MappingPolicy and TwoStageFallbackPolicy, chatbot is stuck on action_default_ask_affirmation

Hi, TLDR: after triggering response my chatbot is stuck on fallbackpolicy, even if confidence = 1.0

nlu.md:

## intent:chitchat_weather
- what is the weather
- tell me weather

domain.yml:

intents:
  - chitchat_weather: {triggers: utter_chitchat_weather}
templates:
  utter_chitchat_weather:
    - text: It's raining

TwoStageFallback config:

- name: TwoStageFallbackPolicy
  nlu_threshold: 0.3
  ambiguity_threshold: 0.3
  core_threshold: 0.3
  fallback_core_action_name: "action_default_ask_affirmation"
  fallback_nlu_action_name: "action_default_ask_affirmation"
  deny_suggestion_intent_name: "out_of_scope"

Fallback action is implemented just like here: Open source Two Stage Fallback Policy with Rasa NLU and Rasa Core

example conversation:

  • hey
  • hey, how are you?
  • what is the weather? (this message triggers utter_chitchat_weather)
  • It’s raining
  • hey
  • Did you mean… (3 buttons)
  • /greet
  • Did you mean… (3 buttons)

And chatbot gets stuck here, no matter which message is passed, even for intent name (/greet), when confidence = 1.0

[{'name': 'greet', 'confidence': 1.0}]

chatbot responds with action_default_ask_affirmation.

The TwoStageFallbackPolicy works fine, there is only problem with intent triggering

Hey, did you provide a story for this case?

I’ve just solved this. Apparently when using both MappingPolicy and FallbackPolicy / TwoStageFallbackPolicy, it’s necessary to add AugmentedMemoizationPolicy instead of MemoizationPolicy.