Warning: This document is for an old version of Rasa Core. The latest version is 0.14.5.

Actions

Contents

• Actions
  • Utter Actions
  • Custom Actions
  • Custom Actions Written in Python
  • Default Actions
  • Execute Actions in Other Code
    • Action Server API
  • Have questions or feedback?

Actions are the things your bot runs in response to user input. There are three kinds of actions in Rasa Core:

1. default actions (action_listen, action_restart, action_default_fallback)
2. utter actions, starting with utter_, which just sends a message to the user (see Bot Responses).
3. custom actions - any other action, these actions can run arbitrary code

Utter Actions

To define an UtterAction, add an utterance template to the domain file, that starts with utter_:

templates:
  utter_my_message:
    - "this is what I want my action to say!"

It is conventional to start the name of an UtterAction with utter_. If this prefix is missing, you can still use the template in your custom actions, but the template can not be directly predicted as its own action. See Bot Responses for more details.

If you use an external NLG service, you don’t need to specify the templates in the domain, but you still need to add the utterance names to the actions list of the domain.

Custom Actions

An action can run any code you want. Custom actions can turn on the lights, add an event to a calendar, check a user’s bank balance, or anything else you can imagine.

Core will call an endpoint you can specify, when a custom action is predicted. This endpoint should be a webserver that reacts to this call, runs the code and optionally returns information to modify the dialogue state.

To specify, your action server use the endpoints.yml:

action_endpoint:
  url: "http://localhost:5055/webhook"

And pass it to the scripts using --endpoints endpoints.yml.

You can create an action server in node.js, .NET, java, or any other language and define your actions there - but we provide a small python sdk to make development there even easier.

Custom Actions Written in Python

For actions written in python, we have a convenient SDK which starts this action server for you.

The only thing your action server needs to install is rasa_core_sdk:

pip install rasa_core_sdk

Note

You do not need to install rasa_core for your action server. E.g. it is recommended to run Rasa Core in a docker container and create a separate container for your action server. In this separate container, you only need to install rasa_core_sdk.

If your actions are defined in a file called actions.py, run this command:

python -m rasa_core_sdk.endpoint --actions actions

In a restaurant bot, if the user says “show me a Mexican restaurant”, your bot could execute the action ActionCheckRestaurants, which might look like this:

from rasa_core_sdk import Action
from rasa_core_sdk.events import SlotSet

class ActionCheckRestaurants(Action):
   def name(self):
      # type: () -> Text
      return "action_check_restaurants"

   def run(self, dispatcher, tracker, domain):
      # type: (CollectingDispatcher, Tracker, Dict[Text, Any]) -> List[Dict[Text, Any]]

      cuisine = tracker.get_slot('cuisine')
      q = "select * from restaurants where cuisine='{0}' limit 1".format(cuisine)
      result = db.query(q)

      return [SlotSet("matches", result if result is not None else [])]

You should add the the action name action_check_restaurants to the actions in your domain file. The action’s run method receives three arguments. You can access the values of slots and the latest message sent by the user using the tracker object, and you can send messages back to the user with the dispatcher object, by calling dispatcher.utter_template, dispatcher.utter_message, or any other rasa_core_sdk.executor.CollectingDispatcher method.

Details of the run method:

Action.run(dispatcher, tracker, domain)

Execute the side effects of this action.

Parameters:

• dispatcher (CollectingDispatcher) – the dispatcher which is used to send messages back to the user. Use dipatcher.utter_message() or any other rasa_core_sdk.executor.CollectingDispatcher method.
• tracker (Tracker) – the state tracker for the current user. You can access slot values using tracker.get_slot(slot_name), the most recent user message is tracker.latest_message.text and any other rasa_core_sdk.Tracker property.
• domain (Dict[Text, Any]) – the bot’s domain

Returns:

A dictionary of

rasa_core_sdk.events.Event instances that is returned through the endpoint

Return type:

List[Dict[Text, Any]]

There is an example of a SlotSet event below, and a full list of possible events in Events.

Default Actions

There are three default actions:

action_listen	stop predicting more actions and wait for user input
action_restart	reset the whole conversation, usually triggered by using /restart
action_default_fallback	undoes the last user message (as if the user did not send it) and utters a message that the bot did not understand. See Fallback Actions.

All the default actions can be overwritten. To do so, add the action name to the list of actions in your domain:

actions:
- action_listen

Rasa Core will then call your action endpoint and treat it as every other custom action.

Execute Actions in Other Code

Rasa Core will send an HTTP POST request to your server containing information on which action to run. Furthermore, this request will contain all information about the conversation.

As a response to the action call from Core, you can modify the tracker, e.g. by setting slots and send responses back to the user. All of the modifications are done using events. There is a list of all possible event types in Events.

Action Server API

Documentation of the action server API as OpenAPI Spec.

Have questions or feedback?

We have a very active support community on Rasa Community Forum that is happy to help you with your questions. If you have any feedback for us or a specific suggestion for improving the docs, feel free to share it by creating an issue on Rasa Core GitHub repository.