lohchar/RAG_Constitution_of_Kenya

Retrieval-Augmented Generation to answer queries based on the Constitution of Kenya. ## RAG_Constitution_of_Kenya Retrieval-Augmented Generation to answer queries based on the Constitution of Kenya. ### Overview - This project is a sophisticated Question-Answering (Q&A) system built to answer queries based on the Constitution of Kenya 2010. - It leverages a powerful technique called Retrieval-Augmented Generation (RAG) to ensure that the answers are accurate and grounded in the provided legal text. - The system uses a Large Language Model (LLM) from Groq, accessed via the langchain-groq library, and a state-of-the-art embedding model to understand and process user questions. ### Technologies Used This project is built with a modern stack of AI and Python libraries: - LLM: Groq (llama-3.3-70b-versatile) - Framework: LangChain - Embedding Model: sentence-transformers (BAAI/bge-base-en-v1.5) - Vector Database: ChromaDB - PDF Loading: pypdf - Environment: Google Colab / Jupyter Notebook ### How It Works The chatbot follows a RAG pipeline to provide accurate, context-aware answers: - *Document Loading:* The Constitution of Kenya 2010 is loaded from a PDF file (The_Constitution_of_Kenya_2010.pdf). - *Text Splitting:* To handle the large document efficiently, the text is split into smaller, manageable chunks. This project uses a ParentDocumentRetriever strategy, which splits the text into larger "parent" chunks and smaller "child" chunks. - This ensures that the LLM receives enough context without being overwhelmed. - *Embedding:* Each text chunk is converted into a numerical vector representation (embedding) using the BAAI/bge-base-en-v1.5 model. - These embeddings capture the semantic meaning of the text. - *Vector Storage:* The embeddings and their corresponding text chunks are stored in a ChromaDB vector store. This database allows for quick and efficient similarity searches. - *Retrieval*: When a user asks a question, the system first embeds the query and then searches the ChromaDB store to find the most relevant text chunks from the Constitution. - * …