Working of Two stage fallback policy

Hello Everyone, I Need the help in understanding the Two Stage Fallback policy. The Example would be a great help. thanks.

Check out the docs on the TwoStageFallbackPolicy here as well as the docs on fallback actions.

We also wrote a blog post about it upon release.

@erohmensing Can you also share a small working POC of TwoStageFallbackPolicy. That will clear out most of our doubts.

The post is good but it doesn’t tell us upfront which action is called for affirmation from user when nlu classification threshold is low.

From the docs, I could only figure out that final fallback action called is based on the intent specified in deny_suggestion_intent_name, when the classification fails for the second time.

Hi @tonysinghmss, our demo bot uses the two stage fallback policy: GitHub - RasaHQ/rasa-demo: Sara - the Rasa Demo Bot: An example of a contextual AI assistant built with the open source Rasa Stack

@erohmensing, just a suggestion, can you people add more description in the docs, some code examples. like I was following

to implement two stage fallback using FallbackClassifer.

but that is not working. From another doc i read, you have mentioned some params for RulePolicy. I added that then only i was able to make this working. Alone Fallback document was not sufficent to make it working.

And may be the code flow you can add, so that we can get better idea how things are working, that will be very helpful.

And can you help me on this?

thanks a lot :slight_smile: