RAG
Also: retrieval-augmented generation
Retrieval-augmented generation: giving a model relevant snippets from your own documents at question time, so it can answer from them rather than from memory.
RAG, short for retrieval-augmented generation, is a way to let a model answer questions about information it was never trained on, such as your own documents. Rather than fine-tuning the model, you retrieve the most relevant passages at the moment you ask, and paste them into the prompt for the model to work from.
The usual recipe is to split your documents into chunks, turn each into an embedding, and store them in a vector database. When a question comes in, you embed it, find the closest chunks, and hand those to the model along with the question.
RAG is often the right first tool when you want a model to know about your own material, because it is simpler than fine-tuning, keeps your data under your control, and updates instantly when your documents change. Its quality depends on retrieving the right passages, which is where most of the practical work lies.