Fix: make sure audio output enable / disable doesn't need a refresh in order for audio to become audible again

cloud.go

@@ -20,6 +20,7 @@ import (
 
 	"github.com/coder/websocket"
 	"github.com/gin-gonic/gin"
+	"github.com/jetkvm/kvm/internal/audio"
 	"github.com/rs/zerolog"
 )
 
@@ -480,6 +481,16 @@ func handleSessionRequest(
 	cancelKeyboardMacro()
 
 	currentSession = session
+
+	// Set up audio relay callback to get current session's audio track
+	// This is needed for audio output to work after enable/disable cycles
+	audio.SetCurrentSessionCallback(func() audio.AudioTrackWriter {
+		if currentSession != nil {
+			return currentSession.AudioTrack
+		}
+		return nil
+	})
+
 	_ = wsjson.Write(context.Background(), c, gin.H{"type": "answer", "data": sd})
 	return nil
 }

main.go

@@ -21,16 +21,6 @@ var (
 	audioSupervisor  *audio.AudioOutputSupervisor
 )
 
-// runAudioServer is now handled by audio.RunAudioOutputServer
-// This function is kept for backward compatibility but delegates to the audio package
-func runAudioServer() {
-	err := audio.RunAudioOutputServer()
-	if err != nil {
-		logger.Error().Err(err).Msg("audio output server failed")
-		os.Exit(1)
-	}
-}
-
 func startAudioSubprocess() error {
 	// Initialize validation cache for optimal performance
 	audio.InitValidationCache()
@@ -47,14 +37,14 @@ func startAudioSubprocess() error {
 	audio.SetAudioInputSupervisor(audioInputSupervisor)
 
 	// Set default OPUS configuration for audio input supervisor (low quality for single-core RV1106)
-	config := audio.Config
+	audioConfig := audio.Config
 	audioInputSupervisor.SetOpusConfig(
-		config.AudioQualityLowInputBitrate*1000, // Convert kbps to bps
-		config.AudioQualityLowOpusComplexity,
-		config.AudioQualityLowOpusVBR,
-		config.AudioQualityLowOpusSignalType,
-		config.AudioQualityLowOpusBandwidth,
-		config.AudioQualityLowOpusDTX,
+		audioConfig.AudioQualityLowInputBitrate*1000, // Convert kbps to bps
+		audioConfig.AudioQualityLowOpusComplexity,
+		audioConfig.AudioQualityLowOpusVBR,
+		audioConfig.AudioQualityLowOpusSignalType,
+		audioConfig.AudioQualityLowOpusBandwidth,
+		audioConfig.AudioQualityLowOpusDTX,
 	)
 
 	// Note: Audio input supervisor is NOT started here - it will be started on-demand
@@ -110,6 +100,12 @@ func startAudioSubprocess() error {
 		},
 	)
 
+	// Check if USB audio device is enabled before starting audio processes
+	if config.UsbDevices == nil || !config.UsbDevices.Audio {
+		logger.Info().Msg("USB audio device disabled - skipping audio supervisor startup")
+		return nil
+	}
+
 	// Start the supervisor
 	if err := audioSupervisor.Start(); err != nil {
 		return fmt.Errorf("failed to start audio supervisor: %w", err)
@@ -137,7 +133,11 @@ func Main(audioServer bool, audioInputServer bool) {
 
 	// If running as audio server, only initialize audio processing
 	if isAudioServer {
-		runAudioServer()
+		err := audio.RunAudioOutputServer()
+		if err != nil {
+			logger.Error().Err(err).Msg("audio output server failed")
+			os.Exit(1)
+		}
 		return
 	}
 

ui/src/hooks/useMicrophone.ts

@@ -113,14 +113,12 @@ export function useMicrophone() {
     // Debounce sync calls to prevent race conditions
     const now = Date.now();
     if (now - lastSyncRef.current < AUDIO_CONFIG.SYNC_DEBOUNCE_MS) {
-      devLog("Skipping sync - too frequent");
       return;
     }
     lastSyncRef.current = now;
     
     // Don't sync if we're in the middle of starting the microphone
     if (isStartingRef.current) {
-      devLog("Skipping sync - microphone is starting");
       return;
     }
     
@@ -197,7 +195,6 @@ export function useMicrophone() {
         audioConstraints.deviceId = { exact: deviceId };
       }
       
-      devLog("Requesting microphone with constraints:", audioConstraints);
       const stream = await navigator.mediaDevices.getUserMedia({
         audio: audioConstraints
       });
@@ -265,7 +262,6 @@ export function useMicrophone() {
       }
 
       // Notify backend that microphone is started
-      devLog("Notifying backend about microphone start...");
       
             // Retry logic for backend failures
       let backendSuccess = false;
@@ -274,7 +270,6 @@ export function useMicrophone() {
       for (let attempt = 1; attempt <= 3; attempt++) {
         // If this is a retry, first try to reset the backend microphone state
         if (attempt > 1) {
-          devLog(`Backend start attempt ${attempt}, first trying to reset backend state...`);
           try {
             // Use RPC for reset (cloud-compatible)
             if (rpcDataChannel?.readyState === "open") {
@@ -290,7 +285,6 @@ export function useMicrophone() {
                       resolve(); // Continue even if both fail
                     });
                   } else {
-                    devLog("RPC microphone reset successful");
                     resolve();
                   }
                 });
@@ -315,7 +309,6 @@ export function useMicrophone() {
           
           // For RPC errors, try again after a short delay
           if (attempt < 3) {
-            devLog(`Retrying backend start in 500ms (attempt ${attempt + 1}/3)...`);
             await new Promise(resolve => setTimeout(resolve, 500));
             continue;
           }
@@ -556,7 +549,6 @@ export function useMicrophone() {
     const autoRestoreMicrophone = async () => {
       // Wait for RPC connection to be ready before attempting any operations
       if (rpcDataChannel?.readyState !== "open") {
-        devLog("RPC connection not ready for microphone auto-restore, skipping");
         return;
       }
       
@@ -565,7 +557,6 @@ export function useMicrophone() {
       
       // If microphone was enabled before page reload and is not currently active, restore it
       if (microphoneWasEnabled && !isMicrophoneActive && peerConnection) {
-        devLog("Auto-restoring microphone after page reload");
         try {
           const result = await startMicrophone();
           if (result.success) {

web.go

@@ -20,6 +20,7 @@ import (
 	gin_logger "github.com/gin-contrib/logger"
 	"github.com/gin-gonic/gin"
 	"github.com/google/uuid"
+	"github.com/jetkvm/kvm/internal/audio"
 	"github.com/jetkvm/kvm/internal/logging"
 	"github.com/pion/webrtc/v4"
 	"github.com/prometheus/client_golang/prometheus"
@@ -233,6 +234,16 @@ func handleWebRTCSession(c *gin.Context) {
 	cancelKeyboardMacro()
 
 	currentSession = session
+
+	// Set up audio relay callback to get current session's audio track
+	// This is needed for audio output to work after enable/disable cycles
+	audio.SetCurrentSessionCallback(func() audio.AudioTrackWriter {
+		if currentSession != nil {
+			return currentSession.AudioTrack
+		}
+		return nil
+	})
+
 	c.JSON(http.StatusOK, gin.H{"sd": sd})
 }