Responses - Rasa Documentation

Defining Responses

Responses go under the responses key in your domain file or in a separate “responses.yml” file. Each response name should start with utter_. For example, you could add responses for greeting and saying goodbye under the response names utter_greet and utter_bye:

domain.yml

intents:
  - greet

responses:
  utter_greet:
  - text: "Hi there!"
  utter_bye:
  - text: "See you!"

If you are using retrieval intents in your assistant, you also need to add responses for your assistant’s replies to these intents:

intents:
  - chitchat

responses:
  utter_chitchat/ask_name:
  - text: Oh yeah, I am called the retrieval bot.

  utter_chitchat/ask_weather:
  - text: Oh, it does look sunny right now in Berlin.

Notice the special format of response names for retrieval intents. Each name starts with utter_, followed by the retrieval intent’s name (here chitchat) and finally a suffix specifying the different response keys (here ask_name and ask_weather). See the documentation for NLU training examples to learn more.

Using Variables in Responses

You can use variables to insert information into responses. Within a response, a variable is enclosed in curly brackets. For example, see the variable name below:

domain.yml

responses:
  utter_greet:
  - text: "Hey, {name}. How are you?"

When the utter_greet response is used, Rasa automatically fills in the variable with the value found in the slot called name. If such a slot doesn’t exist or is empty, the variable gets filled with None. Another way to fill in a variable is within a custom action. In your custom action code, you can supply values to a response to fill in specific variables. If you’re using the Rasa SDK for your action server, you can pass a value for the variable as a keyword argument to dispatcher.utter_message:

dispatcher.utter_message(
    template="utter_greet",
    name="Sara"
)

If you use a different custom action server, supply the values by adding extra parameters to the responses your server returns:

{
  "events":[
    ...
  ],
  "responses":[
    {
      "template":"utter_greet",
      "name":"Sara"
    }
  ]
}

Response Variations

You can make your assistant’s replies more interesting if you provide multiple response variations to choose from for a given response name:

domain.yml

responses:
  utter_greet:
  - text: "Hey, {name}. How are you?"
  - text: "Hey, {name}. How is your day going?"

In this example, when utter_greet gets predicted as the next action, Rasa will randomly pick one of the two response variations to use.

IDs for Responses

New in Rasa 3.6You can now set an ID for any response. This is useful when you want to use the NLG server to generate the response.Type for ID is string.

Example of response variations with ID:

domain.yml

responses:
  utter_greet:
  - id: "greet_1"
    text: "Hey, {name}. How are you?"
  - id: "greet_2"
    text: "Hey, {name}. How is your day going?"

Channel-Specific Response Variations

To specify different response variations depending on which channel the user is connected to, use channel-specific response variations. In the following example, the channel key makes the first response variation channel-specific for the slack channel while the second variation is not channel-specific:

domain.yml

responses:
  utter_ask_game:
  - text: "Which game would you like to play on Slack?"
    channel: "slack"
  - text: "Which game would you like to play?"

Make sure the value of the channel key matches the value returned by the name() method of your input channel. If you are using a built-in channel, this value will also match the channel name used in your credentials.yml file.

When your assistant looks for suitable response variations under a given response name, it will first try to choose from channel-specific variations for the current channel. If there are no such variations, the assistant will choose from any response variations which are not channel-specific. In the above example, the second response variation has no channel specified and can be used by your assistant for all channels other than slack.

For each response, try to have at least one response variation without the channel key. This allows your assistant to properly respond in all environments, such as in new channels, in the shell and in interactive learning.

Conditional Response Variations

Specific response variations can also be selected based on one or more slot values using a conditional response variation. A conditional response variation is defined in the domain or responses YAML files similarly to a standard response variation but with an additional condition key.

Predicate Conditions

New in 3.13You can now use predicate expressions to define conditions using the slots namespace for conditional response variations.This feature was originally released as a beta feature in Rasa Pro 3.12.0 and is now generally available (GA).

The condition key can now specify a predicate statement, similar to the usage of conditions in flows. These predicates enable you to use a variety of logical operators, comparison operators and other constructs. They are evaluated with the pypred library. For example:

domain.yml

responses:
  utter_greet:
    - condition: slots.prior_visits > 1
      text: "Hey, {name}. Nice to see you again! How are you?"

    - condition: not slots.prior_visits
      text: "Welcome. How is your day going?"

In the example above, the first response variation ("Hey, {name}. Nice to see you again! How are you?") will be used whenever the utter_greet action is executed and the prior_visits slot is greater than 1. The second variation, which has a condition that the prior_visits slot is not set, will be used when the slot is not set.

`Name` and `Value` Equality Constraints

Deprecation WarningWriting the condition as a list of dictionaries consisting of name and value constraints is deprecated. This format will be removed in the next major release of Rasa.

The condition key specifies a list of slot name and value constraints. When a response is triggered during a dialogue, the constraints of each conditional response variation are checked against the current dialogue state. If all constraint slot values are equal to the corresponding slot values of the current dialogue state, the response variation is eligible to be used by your conversational assistant.

The comparison of dialogue state slot values and constraint slot values is performed by the equality ”==” operator which requires the type of slot values to match too. For example, if the constraint is specified as value: true, then the slot needs to be filled with a boolean true, not the string "true".

In the following example, we will define one conditional response variation with one constraint, that the logged_in slot is set to true:

domain.yml

slots:
  logged_in:
    type: bool
    mappings:
      - type: controlled
  name:
    type: text

responses:
  utter_greet:
    - condition:
        - type: slot
          name: logged_in
          value: true
      text: "Hey, {name}. Nice to see you again! How are you?"

    - text: "Welcome. How is your day going?"

flows.yml

flows:
  Greet:
    name: Greet
    description: This flow is called to greet customers at the start of the conversation.
    steps:
      - action: logged_in
      - action: utter_Greet

In the example above, the first response variation ("Hey, {name}. Nice to see you again! How are you?") will be used whenever the utter_greet action is executed and the logged_in slot is set to true. The second variation, which has no condition, will be treated as the default and used whenever logged_in is not equal to true.

Variation Selection

During a dialogue, Rasa will choose from all conditional response variations whose constraints are satisfied. If there are multiple eligible conditional response variations, Rasa will pick one at random. For example, consider the following response:

domain.yml

responses:
  utter_greet:
    - condition: slots.logged_in is true
      text: "Hey, {name}. Nice to see you again! How are you?"

    - condition: slots.eligible_for_upgrade is true
      text: "Welcome, {name}. Did you know you are eligible for a free upgrade?"

    - text: "Welcome. How is your day going?"

If logged_in and eligible_for_upgrade are both set to true then both the first and second response variations are eligible to be used, and will be chosen by the conversational assistant with equal probability. You can continue using channel-specific response variations alongside conditional response variations as shown in the example below.

domain.yml

slots:
  logged_in:
    type: bool
    mappings:
      - type: controlled
  name:
    type: text

responses:
  utter_greet:
    - condition: slots.logged_in is true
      text: "Hey, {name}. Nice to see you again on Slack! How are you?"
      channel: slack

    - text: "Welcome. How is your day going?"

Rasa will prioritize the selection of responses in the following order:

conditional response variations with matching channel
default responses with matching channel
conditional response variations with no matching channel
default responses with no matching channel

It is highly recommended to always provide a default response variation without a condition to guard against those cases when no conditional response matches filled slots. If no conditional response variation is eligible to be used and no default is provided, Rasa will respond with the default utterance utter_internal_error_rasa.

Rich Responses

You can make responses rich by adding visual and interactive elements. There are several types of elements that are supported across many channels:

Buttons

You can add buttons to a response to allow the user to select from a list of options. The buttons are displayed as clickable elements in the chat window. Each button in the list of buttons should have two keys:

title: The text displayed on the buttons that the user sees.
payload: The message sent from the user to the assistant when the button is clicked.

Button payloads can be used to either:

trigger intents and pass entities to the assistant.
issue commands to set slots
pass a predefined free-form string message to the assistant. Note that this option should be used if none of the above options are feasible.

In addition, buttons provide the advantage of skipping the NLU pipeline and directly annotating the user message with the intent, entities or set slot commands defined in the payload.

Check your channelKeep in mind that it is up to the implementation of the output channel how to display the defined buttons. For example, some channels have a limit on the number of buttons you can provide. Check your channel’s documentation under Concepts > Channel Connectors for any channel-specific restrictions.

Triggering Intents or Passing Entities

Here is an example of a response that uses buttons to trigger an intent:

domain.yml

responses:
  utter_greet:
  - text: "Hey! How are you?"
    buttons:
    - title: "great"
      payload: "/mood_great"
    - title: "super sad"
      payload: "/mood_sad"

If you would like the buttons to also pass entities to the assistant:

domain.yml

responses:
  utter_greet:
  - text: "Hey! Would you like to purchase motor or home insurance?"
    buttons:
    - title: "Motor insurance"
      payload: '/inform{{"insurance":"motor"}}'
    - title: "Home insurance"
      payload: '/inform{{"insurance":"home"}}'

Passing multiple entities is also possible with:

'/intent_name{{"entity_type_1":"entity_value_1", "entity_type_2": "entity_value_2"}}'

overwrite nlu with buttonsYou can use buttons to overwrite the NLU prediction and trigger a specific intent and entities.Messages starting with / are sent handled by the RegexInterpreter, which expects NLU input in a shortened /intent{entities} format. In the example above, if the user clicks a button, the user input will be classified as either the mood_great or mood_sad intent.You can include entities with the intent to be passed to the RegexInterpreter using the following format:/inform{"ORG":"Rasa", "GPE":"Germany"}The RegexInterpreter will classify the message above with the intent inform and extract the entities Rasa and Germany which are of type ORG and GPE respectively.

escaping curly braces in domain.ymlYou need to write the /intent{entities} shorthand response with double curly braces in domain.yml so that the assistant does not treat it as a variable in a response and interpolate the content within the curly braces.

Issuing Set Slot Commands

New in 3.9.0Starting from Rasa Pro 3.9.0, you can use buttons to issue commands to set slots.

Payload Syntax

To issue set slot commands, you can use the following format in the payload: /SetSlots(slot_name=slot_value). You can define multiple slot key-value pairs in the same command. The slot names that you define in the payload should be slots that are requested via the active flow or flows that the command generator predicts a StartFlow command for in the same turn. Note that there is a limit of 10 slot key-value pairs per command to prevent Regular expression Denial of Service (ReDoS) attacks. Here is an example:

domain.yml

responses:
  utter_contactless_limit:
  - text: "Which card would you like to set the maximum contactless limit for?"
    buttons:
    - title: "credit"
      payload: "/SetSlots(amount=100, card_type=credit)"
    - title: "debit"
      payload: "/SetSlots(amount=100, card_type=debit)"

Note that the SetSlots command is case-sensitive and should be written exactly as shown above. The regular expression used for extracting slot names and values from the payload does not allow the following characters:

=, ,, (, ) in the slot name
,, (, ) in the slot value

You can also use this syntax to start a flow by first setting a slot and then branching on that slot to execute a link or a call step.

All slot types are supported to be filled by buttons except for list slots.

Dynamic Buttons

You can also create a dynamic list of buttons in a reply via a custom action. Maybe the list of responses come from an API or the list of buttons is determined based on the value of another slot or the state of the conversation. This can be done via a collect step and a custom action called action_ask_{slot_name}. For example, let’s say your bot needs to ask the user which of their credit cards they want help with. We would create a response without the buttons and then use a custom action to get the list of cards associated with the user. There is also a cards slot with a list of all the users cards. This was loaded when the user first connected to the bot via action_session_start. There are also slots with the current card name and number.

domain.yml

slots:
  current_card_name:
    type: text
  current_card_number:
    type: text
  cards:
    type: list

responses:
  utter_select_card:
    - text: "Here are the your cards, select the one you are referring to?"

The select_card flow does a collect: current_card_name to request the current card from the user.

flow.yml

flows:
  select_card:
    description: This flow is called when the user has multiple cards and needs to select one.
    name: Select card
    # block this flow from th list of possible flows for the LLM, it should only be called from other flows
    if: False
    steps:
      - collect: current_card_name
        ask_before_filling: true
        next: END

Create a custom action named action_ask_current_card_name which the flow collect will call.

actions.py

class ActionShowSlots(Action):
    def name(self):
        return "action_ask_current_card_name"

    def run(self, dispatcher, tracker, domain):
        events = []
        cards = tracker.get_slot("cards")
        if not cards:
            dispatcher.utter_message(text="No cards found.")
        else:
            buttons = []
            for card in cards:
                buttons.append(
                    {
                        "title": card.get("name"),
                        "payload": f"/SetSlots(current_card_name={card.get('name')}, current_card_number={card.get('number')})"
                    }
                )
            dispatcher.utter_message(response="utter_select_card", buttons=buttons)
        return events

You can read more about the action_ask_{slot_name} here

Images

You can add images to a response by providing a URL to the image under the image key:

domain.yml

  utter_cheer_up:
  - text: "Here is something to cheer you up:"
    image: "https://i.imgur.com/nGF1K8f.jpg"

Custom Output Payloads

You can send any arbitrary output to the output channel using the custom key. The output channel receives the object stored under the custom key as a JSON payload. Here’s an example of how to send a date picker to the Slack Output Channel:

domain.yml

responses:
  utter_take_bet:
  - custom:
      blocks:
      - type: section
        text:
          text: "Make a bet on when the world will end:"
          type: mrkdwn
        accessory:
          type: datepicker
          initial_date: '2019-05-21'
          placeholder:
            type: plain_text
            text: Select a date

Voice-Specific Response Properties

For voice assistants, you can control specific behaviors of individual responses using voice-specific properties.

Controlling Interruptions

Beta FeatureInterruption handling is currently in beta and available for selected Voice Stream Channels only.

You can control whether users can interrupt the assistant while a specific response is being spoken using the allow_interruptions property:

domain.yml

responses:
  utter_current_balance:
    - text: You still have {current_balance} in your account.
      allow_interruptions: false
  
  utter_important_terms:
    - text: Please note the following terms and conditions that apply to your account...
      allow_interruptions: false

  utter_ask_preference:
    - text: What would you like to do today?
      allow_interruptions: true  # This is the default

By default, allow_interruptions is true for all responses. Setting it to false ensures that users must hear the complete message before the assistant will respond to their input. When to use allow_interruptions: false:

Critical information that users must hear completely (account balances, terms and conditions, emergency information)
Legal disclaimers or compliance-related content
Important instructions that shouldn’t be missed

👉 Learn more about interruption handling

Using Responses in Conversations

Calling Responses as Actions

If the name of the response starts with utter_, the response can directly be used as an action, without being listed in the actions section of your domain. You would add the response to the domain:

domain.yml

responses:
  utter_greet:
  - text: "Hey! How are you?"

You can use that same response as an action in your Flows:

flows.yml

flows:
  Greet:
    name: Greet
    description: This flow is called to greet customers at the start of the conversation.
    steps:
      - action: utter_Greet

When the utter_greet action runs, it will send the message from the response back to the user.

Changing responsesIf you want to change the text, or any other part of the response, you need to retrain the assistant before these changes will be picked up.

Calling Responses from Custom Actions

You can use the responses to generate response messages from your custom actions. If you’re using Rasa SDK as your action server, you can use the dispatcher to generate the response message, for example:

actions.py

from rasa_sdk.interfaces import Action

class ActionGreet(Action):
    def name(self):
        return 'action_greet'

    def run(self, dispatcher, tracker, domain):
        dispatcher.utter_message(template="utter_greet")
        return []

If you use a different custom action server, your server should return the following JSON to call the utter_greet response:

{
  "events": [],
  "responses": [
    {
      "template": "utter_greet"
    }
  ]
}

​Defining Responses

​Using Variables in Responses

​Response Variations

​IDs for Responses

​Channel-Specific Response Variations

​Conditional Response Variations

​Predicate Conditions

​Name and Value Equality Constraints

​Variation Selection

​Rich Responses

​Buttons

​Triggering Intents or Passing Entities

​Issuing Set Slot Commands

Payload Syntax

​Dynamic Buttons

​Images

​Custom Output Payloads

​Voice-Specific Response Properties

​Controlling Interruptions

​Using Responses in Conversations

​Calling Responses as Actions

​Calling Responses from Custom Actions

Defining Responses

Using Variables in Responses

Response Variations

IDs for Responses

Channel-Specific Response Variations

Conditional Response Variations

Predicate Conditions

`Name` and `Value` Equality Constraints

Variation Selection

Rich Responses

Buttons

Triggering Intents or Passing Entities

Issuing Set Slot Commands

Dynamic Buttons

Images

Custom Output Payloads

Voice-Specific Response Properties

Controlling Interruptions

Using Responses in Conversations

Calling Responses as Actions

Calling Responses from Custom Actions