Versionizing Rasa-X project using docker

[webgem-jpl (Julien P Lefebvre)]

Hi, I am working on a setup for Rasa X with docker.

When someone changes training data it does change the nlu and stories files accordingly, although if I am correct the annotated conversation are only in database.

When we train a model the model file is created in /models accordingly.

When run locally it is not hard to keep track of the file changes and commit them, but with Rasa-X running on server with docker what is the best way to commit the states of the project ? What about the annotated conversations, is there a way to export and import them as file so that it can be committed using git ?

Thanks

[ricwo]

Hi @webgem-jpl, welcome to the forum!

Changes to the training data are only dumped to .md files when run locally. On docker, the data only exists in database. You can download a file containing all your NLU training data in the UI (NLU training data -> training data -> download). You can also request a markdown dump of all your stories in the UI (Stories -> export stories). Your annotated conversations should appear in your stories tab as well (and can be downloaded).

I hope that helps!