Vocabulary
RAG
Retrieval-augmented generation: fetch relevant notes or docs first, then let the model answer with that material in view.
Also called: retrieval-augmented generation
RAG is a pattern, not a product. Before the model answers, a search step pulls passages that might matter (from your files, a wiki, a ticket system). Those passages are stuffed into the prompt so the model can ground its reply.
It became the default tutorial answer for “how do I make AI know my stuff?” Long context windows changed the tradeoffs: sometimes you can just include the document. RAG still wins when the corpus is huge, changing, or expensive to load every time.
The hard part is rarely the vector database logo. It is chunking, freshness, permissions, and knowing when retrieval returned the wrong neighbor.
Cups that use this term