RAG
Retrieval-Augmented Generation
Fetch the relevant documents first, then let the model answer using them.
Retrieval-augmented generation splits the work in two. First, a search step finds passages relevant to the question. Second, the model gets those passages in its context and writes the answer.
This is the standard answer to "the model does not know our data". It beats fine-tuning for factual freshness because you update the index, not the weights, and every answer can point at the source passage.
RAG quality lives in retrieval, not in the generation step. Chunking, embeddings, and ranking decide whether the right passage is even present in the context. If retrieval returns the wrong documents, a better model cannot rescue the answer.
Related terms
Embedding
A list of numbers that places text near other text with similar meaning.
Vector Database
Storage and search built for embeddings rather than exact values.
Grounding
Tying the answer to something checkable, so it can be verified instead of trusted.
Hallucination
Output that is fluent and wrong. Two different problems hide under one word.
Context Engineering
Deciding what goes into the model's context, and what stays out.
Want this working inside your team?
It starts with one conversation about your workflow.