app.py

import streamlit as st
import random
from app_config import SYSTEM_PROMPT, NLP_MODEL_NAME, NUMBER_OF_VECTORS_FOR_RAG, NLP_MODEL_TEMPERATURE, NLP_MODEL_MAX_TOKENS, VECTOR_MAX_TOKENS,SUMMMERIZE_PROMPT
from functions import get_vectorstore, tiktoken_len,save_audio_file,get_audio_transcription
from langchain.memory import ConversationSummaryBufferMemory
from langchain_core.messages import SystemMessage, HumanMessage, AIMessage
from langchain.chains.summarize import load_summarize_chain
from langchain.prompts import PromptTemplate
from langchain_groq import ChatGroq
from dotenv import load_dotenv
from pathlib import Path
import os
env_path = Path('.') / '.env'
load_dotenv(dotenv_path=env_path)

def response_generator(prompt: str) -> str:
    """this function can be used for general quetion answers which are related to tyrex and tyre recycling 

    Args:
        prompt (string): user query

    Returns:
        string: answer of the query
    """

    try:
        retriever = st.session_state.retriever
        docs = retriever.invoke(prompt)
        my_context = [doc.page_content for doc in docs]
        my_context = '\n\n'.join(my_context)

        system_message = SystemMessage(content = SYSTEM_PROMPT.format(context=my_context, previous_message_summary=st.session_state.rag_memory.moving_summary_buffer))
        chat_messages = (system_message + st.session_state.rag_memory.chat_memory.messages + HumanMessage(content=prompt)).messages

        response = st.session_state.llm.invoke(chat_messages)
        return response.content
    
    except Exception as error:
        print(error)
        return "Oops! something went wrong, please try again."


st.markdown(
    """
<style>
    .st-emotion-cache-janbn0 {
        flex-direction: row-reverse;
        text-align: right;
    }
</style>
""",
    unsafe_allow_html=True,
)

st.header("TubeChat: AI Chatbot for Video Summary and Q&A")
if url := st.text_input("Enter the Youtube Video URL: "):

    if "url" in st.session_state and url!= st.session_state.url:
        st.session_state.clear()

    if "url" not in st.session_state:
    # save audio file
        save_audio_file(url)
        print("save audio file ...")
        st.session_state.url = url
        # save transcrption
        get_audio_transcription()
        print("save transcription")

    print("SYSTEM MESSAGE")
    if "messages" not in st.session_state:
        st.session_state.messages=[{"role": "system", "content": SYSTEM_PROMPT}]
        
    print("SYSTEM MODEL")
    if "llm" not in st.session_state:
        st.session_state.llm = ChatGroq(temperature=NLP_MODEL_TEMPERATURE, groq_api_key=str(os.getenv('GROQ_API_KEY')), model_name=NLP_MODEL_NAME)

    print("rag")
    if "rag_memory" not in st.session_state:
        st.session_state.rag_memory = ConversationSummaryBufferMemory(llm=st.session_state.llm, max_token_limit=NLP_MODEL_MAX_TOKENS - tiktoken_len(SYSTEM_PROMPT) - VECTOR_MAX_TOKENS*NUMBER_OF_VECTORS_FOR_RAG)

    print("retrival")
    if "retriever" not in st.session_state:
        vector_store, docs = get_vectorstore()
        st.session_state.retriever = vector_store.as_retriever(k=NUMBER_OF_VECTORS_FOR_RAG)

        #get summary of given youtube video
        summary_prompt = PromptTemplate(template=SUMMMERIZE_PROMPT,
                            input_variables=["text"])
        chain = load_summarize_chain(st.session_state.llm,chain_type="stuff",prompt = summary_prompt)
        output_summary = chain.run(docs)
        st.session_state.messages.append({"role":"assistant","content":"Video's Summary: \n"+ output_summary})

    print("container")
    # Display chat messages from history
    container =  st.container(height=700)
    for message in st.session_state.messages:
        if message["role"] != "system":
            with container.chat_message(message["role"]):
                st.write(message["content"])

    # When user gives input
    if prompt := st.chat_input("Enter your query here... "):
        with container.chat_message("user"):
            st.write(prompt)
        st.session_state.messages.append({"role":"user" , "content":prompt})
        
        with container.chat_message("assistant"):  
            response = response_generator(prompt=prompt)
            print("******************************************************** Response ********************************************************")
            print("MY RESPONSE IS:", response)
            st.write(response)
        
        print("Response is:", response)
        st.session_state.rag_memory.save_context({'input': prompt}, {'output': response})
        st.session_state.messages.append({"role":"assistant" , "content":response})