7 Types of RAG Systems: A Deep Dive

 

This chapter covers

  • Naïve RAG, Graph RAG, Hybrid RAG, and Memory Augmented RAG, their components, workflow, and use cases.
  • For each RAG system, we will examine a practical example using sample data to help us understand how RAG systems operate.
  • We will see how each RAG differs and when to use them.
  • A comparison table of each RAG system that helps you choose the appropriate RAG implementation for your use case.

Chapter 6 discussed RAG in detail, focusing on its components. We thoroughly examined how a RAG system operates by incorporating external data sources into the model’s query processing. We explored the benefits, architectural review, types of retrievers, advantages, and use cases of RAG components.

In this Chapter, we will examine different types of RAG systems based on the architecture, advantages, and stages discussed in Chapter 6. We will examine each RAG system using one real-world example. This example will help you understand how to implement a RAG system for custom use cases. After discussing each RAG system, we will also look at its pros and cons, and the example with its results will help you understand which use cases you can solve using the different RAG systems.

7.1 Naïve RAG

7.1.1 Core Components of a Naïve RAG System

7.1.2 Building a Naïve RAG system from scratch

7.1.3 Pros, Cons, and Use Cases

7.2 Graph RAG

7.2.1 Components of a Graph RAG system

7.2.2 Workflow of a Graph RAG system

7.2.3 Building a Graph RAG system

7.2.4 Pros, Cons, and Use Cases

7.3 Hybrid RAG

7.3.1 Core Components of a Hybrid RAG System

7.3.2 Workflow of a Hybrid RAG System

7.3.3 Building a Hybrid RAG system

7.3.4 Pros, Cons, and Use Cases

7.4 Memory Augmented RAG System

7.4.1 Components of a Memory RAG system

7.4.2 Workflow of a Memory RAG System

7.4.3 Building a Memory Augmented RAG system.

7.4.4 Pros, Cons, and Use Cases

7.5 Which one to use?

7.6 Summary