Add KServe inference example for Kubeflow ModelKit

gorkem · gorkem · commit d2300ffe65e2 · 2025-12-02T10:01:22.000-05:00
Includes InferenceService manifest demonstrating kit:// URI
integration with KServe for deploying ModelKits.

Signed-off-by: Gorkem Ercan &lt;gorkem.ercan@gmail.com&gt;
diff --git a/build/dockerfiles/kubeflow-components/examples/house-prices-inference-service.yaml b/build/dockerfiles/kubeflow-components/examples/house-prices-inference-service.yaml
@@ -0,0 +1,14 @@
+apiVersion: serving.kserve.io/v1beta1
+kind: InferenceService
+metadata:
+  name: house-prices-model
+  namespace: modelkit-demo
+  annotations:
+    serving.kserve.io/deploymentMode: RawDeployment
+spec:
+  predictor:
+    model:
+      modelFormat:
+        name: sklearn
+      storageUri: kit://jozu.ml/demo/house-prices:latest
+      protocolVersion: v2
diff --git a/build/dockerfiles/kubeflow-components/examples/house-prices-pipeline.py b/build/dockerfiles/kubeflow-components/examples/house-prices-pipeline.py
@@ -5,7 +5,7 @@
 creates a directory with well-named files (model.pkl, predictions.csv, train.py, README.md)
 and the push-modelkit component packages the entire directory as a ModelKit.
 
-Uses pure KFP v2.14.3 components without v1 compatibility.
+Uses KFP v2.14.3
 """
 
 from kfp import dsl, kubernetes
