Function: executePlan()

function executePlan(
  store: PlanStore,
  planId: string,
  options: RunOptions,
): Promise<RunProjection>;

Defined in: src/batteries/orchestration/executor.ts:157

Execute a plan against external input and return the projection folded from the durable event log. External input is materialised as the entry frame's node_settled before any other node runs, so it is addressable by NodeRef exactly like any other node output and is rebuilt from events on resume.

Parameters

Parameter	Type	Description
store	PlanStore	The plan store backing the run.
planId	string	The id of the plan to execute.
options	RunOptions	Per-run input and override dependencies.

Returns

Promise<RunProjection>

The RunProjection folded from the run's event log.

Remarks

ORDER OF OPERATIONS, because it is observable and load-bearing. Everything that can refuse a request happens BEFORE claimRun: the plan is read and folded, the entry node is located, and RunOptions.input is validated against its DeclaredField[]. Only then is the run claimed.

That ordering is not an optimisation. claimRun is irreversible by design — a plan admits one run EVER and the store exposes no release — so a check that ran after it would burn the plan's only run on a request that never invoked a single tool, leaving it permanently run_already_claimed with clonePlan the only recovery. A rejected request must cost the plan nothing.

So a caller can rely on this: if this function throws on invalid input, the plan is still runnable. Fix the input and call again.

The budget is the PLAN'S bounds.maxSteps, not a library constant — bounds are plan content, digested and approved by the operator. Exhausting it settles the run halted with budget_exhausted{settled}, never process_death: the executor knows why it stopped, and a resume re-reports the same cause rather than looping, because the bound has not changed.

Throws

If RunOptions.input violates the entry node's declared fields — before any claim.