RAG 101: Retrieval Augmented Generation

Hey there! I’m writing a blog series to help me document my journey learning in-depth about RAG and a maybe a lot of unrelated things alongside. Again, this is not a learning resource, it’s just me documenting my learning journey but you might find this useful, so feel free to follow along.

Cover image — represent RAG as a visual

Introduction

Any LLM (Large Language Model) you use is trained on some (large) amount of data and their knowledge is limited to this training data. RAG is a technique in AI (Artificial Intelligence) that improves text generation by combining two steps: retrieval and generation. Instead of relying only on pre-trained knowledge, RAG first searches for relevant information from an external database, documents, or knowledge source (retrieval) and then uses that information to generate a more accurate and context-aware response (generation). This approach makes AI more reliable, dynamic, and capable of providing up-to-date answers, even beyond its original training data.

Or think of it this way — Imagine you’re taking a test, you have two options:

1. Try to remember everything from your brain (like a normal LLM)
2. Look up the correct answer in a book and then explain it in your own words.