I want to make my Rasa model modular, i.e. split it at least into NLU and (Dialogue Management & NLG).
The purpose of this is, that I can use e.g. other trained NLUs.
Is there a possibility to pass already parsed input, i.e. {“text”: “This is the input text”, “intent”: “input_intent”, “entities”: […]} to the processor/agent?
Could you explain the usecase that you have here? That might help me point you in the right direction.
In general you can always use a custom action to receive the text, entities and entities in order to handle it with custom python code. That might be sufficient for your usecase.