I think that you should set the facedetect endpoint to be of type “text” instead of json. Use regex to only keep array and provide this result to cord parameter.