Skip to content

RAGsphere: A Unified Library of Retrieval-Augmented Generation Techniques with Implementations, Comparisons, and a Practical Selection Guide

Empowering RAG Practitioners with Insights into Modern Techniques and Guidance for Choosing the Right Approach.

Getting Started Techniques

Welcome

Introduction

Retrieval-Augmented Generation (RAG) is transforming the landscape of generative AI by seamlessly integrating information retrieval with advanced generation capabilities. This repository presents a collection of modern techniques designed to enhance the performance of RAG systems, enabling them to produce responses that are more accurate, context-aware, and informative. Our goal is to provide a valuable resource for researchers and practitioners looking to push the boundaries of what's possible with RAG.

RAG Techniques

Explore the list of Retrieval-Augmented Generation methods that have been considered for this library.

Category Technique Ref. Demo
1 Foundational 🧱 Vector RAG github
2 Advanced Architecture 🛰️ Vector GraphRAG github jupyter
3 Advanced Architecture 🛰️ Vector Cypher GraphRAG github jupyter
4 Advanced Architecture 🛰️ Hybrid GraphRAG github jupyter
5 Advanced Architecture 🛰️ Hybrid Cypher GraphRAG github jupyter
6 Advanced Architecture 🛰️ Text2Cypher github jupyter
7 Advanced Architecture 🛰️ GARAG jupyter
8 Advanced Architecture 🛰️ Naive GraphRAG jupyter
9 Advanced Architecture 🛰️ Microsoft GraphRAG - in progress github arxiv
10 Advanced Architecture 🛰️ LightRAG - in progress github arxiv
11 Advanced Architecture 🛰️ PathRAG - in progress github arxiv
12 Advanced Architecture 🛰️ GNN-RAG - in progress github arxiv
13 Advanced Architecture 🛰️ T-RAG - in progress github arxiv
14 Context Enrichment 🧩
15 Query Enhancement ✨