app.py

import os 
import gradio as gr
import time 
import logging
from langchain.document_loaders import PDFMinerLoader,CSVLoader ,UnstructuredWordDocumentLoader,TextLoader,OnlinePDFLoader
from langchain.text_splitter import CharacterTextSplitter
from langchain.embeddings import SentenceTransformerEmbeddings
from langchain.vectorstores import FAISS
from langchain import HuggingFaceHub
from langchain.chains import RetrievalQA
from langchain.prompts import PromptTemplate
from langchain.docstore.document import Document
from youtube_transcript_api import YouTubeTranscriptApi
from . import chatops 

logger = logging.getLogger(__name__)

DEVICE = 'cpu'
MAX_NEW_TOKENS = 4096
DEFAULT_TEMPERATURE = 0.1
DEFAULT_MAX_NEW_TOKENS = 2048
MAX_INPUT_TOKEN_LENGTH = 4000
DEFAULT_CHAR_LENGTH = 1000

def loading_file():
    return "Loading..."


def get_text_from_youtube_link(video_link,max_video_length=800):
    video_text = ""
    video_id = video_link.split("watch?v=")[1].split("&")[0]
    srt = YouTubeTranscriptApi.get_transcript(video_id)
    for text_data in srt:
        video_text = video_text + " " + text_data.get("text")
    if len(video_text) > max_video_length:
        return video_text[0:max_video_length]
    else:
        return video_text

def process_documents(documents,data_chunk=1500,chunk_overlap=100):
    text_splitter = CharacterTextSplitter(chunk_size=data_chunk, chunk_overlap=chunk_overlap,separator='\n')
    texts = text_splitter.split_documents(documents)
    return texts

def process_youtube_link(link, document_name="youtube-content"):
    try:
        metadata = {"source": f"{document_name}.txt"}
        return [Document(page_content=get_text_from_youtube_link(video_link=link), metadata=metadata)]
    except Exception as err:
        logger.error(f'Error in reading document. {err}')


def youtube_chat(youtube_link,API_key,llm='HuggingFace',temperature=0.1,max_tokens=1096,char_length=1500):
    
    document  = process_youtube_link(link=youtube_link)
    embedding_model = SentenceTransformerEmbeddings(model_name='thenlper/gte-base',model_kwargs={"device": DEVICE})
    texts = process_documents(documents=document)
    global vector_db
    vector_db = FAISS.from_documents(documents=texts, embedding= embedding_model)
    global qa
    qa = RetrievalQA.from_chain_type(llm=chatops.chat_application(llm_service=llm,key=API_key,
                                                            temperature=temperature,
                                                            max_tokens=max_tokens
                                                        ),
                                chain_type='stuff',
                                retriever=vector_db.as_retriever(),
                                #  chain_type_kwargs=chain_type_kwargs,
                                return_source_documents=True
                            )
    return "Youtube link Processing completed ..."


css="""
#col-container {max-width: 700px; margin-left: auto; margin-right: auto;}
"""

title = """
<div style="text-align: center;max-width: 700px;">
    <h1>Chat on You Tube video data • OpenAI/HuggingFace</h1>
    <p style="text-align: center;">Upload a You tube Link, to create its captions and load them as embeddings <br />
    once status is ready, you can start asking questions about the content you uploaded.<br />
    The repo provides you an option to use HuggingFace/OpenAI as LLM's, make sure to add your API Key before proceding.
    </p>
</div>
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.HTML(title)
    
    with gr.Group():
        chatbot = gr.Chatbot(height=300)
    with gr.Row():
        question = gr.Textbox(label="Type your question !",lines=1).style(full_width=True)
        submit_btn = gr.Button(value="Send message", variant="primary", scale = 1)
        clean_chat_btn =  gr.Button("Delete Chat")

    with gr.Column():
        with gr.Box():
            LLM_option = gr.Dropdown(['HuggingFace','OpenAI'],label='Large Language Model Selection',info='LLM Service')
            API_key = gr.Textbox(label="Add API key", type="password",autofocus=True)
            with gr.Accordion(label='Advanced options', open=False):
                    max_new_tokens = gr.Slider(
                        label='Max new tokens',
                        minimum=2048,
                        maximum=MAX_NEW_TOKENS,
                        step=1,
                        value=DEFAULT_MAX_NEW_TOKENS,
                        )
                    temperature = gr.Slider(
                    label='Temperature',
                    minimum=0.1,
                    maximum=4.0,
                    step=0.1,
                    value=DEFAULT_TEMPERATURE,
                    )
                    char_length = gr.Slider(
                        label='Max Character',
                        minimum= DEFAULT_CHAR_LENGTH,
                        maximum = 5*DEFAULT_CHAR_LENGTH,
                        step = 500,
                        value= 1500
                    )

    with gr.Column():
            with gr.Box():
                add_link = gr.Textbox(label="Add your you tube Link",text_align='left',autofocus=True)