actions

Custom Rasa Action to fetch birthdata from a Qanary system

Please recognize the big picture of this project.

Table of Contents

• Purpose
• Test the Rasa Action
  • Test the Rasa Action directly
  • Test the Rasa Action in a Docker container
    • Preparation: Build Docker image
    • Execute test
      • Manually set environment variable inside of the Docker container
      • Automatically set environment variable inside of the Docker container to the host machine
    • Execute the unit test
      • Run test of action in Docker container as a single command

Purpose

In this folder a custom Rasa action is implemented which fetches birthdata from a Qanary system. The custom Rasa action is implemented in the file actions.py. It can be tested in the file actions_test.py (see below).

For computing the results the following Qanary components are used (as defined here):

1. AutomationServiceComponent

2. BirthDataQueryBuilderWikidata

3. SparqlExecuterComponent

The custom Rasa action is implemented as a class BirthDataAction which inherits from the class Action of the Rasa framework. The action returns HTML code that should contain the birthdata of the recognized person(s) visualized in a table.

Test the Rasa Action

Test the Rasa Action directly

The test file is a standard Python unittest. Hence, you can start it directly from the command line.

python -m unittest -v actions_test.py

⚠️	Make sure that your Qanary system is running and the environment variable QANARY_IP is set correctly and the Qanary system has started completely.

Test the Rasa Action in a Docker container

This document describes how to test a Rasa action using a Docker container. Following this approach fast testing of your custom action is possible without the need to (re-)start the Rasa server. Unit tests are executed in a Docker container and the Rasa action is executed in a Docker container. All tests are stored in the file actions_test.py.

Preparation: Build Docker image

docker build -f Dockerfile_action_test -t "rasa-action-test:latest" .

⚠️	This might take some time while building the image for the first time, because the Docker image is build from scratch where installing Rasa and Tensorflow might take some time.

Execute test

Now, the Docker container can be started and the test can be executed. To map the current directory to the Docker container the option --mount is used. Here, the current directory is mapped to the directory /actions in the Docker container.

docker run --rm -it --mount src=$(pwd),target=/actions,type=bind rasa-action-test:latest bash

⚠️	From inside the Docker container external network connections are not possible directly.

The environment variable QANARY_IP needs to be set to the IP address of the machine where the corresponding Qanary system is running, s.t., the Rasa action can fetch the expected data.

Manually set environment variable inside of the Docker container

You can always set manually the environment variable QANARY_IP inside of the Docker container.

export QANARY_IP=...

Automatically set environment variable inside of the Docker container to the host machine

If you want to set the environment variable QANARY_IP automatically to the IP address of the host machine, you can use the following command. This is very useful if you are running the Qanary system on the same machine as the test.

export QANARY_IP=$(ip route show | awk '/default/ {print $3}')

Execute the unit test

⚠️	Make sure that your Qanary system is running and the environment variable QANARY_IP is set correctly and the Qanary system has started completely.

python -m unittest -v actions_test.py

Run test of action in Docker container as a single command

The following command executes the test in a Docker container and sets the environment variable QANARY_IP to the IP address of the host machine (i.e., this only works if you are running the Qanary system on the same machine).

docker run --rm -it --mount src=$(pwd),target=/actions,type=bind rasa-action-test:latest bash -c "export QANARY_IP=\$(ip route show | awk '/default/ {print \$3}') ; python -m unittest -v actions_test.py"