lili 51144f1a3b feat(cheap): T5 Service 壳合成终结事件(run 崩快失败,封 600s 慢失败出口)+ 压缩崩根因坐实与 fail-open 补丁 ...

- 压缩崩根因(venv 2.0.2 代码级证据,4 次实证 'important_discoveries' required):reply 每轮推理前
  compress_context(_agent.py:609)→ generate_structured_output 用 SummarySchema json-schema(五字段全
  required+maxLength,_config.py:11-53)强制 M3 tool_choice 产摘要 → dict 路 jsonschema.validate
  (model/_base.py:566-567)对缺字段/超长抛 ValidationError → 非 retryable 直接 raise(_base.py:426-428)
  → _compress_context_impl 非 overflow 支 raise e from None(_agent.py:461)→ 穿透 reply;ChatService.run
  吞异常不发终结事件(app/_service/_chat.py:166-176)→ driver 600s 慢失败
- ctx_compress_patch:fail-open 包 Agent.compress_context(压缩=旁路优化件,失败只跳过不杀主链;连续 2 次
  断路防每轮白烧压缩计费调用;钉 2.0.2 不改 venv 本体,纪律同 m3_stream_patch);Service 与 CLI 两处 apply
- collector on_reply 增崩溃支:先 flush 收口采集,再 yield 合成 ReplyEndEvent(消费侧先 publish 到 bus →
  driver 立刻收到回合终结按盘面快速失败),原异常原样重抛不掩盖;正常路零行为变化
- 测试 9 用例:成功透传/失败吞掉/2 次断路/成功清零/按实例隔离/apply 幂等;崩溃合成终结+异常上抛+崩溃路
  flush/正常路不追加/合成兜底不遮原异常

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

2026-07-03 14:21:34 -07:00

..

test_a11_m1.py

feat(studio): A11 切片三 M1 前置基座 + M2 worker 判意图

2026-06-29 01:04:55 -07:00

test_a11_m2_classify.py

feat(studio): A11 切片三 M1 前置基座 + M2 worker 判意图

2026-06-29 01:04:55 -07:00

test_a11_m3_deterministic.py

feat(studio): A11 切片三 M5 三结构断言机器门(改对了没/有没有误伤/血缘可查)

2026-06-29 02:34:06 -07:00

test_a11_m4_regenerate.py

feat(studio): A11 切片三 M5 三结构断言机器门(改对了没/有没有误伤/血缘可查)

2026-06-29 02:34:06 -07:00

test_a11_m5_assertions.py

feat(studio): A11 切片三 M5 三结构断言机器门(改对了没/有没有误伤/血缘可查)

2026-06-29 02:34:06 -07:00

test_auto_vs_golden.py

feat(cheap-worker): 同游戏双驱动 play 编排 + auto-vs-golden 批跑(M2 U2)

2026-06-27 08:47:13 -07:00

test_bake_off.py

feat(cheap-worker): 便宜档生成丰富度 + M1 达标门富化重验(切片一 A/B)

2026-06-30 01:39:06 -07:00

test_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

test_budget_gate.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

test_budget_two_stage.py

feat(budget): T3 预算两段式统一(软停线 ¥10/¥50 只许收尾+RMB 硬地板 ×1.5 fail-closed 封上界+CLI/Service 同源工厂) (W-ARCH②,创始人 2026-07-03 裁决)

2026-07-03 14:15:59 -07:00

test_cheap_gates.py

fix(cheap): 阶段一② 终审 merge-before(红线③ unlink×2 防陈旧 verdict 假绿 + setup fail-fast + acquire_ports to_thread + 死码/ReplyEndEvent canary/restricted fail-closed) (切片一 阶段一②/终审fix)

2026-07-02 13:34:57 -07:00

test_cheap_service_app.py

fix(cheap): 阶段一② 终审 merge-before(红线③ unlink×2 防陈旧 verdict 假绿 + setup fail-fast + acquire_ports to_thread + 死码/ReplyEndEvent canary/restricted fail-closed) (切片一 阶段一②/终审fix)

2026-07-02 13:34:57 -07:00

test_cheap_service_driver.py

fix(cheap): 封死 M3 400 流式 tool_calls 聚合炸弹——parallel_tool_calls=False 主防线 + 同 index 异 id 按 id 分桶兜底补丁 (fix400 A/B/C)

2026-07-02 21:09:44 -07:00

test_cheap_trace.py

feat(cheap-worker): 便宜档生成丰富度 + M1 达标门富化重验(切片一 A/B)

2026-06-30 01:39:06 -07:00

test_cheap_verify.py

feat(cheap): 非遗品类 rubric 接线——heritage.rubric.json 规整为约定形状 + _template-feiyi 登记 (rubric 挂点归一②)

2026-07-03 00:08:22 -07:00

test_compare_multi.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

test_compare.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

test_ctx_compress_patch.py

feat(cheap): T5 Service 壳合成终结事件(run 崩快失败,封 600s 慢失败出口)+ 压缩崩根因坐实与 fail-open 补丁

2026-07-03 14:21:34 -07:00

test_dedup.py

feat(cheap-worker): M3b 本机 U1-U3 — 9d trace parity 救活 D11 + D9 vendored + 跨语言反假绿门(切片一 M3b)

2026-06-28 02:17:41 -07:00

test_ensure_play_spec.py

fix(WU-C 5.4): tap-targets harness 认单数/可空 target — 解 M2 Node 退役阻塞

2026-06-27 10:41:16 -07:00

test_genre_puzzle_assets.py

merge: W-GENRE 解谜品类四件套(skill puzzle-game-design + _template-puzzle 9/9 + _fewshot-puzzle 正例 + rubric fixture/金标 spec + 范例签名机器门)——品类路由并集 v1.4.0(剧情6/TRPG7/非遗8/解谜9)

2026-07-02 23:48:41 -07:00

test_genre_route.py

feat(cheap): T4 生产 create 路 genre 最小路由(brief→五类确定性词表·先验序=特异性降序·无命中回退通用) (M3 cutover 放行检查单增补)

2026-07-03 14:17:54 -07:00

test_golden_specs.py

merge: W-GENRE 解谜品类四件套(skill puzzle-game-design + _template-puzzle 9/9 + _fewshot-puzzle 正例 + rubric fixture/金标 spec + 范例签名机器门)——品类路由并集 v1.4.0(剧情6/TRPG7/非遗8/解谜9)

2026-07-02 23:48:41 -07:00

test_m3_stream_patch.py

fix(cheap): 封死 M3 400 流式 tool_calls 聚合炸弹——parallel_tool_calls=False 主防线 + 同 index 异 id 按 id 分桶兜底补丁 (fix400 A/B/C)

2026-07-02 21:09:44 -07:00

test_play_spec_contract.py

feat(cheap): T1 驱动器约定收敛·C5 考卷契约接线(startRitual 起局仪式+driver 显式化+sourceHash 陈旧重生·金标 held-constant 不误杀) (W-S1 修复三单·单①)

2026-07-03 14:02:53 -07:00

test_repair_middleware_spike.py

feat(cheap-worker): 续修 middleware 机制 spike 坐实(拦 finish+注入+续跑)(切片一 阶段一前置)

2026-07-01 05:22:49 -07:00

test_result_out.py

feat(cheap-worker): M3b 本机 U1-U3 — 9d trace parity 救活 D11 + D9 vendored + 跨语言反假绿门(切片一 M3b)

2026-06-28 02:17:41 -07:00

test_roles.py

feat(cheap-worker): 便宜档生成丰富度 + M1 达标门富化重验(切片一 A/B)

2026-06-30 01:39:06 -07:00

test_run.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

test_service_synthetic_reply_end.py

feat(cheap): T5 Service 壳合成终结事件(run 崩快失败,封 600s 慢失败出口)+ 压缩崩根因坐实与 fail-open 补丁

2026-07-03 14:21:34 -07:00

test_studio_smoke.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

test_toolkit.py

feat(cheap-gen): 阶段一B #5 drawButton 改返命中矩形(黄金首件·治 #1 翻车点根因)

2026-06-30 01:39:05 -07:00

test_trace_source.py

feat(cheap-worker): M3b 本机 U1-U3 — 9d trace parity 救活 D11 + D9 vendored + 跨语言反假绿门(切片一 M3b)

2026-06-28 02:17:41 -07:00

test_worker_service.py

feat(cheap-worker): M3b 本机 U1-U3 — 9d trace parity 救活 D11 + D9 vendored + 跨语言反假绿门(切片一 M3b)

2026-06-28 02:17:41 -07:00