Skip to content

evaluate_node

junjo.evaluate_node

Kind: Function

SDK version: 0.64.0

Documentation channel: Next source preview

evaluate_node(*, node: Node, store: BaseStore[StateT], correlation: ExecutionCorrelation | None = None) -> NodeEvaluationResult[StateT]

View source

Execute one real Node with production Node and Store lifecycle evidence.

This helper is a one-shot envelope for live application evals. Construct a fresh node and initialized store for every case. Junjo executes them inside a generated single-Node Workflow so tracing, state transitions, lifecycle dispatch, cancellation, failures, and execution correlation use the same public runtime as production Workflows.

Junjo intentionally does not own datasets, judges, rubrics, thresholds, or reports. Workflow and Agent evals should call their normal execute() methods directly.

Name Type Description Default
node Node Fresh Node instance to evaluate.
store BaseStore[StateT] Fresh initialized Store containing the eval case input.
correlation ExecutionCorrelation | None Optional trusted application identity for this eval run. None
  • NodeEvaluationResult[StateT]: Detached final state and exact execution identities.