🤖 RAG Chatbot for Financial Policies

An AI-powered Retrieval-Augmented Generation (RAG) chatbot that answers questions from financial policy documents with page/section citations. Built with Python, ChromaDB, Sentence Transformers, and Gradio for interactive Q&A.

  • 🔧 Tech Stack: Python, ChromaDB, FAISS, Sentence Transformers, Gradio
  • 💻 GitHub: View Repository
  • 📌 Type: AI/ML Application
  • 🎯 Domain: Natural Language Processing, RAG Systems

📋 Project Overview

ChatBot_Financial is an intelligent chatbot designed to help users query financial policy documents and receive accurate answers with precise citations. Using Retrieval-Augmented Generation (RAG) architecture, the system combines document retrieval with language understanding to provide contextually relevant responses.

🚀 Key Technologies

AI & ML
  • Sentence Transformers
  • RAG Architecture
  • Embedding Generation
Vector Database
  • ChromaDB
  • FAISS Index
  • Similarity Search
Interface
  • Gradio UI
  • Jupyter Notebook
  • Command Line

📁 Project Architecture

📄 document_loader.py
Loads documents, primarily PDFs for processing
✂️ text_splitter.py
Breaks documents into smaller, manageable chunks
🔢 embedding_generator.py
Converts text into numerical embeddings
🗄️ vector_store.py
Manages FAISS vector store for retrieval
🔍 query_processor.py
Processes user questions and fetches answers
💬 conversation_manager.py
Manages chatbot memory and conversation flow

✨ Key Features

  • 📚 Document Q&A - Ask questions about financial policy documents
  • 📍 Page/Section Citations - Answers include precise source references
  • 🧠 RAG Architecture - Combines retrieval with generation for accuracy
  • 🔍 Semantic Search - FAISS-powered efficient similarity search
  • 💬 Interactive UI - Gradio-based web interface for easy interaction
  • 📓 Jupyter Support - Notebook for development and testing

🛠️ How It Works

  • 1. Document Loading - PDF documents are loaded and processed
  • 2. Text Chunking - Documents split into smaller chunks for better retrieval
  • 3. Embedding Generation - Text converted to vector embeddings
  • 4. Index Storage - Embeddings stored in FAISS index for fast retrieval
  • 5. Query Processing - User questions processed and relevant chunks retrieved
  • 6. Response Generation - AI generates answers with citations

"An intelligent RAG system that transforms how users interact with complex financial documents, providing accurate answers with verifiable citations for enhanced trust and transparency."