FitnessEvaluator, ScenarioRunner, WeightedFitnessEvaluator, run_benchmark, and run_batch are exported from the top-level package alongside ApprovalGate and KillSwitch. A test harness has no place in the production surface.
Options:
- Move to an optional
dev / benchmark extra (separate install)
- Move to a separate package entirely
Until then, any consumer doing from agent_control_plane import * or scanning the public API gets benchmark infrastructure mixed in with governance primitives.
FitnessEvaluator, ScenarioRunner, WeightedFitnessEvaluator, run_benchmark, and run_batch are exported from the top-level package alongside ApprovalGate and KillSwitch. A test harness has no place in the production surface.
Options:
dev/benchmarkextra (separate install)Until then, any consumer doing
from agent_control_plane import *or scanning the public API gets benchmark infrastructure mixed in with governance primitives.