package com.sherpaonnxofflinestt

/**
 * TEN-VAD: Voice Activity Detection wrapper
 *
 * Ultra-low latency ML-based speech detection with 10ms frame processing.
 */
class TenVad(
    private val hopSize: Int = 160,  // 10ms at 16kHz
    private val threshold: Float = 0.5f
) {
    private var handle: Long = 0

    init {
        System.loadLibrary("ten_vad_jni")
        handle = nativeCreate(hopSize, threshold)
        if (handle == 0L) {
            throw RuntimeException("Failed to create TEN-VAD instance")
        }
    }

    /**
     * Process audio samples through VAD
     *
     * @param samples Audio samples (int16, 16kHz, mono). Length must equal hopSize.
     * @return Pair of (probability: Float, isSpeech: Boolean) or null on error
     */
    fun process(samples: ShortArray): VadResult? {
        if (handle == 0L) return null
        if (samples.size != hopSize) {
            throw IllegalArgumentException("Audio samples length (${samples.size}) must equal hopSize ($hopSize)")
        }

        val result = nativeProcess(handle, samples) ?: return null
        return VadResult(
            probability = result[0],
            isSpeech = result[1] > 0.5f
        )
    }

    /**
     * Release resources
     */
    fun destroy() {
        if (handle != 0L) {
            nativeDestroy(handle)
            handle = 0
        }
    }

    /**
     * Get TEN-VAD version
     */
    fun getVersion(): String = nativeGetVersion()

    /**
     * Check if VAD is initialized
     */
    fun isInitialized(): Boolean = handle != 0L

    // Native methods
    private external fun nativeCreate(hopSize: Int, threshold: Float): Long
    private external fun nativeProcess(handle: Long, audioData: ShortArray): FloatArray?
    private external fun nativeDestroy(handle: Long)
    private external fun nativeGetVersion(): String

    companion object {
        const val SAMPLE_RATE = 16000
        const val DEFAULT_HOP_SIZE = 160  // 10ms at 16kHz
        const val DEFAULT_THRESHOLD = 0.5f
    }
}

/**
 * VAD processing result
 */
data class VadResult(
    val probability: Float,
    val isSpeech: Boolean
)