Fork

This fork was originally only for packaging purposes. Since then, we've added a number of new features including Dependency Injection via Guice and versions of models.

Getting started

Build the project using [Apache Maven] (http://maven.apache.org/):

mvn clean install

Change the working directory to server and start the embedded Jetty web server:

mvn jetty:run

Congratulations! You now have an openscoring web service running at [localhost:8080] (http://localhost:8080/openscoring/)

REST API

PUT - Deploy a model

Deploy the contents of the PMML file DecisionTreeIris.pmml as a model DecisionTreeIris:

curl -X PUT --data-binary @DecisionTreeIris.pmml -H "Content-type: text/xml" http://localhost:8080/openscoring/model/DecisionTreeIris

For a list of sample PMML files please take a look at [JPMML R/Rattle support module] (https://github.com/jpmml/jpmml/tree/master/pmml-rattle/src/test/resources/pmml) or [JPMML KNIME support module] (https://github.com/jpmml/jpmml/tree/master/pmml-knime/src/test/resources/pmml).

GET - Obtain model information

Get the list of deployed models

Obtain the list of deployed models:

curl -X GET http://localhost:8080/openscoring/model

The response body is the JSON serialized form of a list of model identifiers:

[
	"DecisionTreeIris"
]

Get the description of a deployed model

Obtain the description of the public interface of the model DecisionTreeIris:

curl -X GET http://localhost:8080/openscoring/model/DecisionTreeIris

The response body is the JSON serialized form of an org.openscoring.common.SummaryResponse object:

{
	"activeFields" : ["Sepal.Length", "Sepal.Width", "Petal.Length", "Petal.Width"],
	"predictedFields" : ["Species"],
	"outputFields" : ["Predicted_Species", "Probability_setosa", "Probability_versicolor", "Probability_virginica"]
}

Field definitions are retrieved from the [Mining Schema element] (http://www.dmg.org/v4-1/MiningSchema.html) of the PMML document. The execution fails with HTTP status code "500 Internal Server Error" if the PMML document contains an unsupported model type.

POST - Perform model evaluation

The evaluation can be performed either in single prediction mode or in batch prediction mode (see below). On average, the batch prediction mode is expected to provide better throughput.

Single prediction mode

Send the contents of the JSON file EvaluationRequest.json for evaluation to the model DecisionTreeIris:

curl -X POST --data-binary @EvaluationRequest.json -H "Content-type: application/json" http://localhost:8080/openscoring/model/DecisionTreeIris

The request body is the JSON serialized form of an org.openscoring.common.EvaluationRequest object:

{
	"parameters" : {
		"Sepal.Length" : 5.1,
		"Sepal.Width" : 3.5,
		"Petal.Length" : 1.4,
		"Petal.Width" : 0.2
	}
}

The response body is the JSON serialized form of an org.openscoring.common.EvaluationResponse object:

{
	"result" :
		{
			"Species" : "setosa",
			"Predicted_Species" : "setosa",
			"Probability_setosa" : 1.0,
			"Probability_versicolor" : 0.0,
			"Probability_virginica" : 0.0
		}
}

Batch prediction mode

Send the contents of the JSON file BatchEvaluationRequest.json for evaluation to the model DecisionTreeIris (please note /batch at the end of the URL):

curl -X POST --data-binary @BatchEvaluationRequest.json -H "Content-type: application/json" http://localhost:8080/openscoring/model/DecisionTreeIris/batch

The request body is the JSON serialized form of a list of org.openscoring.common.EvaluationRequest objects. The number of list elements is not restricted.

The response body is the JSON serialized form of a list of org.openscoring.common.EvaluationResponse objects.

DELETE - Undeploy a model

Undeploy the model DecisionTreeIris:

curl -X DELETE http://localhost:8080/openscoring/model/DecisionTreeIris