How to train a new model in Rasa without restarting the http server

Hi team, I want to add new training data as json format while Rasa is running as http server. How to train the model without restarting server. I can use \train api as per said in docs but there it’s also said that nlu data should be in markdown format. I can run rasa train from another terminal or some python script to do that to automate the process but right now I have to stop the rasa server and run it again. Is there any way to add new nlu data, train the model and load it without restarting the server ? Another thing I want to know if I call the \train api, will the new nlu data get append to the existing nlu data or just replace it and trains a new model based on the new data ?