Retrieval as a Graph Step

RAG (Retrieval Augmented Generation) is often taught as a separate architecture. In LangGraph, it's just a subset of nodes.

The RAG Subgraph

1. RetrieveNode: Takes query, returns docs.
2. GradeNode (Optional): LLM checks if docs are relevant.
3. GenerateNode: Takes docs + query, returns answer.

By making retrieval a discrete step, you can verify it. You can inspect the State after Step 1 to see exactly what the chunks were.