Retrieval Augmented Generation (RAG) for LLMs

Table of Contents

1. About The Project
2. Getting Started
   • Prerequisites
   • Installation
3. Usage
4. Architecture
5. Features
6. Deployed Application

About The Project

Welcome to the Insurance Question Answering System! This project leverages a Retrieval-Augmented Generation (RAG) model to provide accurate and comprehensive answers to insurance-related questions. The system integrates a powerful language model with a retrieval mechanism to enhance the quality and relevance of responses.

Getting Started

How to run this project locally

Prerequisites

Install milvus standalone: Follow instruction here to install and setup an milvus database.

Installation

1. Get a free Google Gemini API key here.
2. Clone the repo

   git clone https://github.com/mrpoponep/RAG-For_LLMs

3. Install packages using pip.

    pip install -r requirement.txt

4. Create a local_settings.py file with your API key in it.

   API_KEY='abcdxyz'

Usage

1. Start your milvus database and connect to it.
2. Prepare your own dataset as jsonl file (Example).
3. Run main.ipynb file to upload your embeded data onto the database.
4. Test your RAG and LLM by running Test.ipynb file.\

Architecture

The system consists of the following components:

Features

1. Real data used for trainning: Data used for retrieval are 21,325 answers for 12,889 common question gathered from this website insurance library.
2. Accurate Question Answering: Provides detailed and accurate answers to a wide range of insurance-related questions.
3. Retrieval-Augmented Generation: Combines retrieval of relevant documents with generation of answers for improved accuracy.

Deployed Application

You can access the deployed version of this project here.

Visit the link to try out the Insurance Question Answering System and see it in action!