I appreciate the collaboration. Can someone clarify this question for me? If I have a node in n8n from a vector store and set the number of similarities to 4, as shown in the image, but when the AI interprets it, the response is much shorter and only taken from one vector—are tokens consumed based on all four results or just on the final single vector result