Iterating forms in Rasa Open Source

Forms in Rasa Open Source are a great tool for a chatbot to gather a set of required information. While developing a chatbot myself, I needed to gather the same set of information N times, where N is chosen by the user. The best way of doing that would be to activate a form N consecutive times. I faced some problems trying to do that, so I hope this repo will save you some time.

The repo provides a complete, though minimal example on how to use Rules and Custom Actions to be able to iterate over a form N times.

The example resulting chatbot asks for the number N of iterations, and subsequently starts N times the form my_form consisting of a single required slot lorem-ipsum. At the end of every iteration, the filled slots are saved into the iterations slot for subsequent use.

Consider for instance you have a chatbot to book spots for a conference. The user can book up to 5 spots and the chatbot need to gather name and age of all the participants. In this case, my_form would have name and age required slots.

Important

Refer to my post on Rasa forum for an issue using an earlier version of Rasa.

How to use it

Using virtual environment

1. Please follow the official Rasa docs to create a virtual environment and install Rasa
2. Let the action_endpoint in app/endpoints.yml point to localhost:5055:

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

3. cd app
4. Train the model: rasa train --domain domain
5. Run the action server: rasa run actions
6. In a new shell in the app folder, activate the environment and run Rasa shell: rasa shell and interact with the chatbot

Using docker compose

As of February 2023, the Rasa Docker image is not compatible with M1/M2 powered Mac.

1. Let the action_endpoint in app/endpoints.yml point to the rasa_action_server service:

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

2. cd docker
3. Build the images: docker compose build
4. Train the model: docker compose run --rm rasa_core train --domain domain
5. Run the services: docker compose up -d (for simple shell testing as per the following point, it sufficies to run the action_server only)
6. Run Rasa shell docker compose run --rm rasa_core shell and interact with the chatbot

The docker/docker-compose.yml comes with the possibility of setting up a custom tracker store using postgreSQL, but it is commented by default. If you want to unlock it, you must also uncomment the tracker_store related lines in app/endpoints.yml:

tracker_store:
    type: SQL
    dialect: "postgresql" 
    url: "rasa_tracker_store" 
    db: "rasa" 
    username: "rasa"
    password: "rasa"