am I doing something wrong, or there is a way to throttle the requests to openAI