COVID-19 FAQ Chatbot with Rasa, Docker and Heroku

This is undoubtedly a unique moment in the history of the world we know. Most of us have never experienced something even close. Travelling, socialising, just do normal things were never questioned before.

As today (March 29th) we hardly see the end of this, but commitment, hope and dreams make us endure.

This article presents how to create a FAQ chatbot powered by Rasa, built with Docker and deployed on Heroku.

Rasa is an open source machine learning Python framework to build powerful chatbots which can be deployed on various channels (a Web site, Facebook, Telegram, many more supported).

First download and install Rasa, configure your Python virtual environment and init your project

mkdir covid-19-chatbot
rasa init

First important step is to instruct the ‘MemoizationPolicy’ (which memorises the conversations as they take place) to be forgetful (in config.yml) and only considers the last user message:

- name: MemoizationPolicy
max_history: 1
- name: MappingPolicy

From them on it is pretty straight forward: follow Rasa doc to define stories and intents, you only need to pay attention to the “Response Selector” component.

The ResponseSelector NLU component is designed to make it easier to handle dialogue elements like Small Talk and FAQ messages in a simple manner. By using the ResponseSelector, you only need one story to handle all FAQs, instead of adding new stories every time you want to increase your bot’s scope.

Each intent is associated to a question (which you should provide in different forms to take advantage of the NLP capabilities). Intent names have the ‘faq’ prefix.

## intent: faq/what_is
- what is Corona virus?
- define Corona virus
- what is COVID-19?
- define COVID-19

Link the intent to the Chatbot response still using the ‘faq’ prefix which allows the ResponseSelector to capture and process them.

## what is 
* faq/what_is
- Coronavirus (COVID-19) is a new coronavirus that has not been previously identified. ‘CO’ stands for ‘corona,’ ‘VI’ for ‘virus,’ and ‘D’ for disease.

The very important point here is you only create a single story for all questions (see

## Some question from FAQ
* faq
- respond_faq

Train the model (‘rasa train’) and test it (‘rasa shell’)… obviously add more Q&A to make it more interesting (and more useful).

Dockerizing the Chatbot

Lets create a Docker image to run the Rasa Chatbot (and later deploy it)

# from rasa base image
FROM rasa/rasa:1.8.0
# copy all source and the Rasa generated model
COPY . /app
# inform which port will run on
# script to run rasa core
COPY /app/scripts/
# script to run rasa shell
COPY /app/scripts/
USER root
RUN chmod a+x /app/scripts/
RUN chmod a+x /app/scripts/
ENV shell_mode false
# launch script (rasa shell or rasa run)
CMD sh -c 'if [ "$shell_mode" = false ]; then /app/scripts/; else /app/scripts/; fi'

Now take a look at ‘’: the script conveniently defines the rasa execution and required parameters

rasa run -p $PORT --cors "*" --enable-api --debug

A very interesting thing happens in the startup script:

  • define the PORT where Rasa will run: here we inject the env variables $PORT. This is extremely important when deploying on Heroku as the port is dynamic and assigned at deployment time

Build the image and run it first locally to verify you can talk to it

docker build -t perosa/rasa-covid .docker run -p 5005:5005 -e PORT=5005 -e shell_mode=true -it --rm perosa/rasa-covid

Heroku time

It is now time to push and run it on Heroku: with a Dockerized application in place the obvious choice is the Heroku Container Registry.

First create the Heroku application (using the Dashboard or Heroku CLI), then make sure to logon into the registry

heroku container:login

Tag the image and push it

# tag the image
docker tag perosa/rasa-covid
# push the image
docker push
# deploy
heroku container:release web -a rasa-covid

You can follow the deployment logs on Heroku, if it is all going well you should be able to see “State changed from starting to up”

Ping it with https://{app name} and get a response from your Rasa chatbot :-)

Talk to the Chatbot

It is not fun if you cannot talk to your Chatbot, right?

The quickest and easiest way is to leverage on the REST API: send a POST with a message

curl -d ‘{“sender”: “Beppe”, “message”: “Hi”}’ -H “Content-Type: application/json” -X POST[{“recipient_id”:”Beppe”,”text”:”Hello”},{“recipient_id”:”Beppe”,”text”:”I can answer questions about the Corona virus.”}]

Explore the Rasa channels to choose what is the best home for your Bot and engage your audience.

Hope this was useful (source code is here), ping me for questions and stay safe 🙏

Software Engineering, Data Engineering, Coder. And other hobbies too.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store