0xrsydn/sts2-bot
1
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

feat(policy): add opt-in composite reward scoring with audit tests

Browse source
This commit is contained in:
0xrsydn 2026-09-23 14:17:38 +07:00
commit 32359c1fe7
5 changed files with 233 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

3
run.py
View file

@ -128,6 +128,7 @@ def decision_record(step: int, state_type: str, run_state, decision,
"params": decision.params,
"reason": decision.reason,
"confidence": decision.confidence,
"scoring": decision.scoring,
"error": error,
"jev": getattr(client, "last", None),
}
@ -211,7 +212,7 @@ def main() -> int:
"session can contain several runs and the results "
"cannot be attributed to an arm.")
ap.add_argument("--capture-dir", default="capture")
ap.add_argument("--card-skip-policy", choices=("jev", "combined"), default=None,
ap.add_argument("--card-skip-policy", choices=("jev", "combined", "composite"), default=None,
help="how card-reward skips are decided (default: brain's own)")
args = ap.parse_args()
if args.steps < 1 or args.max_jev_errors < 1 or args.max_duplicate_waits < 0: