A purely local solution cannot leverage cloud computing power for semantic search; implementing semantic search locally requires running an embedding model on the device, placing certain demands on computing resources and storage