Lucy23-2024/FinanceBill-Assistant-KE2025

A bilingual (English/Kiswahili) RAG assistant for the Kenya 2025 Finance Bill, built with LangChain, and Ollama # FinanceBill-Assistant-KE2025 A bilingual (English/Kiswahili) RAG assistant for the Kenya 2025 Finance Bill # Finance Bill RAG Assistant **A Retrieval-Augmented Generation (RAG) assistant that answers questions from a Finance Bill PDF (English & Swahili).** --- ## Table of contents * What is this project? * Features * Prerequisites * Setup — step by step (beginner-friendly) * Project structure * How to run (notebook) * How to run (Streamlit app) * How it works (simple explanation) * Key settings to tune (and why) * Prompts used in this project * Examples & expected behavior * Troubleshooting --- ## What is this project? This repository demonstrates a **beginner-friendly** Retrieval-Augmented Generation (RAG) workflow that: * loads a Finance Bill PDF, * splits it into searchable chunks, * creates embeddings and stores them in a vector database, * retrieves relevant chunks for a user question, * answers the question using an LLM and custom prompts. It’s aimed at learners who are new to LangChain, vector search, and practical RAG setups. --- ## Features * Load PDFs (Finance Bill) and extract text. * Split long documents into smaller chunks for better retrieval. * Create embeddings and persist a vector store (Chroma by default). * Custom Query + Answer prompts (keeps language consistent — English/Swahili). * A Jupyter notebook for exploration and a Streamlit app for end-user interaction. --- ## Prerequisites * **Python 3.12** (use `python --version` to check) * `git` (optional, for cloning) * Basic command-line knowledge (open a terminal) * Optional: account/API key for your chosen LLM (OpenAI) or a local LLM (Ollama) --- ## Setup — step by step (beginner-friendly) 1. **Clone the repo** (or make a local folder and copy files): ```bash # replace with your repo link, or skip if working locally git clone cd ``` 2. **Create & activate a virtual environment** (recommended): ```bash # Linux / macOS python -m venv venv source venv/bin/activate # Windo …