Expertise / RAG

Foundations.

What RAG is, why it beats fine-tuning in most cases, open-book vs closed-book modes.

The RAG architecture map

Every production RAG system has the same core layers. Here's the full architecture map and what decisions each layer forces.

What is RAG?

Retrieval-Augmented Generation (RAG) is the architectural pattern that lets an LLM answer questions using your data. Here's what it actually is and what it isn't.

When not to use RAG

RAG isn't always the right answer. Here are the scenarios where it adds complexity without improving outcomes and what to use instead.

Why RAG over fine-tuning

Fine-tuning teaches a model new behaviors. RAG gives a model new facts. Most of what teams think they want from fine-tuning is actually a RAG problem.
Screenshot of pinecone.io: Pinecone: RAG guide
pinecone.ioPinecone: RAG guide
 Screenshot of docs.llamaindex.ai: LlamaIndex docs
docs.llamaindex.aiLlamaIndex docs

Further reading

Watch

3Blue1Brown - Attention in transformers, visually explained