Having to explicitly choose the API endpoint, depending on the command type, is awkward. A new endpoint, that contained a built-in SQL parser, would be best. It would route the statements to the correct endpoints (or internal calls) as needed.