nirs4all.analysis.playground_execution module

Stateless scientific Playground preview; no jobs, HTTP or training runner.

Each step delegates to owner operators. The response is an exploration result, not a fitted predictive pipeline or an independent evaluation score.

nirs4all.analysis.playground_execution.execute_preview(batch: PreviewBatch, steps: list[PreviewStep] | None = None, *, sampling: dict[str, Any] | None = None, options: dict[str, Any] | None = None, limits: PreviewLimits | None = None, metric_compute: Callable[[...], dict[str, Any]] | None = None) → dict[str, Any]

Return true transformed arrays and inline analysis, preserving row lineage.

metric_compute optionally replaces the owner descriptor callable (not a server callback). The default computes the historical descriptor set. Request/response byte limits remain the application’s transport duty. use_cache is an optional optimization hint; this callable is stateless.