Usually fsm can be uesed in dialoge system ,i dont see why rasa abandon it ?

Hello! You can model simple FSM using RulePolicy create default rasa project via rasa init command add this slot mappings to model state transitions intents: - greet - goodbye actions: - action_state_transition slots: state: type: categorical initial_value: greet influence_conversation: true values: - greet - goodbye mappings: - type: custom new_state: type: categorical initial_value: greet influence_conversation: false values: - greet - goodbye mappings: - type: from_intent intent: goodbye value: goodbye condition: - slot_was_set: - state: greet - type: from_intent intent: greet value: greet condition: - slot_was_set: - state: goodbye responses: utter_greet: - text: "Hey! How are you?" utter_happy: - text: "Great, carry on!" utter_goodbye: - text: "Bye" utter_iamabot: - text: "I am a bot, powered by Rasa." define state handlers as follows: rules: # GREET - rule: Entrypoint condition: - slot_was_set: - state: greet steps: - action: action_state_transition - action: utter_greet - rule: Bye condition: - slot_was_set: - state: greet steps: - intent: goodbye - action: utter_iamabot - action: action_state_transition wait_for_user_input: false - rule: Greet condition: - slot_was_set: - state: greet steps: - intent: greet - action: utter_happy # GOODBYE - rule: Entrypoint condition: - slot_was_set: - state: goodbye steps: - action: action_state_transition - action: utter_goodbye - rule: Bye condition: - slot_was_set: - state: goodbye steps: - intent: goodbye - action: utter_happy - rule: Greet condition: - slot_was_set: - state: goodbye steps: - intent: greet - action: utter_iamabot - action: action_state_transition wait_for_user_input: false add custom action: from typing import Any, Text, Dict, List from rasa_sdk import Action, Tracker from rasa_sdk.events import SlotSet from rasa_sdk.executor import CollectingDispatcher class ActionStateTransition(Action): def name(self) -> Text: return "action_state_transition" def run( self, dispatcher: CollectingDispatcher, tracker: Tracker, domain: Dict[Text, Any] ) -> List[Dict[Text, Any]]: return [SlotSet('state', tracker.get_slot('new_state'))] State Machine is: 2 states greet and goodbye on intent goodbye being in state greet go to state goodbye on intent greet being in state goodbye go to state greet on intent greet being in state greet no transition on intent goodbye being in state goodbye no transition You are also able to use mapping conditions to model simple constrain of transitions and custom mappings to model complex transition logic Thank you!

Why rasa dont support Policy based ont FSM? Is that possible by developing custom policy?

Rasa Open Source

Ulov888 (Ulov888) June 20, 2022, 5:50am 2

@ akelad@ chkosscan i get your help?

Topic		Replies	Views
Dialog Engine ML vs State Machine Rasa Open Source	5	1700	August 13, 2020
Need Guidance on Action Loops Rasa Open Source	1	17	April 12, 2025
How to create custom policy? Rasa Open Source	3	588	August 18, 2020
Can I make decisions based on slots? Rasa Open Source	3	277	August 23, 2023
Rasa will not save slot from entity of intent Getting Started with Rasa	1	492	March 4, 2023

Why rasa dont support Policy based ont FSM? Is that possible by developing custom policy?

Related topics