Use the API at /v1/docs instead of scraping HTML.