Add test for pareto filter (opendatahub-io#962)

* feat: Add test for pareto filter

Signed-off-by: lugi0 <lgiorgi@redhat.com>

* fix: change comment

Signed-off-by: lugi0 <lgiorgi@redhat.com>

* fix: address comments

Signed-off-by: lugi0 <lgiorgi@redhat.com>

* fix: use model from constant list

Signed-off-by: lugi0 <lgiorgi@redhat.com>

---------

Signed-off-by: lugi0 <lgiorgi@redhat.com>

Author: lugi0

tests/model_registry/model_catalog/test_model_artifact_search.py

@@ -6,6 +6,7 @@
     fetch_all_artifacts_with_dynamic_paging,
     validate_model_artifacts_match_criteria_and,
     validate_model_artifacts_match_criteria_or,
+    validate_recommendations_subset,
 )
 from tests.model_registry.model_catalog.constants import (
     VALIDATED_CATALOG_ID,
@@ -204,3 +205,66 @@ def test_filter_query_advanced_artifact_search(
                 )
             else:
                 pytest.fail(f"{logic_type} filter validation failed for model {model_name}")
+
+    @pytest.mark.parametrize(
+        "randomly_picked_model_from_catalog_api_by_source",
+        [
+            pytest.param(
+                {
+                    "catalog_id": VALIDATED_CATALOG_ID,
+                    "header_type": "registry",
+                    "model_name": random.choice(MODEL_NAMEs_ARTIFACT_SEARCH),
+                },
+                id="test_performance_artifacts_recommendations_parameter",
+            ),
+        ],
+        indirect=["randomly_picked_model_from_catalog_api_by_source"],
+    )
+    def test_performance_artifacts_recommendations_parameter(
+        self: Self,
+        enabled_model_catalog_config_map: ConfigMap,
+        model_catalog_rest_url: list[str],
+        model_registry_rest_headers: dict[str, str],
+        randomly_picked_model_from_catalog_api_by_source: tuple[dict, str, str],
+    ):
+        """
+        Test the recommendations query parameter for performance artifacts endpoint.
+
+        Validates that recommendations=true returns a filtered subset of performance
+        artifacts that are optimal based on cost and latency compared to the full set.
+        """
+        _, model_name, catalog_id = randomly_picked_model_from_catalog_api_by_source
+
+        LOGGER.info(f"Testing performance artifacts recommendations parameter for model: {model_name}")
+
+        # Get all performance artifacts (baseline)
+        full_results = fetch_all_artifacts_with_dynamic_paging(
+            url_with_pagesize=(
+                f"{model_catalog_rest_url[0]}sources/{catalog_id}/models/{model_name}/artifacts/performance?pageSize"
+            ),
+            headers=model_registry_rest_headers,
+            page_size=100,
+        )
+
+        # Get recommendations-filtered performance artifacts
+        recommendations_results = fetch_all_artifacts_with_dynamic_paging(
+            url_with_pagesize=(
+                f"{model_catalog_rest_url[0]}sources/{catalog_id}/models/{model_name}/artifacts/performance?"
+                f"recommendations=true&pageSize"
+            ),
+            headers=model_registry_rest_headers,
+            page_size=100,
+        )
+
+        if (full_results and not recommendations_results) or (len(recommendations_results) > len(full_results)):
+            pytest.fail(f"Recommendations parameter functionality failed for model {model_name}")
+
+        # Validate subset relationship
+        validation_passed = validate_recommendations_subset(
+            full_artifacts=full_results["items"],
+            recommendations_artifacts=recommendations_results["items"],
+            model_name=model_name,
+        )
+
+        assert validation_passed, f"Recommendations subset validation failed for model {model_name}"
+        LOGGER.info(f"Successfully validated recommendations parameter functionality for model {model_name}")

tests/model_registry/model_catalog/utils.py

@@ -1272,3 +1272,47 @@ def get_metadata_from_catalog_pod(model_catalog_pod: Pod, model_name: str) -> di
     except Exception as e:
         LOGGER.error(f"Failed to read metadata.json for model '{model_name}': {e}")
         raise
+
+
+def validate_recommendations_subset(
+    full_artifacts: list[dict[str, Any]], recommendations_artifacts: list[dict[str, Any]], model_name: str
+) -> bool:
+    """
+    Validate that recommendations artifacts are a proper subset of all artifacts.
+
+    Args:
+        full_artifacts: All performance artifacts (recommendations=false)
+        recommendations_artifacts: Filtered artifacts (recommendations=true)
+        model_name: Model name for logging
+
+    Returns:
+        bool: True if validation passes
+
+    Raises:
+        AssertionError: If validation fails with descriptive message
+    """
+    LOGGER.info(f"Validating recommendations subset for model '{model_name}'")
+
+    # Convert artifacts to comparable format (using artifact ID for comparison)
+    full_artifact_ids = {artifact.get("id") for artifact in full_artifacts if artifact.get("id")}
+    recommendations_artifact_ids = {artifact.get("id") for artifact in recommendations_artifacts if artifact.get("id")}
+
+    # Check subset relationship: all recommendation IDs should exist in full results
+    missing_in_full = recommendations_artifact_ids - full_artifact_ids
+    if missing_in_full:
+        error_msg = (
+            f"Model '{model_name}': Found {len(missing_in_full)} recommendation artifacts "
+            f"that don't exist in full results: {missing_in_full}"
+        )
+        LOGGER.error(error_msg)
+        raise AssertionError(error_msg)
+
+    # Log success details
+    subset_percentage = (len(recommendations_artifacts) / len(full_artifacts)) * 100
+    LOGGER.info(
+        f"Model '{model_name}': Recommendations validation passed - "
+        f"{len(recommendations_artifacts)}/{len(full_artifacts)} artifacts "
+        f"({subset_percentage:.1f}% of total)"
+    )
+
+    return True