Kogito Serverless Workflow - Patient Onboarding

Description

In this example we will see how to use the Serverless Workflow specification implementation on Kogito to orchestrate a sequence of REST calls in an Event-Driven architecture.

This use case is about onboarding a new patient in a hospital, and based on their symptom, the application will assign to the right expert.

The workflow starts with a CloudEvent informing a new patient entry:

{
   "id":"12345",
   "source":"/hospital/entry",
   "specversion":"1.0",
   "type":"new.patient.events",
   "data":{
      "name":"John",
      "dateOfBirth":"2017-08-15",
      "symptoms":[
         "seizures"
      ]
   }
}

Or you can use cURL and set the CloudEvents parameters in the request headers:

$ curl -X POST \
      -H "content-type: application/json"  \
      -H "ce-specversion: 1.0"  \
      -H "ce-source: /hospital/entry"  \
      -H "ce-type: new.patients.events"  \
      -H "ce-id: 12346"  \
      -d "{ \"name\": \"Mick\", \"dateOfBirth\": \"1983-08-15\", \"symptoms\":[\"seizures\"]}" \
  http://localhost:8080

The workflow will then call a sequence of REST services:

1. StoreNewPatient will store the patient in the internal database for later processing
2. AssignDoctorToPatient will run a set of rules to determine to which doctor this patient should be assigned, based on their symptoms or general characteristics (children always go to Pediatrics, for example)
3. SchedulePatientAppointment will find a time slot for this patient based on the agenda of the assigned doctor

After executing the workflow, the appointments' data can be fetched using the /onboarding/schedule/appointment endpoint.

There's a nice web interface for you to try the example available in the root url (http://localhost:8080 for local environments):

Enter the data into the provided form to see the example in action! Date of Birth must be in YYYY-MM-DD format.

Also, there is the Swagger UI that can also be used to explore the APIs in this example: http://localhost:8080/swagger-ui/

Installing and Running

Prerequisites

You will need:

• Java 17+ installed
• Environment variable JAVA_HOME set accordingly
• Maven 3.9.6+ installed

When using native image compilation, you will also need:

• GraalVm 20.2.0+ installed
• Environment variable GRAALVM_HOME set accordingly
• Note that GraalVM native image compilation typically requires other packages (glibc-devel, zlib-devel and gcc) to be installed too. You also need ‘native-image’ installed in GraalVM (using ‘gu install native-image’). Please refer to GraalVM installation documentation for more details.

Compile and Run in Local Dev Mode

mvn clean package quarkus:dev

Use curl command to send the CloudEvent through HTTP to the application:

$ curl -X POST \
      -H "content-type: application/json"  \
      -H "ce-specversion: 1.0"  \
      -H "ce-source: /hospital/entry"  \
      -H "ce-type: new.patients.events"  \
      -H "ce-id: 12346"  \
      -d "{ \"name\": \"Mick\", \"dateOfBirth\": \"1983-08-15\", \"symptoms\":[\"seizures\"]}" \
  http://localhost:8080

In the application's log, you will see something like:

2020-11-17 12:51:11,581 DEBUG [org.acm.sw.onb.res.PatientResource] (executor-thread-198) Received patient to store in the internal in memory database: Patient{name='Mick', id='null', symptoms=[seizures], dateOfBirth=1983-08-15}
2020-11-17 12:51:11,581 DEBUG [org.acm.sw.onb.res.PatientResource] (executor-thread-198) Patient has been stored in the internal memory: Patient{name='Mick', id='5560ed3d-becb-4f35-8c6a-2816338bf8fd', symptoms=[seizures], dateOfBirth=1983-08-15}
2020-11-17 12:51:11,600 DEBUG [org.acm.sw.onb.res.AppointmentResource] (executor-thread-198) Receive patient to schedule appointments: Patient{name='Mick', id='5560ed3d-becb-4f35-8c6a-2816338bf8fd', symptoms=[seizures], dateOfBirth=1983-08-15}
2020-11-17 12:51:11,600 DEBUG [org.acm.sw.onb.res.AppointmentResource] (executor-thread-198) Processed patient: Patient{name='Mick', id='5560ed3d-becb-4f35-8c6a-2816338bf8fd', symptoms=[seizures], dateOfBirth=1983-08-15}

Then query the appointments' resource to see the assigned doctor:

$ curl http://localhost:8080/onboarding/schedule/appointment | jq '.'

[
  {
    "doctor": {
      "id": "8293dc94-2386-11eb-adc1-0242ac120002",
      "name": "Maria Mind",
      "specialty": "Neurology"
    },
    "patient": {
      "name": "Mick",
      "id": "5560ed3d-becb-4f35-8c6a-2816338bf8fd",
      "symptoms": [
        "seizures"
      ],
      "dateOfBirth": "1983-08-15",
      "assignedDoctor": {
        "id": "8293dc94-2386-11eb-adc1-0242ac120002",
        "name": "Maria Mind",
        "specialty": "Neurology"
      }
    },
    "date": "2020-11-18 09:00"
  }
]

Note: Please make sure you have jq installed in your system to see the formatted output.

Compile and Run in JVM mode

mvn clean package 
java -jar target/quarkus-app/quarkus-run.jar

or on Windows

mvn clean package
java -jar target\quarkus-app\quarkus-run.jar

Compile and Run using Local Native Image

Note that this requires GRAALVM_HOME to point to a valid GraalVM installation

mvn clean package -Pnative

To run the generated native executable, generated in target/, execute

./target/serverless-workflow-functions-quarkus-runner.jar

Deploying in Minikube

We have prepared a knative Maven profile to build the service image locally and all the Knative resources you need to get started.

1. Install Knative
2. Install the KogitoSource via command line.
3. Run eval $(minikube docker-env) to build the image directly into the Minikube registry.
4. Run mvn clean install -Pknative -Dnamespace=<your namespace> to build the image and the Knative resources for your application to run.
5. Apply the objects created for you with kubectl apply -f target/kubernetes/*.yml. It will deploy the objects from knative.yml and kogito.yml generated files.
6. Run curl from the terminal like you did in the previously steps.

Credits

• Disease Prediction through Symptoms dataset from Kagle
• Tihomir Surdilovic for the nice UI