Skip to contents

Evaluates the active workflow packs against a derived workflow context and returns obligations, suggested actions, and planner-ready external holds. Callers can pass result$metadata$external_holds into bg_plan() to keep workflow holds distinct from structural blockers.

Usage

bg_next_actions(project, scope = c("project", "branch"), branch_id = NULL)

Arguments

project

A bg_handle.

scope

One of project or branch.

branch_id

Optional branch id, required for branch-scoped queries.

Value

A bg_next_actions_result plain-data list with context, obligations, actions, and metadata$external_holds.

Details

For the built-in bayesgrove.default_bayesian pack, the returned obligations and actions can include:

  • review_computation_validity and matching computation_review actions for fresh warning/error summaries,

  • review_fit_criticism plus fit_criticism and branch_and_modify actions for branch-scoped fit or diagnostic problems,

  • compare_candidate_branches plus comparison-node creation or model_comparison decision actions when multiple clean fit candidates exist, and

  • accept_or_reject_branch plus branch_disposition actions after a current comparison exists for an active candidate set.

When the optional packs are active, the result can also include prior-rationale recording, prior and posterior predictive review, simulation-based calibration review, model-selection review keyed to model_comparison and stacking_weights summaries, causal-question prompts, and PAD annotation review.