import numpy as np

import librosa
import torch
from .init import pipe

TASK = "transcribe"

class A2T:
    def __init__(self, mic):
        self.mic = mic

    def __transcribe(self, inputs, task: str = None):
        if inputs is  None:
            print("Inputs None")

        transcribed_text = pipe(inputs, generate_kwargs={"task": task}, return_timestamps=True)["text"]
        print(transcribed_text)
        return transcribed_text
        
    def predict(self):
        try:
            if self.mic is not None:
                chunk = self.mic.get_array_of_samples()
                audio = np.array(chunk)
                print(audio)
                return self.__transcribe(inputs=audio, task=TASK)
            else:
                return "please provide audio"
                
        except Exception as e:
            print("Predict error", e)
            return "Oops some kinda error"