Nlu & stories from a database?

Hi there, just wondering if its possible to have nlu (intents) and stories loaded from a database?

Scenario is if you wanted your chatbot to be more open-domain, so have a large list loaded from a database … is there a limit in Rasa around this?

During the training or at inference (runtime)?

I have a similar (I suspect) question - is it possible to update the domain.yml file programmaticaly (via an API) at training time to add intents, entities, slots, etc?

My use case is having a set of data sources in the file and wanting to automatically update objects in the domain.yml file (especially intents) based on the initial parsing of the data sources in Python. A different use case than the original poster, but I suspect that the main requirement to update Rasa control files (domain.yml,, programmatically is the same.

During training

@vimalnar @ryanmark You will want to write your own Training Data Importers. There is also a blog post which explains it a bit more: Customize how Rasa imports training data