from .base import MetaItem, BasePipe
from ..helpers.vadprocessor import VadV2
import numpy as np
from silero_vad import get_speech_timestamps
from typing import List
import logging

# import noisereduce as nr


class VadPipe(BasePipe):
    vac = None
    sample_rate = 16000
    window_size_samples = 512
    chunk_size = 512
    prob_threshold=0.5,
    silence_s=0.5,
    cache_s=0.25,



    @classmethod
    def init(cls):
        if cls.vac is None:
            cls.vac = VadV2(cls.prob_threshold, cls.sample_rate, cls.silence_s * 1000, cls.cache_s * 1000, max_speech_duration_s=15)
    
    def process(self, in_data: MetaItem) -> MetaItem:
        audio_buffer = np.frombuffer(in_data.source_audio)
        vad_audio = self.vac(audio_buffer)
        if vad_audio:
            in_data.audio = vad_audio['audio']
        else:
            in_data.audio = b""
        return in_data


    # def reduce_noise(self, data):
    #     return nr.reduce_noise(y=data, sr=self.sample_rate)