Refactor Kubeflow components to use reference instead of uri

Update component inputs, outputs, and scripts to use reference
instead of uri for OCI alignment. Also remove redundant CI triggers and dependencies.

Signed-off-by: Gorkem Ercan <gorkem.ercan@gmail.com>

Author: gorkem

.github/workflows/kubeflow-components-test.yaml

@@ -5,12 +5,6 @@ on:
     paths:
       - 'build/dockerfiles/kubeflow-components/**'
       - '.github/workflows/kubeflow-components-test.yaml'
-  push:
-    branches:
-      - main
-    paths:
-      - 'build/dockerfiles/kubeflow-components/**'
-      - '.github/workflows/kubeflow-components-test.yaml'
 
 jobs:
   test-scripts:
@@ -25,10 +19,6 @@ jobs:
           sudo apt-get update
           sudo apt-get install -y bats
 
-      - name: Install jq
-        run: |
-          sudo apt-get install -y jq
-
       - name: Run push-modelkit tests
         working-directory: build/dockerfiles/kubeflow-components
         run: bats tests/push-modelkit.bats

build/dockerfiles/kubeflow-components/README.md

@@ -31,16 +31,16 @@ If a `Kitfile` exists in `modelkit_dir`, it is used as-is. Otherwise, one is aut
 
 **Outputs**
 
-- `uri` – Tagged ModelKit URI (e.g., `registry.io/myorg/mymodel:v1`)
-- `digest` – Digest-based ModelKit URI (e.g., `registry.io/myorg/mymodel@sha256:abc…`)
+- `ref` – Tagged ModelKit reference (e.g., `registry.io/myorg/mymodel:v1`)
+- `digest` – Digest-based ModelKit reference (e.g., `registry.io/myorg/mymodel@sha256:abc…`)
 
 ### unpack-modelkit
 
 Pulls a ModelKit from a registry and extracts it.
 
 **Inputs**
 
-- `modelkit_uri` – ModelKit reference (e.g., `registry.io/repo:tag` or `registry.io/repo@sha256:…`)
+- `modelkit_reference` – ModelKit reference (e.g., `registry.io/repo:tag` or `registry.io/repo@sha256:…`)
 - `extract_path` – Directory to extract contents (default: `/tmp/model`)
 
 **Outputs**
@@ -89,7 +89,7 @@ def push_modelkit(
     repository: str,
     tag: str,
     input_modelkit_dir: dsl.Input[dsl.Artifact],
-    output_uri: dsl.Output[dsl.Artifact],
+    output_ref: dsl.Output[dsl.Artifact],
     output_digest: dsl.Output[dsl.Artifact],
     modelkit_name: str = '',
     modelkit_desc: str = '',
@@ -111,7 +111,7 @@ def push_modelkit(
             f'--dataset-uri "{dataset_uri}" '
             f'--code-repo "{code_repo}" '
             f'--code-commit "{code_commit}" '
-            f'&& cp /tmp/outputs/uri "{output_uri.path}" '
+            f'&& cp /tmp/outputs/reference "{output_ref.path}" '
             f'&& cp /tmp/outputs/digest "{output_digest.path}"'
         ],
     )

build/dockerfiles/kubeflow-components/components/push-modelkit/component.yaml

@@ -17,8 +17,8 @@ inputs:
   - {name: code_commit, type: String, optional: true, description: 'Code commit hash'}
 
 outputs:
-  - {name: uri, type: String, description: 'Tagged ModelKit URI (e.g., registry.io/repo:tag)'}
-  - {name: digest, type: String, description: 'Digest-based ModelKit URI (e.g., registry.io/repo@sha256:...)'}
+  - {name: ref, type: String, description: 'Tagged ModelKit reference (e.g., registry.io/repo:tag)'}
+  - {name: digest, type: String, description: 'Digest-based ModelKit reference(e.g., registry.io/repo@sha256:...)'}
 
 implementation:
   container:
@@ -47,5 +47,5 @@ implementation:
       - {inputValue: code_repo}
       - {inputValue: code_commit}
     fileOutputs:
-      uri: /tmp/outputs/uri
+      ref: /tmp/outputs/reference
       digest: /tmp/outputs/digest

build/dockerfiles/kubeflow-components/components/unpack-modelkit/component.yaml

@@ -2,7 +2,7 @@ name: unpack-modelkit
 description: Unpack ModelKit artifacts from a registry
 
 inputs:
-  - {name: modelkit_uri, type: String, description: 'ModelKit reference (e.g., registry.io/repo:tag or registry.io/repo@sha256:...)'}
+  - {name: modelkit_reference, type: String, description: 'ModelKit reference (e.g., registry.io/repo:tag or registry.io/repo@sha256:...)'}
   - {name: extract_path, type: String, default: '/tmp/model', description: 'Directory to extract ModelKit artifacts'}
 
 outputs:
@@ -14,7 +14,7 @@ implementation:
     command:
       - /bin/bash
       - /scripts/unpack-modelkit.sh
-      - {inputValue: modelkit_uri}
+      - {inputValue: modelkit_reference}
       - {inputValue: extract_path}
     fileOutputs:
       model_path: /tmp/outputs/model_path

build/dockerfiles/kubeflow-components/examples/house-prices-pipeline.py

@@ -114,7 +114,7 @@ def push_modelkit(
     registry: str,
     repository: str,
     tag: str,
-    output_uri: dsl.Output[dsl.Artifact],
+    output_ref: dsl.Output[dsl.Artifact],
     output_digest: dsl.Output[dsl.Artifact],
     input_modelkit_dir: dsl.Input[dsl.Artifact],
     modelkit_name: str = '',
@@ -127,8 +127,8 @@ def push_modelkit(
     """Package and push model as ModelKit with attestation.
 
     Outputs:
-        output_uri: Tagged URI (e.g., jozu.ml/repo:tag)
-        output_digest: Digest URI (e.g., jozu.ml/repo@sha256:...)
+        output_ref: Tagged reference (e.g., jozu.ml/repo:tag)
+        output_digest: Digest reference (e.g., jozu.ml/repo@sha256:...)
     """
     # Build command using safe argument passing
     return dsl.ContainerSpec(
@@ -146,7 +146,7 @@ def push_modelkit(
             ${7:+--dataset-uri "$7"} \
             ${8:+--code-repo "$8"} \
             ${9:+--code-commit "$9"} \
-            && cp /tmp/outputs/uri "${10}" \
+            && cp /tmp/outputs/reference "${10}" \
             && cp /tmp/outputs/digest "${11}"
             ''',
             registry,
@@ -159,7 +159,7 @@ def push_modelkit(
             dataset_uri,
             code_repo,
             code_commit,
-            output_uri.path,
+            output_ref.path,
             output_digest.path
         ]
     )

build/dockerfiles/kubeflow-components/scripts/lib/common.sh

@@ -69,8 +69,8 @@ die() {
 
 # Retry logic
 retry() {
-    local max_attempts=${1:-3}
-    local delay=${2:-2}
+    local max_attempts=$1
+    local delay=$2
     shift 2
     local attempt=1
 

build/dockerfiles/kubeflow-components/scripts/push-modelkit.sh

@@ -91,13 +91,12 @@ if [ ! -d "$MODELKIT_DIR" ]; then
     die "ModelKit directory not found: $MODELKIT_DIR"
 fi
 
-# Construct ModelKit URI
-MODELKIT_URI="${REGISTRY}/${REPOSITORY}:${TAG}"
+# Construct ModelKit reference
+MODELKIT_REF="${REGISTRY}/${REPOSITORY}:${TAG}"
 
-log_info "Starting ModelKit push" "{\"uri\":\"$MODELKIT_URI\"}"
+log_info "Starting ModelKit push" "{\"reference\":\"$MODELKIT_REF\"}"
 
 require_cmd kit cosign jq
-require_env DOCKER_CONFIG
 
 # Disable kit update notifications
 kit version --show-update-notifications=false >/dev/null 2>&1 || true
@@ -126,47 +125,41 @@ fi
 
 # Pack the ModelKit
 log_info "Packing ModelKit artifacts"
-kit pack "$WORK_DIR" -t "$MODELKIT_URI" || die "Failed to pack ModelKit"
+kit pack "$WORK_DIR" -t "$MODELKIT_REF" || die "Failed to pack ModelKit"
 
 # Push to registry with retry
 log_info "Pushing to registry"
-retry 3 2 kit push "$MODELKIT_URI" || die "Failed to push ModelKit"
+retry 3 2 kit push "$MODELKIT_REF" || die "Failed to push ModelKit"
 
-# Extract digest from kit inspect
-log_debug "Extracting digest"
-MODELKIT_DIGEST=$(echo "$MODELKIT_URI" | grep -oE '@sha256:[a-f0-9]+' | sed 's/@sha256://' || echo "")
+# Fetch digest from registry
+log_debug "Fetching digest from registry"
 
-if [ -z "$MODELKIT_DIGEST" ]; then
-    log_debug "No digest in URI, fetching from registry"
-
-    set +e
-    INSPECT_OUTPUT=$(kit inspect "$MODELKIT_URI" --remote 2>&1)
-    INSPECT_EXIT_CODE=$?
-    set -e
+set +e
+INSPECT_OUTPUT=$(kit inspect "$MODELKIT_REF" --remote 2>&1)
+INSPECT_EXIT_CODE=$?
+set -e
 
-    log_debug "Kit inspect completed" "{\"exit_code\":$INSPECT_EXIT_CODE}"
+log_debug "Kit inspect completed" "{\"exit_code\":$INSPECT_EXIT_CODE}"
 
-    if [ $INSPECT_EXIT_CODE -eq 0 ]; then
-        # Extract digest from JSON output, filtering out any log lines
-        MODELKIT_DIGEST=$(echo "$INSPECT_OUTPUT" | grep -v '^{"timestamp"' | jq -r '.digest' 2>/dev/null | sed 's/sha256://' || echo "")
-    fi
+if [ $INSPECT_EXIT_CODE -eq 0 ]; then
+    MODELKIT_DIGEST=$(echo "$INSPECT_OUTPUT" | jq -r '.digest' 2>/dev/null || echo "")
+fi
 
-    if [ -z "$MODELKIT_DIGEST" ]; then
-        die "Could not determine ModelKit digest" "{\"reference\":\"$MODELKIT_URI\",\"exit_code\":$INSPECT_EXIT_CODE}"
-    fi
+if [ -z "$MODELKIT_DIGEST" ]; then
+    die "Could not determine ModelKit digest" "{\"reference\":\"$MODELKIT_REF\",\"exit_code\":$INSPECT_EXIT_CODE}"
 fi
 
 log_debug "ModelKit digest: $MODELKIT_DIGEST"
 
-# Construct full URI with digest
-FULL_URI="${REGISTRY}/${REPOSITORY}@sha256:${MODELKIT_DIGEST}"
+# Construct full reference with digest
+FULL_REF="${REGISTRY}/${REPOSITORY}@${MODELKIT_DIGEST}"
 
-log_info "Push completed" "{\"uri\":\"$FULL_URI\"}"
+log_info "Push completed" "{\"reference\":\"$FULL_REF\"}"
 
 # Create in-toto attestation predicate
 ATTESTATION_PREDICATE=$(jq -nc \
-    --arg uri "$FULL_URI" \
-    --arg digest "sha256:$MODELKIT_DIGEST" \
+    --arg reference "$FULL_REF" \
+    --arg digest "$MODELKIT_DIGEST" \
     --arg dataset_uri "$DATASET_URI" \
     --arg code_repo "$CODE_REPO" \
     --arg code_commit "$CODE_COMMIT" \
@@ -175,7 +168,7 @@ ATTESTATION_PREDICATE=$(jq -nc \
         predicateType: "https://kitops.ml/attestation/v1",
         predicate: {
             modelkit: {
-                uri: $uri,
+                reference: $reference,
                 digest: $digest
             },
             metadata: {
@@ -201,7 +194,7 @@ if [ -f "/etc/cosign/cosign.key" ]; then
         --predicate "$PREDICATE_FILE" \
         --tlog-upload=false \
         --yes \
-        "$FULL_URI" 2>&1; then
+        "$FULL_REF" 2>&1; then
         log_info "Signed with cosign"
     else
         log_warn "Failed to sign with cosign, continuing"
@@ -214,16 +207,16 @@ fi
 
 # Output results
 # Write to KFP output files
-echo -n "$MODELKIT_URI" > /tmp/outputs/uri         # Tagged URI (e.g., jozu.ml/repo:tag)
-echo -n "$FULL_URI" > /tmp/outputs/digest          # Digest URI (e.g., jozu.ml/repo@sha256:...)
+echo -n "$MODELKIT_REF" > /tmp/outputs/reference         # Tagged reference (e.g., jozu.ml/repo:tag)
+echo -n "$FULL_REF" > /tmp/outputs/digest          # Digest reference (e.g., jozu.ml/repo@sha256:...)
 
 # Output JSON to stdout
 jq -n \
-    --arg uri "$FULL_URI" \
-    --arg digest "sha256:$MODELKIT_DIGEST" \
+    --arg reference "$FULL_REF" \
+    --arg digest "$MODELKIT_DIGEST" \
     --arg timestamp "$(date -u +"%Y-%m-%dT%H:%M:%SZ")" \
     '{
-        "uri": $uri,
+        "reference": $reference,
         "digest": $digest,
         "timestamp": $timestamp,
         "status": "success"

build/dockerfiles/kubeflow-components/scripts/unpack-modelkit.sh

@@ -1,7 +1,7 @@
 #!/bin/bash
 set -euo pipefail
 
-# Usage: /scripts/unpack-modelkit.sh <modelkit_uri> [extract_path]
+# Usage: /scripts/unpack-modelkit.sh <modelkit_reference> [extract_path]
 # Environment variables: `DOCKER_CONFIG` (path to .docker directory containing config.json)
 # Unpacks ModelKit artifacts to a directory
 
@@ -10,13 +10,13 @@ source "${SCRIPT_DIR}/lib/common.sh"
 
 # Validate arguments
 if [ $# -lt 1 ]; then
-    die "Usage: $0 <modelkit_uri> [extract_path]"
+    die "Usage: $0 <modelkit_reference> [extract_path]"
 fi
 
-MODELKIT_URI="$1"
+MODELKIT_REF="$1"
 EXTRACT_PATH="${2:-/tmp/model}"
 
-log_info "Starting unpack" "{\"modelkit_uri\":\"$MODELKIT_URI\",\"extract_path\":\"$EXTRACT_PATH\"}"
+log_info "Starting unpack" "{\"modelkit_reference\":\"$MODELKIT_REF\",\"extract_path\":\"$EXTRACT_PATH\"}"
 
 require_cmd kit jq
 require_env DOCKER_CONFIG
@@ -30,7 +30,7 @@ mkdir -p "$EXTRACT_PATH"
 
 # Step 1: Unpack ModelKit with retry
 log_info "Unpacking"
-retry 3 2 kit unpack "$MODELKIT_URI" -d "$EXTRACT_PATH" || die "Failed to unpack ModelKit"
+retry 3 2 kit unpack "$MODELKIT_REF" -d "$EXTRACT_PATH" || die "Failed to unpack ModelKit"
 
 log_info "Unpacked successfully" "{\"path\":\"$EXTRACT_PATH\"}"
 
@@ -41,11 +41,11 @@ echo -n "$EXTRACT_PATH" > /tmp/outputs/model_path
 # Output JSON to stdout
 jq -n \
     --arg path "$EXTRACT_PATH" \
-    --arg uri "$MODELKIT_URI" \
+    --arg reference "$MODELKIT_REF" \
     --arg timestamp "$(date -u +"%Y-%m-%dT%H:%M:%SZ")" \
     '{
         "model_path": $path,
-        "modelkit_uri": $uri,
+        "modelkit_reference": $reference,
         "timestamp": $timestamp,
         "status": "success"
     }'

build/dockerfiles/kubeflow-components/tests/push-modelkit.bats

@@ -239,18 +239,18 @@ EOF
 @test "creates output files in /tmp/outputs" {
     run bash "$SCRIPT_PATH" "registry.io" "myorg/mymodel" "v1" --modelkit-dir "$MODEL_DIR"
     [ "$status" -eq 0 ]
-    [ -f "$OUTPUT_DIR/uri" ]
+    [ -f "$OUTPUT_DIR/reference" ]
     [ -f "$OUTPUT_DIR/digest" ]
 }
 
 @test "output files contain correct values" {
     run bash "$SCRIPT_PATH" "registry.io" "myorg/mymodel" "v1" --modelkit-dir "$MODEL_DIR"
     [ "$status" -eq 0 ]
 
-    uri_content=$(cat "$OUTPUT_DIR/uri")
+    ref_content=$(cat "$OUTPUT_DIR/reference")
     digest_content=$(cat "$OUTPUT_DIR/digest")
 
-    [[ "$uri_content" == "registry.io/myorg/mymodel:v1" ]]
+    [[ "$ref_content" == "registry.io/myorg/mymodel:v1" ]]
     [[ "$digest_content" =~ registry.io/myorg/mymodel@sha256:abc123def456 ]]
 }
 
@@ -260,7 +260,7 @@ EOF
 
     # Extract final JSON output
     json_output=$(echo "$output" | awk '/^{$/,/^}$/' | jq -s '.[] | select(.status != null)')
-    echo "$json_output" | jq -e '.uri'
+    echo "$json_output" | jq -e '.reference'
     echo "$json_output" | jq -e '.digest'
     echo "$json_output" | jq -e '.status == "success"'
 }
@@ -303,13 +303,6 @@ EOF
     [[ "$output" =~ "Required command not found: kit" ]]
 }
 
-@test "fails when DOCKER_CONFIG not set" {
-    unset DOCKER_CONFIG
-    run bash "$SCRIPT_PATH" "registry.io" "myorg/mymodel" "v1" --modelkit-dir "$MODEL_DIR"
-    [ "$status" -eq 1 ]
-    [[ "$output" =~ "Required environment variable not set: DOCKER_CONFIG" ]]
-}
-
 @test "retries on push failure and eventually fails" {
     # Replace kit with failing version
     mv "$TEST_DIR/bin/kit" "$TEST_DIR/bin/kit.bak"

build/dockerfiles/kubeflow-components/tests/unpack-modelkit.bats

@@ -96,10 +96,10 @@ teardown() {
     run bash "$SCRIPT_PATH"
     [ "$status" -eq 1 ]
     [[ "$output" =~ "Usage:" ]]
-    [[ "$output" =~ "modelkit_uri" ]]
+    [[ "$output" =~ "modelkit_reference" ]]
 }
 
-@test "succeeds with only modelkit_uri (uses default extract path)" {
+@test "succeeds with only modelkit_reference (uses default extract path)" {
     run bash "$SCRIPT_PATH" "registry.io/myorg/mymodel:v1"
     [ "$status" -eq 0 ]
     [[ "$output" =~ "Unpack workflow completed" ]]
@@ -136,7 +136,7 @@ teardown() {
     # Extract final JSON output (the one with "status" field)
     json_output=$(echo "$output" | awk '/^{$/,/^}$/' | jq -s '.[] | select(.status != null)')
     echo "$json_output" | jq -e '.model_path'
-    echo "$json_output" | jq -e '.modelkit_uri'
+    echo "$json_output" | jq -e '.modelkit_reference'
     echo "$json_output" | jq -e '.status == "success"'
 }