gp3mlpy 0.1.0__py3-none-any.whl

gp3mlpy/__init__.py

@@ -0,0 +1,101 @@
+# ruff: noqa: F401
+"""gp3mlpy: Python port of gp3ml 0.3.0."""
+from .exceptions import GP3MLError, OptionalDependencyError
+from .task_governance import (
+    gp3ml_prohibited_uses, declare_gazepoint_task, assert_gp3ml_use_case, validate_gazepoint_ml_roles,
+)
+from .feature_provenance import (
+    create_gazepoint_feature_manifest, validate_gazepoint_feature_manifest, write_gazepoint_feature_manifest_csv,
+)
+__version__ = "0.1.0"
+r_reference_version = "0.3.0"
+
+from .leakage import audit_gazepoint_ml_leakage, write_gazepoint_ml_leakage_audit_csv
+
+from .splitting import split_gazepoint_ml_data, validate_gazepoint_ml_split, write_gazepoint_ml_split_csv
+
+from .resampling import create_gazepoint_group_folds, audit_gazepoint_group_folds, validate_gazepoint_group_folds, write_gazepoint_group_folds_csv
+
+from .metrics import (
+    gazepoint_classification_metrics,
+    gazepoint_regression_metrics,
+    gazepoint_performance_metrics,
+    bootstrap_gazepoint_metrics,
+)
+from .preprocessing import fit_gazepoint_preprocessor, bake_gazepoint_preprocessor
+from .calibration import fit_gazepoint_calibrator, apply_gazepoint_calibrator, assess_gazepoint_calibration
+from .model_engines import (
+    gp3ml_available_engines,
+    integrate_black_box_model,
+    fit_gazepoint_model,
+    train_gazepoint_classifier,
+)
+from .deep_learning import fit_gazepoint_deep_model
+from .engine_capabilities import gp3ml_engine_capabilities, assert_gp3ml_engine_available
+from .synthetic import simulate_gazepoint_governed_data, create_gazepoint_synthetic_manifest, create_gazepoint_synthetic_task
+from .target_uncertainty import bootstrap_gazepoint_metrics_by_unit, summarize_gazepoint_resample_uncertainty, validate_gazepoint_target_uncertainty, write_gazepoint_target_uncertainty
+from .resample_evaluation import evaluate_gazepoint_group_folds, collect_gazepoint_fold_predictions, summarize_gazepoint_resample_performance, validate_gazepoint_resample_evaluation, write_gazepoint_resample_evaluation
+from .governance_reports import create_gazepoint_model_card, write_gazepoint_model_card, evaluate_external_validation, create_external_validation_report, write_external_validation_report, create_gazepoint_reproducibility_report, write_gazepoint_reproducibility_report
+from .external_validation import (
+    declare_gazepoint_external_dataset,
+    evaluate_gazepoint_external_transportability,
+    validate_gazepoint_transportability,
+    write_gazepoint_transportability_report,
+)
+from .resampling_diagnostics import diagnose_gazepoint_group_folds, validate_gazepoint_fold_diagnostics, write_gazepoint_fold_diagnostics_csv
+from .model_tuning import create_gazepoint_tuning_grid, tune_gazepoint_model, compare_gazepoint_models, select_gazepoint_model, validate_gazepoint_model_tuning, write_gazepoint_model_tuning
+from .nested_resampling import create_gazepoint_nested_folds, audit_gazepoint_nested_resampling, validate_gazepoint_nested_folds, evaluate_gazepoint_nested_resampling, validate_gazepoint_nested_evaluation, write_gazepoint_nested_evaluation
+
+from .roadmap_reporting import (
+    create_gazepoint_release_evidence,
+    create_gazepoint_release_model_card,
+    write_gazepoint_release_model_card,
+)
+
+from .decision_governance import (
+    apply_gazepoint_decision_rule,
+    audit_gazepoint_abstention,
+    create_gazepoint_decision_rule,
+    evaluate_gazepoint_thresholds,
+    select_gazepoint_threshold,
+    validate_gazepoint_decision_rule,
+)
+
+from .conformal import (
+    assess_gazepoint_conformal_coverage,
+    fit_gazepoint_conformal,
+    predict_gazepoint_interval,
+    predict_gazepoint_set,
+    validate_gazepoint_conformal,
+)
+
+from .dataset_shift import (
+    audit_gazepoint_dataset_shift,
+    audit_gazepoint_missingness_shift,
+    summarize_gazepoint_shift,
+)
+
+from .analysis_plan import (audit_gazepoint_plan_deviations, declare_gazepoint_analysis_plan, lock_gazepoint_analysis_plan, validate_gazepoint_analysis_plan, write_gazepoint_analysis_plan)
+
+from .environment import capture_gazepoint_environment, compare_gazepoint_environments, validate_gazepoint_environment
+
+from .reproducibility import audit_gazepoint_reproducibility, normalize_gazepoint_artifact_text, with_gazepoint_reproducible_output, write_gazepoint_reproducibility_audit
+
+from .release_provenance import validate_gazepoint_release_checksums, write_gazepoint_release_checksums
+
+from .governance_profiles import audit_gp3ml_governance_profile, create_gp3ml_governance_profile, write_gp3ml_governance_profile
+
+from .api_contracts import gp3ml_api_contracts, gp3ml_object_schema, validate_gp3ml_object_contract, audit_gp3ml_api_stability, write_gp3ml_api_contracts
+from .interoperability import gp3ml_interop_contracts, create_gazepoint_handoff, validate_gazepoint_handoff, combine_gazepoint_handoffs, as_gp3ml_data
+from .research_workflow import simulate_gazepoint_research_handoffs, validate_gazepoint_research_bundle
+from .model_artifacts import create_gazepoint_model_artifact, restore_gazepoint_model_artifact, validate_gazepoint_model_artifact, test_gazepoint_model_portability
+from .robustness import evaluate_gazepoint_seed_stability, evaluate_gazepoint_feature_stability, evaluate_gazepoint_threshold_stability, evaluate_gazepoint_missingness_sensitivity, audit_gazepoint_model_robustness
+from .ro_crate import write_gazepoint_ro_crate, validate_gazepoint_ro_crate
+from . import plotting as _plotting
+
+# Attach source-derived gp3ml 0.3.0 documentation to every compatibility export.
+from ._reference_docs import REFERENCE_DOCS as _REFERENCE_DOCS
+for _doc_name, _doc_text in _REFERENCE_DOCS.items():
+    _doc_object = globals().get(_doc_name)
+    if callable(_doc_object):
+        _doc_object.__doc__ = _doc_text

gp3mlpy/__init__.pyi

@@ -0,0 +1,207 @@
+from __future__ import annotations
+
+from pathlib import Path
+from collections.abc import Callable, Mapping, Sequence
+from typing import Any, TypeVar
+import numpy as np
+import pandas as pd
+
+T = TypeVar("T")
+
+_GP3MLAPIContractRegistry = Any
+_GP3MLAPIStabilityAudit = Any
+_GP3MLAbstentionAudit = Any
+_GP3MLAnalysisPlan = Any
+_GP3MLAnalysisPlanValidation = Any
+_GP3MLCalibrationAssessment = Any
+_GP3MLCalibrator = Any
+_GP3MLConformal = Any
+_GP3MLConformalCoverage = Any
+_GP3MLConformalValidation = Any
+_GP3MLDatasetShiftAudit = Any
+_GP3MLDecisionRule = Any
+_GP3MLDecisionRuleValidation = Any
+_GP3MLEngine = Any
+_GP3MLEnvironment = Any
+_GP3MLEnvironmentComparison = Any
+_GP3MLExternalDatasetDeclaration = Any
+_GP3MLGovernanceProfile = Any
+_GP3MLGovernanceProfileAudit = Any
+_GP3MLHandoff = Any
+_GP3MLHandoffBundle = Any
+_GP3MLHandoffValidation = Any
+_GP3MLMetricUncertainty = Any
+_GP3MLMissingnessShiftAudit = Any
+_GP3MLModel = Any
+_GP3MLModelArtifact = Any
+_GP3MLModelArtifactValidation = Any
+_GP3MLModelPortability = Any
+_GP3MLModelSelection = Any
+_GP3MLModelTuning = Any
+_GP3MLModelTuningValidation = Any
+_GP3MLNestedEvaluation = Any
+_GP3MLNestedEvaluationValidation = Any
+_GP3MLNestedFolds = Any
+_GP3MLNestedFoldsValidation = Any
+_GP3MLNestedResamplingAudit = Any
+_GP3MLObjectContractValidation = Any
+_GP3MLPlanDeviationAudit = Any
+_GP3MLPreprocessor = Any
+_GP3MLReleaseChecksumManifest = Any
+_GP3MLReleaseChecksumValidation = Any
+_GP3MLReleaseEvidence = Any
+_GP3MLReleaseModelCard = Any
+_GP3MLReproducibilityAudit = Any
+_GP3MLResampleEvaluation = Any
+_GP3MLResampleUncertainty = Any
+_GP3MLResearchBundle = Any
+_GP3MLResearchBundleValidation = Any
+_GP3MLRoleValidation = Any
+_GP3MLShiftSummary = Any
+_GP3MLTargetUncertainty = Any
+_GP3MLTask = Any
+_GP3MLThresholdEvaluation = Any
+_GP3MLTransportabilityReport = Any
+_GP3MLTransportabilityValidation = Any
+_GP3MLTuningGrid = Any
+_GazepointFeatureManifestValidation = Any
+_GazepointFoldDiagnostics = Any
+_GazepointFoldDiagnosticsValidation = Any
+_GazepointGroupFolds = Any
+_GazepointGroupFoldsAudit = Any
+_GazepointGroupFoldsValidation = Any
+_GazepointMLLeakageAudit = Any
+_GazepointMLSplit = Any
+_GazepointMLSplitValidation = Any
+
+
+r_reference_version: str
+__version__: str
+
+def apply_gazepoint_calibrator(calibrator: '_GP3MLCalibrator', probability: 'Sequence[float] | np.ndarray') -> 'np.ndarray': ...
+def apply_gazepoint_decision_rule(rule: '_GP3MLDecisionRule', probability: 'Sequence[float]', positive: 'Any', negative: 'Any', abstain_label: 'str' = ...) -> 'pd.Categorical': ...
+def as_gp3ml_data(x: 'Any', *args: 'Any', **kwargs: 'Any') -> 'pd.DataFrame': ...
+def assert_gp3ml_engine_available(engine: 'str', check_keras_backend: 'bool' = ...) -> 'bool': ...
+def assert_gp3ml_use_case(task: '_GP3MLTask', data: 'pd.DataFrame | None' = ...) -> 'bool': ...
+def assess_gazepoint_calibration(truth: 'Sequence[object] | pd.Series', probability: 'Sequence[float] | np.ndarray', positive: 'str | None' = ..., bins: 'int' = ..., bootstrap: 'int' = ..., conf_level: 'float' = ..., seed: 'int' = ...) -> '_GP3MLCalibrationAssessment': ...
+def assess_gazepoint_conformal_coverage(object: '_GP3MLConformal', truth: 'Sequence[Any]', interval: 'pd.DataFrame | None' = ..., set: 'pd.DataFrame | None' = ..., unit: 'Sequence[Any] | None' = ...) -> '_GP3MLConformalCoverage': ...
+def audit_gazepoint_abstention(truth: 'Sequence[Any]', decision: 'Sequence[Any]', abstain_label: 'str' = ...) -> '_GP3MLAbstentionAudit': ...
+def audit_gazepoint_dataset_shift(development: 'pd.DataFrame', external: 'pd.DataFrame', predictors: 'Sequence[str] | None' = ..., thresholds: 'Mapping[str, float] | None' = ...) -> '_GP3MLDatasetShiftAudit': ...
+def audit_gazepoint_group_folds(x: '_GazepointGroupFolds') -> '_GazepointGroupFoldsAudit': ...
+def audit_gazepoint_missingness_shift(development: 'pd.DataFrame', external: 'pd.DataFrame', predictors: 'Sequence[str] | None' = ..., review_delta: 'float' = ..., fail_delta: 'float' = ...) -> '_GP3MLMissingnessShiftAudit': ...
+def audit_gazepoint_ml_leakage(analysis: 'pd.DataFrame', assessment: 'pd.DataFrame', outcome: 'str', predictors: 'list[str] | tuple[str, ...] | str', participant_id: 'str | None' = ..., trial_id: 'str | None' = ..., stimulus_id: 'str | None' = ..., generalization_target: 'str' = ..., target_derived: 'list[str] | tuple[str, ...] | str | None' = ..., post_outcome: 'list[str] | tuple[str, ...] | str | None' = ...) -> '_GazepointMLLeakageAudit': ...
+def audit_gazepoint_model_robustness(seed_stability: 'Any' = ..., feature_stability: 'Any' = ..., threshold_stability: 'Any' = ..., missingness_stability: 'Any' = ..., relative_sd_review: 'Any' = ..., relative_sd_fail: 'Any' = ...) -> 'Any': ...
+def audit_gazepoint_nested_resampling(x: '_GP3MLNestedFolds') -> '_GP3MLNestedResamplingAudit': ...
+def audit_gazepoint_plan_deviations(plan: '_GP3MLAnalysisPlan', actual: 'Mapping[str, Any]', fields: 'Sequence[str]' = ...) -> '_GP3MLPlanDeviationAudit': ...
+def audit_gazepoint_reproducibility(paths: 'Sequence[str | Path] | str | Path', recursive: 'bool' = ..., extensions: 'Sequence[str]' = ...) -> '_GP3MLReproducibilityAudit': ...
+def audit_gp3ml_api_stability(registry: '_GP3MLAPIContractRegistry | None' = ...) -> '_GP3MLAPIStabilityAudit': ...
+def audit_gp3ml_governance_profile(profile: '_GP3MLGovernanceProfile') -> '_GP3MLGovernanceProfileAudit': ...
+def bake_gazepoint_preprocessor(preprocessor: '_GP3MLPreprocessor', new_data: 'pd.DataFrame') -> 'np.ndarray': ...
+def bootstrap_gazepoint_metrics(task: '_GP3MLTask', truth: 'Sequence[Any] | pd.Series', prediction: 'Sequence[Any] | pd.Series | None' = ..., probability: 'Sequence[float] | np.ndarray | None' = ..., threshold: 'float' = ..., bootstrap: 'int' = ..., conf_level: 'float' = ..., seed: 'int' = ...) -> '_GP3MLMetricUncertainty': ...
+def bootstrap_gazepoint_metrics_by_unit(task: 'Any', truth: 'Any', prediction: 'Any' = ..., probability: 'Any' = ..., participant_id: 'Any' = ..., stimulus_id: 'Any' = ..., unit: 'str' = ..., bootstrap: 'int' = ..., conf_level: 'float' = ..., seed: 'int' = ..., threshold: 'float' = ..., stratify_observations: 'bool' = ...) -> 'Any': ...
+def capture_gazepoint_environment(packages: 'Sequence[str] | str | None' = ..., root: 'str | Path' = ..., include_renv: 'bool' = ...) -> '_GP3MLEnvironment': ...
+def collect_gazepoint_fold_predictions(x: 'Any', include_failed: 'Any' = ...) -> 'Any': ...
+def combine_gazepoint_handoffs(handoffs: 'Any', keys: 'Any' = ..., collision: 'str' = ...) -> '_GP3MLHandoffBundle': ...
+def compare_gazepoint_environments(reference: '_GP3MLEnvironment', current: '_GP3MLEnvironment') -> '_GP3MLEnvironmentComparison': ...
+def compare_gazepoint_models(x: '_GP3MLModelTuning', metrics: 'Any' = ...) -> 'pd.DataFrame': ...
+def create_external_validation_report(validation: 'Any', development_metrics: 'Any' = ..., limitations: 'Any' = ...) -> 'Any': ...
+def create_gazepoint_decision_rule(metric: 'str', direction: 'str' = ..., threshold: 'float | None' = ..., threshold_origin: 'str' = ..., cost_false_positive: 'float' = ..., cost_false_negative: 'float' = ..., abstention_allowed: 'bool' = ..., abstention_interval: 'Sequence[float] | None' = ..., calibration_source: 'str' = ..., training_partition: 'str' = ..., generalization_target: 'str | None' = ..., scientific_justification: 'str | None' = ...) -> '_GP3MLDecisionRule': ...
+def create_gazepoint_feature_manifest(features: 'list[str] | tuple[str, ...] | str', scientific_source: 'Any' = ..., source_table: 'Any' = ..., transformation: 'Any' = ..., availability_stage: 'Any' = ..., prediction_time_available: 'Any' = ..., outcome_derived: 'Any' = ..., post_outcome: 'Any' = ..., identifier: 'Any' = ..., preprocessing_scope: 'Any' = ..., fold_local_required: 'Any' = ..., reviewer_notes: 'Any' = ...) -> 'pd.DataFrame': ...
+def create_gazepoint_group_folds(data: 'pd.DataFrame', outcome: 'str', predictors: 'list[str] | tuple[str, ...] | str', feature_manifest: 'pd.DataFrame', generalization_target: 'str', participant_id: 'str | None' = ..., trial_id: 'str | None' = ..., stimulus_id: 'str | None' = ..., v: 'Any' = ..., repeats: 'int' = ..., seed: 'int' = ..., source_row_id: 'str' = ...) -> '_GazepointGroupFolds': ...
+def create_gazepoint_handoff(data: 'pd.DataFrame', source_package: 'str', source_version: 'str | None' = ..., producer: 'str | None' = ..., keys: 'Any' = ..., outcome: 'str | None' = ..., predictors: 'Any' = ..., feature_manifest: 'Any' = ..., notes: 'Any' = ...) -> '_GP3MLHandoff': ...
+def create_gazepoint_model_artifact(model: 'Any', preprocessor: 'Any' = ..., feature_manifest: 'Any' = ..., task: 'Any' = ..., decision_rule: 'Any' = ..., model_card: 'Any' = ..., reference_data: 'pd.DataFrame | None' = ..., bundle_model: 'bool' = ...) -> '_GP3MLModelArtifact': ...
+def create_gazepoint_model_card(model: '_GP3MLModel', intended_use: 'str', evaluation: 'Any' = ..., calibration: 'Any' = ..., feature_manifest: 'Any' = ..., external_validation: 'Any' = ..., limitations: 'Any' = ..., ethical_review: 'Any' = ...) -> 'Any': ...
+def create_gazepoint_nested_folds(outer_folds: '_GazepointGroupFolds', inner_v: 'int' = ..., inner_repeats: 'int' = ..., seed: 'int' = ..., continue_on_error: 'bool' = ...) -> '_GP3MLNestedFolds': ...
+def create_gazepoint_release_evidence(objects: 'Mapping[str, Any] | None' = ..., files: 'Mapping[str, str | Path] | None' = ..., version: 'str' = ..., notes: 'str | Sequence[str]' = ...) -> '_GP3MLReleaseEvidence': ...
+def create_gazepoint_release_model_card(model: '_GP3MLModel', intended_use: 'str', evaluation: 'Any' = ..., selection: '_GP3MLModelSelection | None' = ..., uncertainty: '_GP3MLTargetUncertainty | _GP3MLResampleUncertainty | None' = ..., calibration: 'Any' = ..., feature_manifest: 'Any' = ..., transportability: '_GP3MLTransportabilityReport | None' = ..., limitations: 'str | Sequence[str] | None' = ..., ethical_review: 'Any' = ..., deployment_status: 'str' = ...) -> '_GP3MLReleaseModelCard': ...
+def create_gazepoint_reproducibility_report(objects: 'Any' = ..., data: 'Any' = ..., seeds: 'Any' = ..., notes: 'Any' = ..., project_path: 'Any' = ...) -> 'Any': ...
+def create_gazepoint_synthetic_manifest(outcome: 'str', predictors: 'Sequence[str]', participant_id: 'str' = ..., stimulus_id: 'str' = ..., trial_id: 'str' = ...) -> 'pd.DataFrame': ...
+def create_gazepoint_synthetic_task(data: 'pd.DataFrame', workflow: 'str' = ..., generalization_target: 'str' = ...) -> 'Any': ...
+def create_gazepoint_tuning_grid(engine: 'Any', engine_grid: 'dict[str, Any] | None' = ..., preprocessor_grid: 'dict[str, Any] | None' = ..., thresholds: 'Any' = ..., complexity: 'Any' = ..., interpretability: 'Any' = ..., labels: 'Any' = ...) -> '_GP3MLTuningGrid': ...
+def create_gp3ml_governance_profile(evidence: 'Mapping[str, object]', framework: 'str' = ...) -> '_GP3MLGovernanceProfile': ...
+def declare_gazepoint_analysis_plan(research_question: 'Any', scientific_purpose: 'Any', outcome: 'Any', outcome_definition: 'Any', predictors: 'Sequence[str]', generalization_target: 'Any', grouping_variables: 'Sequence[str] | str' = ..., eligible_population: 'Any' = ..., exclusion_rules: 'Sequence[str] | str' = ..., preprocessing_plan: 'Any' = ..., candidate_models: 'Any' = ..., primary_metric: 'Any' = ..., secondary_metrics: 'Sequence[str] | str' = ..., calibration_metric: 'Any' = ..., uncertainty_method: 'Any' = ..., threshold_policy: 'Any' = ..., external_validation_required: 'bool' = ..., seed_strategy: 'Any' = ..., prohibited_interpretations: 'Sequence[str] | str | None' = ...) -> '_GP3MLAnalysisPlan': ...
+def declare_gazepoint_external_dataset(data: 'pd.DataFrame', label: 'str', independent: 'bool', origin: 'str', collection_period: 'Any' = ..., participant_id: 'str | None' = ..., stimulus_id: 'str | None' = ..., notes: 'Any' = ...) -> '_GP3MLExternalDatasetDeclaration': ...
+def declare_gazepoint_task(data: 'pd.DataFrame', outcome: 'str', purpose: 'str', task_type: 'str' = ..., unit_id: 'str | None' = ..., participant_id: 'str | None' = ..., stimulus_id: 'str | None' = ..., generalization_target: 'str' = ..., positive: 'str | None' = ..., observed_outcome: 'bool' = ..., sensitive_outcome: 'bool' = ...) -> '_GP3MLTask': ...
+def diagnose_gazepoint_group_folds(x: '_GazepointGroupFolds', imbalance_review: 'float' = ..., imbalance_fail: 'float' = ...) -> '_GazepointFoldDiagnostics': ...
+def evaluate_external_validation(model: '_GP3MLModel', external_data: 'pd.DataFrame', label: 'Any' = ..., threshold: 'Any' = ..., bootstrap: 'Any' = ..., seed: 'Any' = ...) -> 'Any': ...
+def evaluate_gazepoint_external_transportability(model: '_GP3MLModel', development_data: 'pd.DataFrame', external_data: 'pd.DataFrame | None' = ..., declaration: '_GP3MLExternalDatasetDeclaration | None' = ..., development_evaluation: 'Any' = ..., threshold: 'float | None' = ..., bootstrap: 'int' = ..., seed: 'int' = ...) -> '_GP3MLTransportabilityReport': ...
+def evaluate_gazepoint_feature_stability(features: 'Any', evaluator: 'Any', **kwargs: 'Any') -> 'Any': ...
+def evaluate_gazepoint_group_folds(folds: '_GazepointGroupFolds', task: '_GP3MLTask', predictors: 'Sequence[str] | None' = ..., engine: 'Any' = ..., preprocessor_args: 'dict[str, Any] | None' = ..., engine_args: 'dict[str, Any] | None' = ..., threshold: 'float' = ..., seed: 'int' = ..., assess_calibration: 'bool' = ..., calibration_bins: 'int' = ..., calibration_bootstrap: 'int' = ..., keep_models: 'bool' = ..., continue_on_error: 'bool' = ...) -> '_GP3MLResampleEvaluation': ...
+def evaluate_gazepoint_missingness_sensitivity(scenarios: 'Any', evaluator: 'Any', **kwargs: 'Any') -> 'Any': ...
+def evaluate_gazepoint_nested_resampling(nested_folds: '_GP3MLNestedFolds', task: 'Any', tuning_grid: '_GP3MLTuningGrid', selection_metric: 'str', direction: 'str', predictors: 'Any' = ..., minimum_success_prop: 'float' = ..., tie_breakers: 'Any' = ..., selection_rationale: 'str' = ..., seed: 'int' = ..., keep_models: 'bool' = ..., continue_on_error: 'bool' = ...) -> '_GP3MLNestedEvaluation': ...
+def evaluate_gazepoint_seed_stability(seeds: 'Any', evaluator: 'Any', **kwargs: 'Any') -> 'Any': ...
+def evaluate_gazepoint_threshold_stability(evaluation: 'Any', metric: 'Any', direction: 'Any' = ..., tolerance: 'Any' = ...) -> 'Any': ...
+def evaluate_gazepoint_thresholds(truth: 'Sequence[Any]', probability: 'Sequence[float]', positive: 'Any', thresholds: 'Sequence[float] | None', cost_false_positive: 'float' = ..., cost_false_negative: 'float' = ...) -> '_GP3MLThresholdEvaluation': ...
+def fit_gazepoint_calibrator(truth: 'Sequence[object] | pd.Series', probability: 'Sequence[float] | np.ndarray', positive: 'str | None' = ..., method: 'str' = ...) -> '_GP3MLCalibrator': ...
+def fit_gazepoint_conformal(truth: 'Sequence[Any]', prediction: 'Sequence[float] | None' = ..., probability: 'Sequence[float] | None' = ..., task_type: 'str' = ..., positive: 'Any' = ..., level: 'float' = ..., calibration_unit: 'str' = ..., unit: 'Sequence[Any] | None' = ..., generalization_target: 'str | None' = ...) -> '_GP3MLConformal': ...
+def fit_gazepoint_deep_model(data: 'pd.DataFrame', task: '_GP3MLTask', predictors: 'Sequence[str] | None' = ..., preprocessor: '_GP3MLPreprocessor | None' = ..., hidden_units: 'Sequence[int]' = ..., dropout: 'float' = ..., epochs: 'int' = ..., batch_size: 'int' = ..., validation_split: 'float' = ..., optimizer: 'str | Any' = ..., seed: 'int' = ..., verbose: 'int' = ...) -> '_GP3MLModel': ...
+def fit_gazepoint_model(data: 'pd.DataFrame', task: '_GP3MLTask', predictors: 'Sequence[str] | None' = ..., engine: 'str | _GP3MLEngine | None' = ..., preprocessor: '_GP3MLPreprocessor | None' = ..., preprocessor_args: 'dict[str, Any] | None' = ..., engine_args: 'dict[str, Any] | None' = ..., seed: 'int' = ..., threshold: 'float' = ...) -> '_GP3MLModel': ...
+def fit_gazepoint_preprocessor(data: 'pd.DataFrame', predictors: 'Sequence[str]', numeric_imputation: 'str' = ..., center: 'bool' = ..., scale: 'bool' = ..., novel_level: 'str' = ..., remove_zero_variance: 'bool' = ...) -> '_GP3MLPreprocessor': ...
+def gazepoint_classification_metrics(truth: 'Sequence[Any] | pd.Series', probability: 'Sequence[float] | np.ndarray', predicted: 'Sequence[Any] | pd.Series | None' = ..., positive: 'str | None' = ..., threshold: 'float' = ...) -> 'pd.DataFrame': ...
+def gazepoint_performance_metrics(task: '_GP3MLTask', truth: 'Sequence[Any] | pd.Series', prediction: 'Sequence[Any] | pd.Series | None' = ..., probability: 'Sequence[float] | np.ndarray | None' = ..., threshold: 'float' = ...) -> 'pd.DataFrame': ...
+def gazepoint_regression_metrics(truth: 'Sequence[float] | np.ndarray', prediction: 'Sequence[float] | np.ndarray') -> 'pd.DataFrame': ...
+def gp3ml_api_contracts() -> '_GP3MLAPIContractRegistry': ...
+def gp3ml_available_engines() -> 'pd.DataFrame': ...
+def gp3ml_engine_capabilities(check_keras_backend: 'bool' = ...) -> 'pd.DataFrame': ...
+def gp3ml_interop_contracts() -> 'pd.DataFrame': ...
+def gp3ml_object_schema(x: 'Any', recursive: 'bool' = ...) -> 'pd.DataFrame': ...
+def gp3ml_prohibited_uses() -> 'list[str]': ...
+def integrate_black_box_model(name: 'str', fit_fun: 'Callable[..., Any]', predict_fun: 'Callable[..., Any]', supports: 'Sequence[str]' = ..., probability: 'bool' = ..., metadata: 'dict[str, Any] | None' = ..., safety_declaration: 'dict[str, bool] | None' = ...) -> '_GP3MLEngine': ...
+def lock_gazepoint_analysis_plan(plan: '_GP3MLAnalysisPlan', plan_id: 'str | None' = ..., locked_at: 'Any' = ...) -> '_GP3MLAnalysisPlan': ...
+def normalize_gazepoint_artifact_text(x: 'Sequence[str] | str', project_path: 'str | Path | None' = ...) -> 'Any': ...
+def predict_gazepoint_interval(object: '_GP3MLConformal', prediction: 'Sequence[float]') -> 'pd.DataFrame': ...
+def predict_gazepoint_set(object: '_GP3MLConformal', probability: 'Sequence[float]') -> 'pd.DataFrame': ...
+def restore_gazepoint_model_artifact(artifact: 'Any') -> 'Any': ...
+def select_gazepoint_model(x: '_GP3MLModelTuning', metric: 'str', direction: 'str', minimum_success_prop: 'float' = ..., tie_breakers: 'Any' = ..., rationale: 'str | None' = ...) -> '_GP3MLModelSelection': ...
+def select_gazepoint_threshold(evaluation: '_GP3MLThresholdEvaluation', metric: 'str', direction: 'str' = ..., threshold_origin: 'str' = ..., training_partition: 'str' = ..., generalization_target: 'str | None' = ..., scientific_justification: 'str | None' = ..., abstention_allowed: 'bool' = ..., abstention_interval: 'Sequence[float] | None' = ...) -> '_GP3MLDecisionRule': ...
+def simulate_gazepoint_governed_data(n_participants: 'int' = ..., n_stimuli: 'int' = ..., trials_per_cell: 'int' = ..., seed: 'int' = ...) -> 'pd.DataFrame': ...
+def simulate_gazepoint_research_handoffs(n_participants: 'int' = ..., n_stimuli: 'int' = ..., trials_per_stimulus: 'int' = ..., seed: 'int' = ...) -> '_GP3MLResearchBundle': ...
+def split_gazepoint_ml_data(data: 'pd.DataFrame', outcome: 'str', predictors: 'list[str] | tuple[str, ...] | str', feature_manifest: 'pd.DataFrame', generalization_target: 'str', participant_id: 'str | None' = ..., trial_id: 'str | None' = ..., stimulus_id: 'str | None' = ..., assessment_prop: 'float' = ..., seed: 'int' = ..., source_row_id: 'str' = ...) -> '_GazepointMLSplit': ...
+def summarize_gazepoint_resample_performance(x: 'Any', aggregation: 'Any' = ..., conf_level: 'Any' = ...) -> 'Any': ...
+def summarize_gazepoint_resample_uncertainty(evaluation: 'Any', unit: 'str' = ..., conf_level: 'float' = ...) -> 'Any': ...
+def summarize_gazepoint_shift(shift: '_GP3MLDatasetShiftAudit', missingness: '_GP3MLMissingnessShiftAudit | None' = ...) -> '_GP3MLShiftSummary': ...
+def test_gazepoint_model_portability(artifact: 'Any', newdata: 'Any' = ..., tolerance: 'float' = ..., fresh_process: 'bool' = ...) -> '_GP3MLModelPortability': ...
+def train_gazepoint_classifier(data: 'pd.DataFrame', task: '_GP3MLTask', predictors: 'Sequence[str] | None' = ..., engine: 'str | _GP3MLEngine' = ..., **kwargs: 'Any') -> '_GP3MLModel': ...
+def tune_gazepoint_model(folds: 'Any', task: 'Any', tuning_grid: '_GP3MLTuningGrid', predictors: 'Any' = ..., metrics: 'Any' = ..., seed: 'int' = ..., continue_on_error: 'bool' = ..., keep_evaluations: 'bool' = ...) -> '_GP3MLModelTuning': ...
+def validate_gazepoint_analysis_plan(plan: 'Any') -> '_GP3MLAnalysisPlanValidation': ...
+def validate_gazepoint_conformal(object: 'Any') -> '_GP3MLConformalValidation': ...
+def validate_gazepoint_decision_rule(rule: 'Any', require_threshold: 'bool' = ...) -> '_GP3MLDecisionRuleValidation': ...
+def validate_gazepoint_environment(reference: '_GP3MLEnvironment', root: 'str | Path' = ..., include_renv: 'bool' = ...) -> '_GP3MLEnvironmentComparison': ...
+def validate_gazepoint_feature_manifest(x: 'pd.DataFrame') -> '_GazepointFeatureManifestValidation': ...
+def validate_gazepoint_fold_diagnostics(x: '_GazepointFoldDiagnostics') -> '_GazepointFoldDiagnosticsValidation': ...
+def validate_gazepoint_group_folds(x: '_GazepointGroupFolds') -> '_GazepointGroupFoldsValidation': ...
+def validate_gazepoint_handoff(x: 'Any') -> '_GP3MLHandoffValidation': ...
+def validate_gazepoint_ml_roles(data: 'pd.DataFrame', task: '_GP3MLTask', predictors: 'list[str] | tuple[str, ...] | str', feature_manifest: 'pd.DataFrame | None' = ...) -> '_GP3MLRoleValidation': ...
+def validate_gazepoint_ml_split(x: '_GazepointMLSplit') -> '_GazepointMLSplitValidation': ...
+def validate_gazepoint_model_artifact(artifact: 'Any', verify_hash: 'bool' = ...) -> '_GP3MLModelArtifactValidation': ...
+def validate_gazepoint_model_tuning(x: '_GP3MLModelTuning') -> '_GP3MLModelTuningValidation': ...
+def validate_gazepoint_nested_evaluation(x: '_GP3MLNestedEvaluation') -> '_GP3MLNestedEvaluationValidation': ...
+def validate_gazepoint_nested_folds(x: '_GP3MLNestedFolds') -> '_GP3MLNestedFoldsValidation': ...
+def validate_gazepoint_release_checksums(manifest: 'Any', directory: 'str | Path' = ...) -> '_GP3MLReleaseChecksumValidation': ...
+def validate_gazepoint_resample_evaluation(x: 'Any') -> 'Any': ...
+def validate_gazepoint_research_bundle(x: 'Any') -> '_GP3MLResearchBundleValidation': ...
+def validate_gazepoint_ro_crate(path: 'Any') -> 'Any': ...
+def validate_gazepoint_target_uncertainty(x: 'Any') -> 'Any': ...
+def validate_gazepoint_transportability(x: '_GP3MLTransportabilityReport') -> '_GP3MLTransportabilityValidation': ...
+def validate_gp3ml_object_contract(x: 'Any', registry: '_GP3MLAPIContractRegistry | None' = ...) -> '_GP3MLObjectContractValidation': ...
+def with_gazepoint_reproducible_output(code: 'Callable[[], T] | T') -> 'T': ...
+def write_external_validation_report(report: 'Any', path: 'Any', overwrite: 'Any' = ...) -> 'Any': ...
+def write_gazepoint_analysis_plan(plan: '_GP3MLAnalysisPlan', path: 'str | Path', format: 'str' = ...) -> 'str': ...
+def write_gazepoint_feature_manifest_csv(x: 'Any', file: 'str | Path', table: 'str' = ..., overwrite: 'bool' = ..., na: 'str' = ...) -> 'str': ...
+def write_gazepoint_fold_diagnostics_csv(x: '_GazepointFoldDiagnostics', directory: 'str | Path', prefix: 'str' = ..., tables: 'Any' = ..., overwrite: 'bool' = ..., na: 'str' = ...) -> 'dict[str, str]': ...
+def write_gazepoint_group_folds_csv(x: '_GazepointGroupFolds', directory: 'str | Path', prefix: 'str' = ..., tables: 'Any' = ..., include_fold_data: 'bool' = ..., overwrite: 'bool' = ..., na: 'str' = ...) -> 'dict[str, str]': ...
+def write_gazepoint_ml_leakage_audit_csv(x: '_GazepointMLLeakageAudit', file: 'str | Path', table: 'str' = ..., overwrite: 'bool' = ..., na: 'str' = ...) -> 'str': ...
+def write_gazepoint_ml_split_csv(x: '_GazepointMLSplit', directory: 'str | Path', prefix: 'str' = ..., tables: 'list[str] | tuple[str, ...] | str' = ..., overwrite: 'bool' = ..., na: 'str' = ...) -> 'dict[str, str]': ...
+def write_gazepoint_model_card(card: 'Any', path: 'Any', format: 'Any' = ..., overwrite: 'Any' = ...) -> 'Any': ...
+def write_gazepoint_model_tuning(x: '_GP3MLModelTuning', directory: 'str', prefix: 'str' = ..., selection: '_GP3MLModelSelection | None' = ..., overwrite: 'bool' = ...) -> 'dict[str, str]': ...
+def write_gazepoint_nested_evaluation(x: '_GP3MLNestedEvaluation', directory: 'str', prefix: 'str' = ..., overwrite: 'bool' = ...) -> 'dict[str, str]': ...
+def write_gazepoint_release_checksums(files: 'Sequence[str | Path] | str | Path', path: 'str | Path' = ...) -> '_GP3MLReleaseChecksumManifest': ...
+def write_gazepoint_release_model_card(card: '_GP3MLReleaseModelCard', path: 'str | Path', format: 'str' = ..., overwrite: 'bool' = ...) -> 'str': ...
+def write_gazepoint_reproducibility_audit(audit: '_GP3MLReproducibilityAudit', directory: 'str | Path' = ..., prefix: 'str' = ..., overwrite: 'bool' = ...) -> 'Any': ...
+def write_gazepoint_reproducibility_report(report: 'Any', path: 'Any', overwrite: 'Any' = ...) -> 'Any': ...
+def write_gazepoint_resample_evaluation(x: 'Any', directory: 'Any', prefix: 'Any' = ..., overwrite: 'Any' = ...) -> 'Any': ...
+def write_gazepoint_ro_crate(path: 'Any', files: 'Any', name: 'Any', description: 'Any', creator_name: 'Any', creator_orcid: 'Any' = ..., license: 'Any' = ..., doi: 'Any' = ..., copy_files: 'Any' = ...) -> 'Any': ...
+def write_gazepoint_target_uncertainty(x: 'Any', directory: 'Any', prefix: 'Any' = ..., overwrite: 'Any' = ...) -> 'Any': ...
+def write_gazepoint_transportability_report(report: '_GP3MLTransportabilityReport', path: 'str | Path', overwrite: 'bool' = ...) -> 'str': ...
+def write_gp3ml_api_contracts(registry: '_GP3MLAPIContractRegistry | None' = ..., directory: 'str' = ..., prefix: 'str' = ..., overwrite: 'bool' = ...) -> 'dict[str, str]': ...
+def write_gp3ml_governance_profile(audit: '_GP3MLGovernanceProfileAudit', path: 'str | Path') -> 'str': ...

gp3mlpy/_reference_docs.py

@@ -0,0 +1,3 @@
+"""Source-derived public API documentation for the gp3ml 0.3.0 compatibility layer."""
+
+REFERENCE_DOCS = {'apply_gazepoint_calibrator': 'Apply a fitted probability calibrator\n\nApply a fitted probability calibrator\n\nParameters\n----------\ncalibrator\n    A fitted `gp3ml_calibrator object.\nprobability\n    Uncalibrated probabilities to transform.\n\nReturns\n-------\nA numeric vector of calibrated probabilities, clipped to the open unit interval.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'apply_gazepoint_decision_rule': 'Apply a governed classification decision rule\n\nApply a governed classification decision rule\n\nParameters\n----------\nrule\n    A validated decision rule.\nprobability\n    Positive-class probability.\npositive\n    Positive class label.\nnegative\n    Negative class label.\nabstain_label\n    Label used for abstentions.\n\nReturns\n-------\nA factor of governed decisions.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'as_gp3ml_data': 'Extract model-ready data from a gp3ml handoff object\n\nExtract model-ready data from a gp3ml handoff object\n\nParameters\n----------\nx\n    A `gp3ml_handoff or `gp3ml_handoff_bundle.\nargs\n    Additional reserved compatibility arguments.\nkwargs\n    Additional reserved compatibility arguments.\n\nReturns\n-------\nA data frame.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'assert_gp3ml_engine_available': 'Assert that a gp3ml engine is available\n\nAssert that a gp3ml engine is available\n\nParameters\n----------\nengine\n    Engine name.\ncheck_keras_backend\n    Whether to verify a configured Keras backend.\n\nReturns\n-------\nInvisibly `TRUE on success.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'assert_gp3ml_use_case': 'Assert that a task is within the permitted gp3ml scope\n\nAssert that a task is within the permitted gp3ml scope\n\nParameters\n----------\ntask\n    A `gp3ml_task object.\ndata\n    Optional data frame used to validate task columns.\n\nReturns\n-------\nInvisibly returns `TRUE when the task is permitted; otherwise, the function stops with an error.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'assess_gazepoint_calibration': 'Calibration assessment with bootstrap uncertainty\n\nCalibration assessment with bootstrap uncertainty\n\nParameters\n----------\ntruth\n    Observed binary outcome values.\nprobability\n    Predicted positive-class probabilities.\npositive\n    Label representing the positive class.\nbins\n    Number of reliability bins.\nbootstrap\n    Number of bootstrap replicates.\nconf_level\n    Confidence level for percentile intervals.\nseed\n    Deterministic random seed.\n\nReturns\n-------\nA `gp3ml_calibration_assessment object containing calibration summaries, reliability-bin results, bootstrap intervals, and assessment settings.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'assess_gazepoint_conformal_coverage': 'Assess conformal coverage\n\nAssess conformal coverage\n\nParameters\n----------\nobject\n    A `gp3ml_conformal_fit.\ntruth\n    Observed outcomes.\ninterval\n    Regression interval data frame.\nset\n    Classification set data frame.\nunit\n    Optional assessment-unit identifier.\n\nReturns\n-------\nA `gp3ml_conformal_coverage.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'audit_gazepoint_abstention': 'Audit abstention decisions\n\nAudit abstention decisions\n\nParameters\n----------\ntruth\n    Observed binary outcome.\ndecision\n    Decisions returned by `apply_gazepoint_decision_rule().\nabstain_label\n    Abstention label.\n\nReturns\n-------\nA `gp3ml_abstention_audit.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'audit_gazepoint_dataset_shift': 'Audit predictor distribution shift\n\nAudit predictor distribution shift\n\nParameters\n----------\ndevelopment\n    Development/training data.\nexternal\n    Independent or later data to compare.\npredictors\n    Predictors to audit. Defaults to common columns.\nthresholds\n    Named threshold list.\n\nReturns\n-------\nA `gp3ml_dataset_shift_audit.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'audit_gazepoint_group_folds': 'Aggregate leakage audits across group-aware folds\n\nAggregate leakage audits across group-aware folds\n\nParameters\n----------\nx\n    A `gazepoint_group_folds object.\n\nReturns\n-------\nAn object of class `gazepoint_group_folds_audit.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'audit_gazepoint_missingness_shift': 'Audit missingness shift\n\nAudit missingness shift\n\nParameters\n----------\ndevelopment\n    Development data.\nexternal\n    External/new data.\npredictors\n    Predictors to audit.\nreview_delta\n    Review threshold for absolute missingness change.\nfail_delta\n    Fail threshold for absolute missingness change.\n\nReturns\n-------\nA `gp3ml_missingness_shift_audit.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'audit_gazepoint_ml_leakage': 'Audit leakage between predictive-analysis partitions\n\nAudits already-defined analysis and assessment partitions for common forms of leakage and for incompatibility with a declared generalization target. The function does not create data splits, preprocess variables, select features, or fit predictive models.\n\nThe overall status is `"fail" when at least one failing check is present, `"review" when no failing checks are present but at least one review item is present, and `"pass" otherwise. When `participant_id is supplied, trial overlap is evaluated using composite participant-trial units. This permits trial labels such as `"T01" to be reused by different participants without being treated as leakage. Without `participant_id, `trial_id is assumed to be globally unique. The audit can identify structural leakage visible in the supplied partitions and declared variable roles. It cannot prove that preprocessing or feature selection was estimated inside resampling folds. Those operations require separate provenance and resampling safeguards. The function does not determine whether an outcome is scientifically or ethically appropriate. All uses remain subject to the package governance and prohibited-use statements.\n\nParameters\n----------\nanalysis\n    A data frame containing the analysis or training partition.\nassessment\n    A data frame containing the assessment or test partition.\noutcome\n    A single column name identifying the outcome.\npredictors\n    A character vector identifying intended predictor columns.\nparticipant_id\n    An optional participant-identifier column.\ntrial_id\n    An optional trial-identifier column.\nstimulus_id\n    An optional stimulus-identifier column.\ngeneralization_target\n    The predictive generalization target. One of `"new_trials_known_participants", `"new_participants", `"new_stimuli", or `"new_participants_and_new_stimuli".\ntarget_derived\n    Character vector of columns known to have been derived directly from the outcome.\npost_outcome\n    Character vector of columns measured or constructed after the outcome became available.\n\nReturns\n-------\nAn object of class `gazepoint_ml_leakage_audit. The object contains an overall status, partition summary, complete check table, and machine-readable table of non-passing issues.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'audit_gazepoint_model_robustness': 'Audit multiple robustness dimensions\n\nAudit multiple robustness dimensions\n\nParameters\n----------\nseed_stability\n    Optional seed-stability object.\nfeature_stability\n    Optional feature-stability object.\nthreshold_stability\n    Optional threshold-stability object.\nmissingness_stability\n    Optional missingness-stability object.\nrelative_sd_review\n    Relative SD threshold for review.\nrelative_sd_fail\n    Relative SD threshold for fail.\n\nReturns\n-------\nA `gp3ml_model_robustness_audit.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'audit_gazepoint_nested_resampling': 'Audit nested grouped resampling for outer-assessment leakage\n\nAudit nested grouped resampling for outer-assessment leakage\n\nParameters\n----------\nx\n    A `gp3ml_nested_folds object.\n\nReturns\n-------\nA `gp3ml_nested_resampling_audit.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'audit_gazepoint_plan_deviations': 'Audit deviations from a locked analysis plan\n\nAudit deviations from a locked analysis plan\n\nParameters\n----------\nplan\n    A locked analysis plan.\nactual\n    Named list describing the analysis actually performed.\nfields\n    Fields to compare.\n\nReturns\n-------\nA `gp3ml_plan_deviation_audit.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'audit_gazepoint_reproducibility': 'Audit generated artifacts for volatile output\n\nAudit generated artifacts for volatile output\n\nParameters\n----------\npaths\n    Files or directories to audit.\nrecursive\n    Whether directories are searched recursively.\nextensions\n    Text-file extensions to inspect.\n\nReturns\n-------\nA `gp3ml_reproducibility_audit.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'audit_gp3ml_api_stability': 'Audit gp3ml API stability\n\nAudit gp3ml API stability\n\nParameters\n----------\nregistry\n    Contract registry.\n\nReturns\n-------\nA `gp3ml_api_stability_audit object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'audit_gp3ml_governance_profile': 'Audit a governance-evidence profile\n\nAudit a governance-evidence profile\n\nParameters\n----------\nprofile\n    Governance profile.\n\nReturns\n-------\nA `gp3ml_governance_profile_audit.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'bake_gazepoint_preprocessor': 'Apply a fitted preprocessing engine\n\nApply a fitted preprocessing engine\n\nParameters\n----------\npreprocessor\n    A fitted `gp3ml_preprocessor object.\nnew_data\n    Data to transform using the fitted parameters.\n\nReturns\n-------\nA numeric model matrix transformed using only the parameters stored in the fitted preprocessor.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'bootstrap_gazepoint_metrics': 'Bootstrap uncertainty intervals for performance metrics\n\nBootstrap uncertainty intervals for performance metrics\n\nParameters\n----------\ntask\n    A governed `gp3ml_task object.\ntruth\n    Observed outcome values.\nprediction\n    Predicted classes or numeric values.\nprobability\n    Predicted positive-class probabilities.\nthreshold\n    Probability threshold for classification.\nbootstrap\n    Number of bootstrap replicates.\nconf_level\n    Confidence level for percentile intervals.\nseed\n    Deterministic random seed.\n\nReturns\n-------\nA `gp3ml_metric_uncertainty object containing point estimates, percentile intervals, bootstrap draws, resampling settings, and the governed task.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'bootstrap_gazepoint_metrics_by_unit': 'Generalization-target-aligned bootstrap uncertainty\n\nResamples observations or declared clusters while preserving every row that belongs to a sampled cluster. Repeated cluster draws duplicate all associated rows. The returned object records the resampling unit and must not be described as uncertainty for another unit.\n\nParameters\n----------\ntask\n    Governed task.\ntruth\n    Observed outcomes.\nprediction\n    Predicted classes or numeric outcomes.\nprobability\n    Positive-class probabilities.\nparticipant_id\n    Participant identifiers for participant-based methods.\nstimulus_id\n    Stimulus identifiers for stimulus-based methods.\nunit\n    Resampling unit.\nbootstrap\n    Number of replicates.\nconf_level\n    Percentile interval level.\nseed\n    Deterministic seed.\nthreshold\n    Classification threshold.\nstratify_observations\n    Whether the observation-level classification bootstrap preserves class counts.\n\nReturns\n-------\nA `gp3ml_target_uncertainty object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'capture_gazepoint_environment': 'Capture a reproducibility environment record\n\nCapture a reproducibility environment record\n\nParameters\n----------\npackages\n    Packages to record. Defaults to gp3ml and currently loaded namespaces.\nroot\n    Repository/project root used to capture Git SHA.\ninclude_renv\n    Whether to record an existing `renv.lock hash.\n\nReturns\n-------\nA `gp3ml_environment_record.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'collect_gazepoint_fold_predictions': 'Collect predictions from a grouped-fold evaluation\n\nCollect predictions from a grouped-fold evaluation\n\nParameters\n----------\nx\n    A `gp3ml_resample_evaluation.\ninclude_failed\n    Whether failed folds are represented by explicit status rows when they produced no predictions.\n\nReturns\n-------\nA data frame of row-level assessment predictions with fold labels.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'combine_gazepoint_handoffs': "Combine validated cross-package handoffs\n\nCombine validated cross-package handoffs\n\nParameters\n----------\nhandoffs\n    Named list of `gp3ml_handoff objects.\nkeys\n    Optional join keys; defaults to the first handoff's keys.\ncollision\n    How to handle overlapping non-key column names.\n\nReturns\n-------\nA `gp3ml_handoff_bundle.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.", 'compare_gazepoint_environments': 'Compare two environment records\n\nCompare two environment records\n\nParameters\n----------\nreference\n    Reference environment.\ncurrent\n    Current environment.\n\nReturns\n-------\nA `gp3ml_environment_comparison.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'compare_gazepoint_models': 'Compare governed model candidates without selecting a winner\n\nCompare governed model candidates without selecting a winner\n\nParameters\n----------\nx\n    A `gp3ml_model_tuning object.\nmetrics\n    Optional metric names.\n\nReturns\n-------\nA data frame retaining candidate status, failures, complexity, interpretability, and fold-distribution summaries.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_external_validation_report': 'Create an external-validation report object\n\nCreate an external-validation report object\n\nParameters\n----------\nvalidation\n    A `gp3ml_external_validation object.\ndevelopment_metrics\n    Optional development-sample metrics.\nlimitations\n    Character vector describing report limitations.\n\nReturns\n-------\nA `gp3ml_external_validation_report object containing the validation result, optional development metrics, limitations, and prohibited-use information.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_decision_rule': 'Create a governed classification decision rule\n\nCreate a governed classification decision rule\n\nParameters\n----------\nmetric\n    Metric used to justify the threshold.\ndirection\n    Either `"maximize" or `"minimize".\nthreshold\n    Optional probability threshold. Leave `NULL until selected.\nthreshold_origin\n    Origin of the threshold.\ncost_false_positive\n    Non-negative false-positive cost.\ncost_false_negative\n    Non-negative false-negative cost.\nabstention_allowed\n    Whether abstention is permitted.\nabstention_interval\n    Optional length-two probability interval. Probabilities inside the interval are labelled as abstentions.\ncalibration_source\n    Description of the calibration source.\ntraining_partition\n    Description of the data partition used to determine the threshold.\ngeneralization_target\n    Declared generalization target.\nscientific_justification\n    Explicit scientific justification.\n\nReturns\n-------\nA `gp3ml_decision_rule.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_feature_manifest': 'Create a Gazepoint feature-provenance manifest\n\nCreates a structured provenance manifest for intended predictive features. Each row records where a feature originated, when it became available, whether it is outcome-derived or post-outcome, and where any data-dependent preprocessing was estimated.\n\nEach row is treated as an intended predictor. Consequently, outcome-derived, post-outcome, unavailable, and identifier features are treated as failing conditions by `validate_gazepoint_feature_manifest(). The manifest records declared provenance. It does not independently prove that preprocessing was estimated within the stated scope.\n\nParameters\n----------\nfeatures\n    Character vector of unique feature names.\nscientific_source\n    Scientific or measurement source for each feature.\nsource_table\n    Source export, table, or object for each feature.\ntransformation\n    Description of the transformation used to construct each feature.\navailability_stage\n    Availability stage for each feature. One of `"pre_exposure", `"during_exposure", `"post_exposure_pre_outcome", `"at_prediction", `"post_outcome", or `"unknown".\nprediction_time_available\n    Logical vector indicating whether each feature is available at the intended prediction time.\noutcome_derived\n    Logical vector indicating whether each feature was derived directly or indirectly from the outcome.\npost_outcome\n    Logical vector indicating whether each feature was measured or constructed after the outcome became available.\nidentifier\n    Logical vector indicating whether each feature is an identifier or row-location variable.\npreprocessing_scope\n    Scope in which any data-dependent preprocessing was estimated. One of `"none", `"global", `"analysis_partition", `"resampling_fold", or `"unknown".\nfold_local_required\n    Logical vector indicating whether preprocessing for each feature must be estimated separately inside each resampling fold.\nreviewer_notes\n    Optional reviewer-facing notes.\n\nReturns\n-------\nA data frame of class `gazepoint_feature_manifest.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_group_folds': 'Create deterministic group-aware Gazepoint resampling folds\n\nCreates repeated grouped assessment folds that preserve the grouping structure implied by an explicit generalization target. A passing feature-provenance manifest is required, and every analysis-assessment pair is evaluated using the leakage audit.\n\nFor new trials among known participants, participant-trial units are assigned separately within each participant. For simultaneous participant and stimulus generalization, crossed participant-stimulus assessment blocks are created; cross-block rows are excluded from that fold. Each source row appears in assessment exactly once per repeat. This function does not perform preprocessing, feature selection, tuning, nested resampling, or model fitting.\n\nParameters\n----------\ndata\n    A data frame containing the outcome, predictors, and grouping identifiers.\noutcome\n    Name of the outcome column.\npredictors\n    Character vector naming intended predictors.\nfeature_manifest\n    A feature manifest containing all intended predictors.\ngeneralization_target\n    One of `"new_trials_known_participants", `"new_participants", `"new_stimuli", or `"new_participants_and_new_stimuli".\nparticipant_id\n    Optional participant-identifier column.\ntrial_id\n    Optional trial-identifier column.\nstimulus_id\n    Optional stimulus-identifier column.\nv\n    Number of group folds. For simultaneous participant and stimulus generalization, a length-two vector specifies participant and stimulus fold counts.\nrepeats\n    Number of repeated fold assignments.\nseed\n    Integer random seed. The caller\'s random-number state is restored.\nsource_row_id\n    Name of the source-row identifier added to returned partitions.\n\nReturns\n-------\nAn object of class `gazepoint_group_folds.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_handoff': 'Create a lightweight cross-package Gazepoint handoff\n\nCreate a lightweight cross-package Gazepoint handoff\n\nParameters\n----------\ndata\n    Prepared data frame.\nsource_package\n    Upstream source package or `study_design/`custom.\nsource_version\n    Optional source-package version.\nproducer\n    Optional upstream function/workflow label.\nkeys\n    Character vector of row-identifying join keys.\noutcome\n    Optional observed outcome column.\npredictors\n    Optional prepared predictor columns.\nfeature_manifest\n    Optional gp3ml feature-provenance manifest.\nnotes\n    Optional handoff notes.\n\nReturns\n-------\nA `gp3ml_handoff object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_model_artifact': 'Create a portable governed model artifact\n\nCreate a portable governed model artifact\n\nParameters\n----------\nmodel\n    A fitted `gp3ml_model or controlled model object.\npreprocessor\n    Optional preprocessing object.\nfeature_manifest\n    Optional feature manifest.\ntask\n    Optional task; defaults to `model$task.\ndecision_rule\n    Optional decision rule.\nmodel_card\n    Optional model card.\nreference_data\n    Optional deterministic prediction fixture.\nbundle_model\n    Whether to attempt `bundle::bundle() when available.\n\nReturns\n-------\nA `gp3ml_model_artifact.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_model_card': 'Create a governance-focused model card\n\nCreate a governance-focused model card\n\nParameters\n----------\nmodel\n    A fitted `gp3ml_model object.\nintended_use\n    Explicit description of the intended research use.\nevaluation\n    Optional performance-evaluation object.\ncalibration\n    Optional calibration-assessment object.\nfeature_manifest\n    Optional feature-provenance manifest.\nexternal_validation\n    Optional external-validation result.\nlimitations\n    Character vector describing model limitations.\nethical_review\n    Optional ethical-review information.\n\nReturns\n-------\nA `gp3ml_model_card object containing task, model, governance, evaluation, calibration, provenance, external-validation, and limitation metadata.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_nested_folds': 'Create nested grouped resampling from mature outer folds\n\nInner folds are constructed only from each outer analysis partition and preserve the declared participant/stimulus generalization target. The outer assessment partition is never used for inner preprocessing or tuning.\n\nParameters\n----------\nouter_folds\n    A validated `gazepoint_group_folds object.\ninner_v\n    Number of inner folds.\ninner_repeats\n    Number of inner repeats.\nseed\n    Base deterministic seed.\ncontinue_on_error\n    Whether infeasible outer folds are retained as failures instead of stopping immediately.\n\nReturns\n-------\nA `gp3ml_nested_folds object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_release_evidence': 'Create a release evidence manifest\n\nCreate a release evidence manifest\n\nParameters\n----------\nobjects\n    Named analysis objects to fingerprint.\nfiles\n    Named file paths to checksum.\nversion\n    Intended future release version.\nnotes\n    Optional release notes.\n\nReturns\n-------\nA `gp3ml_release_evidence object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_release_model_card': 'Create a release-ready governed model card\n\nExtends the existing model-card structure with explicit model-selection, target-aligned uncertainty, nested-resampling, and transportability fields.\n\nParameters\n----------\nmodel\n    Fitted governed model.\nintended_use\n    Intended scientific use.\nevaluation\n    Grouped or nested evaluation.\nselection\n    Optional `gp3ml_model_selection.\nuncertainty\n    Optional target-aligned uncertainty object.\ncalibration\n    Optional calibration assessment.\nfeature_manifest\n    Optional feature manifest.\ntransportability\n    Optional transportability report.\nlimitations\n    Required limitations.\nethical_review\n    Optional ethical-review information.\ndeployment_status\n    Deployment status; defaults to research review only.\n\nReturns\n-------\nA `gp3ml_release_model_card.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_reproducibility_report': 'Create a reproducibility report\n\nCreate a reproducibility report\n\nParameters\n----------\nobjects\n    Named objects to fingerprint.\ndata\n    Optional data frame to fingerprint.\nseeds\n    Named list of deterministic seeds.\nnotes\n    Optional reproducibility notes.\nproject_path\n    Project directory recorded in the report.\n\nReturns\n-------\nA `gp3ml_reproducibility_report object containing runtime information, object and data fingerprints, seeds, Git metadata, notes, and prohibited uses.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_synthetic_manifest': 'Create a synthetic governed feature manifest\n\nCreate a synthetic governed feature manifest\n\nParameters\n----------\noutcome\n    Name of the observed synthetic outcome.\npredictors\n    Predictor names to declare.\nparticipant_id\n    Participant identifier column.\nstimulus_id\n    Stimulus identifier column.\ntrial_id\n    Trial identifier column.\n\nReturns\n-------\nA `gazepoint_feature_manifest produced by `create_gazepoint_feature_manifest().\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_synthetic_task': 'Create one of the governed synthetic demonstration tasks\n\nCreate one of the governed synthetic demonstration tasks\n\nParameters\n----------\ndata\n    Synthetic data from `simulate_gazepoint_governed_data().\nworkflow\n    Workflow name.\ngeneralization_target\n    Declared generalization target.\n\nReturns\n-------\nA governed `gp3ml_task.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gazepoint_tuning_grid': 'Create an explicit governed tuning grid\n\nCandidate values are fully materialized before evaluation. No hidden metric, default ranking rule, or automatic winner is created.\n\nParameters\n----------\nengine\n    One or more governed engine names.\nengine_grid\n    Named list of engine-argument candidate values.\npreprocessor_grid\n    Named list of preprocessing-argument candidate values.\nthresholds\n    One or more explicit classification thresholds.\ncomplexity\n    Optional complexity labels or numeric scores.\ninterpretability\n    Optional interpretability labels or numeric scores.\nlabels\n    Optional candidate labels.\n\nReturns\n-------\nA `gp3ml_tuning_grid with one row per explicit candidate.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'create_gp3ml_governance_profile': 'Create a governance-evidence profile\n\nCreate a governance-evidence profile\n\nParameters\n----------\nevidence\n    Named list of gp3ml evidence objects.\nframework\n    Governance crosswalk.\n\nReturns\n-------\nA `gp3ml_governance_profile.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'declare_gazepoint_analysis_plan': 'Declare a frozen-analysis-plan contract\n\nDeclare a frozen-analysis-plan contract\n\nParameters\n----------\nresearch_question\n    Research question.\nscientific_purpose\n    Explicit scientific purpose.\noutcome\n    Outcome name.\noutcome_definition\n    Operational definition of the observed outcome.\npredictors\n    Predeclared predictors.\ngeneralization_target\n    Intended generalization target.\ngrouping_variables\n    Grouping columns.\neligible_population\n    Eligibility statement.\nexclusion_rules\n    Character vector of predeclared exclusions.\npreprocessing_plan\n    Preprocessing plan.\ncandidate_models\n    Candidate model specifications or names.\nprimary_metric\n    Primary metric.\nsecondary_metrics\n    Secondary metrics.\ncalibration_metric\n    Calibration metric.\nuncertainty_method\n    Uncertainty method.\nthreshold_policy\n    Threshold/decision policy.\nexternal_validation_required\n    Whether independent validation is required.\nseed_strategy\n    Deterministic seed strategy.\nprohibited_interpretations\n    Character vector of prohibited interpretations.\n\nReturns\n-------\nA mutable `gp3ml_analysis_plan until locked.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'declare_gazepoint_external_dataset': 'Declare an external dataset and its independence status\n\nDeclare an external dataset and its independence status\n\nParameters\n----------\ndata\n    Candidate external-validation data.\nlabel\n    Dataset label.\nindependent\n    Explicit logical declaration of independence from model development and internal resampling.\norigin\n    Human-readable origin or collection source.\ncollection_period\n    Optional collection period.\nparticipant_id\n    Participant identifier column.\nstimulus_id\n    Stimulus identifier column.\nnotes\n    Optional notes.\n\nReturns\n-------\nA `gp3ml_external_dataset_declaration.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'declare_gazepoint_task': 'Declare a governed Gazepoint prediction task\n\nDeclare a governed Gazepoint prediction task\n\nParameters\n----------\ndata\n    A data frame containing the outcome and task identifiers.\noutcome\n    Name of the explicitly observed outcome column.\npurpose\n    One explicit scientific-purpose statement.\ntask_type\n    Either `classification or `regression.\nunit_id\n    Column identifying the prediction unit.\nparticipant_id\n    Optional participant-identifier column.\nstimulus_id\n    Optional stimulus-identifier column.\ngeneralization_target\n    The intended target of generalization.\npositive\n    Positive outcome level for binary classification.\nobserved_outcome\n    Whether the outcome was directly observed.\nsensitive_outcome\n    Whether the outcome is sensitive or prohibited.\n\nReturns\n-------\nA governed `gp3ml_task object describing the outcome, scientific purpose, prediction unit, grouping roles, task type, and generalization target.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'diagnose_gazepoint_group_folds': 'Diagnose group-aware Gazepoint resampling folds\n\nCreates fold-size, repeat-level, grouping, assessment-coverage, outcome-balance, and exclusion diagnostics for an existing `gazepoint_group_folds object.\n\nParameters\n----------\nx\n    A `gazepoint_group_folds object.\nimbalance_review\n    Fold-size ratio above which diagnostics receive a `review status.\nimbalance_fail\n    Fold-size ratio above which diagnostics receive a `fail status.\n\nReturns\n-------\nAn object of class `gazepoint_fold_diagnostics.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'evaluate_external_validation': 'Evaluate an independent external-validation dataset\n\nEvaluate an independent external-validation dataset\n\nParameters\n----------\nmodel\n    A fitted `gp3ml_model object.\nexternal_data\n    Independent external-validation data.\nlabel\n    Label identifying the validation dataset.\nthreshold\n    Classification probability threshold.\nbootstrap\n    Number of calibration bootstrap replicates.\nseed\n    Deterministic random seed.\n\nReturns\n-------\nA `gp3ml_external_validation object containing external predictions, performance metrics, calibration results where applicable, predictor-shift diagnostics, a dataset fingerprint, and task metadata.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'evaluate_gazepoint_external_transportability': 'Evaluate external transportability and validation status\n\nAn internal holdout or a dataset explicitly declared non-independent is labelled `not_externally_validated; it cannot generate an external- validation claim.\n\nParameters\n----------\nmodel\n    Fitted governed model.\ndevelopment_data\n    Data used to characterize development schema and group coverage.\nexternal_data\n    Candidate external data. May be `NULL to create an explicit not-validated status.\ndeclaration\n    External dataset declaration. Required when external data are supplied.\ndevelopment_evaluation\n    Optional grouped development evaluation.\nthreshold\n    Classification threshold.\nbootstrap\n    Calibration bootstrap replicates.\nseed\n    Deterministic seed.\n\nReturns\n-------\nA `gp3ml_transportability_report object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'evaluate_gazepoint_feature_stability': 'Evaluate leave-one-feature-out stability\n\nEvaluate leave-one-feature-out stability\n\nParameters\n----------\nfeatures\n    Predictor names.\nevaluator\n    Function called as `evaluator(excluded_feature = feature, ...).\nkwargs\n    Additional reserved compatibility arguments.\n\nReturns\n-------\nA `gp3ml_stability_evaluation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'evaluate_gazepoint_group_folds': "Evaluate a governed model specification across materialized grouped folds\n\nFits preprocessing and the requested model only on each fold's analysis partition, predicts only on the corresponding assessment partition, retains excluded rows, and records fold-level metrics, leakage audits, warnings, and failures. Row-level predictions are never relabelled as participant- or stimulus-level estimates.\n\nParameters\n----------\nfolds\n    A mature `gazepoint_group_folds object containing materialized folds under `folds$folds.\ntask\n    A governed `gp3ml_task compatible with the fold metadata.\npredictors\n    Optional predictor names. Defaults to the fold metadata.\nengine\n    Model engine name or governed custom engine.\npreprocessor_args\n    Arguments passed to `fit_gazepoint_preprocessor().\nengine_args\n    Arguments passed to `fit_gazepoint_model().\nthreshold\n    Classification threshold.\nseed\n    Base deterministic seed.\nassess_calibration\n    Whether to calculate assessment-fold calibration summaries for classification tasks.\ncalibration_bins\n    Number of reliability bins.\ncalibration_bootstrap\n    Calibration bootstrap replicates. Use zero in fast smoke tests.\nkeep_models\n    Whether fitted fold models are retained.\ncontinue_on_error\n    Whether later folds continue after a failed fold.\n\nReturns\n-------\nA `gp3ml_resample_evaluation object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.", 'evaluate_gazepoint_missingness_sensitivity': 'Evaluate named missingness-sensitivity scenarios\n\nEvaluate named missingness-sensitivity scenarios\n\nParameters\n----------\nscenarios\n    Named list of scenario objects.\nevaluator\n    Function called as `evaluator(scenario = scenario, name = name, ...).\nkwargs\n    Additional reserved compatibility arguments.\n\nReturns\n-------\nA `gp3ml_stability_evaluation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'evaluate_gazepoint_nested_resampling': 'Evaluate nested grouped resampling with inner governed tuning\n\nEvaluate nested grouped resampling with inner governed tuning\n\nParameters\n----------\nnested_folds\n    A `gp3ml_nested_folds object.\ntask\n    Governed task.\ntuning_grid\n    Explicit tuning grid.\nselection_metric\n    Explicit inner selection metric.\ndirection\n    Explicit selection direction.\npredictors\n    Optional predictors.\nminimum_success_prop\n    Minimum inner-fold success proportion.\ntie_breakers\n    Optional secondary metrics.\nselection_rationale\n    Human rationale recorded for each outer fold.\nseed\n    Base deterministic seed.\nkeep_models\n    Whether outer fitted models are retained.\ncontinue_on_error\n    Whether failed outer folds remain in the result.\n\nReturns\n-------\nA `gp3ml_nested_evaluation object retaining inner tuning results, selections, outer predictions, metrics, and failures.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'evaluate_gazepoint_seed_stability': 'Evaluate seed stability\n\nEvaluate seed stability\n\nParameters\n----------\nseeds\n    Integer seeds.\nevaluator\n    Function called as `evaluator(seed = seed, ...).\nkwargs\n    Additional reserved compatibility arguments.\n\nReturns\n-------\nA `gp3ml_stability_evaluation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'evaluate_gazepoint_threshold_stability': 'Evaluate threshold stability around the optimum\n\nEvaluate threshold stability around the optimum\n\nParameters\n----------\nevaluation\n    Threshold evaluation.\nmetric\n    Metric.\ndirection\n    Optimization direction.\ntolerance\n    Fractional tolerance from the optimum.\n\nReturns\n-------\nA `gp3ml_threshold_stability.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'evaluate_gazepoint_thresholds': 'Evaluate explicit classification thresholds\n\nEvaluate explicit classification thresholds\n\nParameters\n----------\ntruth\n    Observed binary outcome.\nprobability\n    Probability of the positive class.\npositive\n    Positive class label.\nthresholds\n    Explicit candidate thresholds.\ncost_false_positive\n    False-positive cost.\ncost_false_negative\n    False-negative cost.\n\nReturns\n-------\nA `gp3ml_threshold_evaluation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'fit_gazepoint_calibrator': 'Fit a probability calibrator\n\nFit a probability calibrator\n\nParameters\n----------\ntruth\n    Observed binary outcome values.\nprobability\n    Uncalibrated positive-class probabilities.\npositive\n    Label representing the positive class.\nmethod\n    Calibration method: Platt scaling or isotonic regression.\n\nReturns\n-------\nA fitted `gp3ml_calibrator object containing the calibration method, fitted model, and outcome labels.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'fit_gazepoint_conformal': 'Fit target-aware split-conformal calibration\n\nThis function provides conservative split-conformal calibration for explicitly observed regression or binary classification outcomes. When a grouped calibration unit is supplied, row scores are aggregated to the maximum score within each calibration unit before the conformal quantile is estimated. This records and respects the calibration unit but does not claim distribution-free coverage under arbitrary dependence.\n\nParameters\n----------\ntruth\n    Observed calibration outcomes.\nprediction\n    Numeric predictions for regression.\nprobability\n    Positive-class probabilities for classification.\ntask_type\n    `"regression" or `"classification".\npositive\n    Positive class label for classification.\nlevel\n    Nominal coverage level.\ncalibration_unit\n    Calibration unit.\nunit\n    Optional group identifier for grouped calibration.\ngeneralization_target\n    Declared generalization target.\n\nReturns\n-------\nA `gp3ml_conformal_fit.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'fit_gazepoint_deep_model': 'Fit an optional governed deep-learning model through keras3\n\nFit an optional governed deep-learning model through keras3\n\nParameters\n----------\ndata\n    Analysis data used to fit the network.\ntask\n    A governed `gp3ml_task object.\npredictors\n    Optional character vector of predictor columns.\npreprocessor\n    Optional fitted preprocessing object.\nhidden_units\n    Integer vector of hidden-layer sizes.\ndropout\n    Dropout proportion applied after hidden layers.\nepochs\n    Number of training epochs.\nbatch_size\n    Training batch size.\nvalidation_split\n    Proportion reserved for internal validation.\noptimizer\n    Keras optimizer name or object.\nseed\n    Deterministic random seed.\nverbose\n    Keras training verbosity.\n\nReturns\n-------\nA governed `gp3ml_model object containing the fitted `keras3 model, training history, preprocessing object, task contract, and training metadata.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'fit_gazepoint_model': 'Fit a governed Gazepoint model\n\nFit a governed Gazepoint model\n\nParameters\n----------\ndata\n    Analysis data used to fit the model.\ntask\n    A governed `gp3ml_task object.\npredictors\n    Optional character vector of predictor columns.\nengine\n    Engine name or controlled custom-engine object.\npreprocessor\n    Optional fitted preprocessing object.\npreprocessor_args\n    Arguments passed to preprocessing fitting.\nengine_args\n    Arguments passed to the model engine.\nseed\n    Deterministic random seed.\nthreshold\n    Classification probability threshold.\n\nReturns\n-------\nA governed `gp3ml_model object containing the fitted engine, preprocessing object, task contract, predictors, and training metadata.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'fit_gazepoint_preprocessor': 'Fit a fold-local preprocessing engine\n\nFit a fold-local preprocessing engine\n\nParameters\n----------\ndata\n    Analysis data used to estimate preprocessing parameters.\npredictors\n    Character vector naming predictor columns.\nnumeric_imputation\n    Numeric imputation method.\ncenter\n    Whether numeric model columns should be centered.\nscale\n    Whether numeric model columns should be scaled.\nnovel_level\n    How novel categorical levels should be handled.\nremove_zero_variance\n    Whether zero-variance columns are removed.\n\nReturns\n-------\nA fitted `gp3ml_preprocessor object containing analysis-partition imputation values, factor levels, model columns, centering values, and scaling values.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'gazepoint_classification_metrics': 'Binary classification metrics\n\nBinary classification metrics\n\nParameters\n----------\ntruth\n    Observed binary outcome values.\nprobability\n    Predicted positive-class probabilities.\npredicted\n    Optional predicted classes.\npositive\n    Label representing the positive class.\nthreshold\n    Probability threshold used for class predictions.\n\nReturns\n-------\nA one-row data frame containing the sample size, threshold, class-performance measures, discrimination metrics, Brier score, and log loss.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'gazepoint_performance_metrics': 'Task-aware performance metrics\n\nTask-aware performance metrics\n\nParameters\n----------\ntask\n    A governed `gp3ml_task object.\ntruth\n    Observed outcome values.\nprediction\n    Predicted classes or numeric values.\nprobability\n    Predicted positive-class probabilities.\nthreshold\n    Probability threshold for classification.\n\nReturns\n-------\nA one-row data frame of classification or regression metrics selected according to the governed task type.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'gazepoint_regression_metrics': 'Regression metrics\n\nRegression metrics\n\nParameters\n----------\ntruth\n    Observed numeric outcome values.\nprediction\n    Predicted numeric outcome values.\n\nReturns\n-------\nA one-row data frame containing the sample size, RMSE, MAE, R-squared value, and prediction correlation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'gp3ml_api_contracts': "gp3ml public API contracts\n\nReturns the package's explicit compatibility contract for the public API. APIs classified as stable in version 0.2.0 remain stable throughout the 0.3.x line. New APIs introduced by the current development milestone are marked experimental until promoted by a later release decision.\n\nReturns\n-------\nA `gp3ml_api_contract_registry object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.", 'gp3ml_available_engines': 'List available model engines\n\nList available model engines\n\nReturns\n-------\nA data frame listing supported model-engine names and whether each optional engine is currently available.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'gp3ml_engine_capabilities': 'Audit gp3ml model-engine capabilities\n\nAudit gp3ml model-engine capabilities\n\nParameters\n----------\ncheck_keras_backend\n    Whether to query the configured Keras backend.\n\nReturns\n-------\nA `gp3ml_engine_capabilities data frame.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'gp3ml_interop_contracts': 'Cross-package interoperability contracts\n\nDescribes a lightweight handoff boundary. Upstream packages remain responsible for their own importing, cleaning, feature derivation, signal processing, sequence processing, and quality control. `gp3ml receives already prepared observed variables together with explicit provenance.\n\nReturns\n-------\nA data frame describing supported handoff sources and responsibilities.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'gp3ml_object_schema': 'Describe the schema of a gp3ml object\n\nDescribe the schema of a gp3ml object\n\nParameters\n----------\nx\n    Object to inspect.\nrecursive\n    Whether to include one level of nested named-list components.\n\nReturns\n-------\nA data frame describing component names, classes, storage types, lengths, and dimensions.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'gp3ml_prohibited_uses': 'Prohibited gp3ml uses\n\nProhibited gp3ml uses\n\nReturns\n-------\nA character vector of prohibited use descriptions.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'integrate_black_box_model': 'Integrate a controlled black-box model engine\n\nIntegrate a controlled black-box model engine\n\nParameters\n----------\nname\n    Unique name for the custom engine.\nfit_fun\n    Function that fits the custom engine.\npredict_fun\n    Function that generates predictions.\nsupports\n    Task types supported by the engine.\nprobability\n    Whether classification probabilities are supported.\nmetadata\n    Optional engine metadata.\nsafety_declaration\n    Named logical safety declarations.\n\nReturns\n-------\nA controlled `gp3ml_engine object containing the custom fit and prediction functions, supported task types, metadata, and explicit safety declarations.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'lock_gazepoint_analysis_plan': 'Lock an analysis plan using SHA-256\n\nLock an analysis plan using SHA-256\n\nParameters\n----------\nplan\n    A valid unlocked plan.\nplan_id\n    Optional stable identifier.\nlocked_at\n    Optional lock time.\n\nReturns\n-------\nA locked `gp3ml_analysis_plan.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'normalize_gazepoint_artifact_text': 'Normalize volatile text in generated research artifacts\n\nNormalize volatile text in generated research artifacts\n\nParameters\n----------\nx\n    Character vector.\nproject_path\n    Optional project path to replace by `<PROJECT>.\n\nReturns\n-------\nCharacter vector with volatile runtime fragments normalized.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'predict_gazepoint_interval': 'Predict conformal regression intervals\n\nPredict conformal regression intervals\n\nParameters\n----------\nobject\n    A regression `gp3ml_conformal_fit.\nprediction\n    Point predictions.\n\nReturns\n-------\nData frame with point prediction, lower, and upper limits.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'predict_gazepoint_set': 'Predict binary conformal prediction sets\n\nPredict binary conformal prediction sets\n\nParameters\n----------\nobject\n    A classification `gp3ml_conformal_fit.\nprobability\n    Positive-class probabilities.\n\nReturns\n-------\nA data frame containing set membership and a readable set label.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'restore_gazepoint_model_artifact': 'Restore a model artifact\n\nRestore a model artifact\n\nParameters\n----------\nartifact\n    A model artifact.\n\nReturns\n-------\nA restored artifact with an unbundled model where needed.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'select_gazepoint_model': 'Select a governed candidate using an explicit metric and direction\n\nThis function records a reviewable decision. It does not refit a model and refuses accuracy as the sole primary metric.\n\nParameters\n----------\nx\n    A `gp3ml_model_tuning object.\nmetric\n    Explicit primary metric.\ndirection\n    Explicit optimization direction.\nminimum_success_prop\n    Minimum successful-fold proportion.\ntie_breakers\n    Optional ordered secondary metric names.\nrationale\n    Required human-readable selection rationale.\n\nReturns\n-------\nA `gp3ml_model_selection object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'select_gazepoint_threshold': 'Select a threshold from a governed threshold evaluation\n\nSelect a threshold from a governed threshold evaluation\n\nParameters\n----------\nevaluation\n    A `gp3ml_threshold_evaluation.\nmetric\n    Metric column to optimize.\ndirection\n    `"maximize" or `"minimize".\nthreshold_origin\n    Must identify an analysis/training source.\ntraining_partition\n    Partition used to select the threshold.\ngeneralization_target\n    Declared target.\nscientific_justification\n    Explicit justification.\nabstention_allowed\n    Whether abstention is allowed.\nabstention_interval\n    Optional abstention interval.\n\nReturns\n-------\nA `gp3ml_decision_rule.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'simulate_gazepoint_governed_data': 'Simulate governed synthetic Gazepoint-derived data\n\nCreates deterministic, non-sensitive synthetic data for package examples, tests, and website articles. The generated outcomes are explicitly observed: a predefined recording-quality review status, an experimentally assigned condition, and a non-sensitive recorded response.\n\nParameters\n----------\nn_participants\n    Number of synthetic participants.\nn_stimuli\n    Number of synthetic stimuli.\ntrials_per_cell\n    Number of trials per participant-stimulus cell.\nseed\n    Deterministic random seed.\n\nReturns\n-------\nA data frame containing identifiers, observed outcomes, and predeclared synthetic predictors.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'simulate_gazepoint_research_handoffs': 'Simulate realistic cross-package Gazepoint research handoffs\n\nGenerates deterministic, shareable, synthetic prepared outputs representing handoffs from `gp3tools, `gpbiometrics, and `gp3sequences. The outcome is an experimentally assigned condition. Biometrics variables are signal-quality summaries only; no health, emotion, stress, cognition, or other mental-state outcome is generated or inferred.\n\nParameters\n----------\nn_participants\n    Number of participants.\nn_stimuli\n    Number of stimuli.\ntrials_per_stimulus\n    Trials per participant-stimulus cell.\nseed\n    Deterministic seed.\n\nReturns\n-------\nA `gp3ml_research_bundle.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'split_gazepoint_ml_data': 'Create a deterministic group-aware Gazepoint holdout split\n\nCreates analysis and assessment partitions that preserve the grouping unit implied by an explicit generalization target.\n\nFor simultaneous participant and stimulus generalization, cross-block rows are placed in the excluded partition. This function does not perform preprocessing, feature selection, resampling, or model fitting.\n\nParameters\n----------\ndata\n    Data frame containing the outcome, predictors, and grouping identifiers.\noutcome\n    Name of the outcome column.\npredictors\n    Character vector of predictor-column names.\nfeature_manifest\n    Feature manifest containing the predictors.\ngeneralization_target\n    Declared predictive-generalization target.\nparticipant_id\n    Optional participant-identifier column.\ntrial_id\n    Optional trial-identifier column.\nstimulus_id\n    Optional stimulus-identifier column.\nassessment_prop\n    Requested assessment proportion.\nseed\n    Integer random seed.\nsource_row_id\n    Name of the source-row identifier added to the returned partitions.\n\nReturns\n-------\nAn object of class `gazepoint_ml_split.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'summarize_gazepoint_resample_performance': 'Summarize repeated grouped-resampling performance\n\nSummarize repeated grouped-resampling performance\n\nParameters\n----------\nx\n    A `gp3ml_resample_evaluation.\naggregation\n    Either fold-distribution summaries or pooled row-level predictions. Pooled rows are explicitly labelled and do not change the generalization unit.\nconf_level\n    Confidence level for fold-distribution quantiles.\n\nReturns\n-------\nA `gp3ml_resample_performance_summary object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'summarize_gazepoint_resample_uncertainty': 'Summarize uncertainty across folds or repeats\n\nSummarize uncertainty across folds or repeats\n\nParameters\n----------\nevaluation\n    A `gp3ml_resample_evaluation or `gp3ml_nested_evaluation.\nunit\n    Distribution unit: individual folds or repeat means.\nconf_level\n    Quantile interval level.\n\nReturns\n-------\nA `gp3ml_resample_uncertainty object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'summarize_gazepoint_shift': 'Summarize dataset shift without collapsing it to one drift score\n\nSummarize dataset shift without collapsing it to one drift score\n\nParameters\n----------\nshift\n    A dataset-shift audit.\nmissingness\n    Optional missingness-shift audit.\n\nReturns\n-------\nA structured summary.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'test_gazepoint_model_portability': 'Test model-artifact serialization and optional fresh-process prediction\n\nTest model-artifact serialization and optional fresh-process prediction\n\nParameters\n----------\nartifact\n    Model artifact.\nnewdata\n    Optional prediction fixture.\ntolerance\n    Numeric prediction tolerance.\nfresh_process\n    Whether to test in a fresh R process using `callr.\n\nReturns\n-------\nA `gp3ml_model_portability_test.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'train_gazepoint_classifier': 'Generic governed binary-classifier training wrapper\n\nGeneric governed binary-classifier training wrapper\n\nParameters\n----------\ndata\n    Analysis data used to train the classifier.\ntask\n    A governed binary-classification task.\npredictors\n    Optional character vector of predictor columns.\nengine\n    Classification engine name or custom engine.\nkwargs\n    Additional reserved compatibility arguments.\n\nReturns\n-------\nA governed classification `gp3ml_model object returned by `fit_gazepoint_model().\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'tune_gazepoint_model': 'Evaluate every governed candidate on the same grouped folds\n\nEvaluate every governed candidate on the same grouped folds\n\nParameters\n----------\nfolds\n    A `gazepoint_group_folds object.\ntask\n    A governed task.\ntuning_grid\n    A `gp3ml_tuning_grid.\npredictors\n    Optional declared predictors.\nmetrics\n    Optional metric names retained in the comparison table.\nseed\n    Base deterministic seed.\ncontinue_on_error\n    Whether failed candidates remain in the result while later candidates continue.\nkeep_evaluations\n    Whether complete candidate evaluations are retained.\n\nReturns\n-------\nA `gp3ml_model_tuning object retaining all candidates and failures.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_analysis_plan': 'Validate an analysis plan\n\nValidate an analysis plan\n\nParameters\n----------\nplan\n    A `gp3ml_analysis_plan.\n\nReturns\n-------\nA validation object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_conformal': 'Validate a conformal fit\n\nValidate a conformal fit\n\nParameters\n----------\nobject\n    A conformal fit.\n\nReturns\n-------\nA validation object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_decision_rule': 'Validate a governed decision rule\n\nValidate a governed decision rule\n\nParameters\n----------\nrule\n    A `gp3ml_decision_rule.\nrequire_threshold\n    Whether a concrete threshold is required.\n\nReturns\n-------\nA validation object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_environment': 'Validate the current environment against a reference record\n\nValidate the current environment against a reference record\n\nParameters\n----------\nreference\n    Reference environment record.\nroot\n    Project root.\ninclude_renv\n    Whether to compare renv lock hashes.\n\nReturns\n-------\nAn environment comparison.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_feature_manifest': 'Validate a Gazepoint feature-provenance manifest\n\nValidates the schema and declared scientific safeguards in a feature manifest. Schema errors stop execution. Substantive concerns are returned as structured `pass, `review, or `fail checks.\n\nA manifest fails when an intended predictor is declared as outcome-derived, post-outcome, unavailable at prediction time, or an identifier. It also fails when fold-local estimation is required but preprocessing is declared outside the resampling fold. Unknown or incomplete provenance is returned for review rather than treated as evidence that a safeguard was satisfied.\n\nParameters\n----------\nx\n    A feature manifest created by `create_gazepoint_feature_manifest() or a compatible data frame.\n\nReturns\n-------\nAn object of class `gazepoint_feature_manifest_validation containing the overall status, complete checks, non-passing issues, and validated manifest.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_fold_diagnostics': 'Validate Gazepoint fold diagnostics\n\nValidate Gazepoint fold diagnostics\n\nParameters\n----------\nx\n    A `gazepoint_fold_diagnostics object.\n\nReturns\n-------\nAn object of class `gazepoint_fold_diagnostics_validation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_group_folds': 'Validate group-aware Gazepoint resampling folds\n\nValidate group-aware Gazepoint resampling folds\n\nParameters\n----------\nx\n    A `gazepoint_group_folds object.\n\nReturns\n-------\nAn object of class `gazepoint_group_folds_validation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_handoff': 'Validate a Gazepoint handoff\n\nValidate a Gazepoint handoff\n\nParameters\n----------\nx\n    A `gp3ml_handoff.\n\nReturns\n-------\nA `gp3ml_handoff_validation object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_ml_roles': 'Validate outcome, predictor, identifier, and grouping roles\n\nValidate outcome, predictor, identifier, and grouping roles\n\nParameters\n----------\ndata\n    A data frame containing outcome, predictors, and identifiers.\ntask\n    A governed `gp3ml_task object.\npredictors\n    Character vector naming intended predictors.\nfeature_manifest\n    Optional Gazepoint feature-provenance manifest.\n\nReturns\n-------\nA `gp3ml_role_validation object containing the overall status, complete check table, non-passing issues, and optional feature-manifest validation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_ml_split': 'Validate a group-aware Gazepoint holdout split\n\nValidate a group-aware Gazepoint holdout split\n\nParameters\n----------\nx\n    An object returned by `split_gazepoint_ml_data().\n\nReturns\n-------\nAn object of class `gazepoint_ml_split_validation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_model_artifact': 'Validate a model artifact\n\nValidate a model artifact\n\nParameters\n----------\nartifact\n    Artifact to validate.\nverify_hash\n    Whether to recompute the SHA-256 payload hash.\n\nReturns\n-------\nA validation object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_model_tuning': 'Validate governed tuning results\n\nValidate governed tuning results\n\nParameters\n----------\nx\n    A `gp3ml_model_tuning object.\n\nReturns\n-------\nA `gp3ml_model_tuning_validation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_nested_evaluation': 'Validate a nested evaluation\n\nValidate a nested evaluation\n\nParameters\n----------\nx\n    A `gp3ml_nested_evaluation.\n\nReturns\n-------\nA `gp3ml_nested_evaluation_validation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_nested_folds': 'Validate nested grouped folds\n\nValidate nested grouped folds\n\nParameters\n----------\nx\n    A `gp3ml_nested_folds object.\n\nReturns\n-------\nA `gp3ml_nested_folds_validation object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_release_checksums': 'Validate SHA-256 release checksums\n\nValidate SHA-256 release checksums\n\nParameters\n----------\nmanifest\n    Checksum manifest or path.\ndirectory\n    Directory containing artifacts.\n\nReturns\n-------\nA validation object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_resample_evaluation': 'Validate a grouped-fold evaluation result\n\nValidate a grouped-fold evaluation result\n\nParameters\n----------\nx\n    A `gp3ml_resample_evaluation.\n\nReturns\n-------\nA `gp3ml_resample_evaluation_validation object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_research_bundle': 'Validate a synthetic cross-package research bundle\n\nValidate a synthetic cross-package research bundle\n\nParameters\n----------\nx\n    A `gp3ml_research_bundle.\n\nReturns\n-------\nA `gp3ml_research_bundle_validation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_ro_crate': 'Validate a gp3ml RO-Crate-oriented export\n\nValidate a gp3ml RO-Crate-oriented export\n\nParameters\n----------\npath\n    Crate directory or `gp3ml_ro_crate.\n\nReturns\n-------\nA validation object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_target_uncertainty': 'Validate target-aligned uncertainty metadata\n\nValidate target-aligned uncertainty metadata\n\nParameters\n----------\nx\n    A `gp3ml_target_uncertainty or `gp3ml_resample_uncertainty.\n\nReturns\n-------\nA `gp3ml_uncertainty_validation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gazepoint_transportability': 'Validate an external transportability report\n\nValidate an external transportability report\n\nParameters\n----------\nx\n    A `gp3ml_transportability_report.\n\nReturns\n-------\nA `gp3ml_transportability_validation.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'validate_gp3ml_object_contract': 'Validate an object against the gp3ml public-object contract\n\nValidate an object against the gp3ml public-object contract\n\nParameters\n----------\nx\n    A gp3ml object.\nregistry\n    Contract registry from `gp3ml_api_contracts().\n\nReturns\n-------\nA `gp3ml_object_contract_validation object.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'with_gazepoint_reproducible_output': 'Evaluate code with deterministic documentation-output settings\n\nEvaluate code with deterministic documentation-output settings\n\nParameters\n----------\ncode\n    Expression to evaluate.\n\nReturns\n-------\nThe value of `code.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_external_validation_report': 'Write an external-validation report\n\nWrite an external-validation report\n\nParameters\n----------\nreport\n    A `gp3ml_external_validation_report object.\npath\n    Destination Markdown file path.\noverwrite\n    Whether an existing file may be replaced.\n\nReturns\n-------\nThe destination path, returned invisibly after the Markdown report is written.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_analysis_plan': 'Write an analysis plan\n\nWrite an analysis plan\n\nParameters\n----------\nplan\n    Analysis plan.\npath\n    Output path.\nformat\n    `"rds", `"json", or `"md".\n\nReturns\n-------\nNormalized output path, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_feature_manifest_csv': 'Write a Gazepoint feature manifest or validation table to CSV\n\nWrites a feature manifest or one table from a validated manifest to a UTF-8 CSV file. Existing files are not replaced unless explicitly permitted.\n\nParameters\n----------\nx\n    A `gazepoint_feature_manifest, compatible data frame, or object returned by `validate_gazepoint_feature_manifest().\nfile\n    A single output path ending in `.csv.\ntable\n    Table to export. One of `"manifest", `"issues", or `"checks". Plain manifest inputs support only `"manifest".\noverwrite\n    Logical. When `FALSE, the default, an existing file causes an error.\nna\n    Character value used for missing values.\n\nReturns\n-------\nThe normalized output path, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_fold_diagnostics_csv': 'Write Gazepoint fold diagnostics to CSV files\n\nWrite Gazepoint fold diagnostics to CSV files\n\nParameters\n----------\nx\n    A `gazepoint_fold_diagnostics object.\ndirectory\n    Output directory.\nprefix\n    File-name prefix.\ntables\n    Diagnostic tables to export.\noverwrite\n    Whether existing files may be overwritten.\nna\n    String used for missing values.\n\nReturns\n-------\nA named character vector of written file paths, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_group_folds_csv': 'Write group-aware resampling tables to CSV\n\nWrite group-aware resampling tables to CSV\n\nParameters\n----------\nx\n    A `gazepoint_group_folds object.\ndirectory\n    Output directory.\nprefix\n    Non-empty filename prefix.\ntables\n    Character vector selecting summary tables.\ninclude_fold_data\n    Logical. Whether every materialized fold partition should also be written.\noverwrite\n    Logical. Whether existing files may be replaced.\nna\n    Character representation of missing values.\n\nReturns\n-------\nA named character vector of normalized output paths, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_ml_leakage_audit_csv': 'Write a Gazepoint ML leakage-audit table to CSV\n\nWrites one machine-readable table from a leakage-audit object to a UTF-8 CSV file. Existing files are not replaced unless explicitly permitted.\n\nParameters\n----------\nx\n    An object returned by `audit_gazepoint_ml_leakage().\nfile\n    A single output file path ending in `.csv.\ntable\n    The audit table to export. One of `"issues", `"checks", or `"partition_summary".\noverwrite\n    Logical. When `FALSE, the default, an existing output file causes an error.\nna\n    Character value used for missing values in the CSV file.\n\nReturns\n-------\nThe normalized output path, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_ml_split_csv': 'Write group-aware split tables to CSV\n\nWrite group-aware split tables to CSV\n\nParameters\n----------\nx\n    A `gazepoint_ml_split object.\ndirectory\n    Output directory.\nprefix\n    Filename prefix.\ntables\n    Tables to export.\noverwrite\n    Whether existing files may be replaced.\nna\n    Character representation of missing values.\n\nReturns\n-------\nA named character vector of normalized file paths, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_model_card': 'Write a model card\n\nWrite a model card\n\nParameters\n----------\ncard\n    A `gp3ml_model_card object.\npath\n    Destination file path.\nformat\n    Output format: Markdown or JSON.\noverwrite\n    Whether an existing file may be replaced.\n\nReturns\n-------\nThe destination path, returned invisibly after the model card is written.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_model_tuning': 'Write governed tuning and selection tables\n\nWrite governed tuning and selection tables\n\nParameters\n----------\nx\n    A `gp3ml_model_tuning object.\ndirectory\n    Output directory.\nprefix\n    Filename prefix.\nselection\n    Optional `gp3ml_model_selection to record.\noverwrite\n    Whether existing files may be replaced.\n\nReturns\n-------\nNamed output paths, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_nested_evaluation': 'Write nested-resampling evaluation tables\n\nWrite nested-resampling evaluation tables\n\nParameters\n----------\nx\n    A `gp3ml_nested_evaluation.\ndirectory\n    Output directory.\nprefix\n    Filename prefix.\noverwrite\n    Whether existing files may be replaced.\n\nReturns\n-------\nNamed paths, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_release_checksums': 'Write SHA-256 checksums for release artifacts\n\nWrite SHA-256 checksums for release artifacts\n\nParameters\n----------\nfiles\n    Release artifact files.\npath\n    Output checksum manifest.\n\nReturns\n-------\nA `gp3ml_release_checksum_manifest.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_release_model_card': 'Write a release-ready governed model card\n\nWrite a release-ready governed model card\n\nParameters\n----------\ncard\n    A `gp3ml_release_model_card.\npath\n    Destination path.\nformat\n    Markdown or JSON.\noverwrite\n    Whether an existing file may be replaced.\n\nReturns\n-------\nThe destination path, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_reproducibility_audit': 'Write a reproducibility-hardening audit\n\nWrite a reproducibility-hardening audit\n\nParameters\n----------\naudit\n    A `gp3ml_reproducibility_audit.\ndirectory\n    Destination directory.\nprefix\n    File prefix.\noverwrite\n    Whether existing files may be replaced.\n\nReturns\n-------\nNamed paths.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_reproducibility_report': 'Write a reproducibility report\n\nWrite a reproducibility report\n\nParameters\n----------\nreport\n    A `gp3ml_reproducibility_report object.\npath\n    Destination Markdown file path.\noverwrite\n    Whether an existing file may be replaced.\n\nReturns\n-------\nThe destination path, returned invisibly after the reproducibility report is written.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_resample_evaluation': 'Write grouped-fold evaluation tables\n\nWrite grouped-fold evaluation tables\n\nParameters\n----------\nx\n    A `gp3ml_resample_evaluation.\ndirectory\n    Output directory.\nprefix\n    Filename prefix.\noverwrite\n    Whether existing files may be replaced.\n\nReturns\n-------\nNamed output paths, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_ro_crate': 'Write a minimal RO-Crate-oriented research object\n\nThis helper writes a conservative RO-Crate-oriented JSON-LD metadata file and SHA-256 file hashes. It does not claim formal RO-Crate conformance; use an independent validator when formal conformance is required.\n\nParameters\n----------\npath\n    Output directory.\nfiles\n    Named or unnamed character vector of files to include.\nname\n    Research-object name.\ndescription\n    Description.\ncreator_name\n    Creator name.\ncreator_orcid\n    Optional ORCID URI or identifier.\nlicense\n    License URI or label.\ndoi\n    Optional DOI.\ncopy_files\n    Whether to copy files into the crate directory.\n\nReturns\n-------\nA `gp3ml_ro_crate.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_target_uncertainty': 'Write target-aligned uncertainty tables\n\nWrite target-aligned uncertainty tables\n\nParameters\n----------\nx\n    A gp3ml uncertainty object.\ndirectory\n    Output directory.\nprefix\n    Filename prefix.\noverwrite\n    Whether existing files may be replaced.\n\nReturns\n-------\nNamed paths, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gazepoint_transportability_report': 'Write an expanded transportability report\n\nWrite an expanded transportability report\n\nParameters\n----------\nreport\n    A `gp3ml_transportability_report.\npath\n    Destination Markdown path.\noverwrite\n    Whether an existing file may be replaced.\n\nReturns\n-------\nThe destination path, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gp3ml_api_contracts': 'Write gp3ml API contracts\n\nWrite gp3ml API contracts\n\nParameters\n----------\nregistry\n    Contract registry.\ndirectory\n    Destination directory.\nprefix\n    File prefix.\noverwrite\n    Whether existing files may be replaced.\n\nReturns\n-------\nNamed character vector of written paths.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.', 'write_gp3ml_governance_profile': 'Write a governance profile audit\n\nWrite a governance profile audit\n\nParameters\n----------\naudit\n    Governance profile audit.\npath\n    Markdown output path.\n\nReturns\n-------\nOutput path, invisibly.\n\nNotes\n-----\nPython port of the frozen gp3ml 0.3.0 R contract. Governance and leakage safeguards are preserved; backend-level numerical parity is tracked separately where Python and R engines differ.'}