Do you always have to run rasa train when making changes to domain.yml for response text?

This might be a silly question, but do you always have to re-run rasa train when you make changes to domain.yml file? For example, sometimes we’re just changing the response text of a utter_[foo].

I’m just asking to learn.

1 Like

If you poke inside the trained model, inside a folder named core, you can see the domain file. If you make changes to the domain file, then it’ll need to be updated inside the model zip.(done during model training)

You can try copying the updated domain.yml file inside the model zip and check whether it’s taking the new response or not. (I haven’t tested this yet.) If it works, then you can skip the training step

Much appreciated the insight.

It would be very very nice if this was avoided, in cases when just responses text changes.