Action_default_fallback is invoked even I have a custom action in my story

Rasa Open Source
1

I am following youtube videos on Rasa open source. I wanted to set my slot using custom action and then retrieve it again by a custom action. Slot is setting correctly but the next action which is to retrieve slot value is overridden by action_default_fallback.

NLU.yml

- intent: tell_name

examples: | - I want to tell you my name. - give name - do you know my name? - i’ll give you my name.

  • intent: repeat_name examples: |

    • what is my name
    • tell me my name
    • do you know my name
    • repeat my name

  • intent: give_name examples: |

    • Rajat
    • Vincent
    • Anand
    • Rahul
    • Ashish

STORY.yml

  • story: name path steps:
    • intent: greet
    • action: utter_greet
    • intent: tell_name
    • action: utter_ask_name
    • intent: give_name
    • action: action_receive_name
    • intent: repeat_name
    • action: action_say_name

DOMAIN.yml version: “2.0”

intents:

  • greet
  • goodbye
  • affirm
  • deny
  • mood_great
  • mood_unhappy
  • bot_challenge
  • tell_name
  • give_name
  • repeat_name

entities:

slots: name: type: text

actions:

  • action_receive_name
  • action_say_name

responses: utter_greet:

  • text: “Hey! How are you?”

utter_cheer_up:

utter_did_that_help:

  • text: “Did that help you?”

utter_happy:

  • text: “Great, carry on!”

utter_goodbye:

  • text: “Bye”

utter_iamabot:

  • text: “I am a bot, powered by Rasa.”

utter_ask_name:

  • text: “What is your name?”

session_config: session_expiration_time: 60 carry_over_slots_to_new_session: true

actions.py

from typing import Any, Text, Dict, List

from rasa_sdk.events import SlotSet from rasa_sdk import Action, Tracker from rasa_sdk.executor import CollectingDispatcher

class ActionReceiveName(Action):

def name(self) -> Text:
    return "action_receive_name"

def run(self, dispatcher: CollectingDispatcher,
        tracker: Tracker,
        domain: Dict[Text, Any]) -> List[Dict[Text, Any]]:
    text = tracker.latest_message['text']
    print("this action class is working")
    dispatcher.utter_message(text=f"I will remember your name : {text}")

    return [SlotSet("name", text)]

class ActionSayName(Action):

def name(self) -> Text:
    return "action_say_name"

def run(self, dispatcher: CollectingDispatcher,
        tracker: Tracker,
        domain: Dict[Text, Any]) -> List[Dict[Text, Any]]:
    name = tracker.get_slot("name")
    print("rajat"+"  "+name)
    if not name:
        dispatcher.utter_message(text="I don't know your name :(")
    else:
        dispatcher.utter_message(text=f"Your name is {name}!")
    return []

Instead of getting name from second custom action, I am getting “You input” prompt.

2

Hi @Rajat. The action_default_fallback is usually triggered when you’re saying an utterance that does not have enough confidence. It shouldn’t remove any slot-values though, so I’m curious what exactly happened in the conversation. Glancing at your custom actions, I’d say your code looks good.

Could you share a snippet of the conversation from a rasa interactive session? That way we want see the confidence levels, as well as the slots that have been set in a conversation.

3

Hi @koaning , I did that and surprisingly that did resolve the issue, I ran rasa interactive and chose the correct action that has to be executed and it made a slight change in stories.yml by adding slot_was_set tag

  • action: action_receive_name
    • slot_was_set:
      • name: rajat

After this assistant worked as desired. I assume that’s the correct solution.

4

Looks like it!

The nice thing about rasa interactive is that it will properly set slots in your stories on your behalf when you correct it. That is why I always recommend using it early since it also acts as a linter of sorts.