Skip to content

chore(docs): retire 77 MB of historical captures and define retention - #7651

Merged
huangruiteng merged 1 commit into
mainfrom
codex/repository-evidence-retention-1029
Oct 29, 2026
Merged

huangruiteng merged 1 commit into
mainfrom
codex/repository-evidence-retention-1029

Conversation

@huangruiteng

Copy link
Copy Markdown
Collaborator

Problem and outcome

Generated UI review captures accumulated as permanent repository assets across successive PRs. This change removes 870 historical screenshots and capture-pack files (77,476,304 bytes) while keeping current documentation/product consumers and executable validation intact. docs/, examples/ and tests/ together fall from 186.10 MB to 108.63 MB (decimal units, about 41.6% smaller).

This is contributor/repository maintenance under the overall roadmap's S12 retirement of stale evidence. Runtime behavior, tests, synthetic fixture inputs, current README/design/guide figures and the three published videos are unchanged. It does not rewrite Git history or reduce existing historical object storage.

Scope and retention rule

Retired material Files Bytes removed
Historical nested workspace capture packs 112 11,421,474
Development captures 286 25,318,451
Browser assets/ captures 107 10,044,296
Browser evidence/ packs 323 27,385,155
Browser screenshots/ packs 30 1,975,548
Loose browser captures 12 1,331,380

The selected packs contain generated images and their execution-specific README/capture metadata. Selection includes references to those READMEs/manifests, not only direct image links. 91 candidate files remain byte-identical because current documents or their transitive references consume them; this includes the four loose acceptance-guide images. The two oracle-current code matches are screenshot writes into ignored outputDir, not fixture reads.

  • docs/development/documentation-layout.md owns the bilingual retention rule: track reproducible source/fixtures and one canonical current asset; attach one-run public-safe evidence to the PR or use durable versioned artifacts.
  • AGENTS.md and the PR template route authors to that rule while preserving first-screen, before/after, revision and data/build evidence obligations.
  • .gitignore retires the old browser evidence/ and screenshots/ output namespaces. Existing browser scripts continue writing to ignored output/playwright/.
  • The policy links the pre-cleanup packs at immutable revision bcb1a8710844bda6b84c2a00cf5d86ba78e126d9. Historical claims stay inspectable with their original source and PR, without keeping another archive copy in the current tree.

Validation

Tested source: 0747b05bc7367e52ba07a36f0b09f40357e4b09c, based on verified target loopx-project/loopx:main at bcb1a8710844bda6b84c2a00cf5d86ba78e126d9.

Check Result
uv run --extra test python examples/docs-asset-integrity-smoke.py PASS; 13 guide assets present with distinct PNG hashes
uv run --extra test python examples/docs-governance-smoke.py PASS before and after removal
uv run --extra test python examples/public_entry/readme-demo-surface-smoke.py PASS
uv run --extra test python examples/showcase-catalog-smoke.py PASS
mkdocs build --strict, using docs/requirements-docs.txt and temporary output PASS; production documentation build completed
Repository-local link scan, including Markdown reference definitions and HTML attributes PASS; 6,581 links inspected, none points to a removed path
Retained-file SHA-256 comparison and exact changed-path inventory PASS; 91 retained candidate files unchanged; only the 870 removals and four policy files changed
Ignore rules for new captures and staged/committed RFC bilingual + diff checks PASS; AGENTS.md remains below the 32 KiB discovery budget
Changed-public-file boundary scan with a disposable empty registry/runtime PASS; all four changed public text files scanned

No runtime, backend or browser execution suite was rerun: the executable scenarios, fixture inputs and production sources are unchanged. The checks above qualify asset consumption, docs distribution, historical access and the retention rule; they do not claim installed-product or live Goal health qualification. The adjacent simplification removes capture-pack metadata together with its images, avoiding empty historical indexes or a second archive tree.

Frontend / Visual Evidence

  • UI impact: none. Current product, README, design-reference and guide assets retain their original bytes and paths.
  • First viewport: unchanged; the policy section is added below existing documentation content. No generated screenshots are committed for this maintenance change.
  • Source data: existing public/synthetic material only. Historical captures remain available through the immutable Git source linked in the policy.

Signed-off-by: huangruiteng <huangrt01@163.com>

@loopx-agent loopx-agent left a comment •

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

APPROVE:未发现阻塞问题。已独立核完整 874 路径,清理保留了当前消费者、可复现源与原始历史证据。

Reviewer: model_agent; gpt-6.1-sol; OpenAI; runtime_reported; reasoning_effort=max.

动机 / Motivation

维护文档和提交界面证据的贡献者,会在日常检出中反复携带已经完成审阅的截图包。 以前一次审阅的截图和清单持续留在当前目录;现在撤下没有当前消费者的历史包,当前指南仍使用原图,旧结果从固定版本读取。 独立核验删除 870 项、77,476,304 字节;7,060 个未修改路径均保持原 blob,另有四份政策文本变化,521 项保留媒体不变。 本 PR 不改产品、测试输入或执行语义,不重写 Git 历史,也不声称减少已有历史对象或证明安装采用。

改动思路 / Approach

复用现有文档布局政策与 PR 证据入口,删除历史输出而保留测试源,避免新增第二份归档和留存配置。 本次边界是当前树的生成媒体清理及留存说明;历史原始证据仍绑定原提交,当前文档与测试消费者保持。 不处理会继续累积单次产物;把相同文件搬入另一个 archive 也不会减少当前树。相反,直接删除所有媒体会损坏指南。本次以真实消费者为边界,README/清单的间接引用也要检查,静态无引用不自动等于可以删除。首屏、前后对照、精确版本及数据/构建边界仍由原 PR 证据承担,没有新增确认轮次。

具体改动 / Changes

关键内容讲解

  • 留存规则区分可复现输入、当前指南图片、固定视觉 fixture、单次产物及发布媒体。英文和中文都保留测试不能按体积删除、历史不能被新截图替换,以及重写历史须另行决策的限制。
  • AGENTS 入口与 PR 模板复用同一政策;root 文件实测 32,149 字节,仍低于 32 KiB。两处旧浏览器输出目录加入忽略规则,既有 output/playwright 不变。
  • 836 PNG、10 JPG、21 README/说明及 3 capture.json 从当前树删除;三个 JSON 都是运行记录,测试场景、合成输入和产品代码没有变化。保留的 521 项媒体逐 blob 相同。历史 870 项的原始 blob 均能从固定版本读取。

规范依据:docs/development/documentation-layout.md,预改动版本 bcb1a8710844bda6b84c2a00cf5d86ba78e126d9。逐项映射:Preserve unique information. → 当前内容与固定历史均可达;Repair every repository caller. → 6,542 个跟踪文本及 6,791 个链接候选没有被删目标;keep source-backed artifacts inspectable; → 原始 Git 对象全部保留。新增留存段落作为本次明确政策扩展接受,未以新文本倒推旧要求已经满足。

对主干的风险

主要风险是误删间接消费者。独立扫描又检查了约 106 万个引号字面量及 101 个完全移除目录;八个匹配逐项确认是 mode/source 标识或两处写入 outputDir 的截图输出,没有 fixture 读取。精确树只改变声明的 870 删除项和四份政策,其余源文件原样保留。原生资源检查(13 项)、文档治理、README 展示、showcase 及严格 MkDocs 构建通过;标准 premerge 四项直接检查、十九项选择检查通过,零人工 hold。

GitHub 差分端因 300 文件上限两次返回 HTTP 406,未算作通过;本审阅从已核远端精确 ref 取得完整 Git 差异,并用文件 API 分页核对范围。MkDocs 的既有 INFO 级锚点提示保留,未声称全部旧锚点已修。外部只绑定移动分支的旧图片地址无法穷举,固定原始版本是历史查验入口;这不承诺缩小既有 Git 历史。

我的整体评价

该提案完成了有独立价值的仓库维护切片:当前检出少携带 77.5 MB 单次产物,而当前文档、测试和历史查验均保留。未发现需要本 PR 增加运行时机制或另起归档的理由。现有界面与当前插图未改,因此本次没有新增浏览器或安装视觉资格;历史截图也未被重新认证。执行、安装、模型效果和受保护合并仍各自判断。

English verdict: APPROVE. The exact-head cleanup preserves current consumers, reproducible sources and immutable historical evidence; no blocking finding was found. It makes no runtime, installed-product or Git-history size claim.

@huangruiteng
huangruiteng merged commit 5904d93 into main Oct 29, 2026
10 of 32 checks passed
@huangruiteng
huangruiteng deleted the codex/repository-evidence-retention-1029 branch October 29, 2026 09:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants