app.py

import streamlit as st
import os
from langchain_community.document_loaders import PDFMinerLoader
from langchain_community.embeddings import SentenceTransformerEmbeddings
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import FAISS
from langchain.chains import RetrievalQA
from langchain_community.llms import HuggingFacePipeline
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, pipeline
import torch

st.title("Custom PDF Chatbot")

# Custom CSS for chat messages
st.markdown("""
    <style>
        .user-message {
            text-align: right;
            background-color: #3c8ce7;
            color: white;
            padding: 10px;
            border-radius: 10px;
            margin-bottom: 10px;
            display: inline-block;
            width: fit-content;
            max-width: 70%;
            margin-left: auto;
            box-shadow: 0px 4px 6px rgba(0, 0, 0, 0.1);
        }
        .assistant-message {
            text-align: left;
            background-color: #d16ba5;
            color: white;
            padding: 10px;
            border-radius: 10px;
            margin-bottom: 10px;
            display: inline-block;
            width: fit-content;
            max-width: 70%;
            margin-right: auto;
            box-shadow: 0px 4px 6px rgba(0, 0, 0, 0.1);
        }
    </style>
""", unsafe_allow_html=True)

def get_file_size(file):
    file.seek(0, os.SEEK_END)
    file_size = file.tell()
    file.seek(0)
    return file_size

# Add a sidebar for model selection
st.sidebar.write("Settings")
st.sidebar.write("-----------")
model_options = ["MBZUAI/LaMini-T5-738M", "google/flan-t5-base", "google/flan-t5-small"]
selected_model = st.sidebar.radio("Choose Model", model_options)
st.sidebar.write("-----------")

uploaded_file = st.sidebar.file_uploader("Upload file", type=["pdf"])

@st.cache_resource
def initialize_qa_chain(filepath, CHECKPOINT):
    loader = PDFMinerLoader(filepath)
    documents = loader.load()
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=500)
    splits = text_splitter.split_documents(documents)

    # Create embeddings 
    embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
    vectordb = FAISS.from_documents(splits, embeddings)

    # Initialize model
    TOKENIZER = AutoTokenizer.from_pretrained(CHECKPOINT)
    BASE_MODEL = AutoModelForSeq2SeqLM.from_pretrained(CHECKPOINT, device_map=torch.device('cpu'), torch_dtype=torch.float32)
    pipe = pipeline(
        'text2text-generation',
        model=BASE_MODEL,
        tokenizer=TOKENIZER,
        max_length=256,
        do_sample=True,
        temperature=0.3,
        top_p=0.95,
    )

    llm = HuggingFacePipeline(pipeline=pipe)

    # Build a QA chain
    qa_chain = RetrievalQA.from_chain_type(
        llm=llm,
        chain_type="stuff",
        retriever=vectordb.as_retriever(),
    )
    return qa_chain

def process_answer(instruction, qa_chain):
    generated_text = qa_chain.run(instruction)
    return generated_text

if uploaded_file is not None:
    file_details = {
        "Filename": uploaded_file.name,
        "File size": get_file_size(uploaded_file)
    }
    os.makedirs("docs", exist_ok=True)
    filepath = os.path.join("docs", uploaded_file.name)
    with st.spinner('Embeddings are in process...'):
       qa_chain = initialize_qa_chain(filepath, selected_model)
else:
    qa_chain = None

# Initialize chat history
if "messages" not in st.session_state:
    st.session_state.messages = []

# Display chat messages from history on app rerun
for message in st.session_state.messages:
    if message["role"] == "user":
        st.markdown(f"<div class='user-message'>{message['content']}</div>", unsafe_allow_html=True)
    else:
        st.markdown(f"<div class='assistant-message'>{message['content']}</div>", unsafe_allow_html=True)

# React to user input
if prompt := st.chat_input("What is up?"):
    # Display user message in chat message container
    st.markdown(f"<div class='user-message'>{prompt}</div>", unsafe_allow_html=True)
    # Add user message to chat history
    st.session_state.messages.append({"role": "user", "content": prompt})

    if qa_chain:
        # Generate response
        response = process_answer({'query': prompt}, qa_chain)
    else:
        # Prompt to upload a file
        response = "Please upload a PDF file to enable the chatbot."

    # Display assistant response in chat message container
    st.markdown(f"<div class='assistant-message'>{response}</div>", unsafe_allow_html=True)
    
    # Add assistant response to chat history
    st.session_state.messages.append({"role": "assistant", "content": response})