Train models on sever in runtime

I’m using Rasa 1.2 and need to update my rasa model in runtime using the http server. I use /model/train endpoint to train my model but I have a couple of questions regarding that:

  1. Is it possible to fix the name of newly trained model? Something similar as --fixed-model-name if I was to train it using the command line.
  2. According to the documentation, for the payload of /model/train endpoint I need to provide the domain, config, nlu and stories all in plain text. Is there any way to pass just the file names for these fields in the payload and not the content of the file? If currently there is no way, is there any plan in near future to support this feature?


Hey @Elham welcome to the community and sorry for the late reply.

  1. No, that’s not possible at the moment.
  2. I’m afraid at the moment you have to pass it in the format provided. You could use Rasa X to make this training process via API easier though