Fix: go lint errors

.github/workflows/golangci-lint.yml

@@ -27,11 +27,64 @@ jobs:
         uses: actions/setup-go@fa96338abe5531f6e34c5cc0bbe28c1a533d5505 # v4.2.1
         with:
           go-version: 1.24.4
+      - name: Setup build environment variables
+        id: build-env
+        run: |
+          # Extract versions from Makefile
+          ALSA_VERSION=$(grep '^ALSA_VERSION' Makefile | cut -d'=' -f2 | tr -d ' ')
+          OPUS_VERSION=$(grep '^OPUS_VERSION' Makefile | cut -d'=' -f2 | tr -d ' ')
+          
+          # Get rv1106-system latest commit
+          RV1106_COMMIT=$(git ls-remote https://github.com/jetkvm/rv1106-system.git HEAD | cut -f1)
+          
+          # Set environment variables
+          echo "ALSA_VERSION=$ALSA_VERSION" >> $GITHUB_ENV
+          echo "OPUS_VERSION=$OPUS_VERSION" >> $GITHUB_ENV
+          echo "RV1106_COMMIT=$RV1106_COMMIT" >> $GITHUB_ENV
+          
+          # Set outputs for use in other steps
+          echo "alsa_version=$ALSA_VERSION" >> $GITHUB_OUTPUT
+          echo "opus_version=$OPUS_VERSION" >> $GITHUB_OUTPUT
+          echo "rv1106_commit=$RV1106_COMMIT" >> $GITHUB_OUTPUT
+          
+          # Set resolved cache path
+          CACHE_PATH="$HOME/.jetkvm/audio-libs"
+          echo "CACHE_PATH=$CACHE_PATH" >> $GITHUB_ENV
+          echo "cache_path=$CACHE_PATH" >> $GITHUB_OUTPUT
+          
+          echo "Extracted ALSA version: $ALSA_VERSION"
+          echo "Extracted Opus version: $OPUS_VERSION"
+          echo "Latest rv1106-system commit: $RV1106_COMMIT"
+          echo "Cache path: $CACHE_PATH"
+      - name: Restore audio dependencies cache
+        id: cache-audio-deps
+        uses: actions/cache/restore@v4
+        with:
+          path: ${{ steps.build-env.outputs.cache_path }}
+          key: audio-deps-${{ runner.os }}-alsa-${{ steps.build-env.outputs.alsa_version }}-opus-${{ steps.build-env.outputs.opus_version }}-rv1106-${{ steps.build-env.outputs.rv1106_commit }}
+      - name: Setup development environment
+        if: steps.cache-audio-deps.outputs.cache-hit != 'true'
+        run: make dev_env
+        env:
+          ALSA_VERSION: ${{ env.ALSA_VERSION }}
+          OPUS_VERSION: ${{ env.OPUS_VERSION }}
       - name: Create empty resource directory
         run: |
           mkdir -p static && touch static/.gitkeep
+      - name: Save audio dependencies cache
+        if: always() && steps.cache-audio-deps.outputs.cache-hit != 'true'
+        uses: actions/cache/save@v4
+        with:
+          path: ${{ steps.build-env.outputs.cache_path }}
+          key: ${{ steps.cache-audio-deps.outputs.cache-primary-key }}
       - name: Lint
         uses: golangci/golangci-lint-action@1481404843c368bc19ca9406f87d6e0fc97bdcfd # v7.0.0
         with:
           args: --verbose
           version: v2.0.2
+        env:
+          CGO_ENABLED: 1
+          ALSA_VERSION: ${{ env.ALSA_VERSION }}
+          OPUS_VERSION: ${{ env.OPUS_VERSION }}
+          CGO_CFLAGS: "-I${{ steps.build-env.outputs.cache_path }}/alsa-lib-${{ steps.build-env.outputs.alsa_version }}/include -I${{ steps.build-env.outputs.cache_path }}/opus-${{ steps.build-env.outputs.opus_version }}/include -I${{ steps.build-env.outputs.cache_path }}/opus-${{ steps.build-env.outputs.opus_version }}/celt"
+          CGO_LDFLAGS: "-L${{ steps.build-env.outputs.cache_path }}/alsa-lib-${{ steps.build-env.outputs.alsa_version }}/src/.libs -lasound -L${{ steps.build-env.outputs.cache_path }}/opus-${{ steps.build-env.outputs.opus_version }}/.libs -lopus -lm -ldl -static"

cmd/main.go

@@ -11,6 +11,7 @@ import (
 func main() {
 	versionPtr := flag.Bool("version", false, "print version and exit")
 	versionJsonPtr := flag.Bool("version-json", false, "print version as json and exit")
+	audioServerPtr := flag.Bool("audio-server", false, "Run as audio server subprocess")
 	flag.Parse()
 
 	if *versionPtr || *versionJsonPtr {
@@ -23,5 +24,5 @@ func main() {
 		return
 	}
 
-	kvm.Main()
+	kvm.Main(*audioServerPtr)
 }

dev_deploy.sh

@@ -159,8 +159,8 @@ else
 	msg_info "▶ Building development binary"
 	make build_dev
 	
-	# Kill any existing instances of the application
-	ssh "${REMOTE_USER}@${REMOTE_HOST}" "killall jetkvm_app_debug || true"
+	# Kill any existing instances of the application (specific cleanup)
+	ssh "${REMOTE_USER}@${REMOTE_HOST}" "killall jetkvm_app || true; killall jetkvm_native || true; killall jetkvm_app_debug || true; sleep 2"
 	
 	# Copy the binary to the remote host
 	ssh "${REMOTE_USER}@${REMOTE_HOST}" "cat > ${REMOTE_PATH}/jetkvm_app_debug" < bin/jetkvm_app
@@ -180,18 +180,18 @@ set -e
 # Set the library path to include the directory where librockit.so is located
 export LD_LIBRARY_PATH=/oem/usr/lib:\$LD_LIBRARY_PATH
 
-# Check if production jetkvm_app is running and save its state
-PROD_APP_RUNNING=false
-if pgrep -f "/userdata/jetkvm/bin/jetkvm_app" > /dev/null; then
-    PROD_APP_RUNNING=true
-    echo "Production jetkvm_app is running, will restore after development session"
-else
-    echo "No production jetkvm_app detected"
-fi
-
-# Kill any existing instances of the application
-pkill -f "/userdata/jetkvm/bin/jetkvm_app" || true
+# Kill any existing instances of the application (specific cleanup)
+killall jetkvm_app || true
+killall jetkvm_native || true
 killall jetkvm_app_debug || true
+sleep 2
+
+# Verify no processes are using port 80
+if netstat -tlnp | grep :80 > /dev/null 2>&1; then
+    echo "Warning: Port 80 still in use, attempting to free it..."
+    fuser -k 80/tcp || true
+    sleep 1
+fi
 
 # Navigate to the directory where the binary will be stored
 cd "${REMOTE_PATH}"
@@ -199,29 +199,7 @@ cd "${REMOTE_PATH}"
 # Make the new binary executable
 chmod +x jetkvm_app_debug
 
-# Create a cleanup script that will restore the production app
-cat > /tmp/restore_jetkvm.sh << RESTORE_EOF
-#!/bin/ash
-set -e
-export LD_LIBRARY_PATH=/oem/usr/lib:\$LD_LIBRARY_PATH
-cd ${REMOTE_PATH}
-if [ "$PROD_APP_RUNNING" = "true" ]; then
-    echo "Restoring production jetkvm_app..."
-    killall jetkvm_app_debug || true
-    nohup /userdata/jetkvm/bin/jetkvm_app > /tmp/jetkvm_app.log 2>&1 &
-    echo "Production jetkvm_app restored"
-else
-    echo "No production app was running before, not restoring"
-fi
-RESTORE_EOF
-
-chmod +x /tmp/restore_jetkvm.sh
-
-# Set up signal handler to restore production app on exit
-trap '/tmp/restore_jetkvm.sh' EXIT INT TERM
-
-# Run the application in the foreground
-echo "Starting development jetkvm_app_debug..."
+# Run the application in the background
 PION_LOG_TRACE=${LOG_TRACE_SCOPES} ./jetkvm_app_debug | tee -a /tmp/jetkvm_app_debug.log
 EOF
 fi

input_rpc.go

@@ -14,7 +14,7 @@ const (
 // Input RPC Direct Handlers
 // This module provides optimized direct handlers for high-frequency input events,
 // bypassing the reflection-based RPC system for improved performance.
-// 
+//
 // Performance benefits:
 // - Eliminates reflection overhead (~2-3ms per call)
 // - Reduces memory allocations
@@ -214,4 +214,4 @@ func isInputMethod(method string) bool {
 	default:
 		return false
 	}
-}
+}

internal/audio/api.go

@@ -1,13 +1,51 @@
 package audio
 
-// StartAudioStreaming launches the in-process audio stream and delivers Opus frames to the provided callback.
-// This is now a wrapper around the non-blocking audio implementation for backward compatibility.
-func StartAudioStreaming(send func([]byte)) error {
-	return StartNonBlockingAudioStreaming(send)
+import (
+	"os"
+	"strings"
+)
+
+// isAudioServerProcess detects if we're running as the audio server subprocess
+func isAudioServerProcess() bool {
+	for _, arg := range os.Args {
+		if strings.Contains(arg, "--audio-server") {
+			return true
+		}
+	}
+	return false
 }
 
-// StopAudioStreaming stops the in-process audio stream.
-// This is now a wrapper around the non-blocking audio implementation for backward compatibility.
-func StopAudioStreaming() {
-	StopNonBlockingAudioStreaming()
+// StartAudioStreaming launches the audio stream.
+// In audio server subprocess: uses CGO-based audio streaming
+// In main process: this should not be called (use StartAudioRelay instead)
+func StartAudioStreaming(send func([]byte)) error {
+	if isAudioServerProcess() {
+		// Audio server subprocess: use CGO audio processing
+		return StartAudioOutputStreaming(send)
+	} else {
+		// Main process: should use relay system instead
+		// This is kept for backward compatibility but not recommended
+		return StartAudioOutputStreaming(send)
+	}
+}
+
+// StopAudioStreaming stops the audio stream.
+func StopAudioStreaming() {
+	if isAudioServerProcess() {
+		// Audio server subprocess: stop CGO audio processing
+		StopAudioOutputStreaming()
+	} else {
+		// Main process: stop relay if running
+		StopAudioRelay()
+	}
+}
+
+// StartNonBlockingAudioStreaming is an alias for backward compatibility
+func StartNonBlockingAudioStreaming(send func([]byte)) error {
+	return StartAudioOutputStreaming(send)
+}
+
+// StopNonBlockingAudioStreaming is an alias for backward compatibility
+func StopNonBlockingAudioStreaming() {
+	StopAudioOutputStreaming()
 }

internal/audio/batch_audio.go

@@ -28,27 +28,23 @@ type BatchAudioProcessor struct {
 
 	// Batch queues and state (atomic for lock-free access)
 	readQueue    chan batchReadRequest
-	writeQueue   chan batchWriteRequest
 	initialized  int32
 	running      int32
 	threadPinned int32
 
 	// Buffers (pre-allocated to avoid allocation overhead)
-	readBufPool  *sync.Pool
-	writeBufPool *sync.Pool
+	readBufPool *sync.Pool
 }
 
 type BatchAudioStats struct {
 	// int64 fields MUST be first for ARM32 alignment
-	BatchedReads      int64
-	BatchedWrites     int64
-	SingleReads       int64
-	SingleWrites      int64
-	BatchedFrames     int64
-	SingleFrames      int64
-	CGOCallsReduced   int64
-	OSThreadPinTime   time.Duration // time.Duration is int64 internally
-	LastBatchTime     time.Time
+	BatchedReads    int64
+	SingleReads     int64
+	BatchedFrames   int64
+	SingleFrames    int64
+	CGOCallsReduced int64
+	OSThreadPinTime time.Duration // time.Duration is int64 internally
+	LastBatchTime   time.Time
 }
 
 type batchReadRequest struct {
@@ -57,22 +53,11 @@ type batchReadRequest struct {
 	timestamp  time.Time
 }
 
-type batchWriteRequest struct {
-	buffer     []byte
-	resultChan chan batchWriteResult
-	timestamp  time.Time
-}
-
 type batchReadResult struct {
 	length int
 	err    error
 }
 
-type batchWriteResult struct {
-	written int
-	err     error
-}
-
 // NewBatchAudioProcessor creates a new batch audio processor
 func NewBatchAudioProcessor(batchSize int, batchDuration time.Duration) *BatchAudioProcessor {
 	ctx, cancel := context.WithCancel(context.Background())
@@ -85,17 +70,11 @@ func NewBatchAudioProcessor(batchSize int, batchDuration time.Duration) *BatchAu
 		batchSize:     batchSize,
 		batchDuration: batchDuration,
 		readQueue:     make(chan batchReadRequest, batchSize*2),
-		writeQueue:    make(chan batchWriteRequest, batchSize*2),
 		readBufPool: &sync.Pool{
 			New: func() interface{} {
 				return make([]byte, 1500) // Max audio frame size
 			},
 		},
-		writeBufPool: &sync.Pool{
-			New: func() interface{} {
-				return make([]byte, 4096) // Max write buffer size
-			},
-		},
 	}
 
 	return processor
@@ -114,7 +93,6 @@ func (bap *BatchAudioProcessor) Start() error {
 
 	// Start batch processing goroutines
 	go bap.batchReadProcessor()
-	go bap.batchWriteProcessor()
 
 	bap.logger.Info().Int("batch_size", bap.batchSize).
 		Dur("batch_duration", bap.batchDuration).
@@ -175,44 +153,6 @@ func (bap *BatchAudioProcessor) BatchReadEncode(buffer []byte) (int, error) {
 	}
 }
 
-// BatchDecodeWrite performs batched audio decode and write operations
-func (bap *BatchAudioProcessor) BatchDecodeWrite(buffer []byte) (int, error) {
-	if atomic.LoadInt32(&bap.running) == 0 {
-		// Fallback to single operation if batch processor is not running
-		atomic.AddInt64(&bap.stats.SingleWrites, 1)
-		atomic.AddInt64(&bap.stats.SingleFrames, 1)
-		return CGOAudioDecodeWrite(buffer)
-	}
-
-	resultChan := make(chan batchWriteResult, 1)
-	request := batchWriteRequest{
-		buffer:     buffer,
-		resultChan: resultChan,
-		timestamp:  time.Now(),
-	}
-
-	select {
-	case bap.writeQueue <- request:
-		// Successfully queued
-	case <-time.After(5 * time.Millisecond):
-		// Queue is full or blocked, fallback to single operation
-		atomic.AddInt64(&bap.stats.SingleWrites, 1)
-		atomic.AddInt64(&bap.stats.SingleFrames, 1)
-		return CGOAudioDecodeWrite(buffer)
-	}
-
-	// Wait for result
-	select {
-	case result := <-resultChan:
-		return result.written, result.err
-	case <-time.After(50 * time.Millisecond):
-		// Timeout, fallback to single operation
-		atomic.AddInt64(&bap.stats.SingleWrites, 1)
-		atomic.AddInt64(&bap.stats.SingleFrames, 1)
-		return CGOAudioDecodeWrite(buffer)
-	}
-}
-
 // batchReadProcessor processes batched read operations
 func (bap *BatchAudioProcessor) batchReadProcessor() {
 	defer bap.logger.Debug().Msg("batch read processor stopped")
@@ -249,42 +189,6 @@ func (bap *BatchAudioProcessor) batchReadProcessor() {
 	}
 }
 
-// batchWriteProcessor processes batched write operations
-func (bap *BatchAudioProcessor) batchWriteProcessor() {
-	defer bap.logger.Debug().Msg("batch write processor stopped")
-
-	ticker := time.NewTicker(bap.batchDuration)
-	defer ticker.Stop()
-
-	var batch []batchWriteRequest
-	batch = make([]batchWriteRequest, 0, bap.batchSize)
-
-	for atomic.LoadInt32(&bap.running) == 1 {
-		select {
-		case <-bap.ctx.Done():
-			return
-
-		case req := <-bap.writeQueue:
-			batch = append(batch, req)
-			if len(batch) >= bap.batchSize {
-				bap.processBatchWrite(batch)
-				batch = batch[:0] // Clear slice but keep capacity
-			}
-
-		case <-ticker.C:
-			if len(batch) > 0 {
-				bap.processBatchWrite(batch)
-				batch = batch[:0] // Clear slice but keep capacity
-			}
-		}
-	}
-
-	// Process any remaining requests
-	if len(batch) > 0 {
-		bap.processBatchWrite(batch)
-	}
-}
-
 // processBatchRead processes a batch of read requests efficiently
 func (bap *BatchAudioProcessor) processBatchRead(batch []batchReadRequest) {
 	if len(batch) == 0 {
@@ -328,56 +232,11 @@ func (bap *BatchAudioProcessor) processBatchRead(batch []batchReadRequest) {
 	bap.stats.LastBatchTime = time.Now()
 }
 
-// processBatchWrite processes a batch of write requests efficiently
-func (bap *BatchAudioProcessor) processBatchWrite(batch []batchWriteRequest) {
-	if len(batch) == 0 {
-		return
-	}
-
-	// Pin to OS thread for the entire batch to minimize thread switching overhead
-	start := time.Now()
-	if atomic.CompareAndSwapInt32(&bap.threadPinned, 0, 1) {
-		runtime.LockOSThread()
-		defer func() {
-			runtime.UnlockOSThread()
-			atomic.StoreInt32(&bap.threadPinned, 0)
-			bap.stats.OSThreadPinTime += time.Since(start)
-		}()
-	}
-
-	batchSize := len(batch)
-	atomic.AddInt64(&bap.stats.BatchedWrites, 1)
-	atomic.AddInt64(&bap.stats.BatchedFrames, int64(batchSize))
-	if batchSize > 1 {
-		atomic.AddInt64(&bap.stats.CGOCallsReduced, int64(batchSize-1))
-	}
-
-	// Process each request in the batch
-	for _, req := range batch {
-		written, err := CGOAudioDecodeWrite(req.buffer)
-		result := batchWriteResult{
-			written: written,
-			err:     err,
-		}
-
-		// Send result back (non-blocking)
-		select {
-		case req.resultChan <- result:
-		default:
-			// Requestor timed out, drop result
-		}
-	}
-
-	bap.stats.LastBatchTime = time.Now()
-}
-
 // GetStats returns current batch processor statistics
 func (bap *BatchAudioProcessor) GetStats() BatchAudioStats {
 	return BatchAudioStats{
 		BatchedReads:    atomic.LoadInt64(&bap.stats.BatchedReads),
-		BatchedWrites:   atomic.LoadInt64(&bap.stats.BatchedWrites),
 		SingleReads:     atomic.LoadInt64(&bap.stats.SingleReads),
-		SingleWrites:    atomic.LoadInt64(&bap.stats.SingleWrites),
 		BatchedFrames:   atomic.LoadInt64(&bap.stats.BatchedFrames),
 		SingleFrames:    atomic.LoadInt64(&bap.stats.SingleFrames),
 		CGOCallsReduced: atomic.LoadInt64(&bap.stats.CGOCallsReduced),
@@ -393,7 +252,7 @@ func (bap *BatchAudioProcessor) IsRunning() bool {
 
 // Global batch processor instance
 var (
-	globalBatchProcessor     unsafe.Pointer // *BatchAudioProcessor
+	globalBatchProcessor      unsafe.Pointer // *BatchAudioProcessor
 	batchProcessorInitialized int32
 )
 
@@ -444,12 +303,3 @@ func BatchCGOAudioReadEncode(buffer []byte) (int, error) {
 	}
 	return CGOAudioReadEncode(buffer)
 }
-
-// BatchCGOAudioDecodeWrite is a batched version of CGOAudioDecodeWrite
-func BatchCGOAudioDecodeWrite(buffer []byte) (int, error) {
-	processor := GetBatchAudioProcessor()
-	if processor != nil && processor.IsRunning() {
-		return processor.BatchDecodeWrite(buffer)
-	}
-	return CGOAudioDecodeWrite(buffer)
-}

internal/audio/buffer_pool.go

@@ -23,14 +23,18 @@ func NewAudioBufferPool(bufferSize int) *AudioBufferPool {
 
 // Get retrieves a buffer from the pool
 func (p *AudioBufferPool) Get() []byte {
-	return p.pool.Get().([]byte)
+	if buf := p.pool.Get(); buf != nil {
+		return *buf.(*[]byte)
+	}
+	return make([]byte, 0, 1500) // fallback if pool is empty
 }
 
 // Put returns a buffer to the pool
 func (p *AudioBufferPool) Put(buf []byte) {
 	// Reset length but keep capacity for reuse
 	if cap(buf) >= 1500 { // Only pool buffers of reasonable size
-		p.pool.Put(buf[:0])
+		resetBuf := buf[:0]
+		p.pool.Put(&resetBuf)
 	}
 }
 
@@ -38,7 +42,7 @@ func (p *AudioBufferPool) Put(buf []byte) {
 var (
 	// Pool for 1500-byte audio frame buffers (Opus max frame size)
 	audioFramePool = NewAudioBufferPool(1500)
-	
+
 	// Pool for smaller control buffers
 	audioControlPool = NewAudioBufferPool(64)
 )
@@ -61,4 +65,4 @@ func GetAudioControlBuffer() []byte {
 // PutAudioControlBuffer returns a buffer to the control pool
 func PutAudioControlBuffer(buf []byte) {
 	audioControlPool.Put(buf)
-}
+}

internal/audio/cgo_audio.go

@@ -8,7 +8,7 @@ import (
 )
 
 /*
-#cgo CFLAGS: -I${SRCDIR}/../../tools/alsa-opus-includes
+#cgo CFLAGS: -I$HOME/.jetkvm/audio-libs/alsa-lib-$ALSA_VERSION/include -I$HOME/.jetkvm/audio-libs/opus-$OPUS_VERSION/include -I$HOME/.jetkvm/audio-libs/opus-$OPUS_VERSION/celt
 #cgo LDFLAGS: -L$HOME/.jetkvm/audio-libs/alsa-lib-$ALSA_VERSION/src/.libs -lasound -L$HOME/.jetkvm/audio-libs/opus-$OPUS_VERSION/.libs -lopus -lm -ldl -static
 #include <alsa/asoundlib.h>
 #include <opus.h>
@@ -39,7 +39,7 @@ static volatile int playback_initialized = 0;
 static int safe_alsa_open(snd_pcm_t **handle, const char *device, snd_pcm_stream_t stream) {
 	int attempts = 3;
 	int err;
-	
+
 	while (attempts-- > 0) {
 		err = snd_pcm_open(handle, device, stream, SND_PCM_NONBLOCK);
 		if (err >= 0) {
@@ -47,7 +47,7 @@ static int safe_alsa_open(snd_pcm_t **handle, const char *device, snd_pcm_stream
 			snd_pcm_nonblock(*handle, 0);
 			return 0;
 		}
-		
+
 		if (err == -EBUSY && attempts > 0) {
 			// Device busy, wait and retry
 			usleep(50000); // 50ms
@@ -63,26 +63,26 @@ static int configure_alsa_device(snd_pcm_t *handle, const char *device_name) {
 	snd_pcm_hw_params_t *params;
 	snd_pcm_sw_params_t *sw_params;
 	int err;
-	
+
 	if (!handle) return -1;
-	
+
 	// Use stack allocation for better performance
 	snd_pcm_hw_params_alloca(&params);
 	snd_pcm_sw_params_alloca(&sw_params);
-	
+
 	// Hardware parameters
 	err = snd_pcm_hw_params_any(handle, params);
 	if (err < 0) return err;
-	
+
 	err = snd_pcm_hw_params_set_access(handle, params, SND_PCM_ACCESS_RW_INTERLEAVED);
 	if (err < 0) return err;
-	
+
 	err = snd_pcm_hw_params_set_format(handle, params, SND_PCM_FORMAT_S16_LE);
 	if (err < 0) return err;
-	
+
 	err = snd_pcm_hw_params_set_channels(handle, params, channels);
 	if (err < 0) return err;
-	
+
 	// Set exact rate for better performance
 	err = snd_pcm_hw_params_set_rate(handle, params, sample_rate, 0);
 	if (err < 0) {
@@ -91,70 +91,70 @@ static int configure_alsa_device(snd_pcm_t *handle, const char *device_name) {
 		err = snd_pcm_hw_params_set_rate_near(handle, params, &rate, 0);
 		if (err < 0) return err;
 	}
-	
+
 	// Optimize buffer sizes for low latency
 	snd_pcm_uframes_t period_size = frame_size;
 	err = snd_pcm_hw_params_set_period_size_near(handle, params, &period_size, 0);
 	if (err < 0) return err;
-	
+
 	// Set buffer size to 4 periods for good latency/stability balance
 	snd_pcm_uframes_t buffer_size = period_size * 4;
 	err = snd_pcm_hw_params_set_buffer_size_near(handle, params, &buffer_size);
 	if (err < 0) return err;
-	
+
 	err = snd_pcm_hw_params(handle, params);
 	if (err < 0) return err;
-	
+
 	// Software parameters for optimal performance
 	err = snd_pcm_sw_params_current(handle, sw_params);
 	if (err < 0) return err;
-	
+
 	// Start playback/capture when buffer is period_size frames
 	err = snd_pcm_sw_params_set_start_threshold(handle, sw_params, period_size);
 	if (err < 0) return err;
-	
+
 	// Allow transfers when at least period_size frames are available
 	err = snd_pcm_sw_params_set_avail_min(handle, sw_params, period_size);
 	if (err < 0) return err;
-	
+
 	err = snd_pcm_sw_params(handle, sw_params);
 	if (err < 0) return err;
-	
+
 	return snd_pcm_prepare(handle);
 }
 
 // Initialize ALSA and Opus encoder with improved safety
 int jetkvm_audio_init() {
 	int err;
-	
+
 	// Prevent concurrent initialization
 	if (__sync_bool_compare_and_swap(&capture_initializing, 0, 1) == 0) {
 		return -EBUSY; // Already initializing
 	}
-	
+
 	// Check if already initialized
 	if (capture_initialized) {
 		capture_initializing = 0;
 		return 0;
 	}
-	
+
 	// Clean up any existing resources first
-	if (encoder) { 
-		opus_encoder_destroy(encoder); 
-		encoder = NULL; 
+	if (encoder) {
+		opus_encoder_destroy(encoder);
+		encoder = NULL;
 	}
-	if (pcm_handle) { 
-		snd_pcm_close(pcm_handle); 
-		pcm_handle = NULL; 
+	if (pcm_handle) {
+		snd_pcm_close(pcm_handle);
+		pcm_handle = NULL;
 	}
-	
+
 	// Try to open ALSA capture device
 	err = safe_alsa_open(&pcm_handle, "hw:1,0", SND_PCM_STREAM_CAPTURE);
 	if (err < 0) {
 		capture_initializing = 0;
 		return -1;
 	}
-	
+
 	// Configure the device
 	err = configure_alsa_device(pcm_handle, "capture");
 	if (err < 0) {
@@ -163,7 +163,7 @@ int jetkvm_audio_init() {
 		capture_initializing = 0;
 		return -1;
 	}
-	
+
 	// Initialize Opus encoder
 	int opus_err = 0;
 	encoder = opus_encoder_create(sample_rate, channels, OPUS_APPLICATION_AUDIO, &opus_err);
@@ -172,10 +172,10 @@ int jetkvm_audio_init() {
 		capture_initializing = 0;
 		return -2;
 	}
-	
+
 	opus_encoder_ctl(encoder, OPUS_SET_BITRATE(opus_bitrate));
 	opus_encoder_ctl(encoder, OPUS_SET_COMPLEXITY(opus_complexity));
-	
+
 	capture_initialized = 1;
 	capture_initializing = 0;
 	return 0;
@@ -186,21 +186,21 @@ int jetkvm_audio_read_encode(void *opus_buf) {
 	short pcm_buffer[1920]; // max 2ch*960
 	unsigned char *out = (unsigned char*)opus_buf;
 	int err = 0;
-	
+
 	// Safety checks
 	if (!capture_initialized || !pcm_handle || !encoder || !opus_buf) {
 		return -1;
 	}
-	
+
 	int pcm_rc = snd_pcm_readi(pcm_handle, pcm_buffer, frame_size);
-	
+
 	// Handle ALSA errors with enhanced recovery
 	if (pcm_rc < 0) {
 		if (pcm_rc == -EPIPE) {
 			// Buffer underrun - try to recover
 			err = snd_pcm_prepare(pcm_handle);
 			if (err < 0) return -1;
-			
+
 			pcm_rc = snd_pcm_readi(pcm_handle, pcm_buffer, frame_size);
 			if (pcm_rc < 0) return -1;
 		} else if (pcm_rc == -EAGAIN) {
@@ -221,12 +221,12 @@ int jetkvm_audio_read_encode(void *opus_buf) {
 			return -1;
 		}
 	}
-	
+
 	// If we got fewer frames than expected, pad with silence
 	if (pcm_rc < frame_size) {
 		memset(&pcm_buffer[pcm_rc * channels], 0, (frame_size - pcm_rc) * channels * sizeof(short));
 	}
-	
+
 	int nb_bytes = opus_encode(encoder, pcm_buffer, frame_size, out, max_packet_size);
 	return nb_bytes;
 }
@@ -234,28 +234,28 @@ int jetkvm_audio_read_encode(void *opus_buf) {
 // Initialize ALSA playback with improved safety
 int jetkvm_audio_playback_init() {
 	int err;
-	
+
 	// Prevent concurrent initialization
 	if (__sync_bool_compare_and_swap(&playback_initializing, 0, 1) == 0) {
 		return -EBUSY; // Already initializing
 	}
-	
+
 	// Check if already initialized
 	if (playback_initialized) {
 		playback_initializing = 0;
 		return 0;
 	}
-	
+
 	// Clean up any existing resources first
-	if (decoder) { 
-		opus_decoder_destroy(decoder); 
-		decoder = NULL; 
+	if (decoder) {
+		opus_decoder_destroy(decoder);
+		decoder = NULL;
 	}
-	if (pcm_playback_handle) { 
-		snd_pcm_close(pcm_playback_handle); 
-		pcm_playback_handle = NULL; 
+	if (pcm_playback_handle) {
+		snd_pcm_close(pcm_playback_handle);
+		pcm_playback_handle = NULL;
 	}
-	
+
 	// Try to open the USB gadget audio device for playback
 	err = safe_alsa_open(&pcm_playback_handle, "hw:1,0", SND_PCM_STREAM_PLAYBACK);
 	if (err < 0) {
@@ -266,7 +266,7 @@ int jetkvm_audio_playback_init() {
 			return -1;
 		}
 	}
-	
+
 	// Configure the device
 	err = configure_alsa_device(pcm_playback_handle, "playback");
 	if (err < 0) {
@@ -275,7 +275,7 @@ int jetkvm_audio_playback_init() {
 		playback_initializing = 0;
 		return -1;
 	}
-	
+
 	// Initialize Opus decoder
 	int opus_err = 0;
 	decoder = opus_decoder_create(sample_rate, channels, &opus_err);
@@ -285,7 +285,7 @@ int jetkvm_audio_playback_init() {
 		playback_initializing = 0;
 		return -2;
 	}
-	
+
 	playback_initialized = 1;
 	playback_initializing = 0;
 	return 0;
@@ -296,21 +296,21 @@ int jetkvm_audio_decode_write(void *opus_buf, int opus_size) {
 	short pcm_buffer[1920]; // max 2ch*960
 	unsigned char *in = (unsigned char*)opus_buf;
 	int err = 0;
-	
+
 	// Safety checks
 	if (!playback_initialized || !pcm_playback_handle || !decoder || !opus_buf || opus_size <= 0) {
 		return -1;
 	}
-	
+
 	// Additional bounds checking
 	if (opus_size > max_packet_size) {
 		return -1;
 	}
-	
+
 	// Decode Opus to PCM
 	int pcm_frames = opus_decode(decoder, in, opus_size, pcm_buffer, frame_size, 0);
 	if (pcm_frames < 0) return -1;
-	
+
 	// Write PCM to playback device with enhanced recovery
 	int pcm_rc = snd_pcm_writei(pcm_playback_handle, pcm_buffer, pcm_frames);
 	if (pcm_rc < 0) {
@@ -318,7 +318,7 @@ int jetkvm_audio_decode_write(void *opus_buf, int opus_size) {
 			// Buffer underrun - try to recover
 			err = snd_pcm_prepare(pcm_playback_handle);
 			if (err < 0) return -2;
-			
+
 			pcm_rc = snd_pcm_writei(pcm_playback_handle, pcm_buffer, pcm_frames);
 		} else if (pcm_rc == -ESTRPIPE) {
 			// Device suspended, try to resume
@@ -333,7 +333,7 @@ int jetkvm_audio_decode_write(void *opus_buf, int opus_size) {
 		}
 		if (pcm_rc < 0) return -2;
 	}
-	
+
 	return pcm_frames;
 }
 
@@ -343,20 +343,20 @@ void jetkvm_audio_playback_close() {
 	while (playback_initializing) {
 		usleep(1000); // 1ms
 	}
-	
+
 	// Atomic check and set to prevent double cleanup
 	if (__sync_bool_compare_and_swap(&playback_initialized, 1, 0) == 0) {
 		return; // Already cleaned up
 	}
-	
-	if (decoder) { 
-		opus_decoder_destroy(decoder); 
-		decoder = NULL; 
+
+	if (decoder) {
+		opus_decoder_destroy(decoder);
+		decoder = NULL;
 	}
-	if (pcm_playback_handle) { 
+	if (pcm_playback_handle) {
 		snd_pcm_drain(pcm_playback_handle);
-		snd_pcm_close(pcm_playback_handle); 
-		pcm_playback_handle = NULL; 
+		snd_pcm_close(pcm_playback_handle);
+		pcm_playback_handle = NULL;
 	}
 }
 
@@ -366,19 +366,19 @@ void jetkvm_audio_close() {
 	while (capture_initializing) {
 		usleep(1000); // 1ms
 	}
-	
+
 	capture_initialized = 0;
-	
-	if (encoder) { 
-		opus_encoder_destroy(encoder); 
-		encoder = NULL; 
+
+	if (encoder) {
+		opus_encoder_destroy(encoder);
+		encoder = NULL;
 	}
-	if (pcm_handle) { 
+	if (pcm_handle) {
 		snd_pcm_drop(pcm_handle); // Drop pending samples
-		snd_pcm_close(pcm_handle); 
-		pcm_handle = NULL; 
+		snd_pcm_close(pcm_handle);
+		pcm_handle = NULL;
 	}
-	
+
 	// Also clean up playback
 	jetkvm_audio_playback_close();
 }
@@ -387,15 +387,15 @@ import "C"
 
 // Optimized Go wrappers with reduced overhead
 var (
-	errAudioInitFailed    = errors.New("failed to init ALSA/Opus")
-	errBufferTooSmall     = errors.New("buffer too small")
-	errAudioReadEncode    = errors.New("audio read/encode error")
-	errAudioDecodeWrite   = errors.New("audio decode/write error")
-	errAudioPlaybackInit  = errors.New("failed to init ALSA playback/Opus decoder")
-	errEmptyBuffer        = errors.New("empty buffer")
-	errNilBuffer          = errors.New("nil buffer")
-	errBufferTooLarge     = errors.New("buffer too large")
-	errInvalidBufferPtr   = errors.New("invalid buffer pointer")
+	errAudioInitFailed   = errors.New("failed to init ALSA/Opus")
+	errBufferTooSmall    = errors.New("buffer too small")
+	errAudioReadEncode   = errors.New("audio read/encode error")
+	errAudioDecodeWrite  = errors.New("audio decode/write error")
+	errAudioPlaybackInit = errors.New("failed to init ALSA playback/Opus decoder")
+	errEmptyBuffer       = errors.New("empty buffer")
+	errNilBuffer         = errors.New("nil buffer")
+	errBufferTooLarge    = errors.New("buffer too large")
+	errInvalidBufferPtr  = errors.New("invalid buffer pointer")
 )
 
 func cgoAudioInit() error {
@@ -416,7 +416,7 @@ func cgoAudioReadEncode(buf []byte) (int, error) {
 	if len(buf) < 1276 {
 		return 0, errBufferTooSmall
 	}
-	
+
 	n := C.jetkvm_audio_read_encode(unsafe.Pointer(&buf[0]))
 	if n < 0 {
 		return 0, errAudioReadEncode
@@ -449,26 +449,27 @@ func cgoAudioDecodeWrite(buf []byte) (int, error) {
 	if buf == nil {
 		return 0, errors.New("nil buffer")
 	}
-	
+
 	// Validate buffer size to prevent potential overruns
 	if len(buf) > 4096 { // Maximum reasonable Opus frame size
 		return 0, errors.New("buffer too large")
 	}
-	
+
 	// Ensure buffer is not deallocated by keeping a reference
 	bufPtr := unsafe.Pointer(&buf[0])
 	if bufPtr == nil {
 		return 0, errors.New("invalid buffer pointer")
 	}
-	
+
 	// Add recovery mechanism for C function crashes
 	defer func() {
 		if r := recover(); r != nil {
 			// Log the panic but don't crash the entire program
 			// This should not happen with proper validation, but provides safety
+			_ = r // Explicitly ignore the panic value
 		}
 	}()
-	
+
 	n := C.jetkvm_audio_decode_write(bufPtr, C.int(len(buf)))
 	if n < 0 {
 		return 0, errors.New("audio decode/write error")
@@ -478,10 +479,10 @@ func cgoAudioDecodeWrite(buf []byte) (int, error) {
 
 // Wrapper functions for non-blocking audio manager
 var (
-	CGOAudioInit           = cgoAudioInit
-	CGOAudioClose          = cgoAudioClose
-	CGOAudioReadEncode     = cgoAudioReadEncode
-	CGOAudioPlaybackInit   = cgoAudioPlaybackInit
-	CGOAudioPlaybackClose  = cgoAudioPlaybackClose
-	CGOAudioDecodeWrite    = cgoAudioDecodeWrite
+	CGOAudioInit          = cgoAudioInit
+	CGOAudioClose         = cgoAudioClose
+	CGOAudioReadEncode    = cgoAudioReadEncode
+	CGOAudioPlaybackInit  = cgoAudioPlaybackInit
+	CGOAudioPlaybackClose = cgoAudioPlaybackClose
+	CGOAudioDecodeWrite   = cgoAudioDecodeWrite
 )

internal/audio/events.go

@@ -2,6 +2,7 @@ package audio
 
 import (
 	"context"
+	"fmt"
 	"strings"
 	"sync"
 	"time"
@@ -249,13 +250,13 @@ func (aeb *AudioEventBroadcaster) startMetricsBroadcasting() {
 	for range ticker.C {
 		aeb.mutex.RLock()
 		subscriberCount := len(aeb.subscribers)
-		
+
 		// Early exit if no subscribers to save CPU
 		if subscriberCount == 0 {
 			aeb.mutex.RUnlock()
 			continue
 		}
-		
+
 		// Create a copy for safe iteration
 		subscribersCopy := make([]*AudioEventSubscriber, 0, subscriberCount)
 		for _, sub := range aeb.subscribers {
@@ -270,7 +271,7 @@ func (aeb *AudioEventBroadcaster) startMetricsBroadcasting() {
 				activeSubscribers++
 			}
 		}
-		
+
 		// Skip metrics gathering if no active subscribers
 		if activeSubscribers == 0 {
 			continue
@@ -286,7 +287,7 @@ func (aeb *AudioEventBroadcaster) startMetricsBroadcasting() {
 				BytesProcessed:  audioMetrics.BytesProcessed,
 				LastFrameTime:   audioMetrics.LastFrameTime.Format("2006-01-02T15:04:05.000Z"),
 				ConnectionDrops: audioMetrics.ConnectionDrops,
-				AverageLatency:  audioMetrics.AverageLatency.String(),
+				AverageLatency:  fmt.Sprintf("%.1fms", float64(audioMetrics.AverageLatency.Nanoseconds())/1e6),
 			},
 		}
 		aeb.broadcast(audioMetricsEvent)
@@ -304,7 +305,7 @@ func (aeb *AudioEventBroadcaster) startMetricsBroadcasting() {
 						BytesProcessed:  micMetrics.BytesProcessed,
 						LastFrameTime:   micMetrics.LastFrameTime.Format("2006-01-02T15:04:05.000Z"),
 						ConnectionDrops: micMetrics.ConnectionDrops,
-						AverageLatency:  micMetrics.AverageLatency.String(),
+						AverageLatency:  fmt.Sprintf("%.1fms", float64(micMetrics.AverageLatency.Nanoseconds())/1e6),
 					},
 				}
 				aeb.broadcast(micMetricsEvent)
@@ -357,9 +358,9 @@ func (aeb *AudioEventBroadcaster) sendToSubscriber(subscriber *AudioEventSubscri
 	err := wsjson.Write(ctx, subscriber.conn, event)
 	if err != nil {
 		// Don't log network errors for closed connections as warnings, they're expected
-		if strings.Contains(err.Error(), "use of closed network connection") || 
-		   strings.Contains(err.Error(), "connection reset by peer") ||
-		   strings.Contains(err.Error(), "context canceled") {
+		if strings.Contains(err.Error(), "use of closed network connection") ||
+			strings.Contains(err.Error(), "connection reset by peer") ||
+			strings.Contains(err.Error(), "context canceled") {
 			subscriber.logger.Debug().Err(err).Msg("websocket connection closed during audio event send")
 		} else {
 			subscriber.logger.Warn().Err(err).Msg("failed to send audio event to subscriber")

internal/audio/input.go

@@ -19,21 +19,21 @@ type AudioInputMetrics struct {
 	LastFrameTime   time.Time
 }
 
-// AudioInputManager manages microphone input stream from WebRTC to USB gadget
+// AudioInputManager manages microphone input stream using IPC mode only
 type AudioInputManager struct {
 	// metrics MUST be first for ARM32 alignment (contains int64 fields)
 	metrics AudioInputMetrics
 
-	inputBuffer chan []byte
-	logger      zerolog.Logger
-	running     int32
+	ipcManager *AudioInputIPCManager
+	logger     zerolog.Logger
+	running    int32
 }
 
-// NewAudioInputManager creates a new audio input manager
+// NewAudioInputManager creates a new audio input manager (IPC mode only)
 func NewAudioInputManager() *AudioInputManager {
 	return &AudioInputManager{
-		inputBuffer: make(chan []byte, 100), // Buffer up to 100 frames
-		logger:      logging.GetDefaultLogger().With().Str("component", "audio-input").Logger(),
+		ipcManager: NewAudioInputIPCManager(),
+		logger:     logging.GetDefaultLogger().With().Str("component", "audio-input").Logger(),
 	}
 }
 
@@ -45,9 +45,10 @@ func (aim *AudioInputManager) Start() error {
 
 	aim.logger.Info().Msg("Starting audio input manager")
 
-	// Start the non-blocking audio input stream
-	err := StartNonBlockingAudioInput(aim.inputBuffer)
+	// Start the IPC-based audio input
+	err := aim.ipcManager.Start()
 	if err != nil {
+		aim.logger.Error().Err(err).Msg("Failed to start IPC audio input")
 		atomic.StoreInt32(&aim.running, 0)
 		return err
 	}
@@ -63,57 +64,114 @@ func (aim *AudioInputManager) Stop() {
 
 	aim.logger.Info().Msg("Stopping audio input manager")
 
-	// Stop the non-blocking audio input stream
-	StopNonBlockingAudioInput()
-
-	// Drain the input buffer
-	go func() {
-		for {
-			select {
-			case <-aim.inputBuffer:
-				// Drain
-			case <-time.After(100 * time.Millisecond):
-				return
-			}
-		}
-	}()
+	// Stop the IPC-based audio input
+	aim.ipcManager.Stop()
 
 	aim.logger.Info().Msg("Audio input manager stopped")
 }
 
-// WriteOpusFrame writes an Opus frame to the input buffer
+// WriteOpusFrame writes an Opus frame to the audio input system with latency tracking
 func (aim *AudioInputManager) WriteOpusFrame(frame []byte) error {
-	if atomic.LoadInt32(&aim.running) == 0 {
-		return nil // Not running, ignore
+	if !aim.IsRunning() {
+		return nil // Not running, silently drop
 	}
 
-	select {
-	case aim.inputBuffer <- frame:
-		atomic.AddInt64(&aim.metrics.FramesSent, 1)
-		atomic.AddInt64(&aim.metrics.BytesProcessed, int64(len(frame)))
-		aim.metrics.LastFrameTime = time.Now()
-		return nil
-	default:
-		// Buffer full, drop frame
+	// Track end-to-end latency from WebRTC to IPC
+	startTime := time.Now()
+	err := aim.ipcManager.WriteOpusFrame(frame)
+	processingTime := time.Since(startTime)
+
+	// Log high latency warnings
+	if processingTime > 10*time.Millisecond {
+		aim.logger.Warn().
+			Dur("latency_ms", processingTime).
+			Msg("High audio processing latency detected")
+	}
+
+	if err != nil {
 		atomic.AddInt64(&aim.metrics.FramesDropped, 1)
-		aim.logger.Warn().Msg("Audio input buffer full, dropping frame")
-		return nil
+		return err
+	}
+
+	// Update metrics
+	atomic.AddInt64(&aim.metrics.FramesSent, 1)
+	atomic.AddInt64(&aim.metrics.BytesProcessed, int64(len(frame)))
+	aim.metrics.LastFrameTime = time.Now()
+	aim.metrics.AverageLatency = processingTime
+	return nil
+}
+
+// GetMetrics returns current audio input metrics
+func (aim *AudioInputManager) GetMetrics() AudioInputMetrics {
+	return AudioInputMetrics{
+		FramesSent:     atomic.LoadInt64(&aim.metrics.FramesSent),
+		FramesDropped:  atomic.LoadInt64(&aim.metrics.FramesDropped),
+		BytesProcessed: atomic.LoadInt64(&aim.metrics.BytesProcessed),
+		AverageLatency: aim.metrics.AverageLatency,
+		LastFrameTime:  aim.metrics.LastFrameTime,
 	}
 }
 
-// GetMetrics returns current microphone input metrics
-func (aim *AudioInputManager) GetMetrics() AudioInputMetrics {
-	return AudioInputMetrics{
-		FramesSent:      atomic.LoadInt64(&aim.metrics.FramesSent),
-		FramesDropped:   atomic.LoadInt64(&aim.metrics.FramesDropped),
-		BytesProcessed:  atomic.LoadInt64(&aim.metrics.BytesProcessed),
-		LastFrameTime:   aim.metrics.LastFrameTime,
-		ConnectionDrops: atomic.LoadInt64(&aim.metrics.ConnectionDrops),
-		AverageLatency:  aim.metrics.AverageLatency,
+// GetComprehensiveMetrics returns detailed performance metrics across all components
+func (aim *AudioInputManager) GetComprehensiveMetrics() map[string]interface{} {
+	// Get base metrics
+	baseMetrics := aim.GetMetrics()
+
+	// Get detailed IPC metrics
+	ipcMetrics, detailedStats := aim.ipcManager.GetDetailedMetrics()
+
+	comprehensiveMetrics := map[string]interface{}{
+		"manager": map[string]interface{}{
+			"frames_sent":        baseMetrics.FramesSent,
+			"frames_dropped":     baseMetrics.FramesDropped,
+			"bytes_processed":    baseMetrics.BytesProcessed,
+			"average_latency_ms": float64(baseMetrics.AverageLatency.Nanoseconds()) / 1e6,
+			"last_frame_time":    baseMetrics.LastFrameTime,
+			"running":            aim.IsRunning(),
+		},
+		"ipc": map[string]interface{}{
+			"frames_sent":        ipcMetrics.FramesSent,
+			"frames_dropped":     ipcMetrics.FramesDropped,
+			"bytes_processed":    ipcMetrics.BytesProcessed,
+			"average_latency_ms": float64(ipcMetrics.AverageLatency.Nanoseconds()) / 1e6,
+			"last_frame_time":    ipcMetrics.LastFrameTime,
+		},
+		"detailed": detailedStats,
 	}
+
+	return comprehensiveMetrics
+}
+
+// LogPerformanceStats logs current performance statistics
+func (aim *AudioInputManager) LogPerformanceStats() {
+	metrics := aim.GetComprehensiveMetrics()
+
+	managerStats := metrics["manager"].(map[string]interface{})
+	ipcStats := metrics["ipc"].(map[string]interface{})
+	detailedStats := metrics["detailed"].(map[string]interface{})
+
+	aim.logger.Info().
+		Int64("manager_frames_sent", managerStats["frames_sent"].(int64)).
+		Int64("manager_frames_dropped", managerStats["frames_dropped"].(int64)).
+		Float64("manager_latency_ms", managerStats["average_latency_ms"].(float64)).
+		Int64("ipc_frames_sent", ipcStats["frames_sent"].(int64)).
+		Int64("ipc_frames_dropped", ipcStats["frames_dropped"].(int64)).
+		Float64("ipc_latency_ms", ipcStats["average_latency_ms"].(float64)).
+		Float64("client_drop_rate", detailedStats["client_drop_rate"].(float64)).
+		Float64("frames_per_second", detailedStats["frames_per_second"].(float64)).
+		Msg("Audio input performance metrics")
 }
 
 // IsRunning returns whether the audio input manager is running
 func (aim *AudioInputManager) IsRunning() bool {
 	return atomic.LoadInt32(&aim.running) == 1
 }
+
+// IsReady returns whether the audio input manager is ready to receive frames
+// This checks both that it's running and that the IPC connection is established
+func (aim *AudioInputManager) IsReady() bool {
+	if !aim.IsRunning() {
+		return false
+	}
+	return aim.ipcManager.IsReady()
+}

internal/audio/input_api.go

@@ -0,0 +1,94 @@
+package audio
+
+import (
+	"sync/atomic"
+	"unsafe"
+)
+
+var (
+	// Global audio input manager instance
+	globalInputManager unsafe.Pointer // *AudioInputManager
+)
+
+// AudioInputInterface defines the common interface for audio input managers
+type AudioInputInterface interface {
+	Start() error
+	Stop()
+	WriteOpusFrame(frame []byte) error
+	IsRunning() bool
+	GetMetrics() AudioInputMetrics
+}
+
+// GetSupervisor returns the audio input supervisor for advanced management
+func (m *AudioInputManager) GetSupervisor() *AudioInputSupervisor {
+	return m.ipcManager.GetSupervisor()
+}
+
+// getAudioInputManager returns the audio input manager
+func getAudioInputManager() AudioInputInterface {
+	ptr := atomic.LoadPointer(&globalInputManager)
+	if ptr == nil {
+		// Create new manager
+		newManager := NewAudioInputManager()
+		if atomic.CompareAndSwapPointer(&globalInputManager, nil, unsafe.Pointer(newManager)) {
+			return newManager
+		}
+		// Another goroutine created it, use that one
+		ptr = atomic.LoadPointer(&globalInputManager)
+	}
+	return (*AudioInputManager)(ptr)
+}
+
+// StartAudioInput starts the audio input system using the appropriate manager
+func StartAudioInput() error {
+	manager := getAudioInputManager()
+	return manager.Start()
+}
+
+// StopAudioInput stops the audio input system
+func StopAudioInput() {
+	manager := getAudioInputManager()
+	manager.Stop()
+}
+
+// WriteAudioInputFrame writes an Opus frame to the audio input system
+func WriteAudioInputFrame(frame []byte) error {
+	manager := getAudioInputManager()
+	return manager.WriteOpusFrame(frame)
+}
+
+// IsAudioInputRunning returns whether the audio input system is running
+func IsAudioInputRunning() bool {
+	manager := getAudioInputManager()
+	return manager.IsRunning()
+}
+
+// GetAudioInputMetrics returns current audio input metrics
+func GetAudioInputMetrics() AudioInputMetrics {
+	manager := getAudioInputManager()
+	return manager.GetMetrics()
+}
+
+// GetAudioInputIPCSupervisor returns the IPC supervisor
+func GetAudioInputIPCSupervisor() *AudioInputSupervisor {
+	ptr := atomic.LoadPointer(&globalInputManager)
+	if ptr == nil {
+		return nil
+	}
+
+	manager := (*AudioInputManager)(ptr)
+	return manager.GetSupervisor()
+}
+
+// Helper functions
+
+// ResetAudioInputManagers resets the global manager (for testing)
+func ResetAudioInputManagers() {
+	// Stop existing manager first
+	if ptr := atomic.LoadPointer(&globalInputManager); ptr != nil {
+		(*AudioInputManager)(ptr).Stop()
+	}
+
+	// Reset pointer
+	atomic.StorePointer(&globalInputManager, nil)
+}

internal/audio/input_ipc.go

@@ -0,0 +1,694 @@
+package audio
+
+import (
+	"context"
+	"encoding/binary"
+	"fmt"
+	"io"
+	"net"
+	"os"
+	"path/filepath"
+	"sync"
+	"sync/atomic"
+	"time"
+)
+
+const (
+	inputMagicNumber uint32 = 0x4A4B4D49 // "JKMI" (JetKVM Microphone Input)
+	inputSocketName         = "audio_input.sock"
+	maxFrameSize            = 4096                 // Maximum Opus frame size
+	writeTimeout            = 5 * time.Millisecond // Non-blocking write timeout
+	maxDroppedFrames        = 100                  // Maximum consecutive dropped frames before reconnect
+)
+
+// InputMessageType represents the type of IPC message
+type InputMessageType uint8
+
+const (
+	InputMessageTypeOpusFrame InputMessageType = iota
+	InputMessageTypeConfig
+	InputMessageTypeStop
+	InputMessageTypeHeartbeat
+	InputMessageTypeAck
+)
+
+// InputIPCMessage represents a message sent over IPC
+type InputIPCMessage struct {
+	Magic     uint32
+	Type      InputMessageType
+	Length    uint32
+	Timestamp int64
+	Data      []byte
+}
+
+// InputIPCConfig represents configuration for audio input
+type InputIPCConfig struct {
+	SampleRate int
+	Channels   int
+	FrameSize  int
+}
+
+// AudioInputServer handles IPC communication for audio input processing
+type AudioInputServer struct {
+	// Atomic fields must be first for proper alignment on ARM
+	bufferSize     int64 // Current buffer size (atomic)
+	processingTime int64 // Average processing time in nanoseconds (atomic)
+	droppedFrames  int64 // Dropped frames counter (atomic)
+	totalFrames    int64 // Total frames counter (atomic)
+
+	listener net.Listener
+	conn     net.Conn
+	mtx      sync.Mutex
+	running  bool
+
+	// Triple-goroutine architecture
+	messageChan chan *InputIPCMessage // Buffered channel for incoming messages
+	processChan chan *InputIPCMessage // Buffered channel for processing queue
+	stopChan    chan struct{}         // Stop signal for all goroutines
+	wg          sync.WaitGroup        // Wait group for goroutine coordination
+}
+
+// NewAudioInputServer creates a new audio input server
+func NewAudioInputServer() (*AudioInputServer, error) {
+	socketPath := getInputSocketPath()
+	// Remove existing socket if any
+	os.Remove(socketPath)
+
+	listener, err := net.Listen("unix", socketPath)
+	if err != nil {
+		return nil, fmt.Errorf("failed to create unix socket: %w", err)
+	}
+
+	// Initialize with adaptive buffer size (start with 1000 frames)
+	initialBufferSize := int64(1000)
+
+	return &AudioInputServer{
+		listener:    listener,
+		messageChan: make(chan *InputIPCMessage, initialBufferSize),
+		processChan: make(chan *InputIPCMessage, initialBufferSize),
+		stopChan:    make(chan struct{}),
+		bufferSize:  initialBufferSize,
+	}, nil
+}
+
+// Start starts the audio input server
+func (ais *AudioInputServer) Start() error {
+	ais.mtx.Lock()
+	defer ais.mtx.Unlock()
+
+	if ais.running {
+		return fmt.Errorf("server already running")
+	}
+
+	ais.running = true
+
+	// Start triple-goroutine architecture
+	ais.startReaderGoroutine()
+	ais.startProcessorGoroutine()
+	ais.startMonitorGoroutine()
+
+	// Accept connections in a goroutine
+	go ais.acceptConnections()
+
+	return nil
+}
+
+// Stop stops the audio input server
+func (ais *AudioInputServer) Stop() {
+	ais.mtx.Lock()
+	defer ais.mtx.Unlock()
+
+	if !ais.running {
+		return
+	}
+
+	ais.running = false
+
+	// Signal all goroutines to stop
+	close(ais.stopChan)
+	ais.wg.Wait()
+
+	if ais.conn != nil {
+		ais.conn.Close()
+		ais.conn = nil
+	}
+
+	if ais.listener != nil {
+		ais.listener.Close()
+	}
+}
+
+// Close closes the server and cleans up resources
+func (ais *AudioInputServer) Close() {
+	ais.Stop()
+	// Remove socket file
+	os.Remove(getInputSocketPath())
+}
+
+// acceptConnections accepts incoming connections
+func (ais *AudioInputServer) acceptConnections() {
+	for ais.running {
+		conn, err := ais.listener.Accept()
+		if err != nil {
+			if ais.running {
+				// Only log error if we're still supposed to be running
+				continue
+			}
+			return
+		}
+
+		ais.mtx.Lock()
+		// Close existing connection if any
+		if ais.conn != nil {
+			ais.conn.Close()
+		}
+		ais.conn = conn
+		ais.mtx.Unlock()
+
+		// Handle this connection
+		go ais.handleConnection(conn)
+	}
+}
+
+// handleConnection handles a single client connection
+func (ais *AudioInputServer) handleConnection(conn net.Conn) {
+	defer conn.Close()
+
+	// Connection is now handled by the reader goroutine
+	// Just wait for connection to close or stop signal
+	for {
+		select {
+		case <-ais.stopChan:
+			return
+		default:
+			// Check if connection is still alive
+			if ais.conn == nil {
+				return
+			}
+			time.Sleep(100 * time.Millisecond)
+		}
+	}
+}
+
+// readMessage reads a complete message from the connection
+func (ais *AudioInputServer) readMessage(conn net.Conn) (*InputIPCMessage, error) {
+	// Read header (magic + type + length + timestamp)
+	headerSize := 4 + 1 + 4 + 8 // uint32 + uint8 + uint32 + int64
+	header := make([]byte, headerSize)
+
+	_, err := io.ReadFull(conn, header)
+	if err != nil {
+		return nil, err
+	}
+
+	// Parse header
+	msg := &InputIPCMessage{}
+	msg.Magic = binary.LittleEndian.Uint32(header[0:4])
+	msg.Type = InputMessageType(header[4])
+	msg.Length = binary.LittleEndian.Uint32(header[5:9])
+	msg.Timestamp = int64(binary.LittleEndian.Uint64(header[9:17]))
+
+	// Validate magic number
+	if msg.Magic != inputMagicNumber {
+		return nil, fmt.Errorf("invalid magic number: %x", msg.Magic)
+	}
+
+	// Validate message length
+	if msg.Length > maxFrameSize {
+		return nil, fmt.Errorf("message too large: %d bytes", msg.Length)
+	}
+
+	// Read data if present
+	if msg.Length > 0 {
+		msg.Data = make([]byte, msg.Length)
+		_, err = io.ReadFull(conn, msg.Data)
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	return msg, nil
+}
+
+// processMessage processes a received message
+func (ais *AudioInputServer) processMessage(msg *InputIPCMessage) error {
+	switch msg.Type {
+	case InputMessageTypeOpusFrame:
+		return ais.processOpusFrame(msg.Data)
+	case InputMessageTypeConfig:
+		return ais.processConfig(msg.Data)
+	case InputMessageTypeStop:
+		return fmt.Errorf("stop message received")
+	case InputMessageTypeHeartbeat:
+		return ais.sendAck()
+	default:
+		return fmt.Errorf("unknown message type: %d", msg.Type)
+	}
+}
+
+// processOpusFrame processes an Opus audio frame
+func (ais *AudioInputServer) processOpusFrame(data []byte) error {
+	if len(data) == 0 {
+		return nil // Empty frame, ignore
+	}
+
+	// Process the Opus frame using CGO
+	_, err := CGOAudioDecodeWrite(data)
+	return err
+}
+
+// processConfig processes a configuration update
+func (ais *AudioInputServer) processConfig(data []byte) error {
+	// Acknowledge configuration receipt
+	return ais.sendAck()
+}
+
+// sendAck sends an acknowledgment message
+func (ais *AudioInputServer) sendAck() error {
+	ais.mtx.Lock()
+	defer ais.mtx.Unlock()
+
+	if ais.conn == nil {
+		return fmt.Errorf("no connection")
+	}
+
+	msg := &InputIPCMessage{
+		Magic:     inputMagicNumber,
+		Type:      InputMessageTypeAck,
+		Length:    0,
+		Timestamp: time.Now().UnixNano(),
+	}
+
+	return ais.writeMessage(ais.conn, msg)
+}
+
+// writeMessage writes a message to the connection
+func (ais *AudioInputServer) writeMessage(conn net.Conn, msg *InputIPCMessage) error {
+	// Prepare header
+	headerSize := 4 + 1 + 4 + 8
+	header := make([]byte, headerSize)
+
+	binary.LittleEndian.PutUint32(header[0:4], msg.Magic)
+	header[4] = byte(msg.Type)
+	binary.LittleEndian.PutUint32(header[5:9], msg.Length)
+	binary.LittleEndian.PutUint64(header[9:17], uint64(msg.Timestamp))
+
+	// Write header
+	_, err := conn.Write(header)
+	if err != nil {
+		return err
+	}
+
+	// Write data if present
+	if msg.Length > 0 && msg.Data != nil {
+		_, err = conn.Write(msg.Data)
+		if err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+// AudioInputClient handles IPC communication from the main process
+type AudioInputClient struct {
+	// Atomic fields must be first for proper alignment on ARM
+	droppedFrames int64 // Atomic counter for dropped frames
+	totalFrames   int64 // Atomic counter for total frames
+
+	conn    net.Conn
+	mtx     sync.Mutex
+	running bool
+}
+
+// NewAudioInputClient creates a new audio input client
+func NewAudioInputClient() *AudioInputClient {
+	return &AudioInputClient{}
+}
+
+// Connect connects to the audio input server
+func (aic *AudioInputClient) Connect() error {
+	aic.mtx.Lock()
+	defer aic.mtx.Unlock()
+
+	if aic.running {
+		return nil // Already connected
+	}
+
+	socketPath := getInputSocketPath()
+	// Try connecting multiple times as the server might not be ready
+	// Reduced retry count and delay for faster startup
+	for i := 0; i < 10; i++ {
+		conn, err := net.Dial("unix", socketPath)
+		if err == nil {
+			aic.conn = conn
+			aic.running = true
+			return nil
+		}
+		// Exponential backoff starting at 50ms
+		delay := time.Duration(50*(1<<uint(i/3))) * time.Millisecond
+		if delay > 500*time.Millisecond {
+			delay = 500 * time.Millisecond
+		}
+		time.Sleep(delay)
+	}
+
+	return fmt.Errorf("failed to connect to audio input server")
+}
+
+// Disconnect disconnects from the audio input server
+func (aic *AudioInputClient) Disconnect() {
+	aic.mtx.Lock()
+	defer aic.mtx.Unlock()
+
+	if !aic.running {
+		return
+	}
+
+	aic.running = false
+
+	if aic.conn != nil {
+		// Send stop message
+		msg := &InputIPCMessage{
+			Magic:     inputMagicNumber,
+			Type:      InputMessageTypeStop,
+			Length:    0,
+			Timestamp: time.Now().UnixNano(),
+		}
+		_ = aic.writeMessage(msg) // Ignore errors during shutdown
+
+		aic.conn.Close()
+		aic.conn = nil
+	}
+}
+
+// SendFrame sends an Opus frame to the audio input server
+func (aic *AudioInputClient) SendFrame(frame []byte) error {
+	aic.mtx.Lock()
+	defer aic.mtx.Unlock()
+
+	if !aic.running || aic.conn == nil {
+		return fmt.Errorf("not connected")
+	}
+
+	if len(frame) == 0 {
+		return nil // Empty frame, ignore
+	}
+
+	if len(frame) > maxFrameSize {
+		return fmt.Errorf("frame too large: %d bytes", len(frame))
+	}
+
+	msg := &InputIPCMessage{
+		Magic:     inputMagicNumber,
+		Type:      InputMessageTypeOpusFrame,
+		Length:    uint32(len(frame)),
+		Timestamp: time.Now().UnixNano(),
+		Data:      frame,
+	}
+
+	return aic.writeMessage(msg)
+}
+
+// SendConfig sends a configuration update to the audio input server
+func (aic *AudioInputClient) SendConfig(config InputIPCConfig) error {
+	aic.mtx.Lock()
+	defer aic.mtx.Unlock()
+
+	if !aic.running || aic.conn == nil {
+		return fmt.Errorf("not connected")
+	}
+
+	// Serialize config (simple binary format)
+	data := make([]byte, 12) // 3 * int32
+	binary.LittleEndian.PutUint32(data[0:4], uint32(config.SampleRate))
+	binary.LittleEndian.PutUint32(data[4:8], uint32(config.Channels))
+	binary.LittleEndian.PutUint32(data[8:12], uint32(config.FrameSize))
+
+	msg := &InputIPCMessage{
+		Magic:     inputMagicNumber,
+		Type:      InputMessageTypeConfig,
+		Length:    uint32(len(data)),
+		Timestamp: time.Now().UnixNano(),
+		Data:      data,
+	}
+
+	return aic.writeMessage(msg)
+}
+
+// SendHeartbeat sends a heartbeat message
+func (aic *AudioInputClient) SendHeartbeat() error {
+	aic.mtx.Lock()
+	defer aic.mtx.Unlock()
+
+	if !aic.running || aic.conn == nil {
+		return fmt.Errorf("not connected")
+	}
+
+	msg := &InputIPCMessage{
+		Magic:     inputMagicNumber,
+		Type:      InputMessageTypeHeartbeat,
+		Length:    0,
+		Timestamp: time.Now().UnixNano(),
+	}
+
+	return aic.writeMessage(msg)
+}
+
+// writeMessage writes a message to the server
+func (aic *AudioInputClient) writeMessage(msg *InputIPCMessage) error {
+	// Increment total frames counter
+	atomic.AddInt64(&aic.totalFrames, 1)
+
+	// Prepare header
+	headerSize := 4 + 1 + 4 + 8
+	header := make([]byte, headerSize)
+
+	binary.LittleEndian.PutUint32(header[0:4], msg.Magic)
+	header[4] = byte(msg.Type)
+	binary.LittleEndian.PutUint32(header[5:9], msg.Length)
+	binary.LittleEndian.PutUint64(header[9:17], uint64(msg.Timestamp))
+
+	// Use non-blocking write with timeout
+	ctx, cancel := context.WithTimeout(context.Background(), writeTimeout)
+	defer cancel()
+
+	// Create a channel to signal write completion
+	done := make(chan error, 1)
+	go func() {
+		// Write header
+		_, err := aic.conn.Write(header)
+		if err != nil {
+			done <- err
+			return
+		}
+
+		// Write data if present
+		if msg.Length > 0 && msg.Data != nil {
+			_, err = aic.conn.Write(msg.Data)
+			if err != nil {
+				done <- err
+				return
+			}
+		}
+		done <- nil
+	}()
+
+	// Wait for completion or timeout
+	select {
+	case err := <-done:
+		if err != nil {
+			atomic.AddInt64(&aic.droppedFrames, 1)
+			return err
+		}
+		return nil
+	case <-ctx.Done():
+		// Timeout occurred - drop frame to prevent blocking
+		atomic.AddInt64(&aic.droppedFrames, 1)
+		return fmt.Errorf("write timeout - frame dropped")
+	}
+}
+
+// IsConnected returns whether the client is connected
+func (aic *AudioInputClient) IsConnected() bool {
+	aic.mtx.Lock()
+	defer aic.mtx.Unlock()
+	return aic.running && aic.conn != nil
+}
+
+// GetFrameStats returns frame statistics
+func (aic *AudioInputClient) GetFrameStats() (total, dropped int64) {
+	return atomic.LoadInt64(&aic.totalFrames), atomic.LoadInt64(&aic.droppedFrames)
+}
+
+// GetDropRate returns the current frame drop rate as a percentage
+func (aic *AudioInputClient) GetDropRate() float64 {
+	total := atomic.LoadInt64(&aic.totalFrames)
+	dropped := atomic.LoadInt64(&aic.droppedFrames)
+	if total == 0 {
+		return 0.0
+	}
+	return float64(dropped) / float64(total) * 100.0
+}
+
+// ResetStats resets frame statistics
+func (aic *AudioInputClient) ResetStats() {
+	atomic.StoreInt64(&aic.totalFrames, 0)
+	atomic.StoreInt64(&aic.droppedFrames, 0)
+}
+
+// startReaderGoroutine starts the message reader goroutine
+func (ais *AudioInputServer) startReaderGoroutine() {
+	ais.wg.Add(1)
+	go func() {
+		defer ais.wg.Done()
+		for {
+			select {
+			case <-ais.stopChan:
+				return
+			default:
+				if ais.conn != nil {
+					msg, err := ais.readMessage(ais.conn)
+					if err != nil {
+						continue // Connection error, retry
+					}
+					// Send to message channel with non-blocking write
+					select {
+					case ais.messageChan <- msg:
+						atomic.AddInt64(&ais.totalFrames, 1)
+					default:
+						// Channel full, drop message
+						atomic.AddInt64(&ais.droppedFrames, 1)
+					}
+				}
+			}
+		}
+	}()
+}
+
+// startProcessorGoroutine starts the message processor goroutine
+func (ais *AudioInputServer) startProcessorGoroutine() {
+	ais.wg.Add(1)
+	go func() {
+		defer ais.wg.Done()
+		for {
+			select {
+			case <-ais.stopChan:
+				return
+			case msg := <-ais.messageChan:
+				// Intelligent frame dropping: prioritize recent frames
+				if msg.Type == InputMessageTypeOpusFrame {
+					// Check if processing queue is getting full
+					queueLen := len(ais.processChan)
+					bufferSize := int(atomic.LoadInt64(&ais.bufferSize))
+
+					if queueLen > bufferSize*3/4 {
+						// Drop oldest frames, keep newest
+						select {
+						case <-ais.processChan: // Remove oldest
+							atomic.AddInt64(&ais.droppedFrames, 1)
+						default:
+						}
+					}
+				}
+
+				// Send to processing queue
+				select {
+				case ais.processChan <- msg:
+				default:
+					// Processing queue full, drop frame
+					atomic.AddInt64(&ais.droppedFrames, 1)
+				}
+			}
+		}
+	}()
+}
+
+// startMonitorGoroutine starts the performance monitoring goroutine
+func (ais *AudioInputServer) startMonitorGoroutine() {
+	ais.wg.Add(1)
+	go func() {
+		defer ais.wg.Done()
+		ticker := time.NewTicker(100 * time.Millisecond)
+		defer ticker.Stop()
+
+		for {
+			select {
+			case <-ais.stopChan:
+				return
+			case <-ticker.C:
+				// Process frames from processing queue
+				for {
+					select {
+					case msg := <-ais.processChan:
+						start := time.Now()
+						err := ais.processMessage(msg)
+						processingTime := time.Since(start).Nanoseconds()
+
+						// Calculate end-to-end latency using message timestamp
+						if msg.Type == InputMessageTypeOpusFrame && msg.Timestamp > 0 {
+							msgTime := time.Unix(0, msg.Timestamp)
+							endToEndLatency := time.Since(msgTime).Nanoseconds()
+							// Use exponential moving average for end-to-end latency tracking
+							currentAvg := atomic.LoadInt64(&ais.processingTime)
+							// Weight: 90% historical, 10% current (for smoother averaging)
+							newAvg := (currentAvg*9 + endToEndLatency) / 10
+							atomic.StoreInt64(&ais.processingTime, newAvg)
+						} else {
+							// Fallback to processing time only
+							currentAvg := atomic.LoadInt64(&ais.processingTime)
+							newAvg := (currentAvg + processingTime) / 2
+							atomic.StoreInt64(&ais.processingTime, newAvg)
+						}
+
+						if err != nil {
+							atomic.AddInt64(&ais.droppedFrames, 1)
+						}
+					default:
+						// No more messages to process
+						goto adaptiveBuffering
+					}
+				}
+
+			adaptiveBuffering:
+				// Adaptive buffer sizing based on processing time
+				avgTime := atomic.LoadInt64(&ais.processingTime)
+				currentSize := atomic.LoadInt64(&ais.bufferSize)
+
+				if avgTime > 10*1000*1000 { // > 10ms processing time
+					// Increase buffer size
+					newSize := currentSize * 2
+					if newSize > 1000 {
+						newSize = 1000
+					}
+					atomic.StoreInt64(&ais.bufferSize, newSize)
+				} else if avgTime < 1*1000*1000 { // < 1ms processing time
+					// Decrease buffer size
+					newSize := currentSize / 2
+					if newSize < 50 {
+						newSize = 50
+					}
+					atomic.StoreInt64(&ais.bufferSize, newSize)
+				}
+			}
+		}
+	}()
+}
+
+// GetServerStats returns server performance statistics
+func (ais *AudioInputServer) GetServerStats() (total, dropped int64, avgProcessingTime time.Duration, bufferSize int64) {
+	return atomic.LoadInt64(&ais.totalFrames),
+		atomic.LoadInt64(&ais.droppedFrames),
+		time.Duration(atomic.LoadInt64(&ais.processingTime)),
+		atomic.LoadInt64(&ais.bufferSize)
+}
+
+// Helper functions
+
+// getInputSocketPath returns the path to the input socket
+func getInputSocketPath() string {
+	if path := os.Getenv("JETKVM_AUDIO_INPUT_SOCKET"); path != "" {
+		return path
+	}
+	return filepath.Join("/var/run", inputSocketName)
+}

internal/audio/input_ipc_manager.go

@@ -0,0 +1,197 @@
+package audio
+
+import (
+	"sync/atomic"
+	"time"
+
+	"github.com/jetkvm/kvm/internal/logging"
+	"github.com/rs/zerolog"
+)
+
+// AudioInputIPCManager manages microphone input using IPC when enabled
+type AudioInputIPCManager struct {
+	// metrics MUST be first for ARM32 alignment (contains int64 fields)
+	metrics AudioInputMetrics
+
+	supervisor *AudioInputSupervisor
+	logger     zerolog.Logger
+	running    int32
+}
+
+// NewAudioInputIPCManager creates a new IPC-based audio input manager
+func NewAudioInputIPCManager() *AudioInputIPCManager {
+	return &AudioInputIPCManager{
+		supervisor: NewAudioInputSupervisor(),
+		logger:     logging.GetDefaultLogger().With().Str("component", "audio-input-ipc").Logger(),
+	}
+}
+
+// Start starts the IPC-based audio input system
+func (aim *AudioInputIPCManager) Start() error {
+	if !atomic.CompareAndSwapInt32(&aim.running, 0, 1) {
+		return nil // Already running
+	}
+
+	aim.logger.Info().Msg("Starting IPC-based audio input system")
+
+	// Start the supervisor which will launch the subprocess
+	err := aim.supervisor.Start()
+	if err != nil {
+		atomic.StoreInt32(&aim.running, 0)
+		return err
+	}
+
+	// Send initial configuration
+	config := InputIPCConfig{
+		SampleRate: 48000,
+		Channels:   2,
+		FrameSize:  960, // 20ms at 48kHz
+	}
+
+	// Wait briefly for the subprocess to be ready (reduced from 1 second)
+	time.Sleep(200 * time.Millisecond)
+
+	err = aim.supervisor.SendConfig(config)
+	if err != nil {
+		aim.logger.Warn().Err(err).Msg("Failed to send initial config to audio input server")
+		// Don't fail startup for config errors
+	}
+
+	aim.logger.Info().Msg("IPC-based audio input system started")
+	return nil
+}
+
+// Stop stops the IPC-based audio input system
+func (aim *AudioInputIPCManager) Stop() {
+	if !atomic.CompareAndSwapInt32(&aim.running, 1, 0) {
+		return // Already stopped
+	}
+
+	aim.logger.Info().Msg("Stopping IPC-based audio input system")
+
+	// Stop the supervisor
+	aim.supervisor.Stop()
+
+	aim.logger.Info().Msg("IPC-based audio input system stopped")
+}
+
+// WriteOpusFrame sends an Opus frame to the audio input server via IPC
+func (aim *AudioInputIPCManager) WriteOpusFrame(frame []byte) error {
+	if atomic.LoadInt32(&aim.running) == 0 {
+		return nil // Not running, silently ignore
+	}
+
+	if len(frame) == 0 {
+		return nil // Empty frame, ignore
+	}
+
+	// Start latency measurement
+	startTime := time.Now()
+
+	// Update metrics
+	atomic.AddInt64(&aim.metrics.FramesSent, 1)
+	atomic.AddInt64(&aim.metrics.BytesProcessed, int64(len(frame)))
+	aim.metrics.LastFrameTime = startTime
+
+	// Send frame via IPC
+	err := aim.supervisor.SendFrame(frame)
+	if err != nil {
+		// Count as dropped frame
+		atomic.AddInt64(&aim.metrics.FramesDropped, 1)
+		aim.logger.Debug().Err(err).Msg("Failed to send frame via IPC")
+		return err
+	}
+
+	// Calculate and update latency (end-to-end IPC transmission time)
+	latency := time.Since(startTime)
+	aim.updateLatencyMetrics(latency)
+
+	return nil
+}
+
+// IsRunning returns whether the IPC manager is running
+func (aim *AudioInputIPCManager) IsRunning() bool {
+	return atomic.LoadInt32(&aim.running) == 1
+}
+
+// IsReady returns whether the IPC manager is ready to receive frames
+// This checks that the supervisor is connected to the audio input server
+func (aim *AudioInputIPCManager) IsReady() bool {
+	if !aim.IsRunning() {
+		return false
+	}
+	return aim.supervisor.IsConnected()
+}
+
+// GetMetrics returns current metrics
+func (aim *AudioInputIPCManager) GetMetrics() AudioInputMetrics {
+	return AudioInputMetrics{
+		FramesSent:      atomic.LoadInt64(&aim.metrics.FramesSent),
+		FramesDropped:   atomic.LoadInt64(&aim.metrics.FramesDropped),
+		BytesProcessed:  atomic.LoadInt64(&aim.metrics.BytesProcessed),
+		ConnectionDrops: atomic.LoadInt64(&aim.metrics.ConnectionDrops),
+		AverageLatency:  aim.metrics.AverageLatency,
+		LastFrameTime:   aim.metrics.LastFrameTime,
+	}
+}
+
+// updateLatencyMetrics updates the latency metrics with exponential moving average
+func (aim *AudioInputIPCManager) updateLatencyMetrics(latency time.Duration) {
+	// Use exponential moving average for smooth latency calculation
+	currentAvg := aim.metrics.AverageLatency
+	if currentAvg == 0 {
+		aim.metrics.AverageLatency = latency
+	} else {
+		// EMA with alpha = 0.1 for smooth averaging
+		aim.metrics.AverageLatency = time.Duration(float64(currentAvg)*0.9 + float64(latency)*0.1)
+	}
+}
+
+// GetDetailedMetrics returns comprehensive performance metrics
+func (aim *AudioInputIPCManager) GetDetailedMetrics() (AudioInputMetrics, map[string]interface{}) {
+	metrics := aim.GetMetrics()
+
+	// Get client frame statistics
+	client := aim.supervisor.GetClient()
+	totalFrames, droppedFrames := int64(0), int64(0)
+	dropRate := 0.0
+	if client != nil {
+		totalFrames, droppedFrames = client.GetFrameStats()
+		dropRate = client.GetDropRate()
+	}
+
+	// Get server statistics if available
+	serverStats := make(map[string]interface{})
+	if aim.supervisor.IsRunning() {
+		serverStats["status"] = "running"
+	} else {
+		serverStats["status"] = "stopped"
+	}
+
+	detailedStats := map[string]interface{}{
+		"client_total_frames":   totalFrames,
+		"client_dropped_frames": droppedFrames,
+		"client_drop_rate":      dropRate,
+		"server_stats":          serverStats,
+		"ipc_latency_ms":        float64(metrics.AverageLatency.Nanoseconds()) / 1e6,
+		"frames_per_second":     aim.calculateFrameRate(),
+	}
+
+	return metrics, detailedStats
+}
+
+// calculateFrameRate calculates the current frame rate
+func (aim *AudioInputIPCManager) calculateFrameRate() float64 {
+	framesSent := atomic.LoadInt64(&aim.metrics.FramesSent)
+	if framesSent == 0 {
+		return 0.0
+	}
+
+	// Return typical Opus frame rate
+	return 50.0
+}
+
+// GetSupervisor returns the supervisor for advanced operations
+func (aim *AudioInputIPCManager) GetSupervisor() *AudioInputSupervisor {
+	return aim.supervisor
+}

internal/audio/input_server_main.go

@@ -0,0 +1,72 @@
+package audio
+
+import (
+	"context"
+	"os"
+	"os/signal"
+	"syscall"
+	"time"
+
+	"github.com/jetkvm/kvm/internal/logging"
+)
+
+// IsAudioInputServerProcess detects if we're running as the audio input server subprocess
+func IsAudioInputServerProcess() bool {
+	return os.Getenv("JETKVM_AUDIO_INPUT_SERVER") == "true"
+}
+
+// RunAudioInputServer runs the audio input server subprocess
+// This should be called from main() when the subprocess is detected
+func RunAudioInputServer() error {
+	logger := logging.GetDefaultLogger().With().Str("component", "audio-input-server").Logger()
+	logger.Info().Msg("Starting audio input server subprocess")
+
+	// Initialize CGO audio system
+	err := CGOAudioPlaybackInit()
+	if err != nil {
+		logger.Error().Err(err).Msg("Failed to initialize CGO audio playback")
+		return err
+	}
+	defer CGOAudioPlaybackClose()
+
+	// Create and start the IPC server
+	server, err := NewAudioInputServer()
+	if err != nil {
+		logger.Error().Err(err).Msg("Failed to create audio input server")
+		return err
+	}
+	defer server.Close()
+
+	err = server.Start()
+	if err != nil {
+		logger.Error().Err(err).Msg("Failed to start audio input server")
+		return err
+	}
+
+	logger.Info().Msg("Audio input server started, waiting for connections")
+
+	// Set up signal handling for graceful shutdown
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	sigChan := make(chan os.Signal, 1)
+	signal.Notify(sigChan, syscall.SIGINT, syscall.SIGTERM)
+
+	// Wait for shutdown signal
+	select {
+	case sig := <-sigChan:
+		logger.Info().Str("signal", sig.String()).Msg("Received shutdown signal")
+	case <-ctx.Done():
+		logger.Info().Msg("Context cancelled")
+	}
+
+	// Graceful shutdown
+	logger.Info().Msg("Shutting down audio input server")
+	server.Stop()
+
+	// Give some time for cleanup
+	time.Sleep(100 * time.Millisecond)
+
+	logger.Info().Msg("Audio input server subprocess stopped")
+	return nil
+}

internal/audio/input_supervisor.go

@@ -0,0 +1,259 @@
+package audio
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"os/exec"
+	"sync"
+	"syscall"
+	"time"
+
+	"github.com/jetkvm/kvm/internal/logging"
+	"github.com/rs/zerolog"
+)
+
+// AudioInputSupervisor manages the audio input server subprocess
+type AudioInputSupervisor struct {
+	cmd            *exec.Cmd
+	cancel         context.CancelFunc
+	mtx            sync.Mutex
+	running        bool
+	logger         zerolog.Logger
+	client         *AudioInputClient
+	processMonitor *ProcessMonitor
+}
+
+// NewAudioInputSupervisor creates a new audio input supervisor
+func NewAudioInputSupervisor() *AudioInputSupervisor {
+	return &AudioInputSupervisor{
+		logger:         logging.GetDefaultLogger().With().Str("component", "audio-input-supervisor").Logger(),
+		client:         NewAudioInputClient(),
+		processMonitor: GetProcessMonitor(),
+	}
+}
+
+// Start starts the audio input server subprocess
+func (ais *AudioInputSupervisor) Start() error {
+	ais.mtx.Lock()
+	defer ais.mtx.Unlock()
+
+	if ais.running {
+		return fmt.Errorf("audio input supervisor already running")
+	}
+
+	// Create context for subprocess management
+	ctx, cancel := context.WithCancel(context.Background())
+	ais.cancel = cancel
+
+	// Get current executable path
+	execPath, err := os.Executable()
+	if err != nil {
+		return fmt.Errorf("failed to get executable path: %w", err)
+	}
+
+	// Create command for audio input server subprocess
+	cmd := exec.CommandContext(ctx, execPath)
+	cmd.Env = append(os.Environ(),
+		"JETKVM_AUDIO_INPUT_SERVER=true", // Flag to indicate this is the input server process
+		"JETKVM_AUDIO_INPUT_IPC=true",    // Enable IPC mode
+	)
+
+	// Set process group to allow clean termination
+	cmd.SysProcAttr = &syscall.SysProcAttr{
+		Setpgid: true,
+	}
+
+	ais.cmd = cmd
+	ais.running = true
+
+	// Start the subprocess
+	err = cmd.Start()
+	if err != nil {
+		ais.running = false
+		cancel()
+		return fmt.Errorf("failed to start audio input server: %w", err)
+	}
+
+	ais.logger.Info().Int("pid", cmd.Process.Pid).Msg("Audio input server subprocess started")
+
+	// Add process to monitoring
+	ais.processMonitor.AddProcess(cmd.Process.Pid, "audio-input-server")
+
+	// Monitor the subprocess in a goroutine
+	go ais.monitorSubprocess()
+
+	// Connect client to the server
+	go ais.connectClient()
+
+	return nil
+}
+
+// Stop stops the audio input server subprocess
+func (ais *AudioInputSupervisor) Stop() {
+	ais.mtx.Lock()
+	defer ais.mtx.Unlock()
+
+	if !ais.running {
+		return
+	}
+
+	ais.running = false
+
+	// Disconnect client first
+	if ais.client != nil {
+		ais.client.Disconnect()
+	}
+
+	// Cancel context to signal subprocess to stop
+	if ais.cancel != nil {
+		ais.cancel()
+	}
+
+	// Try graceful termination first
+	if ais.cmd != nil && ais.cmd.Process != nil {
+		ais.logger.Info().Int("pid", ais.cmd.Process.Pid).Msg("Stopping audio input server subprocess")
+
+		// Send SIGTERM
+		err := ais.cmd.Process.Signal(syscall.SIGTERM)
+		if err != nil {
+			ais.logger.Warn().Err(err).Msg("Failed to send SIGTERM to audio input server")
+		}
+
+		// Wait for graceful shutdown with timeout
+		done := make(chan error, 1)
+		go func() {
+			done <- ais.cmd.Wait()
+		}()
+
+		select {
+		case <-done:
+			ais.logger.Info().Msg("Audio input server subprocess stopped gracefully")
+		case <-time.After(5 * time.Second):
+			// Force kill if graceful shutdown failed
+			ais.logger.Warn().Msg("Audio input server subprocess did not stop gracefully, force killing")
+			err := ais.cmd.Process.Kill()
+			if err != nil {
+				ais.logger.Error().Err(err).Msg("Failed to kill audio input server subprocess")
+			}
+		}
+	}
+
+	ais.cmd = nil
+	ais.cancel = nil
+}
+
+// IsRunning returns whether the supervisor is running
+func (ais *AudioInputSupervisor) IsRunning() bool {
+	ais.mtx.Lock()
+	defer ais.mtx.Unlock()
+	return ais.running
+}
+
+// IsConnected returns whether the client is connected to the audio input server
+func (ais *AudioInputSupervisor) IsConnected() bool {
+	if !ais.IsRunning() {
+		return false
+	}
+	return ais.client.IsConnected()
+}
+
+// GetClient returns the IPC client for sending audio frames
+func (ais *AudioInputSupervisor) GetClient() *AudioInputClient {
+	return ais.client
+}
+
+// GetProcessMetrics returns current process metrics if the process is running
+func (ais *AudioInputSupervisor) GetProcessMetrics() *ProcessMetrics {
+	ais.mtx.Lock()
+	defer ais.mtx.Unlock()
+
+	if ais.cmd == nil || ais.cmd.Process == nil {
+		return nil
+	}
+
+	pid := ais.cmd.Process.Pid
+	metrics := ais.processMonitor.GetCurrentMetrics()
+	for _, metric := range metrics {
+		if metric.PID == pid {
+			return &metric
+		}
+	}
+	return nil
+}
+
+// monitorSubprocess monitors the subprocess and handles unexpected exits
+func (ais *AudioInputSupervisor) monitorSubprocess() {
+	if ais.cmd == nil {
+		return
+	}
+
+	pid := ais.cmd.Process.Pid
+	err := ais.cmd.Wait()
+
+	// Remove process from monitoring
+	ais.processMonitor.RemoveProcess(pid)
+
+	ais.mtx.Lock()
+	defer ais.mtx.Unlock()
+
+	if ais.running {
+		// Unexpected exit
+		if err != nil {
+			ais.logger.Error().Err(err).Msg("Audio input server subprocess exited unexpectedly")
+		} else {
+			ais.logger.Warn().Msg("Audio input server subprocess exited unexpectedly")
+		}
+
+		// Disconnect client
+		if ais.client != nil {
+			ais.client.Disconnect()
+		}
+
+		// Mark as not running
+		ais.running = false
+		ais.cmd = nil
+
+		ais.logger.Info().Msg("Audio input server subprocess monitoring stopped")
+	}
+}
+
+// connectClient attempts to connect the client to the server
+func (ais *AudioInputSupervisor) connectClient() {
+	// Wait briefly for the server to start (reduced from 500ms)
+	time.Sleep(100 * time.Millisecond)
+
+	err := ais.client.Connect()
+	if err != nil {
+		ais.logger.Error().Err(err).Msg("Failed to connect to audio input server")
+		return
+	}
+
+	ais.logger.Info().Msg("Connected to audio input server")
+}
+
+// SendFrame sends an audio frame to the subprocess (convenience method)
+func (ais *AudioInputSupervisor) SendFrame(frame []byte) error {
+	if ais.client == nil {
+		return fmt.Errorf("client not initialized")
+	}
+
+	if !ais.client.IsConnected() {
+		return fmt.Errorf("client not connected")
+	}
+
+	return ais.client.SendFrame(frame)
+}
+
+// SendConfig sends a configuration update to the subprocess (convenience method)
+func (ais *AudioInputSupervisor) SendConfig(config InputIPCConfig) error {
+	if ais.client == nil {
+		return fmt.Errorf("client not initialized")
+	}
+
+	if !ais.client.IsConnected() {
+		return fmt.Errorf("client not connected")
+	}
+
+	return ais.client.SendConfig(config)
+}

internal/audio/ipc.go

@@ -0,0 +1,128 @@
+package audio
+
+import (
+	"encoding/binary"
+	"fmt"
+	"io"
+	"net"
+	"os"
+	"path/filepath"
+	"sync"
+	"time"
+)
+
+const (
+	magicNumber uint32 = 0x4A4B564D // "JKVM"
+	socketName         = "audio_output.sock"
+)
+
+type AudioServer struct {
+	listener net.Listener
+	conn     net.Conn
+	mtx      sync.Mutex
+}
+
+func NewAudioServer() (*AudioServer, error) {
+	socketPath := filepath.Join("/var/run", socketName)
+	// Remove existing socket if any
+	os.Remove(socketPath)
+
+	listener, err := net.Listen("unix", socketPath)
+	if err != nil {
+		return nil, fmt.Errorf("failed to create unix socket: %w", err)
+	}
+
+	return &AudioServer{listener: listener}, nil
+}
+
+func (s *AudioServer) Start() error {
+	conn, err := s.listener.Accept()
+	if err != nil {
+		return fmt.Errorf("failed to accept connection: %w", err)
+	}
+	s.conn = conn
+	return nil
+}
+
+func (s *AudioServer) Close() error {
+	if s.conn != nil {
+		s.conn.Close()
+	}
+	return s.listener.Close()
+}
+
+func (s *AudioServer) SendFrame(frame []byte) error {
+	s.mtx.Lock()
+	defer s.mtx.Unlock()
+
+	if s.conn == nil {
+		return fmt.Errorf("no client connected")
+	}
+
+	// Write magic number
+	if err := binary.Write(s.conn, binary.BigEndian, magicNumber); err != nil {
+		return fmt.Errorf("failed to write magic number: %w", err)
+	}
+
+	// Write frame size
+	if err := binary.Write(s.conn, binary.BigEndian, uint32(len(frame))); err != nil {
+		return fmt.Errorf("failed to write frame size: %w", err)
+	}
+
+	// Write frame data
+	if _, err := s.conn.Write(frame); err != nil {
+		return fmt.Errorf("failed to write frame data: %w", err)
+	}
+
+	return nil
+}
+
+type AudioClient struct {
+	conn net.Conn
+	mtx  sync.Mutex
+}
+
+func NewAudioClient() (*AudioClient, error) {
+	socketPath := filepath.Join("/var/run", socketName)
+	// Try connecting multiple times as the server might not be ready
+	for i := 0; i < 5; i++ {
+		conn, err := net.Dial("unix", socketPath)
+		if err == nil {
+			return &AudioClient{conn: conn}, nil
+		}
+		time.Sleep(time.Second)
+	}
+	return nil, fmt.Errorf("failed to connect to audio server")
+}
+
+func (c *AudioClient) Close() error {
+	return c.conn.Close()
+}
+
+func (c *AudioClient) ReceiveFrame() ([]byte, error) {
+	c.mtx.Lock()
+	defer c.mtx.Unlock()
+
+	// Read magic number
+	var magic uint32
+	if err := binary.Read(c.conn, binary.BigEndian, &magic); err != nil {
+		return nil, fmt.Errorf("failed to read magic number: %w", err)
+	}
+	if magic != magicNumber {
+		return nil, fmt.Errorf("invalid magic number: %x", magic)
+	}
+
+	// Read frame size
+	var size uint32
+	if err := binary.Read(c.conn, binary.BigEndian, &size); err != nil {
+		return nil, fmt.Errorf("failed to read frame size: %w", err)
+	}
+
+	// Read frame data
+	frame := make([]byte, size)
+	if _, err := io.ReadFull(c.conn, frame); err != nil {
+		return nil, fmt.Errorf("failed to read frame data: %w", err)
+	}
+
+	return frame, nil
+}

internal/audio/metrics.go

@@ -0,0 +1,410 @@
+package audio
+
+import (
+	"sync"
+	"time"
+
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promauto"
+)
+
+var (
+	// Audio output metrics
+	audioFramesReceivedTotal = promauto.NewCounter(
+		prometheus.CounterOpts{
+			Name: "jetkvm_audio_frames_received_total",
+			Help: "Total number of audio frames received",
+		},
+	)
+
+	audioFramesDroppedTotal = promauto.NewCounter(
+		prometheus.CounterOpts{
+			Name: "jetkvm_audio_frames_dropped_total",
+			Help: "Total number of audio frames dropped",
+		},
+	)
+
+	audioBytesProcessedTotal = promauto.NewCounter(
+		prometheus.CounterOpts{
+			Name: "jetkvm_audio_bytes_processed_total",
+			Help: "Total number of audio bytes processed",
+		},
+	)
+
+	audioConnectionDropsTotal = promauto.NewCounter(
+		prometheus.CounterOpts{
+			Name: "jetkvm_audio_connection_drops_total",
+			Help: "Total number of audio connection drops",
+		},
+	)
+
+	audioAverageLatencySeconds = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_audio_average_latency_seconds",
+			Help: "Average audio latency in seconds",
+		},
+	)
+
+	audioLastFrameTimestamp = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_audio_last_frame_timestamp_seconds",
+			Help: "Timestamp of the last audio frame received",
+		},
+	)
+
+	// Microphone input metrics
+	microphoneFramesSentTotal = promauto.NewCounter(
+		prometheus.CounterOpts{
+			Name: "jetkvm_microphone_frames_sent_total",
+			Help: "Total number of microphone frames sent",
+		},
+	)
+
+	microphoneFramesDroppedTotal = promauto.NewCounter(
+		prometheus.CounterOpts{
+			Name: "jetkvm_microphone_frames_dropped_total",
+			Help: "Total number of microphone frames dropped",
+		},
+	)
+
+	microphoneBytesProcessedTotal = promauto.NewCounter(
+		prometheus.CounterOpts{
+			Name: "jetkvm_microphone_bytes_processed_total",
+			Help: "Total number of microphone bytes processed",
+		},
+	)
+
+	microphoneConnectionDropsTotal = promauto.NewCounter(
+		prometheus.CounterOpts{
+			Name: "jetkvm_microphone_connection_drops_total",
+			Help: "Total number of microphone connection drops",
+		},
+	)
+
+	microphoneAverageLatencySeconds = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_microphone_average_latency_seconds",
+			Help: "Average microphone latency in seconds",
+		},
+	)
+
+	microphoneLastFrameTimestamp = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_microphone_last_frame_timestamp_seconds",
+			Help: "Timestamp of the last microphone frame sent",
+		},
+	)
+
+	// Audio subprocess process metrics
+	audioProcessCpuPercent = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_audio_process_cpu_percent",
+			Help: "CPU usage percentage of audio output subprocess",
+		},
+	)
+
+	audioProcessMemoryPercent = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_audio_process_memory_percent",
+			Help: "Memory usage percentage of audio output subprocess",
+		},
+	)
+
+	audioProcessMemoryRssBytes = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_audio_process_memory_rss_bytes",
+			Help: "RSS memory usage in bytes of audio output subprocess",
+		},
+	)
+
+	audioProcessMemoryVmsBytes = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_audio_process_memory_vms_bytes",
+			Help: "VMS memory usage in bytes of audio output subprocess",
+		},
+	)
+
+	audioProcessRunning = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_audio_process_running",
+			Help: "Whether audio output subprocess is running (1=running, 0=stopped)",
+		},
+	)
+
+	// Microphone subprocess process metrics
+	microphoneProcessCpuPercent = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_microphone_process_cpu_percent",
+			Help: "CPU usage percentage of microphone input subprocess",
+		},
+	)
+
+	microphoneProcessMemoryPercent = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_microphone_process_memory_percent",
+			Help: "Memory usage percentage of microphone input subprocess",
+		},
+	)
+
+	microphoneProcessMemoryRssBytes = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_microphone_process_memory_rss_bytes",
+			Help: "RSS memory usage in bytes of microphone input subprocess",
+		},
+	)
+
+	microphoneProcessMemoryVmsBytes = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_microphone_process_memory_vms_bytes",
+			Help: "VMS memory usage in bytes of microphone input subprocess",
+		},
+	)
+
+	microphoneProcessRunning = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_microphone_process_running",
+			Help: "Whether microphone input subprocess is running (1=running, 0=stopped)",
+		},
+	)
+
+	// Audio configuration metrics
+	audioConfigQuality = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_audio_config_quality",
+			Help: "Current audio quality setting (0=Low, 1=Medium, 2=High, 3=Ultra)",
+		},
+	)
+
+	audioConfigBitrate = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_audio_config_bitrate_kbps",
+			Help: "Current audio bitrate in kbps",
+		},
+	)
+
+	audioConfigSampleRate = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_audio_config_sample_rate_hz",
+			Help: "Current audio sample rate in Hz",
+		},
+	)
+
+	audioConfigChannels = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_audio_config_channels",
+			Help: "Current audio channel count",
+		},
+	)
+
+	microphoneConfigQuality = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_microphone_config_quality",
+			Help: "Current microphone quality setting (0=Low, 1=Medium, 2=High, 3=Ultra)",
+		},
+	)
+
+	microphoneConfigBitrate = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_microphone_config_bitrate_kbps",
+			Help: "Current microphone bitrate in kbps",
+		},
+	)
+
+	microphoneConfigSampleRate = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_microphone_config_sample_rate_hz",
+			Help: "Current microphone sample rate in Hz",
+		},
+	)
+
+	microphoneConfigChannels = promauto.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "jetkvm_microphone_config_channels",
+			Help: "Current microphone channel count",
+		},
+	)
+
+	// Metrics update tracking
+	metricsUpdateMutex sync.RWMutex
+	lastMetricsUpdate  time.Time
+
+	// Counter value tracking (since prometheus counters don't have Get() method)
+	audioFramesReceivedValue  int64
+	audioFramesDroppedValue   int64
+	audioBytesProcessedValue  int64
+	audioConnectionDropsValue int64
+	micFramesSentValue        int64
+	micFramesDroppedValue     int64
+	micBytesProcessedValue    int64
+	micConnectionDropsValue   int64
+)
+
+// UpdateAudioMetrics updates Prometheus metrics with current audio data
+func UpdateAudioMetrics(metrics AudioMetrics) {
+	metricsUpdateMutex.Lock()
+	defer metricsUpdateMutex.Unlock()
+
+	// Update counters with delta values
+	if metrics.FramesReceived > audioFramesReceivedValue {
+		audioFramesReceivedTotal.Add(float64(metrics.FramesReceived - audioFramesReceivedValue))
+		audioFramesReceivedValue = metrics.FramesReceived
+	}
+
+	if metrics.FramesDropped > audioFramesDroppedValue {
+		audioFramesDroppedTotal.Add(float64(metrics.FramesDropped - audioFramesDroppedValue))
+		audioFramesDroppedValue = metrics.FramesDropped
+	}
+
+	if metrics.BytesProcessed > audioBytesProcessedValue {
+		audioBytesProcessedTotal.Add(float64(metrics.BytesProcessed - audioBytesProcessedValue))
+		audioBytesProcessedValue = metrics.BytesProcessed
+	}
+
+	if metrics.ConnectionDrops > audioConnectionDropsValue {
+		audioConnectionDropsTotal.Add(float64(metrics.ConnectionDrops - audioConnectionDropsValue))
+		audioConnectionDropsValue = metrics.ConnectionDrops
+	}
+
+	// Update gauges
+	audioAverageLatencySeconds.Set(float64(metrics.AverageLatency.Nanoseconds()) / 1e9)
+	if !metrics.LastFrameTime.IsZero() {
+		audioLastFrameTimestamp.Set(float64(metrics.LastFrameTime.Unix()))
+	}
+
+	lastMetricsUpdate = time.Now()
+}
+
+// UpdateMicrophoneMetrics updates Prometheus metrics with current microphone data
+func UpdateMicrophoneMetrics(metrics AudioInputMetrics) {
+	metricsUpdateMutex.Lock()
+	defer metricsUpdateMutex.Unlock()
+
+	// Update counters with delta values
+	if metrics.FramesSent > micFramesSentValue {
+		microphoneFramesSentTotal.Add(float64(metrics.FramesSent - micFramesSentValue))
+		micFramesSentValue = metrics.FramesSent
+	}
+
+	if metrics.FramesDropped > micFramesDroppedValue {
+		microphoneFramesDroppedTotal.Add(float64(metrics.FramesDropped - micFramesDroppedValue))
+		micFramesDroppedValue = metrics.FramesDropped
+	}
+
+	if metrics.BytesProcessed > micBytesProcessedValue {
+		microphoneBytesProcessedTotal.Add(float64(metrics.BytesProcessed - micBytesProcessedValue))
+		micBytesProcessedValue = metrics.BytesProcessed
+	}
+
+	if metrics.ConnectionDrops > micConnectionDropsValue {
+		microphoneConnectionDropsTotal.Add(float64(metrics.ConnectionDrops - micConnectionDropsValue))
+		micConnectionDropsValue = metrics.ConnectionDrops
+	}
+
+	// Update gauges
+	microphoneAverageLatencySeconds.Set(float64(metrics.AverageLatency.Nanoseconds()) / 1e9)
+	if !metrics.LastFrameTime.IsZero() {
+		microphoneLastFrameTimestamp.Set(float64(metrics.LastFrameTime.Unix()))
+	}
+
+	lastMetricsUpdate = time.Now()
+}
+
+// UpdateAudioProcessMetrics updates Prometheus metrics with audio subprocess data
+func UpdateAudioProcessMetrics(metrics ProcessMetrics, isRunning bool) {
+	metricsUpdateMutex.Lock()
+	defer metricsUpdateMutex.Unlock()
+
+	audioProcessCpuPercent.Set(metrics.CPUPercent)
+	audioProcessMemoryPercent.Set(metrics.MemoryPercent)
+	audioProcessMemoryRssBytes.Set(float64(metrics.MemoryRSS))
+	audioProcessMemoryVmsBytes.Set(float64(metrics.MemoryVMS))
+	if isRunning {
+		audioProcessRunning.Set(1)
+	} else {
+		audioProcessRunning.Set(0)
+	}
+
+	lastMetricsUpdate = time.Now()
+}
+
+// UpdateMicrophoneProcessMetrics updates Prometheus metrics with microphone subprocess data
+func UpdateMicrophoneProcessMetrics(metrics ProcessMetrics, isRunning bool) {
+	metricsUpdateMutex.Lock()
+	defer metricsUpdateMutex.Unlock()
+
+	microphoneProcessCpuPercent.Set(metrics.CPUPercent)
+	microphoneProcessMemoryPercent.Set(metrics.MemoryPercent)
+	microphoneProcessMemoryRssBytes.Set(float64(metrics.MemoryRSS))
+	microphoneProcessMemoryVmsBytes.Set(float64(metrics.MemoryVMS))
+	if isRunning {
+		microphoneProcessRunning.Set(1)
+	} else {
+		microphoneProcessRunning.Set(0)
+	}
+
+	lastMetricsUpdate = time.Now()
+}
+
+// UpdateAudioConfigMetrics updates Prometheus metrics with audio configuration
+func UpdateAudioConfigMetrics(config AudioConfig) {
+	metricsUpdateMutex.Lock()
+	defer metricsUpdateMutex.Unlock()
+
+	audioConfigQuality.Set(float64(config.Quality))
+	audioConfigBitrate.Set(float64(config.Bitrate))
+	audioConfigSampleRate.Set(float64(config.SampleRate))
+	audioConfigChannels.Set(float64(config.Channels))
+
+	lastMetricsUpdate = time.Now()
+}
+
+// UpdateMicrophoneConfigMetrics updates Prometheus metrics with microphone configuration
+func UpdateMicrophoneConfigMetrics(config AudioConfig) {
+	metricsUpdateMutex.Lock()
+	defer metricsUpdateMutex.Unlock()
+
+	microphoneConfigQuality.Set(float64(config.Quality))
+	microphoneConfigBitrate.Set(float64(config.Bitrate))
+	microphoneConfigSampleRate.Set(float64(config.SampleRate))
+	microphoneConfigChannels.Set(float64(config.Channels))
+
+	lastMetricsUpdate = time.Now()
+}
+
+// GetLastMetricsUpdate returns the timestamp of the last metrics update
+func GetLastMetricsUpdate() time.Time {
+	metricsUpdateMutex.RLock()
+	defer metricsUpdateMutex.RUnlock()
+	return lastMetricsUpdate
+}
+
+// StartMetricsUpdater starts a goroutine that periodically updates Prometheus metrics
+func StartMetricsUpdater() {
+	go func() {
+		ticker := time.NewTicker(5 * time.Second) // Update every 5 seconds
+		defer ticker.Stop()
+
+		for range ticker.C {
+			// Update audio output metrics
+			audioMetrics := GetAudioMetrics()
+			UpdateAudioMetrics(audioMetrics)
+
+			// Update microphone input metrics
+			micMetrics := GetAudioInputMetrics()
+			UpdateMicrophoneMetrics(micMetrics)
+
+			// Update microphone subprocess process metrics
+			if inputSupervisor := GetAudioInputIPCSupervisor(); inputSupervisor != nil {
+				if processMetrics := inputSupervisor.GetProcessMetrics(); processMetrics != nil {
+					UpdateMicrophoneProcessMetrics(*processMetrics, inputSupervisor.IsRunning())
+				}
+			}
+
+			// Update audio configuration metrics
+			audioConfig := GetAudioConfig()
+			UpdateAudioConfigMetrics(audioConfig)
+			micConfig := GetMicrophoneConfig()
+			UpdateMicrophoneConfigMetrics(micConfig)
+		}
+	}()
+}

internal/audio/mic_contention.go

@@ -10,10 +10,10 @@ import (
 // with reduced contention using atomic operations and conditional locking
 type MicrophoneContentionManager struct {
 	// Atomic fields (must be 64-bit aligned on 32-bit systems)
-	lastOpNano    int64         // Unix nanoseconds of last operation
-	cooldownNanos int64         // Cooldown duration in nanoseconds
-	operationID   int64         // Incremental operation ID for tracking
-	
+	lastOpNano    int64 // Unix nanoseconds of last operation
+	cooldownNanos int64 // Cooldown duration in nanoseconds
+	operationID   int64 // Incremental operation ID for tracking
+
 	// Lock-free state flags (using atomic.Pointer for lock-free updates)
 	lockPtr unsafe.Pointer // *sync.Mutex - conditionally allocated
 }
@@ -27,61 +27,61 @@ func NewMicrophoneContentionManager(cooldown time.Duration) *MicrophoneContentio
 
 // OperationResult represents the result of attempting a microphone operation
 type OperationResult struct {
-	Allowed         bool
+	Allowed           bool
 	RemainingCooldown time.Duration
-	OperationID     int64
+	OperationID       int64
 }
 
 // TryOperation attempts to perform a microphone operation with optimized contention handling
 func (mcm *MicrophoneContentionManager) TryOperation() OperationResult {
 	now := time.Now().UnixNano()
 	cooldown := atomic.LoadInt64(&mcm.cooldownNanos)
-	
+
 	// Fast path: check if we're clearly outside cooldown period using atomic read
 	lastOp := atomic.LoadInt64(&mcm.lastOpNano)
 	elapsed := now - lastOp
-	
+
 	if elapsed >= cooldown {
 		// Attempt atomic update without locking
 		if atomic.CompareAndSwapInt64(&mcm.lastOpNano, lastOp, now) {
 			opID := atomic.AddInt64(&mcm.operationID, 1)
 			return OperationResult{
-				Allowed:         true,
+				Allowed:           true,
 				RemainingCooldown: 0,
-				OperationID:     opID,
+				OperationID:       opID,
 			}
 		}
 	}
-	
+
 	// Slow path: potential contention, check remaining cooldown
 	currentLastOp := atomic.LoadInt64(&mcm.lastOpNano)
 	currentElapsed := now - currentLastOp
-	
+
 	if currentElapsed >= cooldown {
 		// Race condition: another operation might have updated lastOpNano
 		// Try once more with CAS
 		if atomic.CompareAndSwapInt64(&mcm.lastOpNano, currentLastOp, now) {
 			opID := atomic.AddInt64(&mcm.operationID, 1)
 			return OperationResult{
-				Allowed:         true,
+				Allowed:           true,
 				RemainingCooldown: 0,
-				OperationID:     opID,
+				OperationID:       opID,
 			}
 		}
 		// If CAS failed, fall through to cooldown calculation
 		currentLastOp = atomic.LoadInt64(&mcm.lastOpNano)
 		currentElapsed = now - currentLastOp
 	}
-	
+
 	remaining := time.Duration(cooldown - currentElapsed)
 	if remaining < 0 {
 		remaining = 0
 	}
-	
+
 	return OperationResult{
-		Allowed:         false,
+		Allowed:           false,
 		RemainingCooldown: remaining,
-		OperationID:     atomic.LoadInt64(&mcm.operationID),
+		OperationID:       atomic.LoadInt64(&mcm.operationID),
 	}
 }
 
@@ -127,20 +127,20 @@ func GetMicrophoneContentionManager() *MicrophoneContentionManager {
 	if ptr != nil {
 		return (*MicrophoneContentionManager)(ptr)
 	}
-	
+
 	// Initialize on first use
 	if atomic.CompareAndSwapInt32(&micContentionInitialized, 0, 1) {
 		manager := NewMicrophoneContentionManager(200 * time.Millisecond)
 		atomic.StorePointer(&globalMicContentionManager, unsafe.Pointer(manager))
 		return manager
 	}
-	
+
 	// Another goroutine initialized it, try again
 	ptr = atomic.LoadPointer(&globalMicContentionManager)
 	if ptr != nil {
 		return (*MicrophoneContentionManager)(ptr)
 	}
-	
+
 	// Fallback: create a new manager (should rarely happen)
 	return NewMicrophoneContentionManager(200 * time.Millisecond)
 }
@@ -155,4 +155,4 @@ func TryMicrophoneOperation() OperationResult {
 func SetMicrophoneCooldown(cooldown time.Duration) {
 	manager := GetMicrophoneContentionManager()
 	manager.SetCooldown(cooldown)
-}
+}

internal/audio/nonblocking_api.go

@@ -1,115 +0,0 @@
-package audio
-
-import (
-	"sync/atomic"
-	"unsafe"
-)
-
-var (
-	// Use unsafe.Pointer for atomic operations instead of mutex
-	globalNonBlockingManager unsafe.Pointer // *NonBlockingAudioManager
-)
-
-// loadManager atomically loads the global manager
-func loadManager() *NonBlockingAudioManager {
-	ptr := atomic.LoadPointer(&globalNonBlockingManager)
-	if ptr == nil {
-		return nil
-	}
-	return (*NonBlockingAudioManager)(ptr)
-}
-
-// storeManager atomically stores the global manager
-func storeManager(manager *NonBlockingAudioManager) {
-	atomic.StorePointer(&globalNonBlockingManager, unsafe.Pointer(manager))
-}
-
-// compareAndSwapManager atomically compares and swaps the global manager
-func compareAndSwapManager(old, new *NonBlockingAudioManager) bool {
-	return atomic.CompareAndSwapPointer(&globalNonBlockingManager, 
-		unsafe.Pointer(old), unsafe.Pointer(new))
-}
-
-// StartNonBlockingAudioStreaming starts the non-blocking audio streaming system
-func StartNonBlockingAudioStreaming(send func([]byte)) error {
-	manager := loadManager()
-	if manager != nil && manager.IsOutputRunning() {
-		return nil // Already running, this is not an error
-	}
-
-	if manager == nil {
-		newManager := NewNonBlockingAudioManager()
-		if !compareAndSwapManager(nil, newManager) {
-			// Another goroutine created manager, use it
-			manager = loadManager()
-		} else {
-			manager = newManager
-		}
-	}
-
-	return manager.StartAudioOutput(send)
-}
-
-// StartNonBlockingAudioInput starts the non-blocking audio input system
-func StartNonBlockingAudioInput(receiveChan <-chan []byte) error {
-	manager := loadManager()
-	if manager == nil {
-		newManager := NewNonBlockingAudioManager()
-		if !compareAndSwapManager(nil, newManager) {
-			// Another goroutine created manager, use it
-			manager = loadManager()
-		} else {
-			manager = newManager
-		}
-	}
-
-	// Check if input is already running to avoid unnecessary operations
-	if manager.IsInputRunning() {
-		return nil // Already running, this is not an error
-	}
-
-	return manager.StartAudioInput(receiveChan)
-}
-
-// StopNonBlockingAudioStreaming stops the non-blocking audio streaming system
-func StopNonBlockingAudioStreaming() {
-	manager := loadManager()
-	if manager != nil {
-		manager.Stop()
-		storeManager(nil)
-	}
-}
-
-// StopNonBlockingAudioInput stops only the audio input without affecting output
-func StopNonBlockingAudioInput() {
-	manager := loadManager()
-	if manager != nil && manager.IsInputRunning() {
-		manager.StopAudioInput()
-		
-		// If both input and output are stopped, recreate manager to ensure clean state
-		if !manager.IsRunning() {
-			storeManager(nil)
-		}
-	}
-}
-
-// GetNonBlockingAudioStats returns statistics from the non-blocking audio system
-func GetNonBlockingAudioStats() NonBlockingAudioStats {
-	manager := loadManager()
-	if manager != nil {
-		return manager.GetStats()
-	}
-	return NonBlockingAudioStats{}
-}
-
-// IsNonBlockingAudioRunning returns true if the non-blocking audio system is running
-func IsNonBlockingAudioRunning() bool {
-	manager := loadManager()
-	return manager != nil && manager.IsRunning()
-}
-
-// IsNonBlockingAudioInputRunning returns true if the non-blocking audio input is running
-func IsNonBlockingAudioInputRunning() bool {
-	manager := loadManager()
-	return manager != nil && manager.IsInputRunning()
-}

internal/audio/nonblocking_audio.go

@@ -1,564 +0,0 @@
-package audio
-
-import (
-	"context"
-	"errors"
-	// "runtime" // removed: no longer directly pinning OS thread here; batching handles it
-	"sync"
-	"sync/atomic"
-	"time"
-
-	"github.com/jetkvm/kvm/internal/logging"
-	"github.com/rs/zerolog"
-)
-
-// NonBlockingAudioManager manages audio operations in separate worker threads
-// to prevent blocking of mouse/keyboard operations
-type NonBlockingAudioManager struct {
-	// Statistics - MUST be first for ARM32 alignment (int64 fields need 8-byte alignment)
-	stats NonBlockingAudioStats
-
-	// Control
-	ctx    context.Context
-	cancel context.CancelFunc
-	wg     sync.WaitGroup
-	logger *zerolog.Logger
-
-	// Audio output (capture from device, send to WebRTC)
-	outputSendFunc   func([]byte)
-	outputWorkChan   chan audioWorkItem
-	outputResultChan chan audioResult
-
-	// Audio input (receive from WebRTC, playback to device)
-	inputReceiveChan <-chan []byte
-	inputWorkChan    chan audioWorkItem
-	inputResultChan  chan audioResult
-
-	// Worker threads and flags - int32 fields grouped together
-	outputRunning       int32
-	inputRunning        int32
-	outputWorkerRunning int32
-	inputWorkerRunning  int32
-}
-
-type audioWorkItem struct {
-	workType   audioWorkType
-	data       []byte
-	resultChan chan audioResult
-}
-
-type audioWorkType int
-
-const (
-	audioWorkInit audioWorkType = iota
-	audioWorkReadEncode
-	audioWorkDecodeWrite
-	audioWorkClose
-)
-
-type audioResult struct {
-	success bool
-	data    []byte
-	length  int
-	err     error
-}
-
-type NonBlockingAudioStats struct {
-	// int64 fields MUST be first for ARM32 alignment
-	OutputFramesProcessed int64
-	OutputFramesDropped   int64
-	InputFramesProcessed  int64
-	InputFramesDropped    int64
-	WorkerErrors          int64
-	// time.Time is int64 internally, so it's also aligned
-	LastProcessTime time.Time
-}
-
-// NewNonBlockingAudioManager creates a new non-blocking audio manager
-func NewNonBlockingAudioManager() *NonBlockingAudioManager {
-	ctx, cancel := context.WithCancel(context.Background())
-	logger := logging.GetDefaultLogger().With().Str("component", "nonblocking-audio").Logger()
-
-	return &NonBlockingAudioManager{
-		ctx:              ctx,
-		cancel:           cancel,
-		logger:           &logger,
-		outputWorkChan:   make(chan audioWorkItem, 10), // Buffer for work items
-		outputResultChan: make(chan audioResult, 10),   // Buffer for results
-		inputWorkChan:    make(chan audioWorkItem, 10),
-		inputResultChan:  make(chan audioResult, 10),
-	}
-}
-
-// StartAudioOutput starts non-blocking audio output (capture and encode)
-func (nam *NonBlockingAudioManager) StartAudioOutput(sendFunc func([]byte)) error {
-	if !atomic.CompareAndSwapInt32(&nam.outputRunning, 0, 1) {
-		return ErrAudioAlreadyRunning
-	}
-
-	nam.outputSendFunc = sendFunc
-
-	// Enable batch audio processing for performance
-	EnableBatchAudioProcessing()
-
-	// Start the blocking worker thread
-	nam.wg.Add(1)
-	go nam.outputWorkerThread()
-
-	// Start the non-blocking coordinator
-	nam.wg.Add(1)
-	go nam.outputCoordinatorThread()
-
-	nam.logger.Info().Msg("non-blocking audio output started with batch processing")
-	return nil
-}
-
-// StartAudioInput starts non-blocking audio input (receive and decode)
-func (nam *NonBlockingAudioManager) StartAudioInput(receiveChan <-chan []byte) error {
-	if !atomic.CompareAndSwapInt32(&nam.inputRunning, 0, 1) {
-		return ErrAudioAlreadyRunning
-	}
-
-	nam.inputReceiveChan = receiveChan
-
-	// Enable batch audio processing for performance
-	EnableBatchAudioProcessing()
-
-	// Start the blocking worker thread
-	nam.wg.Add(1)
-	go nam.inputWorkerThread()
-
-	// Start the non-blocking coordinator
-	nam.wg.Add(1)
-	go nam.inputCoordinatorThread()
-
-	nam.logger.Info().Msg("non-blocking audio input started with batch processing")
-	return nil
-}
-
-// outputWorkerThread handles all blocking audio output operations
-func (nam *NonBlockingAudioManager) outputWorkerThread() {
-	defer nam.wg.Done()
-	defer atomic.StoreInt32(&nam.outputWorkerRunning, 0)
-
-	atomic.StoreInt32(&nam.outputWorkerRunning, 1)
-	nam.logger.Debug().Msg("output worker thread started")
-
-	// Initialize audio in worker thread
-	if err := CGOAudioInit(); err != nil {
-		nam.logger.Error().Err(err).Msg("failed to initialize audio in worker thread")
-		return
-	}
-	defer CGOAudioClose()
-
-	// Use buffer pool to avoid allocations
-	buf := GetAudioFrameBuffer()
-	defer PutAudioFrameBuffer(buf)
-
-	for {
-		select {
-		case <-nam.ctx.Done():
-			nam.logger.Debug().Msg("output worker thread stopping")
-			return
-
-		case workItem := <-nam.outputWorkChan:
-			switch workItem.workType {
-			case audioWorkReadEncode:
-				n, err := BatchCGOAudioReadEncode(buf)
-					
-					result := audioResult{
-					success: err == nil,
-					length:  n,
-					err:     err,
-				}
-				if err == nil && n > 0 {
-					// Get buffer from pool and copy data
-					resultBuf := GetAudioFrameBuffer()
-					copy(resultBuf[:n], buf[:n])
-					result.data = resultBuf[:n]
-				}
-
-				// Send result back (non-blocking)
-				select {
-				case workItem.resultChan <- result:
-				case <-nam.ctx.Done():
-					return
-				default:
-					// Drop result if coordinator is not ready
-					if result.data != nil {
-						PutAudioFrameBuffer(result.data)
-					}
-					atomic.AddInt64(&nam.stats.OutputFramesDropped, 1)
-				}
-
-			case audioWorkClose:
-				nam.logger.Debug().Msg("output worker received close signal")
-				return
-			}
-		}
-	}
-}
-
-// outputCoordinatorThread coordinates audio output without blocking
-func (nam *NonBlockingAudioManager) outputCoordinatorThread() {
-	defer nam.wg.Done()
-	defer atomic.StoreInt32(&nam.outputRunning, 0)
-
-	nam.logger.Debug().Msg("output coordinator thread started")
-
-	ticker := time.NewTicker(20 * time.Millisecond) // Match frame timing
-	defer ticker.Stop()
-
-	pendingWork := false
-	resultChan := make(chan audioResult, 1)
-
-	for atomic.LoadInt32(&nam.outputRunning) == 1 {
-		select {
-		case <-nam.ctx.Done():
-			nam.logger.Debug().Msg("output coordinator stopping")
-			return
-
-		case <-ticker.C:
-			// Only submit work if worker is ready and no pending work
-			if !pendingWork && atomic.LoadInt32(&nam.outputWorkerRunning) == 1 {
-				if IsAudioMuted() {
-					continue // Skip when muted
-				}
-
-				workItem := audioWorkItem{
-					workType:   audioWorkReadEncode,
-					resultChan: resultChan,
-				}
-
-				// Submit work (non-blocking)
-				select {
-				case nam.outputWorkChan <- workItem:
-					pendingWork = true
-				default:
-					// Worker is busy, drop this frame
-					atomic.AddInt64(&nam.stats.OutputFramesDropped, 1)
-				}
-			}
-
-		case result := <-resultChan:
-			pendingWork = false
-			nam.stats.LastProcessTime = time.Now()
-
-			if result.success && result.data != nil && result.length > 0 {
-				// Send to WebRTC (non-blocking)
-				if nam.outputSendFunc != nil {
-					nam.outputSendFunc(result.data)
-					atomic.AddInt64(&nam.stats.OutputFramesProcessed, 1)
-					RecordFrameReceived(result.length)
-				}
-				// Return buffer to pool after use
-				PutAudioFrameBuffer(result.data)
-			} else if result.success && result.length == 0 {
-				// No data available - this is normal, not an error
-				// Just continue without logging or counting as error
-			} else {
-				atomic.AddInt64(&nam.stats.OutputFramesDropped, 1)
-				atomic.AddInt64(&nam.stats.WorkerErrors, 1)
-				if result.err != nil {
-					nam.logger.Warn().Err(result.err).Msg("audio output worker error")
-				}
-				// Clean up buffer if present
-				if result.data != nil {
-					PutAudioFrameBuffer(result.data)
-				}
-				RecordFrameDropped()
-			}
-		}
-	}
-
-	// Signal worker to close
-	select {
-	case nam.outputWorkChan <- audioWorkItem{workType: audioWorkClose}:
-	case <-time.After(100 * time.Millisecond):
-		nam.logger.Warn().Msg("timeout signaling output worker to close")
-	}
-
-	nam.logger.Info().Msg("output coordinator thread stopped")
-}
-
-// inputWorkerThread handles all blocking audio input operations
-func (nam *NonBlockingAudioManager) inputWorkerThread() {
-	defer nam.wg.Done()
-	// Cleanup CGO resources properly to avoid double-close scenarios
-	// The outputWorkerThread's CGOAudioClose() will handle all cleanup
-	atomic.StoreInt32(&nam.inputWorkerRunning, 0)
-
-	atomic.StoreInt32(&nam.inputWorkerRunning, 1)
-	nam.logger.Debug().Msg("input worker thread started")
-
-	// Initialize audio playback in worker thread
-	if err := CGOAudioPlaybackInit(); err != nil {
-		nam.logger.Error().Err(err).Msg("failed to initialize audio playback in worker thread")
-		return
-	}
-	
-	// Ensure CGO cleanup happens even if we exit unexpectedly
-	cgoInitialized := true
-	defer func() {
-		if cgoInitialized {
-			nam.logger.Debug().Msg("cleaning up CGO audio playback")
-			// Add extra safety: ensure no more CGO calls can happen
-			atomic.StoreInt32(&nam.inputWorkerRunning, 0)
-			// Note: Don't call CGOAudioPlaybackClose() here to avoid double-close
-			// The outputWorkerThread's CGOAudioClose() will handle all cleanup
-		}
-	}()
-
-	for {
-		// If coordinator has stopped, exit worker loop
-		if atomic.LoadInt32(&nam.inputRunning) == 0 {
-			return
-		}
-		select {
-		case <-nam.ctx.Done():
-			nam.logger.Debug().Msg("input worker thread stopping due to context cancellation")
-			return
-
-		case workItem := <-nam.inputWorkChan:
-			switch workItem.workType {
-			case audioWorkDecodeWrite:
-				// Check if we're still supposed to be running before processing
-				if atomic.LoadInt32(&nam.inputWorkerRunning) == 0 || atomic.LoadInt32(&nam.inputRunning) == 0 {
-					nam.logger.Debug().Msg("input worker stopping, ignoring decode work")
-					// Do not send to resultChan; coordinator may have exited
-					return
-				}
-				
-				// Validate input data before CGO call
-				if workItem.data == nil || len(workItem.data) == 0 {
-					result := audioResult{
-						success: false,
-						err:     errors.New("invalid audio data"),
-					}
-					
-					// Check if coordinator is still running before sending result
-					if atomic.LoadInt32(&nam.inputRunning) == 1 {
-						select {
-						case workItem.resultChan <- result:
-						case <-nam.ctx.Done():
-							return
-						case <-time.After(10 * time.Millisecond):
-							// Timeout - coordinator may have stopped, drop result
-							atomic.AddInt64(&nam.stats.InputFramesDropped, 1)
-						}
-					} else {
-						// Coordinator has stopped, drop result
-						atomic.AddInt64(&nam.stats.InputFramesDropped, 1)
-					}
-					continue
-				}
-
-				// Perform blocking CGO operation with panic recovery
-				var result audioResult
-				func() {
-					defer func() {
-						if r := recover(); r != nil {
-							nam.logger.Error().Interface("panic", r).Msg("CGO decode write panic recovered")
-							result = audioResult{
-								success: false,
-								err:     errors.New("CGO decode write panic"),
-							}
-						}
-					}()
-					
-					// Double-check we're still running before CGO call
-					if atomic.LoadInt32(&nam.inputWorkerRunning) == 0 {
-						result = audioResult{success: false, err: errors.New("worker shutting down")}
-						return
-					}
-					
-					n, err := BatchCGOAudioDecodeWrite(workItem.data)
-					
-					result = audioResult{
-						success: err == nil,
-						length:  n,
-						err:     err,
-					}
-				}()
-
-				// Send result back (non-blocking) - check if coordinator is still running
-				if atomic.LoadInt32(&nam.inputRunning) == 1 {
-					select {
-					case workItem.resultChan <- result:
-					case <-nam.ctx.Done():
-						return
-					case <-time.After(10 * time.Millisecond):
-						// Timeout - coordinator may have stopped, drop result
-						atomic.AddInt64(&nam.stats.InputFramesDropped, 1)
-					}
-				} else {
-					// Coordinator has stopped, drop result
-					atomic.AddInt64(&nam.stats.InputFramesDropped, 1)
-				}
-
-			case audioWorkClose:
-				nam.logger.Debug().Msg("input worker received close signal")
-				return
-			}
-		}
-	}
-}
-
-// inputCoordinatorThread coordinates audio input without blocking
-func (nam *NonBlockingAudioManager) inputCoordinatorThread() {
-	defer nam.wg.Done()
-	defer atomic.StoreInt32(&nam.inputRunning, 0)
-
-	nam.logger.Debug().Msg("input coordinator thread started")
-
-	resultChan := make(chan audioResult, 1)
-	// Do not close resultChan to avoid races with worker sends during shutdown
-
-	for atomic.LoadInt32(&nam.inputRunning) == 1 {
-		select {
-		case <-nam.ctx.Done():
-			nam.logger.Debug().Msg("input coordinator stopping")
-			return
-
-		case frame := <-nam.inputReceiveChan:
-			if len(frame) == 0 {
-				continue
-			}
-
-			// Submit work to worker (non-blocking)
-			if atomic.LoadInt32(&nam.inputWorkerRunning) == 1 {
-				workItem := audioWorkItem{
-					workType:   audioWorkDecodeWrite,
-					data:       frame,
-					resultChan: resultChan,
-				}
-
-				select {
-				case nam.inputWorkChan <- workItem:
-					// Wait for result with timeout and context cancellation
-					select {
-					case result := <-resultChan:
-						if result.success {
-							atomic.AddInt64(&nam.stats.InputFramesProcessed, 1)
-						} else {
-							atomic.AddInt64(&nam.stats.InputFramesDropped, 1)
-							atomic.AddInt64(&nam.stats.WorkerErrors, 1)
-							if result.err != nil {
-								nam.logger.Warn().Err(result.err).Msg("audio input worker error")
-							}
-						}
-					case <-nam.ctx.Done():
-						nam.logger.Debug().Msg("input coordinator stopping during result wait")
-						return
-					case <-time.After(50 * time.Millisecond):
-						// Timeout waiting for result
-						atomic.AddInt64(&nam.stats.InputFramesDropped, 1)
-						nam.logger.Warn().Msg("timeout waiting for input worker result")
-						// Drain any pending result to prevent worker blocking
-						select {
-						case <-resultChan:
-						default:
-						}
-					}
-				default:
-					// Worker is busy, drop this frame
-					atomic.AddInt64(&nam.stats.InputFramesDropped, 1)
-				}
-			}
-
-		case <-time.After(250 * time.Millisecond):
-			// Periodic timeout to prevent blocking
-			continue
-		}
-	}
-
-	// Avoid sending close signals or touching channels here; inputRunning=0 will stop worker via checks
-	nam.logger.Info().Msg("input coordinator thread stopped")
-}
-
-// Stop stops all audio operations
-func (nam *NonBlockingAudioManager) Stop() {
-	nam.logger.Info().Msg("stopping non-blocking audio manager")
-
-	// Signal all threads to stop
-	nam.cancel()
-
-	// Stop coordinators
-	atomic.StoreInt32(&nam.outputRunning, 0)
-	atomic.StoreInt32(&nam.inputRunning, 0)
-
-	// Wait for all goroutines to finish
-	nam.wg.Wait()
-
-	// Disable batch processing to free resources
-	DisableBatchAudioProcessing()
-
-	nam.logger.Info().Msg("non-blocking audio manager stopped")
-}
-
-// StopAudioInput stops only the audio input operations
-func (nam *NonBlockingAudioManager) StopAudioInput() {
-	nam.logger.Info().Msg("stopping audio input")
-
-	// Stop only the input coordinator
-	atomic.StoreInt32(&nam.inputRunning, 0)
-
-	// Drain the receive channel to prevent blocking senders
-	go func() {
-		for {
-			select {
-			case <-nam.inputReceiveChan:
-				// Drain any remaining frames
-			case <-time.After(100 * time.Millisecond):
-				return
-			}
-		}
-	}()
-
-	// Wait for the worker to actually stop to prevent race conditions
-	timeout := time.After(2 * time.Second)
-	ticker := time.NewTicker(10 * time.Millisecond)
-	defer ticker.Stop()
-
-	for {
-		select {
-		case <-timeout:
-			nam.logger.Warn().Msg("timeout waiting for input worker to stop")
-			return
-		case <-ticker.C:
-			if atomic.LoadInt32(&nam.inputWorkerRunning) == 0 {
-				nam.logger.Info().Msg("audio input stopped successfully")
-				// Close ALSA playback resources now that input worker has stopped
-				CGOAudioPlaybackClose()
-				return
-			}
-		}
-	}
-}
-
-// GetStats returns current statistics
-func (nam *NonBlockingAudioManager) GetStats() NonBlockingAudioStats {
-	return NonBlockingAudioStats{
-		OutputFramesProcessed: atomic.LoadInt64(&nam.stats.OutputFramesProcessed),
-		OutputFramesDropped:   atomic.LoadInt64(&nam.stats.OutputFramesDropped),
-		InputFramesProcessed:  atomic.LoadInt64(&nam.stats.InputFramesProcessed),
-		InputFramesDropped:    atomic.LoadInt64(&nam.stats.InputFramesDropped),
-		WorkerErrors:          atomic.LoadInt64(&nam.stats.WorkerErrors),
-		LastProcessTime:       nam.stats.LastProcessTime,
-	}
-}
-
-// IsRunning returns true if any audio operations are running
-func (nam *NonBlockingAudioManager) IsRunning() bool {
-	return atomic.LoadInt32(&nam.outputRunning) == 1 || atomic.LoadInt32(&nam.inputRunning) == 1
-}
-
-// IsInputRunning returns true if audio input is running
-func (nam *NonBlockingAudioManager) IsInputRunning() bool {
-	return atomic.LoadInt32(&nam.inputRunning) == 1
-}
-
-// IsOutputRunning returns true if audio output is running
-func (nam *NonBlockingAudioManager) IsOutputRunning() bool {
-	return atomic.LoadInt32(&nam.outputRunning) == 1
-}

internal/audio/output_streaming.go

@@ -0,0 +1,94 @@
+package audio
+
+import (
+	"context"
+	"sync/atomic"
+	"time"
+
+	"github.com/jetkvm/kvm/internal/logging"
+	"github.com/rs/zerolog"
+)
+
+var (
+	outputStreamingRunning int32
+	outputStreamingCancel  context.CancelFunc
+	outputStreamingLogger  *zerolog.Logger
+)
+
+func getOutputStreamingLogger() *zerolog.Logger {
+	if outputStreamingLogger == nil {
+		logger := logging.GetDefaultLogger().With().Str("component", "audio-output").Logger()
+		outputStreamingLogger = &logger
+	}
+	return outputStreamingLogger
+}
+
+// StartAudioOutputStreaming starts audio output streaming (capturing system audio)
+func StartAudioOutputStreaming(send func([]byte)) error {
+	if !atomic.CompareAndSwapInt32(&outputStreamingRunning, 0, 1) {
+		return ErrAudioAlreadyRunning
+	}
+
+	// Initialize CGO audio capture
+	if err := CGOAudioInit(); err != nil {
+		atomic.StoreInt32(&outputStreamingRunning, 0)
+		return err
+	}
+
+	ctx, cancel := context.WithCancel(context.Background())
+	outputStreamingCancel = cancel
+
+	// Start audio capture loop
+	go func() {
+		defer func() {
+			CGOAudioClose()
+			atomic.StoreInt32(&outputStreamingRunning, 0)
+			getOutputStreamingLogger().Info().Msg("Audio output streaming stopped")
+		}()
+
+		getOutputStreamingLogger().Info().Msg("Audio output streaming started")
+		buffer := make([]byte, MaxAudioFrameSize)
+
+		for {
+			select {
+			case <-ctx.Done():
+				return
+			default:
+				// Capture audio frame
+				n, err := CGOAudioReadEncode(buffer)
+				if err != nil {
+					getOutputStreamingLogger().Warn().Err(err).Msg("Failed to read/encode audio")
+					continue
+				}
+				if n > 0 {
+					// Send frame to callback
+					frame := make([]byte, n)
+					copy(frame, buffer[:n])
+					send(frame)
+					RecordFrameReceived(n)
+				}
+				// Small delay to prevent busy waiting
+				time.Sleep(10 * time.Millisecond)
+			}
+		}
+	}()
+
+	return nil
+}
+
+// StopAudioOutputStreaming stops audio output streaming
+func StopAudioOutputStreaming() {
+	if atomic.LoadInt32(&outputStreamingRunning) == 0 {
+		return
+	}
+
+	if outputStreamingCancel != nil {
+		outputStreamingCancel()
+		outputStreamingCancel = nil
+	}
+
+	// Wait for streaming to stop
+	for atomic.LoadInt32(&outputStreamingRunning) == 1 {
+		time.Sleep(10 * time.Millisecond)
+	}
+}

internal/audio/process_monitor.go

@@ -0,0 +1,263 @@
+package audio
+
+import (
+	"bufio"
+	"fmt"
+	"os"
+	"strconv"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/jetkvm/kvm/internal/logging"
+	"github.com/rs/zerolog"
+)
+
+// ProcessMetrics represents CPU and memory usage metrics for a process
+type ProcessMetrics struct {
+	PID           int       `json:"pid"`
+	CPUPercent    float64   `json:"cpu_percent"`
+	MemoryRSS     int64     `json:"memory_rss_bytes"`
+	MemoryVMS     int64     `json:"memory_vms_bytes"`
+	MemoryPercent float64   `json:"memory_percent"`
+	Timestamp     time.Time `json:"timestamp"`
+	ProcessName   string    `json:"process_name"`
+}
+
+// ProcessMonitor monitors CPU and memory usage of processes
+type ProcessMonitor struct {
+	logger         zerolog.Logger
+	mutex          sync.RWMutex
+	monitoredPIDs  map[int]*processState
+	running        bool
+	stopChan       chan struct{}
+	metricsChan    chan ProcessMetrics
+	updateInterval time.Duration
+}
+
+// processState tracks the state needed for CPU calculation
+type processState struct {
+	name         string
+	lastCPUTime  int64
+	lastSysTime  int64
+	lastUserTime int64
+	lastSample   time.Time
+}
+
+// NewProcessMonitor creates a new process monitor
+func NewProcessMonitor() *ProcessMonitor {
+	return &ProcessMonitor{
+		logger:         logging.GetDefaultLogger().With().Str("component", "process-monitor").Logger(),
+		monitoredPIDs:  make(map[int]*processState),
+		stopChan:       make(chan struct{}),
+		metricsChan:    make(chan ProcessMetrics, 100),
+		updateInterval: 2 * time.Second, // Update every 2 seconds
+	}
+}
+
+// Start begins monitoring processes
+func (pm *ProcessMonitor) Start() {
+	pm.mutex.Lock()
+	defer pm.mutex.Unlock()
+
+	if pm.running {
+		return
+	}
+
+	pm.running = true
+	go pm.monitorLoop()
+	pm.logger.Info().Msg("Process monitor started")
+}
+
+// Stop stops monitoring processes
+func (pm *ProcessMonitor) Stop() {
+	pm.mutex.Lock()
+	defer pm.mutex.Unlock()
+
+	if !pm.running {
+		return
+	}
+
+	pm.running = false
+	close(pm.stopChan)
+	pm.logger.Info().Msg("Process monitor stopped")
+}
+
+// AddProcess adds a process to monitor
+func (pm *ProcessMonitor) AddProcess(pid int, name string) {
+	pm.mutex.Lock()
+	defer pm.mutex.Unlock()
+
+	pm.monitoredPIDs[pid] = &processState{
+		name:       name,
+		lastSample: time.Now(),
+	}
+	pm.logger.Info().Int("pid", pid).Str("name", name).Msg("Added process to monitor")
+}
+
+// RemoveProcess removes a process from monitoring
+func (pm *ProcessMonitor) RemoveProcess(pid int) {
+	pm.mutex.Lock()
+	defer pm.mutex.Unlock()
+
+	delete(pm.monitoredPIDs, pid)
+	pm.logger.Info().Int("pid", pid).Msg("Removed process from monitor")
+}
+
+// GetMetricsChan returns the channel for receiving metrics
+func (pm *ProcessMonitor) GetMetricsChan() <-chan ProcessMetrics {
+	return pm.metricsChan
+}
+
+// GetCurrentMetrics returns current metrics for all monitored processes
+func (pm *ProcessMonitor) GetCurrentMetrics() []ProcessMetrics {
+	pm.mutex.RLock()
+	defer pm.mutex.RUnlock()
+
+	var metrics []ProcessMetrics
+	for pid, state := range pm.monitoredPIDs {
+		if metric, err := pm.collectMetrics(pid, state); err == nil {
+			metrics = append(metrics, metric)
+		}
+	}
+	return metrics
+}
+
+// monitorLoop is the main monitoring loop
+func (pm *ProcessMonitor) monitorLoop() {
+	ticker := time.NewTicker(pm.updateInterval)
+	defer ticker.Stop()
+
+	for {
+		select {
+		case <-pm.stopChan:
+			return
+		case <-ticker.C:
+			pm.collectAllMetrics()
+		}
+	}
+}
+
+// collectAllMetrics collects metrics for all monitored processes
+func (pm *ProcessMonitor) collectAllMetrics() {
+	pm.mutex.RLock()
+	pids := make(map[int]*processState)
+	for pid, state := range pm.monitoredPIDs {
+		pids[pid] = state
+	}
+	pm.mutex.RUnlock()
+
+	for pid, state := range pids {
+		if metric, err := pm.collectMetrics(pid, state); err == nil {
+			select {
+			case pm.metricsChan <- metric:
+			default:
+				// Channel full, skip this metric
+			}
+		} else {
+			// Process might have died, remove it
+			pm.RemoveProcess(pid)
+		}
+	}
+}
+
+// collectMetrics collects metrics for a specific process
+func (pm *ProcessMonitor) collectMetrics(pid int, state *processState) (ProcessMetrics, error) {
+	now := time.Now()
+	metric := ProcessMetrics{
+		PID:         pid,
+		Timestamp:   now,
+		ProcessName: state.name,
+	}
+
+	// Read /proc/[pid]/stat for CPU and memory info
+	statPath := fmt.Sprintf("/proc/%d/stat", pid)
+	statData, err := os.ReadFile(statPath)
+	if err != nil {
+		return metric, fmt.Errorf("failed to read stat file: %w", err)
+	}
+
+	// Parse stat file
+	fields := strings.Fields(string(statData))
+	if len(fields) < 24 {
+		return metric, fmt.Errorf("invalid stat file format")
+	}
+
+	// Extract CPU times (fields 13, 14 are utime, stime in clock ticks)
+	utime, _ := strconv.ParseInt(fields[13], 10, 64)
+	stime, _ := strconv.ParseInt(fields[14], 10, 64)
+	totalCPUTime := utime + stime
+
+	// Extract memory info (field 22 is vsize, field 23 is rss in pages)
+	vsize, _ := strconv.ParseInt(fields[22], 10, 64)
+	rss, _ := strconv.ParseInt(fields[23], 10, 64)
+
+	// Convert RSS from pages to bytes (assuming 4KB pages)
+	pageSize := int64(4096)
+	metric.MemoryRSS = rss * pageSize
+	metric.MemoryVMS = vsize
+
+	// Calculate CPU percentage
+	if !state.lastSample.IsZero() {
+		timeDelta := now.Sub(state.lastSample).Seconds()
+		cpuDelta := float64(totalCPUTime - state.lastCPUTime)
+
+		// Convert from clock ticks to seconds (assuming 100 Hz)
+		clockTicks := 100.0
+		cpuSeconds := cpuDelta / clockTicks
+
+		if timeDelta > 0 {
+			metric.CPUPercent = (cpuSeconds / timeDelta) * 100.0
+		}
+	}
+
+	// Calculate memory percentage (RSS / total system memory)
+	if totalMem := pm.getTotalMemory(); totalMem > 0 {
+		metric.MemoryPercent = float64(metric.MemoryRSS) / float64(totalMem) * 100.0
+	}
+
+	// Update state for next calculation
+	state.lastCPUTime = totalCPUTime
+	state.lastUserTime = utime
+	state.lastSysTime = stime
+	state.lastSample = now
+
+	return metric, nil
+}
+
+// getTotalMemory returns total system memory in bytes
+func (pm *ProcessMonitor) getTotalMemory() int64 {
+	file, err := os.Open("/proc/meminfo")
+	if err != nil {
+		return 0
+	}
+	defer file.Close()
+
+	scanner := bufio.NewScanner(file)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if strings.HasPrefix(line, "MemTotal:") {
+			fields := strings.Fields(line)
+			if len(fields) >= 2 {
+				if kb, err := strconv.ParseInt(fields[1], 10, 64); err == nil {
+					return kb * 1024 // Convert KB to bytes
+				}
+			}
+			break
+		}
+	}
+	return 0
+}
+
+// Global process monitor instance
+var globalProcessMonitor *ProcessMonitor
+var processMonitorOnce sync.Once
+
+// GetProcessMonitor returns the global process monitor instance
+func GetProcessMonitor() *ProcessMonitor {
+	processMonitorOnce.Do(func() {
+		globalProcessMonitor = NewProcessMonitor()
+		globalProcessMonitor.Start()
+	})
+	return globalProcessMonitor
+}

internal/audio/relay.go

@@ -0,0 +1,196 @@
+package audio
+
+import (
+	"context"
+	"sync"
+
+	"github.com/jetkvm/kvm/internal/logging"
+	"github.com/pion/webrtc/v4/pkg/media"
+	"github.com/rs/zerolog"
+)
+
+// AudioRelay handles forwarding audio frames from the audio server subprocess
+// to WebRTC without any CGO audio processing. This runs in the main process.
+type AudioRelay struct {
+	client  *AudioClient
+	ctx     context.Context
+	cancel  context.CancelFunc
+	wg      sync.WaitGroup
+	logger  *zerolog.Logger
+	running bool
+	mutex   sync.RWMutex
+
+	// WebRTC integration
+	audioTrack AudioTrackWriter
+	config     AudioConfig
+	muted      bool
+
+	// Statistics
+	framesRelayed int64
+	framesDropped int64
+}
+
+// AudioTrackWriter interface for WebRTC audio track
+type AudioTrackWriter interface {
+	WriteSample(sample media.Sample) error
+}
+
+// NewAudioRelay creates a new audio relay for the main process
+func NewAudioRelay() *AudioRelay {
+	ctx, cancel := context.WithCancel(context.Background())
+	logger := logging.GetDefaultLogger().With().Str("component", "audio-relay").Logger()
+
+	return &AudioRelay{
+		ctx:    ctx,
+		cancel: cancel,
+		logger: &logger,
+	}
+}
+
+// Start begins the audio relay process
+func (r *AudioRelay) Start(audioTrack AudioTrackWriter, config AudioConfig) error {
+	r.mutex.Lock()
+	defer r.mutex.Unlock()
+
+	if r.running {
+		return nil // Already running
+	}
+
+	// Create audio client to connect to subprocess
+	client, err := NewAudioClient()
+	if err != nil {
+		return err
+	}
+	r.client = client
+	r.audioTrack = audioTrack
+	r.config = config
+
+	// Start relay goroutine
+	r.wg.Add(1)
+	go r.relayLoop()
+
+	r.running = true
+	r.logger.Info().Msg("Audio relay started")
+	return nil
+}
+
+// Stop stops the audio relay
+func (r *AudioRelay) Stop() {
+	r.mutex.Lock()
+	defer r.mutex.Unlock()
+
+	if !r.running {
+		return
+	}
+
+	r.cancel()
+	r.wg.Wait()
+
+	if r.client != nil {
+		r.client.Close()
+		r.client = nil
+	}
+
+	r.running = false
+	r.logger.Info().Msg("Audio relay stopped")
+}
+
+// SetMuted sets the mute state
+func (r *AudioRelay) SetMuted(muted bool) {
+	r.mutex.Lock()
+	defer r.mutex.Unlock()
+	r.muted = muted
+}
+
+// IsMuted returns the current mute state (checks both relay and global mute)
+func (r *AudioRelay) IsMuted() bool {
+	r.mutex.RLock()
+	defer r.mutex.RUnlock()
+	return r.muted || IsAudioMuted()
+}
+
+// GetStats returns relay statistics
+func (r *AudioRelay) GetStats() (framesRelayed, framesDropped int64) {
+	r.mutex.RLock()
+	defer r.mutex.RUnlock()
+	return r.framesRelayed, r.framesDropped
+}
+
+// UpdateTrack updates the WebRTC audio track for the relay
+func (r *AudioRelay) UpdateTrack(audioTrack AudioTrackWriter) {
+	r.mutex.Lock()
+	defer r.mutex.Unlock()
+	r.audioTrack = audioTrack
+}
+
+// relayLoop is the main relay loop that forwards frames from subprocess to WebRTC
+func (r *AudioRelay) relayLoop() {
+	defer r.wg.Done()
+	r.logger.Debug().Msg("Audio relay loop started")
+
+	for {
+		select {
+		case <-r.ctx.Done():
+			r.logger.Debug().Msg("Audio relay loop stopping")
+			return
+		default:
+			// Receive frame from audio server subprocess
+			frame, err := r.client.ReceiveFrame()
+			if err != nil {
+				r.logger.Error().Err(err).Msg("Failed to receive audio frame")
+				r.incrementDropped()
+				continue
+			}
+
+			// Forward frame to WebRTC
+			if err := r.forwardToWebRTC(frame); err != nil {
+				r.logger.Warn().Err(err).Msg("Failed to forward frame to WebRTC")
+				r.incrementDropped()
+			} else {
+				r.incrementRelayed()
+			}
+		}
+	}
+}
+
+// forwardToWebRTC forwards a frame to the WebRTC audio track
+func (r *AudioRelay) forwardToWebRTC(frame []byte) error {
+	r.mutex.RLock()
+	audioTrack := r.audioTrack
+	config := r.config
+	muted := r.muted
+	r.mutex.RUnlock()
+
+	if audioTrack == nil {
+		return nil // No audio track available
+	}
+
+	// Prepare sample data
+	var sampleData []byte
+	if muted {
+		// Send silence when muted
+		sampleData = make([]byte, len(frame))
+	} else {
+		sampleData = frame
+	}
+
+	// Write sample to WebRTC track
+	return audioTrack.WriteSample(media.Sample{
+		Data:     sampleData,
+		Duration: config.FrameSize,
+	})
+}
+
+// incrementRelayed atomically increments the relayed frames counter
+func (r *AudioRelay) incrementRelayed() {
+	r.mutex.Lock()
+	r.framesRelayed++
+	r.mutex.Unlock()
+}
+
+// incrementDropped atomically increments the dropped frames counter
+func (r *AudioRelay) incrementDropped() {
+	r.mutex.Lock()
+	r.framesDropped++
+	r.mutex.Unlock()
+}

internal/audio/relay_api.go

@@ -0,0 +1,109 @@
+package audio
+
+import (
+	"sync"
+)
+
+// Global relay instance for the main process
+var (
+	globalRelay *AudioRelay
+	relayMutex  sync.RWMutex
+)
+
+// StartAudioRelay starts the audio relay system for the main process
+// This replaces the CGO-based audio system when running in main process mode
+// audioTrack can be nil initially and updated later via UpdateAudioRelayTrack
+func StartAudioRelay(audioTrack AudioTrackWriter) error {
+	relayMutex.Lock()
+	defer relayMutex.Unlock()
+
+	if globalRelay != nil {
+		return nil // Already running
+	}
+
+	// Create new relay
+	relay := NewAudioRelay()
+
+	// Get current audio config
+	config := GetAudioConfig()
+
+	// Start the relay (audioTrack can be nil initially)
+	if err := relay.Start(audioTrack, config); err != nil {
+		return err
+	}
+
+	globalRelay = relay
+	return nil
+}
+
+// StopAudioRelay stops the audio relay system
+func StopAudioRelay() {
+	relayMutex.Lock()
+	defer relayMutex.Unlock()
+
+	if globalRelay != nil {
+		globalRelay.Stop()
+		globalRelay = nil
+	}
+}
+
+// SetAudioRelayMuted sets the mute state for the audio relay
+func SetAudioRelayMuted(muted bool) {
+	relayMutex.RLock()
+	defer relayMutex.RUnlock()
+
+	if globalRelay != nil {
+		globalRelay.SetMuted(muted)
+	}
+}
+
+// IsAudioRelayMuted returns the current mute state of the audio relay
+func IsAudioRelayMuted() bool {
+	relayMutex.RLock()
+	defer relayMutex.RUnlock()
+
+	if globalRelay != nil {
+		return globalRelay.IsMuted()
+	}
+	return false
+}
+
+// GetAudioRelayStats returns statistics from the audio relay
+func GetAudioRelayStats() (framesRelayed, framesDropped int64) {
+	relayMutex.RLock()
+	defer relayMutex.RUnlock()
+
+	if globalRelay != nil {
+		return globalRelay.GetStats()
+	}
+	return 0, 0
+}
+
+// IsAudioRelayRunning returns whether the audio relay is currently running
+func IsAudioRelayRunning() bool {
+	relayMutex.RLock()
+	defer relayMutex.RUnlock()
+
+	return globalRelay != nil
+}
+
+// UpdateAudioRelayTrack updates the WebRTC audio track for the relay
+func UpdateAudioRelayTrack(audioTrack AudioTrackWriter) error {
+	relayMutex.Lock()
+	defer relayMutex.Unlock()
+
+	if globalRelay == nil {
+		// No relay running, start one with the provided track
+		relay := NewAudioRelay()
+		config := GetAudioConfig()
+		if err := relay.Start(audioTrack, config); err != nil {
+			return err
+		}
+		globalRelay = relay
+		return nil
+	}
+
+	// Update the track in the existing relay
+	globalRelay.UpdateTrack(audioTrack)
+	return nil
+}

internal/audio/supervisor.go

@@ -0,0 +1,429 @@
+//go:build cgo
+// +build cgo
+
+package audio
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"os/exec"
+	"sync"
+	"sync/atomic"
+	"syscall"
+	"time"
+
+	"github.com/jetkvm/kvm/internal/logging"
+	"github.com/rs/zerolog"
+)
+
+const (
+	// Maximum number of restart attempts within the restart window
+	maxRestartAttempts = 5
+	// Time window for counting restart attempts
+	restartWindow = 5 * time.Minute
+	// Delay between restart attempts
+	restartDelay = 2 * time.Second
+	// Maximum restart delay (exponential backoff)
+	maxRestartDelay = 30 * time.Second
+)
+
+// AudioServerSupervisor manages the audio server subprocess lifecycle
+type AudioServerSupervisor struct {
+	ctx     context.Context
+	cancel  context.CancelFunc
+	logger  *zerolog.Logger
+	mutex   sync.RWMutex
+	running int32
+
+	// Process management
+	cmd        *exec.Cmd
+	processPID int
+
+	// Restart management
+	restartAttempts []time.Time
+	lastExitCode    int
+	lastExitTime    time.Time
+
+	// Channels for coordination
+	processDone chan struct{}
+	stopChan    chan struct{}
+
+	// Process monitoring
+	processMonitor *ProcessMonitor
+
+	// Callbacks
+	onProcessStart func(pid int)
+	onProcessExit  func(pid int, exitCode int, crashed bool)
+	onRestart      func(attempt int, delay time.Duration)
+}
+
+// NewAudioServerSupervisor creates a new audio server supervisor
+func NewAudioServerSupervisor() *AudioServerSupervisor {
+	ctx, cancel := context.WithCancel(context.Background())
+	logger := logging.GetDefaultLogger().With().Str("component", "audio-supervisor").Logger()
+
+	return &AudioServerSupervisor{
+		ctx:            ctx,
+		cancel:         cancel,
+		logger:         &logger,
+		processDone:    make(chan struct{}),
+		stopChan:       make(chan struct{}),
+		processMonitor: GetProcessMonitor(),
+	}
+}
+
+// SetCallbacks sets optional callbacks for process lifecycle events
+func (s *AudioServerSupervisor) SetCallbacks(
+	onStart func(pid int),
+	onExit func(pid int, exitCode int, crashed bool),
+	onRestart func(attempt int, delay time.Duration),
+) {
+	s.mutex.Lock()
+	defer s.mutex.Unlock()
+
+	s.onProcessStart = onStart
+	s.onProcessExit = onExit
+	s.onRestart = onRestart
+}
+
+// Start begins supervising the audio server process
+func (s *AudioServerSupervisor) Start() error {
+	if !atomic.CompareAndSwapInt32(&s.running, 0, 1) {
+		return fmt.Errorf("supervisor already running")
+	}
+
+	s.logger.Info().Msg("starting audio server supervisor")
+
+	// Start the supervision loop
+	go s.supervisionLoop()
+
+	return nil
+}
+
+// Stop gracefully stops the audio server and supervisor
+func (s *AudioServerSupervisor) Stop() error {
+	if !atomic.CompareAndSwapInt32(&s.running, 1, 0) {
+		return nil // Already stopped
+	}
+
+	s.logger.Info().Msg("stopping audio server supervisor")
+
+	// Signal stop and wait for cleanup
+	close(s.stopChan)
+	s.cancel()
+
+	// Wait for process to exit
+	select {
+	case <-s.processDone:
+		s.logger.Info().Msg("audio server process stopped gracefully")
+	case <-time.After(10 * time.Second):
+		s.logger.Warn().Msg("audio server process did not stop gracefully, forcing termination")
+		s.forceKillProcess()
+	}
+
+	return nil
+}
+
+// IsRunning returns true if the supervisor is running
+func (s *AudioServerSupervisor) IsRunning() bool {
+	return atomic.LoadInt32(&s.running) == 1
+}
+
+// GetProcessPID returns the current process PID (0 if not running)
+func (s *AudioServerSupervisor) GetProcessPID() int {
+	s.mutex.RLock()
+	defer s.mutex.RUnlock()
+	return s.processPID
+}
+
+// GetLastExitInfo returns information about the last process exit
+func (s *AudioServerSupervisor) GetLastExitInfo() (exitCode int, exitTime time.Time) {
+	s.mutex.RLock()
+	defer s.mutex.RUnlock()
+	return s.lastExitCode, s.lastExitTime
+}
+
+// GetProcessMetrics returns current process metrics if the process is running
+func (s *AudioServerSupervisor) GetProcessMetrics() *ProcessMetrics {
+	s.mutex.RLock()
+	pid := s.processPID
+	s.mutex.RUnlock()
+
+	if pid == 0 {
+		return nil
+	}
+
+	metrics := s.processMonitor.GetCurrentMetrics()
+	for _, metric := range metrics {
+		if metric.PID == pid {
+			return &metric
+		}
+	}
+	return nil
+}
+
+// supervisionLoop is the main supervision loop
+func (s *AudioServerSupervisor) supervisionLoop() {
+	defer func() {
+		close(s.processDone)
+		s.logger.Info().Msg("audio server supervision ended")
+	}()
+
+	for atomic.LoadInt32(&s.running) == 1 {
+		select {
+		case <-s.stopChan:
+			s.logger.Info().Msg("received stop signal")
+			s.terminateProcess()
+			return
+		case <-s.ctx.Done():
+			s.logger.Info().Msg("context cancelled")
+			s.terminateProcess()
+			return
+		default:
+			// Start or restart the process
+			if err := s.startProcess(); err != nil {
+				s.logger.Error().Err(err).Msg("failed to start audio server process")
+
+				// Check if we should attempt restart
+				if !s.shouldRestart() {
+					s.logger.Error().Msg("maximum restart attempts exceeded, stopping supervisor")
+					return
+				}
+
+				delay := s.calculateRestartDelay()
+				s.logger.Warn().Dur("delay", delay).Msg("retrying process start after delay")
+
+				if s.onRestart != nil {
+					s.onRestart(len(s.restartAttempts), delay)
+				}
+
+				select {
+				case <-time.After(delay):
+				case <-s.stopChan:
+					return
+				case <-s.ctx.Done():
+					return
+				}
+				continue
+			}
+
+			// Wait for process to exit
+			s.waitForProcessExit()
+
+			// Check if we should restart
+			if !s.shouldRestart() {
+				s.logger.Error().Msg("maximum restart attempts exceeded, stopping supervisor")
+				return
+			}
+
+			// Calculate restart delay
+			delay := s.calculateRestartDelay()
+			s.logger.Info().Dur("delay", delay).Msg("restarting audio server process after delay")
+
+			if s.onRestart != nil {
+				s.onRestart(len(s.restartAttempts), delay)
+			}
+
+			// Wait for restart delay
+			select {
+			case <-time.After(delay):
+			case <-s.stopChan:
+				return
+			case <-s.ctx.Done():
+				return
+			}
+		}
+	}
+}
+
+// startProcess starts the audio server process
+func (s *AudioServerSupervisor) startProcess() error {
+	execPath, err := os.Executable()
+	if err != nil {
+		return fmt.Errorf("failed to get executable path: %w", err)
+	}
+
+	s.mutex.Lock()
+	defer s.mutex.Unlock()
+
+	// Create new command
+	s.cmd = exec.CommandContext(s.ctx, execPath, "--audio-server")
+	s.cmd.Stdout = os.Stdout
+	s.cmd.Stderr = os.Stderr
+
+	// Start the process
+	if err := s.cmd.Start(); err != nil {
+		return fmt.Errorf("failed to start process: %w", err)
+	}
+
+	s.processPID = s.cmd.Process.Pid
+	s.logger.Info().Int("pid", s.processPID).Msg("audio server process started")
+
+	// Add process to monitoring
+	s.processMonitor.AddProcess(s.processPID, "audio-server")
+
+	if s.onProcessStart != nil {
+		s.onProcessStart(s.processPID)
+	}
+
+	return nil
+}
+
+// waitForProcessExit waits for the current process to exit and logs the result
+func (s *AudioServerSupervisor) waitForProcessExit() {
+	s.mutex.RLock()
+	cmd := s.cmd
+	pid := s.processPID
+	s.mutex.RUnlock()
+
+	if cmd == nil {
+		return
+	}
+
+	// Wait for process to exit
+	err := cmd.Wait()
+
+	s.mutex.Lock()
+	s.lastExitTime = time.Now()
+	s.processPID = 0
+
+	var exitCode int
+	var crashed bool
+
+	if err != nil {
+		if exitError, ok := err.(*exec.ExitError); ok {
+			exitCode = exitError.ExitCode()
+			crashed = exitCode != 0
+		} else {
+			// Process was killed or other error
+			exitCode = -1
+			crashed = true
+		}
+	} else {
+		exitCode = 0
+		crashed = false
+	}
+
+	s.lastExitCode = exitCode
+	s.mutex.Unlock()
+
+	// Remove process from monitoring
+	s.processMonitor.RemoveProcess(pid)
+
+	if crashed {
+		s.logger.Error().Int("pid", pid).Int("exit_code", exitCode).Msg("audio server process crashed")
+		s.recordRestartAttempt()
+	} else {
+		s.logger.Info().Int("pid", pid).Msg("audio server process exited gracefully")
+	}
+
+	if s.onProcessExit != nil {
+		s.onProcessExit(pid, exitCode, crashed)
+	}
+}
+
+// terminateProcess gracefully terminates the current process
+func (s *AudioServerSupervisor) terminateProcess() {
+	s.mutex.RLock()
+	cmd := s.cmd
+	pid := s.processPID
+	s.mutex.RUnlock()
+
+	if cmd == nil || cmd.Process == nil {
+		return
+	}
+
+	s.logger.Info().Int("pid", pid).Msg("terminating audio server process")
+
+	// Send SIGTERM first
+	if err := cmd.Process.Signal(syscall.SIGTERM); err != nil {
+		s.logger.Warn().Err(err).Int("pid", pid).Msg("failed to send SIGTERM")
+	}
+
+	// Wait for graceful shutdown
+	done := make(chan struct{})
+	go func() {
+		_ = cmd.Wait()
+		close(done)
+	}()
+
+	select {
+	case <-done:
+		s.logger.Info().Int("pid", pid).Msg("audio server process terminated gracefully")
+	case <-time.After(5 * time.Second):
+		s.logger.Warn().Int("pid", pid).Msg("process did not terminate gracefully, sending SIGKILL")
+		s.forceKillProcess()
+	}
+}
+
+// forceKillProcess forcefully kills the current process
+func (s *AudioServerSupervisor) forceKillProcess() {
+	s.mutex.RLock()
+	cmd := s.cmd
+	pid := s.processPID
+	s.mutex.RUnlock()
+
+	if cmd == nil || cmd.Process == nil {
+		return
+	}
+
+	s.logger.Warn().Int("pid", pid).Msg("force killing audio server process")
+	if err := cmd.Process.Kill(); err != nil {
+		s.logger.Error().Err(err).Int("pid", pid).Msg("failed to kill process")
+	}
+}
+
+// shouldRestart determines if the process should be restarted
+func (s *AudioServerSupervisor) shouldRestart() bool {
+	if atomic.LoadInt32(&s.running) == 0 {
+		return false // Supervisor is stopping
+	}
+
+	s.mutex.RLock()
+	defer s.mutex.RUnlock()
+
+	// Clean up old restart attempts outside the window
+	now := time.Now()
+	var recentAttempts []time.Time
+	for _, attempt := range s.restartAttempts {
+		if now.Sub(attempt) < restartWindow {
+			recentAttempts = append(recentAttempts, attempt)
+		}
+	}
+	s.restartAttempts = recentAttempts
+
+	return len(s.restartAttempts) < maxRestartAttempts
+}
+
+// recordRestartAttempt records a restart attempt
+func (s *AudioServerSupervisor) recordRestartAttempt() {
+	s.mutex.Lock()
+	defer s.mutex.Unlock()
+
+	s.restartAttempts = append(s.restartAttempts, time.Now())
+}
+
+// calculateRestartDelay calculates the delay before next restart attempt
+func (s *AudioServerSupervisor) calculateRestartDelay() time.Duration {
+	s.mutex.RLock()
+	defer s.mutex.RUnlock()
+
+	// Exponential backoff based on recent restart attempts
+	attempts := len(s.restartAttempts)
+	if attempts == 0 {
+		return restartDelay
+	}
+
+	// Calculate exponential backoff: 2^attempts * base delay
+	delay := restartDelay
+	for i := 0; i < attempts && delay < maxRestartDelay; i++ {
+		delay *= 2
+	}
+
+	if delay > maxRestartDelay {
+		delay = maxRestartDelay
+	}
+
+	return delay
+}

main.go

@@ -2,6 +2,7 @@ package kvm
 
 import (
 	"context"
+	"fmt"
 	"net/http"
 	"os"
 	"os/signal"
@@ -10,12 +11,127 @@ import (
 
 	"github.com/gwatts/rootcerts"
 	"github.com/jetkvm/kvm/internal/audio"
-	"github.com/pion/webrtc/v4/pkg/media"
 )
 
-var appCtx context.Context
+var (
+	appCtx           context.Context
+	isAudioServer    bool
+	audioProcessDone chan struct{}
+	audioSupervisor  *audio.AudioServerSupervisor
+)
 
-func Main() {
+func runAudioServer() {
+	logger.Info().Msg("Starting audio server subprocess")
+
+	// Create audio server
+	server, err := audio.NewAudioServer()
+	if err != nil {
+		logger.Error().Err(err).Msg("failed to create audio server")
+		os.Exit(1)
+	}
+	defer server.Close()
+
+	// Start accepting connections
+	if err := server.Start(); err != nil {
+		logger.Error().Err(err).Msg("failed to start audio server")
+		os.Exit(1)
+	}
+
+	// Initialize audio processing
+	err = audio.StartNonBlockingAudioStreaming(func(frame []byte) {
+		if err := server.SendFrame(frame); err != nil {
+			logger.Warn().Err(err).Msg("failed to send audio frame")
+			audio.RecordFrameDropped()
+		}
+	})
+	if err != nil {
+		logger.Error().Err(err).Msg("failed to start audio processing")
+		os.Exit(1)
+	}
+
+	// Wait for termination signal
+	sigs := make(chan os.Signal, 1)
+	signal.Notify(sigs, syscall.SIGINT, syscall.SIGTERM)
+	<-sigs
+
+	// Cleanup
+	audio.StopNonBlockingAudioStreaming()
+	logger.Info().Msg("Audio server subprocess stopped")
+}
+
+func startAudioSubprocess() error {
+	// Create audio server supervisor
+	audioSupervisor = audio.NewAudioServerSupervisor()
+
+	// Set up callbacks for process lifecycle events
+	audioSupervisor.SetCallbacks(
+		// onProcessStart
+		func(pid int) {
+			logger.Info().Int("pid", pid).Msg("audio server process started")
+
+			// Start audio relay system for main process without a track initially
+			// The track will be updated when a WebRTC session is created
+			if err := audio.StartAudioRelay(nil); err != nil {
+				logger.Error().Err(err).Msg("failed to start audio relay")
+			}
+		},
+		// onProcessExit
+		func(pid int, exitCode int, crashed bool) {
+			if crashed {
+				logger.Error().Int("pid", pid).Int("exit_code", exitCode).Msg("audio server process crashed")
+			} else {
+				logger.Info().Int("pid", pid).Msg("audio server process exited gracefully")
+			}
+
+			// Stop audio relay when process exits
+			audio.StopAudioRelay()
+		},
+		// onRestart
+		func(attempt int, delay time.Duration) {
+			logger.Warn().Int("attempt", attempt).Dur("delay", delay).Msg("restarting audio server process")
+		},
+	)
+
+	// Start the supervisor
+	if err := audioSupervisor.Start(); err != nil {
+		return fmt.Errorf("failed to start audio supervisor: %w", err)
+	}
+
+	// Monitor supervisor and handle cleanup
+	go func() {
+		defer close(audioProcessDone)
+
+		// Wait for supervisor to stop
+		for audioSupervisor.IsRunning() {
+			time.Sleep(100 * time.Millisecond)
+		}
+
+		logger.Info().Msg("audio supervisor stopped")
+	}()
+
+	return nil
+}
+
+func Main(audioServer bool) {
+	// Initialize channel and set audio server flag
+	isAudioServer = audioServer
+	audioProcessDone = make(chan struct{})
+
+	// If running as audio server, only initialize audio processing
+	if isAudioServer {
+		runAudioServer()
+		return
+	}
+
+	// If running as audio input server, only initialize audio input processing
+	if audio.IsAudioInputServerProcess() {
+		err := audio.RunAudioInputServer()
+		if err != nil {
+			logger.Error().Err(err).Msg("audio input server failed")
+			os.Exit(1)
+		}
+		return
+	}
 	LoadConfig()
 
 	var cancel context.CancelFunc
@@ -80,30 +196,10 @@ func Main() {
 	// initialize usb gadget
 	initUsbGadget()
 
-	// Start non-blocking audio streaming and deliver Opus frames to WebRTC
-	err = audio.StartNonBlockingAudioStreaming(func(frame []byte) {
-		// Deliver Opus frame to WebRTC audio track if session is active
-		if currentSession != nil {
-			config := audio.GetAudioConfig()
-			var sampleData []byte
-			if audio.IsAudioMuted() {
-				sampleData = make([]byte, len(frame)) // silence
-			} else {
-				sampleData = frame
-			}
-			if err := currentSession.AudioTrack.WriteSample(media.Sample{
-				Data:     sampleData,
-				Duration: config.FrameSize,
-			}); err != nil {
-				logger.Warn().Err(err).Msg("error writing audio sample")
-				audio.RecordFrameDropped()
-			}
-		} else {
-			audio.RecordFrameDropped()
-		}
-	})
+	// Start audio subprocess
+	err = startAudioSubprocess()
 	if err != nil {
-		logger.Warn().Err(err).Msg("failed to start non-blocking audio streaming")
+		logger.Warn().Err(err).Msg("failed to start audio subprocess")
 	}
 
 	// Initialize session provider for audio events
@@ -163,8 +259,18 @@ func Main() {
 	<-sigs
 	logger.Info().Msg("JetKVM Shutting Down")
 
-	// Stop non-blocking audio manager
-	audio.StopNonBlockingAudioStreaming()
+	// Stop audio subprocess and wait for cleanup
+	if !isAudioServer {
+		if audioSupervisor != nil {
+			logger.Info().Msg("stopping audio supervisor")
+			if err := audioSupervisor.Stop(); err != nil {
+				logger.Error().Err(err).Msg("failed to stop audio supervisor")
+			}
+		}
+		<-audioProcessDone
+	} else {
+		audio.StopNonBlockingAudioStreaming()
+	}
 	//if fuseServer != nil {
 	//	err := setMassStorageImage(" ")
 	//	if err != nil {

prometheus.go

@@ -1,6 +1,7 @@
 package kvm
 
 import (
+	"github.com/jetkvm/kvm/internal/audio"
 	"github.com/prometheus/client_golang/prometheus"
 	versioncollector "github.com/prometheus/client_golang/prometheus/collectors/version"
 	"github.com/prometheus/common/version"
@@ -10,4 +11,7 @@ func initPrometheus() {
 	// A Prometheus metrics endpoint.
 	version.Version = builtAppVersion
 	prometheus.MustRegister(versioncollector.NewCollector("jetkvm"))
+
+	// Start audio metrics collection
+	audio.StartMetricsUpdater()
 }

resource/dev_test.sh

@@ -1,4 +1,4 @@
-#!/bin/sh
+#!/bin/bash
 JSON_OUTPUT=false
 GET_COMMANDS=false
 if [ "$1" = "-json" ]; then

ui/src/components/ActionBar.tsx

@@ -2,7 +2,7 @@ import { MdOutlineContentPasteGo, MdVolumeOff, MdVolumeUp, MdGraphicEq } from "r
 import { LuCable, LuHardDrive, LuMaximize, LuSettings, LuSignal } from "react-icons/lu";
 import { FaKeyboard } from "react-icons/fa6";
 import { Popover, PopoverButton, PopoverPanel } from "@headlessui/react";
-import { Fragment, useCallback, useEffect, useRef, useState } from "react";
+import { Fragment, useCallback, useRef } from "react";
 import { CommandLineIcon } from "@heroicons/react/20/solid";
 
 import { Button } from "@components/Button";
@@ -21,7 +21,7 @@ import ExtensionPopover from "@/components/popovers/ExtensionPopover";
 import AudioControlPopover from "@/components/popovers/AudioControlPopover";
 import { useDeviceUiNavigation } from "@/hooks/useAppNavigation";
 import { useAudioEvents } from "@/hooks/useAudioEvents";
-import api from "@/api";
+
 
 // Type for microphone error
 interface MicrophoneError {
@@ -83,35 +83,10 @@ export default function Actionbar({
   );
 
   // Use WebSocket-based audio events for real-time updates
-  const { audioMuted, isConnected } = useAudioEvents();
+  const { audioMuted } = useAudioEvents();
   
-  // Fallback to polling if WebSocket is not connected
-  const [fallbackMuted, setFallbackMuted] = useState(false);
-  useEffect(() => {
-    if (!isConnected) {
-      // Load initial state
-      api.GET("/audio/mute").then(async resp => {
-        if (resp.ok) {
-          const data = await resp.json();
-          setFallbackMuted(!!data.muted);
-        }
-      });
-      
-      // Fallback polling when WebSocket is not available
-      const interval = setInterval(async () => {
-        const resp = await api.GET("/audio/mute");
-        if (resp.ok) {
-          const data = await resp.json();
-          setFallbackMuted(!!data.muted);
-        }
-      }, 1000);
-      
-      return () => clearInterval(interval);
-    }
-  }, [isConnected]);
-  
-  // Use WebSocket data when available, fallback to polling data otherwise
-  const isMuted = isConnected && audioMuted !== null ? audioMuted : fallbackMuted;
+  // Use WebSocket data exclusively - no polling fallback
+  const isMuted = audioMuted ?? false; // Default to false if WebSocket data not available yet
 
   return (
     <Container className="border-b border-b-slate-800/20 bg-white dark:border-b-slate-300/20 dark:bg-slate-900">

ui/src/components/AudioMetricsDashboard.tsx

@@ -1,6 +1,6 @@
 import { useEffect, useState } from "react";
 import { MdGraphicEq, MdSignalWifi4Bar, MdError, MdMic } from "react-icons/md";
-import { LuActivity, LuClock, LuHardDrive, LuSettings } from "react-icons/lu";
+import { LuActivity, LuClock, LuHardDrive, LuSettings, LuCpu, LuMemoryStick } from "react-icons/lu";
 
 import { AudioLevelMeter } from "@components/AudioLevelMeter";
 import { cx } from "@/cva.config";
@@ -27,6 +27,14 @@ interface MicrophoneMetrics {
   average_latency: string;
 }
 
+interface ProcessMetrics {
+  cpu_percent: number;
+  memory_percent: number;
+  memory_rss: number;
+  memory_vms: number;
+  running: boolean;
+}
+
 interface AudioConfig {
   Quality: number;
   Bitrate: number;
@@ -55,6 +63,16 @@ export default function AudioMetricsDashboard() {
   const [fallbackMicrophoneMetrics, setFallbackMicrophoneMetrics] = useState<MicrophoneMetrics | null>(null);
   const [fallbackConnected, setFallbackConnected] = useState(false);
   
+  // Process metrics state
+  const [audioProcessMetrics, setAudioProcessMetrics] = useState<ProcessMetrics | null>(null);
+  const [microphoneProcessMetrics, setMicrophoneProcessMetrics] = useState<ProcessMetrics | null>(null);
+  
+  // Historical data for histograms (last 60 data points, ~1 minute at 1s intervals)
+  const [audioCpuHistory, setAudioCpuHistory] = useState<number[]>([]);
+  const [audioMemoryHistory, setAudioMemoryHistory] = useState<number[]>([]);
+  const [micCpuHistory, setMicCpuHistory] = useState<number[]>([]);
+  const [micMemoryHistory, setMicMemoryHistory] = useState<number[]>([]);
+  
   // Configuration state (these don't change frequently, so we can load them once)
   const [config, setConfig] = useState<AudioConfig | null>(null);
   const [microphoneConfig, setMicrophoneConfig] = useState<AudioConfig | null>(null);
@@ -124,6 +142,29 @@ export default function AudioMetricsDashboard() {
         setFallbackConnected(false);
       }
 
+      // Load audio process metrics
+      try {
+        const audioProcessResp = await api.GET("/audio/process-metrics");
+        if (audioProcessResp.ok) {
+          const audioProcessData = await audioProcessResp.json();
+          setAudioProcessMetrics(audioProcessData);
+          
+          // Update historical data for histograms (keep last 60 points)
+          if (audioProcessData.running) {
+            setAudioCpuHistory(prev => {
+              const newHistory = [...prev, audioProcessData.cpu_percent];
+              return newHistory.slice(-60); // Keep last 60 data points
+            });
+            setAudioMemoryHistory(prev => {
+              const newHistory = [...prev, audioProcessData.memory_percent];
+              return newHistory.slice(-60);
+            });
+          }
+        }
+      } catch (audioProcessError) {
+        console.debug("Audio process metrics not available:", audioProcessError);
+      }
+
       // Load microphone metrics
       try {
         const micResp = await api.GET("/microphone/metrics");
@@ -135,6 +176,29 @@ export default function AudioMetricsDashboard() {
         // Microphone metrics might not be available, that's okay
         console.debug("Microphone metrics not available:", micError);
       }
+
+      // Load microphone process metrics
+      try {
+        const micProcessResp = await api.GET("/microphone/process-metrics");
+        if (micProcessResp.ok) {
+          const micProcessData = await micProcessResp.json();
+          setMicrophoneProcessMetrics(micProcessData);
+          
+          // Update historical data for histograms (keep last 60 points)
+          if (micProcessData.running) {
+            setMicCpuHistory(prev => {
+              const newHistory = [...prev, micProcessData.cpu_percent];
+              return newHistory.slice(-60); // Keep last 60 data points
+            });
+            setMicMemoryHistory(prev => {
+              const newHistory = [...prev, micProcessData.memory_percent];
+              return newHistory.slice(-60);
+            });
+          }
+        }
+      } catch (micProcessError) {
+        console.debug("Microphone process metrics not available:", micProcessError);
+      }
     } catch (error) {
       console.error("Failed to load audio data:", error);
       setFallbackConnected(false);
@@ -158,6 +222,18 @@ export default function AudioMetricsDashboard() {
     return ((metrics.frames_dropped / metrics.frames_received) * 100);
   };
 
+  const formatMemory = (bytes: number) => {
+    if (bytes === 0) return "0 MB";
+    const mb = bytes / (1024 * 1024);
+    if (mb < 1024) {
+      return `${mb.toFixed(1)} MB`;
+    }
+    const gb = mb / 1024;
+    return `${gb.toFixed(2)} GB`;
+  };
+
+
+
   const getQualityColor = (quality: number) => {
     switch (quality) {
       case 0: return "text-yellow-600 dark:text-yellow-400";
@@ -168,6 +244,53 @@ export default function AudioMetricsDashboard() {
     }
   };
 
+  // Histogram component for displaying historical data
+  const Histogram = ({ data, title, unit, color }: { 
+    data: number[], 
+    title: string, 
+    unit: string, 
+    color: string 
+  }) => {
+    if (data.length === 0) return null;
+    
+    const maxValue = Math.max(...data, 1); // Avoid division by zero
+    const minValue = Math.min(...data);
+    const range = maxValue - minValue;
+    
+    return (
+      <div className="space-y-2">
+        <div className="flex items-center justify-between">
+          <span className="text-sm font-medium text-slate-700 dark:text-slate-300">
+            {title}
+          </span>
+          <span className="text-xs text-slate-500 dark:text-slate-400">
+            {data.length > 0 ? `${data[data.length - 1].toFixed(1)}${unit}` : `0${unit}`}
+          </span>
+        </div>
+        <div className="flex items-end gap-0.5 h-16 bg-slate-50 dark:bg-slate-800 rounded p-2">
+          {data.slice(-30).map((value, index) => { // Show last 30 points
+            const height = range > 0 ? ((value - minValue) / range) * 100 : 0;
+            return (
+              <div
+                key={index}
+                className={cx(
+                  "flex-1 rounded-sm transition-all duration-200",
+                  color
+                )}
+                style={{ height: `${Math.max(height, 2)}%` }}
+                title={`${value.toFixed(1)}${unit}`}
+              />
+            );
+          })}
+        </div>
+        <div className="flex justify-between text-xs text-slate-400 dark:text-slate-500">
+          <span>{minValue.toFixed(1)}{unit}</span>
+          <span>{maxValue.toFixed(1)}{unit}</span>
+        </div>
+      </div>
+    );
+  };
+
   return (
     <div className="space-y-4">
       {/* Header */}
@@ -266,6 +389,97 @@ export default function AudioMetricsDashboard() {
         )}
       </div>
 
+      {/* Subprocess Resource Usage - Histogram View */}
+      <div className="grid grid-cols-1 md:grid-cols-2 gap-4">
+        {/* Audio Output Subprocess */}
+        {audioProcessMetrics && (
+          <div className="rounded-lg border border-slate-200 p-3 dark:border-slate-700">
+            <div className="mb-3 flex items-center gap-2">
+              <LuCpu className="h-4 w-4 text-blue-600 dark:text-blue-400" />
+              <span className="font-medium text-slate-900 dark:text-slate-100">
+                Audio Output Process
+              </span>
+              <div className={cx(
+                "h-2 w-2 rounded-full ml-auto",
+                audioProcessMetrics.running ? "bg-green-500" : "bg-red-500"
+              )} />
+            </div>
+            <div className="space-y-4">
+              <Histogram 
+                data={audioCpuHistory} 
+                title="CPU Usage" 
+                unit="%" 
+                color="bg-blue-500 dark:bg-blue-400" 
+              />
+              <Histogram 
+                data={audioMemoryHistory} 
+                title="Memory Usage" 
+                unit="%" 
+                color="bg-purple-500 dark:bg-purple-400" 
+              />
+              <div className="grid grid-cols-2 gap-2 text-xs">
+                <div className="text-center p-2 bg-slate-50 dark:bg-slate-800 rounded">
+                  <div className="font-medium text-slate-900 dark:text-slate-100">
+                    {formatMemory(audioProcessMetrics.memory_rss)}
+                  </div>
+                  <div className="text-slate-500 dark:text-slate-400">RSS</div>
+                </div>
+                <div className="text-center p-2 bg-slate-50 dark:bg-slate-800 rounded">
+                  <div className="font-medium text-slate-900 dark:text-slate-100">
+                    {formatMemory(audioProcessMetrics.memory_vms)}
+                  </div>
+                  <div className="text-slate-500 dark:text-slate-400">VMS</div>
+                </div>
+              </div>
+            </div>
+          </div>
+        )}
+
+        {/* Microphone Input Subprocess */}
+        {microphoneProcessMetrics && (
+          <div className="rounded-lg border border-slate-200 p-3 dark:border-slate-700">
+            <div className="mb-3 flex items-center gap-2">
+              <LuMemoryStick className="h-4 w-4 text-green-600 dark:text-green-400" />
+              <span className="font-medium text-slate-900 dark:text-slate-100">
+                Microphone Input Process
+              </span>
+              <div className={cx(
+                "h-2 w-2 rounded-full ml-auto",
+                microphoneProcessMetrics.running ? "bg-green-500" : "bg-red-500"
+              )} />
+            </div>
+            <div className="space-y-4">
+              <Histogram 
+                data={micCpuHistory} 
+                title="CPU Usage" 
+                unit="%" 
+                color="bg-green-500 dark:bg-green-400" 
+              />
+              <Histogram 
+                data={micMemoryHistory} 
+                title="Memory Usage" 
+                unit="%" 
+                color="bg-orange-500 dark:bg-orange-400" 
+              />
+              <div className="grid grid-cols-2 gap-2 text-xs">
+                <div className="text-center p-2 bg-slate-50 dark:bg-slate-800 rounded">
+                  <div className="font-medium text-slate-900 dark:text-slate-100">
+                    {formatMemory(microphoneProcessMetrics.memory_rss)}
+                  </div>
+                  <div className="text-slate-500 dark:text-slate-400">RSS</div>
+                </div>
+                <div className="text-center p-2 bg-slate-50 dark:bg-slate-800 rounded">
+                  <div className="font-medium text-slate-900 dark:text-slate-100">
+                    {formatMemory(microphoneProcessMetrics.memory_vms)}
+                  </div>
+                  <div className="text-slate-500 dark:text-slate-400">VMS</div>
+                </div>
+              </div>
+            </div>
+          </div>
+        )}
+      </div>
+
       {/* Performance Metrics */}
       {metrics && (
         <div className="space-y-3">
@@ -412,6 +626,41 @@ export default function AudioMetricsDashboard() {
                   />
                 </div>
               )}
+              
+              {/* Microphone Connection Health */}
+              <div className="mt-3 rounded-md bg-slate-50 p-2 dark:bg-slate-700">
+                <div className="mb-2 flex items-center gap-2">
+                  <MdSignalWifi4Bar className="h-3 w-3 text-purple-600 dark:text-purple-400" />
+                  <span className="text-sm font-medium text-slate-900 dark:text-slate-100">
+                    Connection Health
+                  </span>
+                </div>
+                <div className="space-y-2">
+                  <div className="flex justify-between">
+                    <span className="text-xs text-slate-500 dark:text-slate-400">
+                      Connection Drops:
+                    </span>
+                    <span className={cx(
+                      "text-xs font-medium",
+                      microphoneMetrics.connection_drops > 0 
+                        ? "text-red-600 dark:text-red-400" 
+                        : "text-green-600 dark:text-green-400"
+                    )}>
+                      {formatNumber(microphoneMetrics.connection_drops)}
+                    </span>
+                  </div>
+                  {microphoneMetrics.average_latency && (
+                    <div className="flex justify-between">
+                      <span className="text-xs text-slate-500 dark:text-slate-400">
+                        Avg Latency:
+                      </span>
+                      <span className="text-xs font-medium text-slate-900 dark:text-slate-100">
+                        {microphoneMetrics.average_latency}
+                      </span>
+                    </div>
+                  )}
+                </div>
+              </div>
             </div>
           )}
 

ui/src/components/popovers/AudioControlPopover.tsx

@@ -41,23 +41,7 @@ interface AudioConfig {
   FrameSize: string;
 }
 
-interface AudioMetrics {
-  frames_received: number;
-  frames_dropped: number;
-  bytes_processed: number;
-  last_frame_time: string;
-  connection_drops: number;
-  average_latency: string;
-}
 
-interface MicrophoneMetrics {
-  frames_sent: number;
-  frames_dropped: number;
-  bytes_processed: number;
-  last_frame_time: string;
-  connection_drops: number;
-  average_latency: string;
-}
 
 
 
@@ -94,11 +78,7 @@ export default function AudioControlPopover({ microphone, open }: AudioControlPo
     isConnected: wsConnected 
   } = useAudioEvents();
   
-  // Fallback state for when WebSocket is not connected
-  const [fallbackMuted, setFallbackMuted] = useState(false);
-  const [fallbackMetrics, setFallbackMetrics] = useState<AudioMetrics | null>(null);
-  const [fallbackMicMetrics, setFallbackMicMetrics] = useState<MicrophoneMetrics | null>(null);
-  const [fallbackConnected, setFallbackConnected] = useState(false);
+  // WebSocket-only implementation - no fallback polling
   
   // Microphone state from props
   const {
@@ -115,11 +95,11 @@ export default function AudioControlPopover({ microphone, open }: AudioControlPo
     isToggling,
   } = microphone;
   
-  // Use WebSocket data when available, fallback to polling data otherwise
-  const isMuted = wsConnected && audioMuted !== null ? audioMuted : fallbackMuted;
-  const metrics = wsConnected && audioMetrics !== null ? audioMetrics : fallbackMetrics;
-  const micMetrics = wsConnected && microphoneMetrics !== null ? microphoneMetrics : fallbackMicMetrics;
-  const isConnected = wsConnected ? wsConnected : fallbackConnected;
+  // Use WebSocket data exclusively - no polling fallback
+  const isMuted = audioMuted ?? false;
+  const metrics = audioMetrics;
+  const micMetrics = microphoneMetrics;
+  const isConnected = wsConnected;
   
   // Audio level monitoring - enable only when popover is open and microphone is active to save resources
   const analysisEnabled = (open ?? true) && isMicrophoneActive;
@@ -150,34 +130,15 @@ export default function AudioControlPopover({ microphone, open }: AudioControlPo
     }
   }, [configsLoaded]);
 
-  // Optimize fallback polling - only run when WebSocket is not connected
+  // WebSocket-only implementation - sync microphone state when needed
   useEffect(() => {
-    if (!wsConnected && !configsLoaded) {
-      // Load state once if configs aren't loaded yet
-      loadAudioState();
-    }
-    
-    if (!wsConnected) {
-      loadAudioMetrics();
-      loadMicrophoneMetrics();
-      
-      // Reduced frequency for fallback polling (every 3 seconds instead of 2)
-      const metricsInterval = setInterval(() => {
-        if (!wsConnected) { // Double-check to prevent unnecessary requests
-          loadAudioMetrics();
-          loadMicrophoneMetrics();
-        }
-      }, 3000);
-      return () => clearInterval(metricsInterval);
-    }
-    
     // Always sync microphone state, but debounce it
     const syncTimeout = setTimeout(() => {
       syncMicrophoneState();
     }, 500);
     
     return () => clearTimeout(syncTimeout);
-  }, [wsConnected, syncMicrophoneState, configsLoaded]);
+  }, [syncMicrophoneState]);
 
   const loadAudioConfigurations = async () => {
     try {
@@ -203,60 +164,14 @@ export default function AudioControlPopover({ microphone, open }: AudioControlPo
     }
   };
 
-  const loadAudioState = async () => {
-    try {
-      // Load mute state only (configurations are loaded separately)
-      const muteResp = await api.GET("/audio/mute");
-      if (muteResp.ok) {
-        const muteData = await muteResp.json();
-        setFallbackMuted(!!muteData.muted);
-      }
-    } catch (error) {
-      console.error("Failed to load audio state:", error);
-    }
-  };
-
-  const loadAudioMetrics = async () => {
-    try {
-      const resp = await api.GET("/audio/metrics");
-      if (resp.ok) {
-        const data = await resp.json();
-        setFallbackMetrics(data);
-        // Consider connected if API call succeeds, regardless of frame count
-        setFallbackConnected(true);
-      } else {
-        setFallbackConnected(false);
-      }
-    } catch (error) {
-      console.error("Failed to load audio metrics:", error);
-      setFallbackConnected(false);
-    }
-  };
-
-
-
-  const loadMicrophoneMetrics = async () => {
-    try {
-      const resp = await api.GET("/microphone/metrics");
-      if (resp.ok) {
-        const data = await resp.json();
-        setFallbackMicMetrics(data);
-      }
-    } catch (error) {
-      console.error("Failed to load microphone metrics:", error);
-    }
-  };
-
   const handleToggleMute = async () => {
     setIsLoading(true);
     try {
       const resp = await api.POST("/audio/mute", { muted: !isMuted });
-      if (resp.ok) {
-        // WebSocket will handle the state update, but update fallback for immediate feedback
-        if (!wsConnected) {
-          setFallbackMuted(!isMuted);
-        }
+      if (!resp.ok) {
+        console.error("Failed to toggle mute:", resp.statusText);
       }
+      // WebSocket will handle the state update automatically
     } catch (error) {
       console.error("Failed to toggle mute:", error);
     } finally {

ui/src/hooks/useAudioEvents.ts

@@ -62,7 +62,7 @@ export interface UseAudioEventsReturn {
 }
 
 // Global subscription management to prevent multiple subscriptions per WebSocket connection
-let globalSubscriptionState = {
+const globalSubscriptionState = {
   isSubscribed: false,
   subscriberCount: 0,
   connectionId: null as string | null

ui/src/hooks/useMicrophone.ts

@@ -858,11 +858,15 @@ export function useMicrophone() {
   }, [microphoneSender, peerConnection]);
 
   const startMicrophoneDebounced = useCallback((deviceId?: string) => {
-    debouncedOperation(() => startMicrophone(deviceId).then(() => {}), "start");
+    debouncedOperation(async () => {
+      await startMicrophone(deviceId).catch(console.error);
+    }, "start");
   }, [startMicrophone, debouncedOperation]);
 
   const stopMicrophoneDebounced = useCallback(() => {
-    debouncedOperation(() => stopMicrophone().then(() => {}), "stop");
+    debouncedOperation(async () => {
+      await stopMicrophone().catch(console.error);
+    }, "stop");
   }, [stopMicrophone, debouncedOperation]);
 
   // Make debug functions available globally for console access

web.go

@@ -159,10 +159,6 @@ func setupRouter() *gin.Engine {
 		protected.POST("/storage/upload", handleUploadHttp)
 	}
 
-	protected.GET("/audio/mute", func(c *gin.Context) {
-		c.JSON(200, gin.H{"muted": audio.IsAudioMuted()})
-	})
-
 	protected.POST("/audio/mute", func(c *gin.Context) {
 		type muteReq struct {
 			Muted bool `json:"muted"`
@@ -173,6 +169,8 @@ func setupRouter() *gin.Engine {
 			return
 		}
 		audio.SetAudioMuted(req.Muted)
+		// Also set relay mute state if in main process
+		audio.SetAudioRelayMuted(req.Muted)
 
 		// Broadcast audio mute state change via WebSocket
 		broadcaster := audio.GetAudioEventBroadcaster()
@@ -221,7 +219,7 @@ func setupRouter() *gin.Engine {
 			"bytes_processed":  metrics.BytesProcessed,
 			"last_frame_time":  metrics.LastFrameTime,
 			"connection_drops": metrics.ConnectionDrops,
-			"average_latency":  metrics.AverageLatency.String(),
+			"average_latency":  fmt.Sprintf("%.1fms", float64(metrics.AverageLatency.Nanoseconds())/1e6),
 		})
 	})
 
@@ -286,18 +284,18 @@ func setupRouter() *gin.Engine {
 		// Optimized server-side cooldown using atomic operations
 		opResult := audio.TryMicrophoneOperation()
 		if !opResult.Allowed {
-			running := currentSession.AudioInputManager.IsRunning() || audio.IsNonBlockingAudioInputRunning()
+			running := currentSession.AudioInputManager.IsRunning()
 			c.JSON(200, gin.H{
-				"status":                 "cooldown",
-				"running":                running,
-				"cooldown_ms_remaining":  opResult.RemainingCooldown.Milliseconds(),
-				"operation_id":           opResult.OperationID,
+				"status":                "cooldown",
+				"running":               running,
+				"cooldown_ms_remaining": opResult.RemainingCooldown.Milliseconds(),
+				"operation_id":          opResult.OperationID,
 			})
 			return
 		}
 
 		// Check if already running before attempting to start
-		if currentSession.AudioInputManager.IsRunning() || audio.IsNonBlockingAudioInputRunning() {
+		if currentSession.AudioInputManager.IsRunning() {
 			c.JSON(200, gin.H{
 				"status":  "already running",
 				"running": true,
@@ -312,7 +310,7 @@ func setupRouter() *gin.Engine {
 
 			// Check if it's already running after the failed start attempt
 			// This handles race conditions where another request started it
-			if currentSession.AudioInputManager.IsRunning() || audio.IsNonBlockingAudioInputRunning() {
+			if currentSession.AudioInputManager.IsRunning() {
 				c.JSON(200, gin.H{
 					"status":  "started by concurrent request",
 					"running": true,
@@ -348,18 +346,18 @@ func setupRouter() *gin.Engine {
 		// Optimized server-side cooldown using atomic operations
 		opResult := audio.TryMicrophoneOperation()
 		if !opResult.Allowed {
-			running := currentSession.AudioInputManager.IsRunning() || audio.IsNonBlockingAudioInputRunning()
+			running := currentSession.AudioInputManager.IsRunning()
 			c.JSON(200, gin.H{
-				"status":                 "cooldown",
-				"running":                running,
-				"cooldown_ms_remaining":  opResult.RemainingCooldown.Milliseconds(),
-				"operation_id":           opResult.OperationID,
+				"status":                "cooldown",
+				"running":               running,
+				"cooldown_ms_remaining": opResult.RemainingCooldown.Milliseconds(),
+				"operation_id":          opResult.OperationID,
 			})
 			return
 		}
 
 		// Check if already stopped before attempting to stop
-		if !currentSession.AudioInputManager.IsRunning() && !audio.IsNonBlockingAudioInputRunning() {
+		if !currentSession.AudioInputManager.IsRunning() {
 			c.JSON(200, gin.H{
 				"status":  "already stopped",
 				"running": false,
@@ -369,7 +367,7 @@ func setupRouter() *gin.Engine {
 
 		currentSession.AudioInputManager.Stop()
 
-		// AudioInputManager.Stop() already coordinates a clean stop via StopNonBlockingAudioInput()
+		// AudioInputManager.Stop() already coordinates a clean stop via IPC audio input system
 		// so we don't need to call it again here
 
 		// Broadcast microphone state change via WebSocket
@@ -408,7 +406,7 @@ func setupRouter() *gin.Engine {
 				"bytes_processed":  0,
 				"last_frame_time":  "",
 				"connection_drops": 0,
-				"average_latency":  "0s",
+				"average_latency":  "0.0ms",
 			})
 			return
 		}
@@ -420,7 +418,88 @@ func setupRouter() *gin.Engine {
 			"bytes_processed":  metrics.BytesProcessed,
 			"last_frame_time":  metrics.LastFrameTime.Format("2006-01-02T15:04:05.000Z"),
 			"connection_drops": metrics.ConnectionDrops,
-			"average_latency":  metrics.AverageLatency.String(),
+			"average_latency":  fmt.Sprintf("%.1fms", float64(metrics.AverageLatency.Nanoseconds())/1e6),
+		})
+	})
+
+	// Audio subprocess process metrics endpoints
+	protected.GET("/audio/process-metrics", func(c *gin.Context) {
+		// Access the global audio supervisor from main.go
+		if audioSupervisor == nil {
+			c.JSON(200, gin.H{
+				"cpu_percent":    0.0,
+				"memory_percent": 0.0,
+				"memory_rss":     0,
+				"memory_vms":     0,
+				"running":        false,
+			})
+			return
+		}
+
+		metrics := audioSupervisor.GetProcessMetrics()
+		if metrics == nil {
+			c.JSON(200, gin.H{
+				"cpu_percent":    0.0,
+				"memory_percent": 0.0,
+				"memory_rss":     0,
+				"memory_vms":     0,
+				"running":        false,
+			})
+			return
+		}
+
+		c.JSON(200, gin.H{
+			"cpu_percent":    metrics.CPUPercent,
+			"memory_percent": metrics.MemoryPercent,
+			"memory_rss":     metrics.MemoryRSS,
+			"memory_vms":     metrics.MemoryVMS,
+			"running":        true,
+		})
+	})
+
+	protected.GET("/microphone/process-metrics", func(c *gin.Context) {
+		if currentSession == nil || currentSession.AudioInputManager == nil {
+			c.JSON(200, gin.H{
+				"cpu_percent":    0.0,
+				"memory_percent": 0.0,
+				"memory_rss":     0,
+				"memory_vms":     0,
+				"running":        false,
+			})
+			return
+		}
+
+		// Get the supervisor from the audio input manager
+		supervisor := currentSession.AudioInputManager.GetSupervisor()
+		if supervisor == nil {
+			c.JSON(200, gin.H{
+				"cpu_percent":    0.0,
+				"memory_percent": 0.0,
+				"memory_rss":     0,
+				"memory_vms":     0,
+				"running":        false,
+			})
+			return
+		}
+
+		metrics := supervisor.GetProcessMetrics()
+		if metrics == nil {
+			c.JSON(200, gin.H{
+				"cpu_percent":    0.0,
+				"memory_percent": 0.0,
+				"memory_rss":     0,
+				"memory_vms":     0,
+				"running":        false,
+			})
+			return
+		}
+
+		c.JSON(200, gin.H{
+			"cpu_percent":    metrics.CPUPercent,
+			"memory_percent": metrics.MemoryPercent,
+			"memory_rss":     metrics.MemoryRSS,
+			"memory_vms":     metrics.MemoryVMS,
+			"running":        true,
 		})
 	})
 
@@ -437,9 +516,8 @@ func setupRouter() *gin.Engine {
 
 		logger.Info().Msg("forcing microphone state reset")
 
-		// Force stop both the AudioInputManager and NonBlockingAudioManager
+		// Force stop the AudioInputManager
 		currentSession.AudioInputManager.Stop()
-		audio.StopNonBlockingAudioInput()
 
 		// Wait a bit to ensure everything is stopped
 		time.Sleep(100 * time.Millisecond)
@@ -449,9 +527,8 @@ func setupRouter() *gin.Engine {
 		broadcaster.BroadcastMicrophoneStateChanged(false, true)
 
 		c.JSON(200, gin.H{
-			"status":                    "reset",
-			"audio_input_running":       currentSession.AudioInputManager.IsRunning(),
-			"nonblocking_input_running": audio.IsNonBlockingAudioInputRunning(),
+			"status":              "reset",
+			"audio_input_running": currentSession.AudioInputManager.IsRunning(),
 		})
 	})
 

webrtc.go

@@ -30,10 +30,13 @@ type Session struct {
 	AudioInputManager        *audio.AudioInputManager
 	shouldUmountVirtualMedia bool
 
-	// Microphone operation cooldown to mitigate rapid start/stop races
-	micOpMu          sync.Mutex
-	lastMicOp        time.Time
-	micCooldown      time.Duration
+	// Microphone operation throttling
+	micCooldown time.Duration
+
+	// Audio frame processing
+	audioFrameChan chan []byte
+	audioStopChan  chan struct{}
+	audioWg        sync.WaitGroup
 }
 
 type SessionConfig struct {
@@ -118,8 +121,14 @@ func newSession(config SessionConfig) (*Session, error) {
 	session := &Session{
 		peerConnection:    peerConnection,
 		AudioInputManager: audio.NewAudioInputManager(),
+		micCooldown:       100 * time.Millisecond,
+		audioFrameChan:    make(chan []byte, 1000),
+		audioStopChan:     make(chan struct{}),
 	}
 
+	// Start audio processing goroutine
+	session.startAudioProcessor(*logger)
+
 	peerConnection.OnDataChannel(func(d *webrtc.DataChannel) {
 		scopedLogger.Info().Str("label", d.Label()).Uint16("id", *d.ID()).Msg("New DataChannel")
 		switch d.Label() {
@@ -155,6 +164,11 @@ func newSession(config SessionConfig) (*Session, error) {
 		return nil, err
 	}
 
+	// Update the audio relay with the new WebRTC audio track
+	if err := audio.UpdateAudioRelayTrack(session.AudioTrack); err != nil {
+		scopedLogger.Warn().Err(err).Msg("Failed to update audio relay track")
+	}
+
 	videoRtpSender, err := peerConnection.AddTrack(session.VideoTrack)
 	if err != nil {
 		return nil, err
@@ -190,10 +204,14 @@ func newSession(config SessionConfig) (*Session, error) {
 
 					// Extract Opus payload from RTP packet
 					opusPayload := rtpPacket.Payload
-					if len(opusPayload) > 0 && session.AudioInputManager != nil {
-						err := session.AudioInputManager.WriteOpusFrame(opusPayload)
-						if err != nil {
-							scopedLogger.Warn().Err(err).Msg("Failed to write Opus frame to audio input manager")
+					if len(opusPayload) > 0 {
+						// Send to buffered channel for processing
+						select {
+						case session.audioFrameChan <- opusPayload:
+							// Frame sent successfully
+						default:
+							// Channel is full, drop the frame
+							scopedLogger.Warn().Msg("Audio frame channel full, dropping frame")
 						}
 					}
 				}
@@ -245,7 +263,8 @@ func newSession(config SessionConfig) (*Session, error) {
 				err := rpcUnmountImage()
 				scopedLogger.Warn().Err(err).Msg("unmount image failed on connection close")
 			}
-			// Stop audio input manager
+			// Stop audio processing and input manager
+			session.stopAudioProcessor()
 			if session.AudioInputManager != nil {
 				session.AudioInputManager.Stop()
 			}
@@ -262,6 +281,43 @@ func newSession(config SessionConfig) (*Session, error) {
 	return session, nil
 }
 
+// startAudioProcessor starts the dedicated audio processing goroutine
+func (s *Session) startAudioProcessor(logger zerolog.Logger) {
+	s.audioWg.Add(1)
+	go func() {
+		defer s.audioWg.Done()
+		logger.Debug().Msg("Audio processor goroutine started")
+
+		for {
+			select {
+			case frame := <-s.audioFrameChan:
+				if s.AudioInputManager != nil {
+					// Check if audio input manager is ready before processing frames
+					if s.AudioInputManager.IsReady() {
+						err := s.AudioInputManager.WriteOpusFrame(frame)
+						if err != nil {
+							logger.Warn().Err(err).Msg("Failed to write Opus frame to audio input manager")
+						}
+					} else {
+						// Audio input manager not ready, drop frame silently
+						// This prevents the "client not connected" errors during startup
+						logger.Debug().Msg("Audio input manager not ready, dropping frame")
+					}
+				}
+			case <-s.audioStopChan:
+				logger.Debug().Msg("Audio processor goroutine stopping")
+				return
+			}
+		}
+	}()
+}
+
+// stopAudioProcessor stops the audio processing goroutine
+func (s *Session) stopAudioProcessor() {
+	close(s.audioStopChan)
+	s.audioWg.Wait()
+}
+
 func drainRtpSender(rtpSender *webrtc.RTPSender) {
 	// Lock to OS thread to isolate RTCP processing
 	runtime.LockOSThread()