Interface: SpeculateRefineWinnerConfig<TParentIn, TChildIn, TChildOut, THandle, TVerdict, TCandidate>

Defined in: ast/speculate-capability-node.ts:67

Optional post-arbitration refinement step (refineWinnerWhen) for synthesizing cross-candidate improvements on the winning branch.

Runs refinementExecutable on the winning branch scope while keeping losing candidate branches mounted read-only for git.candidateComparisonTools inspection before final tournament teardown.

Type Parameters

Type Parameter	Default type	Description
TParentIn	unknown	Parent actor input payload type.
TChildIn	unknown	Branch worker input payload type.
TChildOut	unknown	Branch worker output payload type.
THandle	unknown	Runtime handle type of the active BranchableScope.
TVerdict	unknown	Structured verdict output type of the arbiter agent.
TCandidate	unknown	Candidate record type passed to mapInput.

Properties

Property	Modifier	Type	Description	Defined in
mapInput	readonly	BivariantCallback<[parentInput: TParentIn, verdict: TVerdict, candidates: readonly TCandidate[]], TChildIn>	Synchronous mapper constructing the refinement executable's input from (parentInput, verdict, candidates).	ast/speculate-capability-node.ts:82
predicate	readonly	BivariantCallback<[verdict: TVerdict], boolean>	Predicate testing whether the arbiter's verdict requests post-selection refinement.	ast/speculate-capability-node.ts:76
refinementExecutable	readonly	| AgentNode<TChildIn, TChildOut, THandle> | LoopNode<TChildIn, TChildOut, THandle, string>	Agent or loop executable invoked on the winning branch handle while losing branch handles remain readable.	ast/speculate-capability-node.ts:78