NeoSyntropyFramework

Guide / KPI

Score quality without hiding policy

Measure node, group, or workflow outcomes, then use a separate validation gate when a threshold should change execution.

Return a named score and reason

Pythonpython
@functional_kpi_node(id="Completeness", output_key="completeness")
def completeness(ctx):
    present = sum(key in ctx.state for key in ("name", "email", "topic"))
    return KpiResult(
        name="completeness",
        score=present / 3,
        reason=f"present={present}/3",
    )

Calibrate semantic scoring with examples

  • Include one clearly poor, one borderline, and one excellent labeled example.
  • Keep KPI measurement separate from the validation node that enforces a minimum.
  • Track the score distribution before choosing a production threshold.