|
23761
|
5
|
5
|
5
|
57
|
0
|
0
|
refs/heads/feat/llm-shooting-script
|
0
|
|
1787029504
|
Edit
Delete
|
|
23758
|
5
|
5
|
5
|
57
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"9af0a397c {"Commits":[{"Sha1":"9af0a397ce69cf03623a495ee3ef49943598f5a4","Message":"Merge origin/main: 双推理路线合流为「本地优先 / 托管兜底」能力路由\n\n远端 PR #1（本地权重真跑，面向 128 GiB 主机）与本地 a3b0b41（托管 DashScope\nQwen-Image 3.0，面向 24 GiB 主机）各自独立实现了改字，在 /v1/edit 与双后端\nrender-pipeline 上正面冲突——同一事实两份运行时口径（违治理原则 1）。\n本次不做二选一，合流为一条按能力路由的通路。\n\ncontracts（路由单源）\n- 新增 RENDER_KIND_FALLBACK_CAPABILITIES + resolveRenderCapabilityKeys()：\n 主能力=本地 qwen-image-edit，兜底=托管 qwen-image-3-api，双后端禁止各自\n 展开再拼一遍降级逻辑\n- 修掉三方自动合并留下的语义漂移：RENDER_KIND_CAPABILITIES 曾被并成\n 「改字 kind 只映射托管」，与本地优先矛盾\n- 新增 LOCAL_MODEL_KINDS；MATERIAL_MODEL_ROUTING 改字标签体现双通路\n- 契约测试同步断言新语义（主/兜底/解析顺序/无兜底 kind）\n\nsidecar\n- run_edit 改为路由版：本地可跑优先走本地（不外传底图、不产生云端调用费），\n 本地不可用才降到托管；显式传 modelKey 则锁定该条路径不做隐式回退\n- HEAD 的托管实现降为 run_hosted_edit() 分支，EditRequest / run_edit 各唯一\n- 两条都不可用 → 503 NO_EDIT_BACKEND_AVAILABLE 且同时带上两侧 capability，\n 禁止静默降级\n\n双后端\n- sidecarEdit 的 modelKey 由硬编码 'qwen-image-edit' 改为路由解析结果，\n 否则小内存主机上托管兜底永远走不到\n- BLOCKED reason 逐条列出每个候选能力为何不可用\n- 清理两处悬空的 hostedEdit 引用（声明在冲突块、使用在自动合并区，\n 靠 typecheck 兜住）——该分支只在 procedural 路径可达，直接调 sidecarRender\n\n前端\n- 血统标注按证据判定：ktv-poster-master.png 由初始化提交 fc2ce50 引入且\n 远端从未改动，远端将其标为「Qwen-Image-2512 产物」属误标，保留\n 「原型演示静态素材」口径（P3：不冒充模型产物）\n- constants.ts 模型口径一律从 MATERIAL_MODEL_ROUTING 派生，不留第二真源\n\n验收（诚实记录）\n- pnpm check 全项 exit 0\n- pnpm check:runtime 真实 DB+Redis 通过（迁移 deploy/status + 双后端测试）\n- pnpm check:inference 在本机失败：权重实体已删除、仅剩 git-lfs 指针，\n sidecar-health 阶段诚实报 PaddleOCR 权重不完整。属环境限制非代码缺陷，\n 已记为 G16，未因此放宽任何断言\n\nCLAUDE.md 回灌：C10 闭环记录、G14 合并两侧真实证据、G16 新增\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-17T21:19:49-07:00"},{"Sha1":"a3b0b41a799077b4ab46b58ee7290ae27be009f3","Message":"feat: integrate Qwen-Image 3.0 hosted edit via DashScope\n\nQwen-Image 3.0 (released 2026-07-21) ships no public weights (HF-verified;\nlatest open checkpoint remains Qwen-Image-2512), so integration goes through\nthe Aliyun Model Studio hosted API instead of the weights manifest.\n\n- contracts: add qwen-image-3-api capability (kind=hosted-api), HOSTED_EDIT_KINDS,\n buildEditInstruction (prompt stays aligned with OCR gate expectations),\n sidecarCapabilitySchema.kind; fix lingering \"Qwen-Image 2.0\" routing drift\n- sidecar: /v1/edit now really calls DashScope multimodal-generation\n (base64 base image, \u003e2048px downscale, 24h result URL downloaded to disk\n immediately, honest 503 HOSTED_API_KEY_MISSING / 502 HOSTED_API_ERROR;\n zero new Python deps)\n- backends (NestJS+Fastify, isomorphic): render pipeline executes hosted edit\n kinds when credential ready -\u003e Asset(RENDERED, model=real engine id) -\u003e\n server-side OCR gate; other model kinds stay honestly BLOCKED\n- web: EditStudio hosted-edit entry (run-hosted-edit); MODEL_ROUTING_FALLBACK\n now derived from contracts; purge remaining \"Qwen-Image 2.0\" labels\n- acceptance: deterministic no-key BLOCKED assertions + runIf(DASHSCOPE_API_KEY)\n live test; 20B BLOCKED assertion moved to always-deterministic DESIGN_MASTER\n- gates re-run green: pnpm check / check:runtime / check:inference / check:ui\n (refreshed the stale failed ui-acceptance report left by a July port collision)\n- docs: CLAUDE.md C9 closed record, G14 rewrite, baseline refresh\n\nCo-Authored-By: Claude Fable 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-12T22:57:17-07:00"},{"Sha1":"34dc595ae157cf7904da55c885aee4af2c69c469","Message":"chore: add material factory model weights\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-06T02:21:25-07:00"}],"HeadCommit":{"Sha1":"9af0a397ce69cf03623a495ee3ef49943598f5a4","Message":"Merge origin/main: 双推理路线合流为「本地优先 / 托管兜底」能力路由\n\n远端 PR #1（本地权重真跑，面向 128 GiB 主机）与本地 a3b0b41（托管 DashScope\nQwen-Image 3.0，面向 24 GiB 主机）各自独立实现了改字，在 /v1/edit 与双后端\nrender-pipeline 上正面冲突——同一事实两份运行时口径（违治理原则 1）。\n本次不做二选一，合流为一条按能力路由的通路。\n\ncontracts（路由单源）\n- 新增 RENDER_KIND_FALLBACK_CAPABILITIES + resolveRenderCapabilityKeys()：\n 主能力=本地 qwen-image-edit，兜底=托管 qwen-image-3-api，双后端禁止各自\n 展开再拼一遍降级逻辑\n- 修掉三方自动合并留下的语义漂移：RENDER_KIND_CAPABILITIES 曾被并成\n 「改字 kind 只映射托管」，与本地优先矛盾\n- 新增 LOCAL_MODEL_KINDS；MATERIAL_MODEL_ROUTING 改字标签体现双通路\n- 契约测试同步断言新语义（主/兜底/解析顺序/无兜底 kind）\n\nsidecar\n- run_edit 改为路由版：本地可跑优先走本地（不外传底图、不产生云端调用费），\n 本地不可用才降到托管；显式传 modelKey 则锁定该条路径不做隐式回退\n- HEAD 的托管实现降为 run_hosted_edit() 分支，EditRequest / run_edit 各唯一\n- 两条都不可用 → 503 NO_EDIT_BACKEND_AVAILABLE 且同时带上两侧 capability，\n 禁止静默降级\n\n双后端\n- sidecarEdit 的 modelKey 由硬编码 'qwen-image-edit' 改为路由解析结果，\n 否则小内存主机上托管兜底永远走不到\n- BLOCKED reason 逐条列出每个候选能力为何不可用\n- 清理两处悬空的 hostedEdit 引用（声明在冲突块、使用在自动合并区，\n 靠 typecheck 兜住）——该分支只在 procedural 路径可达，直接调 sidecarRender\n\n前端\n- 血统标注按证据判定：ktv-poster-master.png 由初始化提交 fc2ce50 引入且\n 远端从未改动，远端将其标为「Qwen-Image-2512 产物」属误标，保留\n 「原型演示静态素材」口径（P3：不冒充模型产物）\n- constants.ts 模型口径一律从 MATERIAL_MODEL_ROUTING 派生，不留第二真源\n\n验收（诚实记录）\n- pnpm check 全项 exit 0\n- pnpm check:runtime 真实 DB+Redis 通过（迁移 deploy/status + 双后端测试）\n- pnpm check:inference 在本机失败：权重实体已删除、仅剩 git-lfs 指针，\n sidecar-health 阶段诚实报 PaddleOCR 权重不完整。属环境限制非代码缺陷，\n 已记为 G16，未因此放宽任何断言\n\nCLAUDE.md 回灌：C10 闭环记录、G14 合并两侧真实证据、G16 新增\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-17T21:19:49-07:00"},"CompareURL":"luoanwu/image-generation/compare/27c04f193a6a4170b1e0f7839a18d0e7e9b529fb...9af0a397ce69cf03623a495ee3ef49943598f5a4","Len":3}...
|
1787026816
|
Edit
Delete
|
|
23586
|
5
|
5
|
7
|
57
|
0
|
0
|
refs/heads/hljTest
|
0
|
{"Commits":[{"Sha1":"549ea3144 {"Commits":[{"Sha1":"549ea31448ec1e62e8f10039ed3f8c7d384e07f8","Message":"feat(material-factory): 接入阿里云百炼真实生成链路\n\nCo-authored-by: Qwen-Coder \u003cqwen-coder@alibabacloud.com\u003e\n","AuthorEmail":"hlj@g-hi.com","AuthorName":"huangliujian","CommitterEmail":"hlj@g-hi.com","CommitterName":"huangliujian","Timestamp":"2026-08-17T17:34:04+08:00"}],"HeadCommit":{"Sha1":"549ea31448ec1e62e8f10039ed3f8c7d384e07f8","Message":"feat(material-factory): 接入阿里云百炼真实生成链路\n\nCo-authored-by: Qwen-Coder \u003cqwen-coder@alibabacloud.com\u003e\n","AuthorEmail":"hlj@g-hi.com","AuthorName":"huangliujian","CommitterEmail":"hlj@g-hi.com","CommitterName":"huangliujian","Timestamp":"2026-08-17T17:34:04+08:00"},"CompareURL":"luoanwu/image-generation/compare/27c04f193a6a4170b1e0f7839a18d0e7e9b529fb...549ea31448ec1e62e8f10039ed3f8c7d384e07f8","Len":1}...
|
1786959245
|
Edit
Delete
|
|
23584
|
5
|
5
|
7
|
57
|
0
|
0
|
refs/heads/hljTest
|
0
|
|
1786959245
|
Edit
Delete
|
|
22305
|
5
|
7
|
5
|
57
|
0
|
0
|
|
0
|
2|chore: add material factory model weights
|
1786603029
|
Edit
Delete
|
|
22304
|
5
|
5
|
5
|
57
|
0
|
0
|
refs/heads/feat/qwen-image-3-hosted
|
0
|
{"Commits":[{"Sha1":"a3b0b41a7 {"Commits":[{"Sha1":"a3b0b41a799077b4ab46b58ee7290ae27be009f3","Message":"feat: integrate Qwen-Image 3.0 hosted edit via DashScope\n\nQwen-Image 3.0 (released 2026-07-21) ships no public weights (HF-verified;\nlatest open checkpoint remains Qwen-Image-2512), so integration goes through\nthe Aliyun Model Studio hosted API instead of the weights manifest.\n\n- contracts: add qwen-image-3-api capability (kind=hosted-api), HOSTED_EDIT_KINDS,\n buildEditInstruction (prompt stays aligned with OCR gate expectations),\n sidecarCapabilitySchema.kind; fix lingering \"Qwen-Image 2.0\" routing drift\n- sidecar: /v1/edit now really calls DashScope multimodal-generation\n (base64 base image, \u003e2048px downscale, 24h result URL downloaded to disk\n immediately, honest 503 HOSTED_API_KEY_MISSING / 502 HOSTED_API_ERROR;\n zero new Python deps)\n- backends (NestJS+Fastify, isomorphic): render pipeline executes hosted edit\n kinds when credential ready -\u003e Asset(RENDERED, model=real engine id) -\u003e\n server-side OCR gate; other model kinds stay honestly BLOCKED\n- web: EditStudio hosted-edit entry (run-hosted-edit); MODEL_ROUTING_FALLBACK\n now derived from contracts; purge remaining \"Qwen-Image 2.0\" labels\n- acceptance: deterministic no-key BLOCKED assertions + runIf(DASHSCOPE_API_KEY)\n live test; 20B BLOCKED assertion moved to always-deterministic DESIGN_MASTER\n- gates re-run green: pnpm check / check:runtime / check:inference / check:ui\n (refreshed the stale failed ui-acceptance report left by a July port collision)\n- docs: CLAUDE.md C9 closed record, G14 rewrite, baseline refresh\n\nCo-Authored-By: Claude Fable 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-12T22:57:17-07:00"},{"Sha1":"34dc595ae157cf7904da55c885aee4af2c69c469","Message":"chore: add material factory model weights\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-06T02:21:25-07:00"}],"HeadCommit":{"Sha1":"a3b0b41a799077b4ab46b58ee7290ae27be009f3","Message":"feat: integrate Qwen-Image 3.0 hosted edit via DashScope\n\nQwen-Image 3.0 (released 2026-07-21) ships no public weights (HF-verified;\nlatest open checkpoint remains Qwen-Image-2512), so integration goes through\nthe Aliyun Model Studio hosted API instead of the weights manifest.\n\n- contracts: add qwen-image-3-api capability (kind=hosted-api), HOSTED_EDIT_KINDS,\n buildEditInstruction (prompt stays aligned with OCR gate expectations),\n sidecarCapabilitySchema.kind; fix lingering \"Qwen-Image 2.0\" routing drift\n- sidecar: /v1/edit now really calls DashScope multimodal-generation\n (base64 base image, \u003e2048px downscale, 24h result URL downloaded to disk\n immediately, honest 503 HOSTED_API_KEY_MISSING / 502 HOSTED_API_ERROR;\n zero new Python deps)\n- backends (NestJS+Fastify, isomorphic): render pipeline executes hosted edit\n kinds when credential ready -\u003e Asset(RENDERED, model=real engine id) -\u003e\n server-side OCR gate; other model kinds stay honestly BLOCKED\n- web: EditStudio hosted-edit entry (run-hosted-edit); MODEL_ROUTING_FALLBACK\n now derived from contracts; purge remaining \"Qwen-Image 2.0\" labels\n- acceptance: deterministic no-key BLOCKED assertions + runIf(DASHSCOPE_API_KEY)\n live test; 20B BLOCKED assertion moved to always-deterministic DESIGN_MASTER\n- gates re-run green: pnpm check / check:runtime / check:inference / check:ui\n (refreshed the stale failed ui-acceptance report left by a July port collision)\n- docs: CLAUDE.md C9 closed record, G14 rewrite, baseline refresh\n\nCo-Authored-By: Claude Fable 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-12T22:57:17-07:00"},"CompareURL":"luoanwu/image-generation/compare/fc2ce503c94d763b1d077e7c734720a0c58e23a8...a3b0b41a799077b4ab46b58ee7290ae27be009f3","Len":2}...
|
1786602934
|
Edit
Delete
|
|
22303
|
5
|
5
|
5
|
57
|
0
|
0
|
refs/heads/feat/qwen-image-3-hosted
|
0
|
|
1786602934
|
Edit
Delete
|
|
16224
|
5
|
5
|
5
|
57
|
0
|
0
|
refs/heads/main
|
1
|
{"Commits":[{"Sha1":"27c04f193 {"Commits":[{"Sha1":"27c04f193a6a4170b1e0f7839a18d0e7e9b529fb","Message":"Merge pull request 'feat(material-factory): 图像模型真实推理对接 + 双后端渲染管线闭环' (#1) from feat/qwen-image-inference into main\n\nReviewed-on: https://gitea.g-hi.com/luoanwu/image-generation/pulls/1\n","AuthorEmail":"law@g-hi.com","AuthorName":"luoanwu","CommitterEmail":"law@g-hi.com","CommitterName":"luoanwu","Timestamp":"2026-07-07T07:46:33+08:00"},{"Sha1":"ac5f340c0422af2a5c3122289142e8517dc434c5","Message":"fix(sidecar): Qwen-Image-Edit 未指定尺寸时保留底图原生分辨率\n\nQwenImageEditPlus 默认把输入塌到 ~1MP 规范面积，密集中文标题在 1MP + 低步数下\n笔画糊/字形崩。新增 _fit_edit_dims：未显式传 width/height 时按底图原生分辨率\n（保纵横比、向下取整 16 倍数、封顶 QWEN_EDIT_MAX_AREA=1.5MP 兜内存/耗时）驱动，\n调用方仍可显式覆盖。\n\n实测 ktv-poster-master(1672x941) 改字：4 步@1.06MP 文字 melted → 20 步@1.49MP\n（1632x912）文字笔画完整清晰、字形正确，无高分辨率伪影。清晰度决定因素是步数\n（生产默认 40），分辨率为次因；两者叠加达到可用海报质量。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luojuhai@luojuhaideMacBook-Pro.local","AuthorName":"luojuhai","CommitterEmail":"luojuhai@luojuhaideMacBook-Pro.local","CommitterName":"luojuhai","Timestamp":"2026-07-07T07:39:05+08:00"},{"Sha1":"973dde30db19928a0937f02a102979d473cce644","Message":"feat(material-factory): 图像模型真实推理对接 + 双后端渲染管线闭环\n\n把嗨设工坊从「权重就绪、推理是桩」推进到「本机真跑转绿」：\n\nsidecar (services/model-sidecar)\n- /v1/generate 真调 Z-Image-Turbo / Qwen-Image-2512(modelKey 路由)\n- /v1/edit 真调 Qwen-Image-Edit-2511(QwenImageEditPlusPipeline)\n- capability() 依赖齐全时开启 qwen-image-master/edit（128 GiB 主机 loadable）\n- 单 pipeline 常驻、kind 切换释放旧模型，防双 20B 同驻 OOM\n- 修 torchvision 隐藏依赖：Qwen2VLProcessor→Qwen2VLVideoProcessor 硬依赖它，\n 缺失则 from_pretrained ImportError（进 pyproject zimage extra）\n- 加载失败包 try/except → 结构化 503（不再冒 500）\n- SIDECAR_QWEN_STEPS：生产默认 40 步，CI/验收可降步提速（真引擎真像素）\n\n双后端 (NestJS + Fastify 同契约)\n- render-pipeline 按 kind 路由：APPEARANCE/SEMANTIC_EDIT→sidecarEdit，\n DESIGN_MASTER/TURBO_VARIANT→sidecarGenerate；job 状态机真流转 + 乐观锁 + outbox 同 tx\n- 修 undici 300s 传输超时：Node 内置 fetch 的 headersTimeout/bodyTimeout 默认 300s\n 会把分钟级同步推理在 ~302s 中止成假 SidecarUnavailableError（AbortSignal 压不住）。\n 改用 undici@6 request() + 请求级 headersTimeout/bodyTimeout 对齐 timeoutMs\n\n验收\n- DATABASE_URL=... INFERENCE_RUN_QWEN=1 pnpm check:inference 双后端各 6/6 绿\n- Qwen-Image-Edit 真改图 SUCCEEDED（金色中文标题精确改写 + OCR gate 联动）\n- pnpm check 全部静态门禁绿；CLAUDE.md(=AGENTS.md) 回灌 G14/真跑证据\n\n仍打开(诚实缺口)：render-pipeline 同步端点(G13)、异步 worker 化、\n单 MPS 多进程串行化、Qwen nightly/self-hosted 验收。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luojuhai@luojuhaideMacBook-Pro.local","AuthorName":"luojuhai","CommitterEmail":"luojuhai@luojuhaideMacBook-Pro.local","CommitterName":"luojuhai","Timestamp":"2026-07-07T06:35:04+08:00"}],"HeadCommit":{"Sha1":"27c04f193a6a4170b1e0f7839a18d0e7e9b529fb","Message":"Merge pull request 'feat(material-factory): 图像模型真实推理对接 + 双后端渲染管线闭环' (#1) from feat/qwen-image-inference into main\n\nReviewed-on: https://gitea.g-hi.com/luoanwu/image-generation/pulls/1\n","AuthorEmail":"law@g-hi.com","AuthorName":"luoanwu","CommitterEmail":"law@g-hi.com","CommitterName":"luoanwu","Timestamp":"2026-07-07T07:46:33+08:00"},"CompareURL":"luoanwu/image-generation/compare/fc2ce503c94d763b1d077e7c734720a0c58e23a8...27c04f193a6a4170b1e0f7839a18d0e7e9b529fb","Len":3}...
|
1783381594
|
Edit
Delete
|
|
16223
|
5
|
11
|
5
|
57
|
0
|
0
|
|
1
|
1|feat(material-factory): 图像模型真实推理对接 + 双后端渲染管线闭环
|
1783381594
|
Edit
Delete
|
|
16222
|
5
|
7
|
5
|
57
|
0
|
0
|
|
1
|
1|feat(material-factory): 图像模型真实推理对接 + 双后端渲染管线闭环
|
1783381429
|
Edit
Delete
|
|
16221
|
5
|
5
|
5
|
57
|
0
|
0
|
refs/heads/feat/qwen-image-inference
|
1
|
{"Commits":[{"Sha1":"ac5f340c0 {"Commits":[{"Sha1":"ac5f340c0422af2a5c3122289142e8517dc434c5","Message":"fix(sidecar): Qwen-Image-Edit 未指定尺寸时保留底图原生分辨率\n\nQwenImageEditPlus 默认把输入塌到 ~1MP 规范面积，密集中文标题在 1MP + 低步数下\n笔画糊/字形崩。新增 _fit_edit_dims：未显式传 width/height 时按底图原生分辨率\n（保纵横比、向下取整 16 倍数、封顶 QWEN_EDIT_MAX_AREA=1.5MP 兜内存/耗时）驱动，\n调用方仍可显式覆盖。\n\n实测 ktv-poster-master(1672x941) 改字：4 步@1.06MP 文字 melted → 20 步@1.49MP\n（1632x912）文字笔画完整清晰、字形正确，无高分辨率伪影。清晰度决定因素是步数\n（生产默认 40），分辨率为次因；两者叠加达到可用海报质量。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luojuhai@luojuhaideMacBook-Pro.local","AuthorName":"luojuhai","CommitterEmail":"luojuhai@luojuhaideMacBook-Pro.local","CommitterName":"luojuhai","Timestamp":"2026-07-07T07:39:05+08:00"},{"Sha1":"973dde30db19928a0937f02a102979d473cce644","Message":"feat(material-factory): 图像模型真实推理对接 + 双后端渲染管线闭环\n\n把嗨设工坊从「权重就绪、推理是桩」推进到「本机真跑转绿」：\n\nsidecar (services/model-sidecar)\n- /v1/generate 真调 Z-Image-Turbo / Qwen-Image-2512(modelKey 路由)\n- /v1/edit 真调 Qwen-Image-Edit-2511(QwenImageEditPlusPipeline)\n- capability() 依赖齐全时开启 qwen-image-master/edit（128 GiB 主机 loadable）\n- 单 pipeline 常驻、kind 切换释放旧模型，防双 20B 同驻 OOM\n- 修 torchvision 隐藏依赖：Qwen2VLProcessor→Qwen2VLVideoProcessor 硬依赖它，\n 缺失则 from_pretrained ImportError（进 pyproject zimage extra）\n- 加载失败包 try/except → 结构化 503（不再冒 500）\n- SIDECAR_QWEN_STEPS：生产默认 40 步，CI/验收可降步提速（真引擎真像素）\n\n双后端 (NestJS + Fastify 同契约)\n- render-pipeline 按 kind 路由：APPEARANCE/SEMANTIC_EDIT→sidecarEdit，\n DESIGN_MASTER/TURBO_VARIANT→sidecarGenerate；job 状态机真流转 + 乐观锁 + outbox 同 tx\n- 修 undici 300s 传输超时：Node 内置 fetch 的 headersTimeout/bodyTimeout 默认 300s\n 会把分钟级同步推理在 ~302s 中止成假 SidecarUnavailableError（AbortSignal 压不住）。\n 改用 undici@6 request() + 请求级 headersTimeout/bodyTimeout 对齐 timeoutMs\n\n验收\n- DATABASE_URL=... INFERENCE_RUN_QWEN=1 pnpm check:inference 双后端各 6/6 绿\n- Qwen-Image-Edit 真改图 SUCCEEDED（金色中文标题精确改写 + OCR gate 联动）\n- pnpm check 全部静态门禁绿；CLAUDE.md(=AGENTS.md) 回灌 G14/真跑证据\n\n仍打开(诚实缺口)：render-pipeline 同步端点(G13)、异步 worker 化、\n单 MPS 多进程串行化、Qwen nightly/self-hosted 验收。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luojuhai@luojuhaideMacBook-Pro.local","AuthorName":"luojuhai","CommitterEmail":"luojuhai@luojuhaideMacBook-Pro.local","CommitterName":"luojuhai","Timestamp":"2026-07-07T06:35:04+08:00"}],"HeadCommit":{"Sha1":"ac5f340c0422af2a5c3122289142e8517dc434c5","Message":"fix(sidecar): Qwen-Image-Edit 未指定尺寸时保留底图原生分辨率\n\nQwenImageEditPlus 默认把输入塌到 ~1MP 规范面积，密集中文标题在 1MP + 低步数下\n笔画糊/字形崩。新增 _fit_edit_dims：未显式传 width/height 时按底图原生分辨率\n（保纵横比、向下取整 16 倍数、封顶 QWEN_EDIT_MAX_AREA=1.5MP 兜内存/耗时）驱动，\n调用方仍可显式覆盖。\n\n实测 ktv-poster-master(1672x941) 改字：4 步@1.06MP 文字 melted → 20 步@1.49MP\n（1632x912）文字笔画完整清晰、字形正确，无高分辨率伪影。清晰度决定因素是步数\n（生产默认 40），分辨率为次因；两者叠加达到可用海报质量。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luojuhai@luojuhaideMacBook-Pro.local","AuthorName":"luojuhai","CommitterEmail":"luojuhai@luojuhaideMacBook-Pro.local","CommitterName":"luojuhai","Timestamp":"2026-07-07T07:39:05+08:00"},"CompareURL":"luoanwu/image-generation/compare/fc2ce503c94d763b1d077e7c734720a0c58e23a8...ac5f340c0422af2a5c3122289142e8517dc434c5","Len":2}...
|
1783381399
|
Edit
Delete
|
|
16220
|
5
|
5
|
5
|
57
|
0
|
0
|
refs/heads/feat/qwen-image-inference
|
1
|
|
1783381399
|
Edit
Delete
|
|
16153
|
5
|
5
|
5
|
57
|
0
|
0
|
refs/heads/main
|
1
|
{"Commits":[{"Sha1":"fc2ce503c {"Commits":[{"Sha1":"fc2ce503c94d763b1d077e7c734720a0c58e23a8","Message":"feat: initialize HI Atelier material factory\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-06T01:30:42-07:00"}],"HeadCommit":{"Sha1":"fc2ce503c94d763b1d077e7c734720a0c58e23a8","Message":"feat: initialize HI Atelier material factory\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-06T01:30:42-07:00"},"CompareURL":"","Len":1}...
|
1783328698
|
Edit
Delete
|
|
16152
|
5
|
5
|
5
|
57
|
0
|
0
|
refs/heads/main
|
1
|
|
1783328698
|
Edit
Delete
|
|
16151
|
5
|
1
|
5
|
57
|
0
|
0
|
|
1
|
|
1783328662
|
Edit
Delete
|
|
16117
|
5
|
5
|
5
|
56
|
0
|
0
|
refs/heads/main
|
1
|
{"Commits":[{"Sha1":"e65fc3036 {"Commits":[{"Sha1":"e65fc3036083a4118e0c909bd6a58914e6b33fef","Message":"feat: initialize SiteScout workspace\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-06T01:31:18-07:00"}],"HeadCommit":{"Sha1":"e65fc3036083a4118e0c909bd6a58914e6b33fef","Message":"feat: initialize SiteScout workspace\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-06T01:31:18-07:00"},"CompareURL":"","Len":1}...
|
1783326783
|
Edit
Delete
|
|
16116
|
5
|
5
|
5
|
56
|
0
|
0
|
refs/heads/main
|
1
|
|
1783326783
|
Edit
Delete
|
|
16115
|
5
|
1
|
5
|
56
|
0
|
0
|
|
1
|
|
1783326759
|
Edit
Delete
|
|
16106
|
5
|
5
|
5
|
55
|
0
|
0
|
refs/heads/main
|
1
|
{"Commits":[{"Sha1":"8b9d81d39 {"Commits":[{"Sha1":"8b9d81d390dac8db76b755c9d7a25f7439e2a59b","Message":"chore: initialize data analysis scaffold\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-06T01:29:19-07:00"}],"HeadCommit":{"Sha1":"8b9d81d390dac8db76b755c9d7a25f7439e2a59b","Message":"chore: initialize data analysis scaffold\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-06T01:29:19-07:00"},"CompareURL":"","Len":1}...
|
1783326647
|
Edit
Delete
|
|
16105
|
5
|
5
|
5
|
55
|
0
|
0
|
refs/heads/main
|
1
|
|
1783326647
|
Edit
Delete
|
|
16104
|
5
|
1
|
5
|
55
|
0
|
0
|
|
1
|
|
1783326636
|
Edit
Delete
|
|
27546
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"06b8b98f4 {"Commits":[{"Sha1":"06b8b98f483723103dcd01449e6ec6772ad10c13","Message":"docs(governance): 回灌远端绿盘——f17d34a 的 GitHub run 33707057068 三 job 全绿，两远端对齐\n\nCo-Authored-By: Claude Fable 5.1 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-02T19:23:08-07:00"}],"HeadCommit":{"Sha1":"06b8b98f483723103dcd01449e6ec6772ad10c13","Message":"docs(governance): 回灌远端绿盘——f17d34a 的 GitHub run 33707057068 三 job 全绿，两远端对齐\n\nCo-Authored-By: Claude Fable 5.1 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-02T19:23:08-07:00"},"CompareURL":"luoanwu/base-framework/compare/f17d34a463e70366161f4603e53a4b2800e9a241...06b8b98f483723103dcd01449e6ec6772ad10c13","Len":1}...
|
1788402192
|
Edit
Delete
|
|
27545
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"f17d34a46 {"Commits":[{"Sha1":"f17d34a463e70366161f4603e53a4b2800e9a241","Message":"fix(gate): G15-B 独占探针真关闭 + 父子探针互认——首个含 G15-B 的远端 run 红在差分 preflight（C23）\n\n推送 7135258 后 GitHub run 33706371355：Static governance 绿，Runtime and UI acceptance 在\ncheck:runtime 第 4 步 conformance-differential 红：「验收 Redis 逻辑库 0 已有 1 个外部连接：\n172.18.0.1:53838 cmd=client|list」。那条连接是顶层 runner 自己几秒前的独占探针：redisExec 收完\n回复只 socket.end() 半关闭，GitHub 的 Redis service 经 docker 网桥暴露，服务端连接残留，子门禁\n把它当成外部占用。本机探针不 SELECT 始终落 db0、验收用 db9，所以三级全绿也撞不上。\n\n- redisExec：队尾追加 QUIT，收到回复后 destroy()；服务端先关也按成功收口\n- 探针 CLIENT SETNAME governance-acceptance-probe-\u003cpid\u003e，foreignClientsOnDb 按前缀排除，\n 父 runner 与派生子门禁各探一次时互认（名字不带项目名，派生仓零改动沿用）\n- 本机 db10 五项正/负向：无名残留连接仍被拒、具名探针被忽略、探针返回后 CLIENT LIST 不再列出自身、\n 解析器只保留真 worker；本机 check:runtime 冒烟 207 tests 通过\n- doctrine：发布状态快照更新为两远端 7135258、G14 登记第四条只有远端才看得见的缺陷、ci-gate 行、C23 索引\n\nCo-Authored-By: Claude Fable 5.1 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-02T19:17:03-07:00"}],"HeadCommit":{"Sha1":"f17d34a463e70366161f4603e53a4b2800e9a241","Message":"fix(gate): G15-B 独占探针真关闭 + 父子探针互认——首个含 G15-B 的远端 run 红在差分 preflight（C23）\n\n推送 7135258 后 GitHub run 33706371355：Static governance 绿，Runtime and UI acceptance 在\ncheck:runtime 第 4 步 conformance-differential 红：「验收 Redis 逻辑库 0 已有 1 个外部连接：\n172.18.0.1:53838 cmd=client|list」。那条连接是顶层 runner 自己几秒前的独占探针：redisExec 收完\n回复只 socket.end() 半关闭，GitHub 的 Redis service 经 docker 网桥暴露，服务端连接残留，子门禁\n把它当成外部占用。本机探针不 SELECT 始终落 db0、验收用 db9，所以三级全绿也撞不上。\n\n- redisExec：队尾追加 QUIT，收到回复后 destroy()；服务端先关也按成功收口\n- 探针 CLIENT SETNAME governance-acceptance-probe-\u003cpid\u003e，foreignClientsOnDb 按前缀排除，\n 父 runner 与派生子门禁各探一次时互认（名字不带项目名，派生仓零改动沿用）\n- 本机 db10 五项正/负向：无名残留连接仍被拒、具名探针被忽略、探针返回后 CLIENT LIST 不再列出自身、\n 解析器只保留真 worker；本机 check:runtime 冒烟 207 tests 通过\n- doctrine：发布状态快照更新为两远端 7135258、G14 登记第四条只有远端才看得见的缺陷、ci-gate 行、C23 索引\n\nCo-Authored-By: Claude Fable 5.1 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-02T19:17:03-07:00"},"CompareURL":"luoanwu/base-framework/compare/7135258b78d5b623099713b2bd3292af46b6329d...f17d34a463e70366161f4603e53a4b2800e9a241","Len":1}...
|
1788401827
|
Edit
Delete
|
|
27544
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"7135258b7 {"Commits":[{"Sha1":"7135258b78d5b623099713b2bd3292af46b6329d","Message":"chore(reports): refresh clean static evidence @ 286eb95（14 项门禁 + kernel profile）\n\nCo-Authored-By: Claude Fable 5.1 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-02T18:59:16-07:00"},{"Sha1":"286eb956f5ee3634246ef80235af45a628635af9","Message":"chore(reports): 刷新三级证据至 1245400（clean；runtime 207 / UI 9 / 差分 8×2 零差异）+ 地板 testsPassed 203→207、uiTestsPassed 8→9\n\n动态区 runtime/UI 行回灌到 clean @ 1245400；own-test-cases 203。同日 aeed741 顶层 runtime 红 5 例\n已由 C22 修复并在 1245400 重跑转绿。\n\nCo-Authored-By: Claude Fable 5.1 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-02T18:59:08-07:00"},{"Sha1":"1245400747afade64c6cda83d362021e0b9d69c2","Message":"fix(realtime): Redis pub/sub 连接 ready 后才交付——修复非 0 逻辑库上订阅约 50% 建不起来（C22）\n\naeed741 的 clean-SHA 顶层 check:runtime 红 5 例（NestJS realtime.command/subscription\n`waitUntil timed out`），单独复跑 3 轮中 2 轮仍红，不是并行噪声。`DEBUG=ioredis:*` 抓到：\nSUB 连接裸建（lazyConnect:false）后立即被 EventBusService.onModuleInit `subscribe()`，\nioredis 立刻置订阅模式并把 subscribe 排进离线队列，握手期 `SELECT 9` / `INFO` 就绪检查被\n客户端以 \"Connection in subscriber mode\" 拒绝 → 判致命错误重连 → 订阅在窗口内从未建立。\n只有 REDIS_URL 带非 0 db 才发 SELECT，CI（db0）从不复现，本机验收切到 db9 后开始「偶发」。\nFastify plugins/redis.ts 同一形态（窗口更窄）。\n\n- 两后端 pub/sub 改 lazyConnect + 显式 await connect()（NestJS 异步 useFactory / Fastify\n Promise.all），ready 后才交付；启动期 Redis 不可达即拒绝启动；挂 error 监听进结构化日志\n- connectionName = api-\u003cbackend\u003e-realtime-\u003crole\u003e-\u003cpid\u003e，CLIENT LIST 可按角色定位\n- 两后端 http.acceptance 各 +1：CLIENT LIST 断言逻辑库 = REDIS_URL 且 SUB sub=1、status ready\n- 证据：修复后 realtime.command 定向 10/10 绿、0 条 ioredis 未处理错误；负向改回旧写法\n readiness 用例 5 轮中 2 轮红（expected 'reconnecting' to be 'ready'）\n- G18 复跑判责纪律补：偶发红必须至少抓一次协议级失败现场\n\nCo-Authored-By: Claude Fable 5.1 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-02T18:55:51-07:00"},{"Sha1":"aeed74154025d55fe8e9f481d473c153de92dd63","Message":"feat(users): 唯一键冲突映射为稳定 409 DUPLICATE_EMAIL + G21 多代理适配器同源门禁\n\nC21：`@@unique([tenantId,email])` 的 Prisma P2002 此前在 NestJS/Fastify 都没有映射，\n重复邮箱直接 500（修复前两后端新用例均 `expected 500 to be 409` 实锤）；NestJS users 404\n还回显资源 id 与框架元数据，与 orders/Fastify 不同体。\n- contracts 新增 `USER_ERROR_CODES.DUPLICATE_EMAIL`（单源）\n- 两后端幂等窄捕获只认 P2002 → 409 `{ message, code }`；Fastify 伪造 P2025 裸 Error 改为\n 类型化 UserNotFoundError/DuplicateUserEmailError；NestJS users 404 改稳定体\n- 两后端 http.acceptance 各 +1 用例（POST/PATCH 409、outbox 无副作用、他租户同 email 201）\n- conformance 矩阵 +HTTP-CONFLICT-001（8→9）、差分 +DIFF-USER-CONFLICT-001（7→8，0 differences）\n- UI +1 用例：重复邮箱在真实页面显示稳定错误、列表不出现第二条\n\nG21：新增 `pnpm check:agent-adapters`（接入 pnpm check / Kernel profile / 棘轮 / 规则\nGOV-AGENT-ADAPTER-001，registryVersion 3→4）：.claude/.agents 同名技能在 doctrine 名与\n画像路径替换之外逐字节比对，单侧技能须登记理由；补齐 business-module-intake Codex 侧副本，\n技能内对不存在技能的硬引用改为「有则调用、无则手工等价」。负向三条（改一行/删单侧/多文件）均红。\n\n基线收紧：ownTestCases 198→201、conformanceMatrixCases 8→9、新增 agentAdapter* 两项。\nruntime/UI 地板待 clean SHA 实跑后回灌。\n\nCo-Authored-By: Claude Fable 5.1 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-02T18:22:32-07:00"},{"Sha1":"269497f441cafaa4501126c4632b5f02ca513233","Message":"chore(reports): refresh clean static evidence\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-02T09:59:04-07:00"}],"HeadCommit":{"Sha1":"7135258b78d5b623099713b2bd3292af46b6329d","Message":"chore(reports): refresh clean static evidence @ 286eb95（14 项门禁 + kernel profile）\n\nCo-Authored-By: Claude Fable 5.1 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-02T18:59:16-07:00"},"CompareURL":"luoanwu/base-framework/compare/b1695a6780efbc95449a553d313599695c0b22ab...7135258b78d5b623099713b2bd3292af46b6329d","Len":12}...
|
1788401211
|
Edit
Delete
|
|
26941
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"b1695a678 {"Commits":[{"Sha1":"b1695a6780efbc95449a553d313599695c0b22ab","Message":"docs(governance): 回灌首份远端 CI 绿盘证据 + 三级证据刷新至 5f15617\n\n远端证据（G14 部分推进，**未闭环**）\nGitHub 远端已加回（github → laoluojuhai/base-framework，私有仓）。\n在 5f15617 上三个 job 全绿：Static governance / Runtime and UI acceptance /\nAggregate same-SHA governance evidence\nrun: https://github.com/laoluojuhai/base-framework/actions/runs/33455966984\n该 run 的 headSha 与本地 HEAD 逐字符对齐。这是本仓历史上第一次远端 CI 绿——\n此前每一次远端运行都是 failure（最近一次 2026-08-26，正是 check:os-product 并入\npnpm check 那一提交）。\n\n三条「只有远端才看得见」的缺陷已在本轮依次修掉，均已单独提交并在本条登记：\n① checkout 浅克隆 → F5 把本仓证据误判为「来自另一个仓库」\n② 差分门禁依赖磁盘残留的 contracts dist（外加 composite 增量陷阱）\n③ 测试数地板正则未剥 ANSI，CI 下恒读 0\n共同点：本地全绿、远端必红——正是 G14「远端拦截未验证」这条缺口存在的意义。\n\n⚠️ ci-gate 仍不标 GREEN：required checks / 分支保护未配置，受控失败 PR 未做。\n「门禁能跑绿」推不出「违规会被挡住」，缺这两项不得宣称 PR 已被机器拦截。\n\n三级证据刷新\n静态 exit 0、runtime 202 tests / 差分 0 differences、UI 8 用例，\n三份报告 provenance 均为 gitSha=5f15617、worktreeDirty=false。\n本轮回灌过程中 freshness-sha-runtime 断言真的挡了我一次（报告已刷到 74b4644 而\n动态区仍写 cb2a1f0），按其要求重跑取证后才通过——新门禁有牙。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:52:26-07:00"}],"HeadCommit":{"Sha1":"b1695a6780efbc95449a553d313599695c0b22ab","Message":"docs(governance): 回灌首份远端 CI 绿盘证据 + 三级证据刷新至 5f15617\n\n远端证据（G14 部分推进，**未闭环**）\nGitHub 远端已加回（github → laoluojuhai/base-framework，私有仓）。\n在 5f15617 上三个 job 全绿：Static governance / Runtime and UI acceptance /\nAggregate same-SHA governance evidence\nrun: https://github.com/laoluojuhai/base-framework/actions/runs/33455966984\n该 run 的 headSha 与本地 HEAD 逐字符对齐。这是本仓历史上第一次远端 CI 绿——\n此前每一次远端运行都是 failure（最近一次 2026-08-26，正是 check:os-product 并入\npnpm check 那一提交）。\n\n三条「只有远端才看得见」的缺陷已在本轮依次修掉，均已单独提交并在本条登记：\n① checkout 浅克隆 → F5 把本仓证据误判为「来自另一个仓库」\n② 差分门禁依赖磁盘残留的 contracts dist（外加 composite 增量陷阱）\n③ 测试数地板正则未剥 ANSI，CI 下恒读 0\n共同点：本地全绿、远端必红——正是 G14「远端拦截未验证」这条缺口存在的意义。\n\n⚠️ ci-gate 仍不标 GREEN：required checks / 分支保护未配置，受控失败 PR 未做。\n「门禁能跑绿」推不出「违规会被挡住」，缺这两项不得宣称 PR 已被机器拦截。\n\n三级证据刷新\n静态 exit 0、runtime 202 tests / 差分 0 differences、UI 8 用例，\n三份报告 provenance 均为 gitSha=5f15617、worktreeDirty=false。\n本轮回灌过程中 freshness-sha-runtime 断言真的挡了我一次（报告已刷到 74b4644 而\n动态区仍写 cb2a1f0），按其要求重跑取证后才通过——新门禁有牙。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:52:26-07:00"},"CompareURL":"luoanwu/base-framework/compare/5f156175e6f2397c334fcd3eb5cde4e45423b68b...b1695a6780efbc95449a553d313599695c0b22ab","Len":1}...
|
1788223950
|
Edit
Delete
|
|
26940
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"5f156175e {"Commits":[{"Sha1":"5f156175e6f2397c334fcd3eb5cde4e45423b68b","Message":"fix(gate): 测试数解析先剥 ANSI——地板校验在 CI 上恒读 0（假红）\n\n远端第三次真跑：static job 绿，runtime job 报\n\n Tests passed regressed: expected \u003e= 202, got 0.\n\n但下载 CI 上传的证据 artifact 看，test 步骤 status=0、4 个 turbo 任务全成功、\nFastify 明明 58 passed——测试全跑过了，是 runner 的**解析**失效。\n\n根因：GitHub Actions 支持颜色，vitest/turbo 因此输出带色汇总行，\"Tests\" 与数字之间\n夹着 ANSI 转义序列，`/Tests\\s+(\\d+)\\s+passed/` 匹配不到。本地非 TTY 无色所以一直有效——\n这条地板在本地有牙、在 CI 上恒读 0，属于「同一门禁在两个环境语义不同」。\n方向上它是 fail-closed（0 \u003c 202 判红）而非假绿，但仍是假红，会把真实回归淹没在噪声里。\n\ncountPassedTests 改为先 stripAnsi 再匹配。\n验证：用真实 CI 那行带色文本喂解析器，剥离前 0、剥离后 58；\n本地 check:runtime 仍读到 202（exit 0）。\n\nUI runner 不受影响：它读 Playwright 的 JSON 统计，不解析终端文本。\n\n这是加回 GitHub 远端后暴露的第三条「只有远端才看得见」的缺陷\n（前两条：checkout 浅克隆让 F5 误判、差分门禁依赖磁盘残留的 contracts dist）。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:43:13-07:00"}],"HeadCommit":{"Sha1":"5f156175e6f2397c334fcd3eb5cde4e45423b68b","Message":"fix(gate): 测试数解析先剥 ANSI——地板校验在 CI 上恒读 0（假红）\n\n远端第三次真跑：static job 绿，runtime job 报\n\n Tests passed regressed: expected \u003e= 202, got 0.\n\n但下载 CI 上传的证据 artifact 看，test 步骤 status=0、4 个 turbo 任务全成功、\nFastify 明明 58 passed——测试全跑过了，是 runner 的**解析**失效。\n\n根因：GitHub Actions 支持颜色，vitest/turbo 因此输出带色汇总行，\"Tests\" 与数字之间\n夹着 ANSI 转义序列，`/Tests\\s+(\\d+)\\s+passed/` 匹配不到。本地非 TTY 无色所以一直有效——\n这条地板在本地有牙、在 CI 上恒读 0，属于「同一门禁在两个环境语义不同」。\n方向上它是 fail-closed（0 \u003c 202 判红）而非假绿，但仍是假红，会把真实回归淹没在噪声里。\n\ncountPassedTests 改为先 stripAnsi 再匹配。\n验证：用真实 CI 那行带色文本喂解析器，剥离前 0、剥离后 58；\n本地 check:runtime 仍读到 202（exit 0）。\n\nUI runner 不受影响：它读 Playwright 的 JSON 统计，不解析终端文本。\n\n这是加回 GitHub 远端后暴露的第三条「只有远端才看得见」的缺陷\n（前两条：checkout 浅克隆让 F5 误判、差分门禁依赖磁盘残留的 contracts dist）。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:43:13-07:00"},"CompareURL":"luoanwu/base-framework/compare/74b464480301493d008b476e72247f7289ccbffb...5f156175e6f2397c334fcd3eb5cde4e45423b68b","Len":1}...
|
1788223468
|
Edit
Delete
|
|
26939
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"74b464480 {"Commits":[{"Sha1":"74b464480301493d008b476e72247f7289ccbffb","Message":"fix(gate): 差分门禁自包含——不再依赖磁盘上残留的 contracts dist\n\n远端 CI 第二次真跑，static job 首次转绿，runtime job 接着暴露下一条：\n\n Error: Failed to resolve entry for package \"@repo/contracts\".\n\ncheck:conformance:differential 直接调 vitest（不经 turbo），因此拿不到 turbo 声明的\n`test → build` 依赖；而 contracts 是 TS 项目引用，没有 dist 时 vite 解析包入口即失败。\n本地长期通过是因为先跑过 pnpm check 把 dist 留在磁盘上——门禁被环境残留掩盖，\n正是本仓「静态门禁必须自包含」纪律要防的形态。\n\n修复时踩到第二层：contracts 的 tsconfig 是 composite（增量），tsbuildinfo 还在而 dist\n已被删时，`tsc -p` 会认为「一切最新」而**不产出任何文件**——构建退出码 0、dist 依旧为空。\n故重建前先清 tsbuildinfo，并在构建后断言 dist/index.js 真的存在，否则显式报错。\n\n负向验证：`rm -rf packages/contracts/dist packages/contracts/tsconfig.tsbuildinfo`\n后直接跑 check:conformance:differential → 自动重建并通过\n（7 checkpoints × 2 backends，0 differences）。\n\n本地：pnpm check exit 0；pnpm check:runtime exit 0。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:36:59-07:00"}],"HeadCommit":{"Sha1":"74b464480301493d008b476e72247f7289ccbffb","Message":"fix(gate): 差分门禁自包含——不再依赖磁盘上残留的 contracts dist\n\n远端 CI 第二次真跑，static job 首次转绿，runtime job 接着暴露下一条：\n\n Error: Failed to resolve entry for package \"@repo/contracts\".\n\ncheck:conformance:differential 直接调 vitest（不经 turbo），因此拿不到 turbo 声明的\n`test → build` 依赖；而 contracts 是 TS 项目引用，没有 dist 时 vite 解析包入口即失败。\n本地长期通过是因为先跑过 pnpm check 把 dist 留在磁盘上——门禁被环境残留掩盖，\n正是本仓「静态门禁必须自包含」纪律要防的形态。\n\n修复时踩到第二层：contracts 的 tsconfig 是 composite（增量），tsbuildinfo 还在而 dist\n已被删时，`tsc -p` 会认为「一切最新」而**不产出任何文件**——构建退出码 0、dist 依旧为空。\n故重建前先清 tsbuildinfo，并在构建后断言 dist/index.js 真的存在，否则显式报错。\n\n负向验证：`rm -rf packages/contracts/dist packages/contracts/tsconfig.tsbuildinfo`\n后直接跑 check:conformance:differential → 自动重建并通过\n（7 checkpoints × 2 backends，0 differences）。\n\n本地：pnpm check exit 0；pnpm check:runtime exit 0。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:36:59-07:00"},"CompareURL":"luoanwu/base-framework/compare/5af8e60ea25ac2d24a90b8f103092f30d96b909a...74b464480301493d008b476e72247f7289ccbffb","Len":1}...
|
1788223021
|
Edit
Delete
|
|
26938
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"5af8e60ea {"Commits":[{"Sha1":"5af8e60ea25ac2d24a90b8f103092f30d96b909a","Message":"fix(ci): checkout 取全历史——浅克隆让 F5 把本仓证据误判为「来自另一个仓库」\n\n加回 GitHub 远端后跑出第一份真实远端证据，static job 立刻红，但**不是治理违规**：\n\n ✗ [F5] reports/runtime-acceptance.latest.json: provenance.gitSha=cb2a1f0\n 不存在于本仓 git 历史——该证据来自**另一个仓库**\n\nF5 用 `git cat-file -e \u003csha\u003e^{commit}` 校验验收证据是否绑定本仓历史（防移植时照搬\n源项目的「测试跑过了」假证据）。而 actions/checkout@v4 默认 fetch-depth:1 只克隆\ntip 提交——验收证据一旦绑定任何祖先提交（正常工作流必然如此：先跑验收再提交报告），\nF5 就必然误判。三处 checkout 均补 fetch-depth: 0。\n\n本地负向复现：`git clone --depth 1` 后提交数=1，`git cat-file -e cb2a1f0` 找不到，\n与 CI 报错一致；全历史下同一命令通过。\n\n值得记的是：这条缺陷**只有真跑远端 CI 才暴露得出来**——本地永远是全历史，\n门禁正向绿；G14 登记的「远端拦截未验证」正是为防这类盲区而存在。\n上一轮修掉的是 check:os-product 依赖仓外检出（结构性必红），这是它后面藏着的第二个。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:29:21-07:00"}],"HeadCommit":{"Sha1":"5af8e60ea25ac2d24a90b8f103092f30d96b909a","Message":"fix(ci): checkout 取全历史——浅克隆让 F5 把本仓证据误判为「来自另一个仓库」\n\n加回 GitHub 远端后跑出第一份真实远端证据，static job 立刻红，但**不是治理违规**：\n\n ✗ [F5] reports/runtime-acceptance.latest.json: provenance.gitSha=cb2a1f0\n 不存在于本仓 git 历史——该证据来自**另一个仓库**\n\nF5 用 `git cat-file -e \u003csha\u003e^{commit}` 校验验收证据是否绑定本仓历史（防移植时照搬\n源项目的「测试跑过了」假证据）。而 actions/checkout@v4 默认 fetch-depth:1 只克隆\ntip 提交——验收证据一旦绑定任何祖先提交（正常工作流必然如此：先跑验收再提交报告），\nF5 就必然误判。三处 checkout 均补 fetch-depth: 0。\n\n本地负向复现：`git clone --depth 1` 后提交数=1，`git cat-file -e cb2a1f0` 找不到，\n与 CI 报错一致；全历史下同一命令通过。\n\n值得记的是：这条缺陷**只有真跑远端 CI 才暴露得出来**——本地永远是全历史，\n门禁正向绿；G14 登记的「远端拦截未验证」正是为防这类盲区而存在。\n上一轮修掉的是 check:os-product 依赖仓外检出（结构性必红），这是它后面藏着的第二个。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:29:21-07:00"},"CompareURL":"luoanwu/base-framework/compare/75189f8787bc51f9cf16ba7b2fdcb735aea2ec15...5af8e60ea25ac2d24a90b8f103092f30d96b909a","Len":1}...
|
1788222564
|
Edit
Delete
|
|
26937
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"75189f878 {"Commits":[{"Sha1":"75189f8787bc51f9cf16ba7b2fdcb735aea2ec15","Message":"chore(reports): 刷新三级证据至 cb2a1f0（clean）+ 动态区同步回灌\n\n结构化日志接线后在 clean commit 上重跑取证：\n- 静态 pnpm check：exit 0\n- 真实 DB check:runtime：202 tests / 0 failures、差分 0 differences\n （PG base_framework_rt_cb2a1f0 + Redis db15）\n- 浏览器 check:ui：8 用例 / 0 失败 / 0 跳过\n （PG base_framework_ui_cb2a1f0 + Redis db12，端口经 UI_WEB_PORT/UI_API_PORT 覆盖）\n\n三份报告 provenance 均为 gitSha=cb2a1f0、worktreeDirty=false，动态区同步更新。\n本轮回灌由上一提交新增的 freshness-sha-runtime / freshness-sha-ui 断言强制——\n报告动了而文档没跟上会直接红，不再依赖人肉记得。\n\n作用域不变：仅证明「本机在 cb2a1f0 这个 clean commit 上通过三级门禁」，\n推不出远端已发布或 CI 已拦截（G14 仍 OPEN）。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:21:01-07:00"},{"Sha1":"cb2a1f09fde2e86e3a4263b59e3565efaed5b104","Message":"feat(observability): NestJS 接结构化日志，与 Fastify 拉平（G20 第一子项）\n\n此前 Fastify 侧是 pino 结构化 JSON，NestJS 全程默认 ConsoleLogger 输出彩色纯文本——\n同一套写链在两后端产出两种不可对齐的日志格式，生产上无法用同一条查询捞两边。\ncheck:dual-backend 是存在性门禁，看不见这层格式差异（G17 行为级盲区的又一实例）。\nG20「无可观测性」里性价比最高的一半其实早就免费存在，只差把另一半接上。\n\n- 新增 common/logger.ts：pino 支撑的 Nest LoggerService，口径与 api-fastify/src/app.ts\n 逐条对齐（同一个 LOG_LEVEL、同样的凭证脱敏路径、production 单行 JSON）；\n Nest 传入的 context（类名）映射为结构化字段而非拼进消息串——拼进去就又要靠正则解析。\n- 在 create-app.ts 这个装配单源接线，测试与 bootstrap 共用同一份（禁止 main.ts 另起一套）；\n- main.ts 的 4 行 console.log 改走 Logger，否则启动横幅纯文本、其余日志 JSON，同进程两种格式；\n bootstrap catch 保留 console.error 兜底（此时应用可能尚未装配成功）。\n\n顺带修掉一个两后端都有的启动崩溃模式：pino-pretty 是 devDependency，用\n`pnpm install --prod` 部署且 NODE_ENV 未设 production 时它不存在，pino 解析不到\ntransport target 会直接抛错——为一个开发期美化依赖把服务搞得起不来。\n两侧均改为先探测可用性、不可用则降级为 JSON。\n\n验证：\n- production：单行 JSON，context 为结构化字段，Error 序列化为 err{type,message,stack}\n- 开发态：pino-pretty 正常着色\n- 负向：临时移走 node_modules 里的 pino-pretty → 降级输出 JSON 而非崩溃，恢复后复原\n- pnpm check exit 0；pnpm check:runtime exit 0（202 tests、差分 0 differences）\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:18:41-07:00"}],"HeadCommit":{"Sha1":"75189f8787bc51f9cf16ba7b2fdcb735aea2ec15","Message":"chore(reports): 刷新三级证据至 cb2a1f0（clean）+ 动态区同步回灌\n\n结构化日志接线后在 clean commit 上重跑取证：\n- 静态 pnpm check：exit 0\n- 真实 DB check:runtime：202 tests / 0 failures、差分 0 differences\n （PG base_framework_rt_cb2a1f0 + Redis db15）\n- 浏览器 check:ui：8 用例 / 0 失败 / 0 跳过\n （PG base_framework_ui_cb2a1f0 + Redis db12，端口经 UI_WEB_PORT/UI_API_PORT 覆盖）\n\n三份报告 provenance 均为 gitSha=cb2a1f0、worktreeDirty=false，动态区同步更新。\n本轮回灌由上一提交新增的 freshness-sha-runtime / freshness-sha-ui 断言强制——\n报告动了而文档没跟上会直接红，不再依赖人肉记得。\n\n作用域不变：仅证明「本机在 cb2a1f0 这个 clean commit 上通过三级门禁」，\n推不出远端已发布或 CI 已拦截（G14 仍 OPEN）。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:21:01-07:00"},"CompareURL":"luoanwu/base-framework/compare/742f865d0672599ba182b16012b7671b114f273f...75189f8787bc51f9cf16ba7b2fdcb735aea2ec15","Len":2}...
|
1788222064
|
Edit
Delete
|
|
26936
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"742f865d0 {"Commits":[{"Sha1":"742f865d0672599ba182b16012b7671b114f273f","Message":"chore(os-product): OS 两级验收在 clean commit 上重跑并回灌动态区\n\n上一轮 OS 两级因上游检出未显式提供而 SKIPPED，动态区据实标为 STALE。\n本轮显式给 DIGITAL_EMPLOYEE_OS_ROOT 后重跑，两级均绿：\n\n- check:os-product（静态兼容）：juhai.baseframework@0.1.0 / protocol 1.3.0\n / 2 migrations / 1 acceptance suite，0 违规\n- check:os-product:runtime（双宿主真实验收）：NestJS + Fastify / 8 cases / 0 failures\n （DB base_framework_os_product_a9bf413 + Redis db13，Node 22.23.2）\n\n证据绑定：本仓 clean @ a9bf413；上游 OS 本机检出源码 clean @ 89b2ade。\n\n关于上游 clean 的判定依据（不接受「报告说 clean 就是 clean」）：\n上游工作树 git status 有 24 个变更，但已逐条核实全部匹配 reports/*.latest.json；\nreport-provenance 的 worktreeDirty 按文档化语义刻意排除该模式（报告是运行产物，\n前一个门禁刷新它们不应被后续门禁误判为「被验收源码 dirty」）。\n用同一排除规则跑 `git status --porcelain=v1 -- ':(exclude)reports/*.latest.json'`\n结果为空——源码确实 clean，dirty=false 不是假绿。\n\npnpm check exit 0；check:docs-truth 的动态区 SHA 断言仍绿。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:04:22-07:00"}],"HeadCommit":{"Sha1":"742f865d0672599ba182b16012b7671b114f273f","Message":"chore(os-product): OS 两级验收在 clean commit 上重跑并回灌动态区\n\n上一轮 OS 两级因上游检出未显式提供而 SKIPPED，动态区据实标为 STALE。\n本轮显式给 DIGITAL_EMPLOYEE_OS_ROOT 后重跑，两级均绿：\n\n- check:os-product（静态兼容）：juhai.baseframework@0.1.0 / protocol 1.3.0\n / 2 migrations / 1 acceptance suite，0 违规\n- check:os-product:runtime（双宿主真实验收）：NestJS + Fastify / 8 cases / 0 failures\n （DB base_framework_os_product_a9bf413 + Redis db13，Node 22.23.2）\n\n证据绑定：本仓 clean @ a9bf413；上游 OS 本机检出源码 clean @ 89b2ade。\n\n关于上游 clean 的判定依据（不接受「报告说 clean 就是 clean」）：\n上游工作树 git status 有 24 个变更，但已逐条核实全部匹配 reports/*.latest.json；\nreport-provenance 的 worktreeDirty 按文档化语义刻意排除该模式（报告是运行产物，\n前一个门禁刷新它们不应被后续门禁误判为「被验收源码 dirty」）。\n用同一排除规则跑 `git status --porcelain=v1 -- ':(exclude)reports/*.latest.json'`\n结果为空——源码确实 clean，dirty=false 不是假绿。\n\npnpm check exit 0；check:docs-truth 的动态区 SHA 断言仍绿。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:04:22-07:00"},"CompareURL":"luoanwu/base-framework/compare/a9bf41341677134557538182b87a1339aa7b7616...742f865d0672599ba182b16012b7671b114f273f","Len":1}...
|
1788221065
|
Edit
Delete
|
|
26935
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"a9bf41341 {"Commits":[{"Sha1":"a9bf41341677134557538182b87a1339aa7b7616","Message":"docs(governance): 动态区回灌至 clean @ 3bb349d + 把「动态区与报告同步」机器化\n\n回灌\n- 三级证据新鲜度从 `dirty @ a9e18b9` 更正为 **clean @ `3bb349d`**（静态 exit 0、\n runtime 202 tests / 差分 0 differences、UI 8 用例），并记录本轮实际基座库名与端口覆盖；\n- G0 余项（clean commit 三级重跑）标记完成；\n- 基线表 6 行「本地 dirty」改为 clean 绑定；\n- OS 两级本轮 SKIPPED / 未实跑，诚实标为 🟡 STALE——不随其他行一起蹭 clean。\n\n机器化（P5：机器门禁 \u003e 手册条目 \u003e 口头经验）\ncheck:docs-truth 新增 freshness-sha-runtime / freshness-sha-ui 两条断言：\n动态区声明的 SHA 必须等于对应 latest 报告的 provenance.gitSha。\n\n实锤依据：连续两轮报告刷新提交都没回灌动态区（报告 clean @ 3bb349d，\n动态区仍写 dirty @ a9e18b9）；而 CLAUDE.md 自己规定「判定治理状态只认动态区 + reports」，\n两份真源打架时该规矩直接失效。同一轮人工回灌时我又把「本轮根本没跑」的 OS 行\n误标成 clean——人肉同步不可靠，正是 P5 要求机器化的形态。\n\n刻意不含静态级：governance.latest.json 每次 pnpm check 都重新生成并绑定当时 HEAD，\n对它断言等价于「每次提交都必须改文档」、提交报告后必红，那是会误杀的门禁。\nruntime/UI 报告只有显式跑真实验收才会变，「报告动了而文档没跟上」才是真信号。\n\n负向测试（收尾清单要求）：\n① 把 runtime 行 SHA 篡改为 a9e18b9 → freshness-sha-runtime 精确红；\n② 删掉 UI 行的 SHA 声明 → freshness-sha-ui 精确红；\n两次均恢复现场后复跑转绿。pnpm check exit 0。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:00:44-07:00"}],"HeadCommit":{"Sha1":"a9bf41341677134557538182b87a1339aa7b7616","Message":"docs(governance): 动态区回灌至 clean @ 3bb349d + 把「动态区与报告同步」机器化\n\n回灌\n- 三级证据新鲜度从 `dirty @ a9e18b9` 更正为 **clean @ `3bb349d`**（静态 exit 0、\n runtime 202 tests / 差分 0 differences、UI 8 用例），并记录本轮实际基座库名与端口覆盖；\n- G0 余项（clean commit 三级重跑）标记完成；\n- 基线表 6 行「本地 dirty」改为 clean 绑定；\n- OS 两级本轮 SKIPPED / 未实跑，诚实标为 🟡 STALE——不随其他行一起蹭 clean。\n\n机器化（P5：机器门禁 \u003e 手册条目 \u003e 口头经验）\ncheck:docs-truth 新增 freshness-sha-runtime / freshness-sha-ui 两条断言：\n动态区声明的 SHA 必须等于对应 latest 报告的 provenance.gitSha。\n\n实锤依据：连续两轮报告刷新提交都没回灌动态区（报告 clean @ 3bb349d，\n动态区仍写 dirty @ a9e18b9）；而 CLAUDE.md 自己规定「判定治理状态只认动态区 + reports」，\n两份真源打架时该规矩直接失效。同一轮人工回灌时我又把「本轮根本没跑」的 OS 行\n误标成 clean——人肉同步不可靠，正是 P5 要求机器化的形态。\n\n刻意不含静态级：governance.latest.json 每次 pnpm check 都重新生成并绑定当时 HEAD，\n对它断言等价于「每次提交都必须改文档」、提交报告后必红，那是会误杀的门禁。\nruntime/UI 报告只有显式跑真实验收才会变，「报告动了而文档没跟上」才是真信号。\n\n负向测试（收尾清单要求）：\n① 把 runtime 行 SHA 篡改为 a9e18b9 → freshness-sha-runtime 精确红；\n② 删掉 UI 行的 SHA 声明 → freshness-sha-ui 精确红；\n两次均恢复现场后复跑转绿。pnpm check exit 0。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T17:00:44-07:00"},"CompareURL":"luoanwu/base-framework/compare/6449421518ff6b2dd62b6ff862ff44aedcbbea29...a9bf41341677134557538182b87a1339aa7b7616","Len":1}...
|
1788220856
|
Edit
Delete
|
|
26934
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"644942151 {"Commits":[{"Sha1":"6449421518ff6b2dd62b6ff862ff44aedcbbea29","Message":"chore(reports): 刷新三级治理证据至 3bb349d（clean，202 tests / 8 UI）\n\n三级门禁均在 clean 工作区、同一提交 3bb349d 上实跑：\n\n- 静态 pnpm check：13 项门禁 + lint + typecheck，exit 0\n- 真实 DB pnpm check:runtime：202 tests / 0 failures、两后端迁移 deploy+status、\n 双后端同夹具差分 0 differences（PG base_framework_rt_final + Redis db11）\n- 浏览器 pnpm check:ui：8 用例 / 0 失败 / 0 跳过（PG base_framework_ui_final + db12）\n\n三份报告的 provenance 均为 gitSha=3bb349d、worktreeDirty=false，\n不再是此前绑定 a9e18b9 dirty 的 STALE 证据。\n\n新增报告：conformance-matrix、conformance-differential、governance-rules、\ngovernance-profile-kernel、governance-status（四层 profile 当前均为 PARTIAL——\nADR-0010 仍是 Proposed，observe 模式只诊断，不构成发布放行）。\n\n作用域声明：以上仅证明「本机 hillao 在 3bb349d 这个 clean commit 上通过三级门禁」，\n推不出远端已发布或 CI 已拦截（G14 仍 OPEN）。\ncheck:os-product 本轮显式 SKIPPED（上游 OS 检出不在场），未写 osProduct* 指标。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T16:36:45-07:00"},{"Sha1":"3bb349d8ab4351f7f17927b50c94255a7d8abbef","Message":"fix(test): outbox 幂等键用例按租户收窄——移除它自己保证不了的全局隔离假设\n\n`dispatchOutboxBatch` 按设计跨租户全局扫描（outbox dispatcher 的\nwrite-guard-allow 豁免正为此），而 check:runtime 让两后端套件经 turbo 并行\n打同一个库。该用例却断言「全库只有我这一行待投递」——兄弟套件此刻产生的\n未投递行会被一并领走，于是期望 1 实得 2/3，数量随并行时序漂移。\n\n实测：连续两次全量红且失败对象不同（realtime.tenant 一次、outbox 一次），\n换全新库仍红且计数从 2 涨到 3；单独复跑该文件恒绿——典型的共享基座污染\n（G18 已登记族、审计发现「两后端共享 DB/Redis」的又一实证）。\n\n本用例要证的是「信封 id 恒等于 outbox 行 id」这条幂等键稳定性，与其他租户\n有多少行无关，故 publish 回调按 tenantId 过滤后再断言。这是移除伪假设，\n不是放宽断言：跨实例不重复投递由同文件上一条用例独立守护。\n\ncheck:runtime：202 tests、0 failures、差分 0 differences（exit 0）。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T16:34:17-07:00"},{"Sha1":"886537eae83f56d3b662e070575ddf3d9a72307f","Message":"feat(governance): ADR-0010 阶段一/二——治理分层与同 SHA 证据模型\n\n把「一堆各自为政的门禁 + 最近一次运行的报告」升级为分层规则与可聚合证据。\nADR-0010 当前为 Proposed，本批只落 observe 模式的实现，不改变现行放行口径。\n\n规则生命周期\n- governance.rules.json + schema：16 条规则，每条带 owner、作用域、严重度、\n 证据映射、profile 归属、ADR 引用与例外期限；check:governance-rules 校验\n 规则无主/入口缺失/例外过期/profile 循环依赖，registryVersion=3。\n\n四层 profile 入口\n- check:kernel / check:production / check:conformance / check:extension，\n 经 check-governance-profile.mjs 统一编排（非 Kernel 的 profile 先跑 Kernel）。\n- governance:status 生成同 SHA 状态清单：缺失、陈旧、dirty、跨 SHA 或未绑定\n registryVersion 的证据一律不产出 VERIFIED，当前四层均为 PARTIAL——\n 这正是它没在拿旧证据冒充绿盘的证明。\n\n双后端行为身份\n- conformance.matrix.json 锁 8 cases × 2 backends 的 case 身份；\n- check:conformance:differential 用**同一夹具**逐字段比较两端 HTTP/DB/outbox\n 的 7 个 checkpoint。存在性门禁看不见语义漂移，差分才看得见。\n\n证据来源加固\n- report-provenance 增加 registryVersion / run URL / artifact 身份；\n- 验收 runner（check:runtime / check:ui）拒绝 GOV_REPORT_* 覆盖变量并 exit 2\n ——此前可用它们一行伪造出「clean@HEAD 通过」的验收报告；\n- governance-report 的缺失指标改 fail-closed：预期指标 key 消失时不再落到\n `?? -1`/`?? 0` 被当成「变好」静默过闸（C18 同族，发生在 current 侧）。\n\n其他\n- runtime-governance 审计与验收基座守卫抽到 scripts/lib，供多入口复用；\n- check 链移出 check:os-product（它依赖仓外 OS 检出，会让静态门禁不自包含、\n GitHub CI 结构性必红）；缺失时显式 SKIPPED，要真实校验用 check:full；\n- governance.yml 改三段 artifact 汇总；\n- baseline 仅收紧：ownTests 24→28、ownTestCases 146→197、testsPassed 192→202、\n uiTestsPassed 4→8，另加 5 项新地板，无任何放宽。\n\npnpm check：13 项静态门禁 + lint + typecheck 全绿（exit 0）。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T16:28:40-07:00"},{"Sha1":"ec489eae037f3a7bc6b027898804a8b1bde0fc23","Message":"fix(app): 404 不泄露资源 id + 前端断网桥接\n\n两处由治理证据实锤出来的产品缺陷：\n\n1. NestJS 订单 404 回显了资源 id（`Order \u003cid\u003e not found`），跨租户探测据此\n 可确认某 id 是否存在——404 的意义正是不泄露存在性。同时 Nest 默认异常体\n 带 statusCode/error 元数据，与 Fastify 的 `{message}` 形成两套对外契约。\n 现统一为稳定 ApiErrorBody `{ message: \"Order not found\" }`。\n\n2. useRealtime 缺 offline/online 桥接：WS/SSE 的 close/error 只在 TCP 真正\n 断开时触发，而「网络没了但连接还挂着」（拔网线、飞行模式、Playwright\n setOffline）不会立刻断 TCP，服务端 30s 心跳也只是数据帧——状态徽章会停在\n Connected 假绿。现监听 window offline/online：offline 主动断开并入既有退避\n 重连路径；online 清退避定时器立即重连，但不直接置 open——navigator.onLine\n 只有 false 是确定的，恢复后仍由真实连接结果决定状态。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T16:28:00-07:00"}],"HeadCommit":{"Sha1":"6449421518ff6b2dd62b6ff862ff44aedcbbea29","Message":"chore(reports): 刷新三级治理证据至 3bb349d（clean，202 tests / 8 UI）\n\n三级门禁均在 clean 工作区、同一提交 3bb349d 上实跑：\n\n- 静态 pnpm check：13 项门禁 + lint + typecheck，exit 0\n- 真实 DB pnpm check:runtime：202 tests / 0 failures、两后端迁移 deploy+status、\n 双后端同夹具差分 0 differences（PG base_framework_rt_final + Redis db11）\n- 浏览器 pnpm check:ui：8 用例 / 0 失败 / 0 跳过（PG base_framework_ui_final + db12）\n\n三份报告的 provenance 均为 gitSha=3bb349d、worktreeDirty=false，\n不再是此前绑定 a9e18b9 dirty 的 STALE 证据。\n\n新增报告：conformance-matrix、conformance-differential、governance-rules、\ngovernance-profile-kernel、governance-status（四层 profile 当前均为 PARTIAL——\nADR-0010 仍是 Proposed，observe 模式只诊断，不构成发布放行）。\n\n作用域声明：以上仅证明「本机 hillao 在 3bb349d 这个 clean commit 上通过三级门禁」，\n推不出远端已发布或 CI 已拦截（G14 仍 OPEN）。\ncheck:os-product 本轮显式 SKIPPED（上游 OS 检出不在场），未写 osProduct* 指标。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-31T16:36:45-07:00"},"CompareURL":"luoanwu/base-framework/compare/a9e18b9599335b3ad128ac8c761fc93b83de8711...6449421518ff6b2dd62b6ff862ff44aedcbbea29","Len":4}...
|
1788219424
|
Edit
Delete
|
|
26786
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"a9e18b959 {"Commits":[{"Sha1":"a9e18b9599335b3ad128ac8c761fc93b83de8711","Message":"fix(governance): 收口实时顺序与验收缺口\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T22:20:28-07:00"}],"HeadCommit":{"Sha1":"a9e18b9599335b3ad128ac8c761fc93b83de8711","Message":"fix(governance): 收口实时顺序与验收缺口\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T22:20:28-07:00"},"CompareURL":"luoanwu/base-framework/compare/4a707c3d3ec98b8ab45864a1b40b941589433914...a9e18b9599335b3ad128ac8c761fc93b83de8711","Len":1}...
|
1788067980
|
Edit
Delete
|
|
26785
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"4a707c3d3 {"Commits":[{"Sha1":"4a707c3d3ec98b8ab45864a1b40b941589433914","Message":"chore(reports): 刷新治理证据至 b1510b5（clean，192 tests）\n\n静态 11 项门禁 + 治理棘轮 + lint + typecheck 全绿；\n真实 PG + Redis 验收 192 tests / 0 failures，provenance 绑定 clean commit b1510b5。\n\n⚠️ 作用域：本轮证据只覆盖静态与运行态两级。浏览器级 check:ui **未重跑**——\n本机无 Node 运行时（全部门禁在 node:22 容器内挂载工作树执行），而容器内没有\nLinux 版 Playwright 浏览器（宿主缓存的是 macOS 构建）。按事实作用域表，\n不得据此宣称 UI 层已验证。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:40:42-07:00"},{"Sha1":"b1510b5a203f7847dc99a0962cd30f36e6901dce","Message":"docs(standards): 修正与实现相悖的 API 标准 + 幽灵路径机器化（批次④·下）\n\n## api-standard 的四项核心约定与实现系统性相悖\n\nCLAUDE.md 真源地图把 docs/standards/*.md 列为「工程标准口径」，使用入口表还直接路由读者\n去读它。但 api-standard 的四项核心约定没有一项与两后端实际实现一致：\n`/api/v1/` 前缀（实际 `/api`，未版本化）、`{ data: ... }` 响应包裹（实际裸资源）、\n嵌套 `error.{code,statusCode,timestamp,path}` 错误体（实际是顶层扁平的 ApiErrorBody）、\n以及一个从不存在的 `packages/contracts/src/errors/error-codes.ts`。\n\n照着它落地新模块，会做出与 contracts `ApiErrorBody` 及 web `ApiError` 消费方直接冲突的接口。\n而 review-standard 的检查清单还会让 reviewer 去核对那个幽灵文件。\n\n- 按本仓既有的 C14「标准 vs 现状」惯例（frontend-standard / backend-standard / git-standard\n 都已有）在 api-standard 头部加对照表，逐项写明目标态与当前真相，并说明改动方向：\n 要落地目标态先改实现与契约，不要反过来照着文档改接口\n- 修正 review-standard 的检查项指向真实位置\n\n## 幽灵路径机器化（C14-⑤ 同族，扫描面扩到 standards）\n\n新增 docs-truth 断言：standards 引用的 `packages/contracts/src/**.ts` 路径必须真实存在，\n否则须在同一行显式标注「目标态 / 当前不存在」。扫描面用 readdirSync 覆盖 docs/standards 全量，\n不硬编码文件清单——旧的幽灵路径断言正是因为清单式扫描漏掉了 docs/README.md 本身而失守过。\n\n该断言上线即多抓到一处此前没人发现的幽灵路径（naming-standard 的\n`packages/contracts/src/store/store.types.ts`），已一并标注。\n负向测试：去掉 review-standard 的诚实标注即红。\n\n## owner-matrix 补录 OS 产品两张业务表\n\n「先登记后建表」是本仓自己立的规则，却被自己的 C20/G23 战役打破：OS 产品包新建了\n`juhai_baseframework_orders` / `..._order_events` 两张带状态机的表却从未登记。\n补录两行并写明 `products/` 下的对象同样要登记（归属真源是 product.manifest.json 的\npersistence.tenantModels，本表挂索引行）。\n\n验收：naming / docs-truth / governance-docs / 治理棘轮均 exit 0。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:39:02-07:00"},{"Sha1":"033e8159c23f4766493fd140ffa7aa923ebe1182","Message":"fix(ops): 优雅停机接线——Fastify 的 onClose 钩子此前全是死代码（批次④·中）\n\n## Fastify 从不优雅停机\n\nserver.ts 没有任何信号处理，也从不调用 app.close()：进程收到 SIGTERM 走默认处理直接终止。\n后果是 prisma / redis / outbox / queue 四个插件里**精心写好的 onClose 钩子从未被执行过**——\nBullMQ worker 不 close（在途 job 只能靠 stalled 恢复）、outbox 轮询事务可能被拦腰砍断、\nWS/SSE 连接收不到 close 帧。而 NestJS 侧早有 enableShutdownHooks，两后端停机行为实质分叉。\n\n存在性门禁看得见「钩子写了」，看不见「它从来没被调用过」——这正是 G17 说的行为级盲区。\n\n- server.ts 补 SIGTERM/SIGINT → app.close()（Fastify 会按注册逆序执行全部 onClose），\n 带 10s 超时兜底（卡住的清理不能让容器一直不退）与重复信号防抖\n\n## NestJS 的 Redis 连接从不 quit，且注释谎称有人管\n\nREDIS_PUB / REDIS_SUB 由 useFactory 裸建，没有任何销毁钩子；而 event-bus.service.ts\n的注释写着「redisSub 连接由 RedisModule 统一关闭」——注释与行为相反，于是没人会去查。\n真实后果：停机时连接不回收（两个 acceptance 测试的 afterAll 至今还得手动 quit 才能让进程退出）。\n\n- RedisModule 实现 onApplicationShutdown，quit 两个连接\n- 修正 event-bus 的失实注释\n\n## 防回潮\n\ncheck:dual-backend 新增 4 条停机接线绊网（两后端各自的信号处理 / app.close /\nenableShutdownHooks / Redis 关闭钩子），负向测试已验证删掉信号处理即红。\n\n验收：真实 PG + Redis 192 tests / 0 failures；静态门禁 + 棘轮 + typecheck 5/5 + lint 2/2 全绿。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:35:43-07:00"},{"Sha1":"44cbbaeacf5f53e08e6e3f48a6d8bd2c3327d2de","Message":"fix(governance): 写链守护补齐五条已验证绕过（批次④·上）\n\ncheck:write-guard 守的是 P4「所有写显式带 tenant_id + 状态迁移乐观锁」这条最高优先级红线，\n但它自身存在五条可复现的绕过——每一条都能让违规代码拿到绿盘：\n\n1. **折行**：逐行匹配下 `await tx.order\\n .update({...})` 完全隐形。这甚至无需恶意——\n Prettier 对长链式调用本来就这样折行。\n2. **方括号访问**：`tx[\"order\"].update(` 用 `\\w+` 匹配不到，换个写法整条溜过去。\n3. **upsert 不在禁止模式里**：它与 update/delete 同源（同样只接受唯一键定位），\n TOCTOU 缝隙一模一样，却从未被拦。\n4. **where 里的 tenantId 从未被验证**：门禁 reason 一直宣称 where 是 `{ id, tenantId }`，\n 但只检查了方法名与 state 前置条件——`updateMany({ where: { id } })` 照样绿，\n 而那正是跨租户写。\n5. **0 行拒绝可被一行注释满足**：断言对原始源码做 includes，\n `// if (updated.count === 0)` 就能让一段被注释掉的防护报绿。\n\n修法：\n- 新增 scripts/lib/source-text.mjs 作为公共单源（stripComments / blankComments / lineOf）。\n 「注释掉的代码仍满足断言」这个形态 check-schema-sync 在 2026-08-15 已实锤并就地修过，\n 但同类断言散布在多个脚本里各自对原始源码匹配——按 P5，实锤过的教训必须推广，\n 否则只是把同一个洞留在别处。\n- A 部分改为「注释置空但保持字符偏移」后整体跨行匹配，再由偏移反推行号\n （逐行挡不住折行，整体匹配又需要能定位）；禁止模式补 upsert 与方括号访问。\n- B 部分先提取 updateMany 的 where 块再逐项断言（state 前置条件 + tenantId），\n 且全部在剥注释后的源码上跑。\n\n验收：五条绕过逐一注入验证必红、逐一恢复现场；正向 10 项静态门禁 + 治理棘轮全绿。\nCLAUDE.md 基线表 write-guard 行同步为实际覆盖面（此前的描述比实际防线宽）。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:32:47-07:00"},{"Sha1":"c494fc7d2e6b1497647b2ca9601766626dd4775d","Message":"feat(auth): G8 生产级租户验签闭环——回退翻转 + 401 判责 + 降级通道清除（批次③）\n\nCLAUDE.md 里标注「动工前须用户点名」的最大 OPEN 缺口，验签半边落地。\n\n## 回退翻转（G8 的核心）\n\ndemo 期口径是 `Bearer 解析 ?? x-tenant-id`。接入验签后这个 `??` 会变成**降级攻击面**：\n伪造一个签名无效的 Bearer → 验签失败 → 自动落回客户端自报的 demo 头，等于没验。\n且只翻转这一处并不够——裸 `?tenantId=` 是另一条更彻底的通道，它连伪造 token 都不需要。\n\n- jwt 模式下：带 Authorization 或 ?access_token= 即**强制验签，失败一律 401，绝不回退**；\n x-tenant-id / ?tenantId= / `tenant:\u003cid\u003e` 简写整条不承认\n- demo 模式保留原型四级来源（本地与既有验收测试零改动），但**生产跑 demo 即拒绝启动**——\n 漏配鉴权应当是「起不来」，不是「起来了但谁都能进」\n\n## 架构切分：策略在 contracts，验签由后端注入\n\ncontracts 被 web 直接 import，把 jose 与密钥语义拉进去会泄漏到浏览器包。故：\n- `decideTenantResolution`（谁可信、失败怎么判）在 contracts，两后端共用，13 条单测覆盖\n- jose 验签在各后端（NestJS `src/auth/`、Fastify `src/plugins/auth.ts`）\n- `tenantFromBearerToken` 更名 `unsafeTenantClaimFromToken`：函数名里的 unsafe 是给 review 看的，\n 它只该用于浏览器 UI 预填与 demo 模式\n\nNestJS 的判定必须放 Guard 而非 `@TenantId()` 参数装饰器——后者的回调是同步的，\n判定留在那里就永远接不上异步的 JWKS 验签。\n\n## 顺带堵掉的三个额外攻击面（安全审计发现，均在已登记 G8 描述之外）\n\n1. **tenantId 是攻击者可控的自由文本**：无格式/长度约束时可凭空创建租户分区、\n 可用超长值撑爆 `@@unique([tenantId,email])` 的 btree 索引行打出稳定 500、\n 可用换行伪造日志行。现统一过 normalizeTenantId（≤128 + 受控字符集）。\n2. **重复查询参数的类型混淆**：Fastify 把 `?access_token=a\u0026access_token=b` 解析成数组，\n `typeof v === 'string'` 判定直接落空 → 静默跳过 token 分支 → 落到裸 ?tenantId=；\n 而 NestJS 的 `URLSearchParams.get()` 取第一个值。**同一条请求两后端两种身份判定**——\n 既是真实越权通道，也是 G17 行为分叉的活体实证。两侧统一取第一个值。\n3. **凭证进日志**：浏览器 EventSource/WS 无法自定义头，token 只能进 `?access_token=`，\n 而 pino 默认序列化器把完整 url 写进日志——接入真实验签后就是每次建连落盘一枚有效 JWT。\n Fastify logger 加 redact（authorization / cookie / req.url）。\n\n## 判责矩阵补上 401\n\n此前缺凭证一律回 400，把「未认证」伪装成「参数写错了」，客户端无从区分「该去登录」\n和「该改请求」。现 contracts 定义 UNAUTHENTICATED / INVALID_TOKEN / TENANT_FORBIDDEN；\n响应体**只回稳定错误码**，验签失败细节只进日志（区分「签名不对/已过期/issuer 不符」\n对攻击者是免费的探测反馈）。WS 保持 close(1008) 而非握手期 401——两后端同口径，\n且浏览器对握手 401 不暴露任何原因。\n\n## 验收\n\n真实 PG + Redis：**192 tests / 0 failures**（地板 161 → 192，只收紧）。\n两后端各 9 条 jwt 判责用例：正确签名放行 / 无凭证 401 / **伪造签名 + 合法 x-tenant-id 仍 401**\n/ 裸 demo 头 401 / 过期 401 / alg:none 401 / 无租户 claim 401 / health 免鉴权 / 重复参数不落 victim。\n新增 check:dual-backend 的 G8 绊网 5 条，每条均有负向红证据；其中「算法白名单」一条初版\n被自身的类型声明 `algorithms: string[]` 满足而无法变红，已改绑 jwtVerify 实参后重测。\n静态门禁 10/10 + 治理棘轮 + typecheck 5/5 + lint 2/2 全绿。\n\nRLS 未叠加（结构已就绪，阻碍在 CI 超级用户/单例 Prisma/dispatcher 跨租户扫描三处），\nweb 构建期内联 token 与 WS Origin 白名单同样登记为 G8 剩余项。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:27:41-07:00"}],"HeadCommit":{"Sha1":"4a707c3d3ec98b8ab45864a1b40b941589433914","Message":"chore(reports): 刷新治理证据至 b1510b5（clean，192 tests）\n\n静态 11 项门禁 + 治理棘轮 + lint + typecheck 全绿；\n真实 PG + Redis 验收 192 tests / 0 failures，provenance 绑定 clean commit b1510b5。\n\n⚠️ 作用域：本轮证据只覆盖静态与运行态两级。浏览器级 check:ui **未重跑**——\n本机无 Node 运行时（全部门禁在 node:22 容器内挂载工作树执行），而容器内没有\nLinux 版 Playwright 浏览器（宿主缓存的是 macOS 构建）。按事实作用域表，\n不得据此宣称 UI 层已验证。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:40:42-07:00"},"CompareURL":"luoanwu/base-framework/compare/53991090b821b76dff11eee3a2e92806ed3ab173...4a707c3d3ec98b8ab45864a1b40b941589433914","Len":10}...
|
1788064993
|
Edit
Delete
|
|
26782
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/governance/deep-optimization
|
0
|
{"Commits":[{"Sha1":"4a707c3d3 {"Commits":[{"Sha1":"4a707c3d3ec98b8ab45864a1b40b941589433914","Message":"chore(reports): 刷新治理证据至 b1510b5（clean，192 tests）\n\n静态 11 项门禁 + 治理棘轮 + lint + typecheck 全绿；\n真实 PG + Redis 验收 192 tests / 0 failures，provenance 绑定 clean commit b1510b5。\n\n⚠️ 作用域：本轮证据只覆盖静态与运行态两级。浏览器级 check:ui **未重跑**——\n本机无 Node 运行时（全部门禁在 node:22 容器内挂载工作树执行），而容器内没有\nLinux 版 Playwright 浏览器（宿主缓存的是 macOS 构建）。按事实作用域表，\n不得据此宣称 UI 层已验证。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:40:42-07:00"},{"Sha1":"b1510b5a203f7847dc99a0962cd30f36e6901dce","Message":"docs(standards): 修正与实现相悖的 API 标准 + 幽灵路径机器化（批次④·下）\n\n## api-standard 的四项核心约定与实现系统性相悖\n\nCLAUDE.md 真源地图把 docs/standards/*.md 列为「工程标准口径」，使用入口表还直接路由读者\n去读它。但 api-standard 的四项核心约定没有一项与两后端实际实现一致：\n`/api/v1/` 前缀（实际 `/api`，未版本化）、`{ data: ... }` 响应包裹（实际裸资源）、\n嵌套 `error.{code,statusCode,timestamp,path}` 错误体（实际是顶层扁平的 ApiErrorBody）、\n以及一个从不存在的 `packages/contracts/src/errors/error-codes.ts`。\n\n照着它落地新模块，会做出与 contracts `ApiErrorBody` 及 web `ApiError` 消费方直接冲突的接口。\n而 review-standard 的检查清单还会让 reviewer 去核对那个幽灵文件。\n\n- 按本仓既有的 C14「标准 vs 现状」惯例（frontend-standard / backend-standard / git-standard\n 都已有）在 api-standard 头部加对照表，逐项写明目标态与当前真相，并说明改动方向：\n 要落地目标态先改实现与契约，不要反过来照着文档改接口\n- 修正 review-standard 的检查项指向真实位置\n\n## 幽灵路径机器化（C14-⑤ 同族，扫描面扩到 standards）\n\n新增 docs-truth 断言：standards 引用的 `packages/contracts/src/**.ts` 路径必须真实存在，\n否则须在同一行显式标注「目标态 / 当前不存在」。扫描面用 readdirSync 覆盖 docs/standards 全量，\n不硬编码文件清单——旧的幽灵路径断言正是因为清单式扫描漏掉了 docs/README.md 本身而失守过。\n\n该断言上线即多抓到一处此前没人发现的幽灵路径（naming-standard 的\n`packages/contracts/src/store/store.types.ts`），已一并标注。\n负向测试：去掉 review-standard 的诚实标注即红。\n\n## owner-matrix 补录 OS 产品两张业务表\n\n「先登记后建表」是本仓自己立的规则，却被自己的 C20/G23 战役打破：OS 产品包新建了\n`juhai_baseframework_orders` / `..._order_events` 两张带状态机的表却从未登记。\n补录两行并写明 `products/` 下的对象同样要登记（归属真源是 product.manifest.json 的\npersistence.tenantModels，本表挂索引行）。\n\n验收：naming / docs-truth / governance-docs / 治理棘轮均 exit 0。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:39:02-07:00"},{"Sha1":"033e8159c23f4766493fd140ffa7aa923ebe1182","Message":"fix(ops): 优雅停机接线——Fastify 的 onClose 钩子此前全是死代码（批次④·中）\n\n## Fastify 从不优雅停机\n\nserver.ts 没有任何信号处理，也从不调用 app.close()：进程收到 SIGTERM 走默认处理直接终止。\n后果是 prisma / redis / outbox / queue 四个插件里**精心写好的 onClose 钩子从未被执行过**——\nBullMQ worker 不 close（在途 job 只能靠 stalled 恢复）、outbox 轮询事务可能被拦腰砍断、\nWS/SSE 连接收不到 close 帧。而 NestJS 侧早有 enableShutdownHooks，两后端停机行为实质分叉。\n\n存在性门禁看得见「钩子写了」，看不见「它从来没被调用过」——这正是 G17 说的行为级盲区。\n\n- server.ts 补 SIGTERM/SIGINT → app.close()（Fastify 会按注册逆序执行全部 onClose），\n 带 10s 超时兜底（卡住的清理不能让容器一直不退）与重复信号防抖\n\n## NestJS 的 Redis 连接从不 quit，且注释谎称有人管\n\nREDIS_PUB / REDIS_SUB 由 useFactory 裸建，没有任何销毁钩子；而 event-bus.service.ts\n的注释写着「redisSub 连接由 RedisModule 统一关闭」——注释与行为相反，于是没人会去查。\n真实后果：停机时连接不回收（两个 acceptance 测试的 afterAll 至今还得手动 quit 才能让进程退出）。\n\n- RedisModule 实现 onApplicationShutdown，quit 两个连接\n- 修正 event-bus 的失实注释\n\n## 防回潮\n\ncheck:dual-backend 新增 4 条停机接线绊网（两后端各自的信号处理 / app.close /\nenableShutdownHooks / Redis 关闭钩子），负向测试已验证删掉信号处理即红。\n\n验收：真实 PG + Redis 192 tests / 0 failures；静态门禁 + 棘轮 + typecheck 5/5 + lint 2/2 全绿。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:35:43-07:00"},{"Sha1":"44cbbaeacf5f53e08e6e3f48a6d8bd2c3327d2de","Message":"fix(governance): 写链守护补齐五条已验证绕过（批次④·上）\n\ncheck:write-guard 守的是 P4「所有写显式带 tenant_id + 状态迁移乐观锁」这条最高优先级红线，\n但它自身存在五条可复现的绕过——每一条都能让违规代码拿到绿盘：\n\n1. **折行**：逐行匹配下 `await tx.order\\n .update({...})` 完全隐形。这甚至无需恶意——\n Prettier 对长链式调用本来就这样折行。\n2. **方括号访问**：`tx[\"order\"].update(` 用 `\\w+` 匹配不到，换个写法整条溜过去。\n3. **upsert 不在禁止模式里**：它与 update/delete 同源（同样只接受唯一键定位），\n TOCTOU 缝隙一模一样，却从未被拦。\n4. **where 里的 tenantId 从未被验证**：门禁 reason 一直宣称 where 是 `{ id, tenantId }`，\n 但只检查了方法名与 state 前置条件——`updateMany({ where: { id } })` 照样绿，\n 而那正是跨租户写。\n5. **0 行拒绝可被一行注释满足**：断言对原始源码做 includes，\n `// if (updated.count === 0)` 就能让一段被注释掉的防护报绿。\n\n修法：\n- 新增 scripts/lib/source-text.mjs 作为公共单源（stripComments / blankComments / lineOf）。\n 「注释掉的代码仍满足断言」这个形态 check-schema-sync 在 2026-08-15 已实锤并就地修过，\n 但同类断言散布在多个脚本里各自对原始源码匹配——按 P5，实锤过的教训必须推广，\n 否则只是把同一个洞留在别处。\n- A 部分改为「注释置空但保持字符偏移」后整体跨行匹配，再由偏移反推行号\n （逐行挡不住折行，整体匹配又需要能定位）；禁止模式补 upsert 与方括号访问。\n- B 部分先提取 updateMany 的 where 块再逐项断言（state 前置条件 + tenantId），\n 且全部在剥注释后的源码上跑。\n\n验收：五条绕过逐一注入验证必红、逐一恢复现场；正向 10 项静态门禁 + 治理棘轮全绿。\nCLAUDE.md 基线表 write-guard 行同步为实际覆盖面（此前的描述比实际防线宽）。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:32:47-07:00"},{"Sha1":"c494fc7d2e6b1497647b2ca9601766626dd4775d","Message":"feat(auth): G8 生产级租户验签闭环——回退翻转 + 401 判责 + 降级通道清除（批次③）\n\nCLAUDE.md 里标注「动工前须用户点名」的最大 OPEN 缺口，验签半边落地。\n\n## 回退翻转（G8 的核心）\n\ndemo 期口径是 `Bearer 解析 ?? x-tenant-id`。接入验签后这个 `??` 会变成**降级攻击面**：\n伪造一个签名无效的 Bearer → 验签失败 → 自动落回客户端自报的 demo 头，等于没验。\n且只翻转这一处并不够——裸 `?tenantId=` 是另一条更彻底的通道，它连伪造 token 都不需要。\n\n- jwt 模式下：带 Authorization 或 ?access_token= 即**强制验签，失败一律 401，绝不回退**；\n x-tenant-id / ?tenantId= / `tenant:\u003cid\u003e` 简写整条不承认\n- demo 模式保留原型四级来源（本地与既有验收测试零改动），但**生产跑 demo 即拒绝启动**——\n 漏配鉴权应当是「起不来」，不是「起来了但谁都能进」\n\n## 架构切分：策略在 contracts，验签由后端注入\n\ncontracts 被 web 直接 import，把 jose 与密钥语义拉进去会泄漏到浏览器包。故：\n- `decideTenantResolution`（谁可信、失败怎么判）在 contracts，两后端共用，13 条单测覆盖\n- jose 验签在各后端（NestJS `src/auth/`、Fastify `src/plugins/auth.ts`）\n- `tenantFromBearerToken` 更名 `unsafeTenantClaimFromToken`：函数名里的 unsafe 是给 review 看的，\n 它只该用于浏览器 UI 预填与 demo 模式\n\nNestJS 的判定必须放 Guard 而非 `@TenantId()` 参数装饰器——后者的回调是同步的，\n判定留在那里就永远接不上异步的 JWKS 验签。\n\n## 顺带堵掉的三个额外攻击面（安全审计发现，均在已登记 G8 描述之外）\n\n1. **tenantId 是攻击者可控的自由文本**：无格式/长度约束时可凭空创建租户分区、\n 可用超长值撑爆 `@@unique([tenantId,email])` 的 btree 索引行打出稳定 500、\n 可用换行伪造日志行。现统一过 normalizeTenantId（≤128 + 受控字符集）。\n2. **重复查询参数的类型混淆**：Fastify 把 `?access_token=a\u0026access_token=b` 解析成数组，\n `typeof v === 'string'` 判定直接落空 → 静默跳过 token 分支 → 落到裸 ?tenantId=；\n 而 NestJS 的 `URLSearchParams.get()` 取第一个值。**同一条请求两后端两种身份判定**——\n 既是真实越权通道，也是 G17 行为分叉的活体实证。两侧统一取第一个值。\n3. **凭证进日志**：浏览器 EventSource/WS 无法自定义头，token 只能进 `?access_token=`，\n 而 pino 默认序列化器把完整 url 写进日志——接入真实验签后就是每次建连落盘一枚有效 JWT。\n Fastify logger 加 redact（authorization / cookie / req.url）。\n\n## 判责矩阵补上 401\n\n此前缺凭证一律回 400，把「未认证」伪装成「参数写错了」，客户端无从区分「该去登录」\n和「该改请求」。现 contracts 定义 UNAUTHENTICATED / INVALID_TOKEN / TENANT_FORBIDDEN；\n响应体**只回稳定错误码**，验签失败细节只进日志（区分「签名不对/已过期/issuer 不符」\n对攻击者是免费的探测反馈）。WS 保持 close(1008) 而非握手期 401——两后端同口径，\n且浏览器对握手 401 不暴露任何原因。\n\n## 验收\n\n真实 PG + Redis：**192 tests / 0 failures**（地板 161 → 192，只收紧）。\n两后端各 9 条 jwt 判责用例：正确签名放行 / 无凭证 401 / **伪造签名 + 合法 x-tenant-id 仍 401**\n/ 裸 demo 头 401 / 过期 401 / alg:none 401 / 无租户 claim 401 / health 免鉴权 / 重复参数不落 victim。\n新增 check:dual-backend 的 G8 绊网 5 条，每条均有负向红证据；其中「算法白名单」一条初版\n被自身的类型声明 `algorithms: string[]` 满足而无法变红，已改绑 jwtVerify 实参后重测。\n静态门禁 10/10 + 治理棘轮 + typecheck 5/5 + lint 2/2 全绿。\n\nRLS 未叠加（结构已就绪，阻碍在 CI 超级用户/单例 Prisma/dispatcher 跨租户扫描三处），\nweb 构建期内联 token 与 WS Origin 白名单同样登记为 G8 剩余项。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:27:41-07:00"}],"HeadCommit":{"Sha1":"4a707c3d3ec98b8ab45864a1b40b941589433914","Message":"chore(reports): 刷新治理证据至 b1510b5（clean，192 tests）\n\n静态 11 项门禁 + 治理棘轮 + lint + typecheck 全绿；\n真实 PG + Redis 验收 192 tests / 0 failures，provenance 绑定 clean commit b1510b5。\n\n⚠️ 作用域：本轮证据只覆盖静态与运行态两级。浏览器级 check:ui **未重跑**——\n本机无 Node 运行时（全部门禁在 node:22 容器内挂载工作树执行），而容器内没有\nLinux 版 Playwright 浏览器（宿主缓存的是 macOS 构建）。按事实作用域表，\n不得据此宣称 UI 层已验证。\n\nCo-Authored-By: Claude Opus 4.8 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-08-29T20:40:42-07:00"},"CompareURL":"luoanwu/base-framework/compare/53991090b821b76dff11eee3a2e92806ed3ab173...4a707c3d3ec98b8ab45864a1b40b941589433914","Len":10}...
|
1788061409
|
Edit
Delete
|
|
26781
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/governance/deep-optimization
|
0
|
|
1788061409
|
Edit
Delete
|
|
26405
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"53991090b {"Commits":[{"Sha1":"53991090b821b76dff11eee3a2e92806ed3ab173","Message":"chore(reports): 刷新 OS 产品验收证据\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-26T16:22:27-07:00"},{"Sha1":"71e6ef64d1ee41cb39ffae15a73f049c42fb82b1","Message":"feat(os): 增加可装载基础框架产品\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-26T16:20:54-07:00"}],"HeadCommit":{"Sha1":"53991090b821b76dff11eee3a2e92806ed3ab173","Message":"chore(reports): 刷新 OS 产品验收证据\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-26T16:22:27-07:00"},"CompareURL":"luoanwu/base-framework/compare/9513a941af257b779d7d2c095dff71172edcefaf...53991090b821b76dff11eee3a2e92806ed3ab173","Len":2}...
|
1787787571
|
Edit
Delete
|
|
24399
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"9513a941a {"Commits":[{"Sha1":"9513a941af257b779d7d2c095dff71172edcefaf","Message":"chore(reports): 刷新静态门禁证据至 e15fa81\n\nCLAUDE.md 改动后静态子报告的 provenance 仍指向 ccbc413，与当前 checkout\n不一致（report-provenance 门禁会红）。在 clean commit 上重跑 pnpm check，\n14 份静态报告统一记为 gitSha=e15fa81 / worktreeDirty=false。\n\nruntime/ui 报告未重跑：本轮为纯文档改动，未触碰任何被它们覆盖的代码，\n其证据仍诚实绑定 c223f35 的实跑，不做无意义刷新。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-19T08:53:15-07:00"},{"Sha1":"e15fa814108637f2484a5be50900fd35a0f05d9d","Message":"docs(claude): 补齐常用命令与运行时拓扑，收敛验证命令双源\n\n对照 /init 的两条标准补齐本文件的实际缺口，治理正文一字未动（纯增量）。\n\n1) 新增「常用命令」小节\n原先只有文件最末尾一个块（install / generate / 三级门禁），缺三类信息：\n- 开发入口与端口（web:3000 / NestJS:3001 / Fastify:3002）\n- 单个测试怎么跑（单文件 / -t 按用例名 / watch）\n- 13 个 check:* 子门禁只写了聚合入口，未说明可单独执行\n同时记入两条本机实证的判责提示：\n- pnpm 锁定 9.15.9，被 pnpm 10 装过会写 store v11，须全量重装 + prisma:generate\n- G15 基座守卫只校验库名前缀、**不校验端口**，故「端口连得上」推不出\n 「基座正确」（2026-08-19 在下游 Digital Employee OS 实际踩到：误用上游\n 容器 PG:55432/Redis:6382 通过守卫，共用 Redis 导致两次时序偶发红）\n\n2) 新增「运行时拓扑」小节\n「真源地图」回答的是「哪份文件说了算」，不回答「请求怎么流」。补一条从\nHTTP 写请求到前端缓存失效的完整事件通路，以及两条读单个文件发现不了的约定：\nNestJS 全局前缀 /api 但 GET /sse/events 被 exclude；WS 用原生 ws 而非\nsocket.io，故协议与 Fastify 侧兼容。\n\n3) 收敛验证命令双源\n文末旧「验证命令」块与新章节重复——同一事实两处维护正是 P1 禁止的漂移温床。\n改为指回顶部，只保留各级门禁的覆盖范围差异（含 check:ui 不覆盖 Fastify\n链路这条 G17 事实）。\n\nAGENTS.md 符号链接完好，check:governance-docs / check:docs-truth 均绿。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-19T08:52:41-07:00"}],"HeadCommit":{"Sha1":"9513a941af257b779d7d2c095dff71172edcefaf","Message":"chore(reports): 刷新静态门禁证据至 e15fa81\n\nCLAUDE.md 改动后静态子报告的 provenance 仍指向 ccbc413，与当前 checkout\n不一致（report-provenance 门禁会红）。在 clean commit 上重跑 pnpm check，\n14 份静态报告统一记为 gitSha=e15fa81 / worktreeDirty=false。\n\nruntime/ui 报告未重跑：本轮为纯文档改动，未触碰任何被它们覆盖的代码，\n其证据仍诚实绑定 c223f35 的实跑，不做无意义刷新。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-19T08:53:15-07:00"},"CompareURL":"luoanwu/base-framework/compare/ccbc413eaf556c59a5c14707ce87d264c050180f...9513a941af257b779d7d2c095dff71172edcefaf","Len":2}...
|
1787154806
|
Edit
Delete
|
|
24398
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"ccbc413ea {"Commits":[{"Sha1":"ccbc413eaf556c59a5c14707ce87d264c050180f","Message":"docs(migration): 回灌 P9——文档内部数字自相矛盾\n\n下游 Digital Employee OS 移植时实测：只改了 CLAUDE.md 的「证据新鲜度」块\n（49 tests），漏了下方 GOVERNANCE-BASELINE 表（仍是源仓的 11/60/56），\n同一份文件里两个数字互相矛盾，而三级门禁全绿。\n\ncheck:docs-truth 是 targeted 断言集，不覆盖数字类事实，抓不到这类漂移。\n数字矛盾比数字过时更险：过时只是陈旧，矛盾意味着至少有一个是编的，\n而读者无从判断是哪个。\n\n按 §5 回灌纪律（「换个移植目标还会再踩吗」→ 会 → 回灌上游），三处同步补齐：\n- 踩坑档案新增 P9\n- 步骤 8 显式点名 GOVERNANCE-BASELINE 表每个数字与发布状态快照，\n 并加警告说明 check:docs-truth 的覆盖边界\n- 验收标准把「证据日期」扩为「证据日期 + 基线表每个数字」\n\n暂未机器化：数字类断言仍靠人工核对，属手册明记的已知缺口。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-19T05:10:13-07:00"}],"HeadCommit":{"Sha1":"ccbc413eaf556c59a5c14707ce87d264c050180f","Message":"docs(migration): 回灌 P9——文档内部数字自相矛盾\n\n下游 Digital Employee OS 移植时实测：只改了 CLAUDE.md 的「证据新鲜度」块\n（49 tests），漏了下方 GOVERNANCE-BASELINE 表（仍是源仓的 11/60/56），\n同一份文件里两个数字互相矛盾，而三级门禁全绿。\n\ncheck:docs-truth 是 targeted 断言集，不覆盖数字类事实，抓不到这类漂移。\n数字矛盾比数字过时更险：过时只是陈旧，矛盾意味着至少有一个是编的，\n而读者无从判断是哪个。\n\n按 §5 回灌纪律（「换个移植目标还会再踩吗」→ 会 → 回灌上游），三处同步补齐：\n- 踩坑档案新增 P9\n- 步骤 8 显式点名 GOVERNANCE-BASELINE 表每个数字与发布状态快照，\n 并加警告说明 check:docs-truth 的覆盖边界\n- 验收标准把「证据日期」扩为「证据日期 + 基线表每个数字」\n\n暂未机器化：数字类断言仍靠人工核对，属手册明记的已知缺口。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-19T05:10:13-07:00"},"CompareURL":"luoanwu/base-framework/compare/4d95adb2ebe583e4dddbd233cbb4ffb574b17862...ccbc413eaf556c59a5c14707ce87d264c050180f","Len":1}...
|
1787141425
|
Edit
Delete
|
|
24397
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"4d95adb2e {"Commits":[{"Sha1":"4d95adb2ebe583e4dddbd233cbb4ffb574b17862","Message":"chore(reports): refresh clean publish evidence for realtime work\n\n三级门禁在 clean commit c223f35 上重跑，所有 reports/*.latest.json 的\nprovenance 记为 gitSha=c223f35 / worktreeDirty=false——证据自此绑定具体\n提交，而非「某次本地 dirty 运行」。\n\nCLAUDE.md 证据新鲜度同步改为绑定该提交，并显式声明这只证明「该提交在\n本机通过三级门禁」，推不出远端 CI 已拦截或已部署（G14 仍 OPEN）。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-19T04:35:15-07:00"},{"Sha1":"c223f35e380114edfe1331619928f405cf77eee2","Message":"feat(realtime): 可靠投递 + 定向推送 + 双向控制帧 + 连接可观测\n\n把实时通讯从单向广播推进到「可靠 + 定向 + 双向 + 可观测」四块，\n两后端对等实现，判定逻辑统一收敛到 @repo/contracts 单源。\n\nA 可靠投递闭环（G16 主体闭环）\n- 信封加 id（= outbox 行 id，稳定幂等键）与 seq（自增重放游标）；\n 前端去重从内容指纹改为 envelope.id——指纹依赖「重投时 ts 不变」\n 这个不成立的假设，补发路径上必然误判。\n- dispatcher 改 SELECT ... FOR UPDATE SKIP LOCKED 领取，publish 在同一\n 事务内完成；先提交再 publish 会从 at-least-once 退化成 at-most-once。\n- 断线重放：SSE 走标准 Last-Event-ID，WS 走 ?since=；补不齐时回\n replay.gap 让客户端全量重取，绝不静默截断。\n- 新增 outbox.seq 迁移（显式建序列，DEFAULT 在 SQL 里可见）。\n\nB 定向推送（订阅过滤）\n- matchesSubscription / parseSubscription 单源在 contracts；\n types 前缀 + ids 维度，未声明 = 收本租户全部。\n- 定位是租户内的带宽减法，不是权限边界；租户隔离仍是唯一安全边界。\n- ping / replay.gap / connection.* 为连接级消息，无视订阅条件必送。\n\nC 双向通讯（刻意限定为连接控制平面）\n- WS 接受 subscribe / ping 控制帧，应答 connection.ack / connection.error。\n- 红线：控制帧不承载任何业务写入。HTTP 写链上挂着 Zod、状态机守卫、\n 租户绑定、outbox 同事务（P4 四同边界），WS 若能写业务，要么重实现\n 一遍那套保证（P1 双真源），要么绕过（写链治理失守）。\n- 三道闸按成本排序：限流 → 大小/解析 → 应用。\n\nD 连接可观测与背压\n- GET /api/realtime/stats 按请求方租户收口：租户只从鉴权上下文取，\n 不接受 ?tenantId= 覆盖，也不返回全局/分租户明细。\n- 单租户连接上限背压：超限 WS 回 1013、SSE 回 429；被拒连接不登记计数。\n\n验收（本地工作区，PG:55432 + Redis:6382）\n- pnpm check 全绿；pnpm check:runtime 155 tests / 0 failures；\n pnpm check:ui 4 用例 / 0 失败。\n- 棘轮收紧：ownTests 11→24、ownTestCases 60→146、testsPassed 56→155。\n- 负向测试已验证有牙：退回裸 findMany → 20 行被投递 40 次；摘掉订阅\n 过滤 → types/ids 断言红；stats 接受调用方指定租户 → 越权探测红。\n\n治理回灌：CLAUDE.md 真源地图新增 5 条实时口径、基线表新增 5 行指标、\nG16 改为「部分闭环」并列明仍缺的死信告警与多实例端到端验收；\n战役叙事外迁 docs/governance-experience.md。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-19T04:33:54-07:00"}],"HeadCommit":{"Sha1":"4d95adb2ebe583e4dddbd233cbb4ffb574b17862","Message":"chore(reports): refresh clean publish evidence for realtime work\n\n三级门禁在 clean commit c223f35 上重跑，所有 reports/*.latest.json 的\nprovenance 记为 gitSha=c223f35 / worktreeDirty=false——证据自此绑定具体\n提交，而非「某次本地 dirty 运行」。\n\nCLAUDE.md 证据新鲜度同步改为绑定该提交，并显式声明这只证明「该提交在\n本机通过三级门禁」，推不出远端 CI 已拦截或已部署（G14 仍 OPEN）。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-08-19T04:35:15-07:00"},"CompareURL":"luoanwu/base-framework/compare/03ef2eb6d9d8f94868d95821c22f3c23f7dc7f38...4d95adb2ebe583e4dddbd233cbb4ffb574b17862","Len":2}...
|
1787139335
|
Edit
Delete
|
|
20056
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"03ef2eb6d {"Commits":[{"Sha1":"03ef2eb6d9d8f94868d95821c22f3c23f7dc7f38","Message":"chore(reports): refresh clean publish evidence\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-30T07:43:05-07:00"}],"HeadCommit":{"Sha1":"03ef2eb6d9d8f94868d95821c22f3c23f7dc7f38","Message":"chore(reports): refresh clean publish evidence\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-30T07:43:05-07:00"},"CompareURL":"luoanwu/base-framework/compare/94e3f1b529034c902d205637ed69033238b5b594...03ef2eb6d9d8f94868d95821c22f3c23f7dc7f38","Len":1}...
|
1785422600
|
Edit
Delete
|
|
20055
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"94e3f1b52 {"Commits":[{"Sha1":"94e3f1b529034c902d205637ed69033238b5b594","Message":"docs(governance): record Gitea publish baseline\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-30T07:41:37-07:00"}],"HeadCommit":{"Sha1":"94e3f1b529034c902d205637ed69033238b5b594","Message":"docs(governance): record Gitea publish baseline\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-30T07:41:37-07:00"},"CompareURL":"luoanwu/base-framework/compare/da8f52db37f5b42d035f92ad27a45067bc57dfa9...94e3f1b529034c902d205637ed69033238b5b594","Len":1}...
|
1785422513
|
Edit
Delete
|
|
20054
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"da8f52db3 {"Commits":[{"Sha1":"da8f52db37f5b42d035f92ad27a45067bc57dfa9","Message":"feat(governance): harden report provenance and UI acceptance\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-30T07:40:06-07:00"},{"Sha1":"4940be24ba6bfff2eefb2284f8ce66c9cbd1f326","Message":"docs(governance): record remote branch consolidation\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-22T01:35:33-07:00"}],"HeadCommit":{"Sha1":"da8f52db37f5b42d035f92ad27a45067bc57dfa9","Message":"feat(governance): harden report provenance and UI acceptance\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-30T07:40:06-07:00"},"CompareURL":"luoanwu/base-framework/compare/02a656716b10d2a5a4937ba01086cbd0a9ff4b85...da8f52db37f5b42d035f92ad27a45067bc57dfa9","Len":2}...
|
1785422433
|
Edit
Delete
|
|
18224
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"02a656716 {"Commits":[{"Sha1":"02a656716b10d2a5a4937ba01086cbd0a9ff4b85","Message":"chore(reports): refresh post-merge acceptance evidence\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-22T01:34:39-07:00"}],"HeadCommit":{"Sha1":"02a656716b10d2a5a4937ba01086cbd0a9ff4b85","Message":"chore(reports): refresh post-merge acceptance evidence\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-22T01:34:39-07:00"},"CompareURL":"luoanwu/base-framework/compare/e109e80574304306c9a5911d5e923d2dc7e6c0f4...02a656716b10d2a5a4937ba01086cbd0a9ff4b85","Len":1}...
|
1784709310
|
Edit
Delete
|
|
18223
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"e109e8057 {"Commits":[{"Sha1":"e109e80574304306c9a5911d5e923d2dc7e6c0f4","Message":"docs(governance): record verified publish baseline\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-22T01:27:34-07:00"}],"HeadCommit":{"Sha1":"e109e80574304306c9a5911d5e923d2dc7e6c0f4","Message":"docs(governance): record verified publish baseline\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-22T01:27:34-07:00"},"CompareURL":"luoanwu/base-framework/compare/4e0ae0a55bf748faaa3a1f2b937b4375c1cf7fa8...e109e80574304306c9a5911d5e923d2dc7e6c0f4","Len":1}...
|
1784709216
|
Edit
Delete
|
|
18222
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"4e0ae0a55 {"Commits":[{"Sha1":"4e0ae0a55bf748faaa3a1f2b937b4375c1cf7fa8","Message":"chore(reports): refresh publish acceptance evidence\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-22T01:26:24-07:00"}],"HeadCommit":{"Sha1":"4e0ae0a55bf748faaa3a1f2b937b4375c1cf7fa8","Message":"chore(reports): refresh publish acceptance evidence\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-22T01:26:24-07:00"},"CompareURL":"luoanwu/base-framework/compare/d846832ebbe1d481019744f038672b935ad09937...4e0ae0a55bf748faaa3a1f2b937b4375c1cf7fa8","Len":1}...
|
1784708793
|
Edit
Delete
|
|
18221
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"d846832eb {"Commits":[{"Sha1":"d846832ebbe1d481019744f038672b935ad09937","Message":"fix(web): scope E2E user-CRUD assertion to users panel, not full page\n\nPorted from the claude/gallant-diffie-a541dd worktree, whose committed\nhistory was already fully contained in main but whose working tree held\nthis un-landed fix.\n\nNarrow the delete-then-hidden assertion in acceptance.spec.ts to\ngetByTestId(\"users-panel\") instead of the whole page. After d799b17's\noutbox fix, user.created events reliably reach the EventFeed and echo\nthe created user's name in the event JSON, so a page-wide getByText\ncould still match the name via the event log after the row itself was\nremoved, producing a false \"still visible\" read on flaky runs.\n\nRecords this as C16 in CLAUDE.md / governance-experience.md (C14 was\nalready taken by the docs-truth work landed in 2fa1666).\n\nCo-Authored-By: Claude Sonnet 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-22T01:23:10-07:00"}],"HeadCommit":{"Sha1":"d846832ebbe1d481019744f038672b935ad09937","Message":"fix(web): scope E2E user-CRUD assertion to users panel, not full page\n\nPorted from the claude/gallant-diffie-a541dd worktree, whose committed\nhistory was already fully contained in main but whose working tree held\nthis un-landed fix.\n\nNarrow the delete-then-hidden assertion in acceptance.spec.ts to\ngetByTestId(\"users-panel\") instead of the whole page. After d799b17's\noutbox fix, user.created events reliably reach the EventFeed and echo\nthe created user's name in the event JSON, so a page-wide getByText\ncould still match the name via the event log after the row itself was\nremoved, producing a false \"still visible\" read on flaky runs.\n\nRecords this as C16 in CLAUDE.md / governance-experience.md (C14 was\nalready taken by the docs-truth work landed in 2fa1666).\n\nCo-Authored-By: Claude Sonnet 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-22T01:23:10-07:00"},"CompareURL":"luoanwu/base-framework/compare/2fa1666f6bb69b7784199199afde715a54856ac8...d846832ebbe1d481019744f038672b935ad09937","Len":1}...
|
1784708606
|
Edit
Delete
|
|
18220
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
0
|
{"Commits":[{"Sha1":"2fa1666f6 {"Commits":[{"Sha1":"2fa1666f6bb69b7784199199afde715a54856ac8","Message":"feat(governance): machineize docs-truth gate, fix Fastify Redis parsing, harden test-base guard\n\n- Add check:docs-truth (C14 machineized): targeted assertions catching README/standards\n drift from runtime truth (dual-backend positioning, tenant-aware envelope shape, NestJS\n HTTP engine, ghost assets like commitlint/RHF), wired into pnpm check + governance ratchet.\n- Fix Fastify BullMQ Redis URL parsing to include password/db (C15), matching NestJS's\n parseRedisConnection; adds a check:dual-backend tripwire so this can't silently regress.\n Without this, authenticated/non-zero-db Redis environments would split pub/sub and\n BullMQ onto different logical Redis instances.\n- Close G15: check:runtime / check:ui preflight now requires explicit REDIS_URL (no more\n silent localhost:6379 default) and enforces a base_framework* db name convention, with\n GOVERNANCE_DB_NAME_OVERRIDE=1 as an explicit escape hatch.\n- Correct stale doc claims across README/docs/standards to match actual code (Fastify is a\n full dual-backend peer per ADR-0009, not an optional fallback; envelope carries tenantId;\n NestJS uses platform-express not the Fastify adapter; annotate unimplemented commitlint/RHF\n plans as not-yet-landed).\n- Refresh governance reports/baseline after re-running pnpm check and check:runtime (56/56).\n\nCo-Authored-By: Claude Sonnet 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-22T01:11:05-07:00"},{"Sha1":"8501dc9731b4f0f44ea36d2ffe557577354ebb31","Message":"Refactor governance documentation structure\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-21T08:01:24-07:00"},{"Sha1":"d799b176f2db4f3e538563eeb8228543d24ca639","Message":"feat(runtime): close outbox failure governance gaps\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-05T07:03:20-07:00"},{"Sha1":"43b8c8c943c406733084427f3ad27d836ae75681","Message":"Update governance baseline for UI acceptance\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-04T23:37:30-07:00"}],"HeadCommit":{"Sha1":"2fa1666f6bb69b7784199199afde715a54856ac8","Message":"feat(governance): machineize docs-truth gate, fix Fastify Redis parsing, harden test-base guard\n\n- Add check:docs-truth (C14 machineized): targeted assertions catching README/standards\n drift from runtime truth (dual-backend positioning, tenant-aware envelope shape, NestJS\n HTTP engine, ghost assets like commitlint/RHF), wired into pnpm check + governance ratchet.\n- Fix Fastify BullMQ Redis URL parsing to include password/db (C15), matching NestJS's\n parseRedisConnection; adds a check:dual-backend tripwire so this can't silently regress.\n Without this, authenticated/non-zero-db Redis environments would split pub/sub and\n BullMQ onto different logical Redis instances.\n- Close G15: check:runtime / check:ui preflight now requires explicit REDIS_URL (no more\n silent localhost:6379 default) and enforces a base_framework* db name convention, with\n GOVERNANCE_DB_NAME_OVERRIDE=1 as an explicit escape hatch.\n- Correct stale doc claims across README/docs/standards to match actual code (Fastify is a\n full dual-backend peer per ADR-0009, not an optional fallback; envelope carries tenantId;\n NestJS uses platform-express not the Fastify adapter; annotate unimplemented commitlint/RHF\n plans as not-yet-landed).\n- Refresh governance reports/baseline after re-running pnpm check and check:runtime (56/56).\n\nCo-Authored-By: Claude Sonnet 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-22T01:11:05-07:00"},"CompareURL":"luoanwu/base-framework/compare/f98acd04c6fe50956d8cf89552a352c8893f037d...2fa1666f6bb69b7784199199afde715a54856ac8","Len":4}...
|
1784707902
|
Edit
Delete
|
|
15804
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
1
|
{"Commits":[{"Sha1":"f98acd04c {"Commits":[{"Sha1":"f98acd04c6fe50956d8cf89552a352c8893f037d","Message":"chore(reports): refresh runtime acceptance evidence\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-02T21:23:52-07:00"},{"Sha1":"1dc0c63efdf29c804f63ca2849e4e570a9cb9120","Message":"feat: harden runtime HTTP acceptance\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-02T21:23:06-07:00"},{"Sha1":"f8ff9b87e68449c3a4d226bf55bbb73180d7c4e8","Message":"feat: harden base framework governance\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-02T01:31:50-07:00"},{"Sha1":"4146021846591d0eb7b5cd8b797c198c00b6aec2","Message":"Add Capacitor export config and absolute API base\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-06-22T20:52:41-07:00"},{"Sha1":"3ad7df35c22d175062b25f5a0456046e65582136","Message":"feat(skills): React+Next.js 前端技能四件套纳入本仓 .claude/skills/\n\n把对应 Vue3+Element Plus 参考样板深化复刻的 React 系技能落入仓库版本管理：\n- frontend-view-react：Next.js App Router + shadcn/ui 列表/表单/详情页 + 完善页面体检清单\n- frontend-api-react：feature api 纯函数 + TanStack Query Hook + Query Key 工厂\n- fullstack-7layer-react：全栈七层（后端 1~5 同源，前端 6~7 用 React 栈）\n- zustand-store：仅 UI/全局态，落实「服务端数据绝不进 Zustand」红线\n\n锚定 docs/standards/frontend-standard.md 与 apps/web 实际运行栈；contracts 入参 Zod\n单源贯穿前端 RHF。CLAUDE.md 同步标注技能已随仓库管理。\n\nCo-Authored-By: Claude Opus 4.8 (1M context) \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-06-14T19:40:51-07:00"}],"HeadCommit":{"Sha1":"f98acd04c6fe50956d8cf89552a352c8893f037d","Message":"chore(reports): refresh runtime acceptance evidence\n","AuthorEmail":"luoguoguo@gmail.com","AuthorName":"luoguoguo","CommitterEmail":"luoguoguo@gmail.com","CommitterName":"luoguoguo","Timestamp":"2026-07-02T21:23:52-07:00"},"CompareURL":"luoanwu/base-framework/compare/7e269a0aa062715c1751b544cce7e7a8e8bd8bc7...f98acd04c6fe50956d8cf89552a352c8893f037d","Len":10}...
|
1783052672
|
Edit
Delete
|
|
15803
|
5
|
5
|
5
|
54
|
0
|
0
|
refs/heads/main
|
1
|
|
1783052672
|
Edit
Delete
|