The RAG Challenge

Introduction

This project explores Retrieval-Augmented Generation (RAG) using an extended knowledge base inspired by a fictional insurance company, InsureLLM.

Project Structure

Root directory:

• uv run ingest.py (from implementation/): Ingest the latest knowledge base data.
• uv run app.py: Run the Q&A Chatbot for interactive conversations.
• uv run evaluator.py: Evaluate chatbot performance using the available evaluation scripts.

implementation directory:

• answer.py: Handles question answering. Customize or enhance fetch_context() and answer_question() as you see fit!
• ingest.py: Responsible for ingestion and preprocessing of the knowledge base. This is fully open to your improvements and ideas.

evaluation directory:

• Contains scripts for automated evaluation on test data (do not modify for consistency in evaluation).

Objective

• Understand the project structure and current implementation.
• Re-implement or improve on ingest.py and answer.py with your approaches.
• Strive for the most effective, high-performing Q&A chatbot!

Good luck!