Basic Forms in Rasa Open Source 2.x

Video#

There are many use-cases to collect information from a user before taking an action on their behalf. If we'd like to, for example, order a pizza then we'd need to know the kind and the size of the pizza. To make the collection of information easy, Rasa has a "Forms"-feature which we dive into in this segment.

How does a Rasa-Form work?#

Conceptually, a form can be seen as a loop that keeps on asking the user for information until enough slot values have been filled. An example of such a flow is shown in the diagram below.

The idea behind a Rasa form is that we don't need to ask a question if the user has already given us the required information. If the user tells us they want to buy a "veggie pizza", then we don't need to ask about the kind of pizza they're interested in anymore.

Code#

Github#

We will share code snippets below. But we also have a Github repository with the entire Rasa project. If you'd prefer to explore the code that way, simply go to the repository here.

That said, to set up a basic form, you'll need to change your domain.yml file to configure the form and slots but you'll also need to update your rules.yml file so that Rasa understands when you trigger a form loop.

Example domain.yml#

The example below demonstrates how to set up a simple for with two slots to fill.

# This is the form definition.
# Note that we refer to slots defined below.
forms:
  simple_pizza_form:
    required_slots:
      pizza_size:
      - entity: pizza_size
        type: from_entity
      pizza_type:
      - entity: pizza_type
        type: from_entity

# These are slot definitions.
# Note that we also have entities with the same name.
slots:
  pizza_size:
    type: text
    influence_conversation: true
  pizza_type:
    type: text
    influence_conversation: true

# These are simple entity definitions
entities:
- pizza_size
- pizza_type

# We also need to add relevant intents 
intents:
- buy_pizza
- inform

# We also need to add responses for our form.
responses:
  utter_submit:
  - text: I will now order a pizza for you!
  utter_pizza_slots:
  - text: I will order a {pizza_size} {pizza_type} pizza.
  utter_ask_pizza_size:
  - text: What size would you like your pizza to be?
  utter_ask_pizza_type:
  - text: What kind of pizza would you like to buy?

You may notice that there are a few responses that follow the utter_ask_<slot_name> naming convention. These reponses are picked up by our form whenever we'd like the user to give us the information required to fill in a slot name.

Example rules.yml#

Given that we have a form defined, we also need Rasa to be able to trigger the loop at the appropriate time in the conversation. You can use rules to define this.

- rule: Activate Pizza Form
  steps:
  - intent: buy_pizza
  - action: simple_pizza_form
  - active_loop: simple_pizza_form

- rule: Submit Pizza Form
  # Ensure that we're starting from an active loop
  condition:
  - active_loop: simple_pizza_form
  steps:
  - action: simple_pizza_form
  # If there are no more requested slots
  - active_loop: null
  - slot_was_set:
    - requested_slot: null
  # Then submit the form
  - action: utter_submit
  - action: utter_pizza_slots

Don't forget nlu.yml#

Since we're dealing with entities, we shouldn't forget to add a few examples to our nlu.yml file. We need examples of the entities but also for the intents that we're using in our rules.yml file.

- intent: buy_pizza
  examples: |
    - i'd like to buy a pizza
    - i want a pizza
    - can i buy a pizza
    - I'm interested in a savory round flattened bread of Italian origin
    - i want to buy a pizza
- intent: inform
  examples: |
    - i'd like a [large](pizza_size) pizza
    - i want to order a [xl](pizza_size) [hawai](pizza_type) pizza
    - [medium](pizza_size) pizza
    - [xl](pizza_size)
    - [small](pizza_size)
    - [s](pizza_size)
    - [pepperoni](pizza_type)

Invalid Data#

It might be the case that the user gives us information that isn't valid. A user might ask us to make a "fruit" pizza, which is something that we cannot provide.

In these situations we'd like to validate and catch this invalid utterances so that we can ask the user to try again.

To validate data, we can use Custom Actions.

Adding a Custom Action#

If we want to add a custom action, we shouldn't forget to add it to our domain.yml file.

actions:
  - validate_simple_pizza_form

Note the naming convention here. We need to adhere to the validate_<form_name> convention.

Python Code#

An example validator for our form can be seen below.

from typing import Text, List, Any, Dict

from rasa_sdk import Tracker, FormValidationAction, Action
from rasa_sdk.events import EventType
from rasa_sdk.executor import CollectingDispatcher
from rasa_sdk.types import DomainDict

ALLOWED_PIZZA_SIZES = [
    "small", "medium", "large", "extra-large", "extra large", 
    "s", "m", "l", "xl"
]
ALLOWED_PIZZA_TYPES = [
    "mozzarella", "fungi", "veggie", "pepperoni", "hawaii"
]

class ValidateSimplePizzaForm(FormValidationAction):
    def name(self) -> Text:
        return "validate_simple_pizza_form"

    def validate_pizza_size(
        self,
        slot_value: Any,
        dispatcher: CollectingDispatcher,
        tracker: Tracker,
        domain: DomainDict,
    ) -> Dict[Text, Any]:
        """Validate `pizza_size` value."""

        if slot_value.lower() not in ALLOWED_PIZZA_SIZES:
            dispatcher.utter_message(text=f"We only accept pizza sizes: s/m/l/xl.")
            return {"pizza_size": None}
        dispatcher.utter_message(text=f"OK! You want to have a {slot_value} pizza.")
        return {"pizza_size": slot_value}

    def validate_pizza_type(
        self,
        slot_value: Any,
        dispatcher: CollectingDispatcher,
        tracker: Tracker,
        domain: DomainDict,
    ) -> Dict[Text, Any]:
        """Validate `pizza_type` value."""

        if slot_value not in ALLOWED_PIZZA_TYPES:
            msg = "I don't recognize that pizza."
            msg += f"We only serve {'/'.join(ALLOWED_PIZZA_TYPES)}."
            dispatcher.utter_message(msg)
            return {"pizza_type": None}
        dispatcher.utter_message(text=f"OK! You want to have a {slot_value} pizza.")
        return {"pizza_type": slot_value}

There's a few things to note.

• Note that our ValidateSimplePizzaForm class inherits from FormValidationAction, which is different from the standard Action class that custom actions inherit from.
• The name method defines the name of the validator, which needs to correspond with the name in our domain.yml file and needs to adhere to the validate_<form_name> naming convention.
• For each slot we have a validate_<slot_name> method. Again, we need to follow a naming convention.
• Note that each validator method returns a dictionary. You can set a slot to None which will invalidate it.

Exploring Forms#

If you want to explore your forms, we recommend using rasa interactive. This gives you an interactive shell that also shows the status of the slots, which is very useful when you're debugging a form.

Links#

• Rasa Forms Documentation

Exercises#

Try to answer the following questions to test your knowledge.

• In what ways is a form validator different from a custom action?
• When would you use a form to set slots? When might a custom action be preferable?