PDF Question & Answer Application

A Streamlit application that enables users to ask questions about PDF documents using LangChain and Ollama LLMs. The application leverages PDF text extraction, vector embeddings, and retrieval-based question answering to provide accurate responses based on document content.

Features

• PDF document loading and processing
• Text chunking with overlap for better context preservation
• Vector embeddings using Ollama
• Fast similarity search using FAISS vector database
• Interactive Q&A interface using Streamlit
• Retrieval-augmented generation for accurate answers

Prerequisites

• Python 3.8+
• Ollama installed and running locally

Installation

1. Clone the repository:

git clone <repository-url>
cd <repository-name>

2. Install required dependencies:

pip install langchain-community langchain streamlit faiss-cpu

3. Make sure Ollama is installed and running with the llama3 model:

ollama pull llama3

Project Structure

.
├── data/
│   └── IJEPA.pdf    # Your PDF document
├── app.py           # Main application code
└── README.md        # This file

Usage

1. Place your PDF file in the data/ directory

2. Run the Streamlit application:

streamlit run app.py

3. Enter your questions in the search box to get answers based on the PDF content

Code Explanation

Document Loading and Processing

from langchain_community.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter

loader = PyPDFLoader("data/IJEPA.pdf")
docs = loader.load()

text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
documents = text_splitter.split_documents(docs)

Vector Store Creation

from langchain_community.embeddings import OllamaEmbeddings
from langchain_community.vectorstores import FAISS

db = FAISS.from_documents(documents[:30], OllamaEmbeddings(model="llama3"))

Question-Answering Chain

from langchain_core.prompts import ChatPromptTemplate
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain_community.llms import Ollama

prompt = ChatPromptTemplate.from_template("""
Answer the following question based only on the provided context.
Think step by step before providing a detailed answer.
I will tip you $1000 if the user finds the answer helpful.
<context>
{context}
</context>
Question:{input}""")

llm = Ollama(model="llama3")
document_chain = create_stuff_documents_chain(llm, prompt)

Retrieval Chain and User Interface

from langchain.chains import create_retrieval_chain

retriever = db.as_retriever()
retrievar_chain = create_retrieval_chain(retriever, document_chain)

prompt = st.text_input("Do Search here")
if prompt:
    respose = retrievar_chain.invoke({"input":prompt})
    st.write(respose['answer'])

Configuration

The application can be customized by modifying these parameters:

• chunk_size: Size of text chunks (default: 1000)
• chunk_overlap: Overlap between chunks (default: 200)
• documents[:30]: Number of documents to process (adjust as needed)

Limitations

• Currently processes only the first 30 documents from the PDF
• Requires Ollama to be running locally
• Performance depends on the size of the PDF and available system resources

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.