orchestrata_volks.harness
Reference › components › volks › orchestrata_volks › harness
Volk Test Harness — scenario-driven evaluation system for autonomous agents.
The Volk Test Harness of orchestrata_volks: a scenario-driven evaluation system for autonomous agents. scenario loads, validates and sets up scenarios; harness_runner (474 lines) runs them through an executor (direct_call_executor or c_x_packet_executor); evaluators in evaluators score the results into scenario_result and trial_result. self_healing_loop retries with healing; history is kept by json_history_tracker and graph_history_tracker with trend_queries. Isolation is incus_isolation, worktree_isolation, isolation_manager and snapshot_manager. reporting formats results, config loads configuration, and cli (429 lines) is the argparse-based command line. See the harness narrative.
42 modules, 4,373 lines of code.
Subpackages
| Package | Modules | LOC |
|---|---|---|
evaluators |
7 | 710 |