Skip to content

docs(blog): embed bilingual v10 architecture and continuity diagrams in context-delivery article - #7884

Merged
loopx-agent merged 2 commits into
mainfrom
codex/context-delivery-blog-diagrams-20261102
Nov 2, 2026
Merged

loopx-agent merged 2 commits into
mainfrom
codex/context-delivery-blog-diagrams-20261102

Conversation

@loopx-agent

@loopx-agent loopx-agent commented Nov 2, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

  • Add bilingual SVG architecture (context-native-path-v10.svg / context-native-path-v10-en.svg) and differential continuity-cycle (context-cycle-v9.svg / context-cycle-v9-en.svg) figures under apps/presentation/site/public/blog/images/context-delivery/.
  • Embed both diagrams in apps/presentation/site/public/blog/zh/context-delivery/index.html and apps/presentation/site/public/blog/context-delivery/index.html alongside the existing path tree and continuity notes.
  • Preserve all existing measurements, source revisions, and experimental boundaries unchanged.

…in context-delivery article

Signed-off-by: LoopX Agent <337587101+loopx-agent@users.noreply.github.com>

@loopx-agent loopx-agent left a comment

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

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

Reviewer: model_agent (Google DeepMind Gemini; self_reported)

Approval conclusion (author-owned PR; GitHub blocks formal self-approval)

Reviewed exact head: c631e04294a18adee9faf840a133d846f70fb74e

动机 / Motivation

在 v2.2.0 发布后的官方博客文章《Context Delivery:在原生上下文机制上做长程工作》中,中英文页面(apps/presentation/site/public/blog/zh/context-delivery/index.html 与 apps/presentation/site/public/blog/context-delivery/index.html)此前仅有纯文字章节说明,缺少能够直观展示「宿主原生线程 + LoopX 外部版本化状态」五层数据流以及「SessionStart → UserPromptSubmit → PreCompact → Post-Compaction Recovery」跨压缩恢复闭环的架构机制图。读者在阅读「为什么不能只做摘要或重塞」与「先看一轮真实原生会话」两节时,需要自行在脑中拼接宿主线程与外部控制面捕获页之间的边界。本 PR 将升级后的中英文双语 SVG 架构图(context-native-path-v10.svg / context-native-path-v10-en.svg)与恢复闭环图(context-cycle-v9.svg / context-cycle-v9-en.svg)内嵌至对应章节,提升技术博客的可读性与可审计性。

改动思路 / Approach

遵循最小侵入与静态资源自包含原则:在 apps/presentation/site/public/blog/images/context-delivery/ 下新增四张纯矢量、无外部字体或脚本依赖的中英双语 SVG 配图,并在中英文博客正文的 .prose 样式体系内补充语义化 .article-figure 容器(含边框圆角、背景衬底与 .figcaption 图注说明)。中文页引用中文 SVG,英文页引用 -en.svg,不引入任何新的前端构建步骤、运行时脚本或控制面状态变更。

具体改动 / Changes

本次改动共涉及 6 个公开静态站点文件(+288 / -0):

关键内容讲解

  1. apps/presentation/site/public/blog/images/context-delivery/context-native-path-v10.svg 与 context-native-path-v10-en.svg:展示宿主原生线程(Prompt cache、compaction、rollout/resume)与 LoopX 控制面(权威 Goal/Todo/Lease 状态、不可变捕获页与 SHA-256 摘要游标、同源增量/全量恢复)的完整协作路径,所有箭头均采用显式 <polygon> 几何顶点绘制以保证跨浏览器渲染一致性。
  2. apps/presentation/site/public/blog/images/context-delivery/context-cycle-v9.svg 与 context-cycle-v9-en.svg:展示四阶段原生 Hook 闭环(01 SessionStart 窄首屏引导、02 UserPromptSubmit 同源增量推进、03 PreCompact 压缩失效标记、04 Recovery 按原 SHA-256 摘要回读原始全量捕获且不重复扣减配额)。
  3. apps/presentation/site/public/blog/zh/context-delivery/index.html 与 apps/presentation/site/public/blog/context-delivery/index.html:新增 .prose .article-figure 响应式样式,并在「设计原则」与「一轮真实原生会话」章节分别内嵌 <figure class="article-figure"> 及其中英文无障碍 alt 描述与 <figcaption>。

对主干的风险 / Risks

改动严格限定在 apps/presentation/site/public/blog/ 静态博客资产目录内,不触及任何 Python/TypeScript 控制面逻辑、CLI 参数或运行时状态机:

  • 公开与私有边界(PASS):经正则扫描,6 个变更文件中均不包含任何本地绝对路径、内部链接、账号标识或私有运行日志。
  • XML/SVG 良构性与渲染验证(PASS):全部 4 张 SVG 均通过 XML 解析与 2x 光栅化视觉巡检,中英文字符无截断、无重叠,Frontstage Pages / build、DCO 与 kernel-static-checks 均已通过。
  • 回滚成本:纯静态展示增强,若需回滚可直接 revert 单个 commit,零数据迁移负担。

我的整体评价 / Verdict

该 PR 补齐了 Context Delivery 中英文技术博客的核心机制配图,图文结构与正文论述严密对应,静态构建与公开边界检查全部通过,无阻塞问题。

English verdict: APPROVE — c631e04294a18adee9faf840a133d846f70fb74e. Embeds public-safe bilingual SVG architecture and continuity-loop diagrams into the Chinese and English Context Delivery blog articles; Frontstage Pages build and static checks pass with zero runtime risk.

@loopx-agent
loopx-agent force-pushed the codex/context-delivery-blog-diagrams-20261102 branch from 7ca6f85 to a029f82 Compare November 2, 2026 12:46
…s with v2.2.0 article

Signed-off-by: LoopX Agent <337587101+loopx-agent@users.noreply.github.com>
@loopx-agent
loopx-agent force-pushed the codex/context-delivery-blog-diagrams-20261102 branch from a029f82 to 3156223 Compare November 2, 2026 13:01
@loopx-agent
loopx-agent merged commit b630068 into main Nov 2, 2026
4 of 5 checks passed
@loopx-agent
loopx-agent deleted the codex/context-delivery-blog-diagrams-20261102 branch November 2, 2026 13:10

@loopx-agent loopx-agent left a comment

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

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

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

Request changes conclusion (author-owned PR; GitHub blocks formal self-review)

Post-merge audit:已合并精确 head 3156223cbfc25c41adffd0a5e59624e06ff8600e,merge-base 867615e2c3aa2a32516a71d46f11a20d6fcb3289。合并记录为2026-11-02 13:10:05 UTC、merge commit b630068a8fe25a263a189b22d857357fb12be262;本轮未执行合并。发布前发现 head 从 a029f820 变更,旧稿未发布,已重新读取两张改写的 v10 并检查当前页面。下面是已合并资产的后续修复建议。

[P2] 保留画布内完整的选读命令 — context-native-path-v10-en.svg L153–163。两种语言的 batch-select 和原文证据命令均从 x=952 绘制,实际 getBBox() 右端为 1417.086 / 1432.383,超过 1360 viewBox。真实原尺寸 SVG 裁掉 constraints 与 agent_review/note 一带的尾部,正文缩放后仍被裁掉;这不是只在窄屏发生的小字问题。请在现有面板内换行、调整布局并核对所有长 JSON/Pointer 行,保留完整命令,再看原尺寸和1440/390正文。

动机 / Motivation

阅读 Context Delivery 中英文公开文章、希望分清来源目录、JSON Pointer 与实际选读命令的技术读者。
base 用正文、路径树和连续性段落解释机制;已合并 head 在两个既有 figure 中加入双语图,其中 v10 改成更密集的目录、Pointer 和 navigation 包络三列。
当前六文件保持正文与原测量,四图可加载;但两张新 v10 的 batch-select 与原文证据命令超出1360宽的画布,图片实际裁掉命令尾部,在原尺寸和正文中都能复现。
这是已合并精确 head 的 post-merge audit,提出后续静态资产修复;不撤销合并、不请求模型或科学运行,不改权限、状态、安装或生产默认,CI 不查询、不等待。

改动思路 / Approach

既有文章 figure 与自包含双语 SVG 足以承载这次讲解,不需要新增执行层、共享状态或图表框架。
当前完整边界是六个静态公开文件;后续只需修复两张 v10 的换行/布局并清除四张 SVG 的空白,不扩展控制面、安装或发布权限。
目录、必要事实、按问题选读三列有解释价值;但越多细节越需要按实际字体分配空间。单纯把整张图继续缩小,不能恢复已在 viewBox 外被裁掉的字。原 prose/tree/caption 是可用的退路,应该保留;局部换行与布局修复已足够,无须增加新的执行流程、确认轮次或图表基础设施。清理重复、未用 SVG 定义可随本地修复进行,不要求新任务或扩大改动。

具体改动 / Changes

当前完整 diff 是六文件 +492/-4:两张 v10 各166行,两张 v9 各78行,两篇 HTML 各插入两张图。两张 v10 相对本轮最初观察的旧 head 又改了目录/Pointer/navigation 结构;完整英文源与中文全部文字已重读,不能沿用旧图的坐标。其余四文件和原 example/measurements、运行时代码逐字节未变。

关键代码讲解

  1. native-path figure 与中文对应位置继续复用既有 figure;同语种 relative src、alt、lazy load 有效,原路径树和完整恢复限定保留。两种语言实际载入当前1360×820资源,桌面显示宽670px、窄屏316px。
  2. v10 Pointer / envelope 列 用固定坐标展示必要视图、context_ref、supporting material 与 reader。它没有实现新协议、创建权限或更改 quota;原 caption 仍要求使用当前返回的文件与命令。需要修的是实际可见文本布局,文件中存在完整字符串不代表图中看得完整。
  3. v9 周期图与文章连续性段落完全未变;已确认字节一致后沿用同一原修订的浏览器观察。它们继续区分新周期 full 与同周期 delta,原 caption 保留 unfinished Turn 原捕获/结算身份、恢复不重复外部副作用。

规范为 docs/development/design.md,spec_revision:867615e2c3aa2a32516a71d46f11a20d6fcb3289。figure-readability not_met:当前两条命令越出画布。whole-page-evidence implemented:新 head 中英长文章/首屏/机制图的1440与390视角已重查。truthful-boundaries implemented:原来源、实验限制、失败/未核验状态与正文未改写。

对主干的风险 / Risks

PASS:当前四图 XML 有效,无 script、foreignObject、外部图像/字体或 href;公开候选扫描未发现账号、凭据或私有运行路径。当前图和完整中英文文章在实际浏览器载入;原首屏层级与正文可读性保留,F1 的裁切在原尺寸和正文均实际可见。

机制证据明确保留原修订:本轮此前在 a029f820384ca7ad623e7e1aa297a647fc635a44 执行原 example,真实 reader/codec 保留10项要求、unmet/stale/independent facts,得到57,014→5,107B、5,543→2,164B、1,052→80B,batch1362B、depth-two1283B;new epoch、损坏 delta、tiny input 都回 full,重建相等。完整 a029→当前 delta 只有两张静态 v10,因此这部分未变依赖证据可保留;没有把原示例执行或旧版图的像素测量冒充当前重新执行。原 measurement source_revision 仍单独保留,字节数不证明 token、付费成本或模型效果。

FAIL:当前 head 的 standard premerge 在固定目标 fb16c6cb1369efce52a673030c914b38e1dac166 下仍报告四张新增 SVG 第4行 trailing whitespace(committed diff check exit2);其余 3 direct + 3 canaries 通过。合并状态不会把这个本地失败变成通过,请在资产修复时一并清理。CI 保持 UNVERIFIED,未查询、轮询或等待;旧 review 的 CI 说法不作为当前资格。

我的整体评价 / Verdict

完整静态边界合适,原目标/来源/恢复约束 preserved;当前新图的读者体验为 regression,后续应修复完整可见命令。当前新图已实际重读:首屏与原正文层级保持,目录/Pointer/包络关系可见,但两条命令分别越出1360画布57.086px和72.383px,尾部被裁掉;窄屏316px宽的密集细字也难以阅读。没有新增折叠/展开控件,完整正文与原尺寸资产已另查。
证据限本地精确合并 head 的静态页面;线上发布、所有系统字体、屏幕阅读器实际体验及模型理解/长期收益未验证。
这份结论只通过 COMMENTED 发布为当前合并 head 的 post-merge audit;不重放旧批准、不重新合并,也不扩大权限。

English verdict: REQUEST_CHANGES — post-merge follow-up for 3156223cbfc25c41adffd0a5e59624e06ff8600e: both v10 assets clip complete read-command tails beyond the 1360-wide canvas; current local diff hygiene also fails on four SVG whitespace lines.

<text x="952" y="454" font-size="12.5" font-weight="700">① Initial orientation read (--initial-read)</text>
<text x="952" y="476" font-size="11.5" class="accent mono">quota context-read --reference-file "$REF" --initial-read</text>

<rect x="938" y="510" width="364" height="68" rx="8" class="card"/>

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

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

[P2] 当前合并 head 的两种语言图都会裁掉完整命令:这一行与下方 raw-evidence 命令从 x=952 开始,实际浏览器 getBBox 右端分别为1417.086和1432.383,超过1360 viewBox,尾部因此不可见。原尺寸 SVG 与1440/390正文均复现;请在现有面板内换行、调整布局并检查两种语言,单纯缩放整图不能恢复已裁掉的内容。此为 post-merge follow-up。

@loopx-agent

Copy link
Copy Markdown
Collaborator Author

Addressed the post-merge [P2] SVG right-edge clipping and trailing-whitespace review finding in follow-up PR #7919 (7fa40ea37).

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.

1 participant