Anybody played around with an LLM and some context with RAG?

My understanding is that documents are fed to the LLM as context for a prompt, and it can answer questions on them.

Semantic search using this technology on huge text documents will be so cool, as opposed to regular keyword search 👀 🔍