feat(voice): zentraler Stimmen-Store + Auto-Provisioning der f5tts-Flotte

Stimmen lagen bisher nur pro Box (xtts/voices/). Jetzt haelt der Diagnostic-
Server sie zentral als /shared/voices/{name}.tar.gz und versorgt jede f5tts-Box
automatisch.

- diagnostic/server.js (Bibliothekar):
  * zentraler Store + Helfer (list/read/write/delete tar.gz)
  * reconcileVoices() beim worker_hello einer f5tts-Box: push fehlende zentrale
    Stimmen (xtts_import_voice, targetInstance) + pull box-eigene, zentral
    fehlende (xtts_export_voice) → seedet den Store aus der ersten Box
  * autonome RVS-Listener: xtts_voice_saved → zentraler Ingest (auch App-Uploads);
    xtts_voice_exported (nur eigene requestId) → in Store schreiben;
    xtts_delete_voice → zentrale Kopie mitloeschen
  * Delete-Action loescht zentrale Kopie direkt mit
- xtts/f5tts/bridge.py: worker_hello traegt jetzt voices:[names] (ohne
  default_ref) fuer den Abgleich; handle_export_voice echot requestId zurueck
  (Korrelation zentral vs. browser-initiiert)

Wiederverwendet den vorhandenen export/import-tar.gz-Transport + Stage-3-
targetInstance-Filter. Playback (preview_voice) und Lastrouting (freieste Box)
existierten bereits. Keine App-/aria-bridge-Aenderung.

Deploy: diagnostic + f5tts-Boxen neu bauen.

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

xtts/f5tts/bridge.py

@@ -675,6 +675,18 @@ async def handle_voice_upload(ws, payload: dict) -> None:
         await _send(ws, "xtts_voice_saved", {"name": name, "error": str(e)[:200]})
 
 
+def _local_voice_names() -> list:
+    """Namen der lokal vorhandenen Nutzer-Stimmen (ohne den Box-lokalen
+    default_ref-Fallback) — fuer die Provisioning-Reconciliation im worker_hello."""
+    names = []
+    if VOICES_DIR.exists():
+        for wav in sorted(VOICES_DIR.glob("*.wav")):
+            if wav.stem == "default_ref":
+                continue
+            names.append(wav.stem)
+    return names
+
+
 async def handle_list_voices(ws) -> None:
     try:
         voices = []
@@ -711,13 +723,16 @@ async def handle_delete_voice(ws, payload: dict) -> None:
 async def handle_export_voice(ws, payload: dict) -> None:
     """Packt eine Stimme (.wav + .txt) als tar.gz und sendet sie base64 zurueck."""
     name = (payload.get("name") or "").strip()
+    # requestId aus dem Request zuruueckspiegeln — der Diagnostic-Bibliothekar
+    # korreliert damit zentrale Exports gegen browser-initiierte.
+    req_id = payload.get("requestId", "") or ""
     if not name:
-        await _send(ws, "xtts_voice_exported", {"ok": False, "error": "name fehlt"})
+        await _send(ws, "xtts_voice_exported", {"ok": False, "requestId": req_id, "error": "name fehlt"})
         return
     try:
         wav, txt = voice_paths(name)
         if not wav.exists():
-            await _send(ws, "xtts_voice_exported", {"ok": False, "name": name, "error": "Stimme nicht gefunden"})
+            await _send(ws, "xtts_voice_exported", {"ok": False, "requestId": req_id, "name": name, "error": "Stimme nicht gefunden"})
             return
         import io, tarfile
         buf = io.BytesIO()
@@ -727,10 +742,10 @@ async def handle_export_voice(ws, payload: dict) -> None:
                 tar.add(txt, arcname=txt.name)
         data = base64.b64encode(buf.getvalue()).decode("ascii")
         logger.info("Voice exportiert: %s (%d KB tar.gz)", name, len(buf.getvalue()) // 1024)
-        await _send(ws, "xtts_voice_exported", {"ok": True, "name": name, "data": data})
+        await _send(ws, "xtts_voice_exported", {"ok": True, "requestId": req_id, "name": name, "data": data})
     except Exception as e:
         logger.exception("handle_export_voice Fehler")
-        await _send(ws, "xtts_voice_exported", {"ok": False, "name": name, "error": str(e)[:200]})
+        await _send(ws, "xtts_voice_exported", {"ok": False, "requestId": req_id, "name": name, "error": str(e)[:200]})
 
 
 async def handle_import_voice(ws, payload: dict) -> None:
@@ -827,6 +842,7 @@ async def _worker_register(ws, *, model: str = "", busy_fn=None) -> None:
         await _send(ws, "worker_hello", {
             "instanceId": INSTANCE_ID, "service": WORKER_SERVICE,
             "node": NODE_NAME, "gpus": GPU_IDS, "model": model,
+            "voices": _local_voice_names(),   # fuer Voice-Provisioning-Reconciliation
         })
         while True:
             await asyncio.sleep(WORKER_PING_INTERVAL_S)