PDF Genie

Use Case:

Want to know something from a 100+ page PDF? Your personal AI assistant can scan the document, analyze its content, and provide an instant response based on the relevant context

Sample output screenshots using Streamlit :

input file used for this ml model => "IRS tax filling instructions 2024.pdf" (100+ pages)

Running Locally:

Spinning up Streamlit:

Question 1 through Streamlit local UI:

Question 2 through Streamlit local UI:

Overview

PDF Genie is a PDF search engine pipeline that processes PDF documents, extracts text, and performs various operations using the LangChain library and other tools. It is designed to build a Retrieval-Augmented Generation (RAG) system, which combines information retrieval and natural language generation to provide accurate and contextually relevant answers from PDF documents.

Technologies used

• "llama3.2:4B" ML Model
• "nomic-embed-text" Embedding Model
• Ollama
• Chroma Vector database
• LangChain
• Streamlit
• RAG System with Local LLM (LargeLanguageModel)

Features

• Load and process PDF documents
• Split text using recursive character text splitter
• Create vector stores for efficient text retrieval
• Use embeddings for text representation
• Perform multi-query retrieval
• Generate responses using a language model

Requirements

• Python 3.7+
• langchain_community
• langchain_text_splitters
• langchain_community.vectorstores
• langchain_ollama
• langchain.prompts
• langchain_core
• ollama
• urllib3