[flashduty]

Doc Review — 2026-10-19

Mode: diff — since 2026-10-16T08:12:21Z (previous successful run) · scope all · window ends 2026-10-19T08:00Z
Findings: 30 total — 4 high, 20 medium, 6 low · 29 applied, 1 handed off, 1 skipped at analysis · 41 files changed (+1227 / −225), 2 new pages

mint broken-links → success no broken links found (all files parsed).

Changes

Finding	Sev	Type	Doc (zh, en mirrored)	Action
f001	high	missing_coverage	on-call/integration/instant-messaging/lark.mdx	New ## 选择接入方式 + QR / Flashduty-managed app flow, 7-state status table, 升级为自动管理
f002	medium	missing_coverage	on-call/integration/instant-messaging/lark.mdx	New ## 在飞书完成配置 (5-step checklist, self-tick vs manual confirm, permission JSON)
f003	medium	stale_content	on-call/integration/instant-messaging/lark.mdx	Removed 网页配置 / 消息卡片请求网址 / 重定向 URL; callbacks are only card.action.trigger
f004	high	config_drift	on-call/integration/instant-messaging/wecom.mdx	Rewrote self-built app setup: hosted access domain, 2 CNAMEs, cert issuance, WW_verify (6 numbered steps)
f005	medium	config_drift	on-call/integration/instant-messaging/wecom.mdx	Token / EncodingAESKey now generated by Flashduty and shown once (direction reversed)
f006	medium	config_drift	on-call/integration/instant-messaging/wecom.mdx	Rewrote third-party app: QR enterprise binding (install / confirm), Corp ID read-only
f007	medium	stale_content	on-call/integration/instant-messaging/microsoft-teams.mdx	New 通知位置 section (团队 / 群聊 tabs), Link Now, add / unlink dialogs
f008	medium	missing_coverage	on-call/integration/webhooks/custom-actions.mdx	New ## 编辑已保存的操作 — masked secrets, retype-to-change, destination guard
f009	high	missing_coverage	on-call/configuration/templates.mdx	New ### 发送测试 — channel matrix, 5 / min, [测试] marker, result + 9 error codes, POST /template/preview/send
f010	medium	missing_coverage	on-call/configuration/templates.mdx	New ### 表格转原生表格 — switch + native table limits + 6 fallback reasons
f011	low	missing_coverage	on-call/configuration/templates.mdx	New ### 预览会提示什么 — format, table counts, byte budget, notice codes
f012	medium	missing_coverage	on-call/configuration/templates.mdx	Read-only state Tip: 系统预置 · 只读, 复制为新模板, 模板设置, 更多操作
f013	medium	missing_coverage	on-call/configuration/custom-forms.mdx	New ## 在 IM 里关闭故障 — guide reply with close=1 link, Feishu ephemeral + toast
f015	high	missing_coverage	NEW rum/analytics/custom-dashboards.mdx	New page (14 sections: entry & routes, panel types, queries, variables, save-to-dashboard, drill-down, versions, trash, permission, limits) + nav entry
f016	medium	feature_drift	rum/explorer/analytics.mdx	New 分组字段的基数限制 (event-ID fields not groupable, count / sum only, ≤ 7 days)
f017	medium	missing_coverage	rum/explorer/analytics.mdx	New 保存到看板 (which panel types can be saved, append vs new dashboard)
f018	low	missing_coverage	rum/explorer/data-query.mdx	New 路径字段与 URL 归一化 (hash routes retained in path / path_group)
f019	medium	stale_content	rum/error-tracking/issue-status.mdx	Resolve modes renamed into an 已解决 submenu; error-viewing.mdx + the 2026-10-16 changelog entry aligned
f020	medium	missing_coverage	platform/permission-design.mdx	Added 看板管理 / Dashboards Manage row + fixed the permission-matrix line
f021	medium	missing_coverage	NEW monitors/engine/self-monitoring.mdx	New page (metric tables, PromQL per group, 7 recommended alerts) + nav entry
f022	medium	missing_coverage	monitors/alert-rules/mysql.mdx	v0.76.0 variable support rewritten as a save / enable gate matching the Prometheus page
f023	low	missing_coverage	monitors/quickstart/quickstart.mdx	Added 8 KiB rendered-value and 255-char render-failure limits
f024	medium	missing_coverage	ai-sre/sessions.mdx	Compaction originals retrievable via session_transcript + .history/ layout
f025	low	missing_coverage	ai-sre/skills.mdx	Added session_transcript to the allowed-tools tool enumeration
f026	medium	missing_coverage	developer/cli.mdx	New integration subsection: ****** masking, retype-to-change, destination guard
f027	low	missing_coverage	ai-sre/im.mdx	IM acknowledgement on async dispatch made concrete
f028	medium	feature_drift	on-call/statuspage/create-manage-page.mdx	Custom-domain uniqueness now covers account-held WeCom domains + release semantics
f029	medium	missing_coverage	changelog/changelog.mdx	Added <Update label="2026-10-19"> with 5 feature subsections
f030	low	stale_content	changelog/changelog.mdx	Fixed 「弹窗」 → 「就地展开的搜索框」 in the 2026-10-14 entry

New pages are registered in docs.json for both zh and en (rum/analytics/custom-dashboards after miniprogram; monitors/engine/self-monitoring after engine-lost-alert). No path was renamed or moved, so no redirect is needed.

Review checklist

• Config field names, defaults and limits match current code — every value in this PR was read from origin/main of the source repo, with the file quoted in the finding (see .doc-review/findings-2026-10-19-081500.html for per-finding evidence)
• WeCom / Lark pages keep the manual path and the new QR path distinguishable; the on-page step numbers the console references (「编号和右侧文档一致」) were kept 1–6
• Template test-send limits: per-cell limit is 2000 bytes (len(value) > FeishuTableMaxCellSize in fc-event logic/template/markdown_table.go:212), not characters — changelog and page now agree
• Tone and structure match neighbouring pages (Lark follows the rewritten DingTalk page)
• Two pre-existing drifts observed but not touched (out of window): zh/on-call/integration/webhooks/custom-actions.mdx lacks the TLS-verification steps that en/ has; zh/monitors/dashboards/dashboards.mdx says 「控制台暂无回收站页面」 which is true for Monit but not for RUM

Cross-repo follow-ups

api-review (generated OpenAPI surface — do not hand-edit):

• f014: POST /template/preview/send is missing from api-reference/on-call.openapi.{zh,en}.json and therefore from openapi/api-catalog.mdx. Route: fc-event cmd/server/routes.go:336 (tmplG.POST("preview/send", template.PreviewSend)), handler cmd/server/controller/template/preview_send.go. The catalog page renders from the spec, so a hand-written row would be a broken link.
• POST /rum/field/list responses now carry a high_cardinality property (used by f016) — same generated-spec surface, api-reference/rum.openapi.{zh,en}.json.

flashduty-website:

• Website changelog (content/changelog/) has no entry newer than 2026-10-05, while the docs changelog now covers 10-12, 10-14, 10-15, 10-16 and 10-19. This window's features (RUM custom dashboards, template send-test + native tables, WeCom hosted access domain, Lark managed app) therefore have no website entry. Owner should decide whether to backfill.
• Integrations catalog: no new integration shipped this window (WeCom / Feishu / Teams setup flows changed, the integrations themselves already existed), so the "60+" count claim is unaffected.

Mapping warnings

14 warnings, all from mapping.yaml going stale against the repo layout. The load-bearing ones:

• fc-foundation-app src/Packages/{saas,monit,rum,ai-sre} are git submodules that are empty in a plain clone, so every frontend path in the mapping resolves to nothing. Repos must be mapped to the standalone clones: fc-saas-web → pages, fc-saas-monit → pages, fc-saas-rum → pages, ai-sre-web → src (repo root is the package — drop the src/Packages/<pkg>/ prefix). This round was run that way and found 118 / 60 / 59 changed frontend files that the mapping would have hidden.
• fc-datasource is mapped only to model/{dingtalk,slack,teams,wecom} + structs/plugin.go, i.e. vendor structs; all IM setup / authorize / bind logic lives in cmd/datasource/controller and logic/data_source and had never been audited. deploy/migrations is unmapped too.
• fc-event cmd/engine/controller/notify is unmapped — it is exactly where the notification-template behaviour of f009/f010 lives.
• pages/Template (notification templates) has no module entry at all, although zh/on-call/configuration/templates.mdx is generated from it.
• fc-safari: logic/tool/builtin, logic/agent, types/prompts unmapped — this window's whole AI SRE user-visible delta (builtin session_transcript, compaction recall note) landed there.
• fc-oncall: cron, queue, naming unmapped — duty / scheduled-notification timing lives there.
• fc-statuspage: only logic/ mapped; the customer-facing domain serving path is cmd/server/controller/page/ + deploy/k8s/.

Skipped

• platform-aisre-2 — private-deployment appSettings.login.defaultType='ldap' / login.ldapAccountId: delivery configuration injected by the operator, not a console setting. Every other private-deployment KV (enableOncall, enableRum, rumIngestEndpoint, …) is likewise undocumented, so this is treated as a deliberate boundary. If product wants it public, it belongs in platform/configure-sso.mdx; evidence: fc-foundation-app src/pages/LoginPage/const.ts:63-70, Situations/SSOLogin/index.tsx:109-136.
• f014 — handed off to api-review (above), not applied here.
• Rejected during analysis (not findings): fc-oncall's cron / Kafka ack / leader-election changes (no documented claim contradicted); monit-frontend's dashboard 回收站 page (host-gated: MONIT_HOST.api does not wire listTrash / restoreDashboard, so Monit really has no trash page — only RUM does); panel_config_invalid panel state (no backend emits that reason yet); ConditionEditDialog 「多个名称请使用英文逗号分隔」 copy, which no backend splits on (possible product bug, not a docs gap); meow-ui design-system churn; fc-pgy (test files only).