nemo-platform-plugin 0.1.0__py3-none-any.whl

nemo_platform/types/evaluation/response_relevancy_metric_param_param.py

@@ -0,0 +1,75 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from __future__ import annotations
+
+from typing import Dict, List, Union
+from typing_extensions import Literal, Required, TypeAlias, TypedDict
+
+from .model_ref import ModelRef
+from .model_param import ModelParam
+from .inference_params_param import InferenceParamsParam
+
+__all__ = ["ResponseRelevancyMetricParamParam", "EmbeddingsModel", "JudgeModel"]
+
+EmbeddingsModel: TypeAlias = Union[ModelParam, ModelRef]
+
+JudgeModel: TypeAlias = Union[ModelParam, ModelRef]
+
+
+class ResponseRelevancyMetricParamParam(TypedDict, total=False):
+    """Request type for ResponseRelevancy metrics."""
+
+    embeddings_model: Required[EmbeddingsModel]
+    """The embeddings model configuration."""
+
+    judge_model: Required[JudgeModel]
+    """The judge model configuration."""
+
+    description: str
+    """Human-readable description of the metric."""
+
+    ignore_request_failure: bool
+    """
+    If True, request failures to the judge model are ignored and the metric result
+    is marked as NaN. Parse/output formatting failures are always converted to NaN.
+    """
+
+    inference: InferenceParamsParam
+    """Parameters for model inference.
+
+    Extra fields can be supplied for additional options applied to the inference
+    request directly. Fields not supported by the model may cause inference errors
+    during evaluation.
+    """
+
+    input_template: Dict[str, object]
+    """Optional Jinja template for rendering the input payload for RAGAS evaluation."""
+
+    labels: Dict[str, str]
+    """Labels are key-value pairs that can be used for grouping and filtering."""
+
+    strictness: int
+    """Number of parallel questions generated. NIM can only generate 1."""
+
+    supported_job_types: List[Literal["online", "offline"]]
+    """
+    A metric can evaluate model outputs for online evaluations or pre-generated
+    outputs for offline evaluations.
+    """
+
+    type: Literal["response_relevancy"]

nemo_platform/types/evaluation/response_relevancy_metric_response.py

@@ -0,0 +1,93 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import Dict, List, Union, Optional
+from datetime import datetime
+from typing_extensions import Literal, TypeAlias
+
+from .model import Model
+from ..._models import BaseModel
+from .model_ref import ModelRef
+from .inference_params import InferenceParams
+
+__all__ = ["ResponseRelevancyMetricResponse", "EmbeddingsModel", "JudgeModel"]
+
+EmbeddingsModel: TypeAlias = Union[Model, ModelRef]
+
+JudgeModel: TypeAlias = Union[Model, ModelRef]
+
+
+class ResponseRelevancyMetricResponse(BaseModel):
+    """Response type for ResponseRelevancy metrics."""
+
+    embeddings_model: EmbeddingsModel
+    """The embeddings model configuration."""
+
+    judge_model: JudgeModel
+    """The judge model configuration."""
+
+    id: Optional[str] = None
+    """Entity name within the workspace"""
+
+    created_at: Optional[datetime] = None
+
+    description: Optional[str] = None
+    """Human-readable description of the metric."""
+
+    ignore_request_failure: Optional[bool] = None
+    """
+    If True, request failures to the judge model are ignored and the metric result
+    is marked as NaN. Parse/output formatting failures are always converted to NaN.
+    """
+
+    inference: Optional[InferenceParams] = None
+    """Parameters for model inference.
+
+    Extra fields can be supplied for additional options applied to the inference
+    request directly. Fields not supported by the model may cause inference errors
+    during evaluation.
+    """
+
+    input_template: Optional[Dict[str, object]] = None
+    """Optional Jinja template for rendering the input payload for RAGAS evaluation."""
+
+    labels: Optional[Dict[str, str]] = None
+    """Labels are key-value pairs that can be used for grouping and filtering."""
+
+    name: Optional[str] = None
+    """Entity name within the workspace"""
+
+    parent: Optional[str] = None
+
+    project: Optional[str] = None
+    """The name of the project associated with this entity."""
+
+    strictness: Optional[int] = None
+    """Number of parallel questions generated. NIM can only generate 1."""
+
+    supported_job_types: Optional[List[Literal["online", "offline"]]] = None
+    """
+    A metric can evaluate model outputs for online evaluations or pre-generated
+    outputs for offline evaluations.
+    """
+
+    type: Optional[Literal["response_relevancy"]] = None
+
+    updated_at: Optional[datetime] = None
+
+    workspace: Optional[str] = None
+    """Workspace identifier"""

nemo_platform/types/evaluation/retriever_pipeline.py

@@ -0,0 +1,34 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import Union
+from typing_extensions import TypeAlias
+
+from .model import Model
+from ..._models import BaseModel
+from .model_ref import ModelRef
+
+__all__ = ["RetrieverPipeline", "EmbeddingsModel"]
+
+EmbeddingsModel: TypeAlias = Union[Model, ModelRef]
+
+
+class RetrieverPipeline(BaseModel):
+    """Pipeline configuration for retriever-based evaluations."""
+
+    embeddings_model: EmbeddingsModel
+    """The embeddings model configuration."""

nemo_platform/types/evaluation/retriever_pipeline_param.py

@@ -0,0 +1,35 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from __future__ import annotations
+
+from typing import Union
+from typing_extensions import Required, TypeAlias, TypedDict
+
+from .model_ref import ModelRef
+from .model_param import ModelParam
+
+__all__ = ["RetrieverPipelineParam", "EmbeddingsModel"]
+
+EmbeddingsModel: TypeAlias = Union[ModelParam, ModelRef]
+
+
+class RetrieverPipelineParam(TypedDict, total=False):
+    """Pipeline configuration for retriever-based evaluations."""
+
+    embeddings_model: Required[EmbeddingsModel]
+    """The embeddings model configuration."""

nemo_platform/types/evaluation/rouge_metric.py

@@ -0,0 +1,76 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import Dict, List, Optional
+from datetime import datetime
+from typing_extensions import Literal
+
+from ..._models import BaseModel
+
+__all__ = ["RougeMetric"]
+
+
+class RougeMetric(BaseModel):
+    """Persisted ROUGE metric."""
+
+    id: str
+
+    created_at: datetime
+
+    created_by: Optional[str] = None
+
+    entity_id: str
+    """Alias for id for backwards compatibility."""
+
+    parent: str
+    """Parent entity ID for nested entities."""
+
+    reference: str
+    """The template for the ground truth reference to evaluate the ROUGE metric with."""
+
+    updated_at: datetime
+
+    updated_by: Optional[str] = None
+
+    workspace: str
+    """Workspace identifier"""
+
+    candidate: Optional[str] = None
+    """The template for the candidate to evaluate the ROUGE metric on.
+
+    If not provided, the output text from the model is used.
+    """
+
+    description: Optional[str] = None
+    """Human-readable description of the metric."""
+
+    labels: Optional[Dict[str, str]] = None
+    """Labels are key-value pairs that can be used for grouping and filtering."""
+
+    name: Optional[str] = None
+    """Entity name within the workspace"""
+
+    project: Optional[str] = None
+    """The name of the project associated with this entity."""
+
+    supported_job_types: Optional[List[Literal["online", "offline"]]] = None
+    """
+    A metric can evaluate model outputs for online evaluations or pre-generated
+    outputs for offline evaluations.
+    """
+
+    type: Optional[Literal["rouge"]] = None

nemo_platform/types/evaluation/rouge_metric_param.py

@@ -0,0 +1,53 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import Dict, List, Optional
+from typing_extensions import Literal
+
+from ..._models import BaseModel
+
+__all__ = ["RougeMetricParam"]
+
+
+class RougeMetricParam(BaseModel):
+    """Request type for ROUGEMetric.
+
+    ROUGE metric for overlap-based summarization quality scoring.
+    """
+
+    reference: str
+    """The template for the ground truth reference to evaluate the ROUGE metric with."""
+
+    candidate: Optional[str] = None
+    """The template for the candidate to evaluate the ROUGE metric on.
+
+    If not provided, the output text from the model is used.
+    """
+
+    description: Optional[str] = None
+    """Human-readable description of the metric."""
+
+    labels: Optional[Dict[str, str]] = None
+    """Labels are key-value pairs that can be used for grouping and filtering."""
+
+    supported_job_types: Optional[List[Literal["online", "offline"]]] = None
+    """
+    A metric can evaluate model outputs for online evaluations or pre-generated
+    outputs for offline evaluations.
+    """
+
+    type: Optional[Literal["rouge"]] = None

nemo_platform/types/evaluation/rouge_metric_param_param.py

@@ -0,0 +1,53 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from __future__ import annotations
+
+from typing import Dict, List
+from typing_extensions import Literal, Required, TypedDict
+
+__all__ = ["RougeMetricParamParam"]
+
+
+class RougeMetricParamParam(TypedDict, total=False):
+    """Request type for ROUGEMetric.
+
+    ROUGE metric for overlap-based summarization quality scoring.
+    """
+
+    reference: Required[str]
+    """The template for the ground truth reference to evaluate the ROUGE metric with."""
+
+    candidate: str
+    """The template for the candidate to evaluate the ROUGE metric on.
+
+    If not provided, the output text from the model is used.
+    """
+
+    description: str
+    """Human-readable description of the metric."""
+
+    labels: Dict[str, str]
+    """Labels are key-value pairs that can be used for grouping and filtering."""
+
+    supported_job_types: List[Literal["online", "offline"]]
+    """
+    A metric can evaluate model outputs for online evaluations or pre-generated
+    outputs for offline evaluations.
+    """
+
+    type: Literal["rouge"]

nemo_platform/types/evaluation/rouge_metric_response.py

@@ -0,0 +1,69 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import Dict, List, Optional
+from datetime import datetime
+from typing_extensions import Literal
+
+from ..._models import BaseModel
+
+__all__ = ["RougeMetricResponse"]
+
+
+class RougeMetricResponse(BaseModel):
+    """Response type for ROUGEMetric."""
+
+    reference: str
+    """The template for the ground truth reference to evaluate the ROUGE metric with."""
+
+    id: Optional[str] = None
+    """Entity name within the workspace"""
+
+    candidate: Optional[str] = None
+    """The template for the candidate to evaluate the ROUGE metric on.
+
+    If not provided, the output text from the model is used.
+    """
+
+    created_at: Optional[datetime] = None
+
+    description: Optional[str] = None
+    """Human-readable description of the metric."""
+
+    labels: Optional[Dict[str, str]] = None
+    """Labels are key-value pairs that can be used for grouping and filtering."""
+
+    name: Optional[str] = None
+    """Entity name within the workspace"""
+
+    parent: Optional[str] = None
+
+    project: Optional[str] = None
+    """The name of the project associated with this entity."""
+
+    supported_job_types: Optional[List[Literal["online", "offline"]]] = None
+    """
+    A metric can evaluate model outputs for online evaluations or pre-generated
+    outputs for offline evaluations.
+    """
+
+    type: Optional[Literal["rouge"]] = None
+
+    updated_at: Optional[datetime] = None
+
+    workspace: Optional[str] = None
+    """Workspace identifier"""

nemo_platform/types/evaluation/row_score.py

@@ -0,0 +1,59 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import TYPE_CHECKING, Dict, List, Optional
+
+from pydantic import Field as FieldInfo
+
+from ..._models import BaseModel
+from .metric_output import MetricOutput
+
+__all__ = ["RowScore"]
+
+
+class RowScore(BaseModel):
+    """Normalized row-level score payload for metric/benchmark job results."""
+
+    item: Dict[str, object]
+    """Input item metadata for the evaluated row."""
+
+    metrics: Dict[str, List[MetricOutput]]
+    """Metric-level row outputs by metric key."""
+
+    requests: List[Dict[str, object]]
+    """Request details captured during evaluation."""
+
+    sample: Dict[str, object]
+    """Sample output payload for the evaluated row."""
+
+    metric_errors: Optional[Dict[str, str]] = None
+    """Full row-level error text keyed by metric for summary rendering."""
+
+    row_index: Optional[int] = None
+    """Stable row position used for result alignment."""
+
+    if TYPE_CHECKING:
+        # Some versions of Pydantic <2.8.0 have a bug and don’t allow assigning a
+        # value to this field, so for compatibility we avoid doing it at runtime.
+        __pydantic_extra__: Dict[str, object] = FieldInfo(init=False)  # pyright: ignore[reportIncompatibleVariableOverride]
+
+        # Stub to indicate that arbitrary properties are accepted.
+        # To access properties that are not valid identifiers you can use `getattr`, e.g.
+        # `getattr(obj, '$type')`
+        def __getattr__(self, attr: str) -> object: ...
+    else:
+        __pydantic_extra__: Dict[str, object]

nemo_platform/types/evaluation/rubric.py

@@ -0,0 +1,40 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import Optional
+
+from ..._models import BaseModel
+
+__all__ = ["Rubric"]
+
+
+class Rubric(BaseModel):
+    label: str
+    """The label to use for the level of the rubric grading criteria.
+
+    (e.g., "helpful", "not_helpful", "positive")
+    """
+
+    value: float
+    """The score value to assign for the criteria used for aggregation and ranking."""
+
+    description: Optional[str] = None
+    """Describe the semantic meaning of each criteria for the given rubric.
+
+    If no judge template is set, the input description for labels are included in
+    the generated judge prompt.
+    """

nemo_platform/types/evaluation/rubric_param.py

@@ -0,0 +1,40 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from __future__ import annotations
+
+from typing_extensions import Required, TypedDict
+
+__all__ = ["RubricParam"]
+
+
+class RubricParam(TypedDict, total=False):
+    label: Required[str]
+    """The label to use for the level of the rubric grading criteria.
+
+    (e.g., "helpful", "not_helpful", "positive")
+    """
+
+    value: Required[float]
+    """The score value to assign for the criteria used for aggregation and ranking."""
+
+    description: str
+    """Describe the semantic meaning of each criteria for the given rubric.
+
+    If no judge template is set, the input description for labels are included in
+    the generated judge prompt.
+    """