Black Friday50% off your first month on any monthly plan with EARLY50 See plans
All 29 lessons
  1. Foundations: what RAG is and a working pipeline

    1. 1. Getting started: what RAG is and when to use it
    2. 2. Types of RAG: a map from naive to agentic
    3. 3. Lab: build the smallest RAG pipeline end to end
    4. 4. How embeddings and vector search find meaning
    5. 5. Document types: what each one breaks, and how to prepare it
  2. Better retrieval: chunking and search

    1. 6. Chunking documents so answers survive
    2. 7. Chunking strategies compared: which to use when
    3. 8. Keyword search vs vector search, and fusing both
    4. 9. Retrieval algorithms compared: which to use when
    5. 10. Reranking the top results before the model reads them
    6. 11. Rewriting the question before you search
  3. Answers you can trust, measure and debug

    1. 12. Lab: grounded answers with citations and refusals
    2. 13. Measuring a RAG system with an evaluation set
    3. 14. Lab: evaluation methods and frameworks: Ragas and DeepEval
    4. 15. Troubleshooting a RAG system, symptom by symptom
  4. Production in a company

    1. 16. Access control: only retrieve what the user may see
    2. 17. Defending against prompt injection in retrieved text
    3. 18. Running RAG in production: freshness, cost and monitoring
    4. 19. Real-world RAG: what worked, what failed, and why
  5. Agentic RAG

    1. 20. Lab: from a fixed pipeline to an agent with a search tool
    2. 21. Self-correcting retrieval: grade, retry, fall back
    3. 22. GraphRAG and routing across many sources
    4. 23. Lab: numbers, tables and financial documents
  6. Frameworks: LangChain, LangGraph and LangSmith

    1. 24. The RAG toolbox: what each library does and why you import it
    2. 25. The framework landscape: LangChain, LangGraph, LlamaIndex, Haystack
    3. 26. Lab: Ask Brightline in LangChain
    4. 27. Lab: agentic RAG as a LangGraph graph
    5. 28. Lab: tracing, evaluating and monitoring with LangSmith
    6. 29. Wrap-up: your production agentic RAG checklist
See the leaderboard

12 min

Types of RAG: a map from naive to agentic

Name the main kinds of RAG, from naive pipelines to graph, multimodal and agentic systems, and what problem each one solves.

By the end of this lesson you can

  • Name the main types of RAG and the problem each one solves
  • Tell naive, advanced, modular and agentic RAG apart
  • Pick a starting type for a new use case, including no retrieval at all

From earlier lessons · 1 of 2

What are the three stages of every RAG request?

Pick one answer.

From earlier lessons · 2 of 2

When can you skip RAG entirely?

Pick one answer.

Your manager forwards three vendor pitches. One sells "advanced RAG", one "GraphRAG", one "agentic RAG". Each claims the others are obsolete. They're not competing products. They're layers you add, one problem at a time.

Why there are so many kinds

Every type of RAG exists because the basic pipeline fails in one specific way. Learn the failure and you know the type.

The widely cited survey by Gao et al. (2023) sorts the field into three stages. Naive RAG is the basic retrieve-then-read pipeline: index, retrieve, generate. Advanced RAG improves the steps before and after retrieval. Modular RAG treats each step as a swappable module, so different questions can take different paths.

[74]

The map

Rendering diagram…
Each type adds one capability to fix one failure. Solid arrows build on each other; dotted ones are alternatives to text retrieval.

Each type in plain words

  1. Naive RAG. Chunk, embed, retrieve the top k, answer. It's the baseline in lesson 3. Use it first, and measure.
  2. Advanced RAG adds steps before retrieval (better chunking, metadata, query rewriting) and after it (reranking, compressing context). Use it when retrieval misses or adds noise. Lessons 6-11 build it.
  3. Hybrid RAG combines keyword and vector search and fuses the results. Use it when questions contain codes, names or jargon. It's lesson 8.
  4. Modular RAG builds the pipeline from swappable parts, such as a router, memory, a search module or a fusion step, so each question can take its own path. Lessons 22 and 26 use it.
  5. Adaptive RAG (Jeong et al., 2024) has a small classifier judge how hard each question is. It then sends it to no retrieval, one search, or several rounds. Use it when easy and hard questions mix, to save cost on the easy ones.
  6. GraphRAG builds an entity graph and pre-written community summaries, for questions about a whole corpus, such as main themes. It's lesson 22.
  7. Corrective RAG and Self-RAG grade what was retrieved and what was written, and retry or refuse. Use them when wrong answers are costly. It's lesson 21.
  8. Agentic RAG puts an agent in charge: it plans searches, picks sources and tools, and checks its work. Lessons 20 and 27 build it.
  9. Multimodal RAG retrieves page images, charts and slides directly with a vision model. Use it for visually rich documents. It's lesson 5.
  10. No retrieval. If the whole knowledge base fits in the prompt, long context with prompt caching can beat all of the above. That's lesson 1.

[4][6][7][69][76][77]

Agentic RAG, a closer look

The agentic RAG survey by Singh et al. describes these systems as embedding autonomous agents into the RAG pipeline. That enables dynamic retrieval, iterative refinement of context, and adaptive workflows. It groups them into a few architectures:

  • A single router agent picks among a few sources or tools.
  • Multi-agent systems give one specialist agent to each source or task, with a coordinator.
  • Hierarchical systems have top-level agents delegate to lower ones.
  • Corrective, adaptive and graph-based variants add grading, difficulty routing or graph reasoning to an agent.

Most company systems need only the first: one agent with a few well-described tools.

[75]

Think of a library help desk. Naive RAG is a clerk who fetches the three books whose titles best match your question. Advanced RAG is a clerk who first rephrases your question, checks two catalogues, and sorts the books by how well they fit. Agentic RAG is a research librarian. They plan the search, follow a lead to another archive and notice a gap. Then they come back with a sourced answer, or tell you none exists.

Match

Match each situation to the type to start with

Pick an item on the left, then its match on the right.

Worth keeping

  1. Every RAG type fixes one failure of the basic pipeline; learn the failure to know the type.
  2. Start with naive RAG and measure; add advanced steps only where it misses.
  3. Agentic RAG puts an agent in charge of searching and checking, at the cost of more model calls.
  4. Sometimes the best RAG is none: long context with caching for small, stable knowledge.

That's the map. Next lesson you build the first box on it, naive RAG, in about 40 lines of Python. One question it gets wrong will lead to everything else.

Leaderboard

Loading…

50 XP for each lesson you finish, 10 for each right answer. Names show only for learners who turn on Show me on leaderboards.

Sources for this lesson (8)

Each fact in the lesson is marked with the number of the page it comes from.

  1. [4]Edge et al. (2024), From Local to Global: A Graph RAG Approach
  2. [6]Asai et al. (2023), Self-RAG
  3. [7]Yan et al. (2024), Corrective Retrieval Augmented Generation
  4. [69]Faysse et al. (2024), ColPali
  5. [74]Gao et al. (2023), RAG for Large Language Models: A Survey
  6. [75]Singh et al. (2025), Agentic RAG: A Survey
  7. [76]Jeong et al. (2024), Adaptive-RAG
  8. [77]Sawarkar et al. (2024), Blended RAG