Hi, can anybody tell me that why or how much 2 min 15 sec call in assistant is processing 1,030,189 tokens making it unbearable cost.... i added this capability in attached mcp to load the menu json in the start of the call, which is almost 7,300 tokens. and system prompt is 600 tokens. how can it reach more than a million token processing within 2 min and 15 sec.? What am i doing wrong , if anybody can help?