QA_LLM_RAG_Text_to_Speech/streamlit_speakup.py

import streamlit as st
import weaviate
from langchain_community.vectorstores import Weaviate
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain.chains import RetrievalQA
from dotenv import load_dotenv
import os
from utils.hf_lazyclass import LazyHuggingFaceEndpoint
from utils.tts_speech import text_to_speech_file

load_dotenv()
hf_token = os.getenv("HUGGINGFACEHUB_API_TOKEN")

client = weaviate.Client(
    url="http://localhost:8081",  
)
repo_id = "mistralai/Mistral-7B-Instruct-v0.2"

llm = LazyHuggingFaceEndpoint(
    repo_id=repo_id, max_new_tokens=128, temperature=0.5, huggingfacehub_api_token = hf_token 
)

model_name = "sentence-transformers/all-mpnet-base-v2"
model_kwargs = {'device': 'cpu'}
encode_kwargs = {'normalize_embeddings': False}

hf = HuggingFaceEmbeddings(
    model_name=model_name,
    model_kwargs=model_kwargs,
    encode_kwargs=encode_kwargs
)

response = client.schema.get()

weaviate_vectorstore = Weaviate(client=client, index_name=response['classes'][0]['class'], text_key="intro",by_text = False, embedding=hf)
retriever = weaviate_vectorstore.as_retriever()

qa_chain = RetrievalQA.from_chain_type(
    llm=llm, chain_type="stuff", retriever = retriever
)

st.title('Insurance Handbook QA with Voice')

st.write("""
This is a simple application where we utilize RAG and Text-to-Speech to answer all your queries regarding Insurance.

In this application, we use the following tech stack:

1. Weaviate Vector Database with Docker Hosting
2. LangChain LLM Framework
3. HuggingFace Embedding Model all-mpnet-base-v2
4. HuggingFace Generative Model Mistral-7B-Instruct-v0.2
5. Elevenlabs Text-to-Speech Model
6. Streamlit for Front-End           
""")

if 'prompt' not in st.session_state:
    st.session_state.prompt = ''
  
if 'audiofile' not in st.session_state:
    st.session_state.audiofile = ''  

query  = st.text_input("Ask Your Insurance Question👇", "")
if st.button("Answer my Question"):
    st.session_state.prompt = query
    response = qa_chain.invoke(query)

    st.session_state.audiofile = text_to_speech_file(response['result'])
    st.audio(st.session_state.audiofile, format="audio/mpeg", loop = False)