Hi all,
I had this issue when upgrading rasa-x 0.41.1/2.7.1/2.7.1 implemented on kubernetes to 0.41.2/2.8.0/2.8.0.
I think the problem is linked to :
And the response is in :
How Rasa-x environment uses the tokens listed under rasa-x and rasa tags of the kubernete’s value.yml file is a bit confusing for me. The solution was to copy the token (reachable when you want to upload a model in rasa x) inside the tokens of rasa and rasa-x inside the values.yml file. I was thinking that those values were specifications for the cluster, it seems that it generates those tokens and we have to update them in the value.yml file and then upgrade the cluster with : helm upgrade --values ./values.yml --namespace NAMESPACE --reuse-values RELEASE_NAME rasa-x/rasa-x
Then you erase your models and train a new one, it will be working
Is that behaviour caused by the call of other rasa versions than the helm chart specifies (2.0.0 is linked to 0.41.1 and 2.7.1) ? Talking about that, what are the risks of doing such a thing ?