From 1ce65407118e50e1a52f478d7e709460787d1b92 Mon Sep 17 00:00:00 2001
From: dschlueter <dschlueter@kitux.de>
Date: Thu, 18 Jun 2026 09:45:17 +0200
Subject: [PATCH] feat(service): no_playback-Flag (nur WAV erzeugen, kein
 lokales Abspielen)

Fuer Remote-/Gateway-Nutzung: mit no_playback=true generiert der Worker nur die
WAV (keep_audio) und startet keinen PlaybackWorker -> der Server spielt nicht laut
ab. Default false (lokales Abspielen unveraendert).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---
 tts_service.py | 26 +++++++++++++++++---------
 1 file changed, 17 insertions(+), 9 deletions(-)

diff --git a/tts_service.py b/tts_service.py
index f71abc8..d073314 100644
--- a/tts_service.py
+++ b/tts_service.py
@@ -99,6 +99,7 @@ class SpeakJob:
     pronunciation_dict: Optional[dict]
     session_id: Optional[str]
     keep_audio: bool = False
+    no_playback: bool = False  # nur WAV erzeugen, NICHT lokal abspielen (Remote-/Gateway-Nutzung)
     status: JobStatus = field(default=JobStatus.pending)
     text_preview: str = field(default="")
     chunks_total: int = 0
@@ -144,13 +145,16 @@ def _worker() -> None:
             job.chunks_total = len(chunks)
             job.text_preview = job.text[:80]
 
-            playback = tts.PlaybackWorker(
-                sample_rate=sr,
-                device=job.audio_device or "pulse",
-                speed=job.speed,
-                stop_event=tts.STOP_REQUESTED,
-            )
-            playback.start()
+            # no_playback: nur generieren (kein lokales Abspielen) -> Remote/Gateway.
+            playback = None
+            if not job.no_playback:
+                playback = tts.PlaybackWorker(
+                    sample_rate=sr,
+                    device=job.audio_device or "pulse",
+                    speed=job.speed,
+                    stop_event=tts.STOP_REQUESTED,
+                )
+                playback.start()
 
             wavs: list[torch.Tensor] = []
             try:
@@ -159,10 +163,12 @@ def _worker() -> None:
                         break
                     wav = tts.generate_chunk(model, model_kind, chunk, job.lang, job.voice)
                     wavs.append(wav)
-                    playback.put(wav)
+                    if playback is not None:
+                        playback.put(wav)
                     job.chunks_done += 1
             finally:
-                playback.stop()
+                if playback is not None:
+                    playback.stop()
 
             if wavs:
                 final = wavs[0] if len(wavs) == 1 else torch.cat(wavs, dim=-1)
@@ -215,6 +221,7 @@ class SpeakRequest(BaseModel):
     session_id: Optional[str] = None
     pronunciation_dict: Optional[dict] = None
     keep_audio: bool = False  # WAV im Cache behalten für GET /audio/{job_id}
+    no_playback: bool = False  # nur generieren (kein lokales Abspielen) -> fuer Remote/Gateway
 
 
 def _job_to_dict(j: SpeakJob) -> dict:
@@ -284,6 +291,7 @@ def speak(req: SpeakRequest):
         pronunciation_dict=req.pronunciation_dict,
         session_id=req.session_id,
         keep_audio=req.keep_audio,
+        no_playback=req.no_playback,
     )
     _job_queue.put(job)