Continuing the discussion from How to change model runtime?:
If you are using Rasa as a server, you can change the model with the HTTP API..
However, if you wish to dynamically swap out a model as stated in the previous post (e.g. when processing a new message), you’d have to write some custom implementation for that, since Rasa only supports one NLU model per conversation. Could you state what problem you’d like this functionality to solve? Perhaps there is another solution