import os
import warnings
import gradio as gr
from transformers import pipeline
import re



# Initialize the speech recognition pipeline and transliterator
 
#p1 = pipeline(task="automatic-speech-recognition", model="cdactvm/w2v-bert-2.0-odia_v1")
#p2 = pipeline(task="automatic-speech-recognition", model="cdactvm/w2v-bert-2.0-hindi_v1")

HF_TOKEN = os.getenv('HW_TOKEN')
hf_writer = gr.HuggingFaceDatasetSaver(HF_TOKEN, "save_audio")


cur_line=0

def readFile():
    f=open('prompt.txt')
    line_num=0
    lines=f.readlines()
    line_num = len(lines)
    return  line_num,lines

totlines,file_content=readFile()

callback = gr.CSVLogger()
        
def readPromt():
    global cur_line
    cur_line+=1
    global file_content
    print (cur_line)
    return file_content[cur_line]
    
def readNext():
    
    global totlines
    print(totlines)
    global cur_line
    if cur_line<totlines-1:
        cur_line+=1
        global file_content
        print (cur_line)
    return [file_content[cur_line],None]

def readPrevious():
    global cur_line
    if cur_line>=0:
        cur_line-=1
    #cur_line=current_line
        global file_content
        print (cur_line)
    return [file_content[cur_line],None]

demo = gr.Blocks()

with demo:
    #dr=gr.Dropdown(["Hindi","Odiya"],value="Odiya",label="Select Language")
    #audio_file = gr.Audio(sources=["microphone","upload"],type="filepath")
    text = gr.Textbox(readPromt())
    upfile =  gr.Audio(
        sources=["microphone","upload"], type="filepath",  label="Record"
    )
    #upfile =  gr.inputs.Audio(source="upload", type="filepath", label="Upload")
    
    with gr.Row():
        b1 = gr.Button("Save")
        b2 = gr.Button("Next")
        b3 = gr.Button("Previous")
        #b4=gr.Button("Clear")
        b2.click(readNext,inputs=None,outputs=[text,upfile])
        b3.click(readPrevious,inputs=None,outputs=[text,upfile])
        #b4.click(lambda: None, outputs=upfile)
   # b1.click(sel_lng, inputs=[dr,mic,upfile], outputs=text)
    #b2.click(text_to_sentiment, inputs=text, outputs=label)
    #callback.setup([text, upfile], "flagged_data_points")
    #callback.setup([text, upfile], hf_writer)
    #b1.click(lambda *args: callback.flag(args), [text, upfile], None, preprocess=False)
    flagging_callback=hf_writer
    b1.click(lambda *args:  flagging_callback, [text, upfile], None, preprocess=False)
demo.launch()