import streamlit as st
import torch
from urllib.parse import urlparse, parse_qs
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
# https://pypi.org/project/youtube-transcript-api/
from youtube_transcript_api import YouTubeTranscriptApi

def get_video_id(url: str) -> str:
    """
    Examples:
    - http://youtu.be/SA2iWivDJiE
    - http://www.youtube.com/watch?v=_oPAwA_Udwc&feature=feedu
    - http://www.youtube.com/embed/SA2iWivDJiE
    - http://www.youtube.com/v/SA2iWivDJiE?version=3&hl=en_US
    """
    query = urlparse(url)
    if query.hostname == 'youtu.be':
        return query.path[1:]
    if query.hostname in ('www.youtube.com', 'youtube.com'):
        if query.path == '/watch':
            p = parse_qs(query.query)
            return p['v'][0]
        if query.path[:7] == '/embed/':
            return query.path.split('/')[2]
        if query.path[:3] == '/v/':
            return query.path.split('/')[2]
    return None


def get_youtube_subtitle(video_id: str) -> str:
    try:  
        parse = YouTubeTranscriptApi.get_transcript(video_id, languages=['ru'])
        result = ''
        for i in parse:
            if (i['text'][0] =='[') & (i['text'][-1] ==']'): continue
            result += ' ' + i['text']
        result = result.strip()[0].upper() + result.strip()[1:]
        return result.strip()
    except:
        return None

device = "cuda" if torch.cuda.is_available() else "cpu"
m_name = '/home/user/app/model'
#m_name = '../model'
tokenizer = AutoTokenizer.from_pretrained(m_name)
model = AutoModelForSeq2SeqLM.from_pretrained(m_name)
model.to(device)
        
if __name__ == "__main__":
    st.header("Annotation of subtitles from YouTube")

    url = st.text_input('Enter the URL of the Youtube video', 'https://www.youtube.com/watch?v=HGSVsK32rKA ')
    st.text("""
            Example:
            https://www.youtube.com/watch?v=HGSVsK32rKA
            https://www.youtube.com/watch?v=fSpARfZ3I50
            https://www.youtube.com/watch?v=3lEMopaRSjw
            """
           )
    video_id = get_video_id(url)

    if video_id is not None:
        subtitle = get_youtube_subtitle(video_id)
        if subtitle is not None:
            st.subheader('Subtitles')
            st.markdown(subtitle)

            inputs = tokenizer( [subtitle],
                                max_length=1000,
                                padding="max_length",
                                truncation=True,
                                return_tensors="pt",
                                )["input_ids"]


            if st.button('Compute summary', help='Click me'):
                outputs = model.generate(inputs.to(device), max_new_tokens=100, do_sample=False)
                summary = tokenizer.decode(outputs[0], skip_special_tokens=True)
                st.subheader('Summary')
                st.markdown(summary)
        else:
            st.markdown(':red[Subtitles are disabled for this video]')
    else:
        st.markdown(':red[Video clip is not detected]')