1 rollouts.presentation.recipe_experiment

rollouts.presentation.recipe_experiment

Present one retained recipe-experiment snapshot without acquiring stores.

The Oracle report owner validates the plan and shards; this module only renders its typed factual attempts and conditional matched-prefix contrasts. Missing endpoints remain unavailable rather than becoming zero-gain observations.

1.1 Functions

Name	Description
recipe_contrast_figure	Show conditional scene-mean effects and factual prefix availability.
recipe_experiment_html	Export a self-contained educational report from one retained snapshot.

1.1.1 recipe_contrast_figure

rollouts.presentation.recipe_experiment.recipe_contrast_figure(
    contrasts,
    *,
    baseline,
    challenger,
)

Show conditional scene-mean effects and factual prefix availability.

The upper panel uses only target/replica pairs with both factual endpoints at each acquisition prefix. The lower panel shows how many chains reached each prefix in each profile and how many were shared. These are descriptive counts and effects, not uncertainty intervals or lookahead estimates.

1.1.1.1 Raises

Name	Type	Description
	ValueError	If the requested declared pair is absent or a prefix is duplicated, which would make the displayed estimand ambiguous.

1.1.2 recipe_experiment_html

rollouts.presentation.recipe_experiment.recipe_experiment_html(snapshot)

Export a self-contained educational report from one retained snapshot.

Every planned attempt remains visible, including rejected, failed, active, interrupted and unreached units. Each declared prefix receives a row even when its effect is unavailable. The embedded Plotly figures and all tables use the supplied snapshot; this function performs no store or plan scan.

1.1.2.1 Raises

Name	Type	Description
	ValueError	If the supplied report and typed records disagree on the snapshot identity, planned population, attempts or contrasts.