RAG and agents
Note
This section is in progress. RAG is named in the Generative AI overview and has a child page that is still an outline. There is no agents article yet.
Why this section exists
A frozen language model only knows what it was trained on. Retrieval-Augmented Generation (RAG) is one way to give it fresh, inspectable context: retrieve, then generate. Agents add a loop — tools, memory, and a next step — on top of that.
I am not starting with a hedge-fund fantasy. The first honest pages here should be a small pipeline a student can follow.
Already here
- Generative AI RAG — a short outline of what that page will become (it used to be the Quarto starter text).
- A January 2025 bookmark, Lab#RAG001, points at an external agents write-up. That is a link list, not a classroom lab.
Coming next
- Why a frozen model is not enough.
- Embeddings and retrieval, in plain language.
- One honest retrieve → augment → generate diagram.
- A first agents page: tools, memory, the loop.