fix(android): talk mode stability — thread safety, TTS fallback, mic cooldown

Bug fixes:
- @Synchronized on ElevenLabsStreamingTts.sendText/finish to prevent
  sentFullText/sentTextLength races across OkHttp and caller threads
- Pre-set pendingRunId via onRunIdKnown callback before chat.send to
  eliminate race where gateway events arrive before runId is stored
- Track drain coroutine as Job; cancel prior on rapid mic toggle to
  prevent duplicate TTS and stale transcript sends
- Mic button disabled during 2s drain cooldown (micCooldown StateFlow)

Codex review fixes:
- Gate agent streaming TTS on sessionKey to prevent cross-session
  audio leaks (P1)
- Clear ElevenLabs credentials when talk.provider is not elevenlabs;
  gate streaming TTS on activeProviderIsElevenLabs (P2)

System TTS fallback fixes:
- Null streamingTts immediately in finishStreamingTts so next response
  gets a fresh TTS instance
- Add hasReceivedAudio flag to ElevenLabsStreamingTts to detect when
  WebSocket connects but returns no audio (invalid key, network error)
- Fall back to playTtsForText when streaming TTS produced no audio
- Track ttsJob to cleanly cancel prior playTtsForText on new response
- Re-throw CancellationException instead of cascading into fallback
  attempts that also get cancelled

apps/android/app/src/main/java/ai/openclaw/android/MainViewModel.kt

@@ -47,6 +47,7 @@ class MainViewModel(app: Application) : AndroidViewModel(app) {
   val locationPreciseEnabled: StateFlow<Boolean> = runtime.locationPreciseEnabled
   val preventSleep: StateFlow<Boolean> = runtime.preventSleep
   val micEnabled: StateFlow<Boolean> = runtime.micEnabled
+  val micCooldown: StateFlow<Boolean> = runtime.micCooldown
   val micStatusText: StateFlow<String> = runtime.micStatusText
   val micLiveTranscript: StateFlow<String?> = runtime.micLiveTranscript
   val micIsListening: StateFlow<Boolean> = runtime.micIsListening

apps/android/app/src/main/java/ai/openclaw/android/NodeRuntime.kt

@@ -344,8 +344,11 @@ class NodeRuntime(context: Context) {
     MicCaptureManager(
       context = appContext,
       scope = scope,
-      sendToGateway = { message ->
+      sendToGateway = { message, onRunIdKnown ->
         val idempotencyKey = UUID.randomUUID().toString()
+        // Notify MicCaptureManager of the idempotency key *before* the network
+        // call so pendingRunId is set before any chat events can arrive.
+        onRunIdKnown(idempotencyKey)
         val params =
           buildJsonObject {
             put("sessionKey", JsonPrimitive(resolveMainSessionKey()))
@@ -375,6 +378,9 @@ class NodeRuntime(context: Context) {
   val micEnabled: StateFlow<Boolean>
     get() = micCapture.micEnabled
 
+  val micCooldown: StateFlow<Boolean>
+    get() = micCapture.micCooldown
+
   val micQueuedMessages: StateFlow<List<String>>
     get() = micCapture.queuedMessages
 

apps/android/app/src/main/java/ai/openclaw/android/ui/VoiceTabScreen.kt

@@ -80,7 +80,9 @@ fun VoiceTabScreen(viewModel: MainViewModel) {
 
   val gatewayStatus by viewModel.statusText.collectAsState()
   val micEnabled by viewModel.micEnabled.collectAsState()
+  val micCooldown by viewModel.micCooldown.collectAsState()
   val speakerEnabled by viewModel.speakerEnabled.collectAsState()
+  val micStatusText by viewModel.micStatusText.collectAsState()
   val micLiveTranscript by viewModel.micLiveTranscript.collectAsState()
   val micQueuedMessages by viewModel.micQueuedMessages.collectAsState()
   val micConversation by viewModel.micConversation.collectAsState()
@@ -244,6 +246,7 @@ fun VoiceTabScreen(viewModel: MainViewModel) {
           }
           Button(
             onClick = {
+              if (micCooldown) return@Button
               if (micEnabled) {
                 viewModel.setMicEnabled(false)
                 return@Button
@@ -255,13 +258,16 @@ fun VoiceTabScreen(viewModel: MainViewModel) {
                 requestMicPermission.launch(Manifest.permission.RECORD_AUDIO)
               }
             },
+            enabled = !micCooldown,
             shape = CircleShape,
             contentPadding = PaddingValues(0.dp),
             modifier = Modifier.size(60.dp),
             colors =
               ButtonDefaults.buttonColors(
-                containerColor = if (micEnabled) mobileDanger else mobileAccent,
+                containerColor = if (micCooldown) mobileTextSecondary else if (micEnabled) mobileDanger else mobileAccent,
                 contentColor = Color.White,
+                disabledContainerColor = mobileTextSecondary,
+                disabledContentColor = Color.White.copy(alpha = 0.5f),
               ),
           ) {
             Icon(
@@ -282,6 +288,7 @@ fun VoiceTabScreen(viewModel: MainViewModel) {
         when {
           queueCount > 0 -> "$queueCount queued"
           micIsSending -> "Sending"
+          micCooldown -> "Cooldown"
           micEnabled -> "Listening"
           else -> "Mic off"
         }

apps/android/app/src/main/java/ai/openclaw/android/voice/ElevenLabsStreamingTts.kt

@@ -63,6 +63,8 @@ class ElevenLabsStreamingTts(
   private var client: OkHttpClient? = null
   @Volatile private var stopped = false
   @Volatile private var finished = false
+  @Volatile var hasReceivedAudio = false
+    private set
   private var drainJob: Job? = null
 
   // Track text already sent so we only send incremental chunks
@@ -77,6 +79,7 @@ class ElevenLabsStreamingTts(
   fun start() {
     stopped = false
     finished = false
+    hasReceivedAudio = false
     sentTextLength = 0
     trackStarted = false
     wsReady = false
@@ -199,6 +202,7 @@ class ElevenLabsStreamingTts(
   /**
    * Returns true if text was accepted, false if text diverged (caller should restart).
    */
+  @Synchronized
   fun sendText(fullText: String): Boolean {
     if (stopped) return false
     if (finished) return true  // Already finishing — not a diverge, don't restart
@@ -233,6 +237,7 @@ class ElevenLabsStreamingTts(
    * Signal that no more text is coming. Sends EOS to ElevenLabs.
    * The WebSocket will close after generating remaining audio.
    */
+  @Synchronized
   fun finish() {
     if (stopped || finished) return
     finished = true
@@ -278,6 +283,7 @@ class ElevenLabsStreamingTts(
     if (!trackStarted) {
       track.play()
       trackStarted = true
+      hasReceivedAudio = true
       Log.d(TAG, "AudioTrack started on first chunk")
     }
 

apps/android/app/src/main/java/ai/openclaw/android/voice/MicCaptureManager.kt

@@ -11,7 +11,6 @@ import android.util.Log
 import android.speech.RecognitionListener
 import android.speech.RecognizerIntent
 import android.speech.SpeechRecognizer
-import android.util.Log
 import androidx.core.content.ContextCompat
 import java.util.UUID
 import kotlinx.coroutines.CoroutineScope
@@ -40,7 +39,12 @@ data class VoiceConversationEntry(
 class MicCaptureManager(
   private val context: Context,
   private val scope: CoroutineScope,
-  private val sendToGateway: suspend (String) -> String?,
+  /**
+   * Send [message] to the gateway and return the run ID.
+   * [onRunIdKnown] is called with the idempotency key *before* the network
+   * round-trip so [pendingRunId] is set before any chat events can arrive.
+   */
+  private val sendToGateway: suspend (message: String, onRunIdKnown: (String) -> Unit) -> String?,
   private val speakAssistantReply: suspend (String) -> Unit = {},
 ) {
   companion object {
@@ -58,6 +62,9 @@ class MicCaptureManager(
   private val _micEnabled = MutableStateFlow(false)
   val micEnabled: StateFlow<Boolean> = _micEnabled
 
+  private val _micCooldown = MutableStateFlow(false)
+  val micCooldown: StateFlow<Boolean> = _micCooldown
+
   private val _isListening = MutableStateFlow(false)
   val isListening: StateFlow<Boolean> = _isListening
 
@@ -88,6 +95,7 @@ class MicCaptureManager(
 
   private var recognizer: SpeechRecognizer? = null
   private var restartJob: Job? = null
+  private var drainJob: Job? = null
   private var pendingRunTimeoutJob: Job? = null
   private var stopRequested = false
 
@@ -98,8 +106,11 @@ class MicCaptureManager(
       start()
       sendQueuedIfIdle()
     } else {
-      // Give the recognizer time to finish processing buffered audio
-      scope.launch {
+      // Give the recognizer time to finish processing buffered audio.
+      // Cancel any prior drain to prevent duplicate sends on rapid toggle.
+      drainJob?.cancel()
+      _micCooldown.value = true
+      drainJob = scope.launch {
         delay(2000L)
         stop()
         // Capture any partial transcript that didn't get a final result from the recognizer
@@ -108,6 +119,8 @@ class MicCaptureManager(
           sessionSegments.add(partial)
         }
         flushSessionToQueue()
+        drainJob = null
+        _micCooldown.value = false
         sendQueuedIfIdle()
       }
     }
@@ -296,8 +309,13 @@ class MicCaptureManager(
 
     scope.launch {
       try {
-        val runId = sendToGateway(next)
-        pendingRunId = runId
+        val runId = sendToGateway(next) { earlyRunId ->
+          // Called with the idempotency key before chat.send fires so that
+          // pendingRunId is populated before any chat events can arrive.
+          pendingRunId = earlyRunId
+        }
+        // Update to the real runId if the gateway returned a different one.
+        if (runId != null && runId != pendingRunId) pendingRunId = runId
         if (runId == null) {
           pendingRunTimeoutJob?.cancel()
           pendingRunTimeoutJob = null