Rasa and Pytorch model session persistence in production

Hi, i have a complex question on behavior of Pytorch model and Rasa.

I have a Rasa chatbot, standard with trained nlu and domain data. Also, i have my custom QA model built with Pytorch popular Ai lib for DL. I have created a custom action to trigger my model when some intent is detected - it works.

My model uses a context of dialogue, i store it in history list. Every time action is triggered i store question-answer pair in history because every next invocation of model uses this history to generate contextual-aware answer.

Question: how do i implement session for every user for my action? I need a separate history to alter my model generated text for every enduser that talk to bot. (Maybe session already implemented - but i am not sure how to test) Should i store history inside or outside of action class?

tracker is the place to store user dependent data. For example, you can create unfeaturized slot and write your history in there

Is there any tutorial?

for creating a slot?