Grouping your training data in Rasa v.2

Did you know you can group your domain.yml,nlu.yml,stories.nlu and even your actions.py in the same directory making it really easy to find everything and work one ability / skill / scenario easily? This allows you to distribute the workload to multiple developers / subject matter experts to build a library of abilities without them stepping on each other.

You can even break apart the individual keys into separate files if that makes sense to you as well. For example you could create an intents.yml, entities.yml, slots.yml, responses.yml, forms.yml etc. It’s super flexible.

Let me know what you think and how you organize your data.

Hi, I really like this idea and tested it locally and it works almost perfectly. (There is a little bug running rasa test nlu).

But I am concerned with Rasa X integration cos it expects a typical project structure and uses normal rasa commands. Have you tested it with Rasa X and is everything was fine?

I was looking for this project structure cos I am now looking to move to rasa form different instrument and this is almost key feature for developers. I have several developers working on аssistant and it quite big with a lot of wide spread function, so even keeping domain in one file will be a huge pain.

Unfortunately right now, RasaX doesn’t work well with this structure.

I’ve tried a few different things and it doesn’t seem to find everything when it trains and then when you alter things, it really wants to dump everything into a domain.yml file in the root of the project.

I’ve mentioned this and hopefully in a future version this will catch on and it’ll be supported.