Use fastApi instead of HTTPServer so we have the same framework in the ml and the cloudCompile subsystems