go-ts/pkg/audio/capture.go

//go:build windows

package audio

import (
	"encoding/binary"
	"fmt"
	"sync"
	"time"
	"unsafe"

	"github.com/go-ole/go-ole"
	"github.com/moutend/go-wca/pkg/wca"
)

// Capturer handles WASAPI audio capture from microphone
type Capturer struct {
	client        *wca.IAudioClient
	captureClient *wca.IAudioCaptureClient
	waveFormat    *wca.WAVEFORMATEX
	bufferSize    uint32
	running       bool
	mu            sync.Mutex
	stopChan      chan struct{}

	// Callback for captured audio (called with 960-sample frames)
	onAudio func(samples []int16)

	// Sample accumulation buffer
	sampleBuffer []int16
	bufferMu     sync.Mutex

	// Current audio level (0-100)
	currentLevel int
	levelMu      sync.RWMutex
}

// NewCapturer creates a new WASAPI audio capturer
func NewCapturer() (*Capturer, error) {
	// Initialize COM (may already be initialized)
	ole.CoInitializeEx(0, ole.COINIT_APARTMENTTHREADED)

	// Get default capture endpoint (microphone)
	var deviceEnumerator *wca.IMMDeviceEnumerator
	if err := wca.CoCreateInstance(
		wca.CLSID_MMDeviceEnumerator,
		0,
		wca.CLSCTX_ALL,
		wca.IID_IMMDeviceEnumerator,
		&deviceEnumerator,
	); err != nil {
		return nil, fmt.Errorf("failed to create device enumerator: %w", err)
	}
	defer deviceEnumerator.Release()

	var device *wca.IMMDevice
	if err := deviceEnumerator.GetDefaultAudioEndpoint(wca.ECapture, wca.EConsole, &device); err != nil {
		return nil, fmt.Errorf("failed to get default capture device: %w", err)
	}
	defer device.Release()

	// Activate audio client
	var audioClient *wca.IAudioClient
	if err := device.Activate(wca.IID_IAudioClient, wca.CLSCTX_ALL, nil, &audioClient); err != nil {
		return nil, fmt.Errorf("failed to activate audio client: %w", err)
	}

	// Set up format for 48kHz mono 16-bit (TeamSpeak format)
	waveFormat := &wca.WAVEFORMATEX{
		WFormatTag:      wca.WAVE_FORMAT_PCM,
		NChannels:       1,
		NSamplesPerSec:  48000,
		WBitsPerSample:  16,
		NBlockAlign:     2,
		NAvgBytesPerSec: 96000,
		CbSize:          0,
	}

	// Initialize in shared mode - 100ms buffer
	duration := wca.REFERENCE_TIME(100 * 10000) // 100ms in 100-nanosecond units
	if err := audioClient.Initialize(
		wca.AUDCLNT_SHAREMODE_SHARED,
		wca.AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM|wca.AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY,
		duration,
		0,
		waveFormat,
		nil,
	); err != nil {
		audioClient.Release()
		return nil, fmt.Errorf("failed to initialize audio client: %w", err)
	}

	// Get buffer size
	var bufferSize uint32
	if err := audioClient.GetBufferSize(&bufferSize); err != nil {
		audioClient.Release()
		return nil, fmt.Errorf("failed to get buffer size: %w", err)
	}

	// Get capture client
	var captureClient *wca.IAudioCaptureClient
	if err := audioClient.GetService(wca.IID_IAudioCaptureClient, &captureClient); err != nil {
		audioClient.Release()
		return nil, fmt.Errorf("failed to get capture client: %w", err)
	}

	return &Capturer{
		client:        audioClient,
		captureClient: captureClient,
		waveFormat:    waveFormat,
		bufferSize:    bufferSize,
		stopChan:      make(chan struct{}),
		sampleBuffer:  make([]int16, 0, captureFrameSamples*50), // ~1 second buffer
	}, nil
}

// SetCallback sets the callback for captured audio (receives 960-sample frames)
func (c *Capturer) SetCallback(fn func(samples []int16)) {
	c.mu.Lock()
	c.onAudio = fn
	c.mu.Unlock()
}

// Start begins audio capture
func (c *Capturer) Start() error {
	c.mu.Lock()
	if c.running {
		c.mu.Unlock()
		return nil
	}
	c.running = true
	c.stopChan = make(chan struct{}) // Recreate channel for each start
	c.mu.Unlock()

	if err := c.client.Start(); err != nil {
		return fmt.Errorf("failed to start audio client: %w", err)
	}

	go c.captureLoop()
	return nil
}

// Stop stops audio capture
func (c *Capturer) Stop() {
	c.mu.Lock()
	if !c.running {
		c.mu.Unlock()
		return
	}
	c.running = false
	c.mu.Unlock()

	close(c.stopChan)
	c.client.Stop()
}

// Close releases all resources
func (c *Capturer) Close() {
	c.Stop()
	if c.captureClient != nil {
		c.captureClient.Release()
	}
	if c.client != nil {
		c.client.Release()
	}
}

// GetLevel returns the current audio input level (0-100)
func (c *Capturer) GetLevel() int {
	c.levelMu.RLock()
	defer c.levelMu.RUnlock()
	return c.currentLevel
}

// IsRunning returns whether capture is active
func (c *Capturer) IsRunning() bool {
	c.mu.Lock()
	defer c.mu.Unlock()
	return c.running
}

func (c *Capturer) captureLoop() {
	ticker := time.NewTicker(10 * time.Millisecond) // Check more often than 20ms
	defer ticker.Stop()

	for {
		select {
		case <-c.stopChan:
			return
		case <-ticker.C:
			c.readFromBuffer()
		}
	}
}

func (c *Capturer) readFromBuffer() {
	// Read all available packets
	for {
		var packetLength uint32
		if err := c.captureClient.GetNextPacketSize(&packetLength); err != nil {
			return
		}

		if packetLength == 0 {
			break
		}

		var buffer *byte
		var numFrames uint32
		var flags uint32
		if err := c.captureClient.GetBuffer(&buffer, &numFrames, &flags, nil, nil); err != nil {
			return
		}

		if numFrames == 0 {
			c.captureClient.ReleaseBuffer(numFrames)
			continue
		}

		samples := make([]int16, numFrames)
		bufSlice := unsafe.Slice(buffer, numFrames*2)

		for i := uint32(0); i < numFrames; i++ {
			samples[i] = int16(binary.LittleEndian.Uint16(bufSlice[i*2:]))
		}

		c.captureClient.ReleaseBuffer(numFrames)

		// Skip silent buffers
		if flags&wca.AUDCLNT_BUFFERFLAGS_SILENT != 0 {
			continue
		}

		// Add to sample buffer
		c.bufferMu.Lock()
		c.sampleBuffer = append(c.sampleBuffer, samples...)

		// Calculate level from latest samples
		level := CalculateRMSLevel(samples)
		c.levelMu.Lock()
		c.currentLevel = level
		c.levelMu.Unlock()

		// Send complete 960-sample frames
		for len(c.sampleBuffer) >= captureFrameSamples {
			frame := make([]int16, captureFrameSamples)
			copy(frame, c.sampleBuffer[:captureFrameSamples])
			c.sampleBuffer = c.sampleBuffer[captureFrameSamples:]

			// Call callback with the frame
			c.mu.Lock()
			callback := c.onAudio
			c.mu.Unlock()

			if callback != nil {
				callback(frame)
			}
		}
		c.bufferMu.Unlock()
	}
}