The API, essentially, works in three steps, including calling the API, asking it to generate the number of questions, and then setting natural language guidelines to assist the synthetic generation.