Skip to content

[PR] Blog page: per-article GitHub discussion button (exact link from recorded data) + EN/中文 pin - #5396

Merged
ranxianglei merged 5 commits into
masterfrom
2026-11-01_blog-comment-link
Nov 1, 2026
Merged

ranxianglei merged 5 commits into
masterfrom
2026-11-01_blog-comment-link

Conversation

@ranxianglei

@ranxianglei ranxianglei commented Nov 1, 2026 •

Copy link
Copy Markdown
Owner

What

Blog detail page header now carries two small links next to "back":

  1. Comment button → exact discussion thread. The link is data, not inference: the blog pipeline workflow creates one GitHub Discussion per post and records its number in website/articles.json; the feed passes it through as item.discussion and the client links straight to github.com/.../discussions/<n>. No recorded number = no button. The earlier "search all discussions by title token" version is gone — no guessed URLs, no dead threads.
  2. EN | 中文 pins. Plain ?lang=en / ?lang=zh links on the article itself; the language pin already outranks the stored UI choice (dsh设置位置提供真的设置面板 #2321), so clicking them re-renders the same article in the other language with zero new state code.

Behavior contract (pinned in tests/web-blog-route.test.ts)

  • item with discussion: 123 → anchor to …/discussions/123, new tab + noopener.
  • item without discussion → article renders, no discussion anchor anywhere.
  • language pins present on every article detail view.

Deps / rollout

Server-side passthrough of the discussion field lands in the blog fetch-fallbacks PR; the numbers appear in articles.json only after the pipeline workflow runs for a post — until then the button is simply hidden (old posts unaffected). No storage, no config knobs.

@github-actions

github-actions Bot commented Nov 1, 2026 •

Copy link
Copy Markdown
Contributor

📦 Built Package Artifact

Branch: 2026-11-01_blog-comment-link (b78195b)

Latest build — install this exact version

npm install -g billion-context@0.0.7757-pr.5396.fb7f1c4

This exact version is immutable and reproducible. The rolling pr tag tracks the newest test build across all PRs.

Recent builds (commit → version)

commit version build
b78195b billion-context@0.0.7757-pr.5396.fb7f1c4 build
eac5297 billion-context@0.0.7674-pr.5396.df8f7a6 build
6f5b484 billion-context@0.0.7666-pr.5396.9421b24 build
f001dc2 billion-context@0.0.7661-pr.5396.24073a2 build
c7beb52 billion-context@0.0.7635-pr.5396.9d0b302 build
cc22c63 billion-context@0.0.7634-pr.5396.94cf55a build
e93e52e billion-context@0.0.7628-pr.5396.4e70214 build

Download artifact

  1. Download the artifact from the Actions run
  2. Extract the tarball and install:
tar xzf billion-context-pr5396.tgz
npm install -g package

Install instructions are refreshed on each push.

@ranxianglei ranxianglei changed the title Blog page: link each article to its GitHub comment thread [PR] Blog page: link each article to its GitHub comment thread Nov 1, 2026
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

查重: 不重复。 #5398 是「给博客文章加『去 GitHub 评论』链接」这个 UI 改动本身;相关但范围不同的条目:#5394 是单篇文章的评论区内容 issue(标题就是 blog:frequent-compression),#4906 是博客自动拉取功能(已关)。三者不重叠。

标签: scope/ui · p3 —— web 面板展示面(UI 功能);残留一个边缘项(旧文章缺评论区 → 空搜索),按 p3 边缘场景记。

我核对了分支 2026-11-01_blog-comment-link(e93e52e)的代码与 CI,没有重做,结论如下。代码干净、CI 绿(summary: total 30 / failed 0 / pending 2——快照时还有 ubuntu 22/2、ubuntu 24/2 两个 test shard 在跑,其余全绿,含 static-gates / gate / version-guard / lock-sync / docs-sync;mergeable_state: blocked 是这俩 pending + 待 review 造成,不是测试红)。两处小发现都不挡合并:

发现 1 — commit message / PR 正文引用了错的 issue(文档/归属错误)
正文和 commit 都写「see issue #5393」作为 blog:<id> 约定的出处,但 #5393 实际是「Claude 经 Antigravity(Google v1internal)part 顺序的内核决策申请」(#5392 的 kernel-change),跟博客完全无关。真正的示例是 #5394(标题 blog:frequent-compression — Comment Section)。建议把引用改成 #5394,或干脆只描述 blog:<post-id> 这个 token、不挂具体编号,免得后人被误导。不影响功能;改 PR 正文即可,不用动代码。

发现 2 — 旧文章没有对应评论区 → 链接打开空搜索(边缘 UX)
现在两篇内置文章里只有 frequent-compression 有评论区(#5394);更早的 meet-bili(2026-10-08)没有 blog:meet-bili 这个 issue。所以点 meet-bili 详情页的「去 GitHub 评论」会打开一个空结果的 GitHub 搜索页——不报错,只是没内容。两个选择:(a) 接受兜底(搜索式链接本来就容忍无结果);(b) 给 meet-bili 也建一个 blog:meet-bili 评论区 issue 补齐。建议 (b),顺手补一篇就行;要的话我来建。


八问

  1. 是什么问题:给博客文章详情页加第二个头部链接「去 GitHub 评论」,点开该文章对应的 GitHub 讨论区(issue 列表按标题里的 blog:<post-id> token 过滤)。
  2. 什么原因导致的:纯新增功能,不是修 bug——之前详情页只有「← 全部文章」一个返回链接,没有通往讨论区的入口。
  3. 影响范围与严重程度:web 面板博客详情页;纯增量、不改任何既有行为。当前唯一边角是旧文章 meet-bili 缺评论区 → 空搜索(p3,见发现 2)。
  4. 怎么修复的:一个 URL helper(blogIssueUrl(id) 拼 issues?q=is:issue in:title "<slug>")+ 详情页模板多拼一个 <a target="_blank" rel="noopener"> + 两条 i18n(zh/en)。用「按标题 token 搜索」而非硬编码 issue 号,天然容错。
  5. 是否有回归问题:无。只在一段字符串拼接处追加一个 anchor,复用既有 bl-back 样式;i18n 两个 locale 都补齐(docs-sync 已过);item.id 是真实字段(blogfeed.ts 中 id=slug,路由也按它匹配);URL 经 encodeURIComponent 且放在双引号属性内,无注入/转义风险。
  6. 是否破坏缓存:不涉及(纯前端渲染,无缓存/游标/去重键)。
  7. 是否改了设计方向:不涉及(客户端内增量,无 server 变更,与 PR 描述一致)。
  8. 是否引入新配置:无(无新 env/config/CLI/key)。

判定:可合并。代码与 CI 都过关;两处发现都是非阻塞小项——发现 1 建议顺手把 PR 正文的 #5393 改成 #5394,发现 2 建议给 meet-bili 补个评论区 issue(或接受空搜索兜底)。要我做这两件小事就说一声。

@ranxianglei ranxianglei left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

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

(ai: vllm-qwen/qwen3.8-27b)

Verdict: APPROVE (posted as a review comment — GitHub forbids approving one's own PR) — after two small direct fixes, pushed as labeled commit c7beb5279 on the PR branch.

Verified independently:

  • Direction. The convention issue (GH #5393, "blog:frequent-compression — Comment Section") states in its OP that every post's thread issue carries blog:<post-id> in the title; the button opens a repo-scoped issue search on the post id. Live-verified the query shape via the GitHub search API: is:issue in:title "frequent-compression" → exactly 1 hit (the thread issue itself).
  • Diff scope (at e93e52e). src/web/client.ts (+7/-1) and src/web/i18n.ts (+2) only. No server change, no kernel/, no config surface, no persistence/wire/cache impact.
  • Security. item.id passes through encodeURIComponent before reaching the href (attribute-breakout safe); target="_blank" + rel="noopener" present.
  • Regression inventory of the touched surface. Back link, date/badge row, body render and the not-found branch are all unchanged — the change is purely additive. The i18n catalog-parity test pins bl.comment in both locales; web-blog-route runs the real WEB_CLIENT in a vm sandbox.
  • Tests. The new anchor had no pin — added one asserting the exact GitHub search URL plus target/rel in tests/web-blog-route.test.ts. Targeted local run on the final head: web-blog-route + web-i18n + web-blogfeed = 27/27; the full-suite verdict is CI's job on the new head.

Fixed directly (commit c7beb5279):

  1. src/web/client.ts:993 — the two header links rendered glued together (</a><a> with no whitespace; .bl-back has no margin/padding). Added a space, matching the existing pattern in the not-found branch (:987).
  2. tests/web-blog-route.test.ts — regression pin for the new anchor (see above).

Non-blocking note: the meet-bili post predates the convention and has no thread issue, so its button opens an empty search list — graceful by design of the search-based lookup.

Numbering note: references above follow GitHub numbering — this PR is GH #5396 and the convention issue is GH #5393 (the PR body's "#5393" citation is correct).

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

查重: 不重复——无其他在途 PR 触碰博客详情页头部或 i18n bl.* 键,本 PR 是「去 GitHub 评论」入口的唯一实现。

审核完成,结论:APPROVE / 可合(等新 head c7beb52 的 CI 转绿即可合并)。GitHub 侧审核已发:#5396 (review) 禁止自己 approve 自己的 PR,故以 review comment 形式发布完整结论)。

八问:

  1. 是什么问题:功能请求——面板博客详情页加一个「去 GitHub 评论」按钮,跳到该文章的评论区 issue(用 issue 标题里的 blog:<文章id> token 定位)。
  2. 什么原因导致的:非缺陷。owner 在 GH#5393(OP 即评论区)确立了「之后每篇新文章评论区都用 blog:<文章id> 格式建帖」的约定,本 PR 实现面板侧入口。
  3. 影响范围与严重程度:纯面板 UI 层新增——每篇博客详情页多一个外链;不涉及压缩、wire、存储、配置面。无既有用户受影响(功能新增)。
  4. 怎么修复的:纯客户端——一个 URL helper(repo-scoped GitHub issue 搜索,id 经 encodeURIComponent)+ 一个锚点(target=_blank rel=noopener)+ 两条 i18n 字符串;服务端零改动。查询形状已用 GitHub 搜索 API 实测:in:title "frequent-compression" 恰好命中 1 条 = 该文章评论区本身。
  5. 是否有回归问题:审了触碰路径的既有行为——返回链接、日期/badge 行、正文渲染、not-found 分支全部不变,纯增量;i18n catalog-parity 测试钉住双语言键。审出两处小问题已直接修掉并 push(labeled commit c7beb52):① 头部两个链接之间没空格且 .bl-back 无 margin,渲染会贴在一起(← 全部文章去 GitHub 评论)→ 补空格(与 not-found 分支既有写法一致);② 新锚点此前没有任何测试钉住 → 在 web-blog-route 的 vm 沙盒测试里补断言,钉住精确 URL + target/rel。本地定向跑 27/27(web-blog-route/web-i18n/web-blogfeed),全量以 CI 为准。
  6. 是否破坏缓存:不涉及(纯前端渲染层,不碰前缀缓存/会话缓存)。
  7. 是否改了设计方向:没有。
  8. 是否引入新的配置:无。

两条备注:

判定:可合。新 head c7beb52 CI 转绿后 owner 自行合并(合并权在 owner)。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

复审确认(第二轮唤醒):head c7beb52 的 CI 已全绿——47 项 check 全部 success/skipped、0 失败(因镜像缓存条目返回了错误 sha 0a84db6 且 0 条 check,本次按「镜像无有效缓存」例外单次直连核实;该缓存异常报给运维留意)。PR 分支自上一轮审核以来无新提交、无新人类评论。结论不变:APPROVE / 可合,审核全文见上一轮回复与 GitHub review(PR #5396)。合并权在 owner,GitHub 侧 mergeable_state=blocked 即等人工合并(同一账号不能 approve 自己的 PR)。

@ranxianglei ranxianglei left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

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

Verdict: REQUEST_CHANGES (GitHub forbids the formal event on own PRs — this comment carries the full review)

(ai: vllm-qwen/qwen3.8-27b)

Round-3 review of new head f001dc239 (delta since c7beb5279: the button target pivoted from issue search to Discussions search, citing a "#5411 convention"). The direction — moving comment sections from issues to Discussions — is your call and I don't contest it; but as of this head the change ships broken. Three findings, the first two blocking:

F1 (blocking) — the "#5411 convention" citation is dead (src/web/client.ts:959, commit message). Verified against GitHub: issue/PR #5411 -> 404, discussion #5411 -> 404. (The mirror-side #5411 is an unrelated auto-update/rotation decision request.) The convention's actual origin is GH#5393, which is now closed. A code comment must cite something that exists: create the convention-establishing discussion first and cite its real number, or drop the number and state the convention inline.

F2 (blocking) — every button now links to an empty page. Verified: the repo has exactly 10 discussions total and none carries a blog:<id> token — no per-post discussion threads exist yet. After merge, 「去 GitHub 评论」opens zero results for ALL posts. Notably, frequent-compression's real thread is issue #5393 (now closed, but still surfaced by the OLD issue-search URL, which had no is:open filter) — so this pivot takes the one working link and makes it dead. This is a behavior drift that needs disclosure in the PR body: old = resolves to the real thread where one exists -> new = always empty until threads are created. Acceptable resolutions: (a) create the discussion threads before merge (port frequent-compression's content out of #5393) — I'll re-verify the search hits then; or (b) explicitly accept the interim dead-link state in the PR body. Side note: is:open in the query also means a later-closed thread silently empties its button — deliberate?

F3 (doc) — PR body is stale. It still says "It opens the search-filtered issue list ... (see issue #5393)" — that describes the pre-f001dc239 behavior. Refresh it to describe the Discussions convention once settled.

Non-blocking checks all pass: test pin matches the helper output byte-for-byte incl. is:open (tests/web-blog-route.test.ts:118); encodeURIComponent + rel="noopener" intact; i18n parity unaffected; no server/config/kernel impact; CI green on this head (47 runs: 35 success / 12 skipped / 0 failed — verified directly because the mirror cache entry again carried a wrong sha, same anomaly as last round).

Please resolve F1-F2 (and refresh the body per F3); I'll re-review immediately after.

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

第三轮审核结论(新 head f001dc239):判定由 APPROVE 改为 REQUEST_CHANGES。owner 推了新提交,把按钮目标从「issue 搜索」改为「repo Discussions 搜索」(引用 "#5411 convention"),独立核实后发现两个阻塞问题。GitHub 侧审核已发(PR #5396,GitHub 禁止对自己的 PR 发正式 REQUEST_CHANGES 事件,故以带判定行的 review comment 发布):#5396 (review)

八问:

  1. 是什么问题:本轮审新增提交 f001dc239——「去 GitHub 评论」按钮从链到 issue 搜索改为链到 repo Discussions 搜索(discussions_q=is:open in:title "<id>")。
  2. 什么原因导致的(发现):引用的 "#5411 convention" 在任何编号体系下都不存在——GH issue/PR #5411 → 404、GH Discussion #5411 → 404、镜像侧 #5411 是不相关的自更新决策请求;且 repo 目前总共只有 10 个 discussion,没有一个带 blog:<id> token——按钮目标全部是空页。
  3. 影响范围与严重程度:合并后每篇博客详情页的按钮都会打开空搜索结果页,包括 frequent-compression(它的真实评论区是 issue blog:frequent-compression — Comment Section | blog:frequent-compression — 评论区 #5393,现已关闭,但旧 URL 无 is:open 过滤仍能找到它)。严重度:UI 功能整体不可用(功能本身 p3,但等于上线死链)。
  4. 怎么修复的(建议):两条路——(a) 先建好对应的 Discussion 评论区(至少把 frequent-compression 的内容从已关的 blog:frequent-compression — Comment Section | blog:frequent-compression — 评论区 #5393 迁过来),引用改成真实存在的编号,合入前我复验搜索命中;(b) 在 PR body 里明确接受过渡期死链状态。无论哪条:src/web/client.ts:959 的引用必须指向真实条目,PR body(仍写着 "search-filtered issue list ... see issue blog:frequent-compression — Comment Section | blog:frequent-compression — 评论区 #5393")必须同步更新为 Discussions 约定。
  5. 是否有回归问题:有行为漂移(旧值→新值):旧 = 有评论区的文章链接解析到真实评论区(第一轮实测 frequent-compression 恰好 1 条命中)→ 新 = 100% 空结果。属 owner 有意方向变更,按纪律需在 PR body 披露;另注意查询里的 is:open:将来某评论区若被关闭,其按钮会静默变空——是否有意请确认。
  6. 是否破坏缓存:不涉及(纯前端渲染层)。
  7. 是否改了设计方向:改了——评论区载体从 Issue 转向 Discussion。方向本身不置喙(owner 决策),但当前实现与现实不符(Discussions 尚不存在)。
  8. 是否引入新的配置:无。

其余核查均通过:测试对新 URL 逐字节钉住(含 is:open,tests/web-blog-route.test.ts:118)、encodeURIComponent + rel="noopener" 完好、i18n parity 不受影响、未触碰 kernel(硬性授权规则未触发);CI 本 head 全绿(47 项:35 success / 12 skipped / 0 失败——因镜像缓存条目又返回了错误 sha 73b7e19(无关提交),按例外单次直连核实,同类异常第二次出现,报运维)。

判定:REQUEST_CHANGES,等 owner 解决 F1(死引用)+ F2(评论区不存在或明示接受死链)并更新 PR body,我随到随复审。

@ranxianglei ranxianglei changed the title [PR] Blog page: link each article to its GitHub comment thread Blog page: per-article GitHub discussion button (exact link from recorded data) + EN/中文 pin Nov 1, 2026
@ranxianglei ranxianglei changed the title Blog page: per-article GitHub discussion button (exact link from recorded data) + EN/中文 pin [PR] Blog page: per-article GitHub discussion button (exact link from recorded data) + EN/中文 pin Nov 1, 2026

@ranxianglei ranxianglei left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

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

(ai: vllm-qwen/qwen3.8-27b)

Verdict: APPROVE (GitHub forbids the formal event on own PRs — this comment carries the review)

Round-4 review of head 6f5b48483 (pivot to exact recorded-number links + EN|中文 pins). Round-3 findings disposition: F1 dead citation — resolved (comment rewritten); F2 dead-link — resolved structurally (exact link from data, or no button: honest degradation); F3 stale body — resolved (body now describes the new design and discloses the rollout split). Independent verification of this head:

  • Language pins work as claimed. The pin is parsed once at init (#2321, src/web/client.ts:146-160) and outranks the stored choice; the anchors resolve to a query-differing URL, which is a FULL navigation (only fragment-only diffs are same-document), so the panel reloads with the pin, lands back on the article via the preserved hash, and the router decodes the slug correctly (route() decodeURIComponent at :4886, matching the list-row encoding at :1015). Same reload semantics as the existing #language-toggle handler (:4907). "Zero new state code" holds.
  • Tests have teeth: positive case pins the exact discussions/ URL incl. target=_blank rel=noopener; the new negative case pins "no recorded number → no discussion anchor anywhere"; i18n catalog parity (web-i18n) covers the label. Local targeted run 7/7; CI green on this head (47 runs: 35 success / 12 skipped — direct fetch, mirror cache entry unreliable again this session).
  • Staged delivery accepted. As of this head the comment button renders for NO post (server passthrough deferred to the fetch-fallbacks PR per the body; no numbers recorded yet). That is disclosed, and what the tests pin is the client-layer contract (item.discussion → exact link), so the staging is coherent. Two follow-ups belong in that other work, not here: (a) wire the passthrough for BOTH item kinds — posts come from website/blog/*.md frontmatter, not articles.json, so "recorded in articles.json" only covers kind=article unless the fallback PR also reads post frontmatter; (b) create the actual Discussion threads (frequent-compression first — port it out of closed #5393) so at least one button goes live.

One small defect fixed directly on the branch (labeled commit eac5297, comment-only, zero runtime impact): the helper comment stated as present fact a pipeline that does not exist yet (no such workflow in .github/workflows/, no website/articles.json on master, no feed passthrough in src/web/blogfeed.ts — BlogItem has no discussion field at src/web/blogfeed.ts:63-75). Comment inconsistent with code facts; reworded to state the actual client contract and where the data dependency lives.

No kernel/config/storage/wire surface touched. Behavior additions vs master: two header links on every article detail view (EN|中文 pins active immediately; comment button hidden until recorded data lands) — both disclosed in the PR body.

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

第四轮审核结论(新 head 6f5b48483,「exact link from recorded data + EN|中文 pin」):判定回到 APPROVE / 可合(新 head CI 转绿后)。GitHub 侧审核已发(PR #5396):#5396 (review)

八问:

  1. 是什么问题:本轮审 owner 对第三轮 REQUEST_CHANGES 的重构——按钮改为「从记录数据取精确 Discussion 编号直链,无编号则不显示按钮」,并新增文章页内 EN|中文 语言 pin。
  2. 什么原因导致的(第三轮发现为何消失):搜索式定位被彻底移除,死引用(#5411)随注释重写而消失;死链问题被结构性解决——数据没有就不渲染按钮,诚实降级。PR body 已同步更新并披露分阶段计划(服务端透传在 fetch-fallbacks PR 落地,编号要等 pipeline 建帖后写入)。
  3. 影响范围与严重程度:合并时用户可见变化 = 每篇博客详情页头部多两个 EN|中文 链接(立即生效);评论按钮在数据落地前对所有文章都不显示(与 master 相比无回归,master 本来就没有按钮)。严重度:无缺陷面。
  4. 怎么修复的(设计):客户端契约 = feed item 带 discussion 正整数 → 渲染 …/discussions/<n> 精确链接(target=_blank rel=noopener),否则无按钮;语言 pin 复用 dsh设置位置提供真的设置面板 #2321 的 ?lang= 机制(pin 优先于存储选择),点击是 query 变化的整页导航 → 重载后面板按 pin 渲染、hash 保留落回原文。独立验证了这套机制成立(路由 decodeURIComponent 与列表行编码一致,client.ts:4886/:1015;与既有 #language-toggle 同语义)。测试有牙:正例钉精确 URL、负例钉「无编号→无锚点」、i18n parity 覆盖标签;本地定向 7/7,CI 本 head 全绿(47:35 success/12 skipped,直连核实——镜像缓存本会话持续返回错误 sha,第三次,已报运维)。
  5. 是否有回归问题:审过触碰路径既有行为——返回列表、日期/badge、正文渲染、not-found 分支均不变;i18n catalog 只增键。无回归。一处小缺陷已直接修掉(labeled commit eac52975c,纯注释零运行时影响):client.ts:959 的注释把尚不存在的 pipeline 当既成事实陈述(repo 里无该 workflow、master 无 articles.json、blogfeed.ts 的 BlogItem 根本没有 discussion 字段,:63-75)——注释与代码事实不符,已改写为陈述真实契约与依赖所在。
  6. 是否破坏缓存:不涉及。
  7. 是否改了设计方向:评论区载体定为 GitHub Discussions + 编号走数据管道(非推断)。方向是 owner 决策,本轮不再置喙;留两条后续给 fetch-fallbacks PR:(a) 透传要同时覆盖两种 item——posts 来自 website/blog/*.md frontmatter 而非 articles.json,(b) 实际建帖(frequent-compression 先从已关的 blog:frequent-compression — Comment Section | blog:frequent-compression — 评论区 #5393 迁出)让至少一个按钮活起来。
  8. 是否引入新的配置:无。

判定:APPROVE / 可合。我的注释修正 commit eac52975c 会触发新一轮 CI,转绿后即可由 owner 合并(合并权在 owner)。

dog and others added 5 commits November 1, 2026 23:49
… to the post's issue thread (#5393 convention: blog:<id> in title)
…n the two detail-header links (they rendered glued); pin the new anchor in web-blog-route
… pins

The comment button no longer guesses a repo-wide Discussions search URL: it renders only when the feed carries a recorded thread number (item.discussion, written back by the blog pipeline workflow into articles.json) and links straight to /discussions/<n>. A post without a recorded thread shows no button at all. Detail header also gains plain ?lang= links that double as a per-article EN|中文 switch (the pin outranks the stored UI language, #2321).
…as fact — say what the client contract is and where the data dependency lives (review nit, comment-only)
@ranxianglei
ranxianglei force-pushed the 2026-11-01_blog-comment-link branch from eac5297 to b78195b Compare November 1, 2026 15:49
@ranxianglei
ranxianglei merged commit 0d0c0b9 into master Nov 1, 2026
47 checks passed
@github-actions

github-actions Bot commented Nov 1, 2026

Copy link
Copy Markdown
Contributor

📦 Merged Build Published

This merge to master was published as 0.0.994-master.0d0c0b9 (rolling master tag).

npm install -g billion-context@0.0.994-master.0d0c0b9  # this exact build
npm install -g billion-context@master  # always the newest merged build

npm view billion-context@master prints the current version at any time.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant