lili 0d582b2956 feat(cheap-worker): ≥80% 达标门 bake-off 骨架 + 聚合判定单测(M1 U3) ...

新建 bake_off.py:tap-targets 三品类各 n 款 run_studio(run_gates=True)生成→自动 spec
驱动九门→取 verdict.pass,按品类聚合 pass_count/total、逐品类 ≥0.8 判达标;任一品类缺样本
报覆盖不足、任一 <0.8 整体未达标(绝不总体平均掩盖卡死品类)。判定纯走确定性九门 verdict.pass、
零 LLM。真跑复用 compare_node 端口池+线程前台有界并发(conc≤15)、gameId 品类前缀隔离、报告不覆盖。

验证:test_bake_off.py 11/11(判定 meets/below/insufficient/边界0.8 + 假绿防护不被平均 +
覆盖不足 + 上限夹取 + 零LLM确定性)。20 款 M3 达标门真跑(收口硬判据)待执行。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-27 04:47:14 -07:00

..

fixtures/golden-specs

feat(cheap-worker): U1 三品类金标 play-spec fixtures

2026-06-26 10:43:39 -07:00

tests

feat(cheap-worker): ≥80% 达标门 bake-off 骨架 + 聚合判定单测(M1 U3)

2026-06-27 04:47:14 -07:00

_bootstrap.py

feat(cheap-worker): 便宜档 Node→Python 重写进 AgentScope spike — 范式可移植坐实

2026-06-26 02:42:14 -07:00

bake_off.py

feat(cheap-worker): ≥80% 达标门 bake-off 骨架 + 聚合判定单测(M1 U3)

2026-06-27 04:47:14 -07:00

cheap_roles.py

feat(cheap-worker): 便宜档 Node→Python 重写进 AgentScope spike — 范式可移植坐实

2026-06-26 02:42:14 -07:00

cheap_run.py

feat(cheap-worker): 自动 play-spec 加厚到金标同质 + 防自证(M1 U2)

2026-06-27 04:40:04 -07:00

cheap_studio.py

feat(cheap-worker): 便宜档预算闸 <¥10 fail-closed 接线 + 成本落盘(M1 U4)

2026-06-27 02:01:20 -07:00

cheap_toolkit.py

feat(cheap-worker): 便宜档 Node→Python 重写进 AgentScope spike — 范式可移植坐实

2026-06-26 02:42:14 -07:00

compare_node.py

feat(cheap-worker): compare_node 加有界并发(端口池+线程,可选)

2026-06-26 11:00:00 -07:00

requirements.txt

feat(cheap-worker): 便宜档 Node→Python 重写进 AgentScope spike — 范式可移植坐实

2026-06-26 02:42:14 -07:00