Skip to main content

Introduction

TL;DR
  • This course goes from traditional RAG → advanced → multimodal → agentic.
  • Built on LangChain (pipelines), LangGraph (agents), LangSmith (tracing).
  • Learn a concept → study a real project → rebuild it yourself.

The opening section sets the scope: this is an end-to-end RAG course that goes from traditional RAG all the way to advanced, multimodal, and agentic systems, built with LangChain, LangGraph, and LangSmith.

What the course builds toward​

  • Traditional RAG — load → parse → chunk → embed → store → retrieve → generate.
  • Advanced retrieval — hybrid search, query enhancement, advanced chunking.
  • Multimodal RAG — retrieving over text and images.
  • Agentic RAG — agents that decide when and how to retrieve, plus corrective, adaptive, autonomous, and multi-agent variants (LangGraph).
  • Evaluation & production — LangSmith tracing, guardrails, LLM gateways, GraphDB, and a final end-to-end project.

How I'm taking these notes​

Each note maps to a course topic, in order, so I can study the video and reinforce it here. I write the concept in my own words, keep the code I actually understand, and note the pitfalls.

Cheat sheet​

  • Stack: LangChain (pipelines) · LangGraph (agents) · LangSmith (tracing/eval).
  • Arc: traditional RAG → advanced retrieval → multimodal → agentic → production.
  • Method: learn the concept, study a real project, rebuild it yourself.

Master the traditional RAG pipeline first, then layer on advanced retrieval, then graduate to agentic RAG — the same order these notes follow.

Quick self-check

What three layers does this course build through?

Traditional RAG → advanced/multimodal RAG → agentic RAG.

What does each tool in the stack do?

LangChain builds the pipelines, LangGraph builds the agents, LangSmith traces and evaluates them.

Related: Glossary · Introduction to RAG →