Retrieval-augmented generation (RAG)

What Retrieval-augmented generation (RAG) means

A language model on its own answers from patterns learned in training. With retrieval-augmented generation, the system first searches a collection of documents (for example legislation or your own files), picks the most relevant passages and places them in front of the model with the question.

RAG makes it possible to cite the passages an answer relied on and to work with material newer than the model’s training. It does not guarantee correctness: retrieval can miss the right passage, and the model can still misread what it was given.

In Consilena

Consilena searches its law library and your documents before answering, and each claim carries a numbered source that opens the passage it relied on.