Added Kubeflow Pipeline component for ModelKit packaging and OCI push

build/kfp-modelkit-component/Dockerfile

@@ -0,0 +1,19 @@
+# Dockerfile for Kubeflow ModelKit Component
+FROM python:3.10-slim
+
+# Install system dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    git \
+    && rm -rf /var/lib/apt/lists/*
+
+# Install ModelKit CLI (replace with actual install if available via pip or other means)
+# Example: pip install modelkit-cli
+# If ModelKit is only available via binary, add COPY or curl/wget here
+# For now, placeholder:
+RUN pip install --no-cache-dir ml-metadata kfp
+# TODO: Add ModelKit CLI installation here
+
+# Copy component code
+COPY component.py /component.py
+
+ENTRYPOINT ["python", "/component.py"]

build/kfp-modelkit-component/component.py

@@ -0,0 +1,69 @@
+import os
+import subprocess
+import json
+from typing import Optional
+from ml_metadata import metadata_store
+from ml_metadata.proto import metadata_store_pb2
+
+# This script is intended to be used as a Kubeflow Pipeline component entrypoint.
+# It expects the following environment variables or arguments:
+#   MODEL_DIR: Path to model artifacts
+#   REGISTRY_URI: OCI registry URI (e.g., oci://my-registry/modelkit:tag)
+#   PIPELINE_RUN_ID: Kubeflow pipeline run ID
+#   EXPERIMENT_NAME: Kubeflow experiment name
+#   KUBEFLOW_METADATA_HOST: MLMD gRPC host (default: metadata-grpc-service.kubeflow)
+#   DOCKER_CONFIG_PATH: Path to Docker config for registry auth (default: /kaniko/.docker)
+
+def main():
+    model_dir = os.environ.get('MODEL_DIR')
+    registry_uri = os.environ.get('REGISTRY_URI')
+    pipeline_run_id = os.environ.get('PIPELINE_RUN_ID')
+    experiment_name = os.environ.get('EXPERIMENT_NAME')
+    kubeflow_metadata_host = os.environ.get('KUBEFLOW_METADATA_HOST', 'metadata-grpc-service.kubeflow')
+    docker_config_path = os.environ.get('DOCKER_CONFIG_PATH', '/kaniko/.docker')
+
+    if not model_dir or not registry_uri or not pipeline_run_id or not experiment_name:
+        raise ValueError("Missing required environment variables.")
+
+    # 1. Connect to ML Metadata
+    store = metadata_store.MetadataStore(
+        metadata_store_pb2.ConnectionConfig(
+            host=kubeflow_metadata_host,
+            port=8080,
+        )
+    )
+    # 2. Extract run/experiment metadata
+    runs = store.get_executions_by_type('kfp-run')
+    run = next((r for r in runs if r.custom_properties.get('run_id', None) and r.custom_properties['run_id'].string_value == pipeline_run_id), None)
+    metrics = run.custom_properties.get('metrics', {}).string_value if run and 'metrics' in run.custom_properties else '{}'
+
+    # 3. Prepare ModelKit metadata
+    metadata = {
+        "pipeline_run_id": pipeline_run_id,
+        "experiment_name": experiment_name,
+        "metrics": metrics,
+    }
+    with open('modelkit-metadata.json', 'w') as f:
+        json.dump(metadata, f)
+
+    # 4. Package model as ModelKit (assumes modelkit CLI is installed)
+    subprocess.run([
+        'modelkit', 'pack',
+        '--input', model_dir,
+        '--metadata', 'modelkit-metadata.json',
+        '--output', 'modelkit.tar'
+    ], check=True)
+
+    # 5. Push to OCI registry
+    env = os.environ.copy()
+    env['DOCKER_CONFIG'] = docker_config_path
+    subprocess.run([
+        'modelkit', 'push',
+        '--input', 'modelkit.tar',
+        '--destination', registry_uri
+    ], check=True, env=env)
+
+    print(f"ModelKit pushed to {registry_uri}")
+
+if __name__ == "__main__":
+    main()

build/kfp-modelkit-component/component.yaml

@@ -0,0 +1,11 @@
+
+        value: {inputValue: registry_uri}
+      - name: PIPELINE_RUN_ID
+        value: {inputValue: pipeline_run_id}
+      - name: EXPERIMENT_NAME
+        value: {inputValue: experiment_name}
+      - name: KUBEFLOW_METADATA_HOST
+        value: {inputValue: kubeflow_metadata_host}
+      - name: DOCKER_CONFIG_PATH
+        value: {inputValue: docker_config_path}
+    # Add volume mounts for secrets if needed