Building a Conversational Web Application for PDF Documents using Mistral-7B-v0.1

In this we'll walk through the development of a web application that allows users to interact with PDF documents via a conversational AI interface. This application leverages modern AI tools and frameworks to process and query text extracted from multiple PDF files. We'll cover the design choices, key components, and code implementation.

Introduction

Our program seeks to give users a simple means of interacting with data found in PDF documents. Users can ask questions about the material by uploading PDFs, and the program will employ a conversational AI model to deliver pertinent answers. Users that need to swiftly extract information from complex documents will profit from this approach. We are going to use mistral.ai for our development case. A ground-breaking team called Mistral AI developed big language models that are sophisticated yet tiny enough to be used on your desktop computer. In contrast to the handicap practices of large organizations for NLP developers working in large quantities, they offer fair support and model weights for download.

Technology Stack

• Streamlit: A powerful library for building interactive web applications with Python.
• LangChain: A framework simplifying the integration of language models and vector databases.
• HuggingFace Transformers: Provides pre-trained models for natural language processing tasks.
• FAISS: A library for efficient similarity search and clustering of dense vectors.
• PyPDF2: A library for reading text from PDF files.

Application Components

1. PDF Processing: Extracts text from PDF files.
2. Text Chunking: Splits text into manageable chunks.
3. Vector Store Creation: Embed text chunks and store them in a FAISS index for efficient retrieval.
4. Conversational AI: Handles user queries and provides relevant responses based on the indexed text.
5. Web Interface: Built with Streamlit, users can upload PDFs and interact with the AI model.

Code Overview

1. PDF Processing

The get_pdf_text function handles the extraction of text from PDF documents.

def get_pdf_text(pdf_docs):

text = ""

for pdf in pdf_docs:

pdf_reader = PdfReader(pdf)

for page in pdf_reader.pages:

text += page.extract_text()

return text

2. Text Chunking

The get_text_chunks function splits the extracted text into smaller chunks for easier processing.

def get_text_chunks(text):

text_splitter = RecursiveCharacterTextSplitter(

chunk_size=1000,

chunk_overlap=200,

length_function=len

)

chunks = text_splitter.split_text(text)

return chunks

Using the RecursiveCharacterTextSplitter, we break the text into chunks of 1000 characters with an overlap of 200 characters. This approach helps in managing significant texts and improves retrieval accuracy.

3. Vector Store Creation

The get_vectorstore function creates a vector store for efficient similarity search.

def get_vectorstore(text_chunks):

model_id = "sentence-transformers/all-MiniLM-L6-v2"

model_kwargs = {'device': 'cpu'}

encode_kwargs = {'normalize_embeddings': False}

embeddings = HuggingFaceEmbeddings(

model_name=model_id,

model_kwargs=model_kwargs,

encode_kwargs=encode_kwargs

)

vectorstore = FAISS.from_texts(texts=text_chunks, embedding=embeddings)

return vectorstore

4. Conversational AI

The get_conversation_chain the function sets up the conversational AI using HuggingFace models.

def get_conversation_chain(vectorstore):

memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)

llm = HuggingFaceHub(repo_id="mistralai/Mistral-7B-v0.1", model_kwargs={"temperature":0.5, "max_length":512})

conversation_chain = ConversationalRetrievalChain.from_llm(

llm=llm,

chain_type="stuff",

retriever=vectorstore.as_retriever(search_kwargs={"k": 3}),

memory=memory

)

return conversation_chain

We use HuggingFaceHub to load the conversational AI model and set up the ConversationalRetrievalChain to handle user queries based on the indexed text.

5. Web Interface

The main function defines the web interface using Streamlit.

def main():

st.set_page_config(page_title="Chat with multiple PDFs", page_icon=":books:")

st.write(css, unsafe_allow_html=True)

if "conversation" not in st.session_state:

st.session_state.conversation = None

if "chat_history" not in st.session_state:

st.session_state.chat_history = None

st.header("Chat with multiple PDFs :books:")

user_question = st.text_input("Ask a question about your documents:")

if user_question:

handle_userinput(user_question)

with st.sidebar:

st.subheader("Your documents")

pdf_docs = st.file_uploader("Upload your PDFs here and click on 'Process'", accept_multiple_files=True)

if st.button("Process"):

with st.spinner("Processing"):

st.write("Processing your documents...")

raw_text = get_pdf_text(pdf_docs)

text_chunks = get_text_chunks(raw_text)

vectorstore = get_vectorstore(text_chunks)

st.session_state.conversation = get_conversation_chain(vectorstore)

st.success("Documents processed successfully!")

The Streamlit interface allows users to upload PDF files and ask questions. The uploaded PDFs are processed, and the conversation chain is set up to handle user interactions.

Finally, you need to create a htmlTemplates.py file to implement your web application.

css = '''

<style>

.chat-message {

padding: 1.5rem; border-radius: 0.5rem; margin-bottom: 1rem; display: flex

}

.chat-message.user {

background-color: #2b313e

}

.chat-message.bot {

background-color: #475063

}

.chat-message .avatar {

width: 20%;

}

.chat-message .avatar img {

max-width: 78px;

max-height: 78px;

border-radius: 50%;

object-fit: cover;

}

.chat-message .message {

width: 80%;

padding: 0 1.5rem;

color: #fff;

}

'''

bot_template = '''

<div class="chat-message bot">

<div class="avatar">

<img src="https://i.ibb.co/jfWFfnk/chatbot-2-logo.jpg">

</div>

<div class="message">{{MSG}}</div>

</div>

'''

user_template = '''

<div class="chat-message user">

<div class="avatar">

<img src="https://i.ibb.co/kKfsbY0/human-chat-head.jpg">

</div>

<div class="message">{{MSG}}</div>

</div>

'''

Conclusions

This application demonstrates how to build a web-based conversational interface for querying information from PDF documents. By integrating various tools and technologies, we can provide a seamless experience for users to interact with document content meaningfully.