app.py

from langchain.vectorstores import Chroma
from langchain.embeddings import SentenceTransformerEmbeddings
from langchain.llms import HuggingFaceHub
from langchain.chains import RetrievalQA
import streamlit as st


persist_direc ="chroma_db"
embeddings = SentenceTransformerEmbeddings(model_name="multi-qa-MiniLM-L6-cos-v1")

vectordb = Chroma(persist_directory=persist_direc, embedding_function=embeddings)


print("Db Fetched!")







st.header("Chatbot for IIT Madras BS Degree: Ask questions to the Handbook")
st.caption("Developed by Indranil Bhattacharyya (21F1005840)")
st.text("Please note that this is an experimental app, if you find something inappropiate report it to the developer")

with st.sidebar:
    st.write("Chat history")
    l = len(st.session_state)

    if l>0:
        for i in st.session_state:
            
            with st.chat_message("human"):
                
                st.write(i)

            with st.chat_message('assitant'):
                st.write(st.session_state[i])
            st.divider() 
    model_name = st.radio("Choose the LLM for the chat:", ["google/flan-t5-large", 
                                                           "google/flan-t5-xxl", 
                                                           "google/flan-ul2",
                                                           "facebook/m2m100_1.2B"])
    

repo_id = model_name
llm = HuggingFaceHub(
    repo_id=repo_id, model_kwargs={"temperature": 1},
    huggingfacehub_api_token="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"
)

qa_chain = RetrievalQA.from_chain_type(
    llm,
    retriever=vectordb.as_retriever()
)               



    
c = 0
prompt = st.chat_input("Say something")
if prompt:
    c += 1
    with st.spinner('Wait for it...'):
    
        result = qa_chain({"query":prompt})
    with st.chat_message("user"):
        st.write(prompt)

    with st.chat_message('assistant'):
        st.write(result["result"])

    key_name = 'question' + str(c)
    st.session_state[prompt] = result["result"]