Merge #4016 (re-arm browser voice mode when speechSynthesis drops, #3983) onto master

static/boot.js

@@ -853,8 +853,48 @@ window._micPendingSend=window._micPendingSend||false;
   // a different session's last assistant reply if the user navigated away
   // between send and stream completion. (Opus pre-release advisor.)
   let _voiceModeThinkingSid=null;
+  let _browserTtsKeepAlive=null;
+  let _browserTtsWatchdog=null;
+  let _browserTtsSuppressNextErrorRearm=false;
   const SILENCE_MS=1800; // auto-send after 1.8s silence
 
+  function _clearBrowserTtsRecovery(){
+    if(_browserTtsKeepAlive){
+      clearInterval(_browserTtsKeepAlive);
+      _browserTtsKeepAlive=null;
+    }
+    if(_browserTtsWatchdog){
+      clearTimeout(_browserTtsWatchdog);
+      _browserTtsWatchdog=null;
+    }
+  }
+
+  function _armBrowserTtsRecovery(clean, rate){
+    _clearBrowserTtsRecovery();
+    _browserTtsSuppressNextErrorRearm=false;
+    const safeRate=(Number.isFinite(rate)&&rate>0)?rate:1;
+    // Chromium can drop utter.onend on later turns, so force a recovery path.
+    const watchdogMs=Math.max(4000,Math.round((String(clean||'').length/(12*safeRate))*1000)+10000);
+    _browserTtsWatchdog=setTimeout(()=>{
+      if(!_voiceModeActive||_voiceModeState!=='speaking') return;
+      _browserTtsSuppressNextErrorRearm=true;
+      try{ speechSynthesis.cancel(); }catch(_){}
+      _clearBrowserTtsRecovery();
+      _startListening();
+    },watchdogMs);
+    _browserTtsKeepAlive=setInterval(()=>{
+      if(!_voiceModeActive||_voiceModeState!=='speaking'){
+        _clearBrowserTtsRecovery();
+        return;
+      }
+      if(!speechSynthesis.speaking) return;
+      try{
+        speechSynthesis.pause();
+        speechSynthesis.resume();
+      }catch(_){}
+    },10000);
+  }
+
   function _setState(state){
     _voiceModeState=state;
     indicator.className='voice-mode-indicator '+state;
@@ -867,6 +907,7 @@ window._micPendingSend=window._micPendingSend||false;
 
   function _startListening(){
     if(!_voiceModeActive) return;
+    _clearBrowserTtsRecovery();
     _setState('listening');
 
     _recognition=new SpeechRecognition();
@@ -1057,14 +1098,27 @@ window._micPendingSend=window._micPendingSend||false;
     if(!isNaN(savedPitch)) utter.pitch=Math.min(2,Math.max(0,savedPitch));
 
     utter.onend=()=>{
+      _browserTtsSuppressNextErrorRearm=false;
+      _clearBrowserTtsRecovery();
       // After speaking, go back to listening
-      if(_voiceModeActive) setTimeout(()=>_startListening(),500);
+      if(_voiceModeActive&&_voiceModeState==='speaking') setTimeout(()=>_startListening(),500);
     };
     utter.onerror=()=>{
+      _clearBrowserTtsRecovery();
+      if(_browserTtsSuppressNextErrorRearm){
+        _browserTtsSuppressNextErrorRearm=false;
+        return;
+      }
       if(_voiceModeActive) setTimeout(()=>_startListening(),1000);
     };
 
-    speechSynthesis.speak(utter);
+    _armBrowserTtsRecovery(clean, utter.rate);
+    try{
+      speechSynthesis.speak(utter);
+    }catch(_){
+      _clearBrowserTtsRecovery();
+      if(_voiceModeActive) setTimeout(()=>_startListening(),1000);
+    }
   }
 
   // Hook into response completion — observe when the agent finishes
@@ -1121,10 +1175,12 @@ window._micPendingSend=window._micPendingSend||false;
     _voiceModeActive=false;
     _voiceModeState='idle';
     _voiceModeThinkingSid=null;
+    _browserTtsSuppressNextErrorRearm=false;
     modeBtn.classList.remove('active');
     _setButtonTooltip(modeBtn, t('voice_mode_toggle'));
     bar.style.display='none';
     clearTimeout(_silenceTimer);
+    _clearBrowserTtsRecovery();
     try{ if(_recognition) _recognition.abort(); }catch(_){}
     _recognition=null;
     if(typeof stopTTS==='function') stopTTS();