Building Agentic AI: Workflows, Fine-Tuning, Optimization, and Deployment

2

First Steps with LLM Workflows

Introduction

You’ve made it this far, which means you’re ready for a case study—so let’s get cooking. Our first major case study will be an LLM workflow with a type of agentic behavior called Retrieval Augmented Generation (RAG). The primary function of a RAG system is to answer questions for a user given the ability to look up information from some store of data (which I’ll also refer to as documents). The idea is that the LLM doesn’t have enough context to answer these questions alone and often needs to use external information when answering queries.

A RAG system generally has three parts:

• An indexer: A mechanism to compress raw text data into vectors; stored in a database.

• A retriever: Closely tied to the ...