RAG supplies the model with passages fetched from an external store at inference time, so answers can rest on material the model was never trained on. For agents, retrieval is a tool like any other, which means the agent chooses when and what to retrieve.

Retrieval Augmented Generation

Pipeline RAG retrieves once, before generation. An agentic system retrieves repeatedly, refining the query from what the last retrieval revealed. The second pattern is markedly better on questions whose right query is only obvious after you have read something.

Agentic Retrieval Versus Pipeline RAG

A topic of the Building AI Agents course. Planned for 2 hours at Medium difficulty.

University of Michigan - Ann Arbor

A category of the Building AI Agents course, covering 4 topics.

Memory, Retrieval and Context Engineering

A topic of the Building AI Agents course. Planned for 1.5 hours at Medium difficulty.

The Context Window as a Budget

Retrieval Augmented Generation for Agents

Short Term and Long Term Memory

A topic of the Building AI Agents course. Planned for 1.5 hours at Hard difficulty.

Learn Before

Related

Learn After