This example contains a simple workflow service that illustrate consecutive invocation of two knative python function calls. The first function initialize a tensor of the specified size with random content using torch and return its contents. The second performs an addition of all the element returned by the previous step using numpy library.

You will need:

• Java 11+ installed
• Maven 3.8.6+ installed
• Docker
• Minikube
• Knative CLI

Open a terminal, go to test directory and type

kn func build

Once it finished, make sure minikube is running with knative profile activated (if you have setup it before, you just need to start minikube with minikube start -p profile and configure the tunnel running minikube tunnel --profile knative)

Once minikube is running, to load the image from your local docker into minikube registry type in a terminal

minikube image load dev.local/test -p knative

Now run that image as a knative service called test

kn service create test --image=dev.local/test --pull-policy=IfNotPresent

Open a terminal, go to test directory and type

kn func build

Once it finished, make sure minikube is running with knative profile activated (if you have setup it before, you just need to start minikube with minikube start -p profile and configure the tunnel running minikube tunnel --profile knative)

Once minikube is running, to load the image from your local docker into minikube registry type in a terminal

minikube image load dev.local/receiver -p knative

Now run that image as a knative service called receiver

kn service create receiver --image=dev.local/receiver --pull-policy=IfNotPresent

Open a terminal, go to workflow directory and run

mvn clean package

In some terminals, you need to ensure the local image is loaded into minikube by running

minikube image load dev.local/serverless-workflow-knative-python-quarkus:1.0-SNAPSHOT -p knative

and update the service accordingly

kn service update serverless-workflow-knative-python-quarkus --image=dev.local/serverless-workflow-knative-python-quarkus:1.0-SNAPSHOT --pull-policy=IfNotPresent

Once done, your workflow service should be available in knative, you need to find out the uri

 kn service list | grep serverless-workflow-knative-python-quarkus

The URI of the service will be the one in the second column

To invoke the flow, you need to execute the following REST invocation, replacing the uri by the one resolved in the previous step and specifying the x and y dimension of the tensor.

curl -X POST -H 'Content-Type:application/json' -H 'Accept:application/json' -d '{"x":3,"y":3}' <uri>/TensorTest

The result is a float number with the sum of the randomly generated matrix.

{"id":"e80c8f2f-3753-45f0-b477-15812a3fe982","workflowdata":6.1767255663871765}