From 028e3b2240555d49ae77c6a8a090bc023f32a2e2 Mon Sep 17 00:00:00 2001
From: duffyduck <info@hacker-net.de>
Date: Wed, 22 Apr 2026 19:32:40 +0200
Subject: [PATCH] fix: Voice-Auswahl funktioniert endlich + Diagnostic setzt
 alle Apps zurueck
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

XTTS-Bridge: im daswer123 local-Mode erwartet der Server speaker_wav als
Basename (z.B. "Maia"), nicht als Pfad. Wir haben bisher "/voices/Maia.wav"
geschickt, was der Server stumm verwirft und Default nimmt. Jetzt: speaker
name pur senden + Warnlog wenn File fehlt.

App: ChatScreen + SettingsScreen horchen auf type "config" vom RVS —
wenn in Diagnostic die globale XTTS-Voice gewechselt wird, werden alle
Apps auf den neuen Wert zurueckgesetzt (wie vom User gewuenscht).
Lokale App-Wahl bleibt sonst intakt und gewinnt pro Request.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---
 android/src/screens/ChatScreen.tsx     |  9 +++++++++
 android/src/screens/SettingsScreen.tsx |  7 +++++++
 xtts/bridge.js                         | 18 ++++++++++++++----
 3 files changed, 30 insertions(+), 4 deletions(-)

diff --git a/android/src/screens/ChatScreen.tsx b/android/src/screens/ChatScreen.tsx
index 9bf1240..72f7689 100644
--- a/android/src/screens/ChatScreen.tsx
+++ b/android/src/screens/ChatScreen.tsx
@@ -325,6 +325,15 @@ const ChatScreen: React.FC = () => {
         const tool = (message.payload.tool as string) || '';
         setAgentActivity({ activity, tool });
       }
+
+      // Voice-Config aus Diagnostic — setzt die lokale App-Stimme auf den
+      // gerade in Diagnostic gewaehlten Wert zurueck. User-Wahl in der App
+      // wird dadurch ueberschrieben.
+      if (message.type === ('config' as any)) {
+        const newVoice = ((message.payload as any).xttsVoice as string) ?? '';
+        localXttsVoiceRef.current = newVoice;
+        AsyncStorage.setItem('aria_xtts_voice', newVoice);
+      }
     });
 
     const unsubState = rvs.onStateChange((state) => {
diff --git a/android/src/screens/SettingsScreen.tsx b/android/src/screens/SettingsScreen.tsx
index b55937e..26b558c 100644
--- a/android/src/screens/SettingsScreen.tsx
+++ b/android/src/screens/SettingsScreen.tsx
@@ -265,6 +265,13 @@ const SettingsScreen: React.FC = () => {
         }
         rvs.send('xtts_list_voices' as any, {});
       }
+
+      // Diagnostic-Voice-Wechsel → lokale App-Stimme auf den neuen Default zuruecksetzen
+      if (message.type === ('config' as any)) {
+        const newVoice = ((message.payload as any).xttsVoice as string) ?? '';
+        setXttsVoice(newVoice);
+        AsyncStorage.setItem('aria_xtts_voice', newVoice);
+      }
     });
 
     return () => {
diff --git a/xtts/bridge.js b/xtts/bridge.js
index edfc698..c1ff8fa 100644
--- a/xtts/bridge.js
+++ b/xtts/bridge.js
@@ -151,8 +151,18 @@ async function _runTTSRequest(payload) {
   log(`TTS-Request (streaming): "${cleanText.slice(0, 80)}..." (${cleanText.length} chars, voice: ${voice || "default"})`);
 
   try {
-    const voiceSample = voice ? path.join(VOICES_DIR, `${voice}.wav`) : null;
-    const hasCustomVoice = voiceSample && fs.existsSync(voiceSample);
+    // Im local-Mode erwartet daswer123 XTTS speaker_wav als Basename (ohne .wav,
+    // ohne Pfad) — der Server prefixt EXAMPLE_FOLDER selbst. Wir checken hier
+    // nur das physische File ab um Warnungen zu loggen; runter ans API geht
+    // nur der Name.
+    const voiceFilePath = voice ? path.join(VOICES_DIR, `${voice}.wav`) : null;
+    const hasCustomVoice = voiceFilePath && fs.existsSync(voiceFilePath);
+    const speakerName = hasCustomVoice ? voice : "";
+    if (voice && !hasCustomVoice) {
+      log(`WARNUNG: Voice "${voice}" angefordert, aber ${voiceFilePath} existiert nicht — nehme Default`);
+    } else if (hasCustomVoice) {
+      log(`Voice "${voice}" verwendet (speaker_wav="${speakerName}")`);
+    }
 
     let chunkIndex = 0;
     let pcmMeta = null;
@@ -190,7 +200,7 @@ async function _runTTSRequest(payload) {
       await streamXTTSAsPCM(
         cleanText,
         language || "de",
-        hasCustomVoice ? voiceSample : null,
+        speakerName,
         onChunk,
       );
     } catch (streamErr) {
@@ -198,7 +208,7 @@ async function _runTTSRequest(payload) {
       await streamXTTSBatch(
         cleanText,
         language || "de",
-        hasCustomVoice ? voiceSample : null,
+        speakerName,
         onChunk,
       );
     }