ibm-watsonx-gov 1.3.3__cp313-cp313-macosx_11_0_arm64.whl

ibm_watsonx_gov/metrics/ndcg/ndcg_decorator.py

@@ -0,0 +1,61 @@
+# ----------------------------------------------------------------------------------------------------
+# IBM Confidential
+# Licensed Materials - Property of IBM
+# 5737-H76, 5900-A3Q
+# © Copyright IBM Corp. 2025  All Rights Reserved.
+# US Government Users Restricted Rights - Use, duplication or disclosure restricted by
+# GSA ADPSchedule Contract with IBM Corp.
+# ----------------------------------------------------------------------------------------------------
+from functools import partial
+from typing import Callable, Optional
+
+from wrapt import decorator
+
+from ibm_watsonx_gov.config.agentic_ai_configuration import \
+    AgenticAIConfiguration
+from ibm_watsonx_gov.entities.enums import EvaluatorFields
+from ibm_watsonx_gov.entities.metric import GenAIMetric
+from ibm_watsonx_gov.metrics.base_metric_decorator import BaseMetricDecorator
+from ibm_watsonx_gov.metrics.context_relevance.context_relevance_metric import \
+    ContextRelevanceMetric
+from ibm_watsonx_gov.metrics.ndcg.ndcg_metric import NDCGMetric
+
+
+class NDCGDecorator(BaseMetricDecorator):
+    def evaluate_ndcg(self,
+                      func: Optional[Callable] = None,
+                      *,
+                      configuration: Optional[AgenticAIConfiguration] = None,
+                      metrics: list[GenAIMetric] = []
+                      ) -> dict:
+        """
+        An evaluation decorator for computing ndcg metric on an agentic node.
+        """
+        if func is None:
+            return partial(self.evaluate_ndcg, configuration=configuration, metrics=metrics)
+
+        if not metrics:
+            metrics = [NDCGMetric()]
+
+        @decorator
+        def wrapper(func, instance, args, kwargs):
+
+            try:
+                self.validate(func=func, metrics=metrics,
+                              valid_metric_types=(NDCGMetric, ContextRelevanceMetric))
+
+                metric_inputs = [EvaluatorFields.INPUT_FIELDS]
+                metric_outputs = [EvaluatorFields.CONTEXT_FIELDS]
+
+                original_result = self.compute_helper(func=func, args=args, kwargs=kwargs,
+                                                      configuration=configuration,
+                                                      metrics=metrics,
+                                                      metric_inputs=metric_inputs,
+                                                      metric_outputs=metric_outputs)
+
+                return original_result
+            except Exception as ex:
+                raise Exception(
+                    f"There was an error while evaluating ndcg metric on {func.__name__},") from ex
+
+        return wrapper(func)

ibm_watsonx_gov/metrics/ndcg/ndcg_metric.py

@@ -0,0 +1,166 @@
+# ----------------------------------------------------------------------------------------------------
+# IBM Confidential
+# Licensed Materials - Property of IBM
+# 5737-H76, 5900-A3Q
+# © Copyright IBM Corp. 2025  All Rights Reserved.
+# US Government Users Restricted Rights - Use, duplication or disclosure restricted by
+# GSA ADPSchedule Contract with IBM Corp.
+# ----------------------------------------------------------------------------------------------------
+
+from typing import Annotated, Any, Literal
+
+import numpy as np
+import pandas as pd
+from pydantic import Field, TypeAdapter, field_validator
+from sklearn.metrics import ndcg_score
+
+from ibm_watsonx_gov.config import AgenticAIConfiguration, GenAIConfiguration
+from ibm_watsonx_gov.entities.enums import MetricGroup, TaskType
+from ibm_watsonx_gov.entities.evaluation_result import (AggregateMetricResult,
+                                                        RecordMetricResult)
+from ibm_watsonx_gov.entities.metric import GenAIMetric
+from ibm_watsonx_gov.entities.metric_threshold import MetricThreshold
+from ibm_watsonx_gov.metrics.context_relevance.context_relevance_metric import (
+    CONTEXT_RELEVANCE, ContextRelevanceMetric, ContextRelevanceResult)
+
+NDCG = "ndcg"
+NDCG_DISPLAY_NAME = "Normalized Discounted Cumulative Gain"
+
+
+class NDCGResult(RecordMetricResult):
+    name: str = NDCG
+    display_name: str = NDCG_DISPLAY_NAME
+    group: MetricGroup = MetricGroup.RETRIEVAL_QUALITY
+
+
+class NDCGMetric(GenAIMetric):
+    """
+    Defines the NDCG(Normalized Discounted Cumulative Gain) metric class.
+
+    The Normalized Discounted Cumulative Gain metric measures the ranking quality of the retrieved contexts.
+    The Context Relevance metric is computed as a pre requisite to compute this metric.
+
+    Examples:
+        1. Create NDCG metric with default parameters and compute using metrics evaluator.
+            .. code-block:: python
+
+                metric = NDCGMetric()
+                result = MetricsEvaluator().evaluate(data={"input_text": "...", "context": "..."},
+                                                    metrics=[metric])
+                # A list of contexts can also be passed as shown below
+                result = MetricsEvaluator().evaluate(data={"input_text": "...", "context": ["...", "..."]},
+                                                    metrics=[metric])
+
+        2. Create NDCG metric with a custom threshold.
+            .. code-block:: python
+
+                threshold  = MetricThreshold(type="lower_limit", value=0.5)
+                metric = NDCGMetric(method=method, threshold=threshold)
+
+        3. Create NDCG metric with llm_as_judge method.
+            .. code-block:: python
+
+                # Define LLM Judge using watsonx.ai
+                # To use other frameworks and models as llm_judge, see :module:`ibm_watsonx_gov.entities.foundation_model`
+                llm_judge = LLMJudge(model=WxAIFoundationModel(
+                                            model_id="ibm/granite-3-3-8b-instruct",
+                                            project_id="<PROJECT_ID>"
+                                    ))
+                cr_metric = ContextRelevanceMetric(llm_judge=llm_judge)
+                ap_metric = NDCGMetric()
+                result = MetricsEvaluator().evaluate(data={"input_text": "...", "context": ["...", "..."]},
+                                                    metrics=[cr_metric, ap_metric])
+    """
+    name: Annotated[Literal["ndcg"],
+                    Field(title="Name",
+                          description="The ndcg metric name.",
+                          default=NDCG, frozen=True)]
+    display_name: Annotated[Literal["Normalized Discounted Cumulative Gain"],
+                            Field(title="Display Name",
+                                  description="The ndcg metric display name.",
+                                  default=NDCG_DISPLAY_NAME, frozen=True)]
+    tasks: Annotated[list[TaskType],
+                     Field(title="Tasks",
+                           description="The list of supported tasks.",
+                           default=[TaskType.RAG])]
+    thresholds: Annotated[list[MetricThreshold],
+                          Field(title="Thresholds",
+                                description="The metric thresholds.",
+                                default=[MetricThreshold(type="lower_limit", value=0.7)])]
+    metric_dependencies: Annotated[list[GenAIMetric],
+                                   Field(title="Metric dependencies",
+                                         description="The list of metric dependencies",
+                                         default=[ContextRelevanceMetric()])]
+    group: Annotated[MetricGroup,
+                     Field(title="Group",
+                           description="The metric group.",
+                           default=MetricGroup.RETRIEVAL_QUALITY, frozen=True)]
+
+    @field_validator("metric_dependencies", mode="before")
+    @classmethod
+    def metric_dependencies_validator(cls, value: Any):
+        if value:
+            value = [TypeAdapter(Annotated[ContextRelevanceMetric, Field(
+                discriminator="name")]).validate_python(
+                m) for m in value]
+        return value
+
+    def evaluate(
+            self,
+            data: pd.DataFrame,
+            configuration: GenAIConfiguration | AgenticAIConfiguration,
+            metrics_result: list[AggregateMetricResult],
+            **kwargs,
+    ) -> AggregateMetricResult:
+        record_level_metrics = []
+        scores = []
+
+        context_relevance_result: list[ContextRelevanceResult] = next(
+            (metric_result.record_level_metrics for metric_result in metrics_result if metric_result.name == CONTEXT_RELEVANCE), None)
+
+        if context_relevance_result is None:
+            raise Exception(
+                f"Failed to evaluate {self.name} metric. Missing context relevance metric result")
+
+        for relevance_result in context_relevance_result:
+            score = self.__compute(
+                relevance_result.additional_info.get(
+                    "contexts_values", []),)
+            scores.append(score)
+            record_level_metrics.append(
+                NDCGResult(
+                    method="",
+                    provider="",
+                    record_id=relevance_result.record_id,
+                    value=score,
+                    thresholds=self.thresholds
+                )
+            )
+
+        mean = sum(scores) / len(scores)
+        aggregate_metric_score = AggregateMetricResult(
+            name=self.name,
+            display_name=self.display_name,
+            method="",
+            provider="",
+            group=self.group,
+            min=min(scores),
+            max=max(scores),
+            mean=mean,
+            value=mean,
+            total_records=len(record_level_metrics),
+            record_level_metrics=record_level_metrics,
+            thresholds=self.thresholds
+        )
+
+        return aggregate_metric_score
+
+    def __compute(self, relevance_scores: list[float]) -> float:
+        if len(relevance_scores) < 2:
+            return 1.0
+
+        true_relevance = np.sort(relevance_scores)[::-1]
+
+        ndcg_value = ndcg_score([true_relevance], [relevance_scores])
+
+        return round(ndcg_value, 4)

ibm_watsonx_gov/metrics/output_token_count/__init__.py

@@ -0,0 +1,8 @@
+# ----------------------------------------------------------------------------------------------------
+# IBM Confidential
+# Licensed Materials - Property of IBM
+# 5737-H76, 5900-A3Q
+# © Copyright IBM Corp. 2025  All Rights Reserved.
+# US Government Users Restricted Rights - Use, duplication or disclosure restricted by
+# GSA ADPSchedule Contract with IBM Corp.
+# ----------------------------------------------------------------------------------------------------

ibm_watsonx_gov/metrics/output_token_count/output_token_count_decorator.py

@@ -0,0 +1,58 @@
+# ----------------------------------------------------------------------------------------------------
+# IBM Confidential
+# Licensed Materials - Property of IBM
+# 5737-H76, 5900-A3Q
+# © Copyright IBM Corp. 2025  All Rights Reserved.
+# US Government Users Restricted Rights - Use, duplication or disclosure restricted by
+# GSA ADPSchedule Contract with IBM Corp.
+# ----------------------------------------------------------------------------------------------------
+from functools import partial
+from typing import Callable, Optional
+
+from wrapt import decorator
+
+from ibm_watsonx_gov.config.agentic_ai_configuration import \
+    AgenticAIConfiguration
+from ibm_watsonx_gov.entities.enums import EvaluatorFields
+from ibm_watsonx_gov.entities.metric import GenAIMetric
+from ibm_watsonx_gov.metrics.base_metric_decorator import BaseMetricDecorator
+from ibm_watsonx_gov.metrics.output_token_count.output_token_count_metric import \
+    OutputTokenCountMetric
+
+
+class OutputTokenCountDecorator(BaseMetricDecorator):
+    def evaluate_output_token_count(self,
+                                    func: Optional[Callable] = None,
+                                    *,
+                                    configuration: Optional[AgenticAIConfiguration] = None,
+                                    metrics: list[GenAIMetric] = []
+                                    ) -> dict:
+        """
+        An evaluation decorator for computing total output token.
+        """
+        if func is None:
+            return partial(self.evaluate_output_token_count, configuration=configuration, metrics=metrics)
+
+        if not metrics:
+            metrics = [OutputTokenCountMetric()]
+
+        @decorator
+        def wrapper(func, instance, args, kwargs):
+
+            try:
+                self.validate(func=func, metrics=metrics,
+                              valid_metric_types=(OutputTokenCountMetric,))
+
+                metric_inputs = [EvaluatorFields.OUTPUT_TOKEN_COUNT_FIELDS]
+
+                original_result = self.compute_helper(func=func, args=args, kwargs=kwargs,
+                                                      configuration=configuration,
+                                                      metrics=metrics,
+                                                      metric_inputs=metric_inputs)
+
+                return original_result
+            except Exception as ex:
+                raise Exception(
+                    f"There was an error while tracking total output token count on {func.__name__},") from ex
+
+        return wrapper(func)

ibm_watsonx_gov/metrics/output_token_count/output_token_count_metric.py

@@ -0,0 +1,112 @@
+# ----------------------------------------------------------------------------------------------------
+# IBM Confidential
+# Licensed Materials - Property of IBM
+# 5737-H76, 5900-A3Q
+# © Copyright IBM Corp. 2025  All Rights Reserved.
+# US Government Users Restricted Rights - Use, duplication or disclosure restricted by
+# GSA ADPSchedule Contract with IBM Corp.
+# ----------------------------------------------------------------------------------------------------
+
+from typing import Annotated, Literal
+
+import pandas as pd
+import re
+from pydantic import Field
+
+from ibm_watsonx_gov.config import AgenticAIConfiguration, GenAIConfiguration
+from ibm_watsonx_gov.entities.enums import MetricGroup, TaskType
+from ibm_watsonx_gov.entities.evaluation_result import (AggregateMetricResult,
+                                                        RecordMetricResult)
+from ibm_watsonx_gov.entities.metric import GenAIMetric
+from ibm_watsonx_gov.utils.validation_util import validate_field
+
+OUTPUT_TOKEN_COUNT = "output_token_count"
+OUTPUT_TOKEN_COUNT_DISPLAY_NAME = "Output Token Count"
+
+
+class OutputTokenCountResult(RecordMetricResult):
+    name: str = OUTPUT_TOKEN_COUNT
+    display_name: str = OUTPUT_TOKEN_COUNT_DISPLAY_NAME
+
+
+class OutputTokenCountMetric(GenAIMetric):
+    """
+    Defines the output token count metric class.
+
+    The output token count metric keep track of LLM output token count.
+
+    Examples:
+        1. Create Cost metric with default parameters and compute using metrics evaluator.
+            .. code-block:: python
+
+                metric = OutputTokenCountMetric()
+                result = MetricsEvaluator().evaluate(data={"completion_tokens": "..."},
+                                                    metrics=[metric])
+    """
+    name: Annotated[Literal["output_token_count"],
+                    Field(title="name",
+                          description="The output token count metric name.",
+                          default=OUTPUT_TOKEN_COUNT, frozen=True)]
+    display_name: Annotated[Literal["Output Token Count"],
+                            Field(title="Display Name",
+                                  description="The output token count metric display name.",
+                                  default=OUTPUT_TOKEN_COUNT_DISPLAY_NAME, frozen=True)]
+    tasks: Annotated[list[TaskType],
+                     Field(title="Tasks",
+                           description="The list of supported tasks.",
+                           default=TaskType.values(), frozen=True)]
+    group: Annotated[MetricGroup,
+                     Field(title="Group",
+                           description="The metric group.",
+                           default=MetricGroup.USAGE, frozen=True)]
+
+    def evaluate(
+            self,
+            data: pd.DataFrame,
+            configuration: GenAIConfiguration | AgenticAIConfiguration,
+            **kwargs,
+    ) -> list[AggregateMetricResult]:
+        from ibm_watsonx_gov.utils.aggregation_util import get_summaries
+
+        validate_field("output_token_count_fields", configuration)
+        record_ids = data[configuration.record_id_field].to_list()
+        data = data.fillna(0)
+        output_tokens = self._evaluate(data, configuration)
+        record_level_metrics = [
+            OutputTokenCountResult(record_id=record_id,
+                                   value=token, group=MetricGroup.USAGE.value)
+            for token, record_id in zip(output_tokens, record_ids)
+        ]
+        summary = get_summaries(output_tokens)
+        aggregate_metric_scores = AggregateMetricResult(
+            name=self.name,
+            display_name=self.display_name,
+            group=self.group,
+            min=summary.get("min"),
+            max=summary.get("max"),
+            mean=summary.get("mean"),
+            value=summary.get("mean"),
+            total_records=len(record_level_metrics),
+            record_level_metrics=record_level_metrics,
+        )
+
+        return aggregate_metric_scores
+
+    def _evaluate(self, data: pd.DataFrame, config: GenAIConfiguration | AgenticAIConfiguration) -> list:
+        """
+        Track total output token.
+        """
+        matched_cols = []
+
+        for pattern in config.output_token_count_fields:
+            # Compile regex pattern for safety and performance
+            regex = re.compile(pattern)
+            # Filter columns matching this pattern
+            matched = [col for col in data.columns if regex.fullmatch(col)]
+            matched_cols.extend(matched)
+
+        # Remove duplicates in case multiple patterns match the same column
+        matched_cols = list(set(matched_cols))
+
+        # Sum across these columns row-wise
+        return data[matched_cols].sum(axis=1).tolist()

ibm_watsonx_gov/metrics/pii/__init__.py

@@ -0,0 +1,16 @@
+# ----------------------------------------------------------------------------------------------------
+# IBM Confidential
+# Licensed Materials - Property of IBM
+# 5737-H76, 5900-A3Q
+# © Copyright IBM Corp. 2025  All Rights Reserved.
+# US Government Users Restricted Rights - Use, duplication or disclosure restricted by
+# GSA ADPSchedule Contract with IBM Corp.
+# ----------------------------------------------------------------------------------------------------
+
+from ibm_watsonx_gov.metrics.pii.input_pii_metric import InputPIIMetric
+from ibm_watsonx_gov.metrics.pii.output_pii_metric import OutputPIIMetric
+from ibm_watsonx_gov.metrics.pii.pii_metric import PIIMetric
+
+__all__ = ["PIIMetric", "InputPIIMetric", "OutputPIIMetric"]
+
+# Made with Bob

ibm_watsonx_gov/metrics/pii/input_pii_metric.py

@@ -0,0 +1,102 @@
+# ----------------------------------------------------------------------------------------------------
+# IBM Confidential
+# Licensed Materials - Property of IBM
+# 5737-H76, 5900-A3Q
+# © Copyright IBM Corp. 2025  All Rights Reserved.
+# US Government Users Restricted Rights - Use, duplication or disclosure restricted by
+# GSA ADPSchedule Contract with IBM Corp.
+# ----------------------------------------------------------------------------------------------------
+
+from typing import Annotated, Literal
+
+import pandas as pd
+from pydantic import Field
+
+from ibm_watsonx_gov.config.gen_ai_configuration import GenAIConfiguration
+from ibm_watsonx_gov.entities.enums import MetricGroup, TaskType
+from ibm_watsonx_gov.entities.evaluation_result import AggregateMetricResult
+from ibm_watsonx_gov.entities.metric import GenAIMetric
+from ibm_watsonx_gov.entities.metric_threshold import MetricThreshold
+from ibm_watsonx_gov.providers.detectors_provider import DetectorsProvider
+from ibm_watsonx_gov.utils.async_util import run_in_event_loop
+
+INPUT_PII = "input_pii"
+
+
+class InputPIIMetric(GenAIMetric):
+    """
+    Defines the Input PII metric class.
+
+    The Input PII metric measures if your model input data contains any personally identifiable information.
+    It is computed using the Watson Natural Language Processing entity extraction model on the input data.
+
+    Examples:
+        1. Create Input PII metric with default parameters and compute using metrics evaluator.
+            .. code-block:: python
+
+                metric = InputPIIMetric()
+                result = MetricsEvaluator().evaluate(data={"input_text": "...", metrics=[metric])
+
+        2. Create Input PII metric with a custom threshold.
+            .. code-block:: python
+
+                threshold  = MetricThreshold(type="lower_limit", value=0.5)
+                metric = InputPIIMetric(threshold=threshold)
+    """
+    name: Annotated[Literal["input_pii"],
+                    Field(title="Name",
+                          description="The input pii metric name.",
+                          default=INPUT_PII, frozen=True)]
+    display_name: Annotated[Literal["Input PII"],
+                            Field(title="Display Name",
+                                  description="The input pii metric display name.",
+                                  default="Input PII", frozen=True)]
+    tasks: Annotated[list[TaskType],
+                     Field(title="Tasks",
+                           description="The list of supported tasks.",
+                           default=TaskType.values(), frozen=True)]
+    thresholds: Annotated[list[MetricThreshold],
+                          Field(title="Thresholds",
+                                description="The metric thresholds.",
+                                default=[MetricThreshold(type="upper_limit", value=0.1)])]
+    group: Annotated[MetricGroup,
+                     Field(title="Group",
+                           description="The metric group.",
+                           default=MetricGroup.CONTENT_SAFETY, frozen=True)]
+
+    async def evaluate_async(
+            self,
+            data: pd.DataFrame | dict,
+            configuration: GenAIConfiguration,
+            **kwargs
+    ) -> AggregateMetricResult:
+
+        provider = DetectorsProvider(configuration=configuration,
+                                     metric_name="pii",
+                                     metric_display_name=self.display_name,
+                                     metric_method=self.method,
+                                     metric_group=self.group,
+                                     thresholds=self.thresholds,
+                                     **kwargs)
+        aggregated_metric_result = await provider.evaluate_async(data=data)
+        # Update the metric name to input_pii
+        aggregated_metric_result.name = self.name
+        for record in aggregated_metric_result.record_level_metrics:
+            record.name = self.name
+        return aggregated_metric_result
+
+    def evaluate(
+        self,
+        data: pd.DataFrame | dict,
+        configuration: GenAIConfiguration,
+        **kwargs,
+    ):
+        # If ran in sync mode, block until it is done
+        return run_in_event_loop(
+            self.evaluate_async,
+            data=data,
+            configuration=configuration,
+            **kwargs,
+        )
+
+# Made with Bob

ibm_watsonx_gov/metrics/pii/output_pii_metric.py

@@ -0,0 +1,107 @@
+# ----------------------------------------------------------------------------------------------------
+# IBM Confidential
+# Licensed Materials - Property of IBM
+# 5737-H76, 5900-A3Q
+# © Copyright IBM Corp. 2025  All Rights Reserved.
+# US Government Users Restricted Rights - Use, duplication or disclosure restricted by
+# GSA ADPSchedule Contract with IBM Corp.
+# ----------------------------------------------------------------------------------------------------
+
+from typing import Annotated, Literal
+
+import pandas as pd
+from pydantic import Field
+
+from ibm_watsonx_gov.config.gen_ai_configuration import GenAIConfiguration
+from ibm_watsonx_gov.entities.enums import MetricGroup, TaskType
+from ibm_watsonx_gov.entities.evaluation_result import AggregateMetricResult
+from ibm_watsonx_gov.entities.metric import GenAIMetric
+from ibm_watsonx_gov.entities.metric_threshold import MetricThreshold
+from ibm_watsonx_gov.providers.detectors_provider import DetectorsProvider
+from ibm_watsonx_gov.utils.async_util import run_in_event_loop
+
+OUTPUT_PII = "output_pii"
+
+
+class OutputPIIMetric(GenAIMetric):
+    """
+    Defines the Output PII metric class.
+
+    The Output PII metric measures if your model output data contains any personally identifiable information.
+    It is computed using the Watson Natural Language Processing entity extraction model on the output data.
+
+    Examples:
+        1. Create Output PII metric with default parameters and compute using metrics evaluator.
+            .. code-block:: python
+
+                metric = OutputPIIMetric()
+                result = MetricsEvaluator().evaluate(data={"generated_text": "...", metrics=[metric])
+
+        2. Create Output PII metric with a custom threshold.
+            .. code-block:: python
+
+                threshold  = MetricThreshold(type="lower_limit", value=0.5)
+                metric = OutputPIIMetric(threshold=threshold)
+    """
+    name: Annotated[Literal["output_pii"],
+                    Field(title="Name",
+                          description="The output pii metric name.",
+                          default=OUTPUT_PII, frozen=True)]
+    display_name: Annotated[Literal["Output PII"],
+                            Field(title="Display Name",
+                                  description="The output pii metric display name.",
+                                  default="Output PII", frozen=True)]
+    tasks: Annotated[list[TaskType],
+                     Field(title="Tasks",
+                           description="The list of supported tasks.",
+                           default=TaskType.values(), frozen=True)]
+    thresholds: Annotated[list[MetricThreshold],
+                          Field(title="Thresholds",
+                                description="The metric thresholds.",
+                                default=[MetricThreshold(type="upper_limit", value=0.1)])]
+    group: Annotated[MetricGroup,
+                     Field(title="Group",
+                           description="The metric group.",
+                           default=MetricGroup.CONTENT_SAFETY, frozen=True)]
+
+    async def evaluate_async(
+            self,
+            data: pd.DataFrame | dict,
+            configuration: GenAIConfiguration,
+            **kwargs
+    ) -> AggregateMetricResult:
+
+        # Create a modified configuration that uses output_fields as input_fields
+        # This allows DetectorsProvider to process output data
+        modified_config = configuration.model_copy(deep=True)
+        modified_config.input_fields = configuration.output_fields
+
+        provider = DetectorsProvider(configuration=modified_config,
+                                     metric_name="pii",
+                                     metric_display_name=self.display_name,
+                                     metric_method=self.method,
+                                     metric_group=self.group,
+                                     thresholds=self.thresholds,
+                                     **kwargs)
+        aggregated_metric_result = await provider.evaluate_async(data=data)
+        # Update the metric name to output_pii
+        aggregated_metric_result.name = self.name
+        for record in aggregated_metric_result.record_level_metrics:
+            record.name = self.name
+        return aggregated_metric_result
+
+    def evaluate(
+        self,
+        data: pd.DataFrame | dict,
+        configuration: GenAIConfiguration,
+        **kwargs,
+    ):
+        # If ran in sync mode, block until it is done
+        return run_in_event_loop(
+            self.evaluate_async,
+            data=data,
+            configuration=configuration,
+            **kwargs,
+        )
+
+# Made with Bob