PDF Chatbot

Introduction

The PDF Chatbot project aims to develop a sophisticated chatbot application that allows users to interact with PDF documents using natural language queries. The system will leverage advanced technologies such as Python, Streamlit, LangChain, Google Gemini API, and FAISS to provide accurate and contextually relevant responses based on the content of uploaded PDF files.

System Overview

The PDF Chatbot is designed to facilitate seamless interaction with PDF documents by enabling users to upload multiple files, extract and process text, and receive answers to their queries. The system will utilize Retrieval-Augmented Generation (RAG) to enhance the accuracy of responses by referencing specific text chunks from the PDFs. A modern Streamlit UI will ensure a user-friendly experience, while robust error handling and loading indicators will enhance system reliability.

Functional Requirements

• As a User, I should be able to upload multiple PDF files to the system.
• As a User, I should be able to extract text from the uploaded PDFs.
• As a User, I should be able to have the documents split into manageable chunks for processing.
• As a User, I should be able to have the system generate embeddings from the document chunks and store them in FAISS.
• As a User, I should be able to ask questions and receive answers based on the uploaded PDFs using the Gemini API.
• As a User, I should be able to see the source text chunks that were used to generate the answer.
• As a User, I should experience a clean and modern UI provided by Streamlit.
• As a User, I should encounter error handling and loading indicators to ensure smooth operation.
• As a Developer, I should be able to generate a requirements.txt file for dependencies.
• As a Developer, I should be able to generate a README.md with setup instructions.
• As a Developer, I should be able to organize the code into multiple files following best practices.
• As a Developer, I should be able to understand the project architecture through documentation.

User Personas

• User: Individuals who interact with the PDF Chatbot to upload PDFs and receive answers to their queries.
• Developer: Individuals responsible for maintaining and enhancing the PDF Chatbot system.

Core User Flows

• User uploads PDF files -> System extracts text -> Documents are split into chunks -> Embeddings are generated and stored in FAISS -> User asks a question -> System uses Gemini API to generate an answer -> Source text chunks are displayed.
• Developer sets up the environment -> Generates requirements.txt and README.md -> Organizes code into multiple files -> Documents project architecture.

Visuals Colors and Theme

• primary: #2A9D8F (Teal)
• primary_light: #A8DADC (Light Teal)
• secondary: #E63946 (Crimson)
• accent: #F4A261 (Orange)
• highlight: #E9C46A (Gold)
• bg: #F1FAEE (Off White)
• surface: rgba(38, 70, 83, 0.8) (Dark Teal)
• text: #1D3557 (Navy Blue)
• text_muted: #457B9D (Muted Blue)
• border: rgba(233, 69, 96, 0.2) (Light Crimson)

Signature Design Concept

The homepage of the PDF Chatbot will feature an interactive "Document Galaxy" concept. Users will navigate through a 3D galaxy where each star represents an uploaded PDF. Hovering over a star will reveal the document's title, and clicking on it will zoom into a detailed view where users can interact with the document's content. The galaxy will be rendered using @react-three/fiber and @react-three/drei, providing a visually stunning and engaging experience. The transition between stars will be smooth, with animations powered by framer-motion, creating a sense of exploration and discovery.

Interaction Model & Motion Direction

The landing page will employ a "parallax" interaction model, creating a layered depth effect as users scroll through the page. Decorative elements will move at different speeds to enhance the sense of depth, while the main content will scroll naturally. This approach will provide a visually rich first impression, ideal for engaging users in the PDF Chatbot's capabilities.

Non-Functional Requirements

• The system must handle concurrent uploads and queries efficiently.
• The UI should be responsive and accessible across various devices and screen sizes.
• The system should ensure data privacy and security for uploaded documents.
• The system should provide quick response times for user queries.

Tech Stack

• Frontend: Streamlit
• Backend: Python
• Database: FAISS for embeddings storage
• AI Models: Google Gemini API
• AI Tools: LangChain
• Local Orchestration: Docker, docker-compose

Assumptions and Constraints

• The system assumes users have a basic understanding of how to interact with web applications.
• The system is constrained to using the specified technologies and tools for development.
• The system assumes the availability of the Google Gemini API for processing queries.

Glossary

• PDF: Portable Document Format, a file format for capturing and sending electronic documents.
• FAISS: Facebook AI Similarity Search, a library for efficient similarity search and clustering of dense vectors.
• RAG: Retrieval-Augmented Generation, a method for improving the accuracy of AI-generated responses by referencing specific data sources.
• Streamlit: An open-source app framework for Machine Learning and Data Science projects.
• Embeddings: Numerical representations of text data used for similarity search and machine learning tasks.