allows discovery-only mode when we can't write
-        can_register = os.access(self.services_dir, os.W_OK) if os.path.exists(self.services_dir) else False
-
-        if can_register:
-            try:
-                self.register()
-                self.update_status('running')
-                self.start_heartbeat()
-                print(f'[ServiceRegistration] Started {self.service_name}-{self.current_hostname} at {self.base_url}')
-            except Exception as e:
-                print(f'[ServiceRegistration] Registration failed (read-only?): {e}')
-                print(f'[ServiceRegistration] Running in discovery-only mode')
-        else:
-            print(f'[ServiceRegistration] Services directory not writable, running in discovery-only mode')
-
-        self._is_started = True
-
-    def stop(self) -> None:
-        """Full shutdown sequence."""
-        if not self._is_started:
-            return
-
-        self.unregister()
-        self._is_started = False
-
-    # ========================================================================
-    # Getters
-    # ========================================================================
-
-    def get_service_file_path(self) -> str:
-        return self.service_file_path
-
-    def get_services_dir(self) -> str:
-        return self.services_dir
-
-    def is_started(self) -> bool:
-        return self._is_started
-
-
-# Singleton instance
-_service_registration: Optional[ServiceRegistration] = None
-
-
-def get_service_discovery() -> ServiceRegistration:
-    """Get or create the service registration singleton.
-
-    Note: Function name kept as 'get_service_discovery' for backward compatibility.
-    """
-    global _service_registration
-    if _service_registration is None:
-        _service_registration = ServiceRegistration()
-    return _service_registration
-
-
-def init_service_discovery(
-    api_url: str = None,
-    base_url: str = None,
-    **kwargs
-) -> ServiceRegistration:
-    """Initialize and start the service registration singleton.
-
-    Note: Function name kept as 'init_service_discovery' for backward compatibility.
-    """
-    global _service_registration
-    _service_registration = ServiceRegistration(api_url=api_url, base_url=base_url, **kwargs)
-    _service_registration.start()
-    return _service_registration

www/index.html

@@ -292,6 +292,14 @@
         footer a:hover { text-decoration: underline; }
 
         /* Services Navigation */
+        /* <meta-service-menu> is in shadow DOM and cannot read the vars
+           above; map this UI's palette onto the four tokens it does read. */
+        meta-service-menu {
+            --mm-nav-fg: var(--text-primary);
+            --mm-nav-bg: var(--bg-tertiary);
+            --mm-nav-border: var(--border-color);
+            --mm-nav-accent: var(--accent-primary);
+        }
         .services-nav {
             background: var(--bg-secondary);
             border-radius: 12px;
@@ -362,16 +370,13 @@
             .services-nav { flex-direction: column; align-items: flex-start; }
         }
     </style>
+    <script type="module" src="meta-service-menu.js"></script>
 </head>
 <body>
     <div class="container">
-        <!-- Services Navigation -->
+        <!-- Neighbouring meta-* services (meta-discovery v1). The element
-        <nav class="services-nav" id="servicesNav" style="display: none;">
+             fetches /api/neighbors from this origin and renders a burger menu. -->
-            <span class="services-nav-label">Services:</span>
+        <meta-service-menu id="servicesNav" current="meta-stremio"></meta-service-menu>
-            <div class="services-nav-items" id="servicesNavItems">
-                <!-- Populated by JavaScript -->
-            </div>
-        </nav>
 
         <header class="header">
             <div>
@@ -548,78 +553,18 @@
             }
         }
 
-        // Service icons mapping
+        // Service nav is <meta-service-menu> now — no local icon map,
-        const serviceIcons = {
+        // name formatter or poller lives here any more.
-            'meta-sort': '📁',
-            'meta-fuse': '🗂️',
-            'meta-stremio': '🎬',
-            'meta-core': '⚙️',
-            'meta-dup': '🔍',
-            'meta-orbit': '🌐',
-            'default': '📦'
-        };
-
-        // Fetch and display discovered services
-        async function fetchServices() {
-            try {
-                const res = await fetch('/api/services');
-                const data = await res.json();
-
-                const navEl = document.getElementById('servicesNav');
-                const itemsEl = document.getElementById('servicesNavItems');
-
-                if (data.services && data.services.length > 0) {
-                    // Sort services by name
-                    const services = data.services.sort((a, b) => a.name.localeCompare(b.name));
-
-                    // Build navigation items
-                    itemsEl.innerHTML = services.map(svc => {
-                        const icon = serviceIcons[svc.name] || serviceIcons.default;
-                        const isActive = svc.name === data.current;
-                        const activeClass = isActive ? ' active' : '';
-
-                        // Use external URL for other services, current path for self
-                        const href = isActive ? '#' : svc.baseUrl;
-
-                        return `
-                            <a href="${href}" class="service-link${activeClass}" onclick="event.preventDefault(); ${isActive ? '' : `window.location.href='${svc.baseUrl}'`}">
-                                <span class="service-icon">${icon}</span>
-                                <span>${formatServiceName(svc.name)}</span>
-                                <span class="service-status"></span>
-                            </a>
-                        `;
-                    }).join('');
-
-                    navEl.style.display = 'flex';
-                } else {
-                    navEl.style.display = 'none';
-                }
-            } catch (e) {
-                console.error('Failed to fetch services:', e);
-                document.getElementById('servicesNav').style.display = 'none';
-            }
-        }
-
-        function formatServiceName(name) {
-            // Convert 'meta-stremio' to 'Meta Stremio'
-            return name.split('-').map(word =>
-                word.charAt(0).toUpperCase() + word.slice(1)
-            ).join(' ');
-        }
 
         // Initial fetch
         fetchStats();
         fetchMetrics();
-        fetchServices();
 
         // Refresh metrics every 5 seconds
         setInterval(fetchMetrics, 5000);
 
         // Refresh stats every 30 seconds
         setInterval(fetchStats, 30000);
-
-        // Refresh services every 60 seconds
-        setInterval(fetchServices, 60000);
     </script>
 </body>
 </html>

www/meta-service-menu.js

@@ -0,0 +1,303 @@
+/**
+ * <meta-service-menu> — the neighbouring meta-* services, as a burger menu.
+ *
+ * Replaces the eight divergent "services bar" implementations (four copies of
+ * a React `ServiceNav.tsx`, two mirrored `services.js`, two bespoke inline
+ * ones) with a single framework-free custom element. It drops into the React
+ * dashboards as a plain tag — no React dependency — and into the vanilla UIs
+ * as an ES module.
+ *
+ * Data comes from the host service's own `GET /api/neighbors`, same-origin.
+ * Never cross-origin: the dev stack's self-signed certs and the OIDC redirects
+ * make a fetch to a neighbour a dead end. Links are plain navigations.
+ *
+ * ⚠ MIRRORED FILE. Byte-identical copies live in every meta-* UI tree; see
+ * `scripts/check-mirrors.sh`. Edit one → edit all.
+ *
+ * Shadow DOM is not optional here. The eight UIs use four incompatible token
+ * vocabularies, and two of them define *different colours under the same
+ * names* (`--bg-secondary` is #12121a in meta-sort, #16213e in meta-core,
+ * #161b22 in meta-dup). meta-search and meta-share are also the only
+ * light-mode-capable UIs, so a hardcoded dark palette would break them.
+ * The element therefore styles itself from four tokens with dark defaults:
+ *
+ *     --mm-nav-fg, --mm-nav-bg, --mm-nav-border, --mm-nav-accent
+ *
+ * Each host maps its own variables onto those, e.g.
+ *     meta-service-menu { --mm-nav-bg: var(--bg-secondary); }
+ *
+ * Attributes:
+ *   current   name of the host service, so it can be marked and not linked
+ *   endpoint  override the default "/api/neighbors"
+ *   label     button tooltip (default "Services")
+ *
+ * Property:
+ *   rewriteUrl  optional (name, url, neighbour) => url, for hosts that must
+ *               translate internal hostnames into browser-reachable ones
+ *               (meta-gateway does this for its dev/prod split).
+ */
+
+const ICONS = {
+  "meta-core": "⚙️",
+  "meta-sort": "📁",
+  "meta-fuse": "🗂️",
+  "meta-stremio": "🎬",
+  "meta-dup": "🔍",
+  "meta-search": "🔎",
+  "meta-share": "🔗",
+  "meta-gateway": "🌉",
+  "meta-watch": "📺",
+  "meta-listen": "🎧",
+  "meta-read": "📖",
+};
+const DEFAULT_ICON = "📦";
+const POLL_MS = 30000;
+
+function titleCase(name) {
+  return String(name || "")
+    .split("-")
+    .map((w) => w.charAt(0).toUpperCase() + w.slice(1))
+    .join(" ");
+}
+
+const TEMPLATE = `
+<style>
+  :host {
+    --_fg: var(--mm-nav-fg, #e8eaed);
+    --_bg: var(--mm-nav-bg, #1a1a2e);
+    --_border: var(--mm-nav-border, rgba(255, 255, 255, 0.14));
+    --_accent: var(--mm-nav-accent, #4ecdc4);
+    position: relative;
+    display: inline-block;
+    font-family: inherit;
+    font-size: 14px;
+    line-height: 1.4;
+  }
+  button.trigger {
+    display: inline-flex;
+    align-items: center;
+    justify-content: center;
+    gap: 0.4rem;
+    width: 2.1rem;
+    height: 2.1rem;
+    padding: 0;
+    cursor: pointer;
+    color: var(--_fg);
+    background: transparent;
+    border: 1px solid var(--_border);
+    border-radius: 8px;
+    font: inherit;
+    transition: border-color 0.15s, background 0.15s;
+  }
+  button.trigger:hover { border-color: var(--_accent); }
+  button.trigger:focus-visible {
+    outline: 2px solid var(--_accent);
+    outline-offset: 2px;
+  }
+  button.trigger[aria-expanded="true"] { border-color: var(--_accent); }
+  .bars { display: block; width: 15px; }
+  .bars span {
+    display: block;
+    height: 2px;
+    margin: 3px 0;
+    border-radius: 2px;
+    background: currentColor;
+  }
+  .panel {
+    position: absolute;
+    top: calc(100% + 6px);
+    right: 0;
+    z-index: 1000;
+    min-width: 13rem;
+    max-height: 70vh;
+    overflow-y: auto;
+    padding: 0.3rem;
+    background: var(--_bg);
+    border: 1px solid var(--_border);
+    border-radius: 10px;
+    box-shadow: 0 10px 30px rgba(0, 0, 0, 0.35);
+  }
+  .panel[hidden] { display: none; }
+  a.item, .item {
+    display: flex;
+    align-items: center;
+    gap: 0.55rem;
+    padding: 0.45rem 0.6rem;
+    border-radius: 6px;
+    color: var(--_fg);
+    text-decoration: none;
+    white-space: nowrap;
+  }
+  a.item:hover { background: color-mix(in srgb, var(--_accent) 18%, transparent); }
+  .item.active { color: var(--_accent); font-weight: 600; cursor: default; }
+  .item.dead { opacity: 0.55; cursor: not-allowed; }
+  .icon { font-size: 1rem; line-height: 1; }
+  .dot {
+    width: 6px;
+    height: 6px;
+    margin-left: auto;
+    border-radius: 50%;
+    background: var(--_accent);
+  }
+  .empty {
+    padding: 0.6rem;
+    opacity: 0.7;
+    font-size: 0.85em;
+    white-space: nowrap;
+  }
+  /* Respect a host that opts out of motion. */
+  @media (prefers-reduced-motion: reduce) {
+    button.trigger { transition: none; }
+  }
+</style>
+<button class="trigger" type="button" aria-haspopup="menu" aria-expanded="false">
+  <span class="bars" aria-hidden="true"><span></span><span></span><span></span></span>
+</button>
+<div class="panel" role="menu" hidden></div>
+`;
+
+class MetaServiceMenu extends HTMLElement {
+  constructor() {
+    super();
+    this.attachShadow({ mode: "open" });
+    this.shadowRoot.innerHTML = TEMPLATE;
+    this._btn = this.shadowRoot.querySelector("button.trigger");
+    this._panel = this.shadowRoot.querySelector(".panel");
+    this._neighbors = [];
+    this._timer = null;
+    this._open = false;
+
+    this._onDocClick = (e) => {
+      if (!this._open) return;
+      if (e.composedPath().includes(this)) return;
+      this._setOpen(false);
+    };
+    this._onKey = (e) => {
+      if (e.key === "Escape" && this._open) {
+        this._setOpen(false);
+        this._btn.focus();
+      }
+    };
+  }
+
+  connectedCallback() {
+    this._btn.title = this.getAttribute("label") || "Services";
+    this._btn.setAttribute("aria-label", this._btn.title);
+    this._btn.addEventListener("click", () => this._setOpen(!this._open));
+    document.addEventListener("click", this._onDocClick);
+    document.addEventListener("keydown", this._onKey);
+
+    this.refresh();
+    this._timer = setInterval(() => this.refresh(), POLL_MS);
+  }
+
+  disconnectedCallback() {
+    if (this._timer) clearInterval(this._timer);
+    this._timer = null;
+    document.removeEventListener("click", this._onDocClick);
+    document.removeEventListener("keydown", this._onKey);
+  }
+
+  get endpoint() {
+    return this.getAttribute("endpoint") || "/api/neighbors";
+  }
+
+  get current() {
+    return this.getAttribute("current") || "";
+  }
+
+  async refresh() {
+    try {
+      const res = await fetch(this.endpoint, {
+        headers: { Accept: "application/json" },
+      });
+      if (!res.ok) throw new Error(String(res.status));
+      const data = await res.json();
+      // `services` is the legacy alias some backends still emit.
+      this._neighbors = data.neighbors || data.services || [];
+    } catch {
+      // Silent: the menu is navigation, not a health indicator. Keep whatever
+      // we last had rather than blanking the list on one failed poll.
+    }
+    this._render();
+  }
+
+  _setOpen(open) {
+    this._open = open;
+    this._panel.hidden = !open;
+    this._btn.setAttribute("aria-expanded", String(open));
+    if (open) this.refresh();
+  }
+
+  _render() {
+    const panel = this._panel;
+    panel.textContent = "";
+
+    const seen = new Set();
+    const rows = [];
+    for (const n of this._neighbors) {
+      const name = n.name || "";
+      if (!name || seen.has(name)) continue;
+      seen.add(name);
+      rows.push(n);
+    }
+    // Always show the host itself, even before its own echo comes back.
+    if (this.current && !seen.has(this.current)) {
+      rows.push({ name: this.current, baseUrl: "" });
+    }
+    rows.sort((a, b) => String(a.name).localeCompare(String(b.name)));
+
+    if (rows.length === 0) {
+      const empty = document.createElement("div");
+      empty.className = "empty";
+      empty.textContent = "No services found";
+      panel.appendChild(empty);
+      return;
+    }
+
+    for (const n of rows) {
+      const isActive = n.name === this.current;
+      let url = n.baseUrl || "";
+      if (typeof this.rewriteUrl === "function") {
+        try {
+          url = this.rewriteUrl(n.name, url, n) || "";
+        } catch {
+          /* a bad hook must not blank the menu */
+        }
+      }
+
+      // A neighbour with no reachable URL is shown but not linked — better
+      // than a link that goes nowhere.
+      const el = document.createElement(isActive || !url ? "span" : "a");
+      el.className =
+        "item" + (isActive ? " active" : "") + (!isActive && !url ? " dead" : "");
+      if (el.tagName === "A") {
+        el.href = url;
+        el.setAttribute("role", "menuitem");
+      }
+      if (!isActive && !url) el.title = "No reachable URL announced";
+
+      const icon = document.createElement("span");
+      icon.className = "icon";
+      icon.textContent = ICONS[n.name] || DEFAULT_ICON;
+
+      const label = document.createElement("span");
+      label.textContent = titleCase(n.name);
+
+      el.append(icon, label);
+      if (n.status === "running" || isActive) {
+        const dot = document.createElement("span");
+        dot.className = "dot";
+        dot.title = n.status || "running";
+        el.appendChild(dot);
+      }
+      panel.appendChild(el);
+    }
+  }
+}
+
+if (!customElements.get("meta-service-menu")) {
+  customElements.define("meta-service-menu", MetaServiceMenu);
+}
+
+export default MetaServiceMenu;