Rather than vectorizing and storing your entire data corpus, vectorize only the subset relevant to each query to keep storage and compute costs manageable.
