Rasa and Docker Containers

Let's discuss how you can use Docker together with Rasa.


Code Used

Let's start a new Rasa project locally via;

python -m pip install rasa
rasa init 
rasa train

You can now run the assistant locally via;

rasa run --enable-api

This should allow you to call the trained model via a web API. You can point traffic to http://localhost:5000/webhook and you should be able to get responses. You can do this with a tool like postman or insomnia but you can also use curl from the command line.

curl --request POST \
  --url http://localhost:5005/model/parse \
  --header 'Content-Type: application/json' \
  --data '{"text": "hi there"}'


Let's see if we can build a Docker image for Rasa. That way, we might more easily share our code for deployment. To do this we'll need to write a Dockerfile first.

# use a python container as a starting point
FROM python:3.7-slim
# install dependencies of interest
RUN python -m pip install rasa
# set workdir and copy data files from disk
# note the latter command uses .dockerignore
COPY . .
# train a new rasa model
RUN rasa train nlu
# set the user to run, don't run as root
USER 1001
# set entrypoint for interactive shells
# command to run when container is called to run
CMD ["run", "--enable-api", "--port", "8080"]

To make sure that we don't copy files into the container that we won't need, we'll add a .dockerignore file.


This way, docker won't copy all the files in the container run the COPY . . command.

Building and Running

To build the image, we'll need to run the following command;

docker build -t koaning/rasa-demo .

Note that we're building the image with the tag koaning/rasa-demo. This is a convention that we'll use to identify the image later.

Once the image is built, we can run the container with the following command;

docker run -it -p 8080:8080 koaning/rasa-demo

This command does a few things.

  • The koaning/rasa-demo is a reference to the container that we built


  • The -it flag means that we want to run the container interactively.
  • The -p 8080:8080 means that we want to route the port 8080 from the

container to port 8080 on the host.

When the container spins up you should be able to once again communicate with the Rasa model via the web API.

curl --request POST \
  --url http://localhost:5005/model/parse \
  --header 'Content-Type: application/json' \
  --data '{"text": "hi there"}'

The difference now is that the model is running inside of a docker container. You can imagine that we're able to share this container and as long as you're able to run docker ... you're also able to run Rasa!


You can do even more with the container that we have right now. For example, you could run the rasa shell via our entrypoint.

docker run -it koaning/rasa-demo shell

Note that this command is equivalent to running rasa shell from the command line.


You can also push the container to a repository so it's available to other people.

docker push koaning/rasa-demo

Next steps.

In the next video we'll use the docker containers that Rasa provides on dockerhub instead of writing our own from scratch.


Try to answer the following questions to test your knowledge.

  1. What might go wrong if you don't have a .dockerignore file?
  2. How big might a docker image be with Rasa in it? What are the main contributing factors in the size?

2016-2024 © Rasa.