When we use the AI assistant nodes to read our data tables, my data table is massive, does this impact my token costs using my open AI assistant, or is this facilitated in buildship?