Selecting the most relevant documents, chunks, or vectors from a corpus given a user query, typically via embeddings and similarity search.