plexus.cli.procedure.model_performance_frontier module

class plexus.cli.procedure.model_performance_frontier.FrontierVariant(label: 'str', yaml_content: 'str', is_current: 'bool', model_provider: 'Any' = None, model_name: 'Any' = None, base_model_name: 'Any' = None, reasoning_effort: 'Any' = None, verbosity: 'Any' = None, temperature: 'Any' = None, max_tokens: 'Any' = None)

Bases: object

__init__(label: str, yaml_content: str, is_current: bool, model_provider: Any = None, model_name: Any = None, base_model_name: Any = None, reasoning_effort: Any = None, verbosity: Any = None, temperature: Any = None, max_tokens: Any = None) None
as_dict() dict[str, Any]
base_model_name: Any = None
is_current: bool
label: str
max_tokens: Any = None
model_name: Any = None
model_provider: Any = None
reasoning_effort: Any = None
temperature: Any = None
verbosity: Any = None
yaml_content: str
plexus.cli.procedure.model_performance_frontier.apply_override(data: Any, path: str, value: Any) None
plexus.cli.procedure.model_performance_frontier.build_result_row(variant: Mapping[str, Any], *, feedback_evaluation: Mapping[str, Any] | None, regression_evaluation: Mapping[str, Any] | None) dict[str, Any]
plexus.cli.procedure.model_performance_frontier.build_variants(yaml_content: str, candidate_matrix: Mapping[str, Any], *, include_current: bool | None = None) list[dict[str, Any]]

Build deterministic, de-duplicated score YAML variants.

The normal matrix only mutates score-root model fields. Node-level or other targeted edits are allowed only through explicit extra_overrides paths.

plexus.cli.procedure.model_performance_frontier.compact_report_envelope(*, artifact_paths: list[str], rows: Iterable[Mapping[str, Any]]) dict[str, Any]
plexus.cli.procedure.model_performance_frontier.mark_pareto_frontier(rows: Iterable[Mapping[str, Any]]) list[dict[str, Any]]
plexus.cli.procedure.model_performance_frontier.render_artifacts(rows: Iterable[Mapping[str, Any]], *, title: str = 'Model Performance Frontier') dict[str, str]