Skip to slide
095 / 221

CH 07 · Retrieval: RAG vs Tools vs Long Context · 07/07

Read as one page

The bottom line

Don't reach for a vector database reflexively because "RAG" is the default story. For many agents, especially those working with a user's own identified documents and needing exact, citable, current answers, giving the model good read/find/list tools is simpler, more accurate, and more maintainable. Add semantic search as a tool when, and only when, corpus scale demands it.

Finished this idea?

Your learning progress stays in this browser.