docs(analytics): 把 analytics 落地页写实，并清掉散在 8 页上的 cube 词汇残留 (#976, #977)

.changeset/analytics-landing-page-cube-vocabulary.md

@@ -0,0 +1,41 @@
+---
+'hotcrm': patch
+---
+
+Write the analytics landing page to source, and retire the "four built-in cubes"
+vocabulary from the eight pages that still carried it.
+
+`content/docs/analytics/index.mdx` is the first screen a reader of the analytics
+docs sees, and every count on it named something else. It advertised **4 cubes**
+(*Sales*, *Pipeline*, *Service*, *Marketing* — four names that exist nowhere in
+the app; the semantic layer is the nine datasets in `src/datasets/`, which the
+analytics service compiles into cubes internally, ADR-0021), **4 dashboards**
+(there are five — *Sales Activity* was missing), **10+ reports across leads,
+deals, cases, contracts** (there are exactly ten, and none of them is a contract
+report: no dataset reads `crm_contract`, so none can be built), a report called
+*Pipeline by Stage* (a shared dashboard tile and a chart title — the report is
+**Opportunities by Stage**) and an **Analytics** navigation group (the group is
+called **Insights**, and it holds CRM Overview, Forecasts, Pipeline Coverage,
+Lead Inflow and SLA Performance — nothing in it is called *Dashboards*,
+*Reports* or *Cubes*). The "AI Copilot reads directly from cubes" claim now says
+what this app declares — no skill names a dataset, cube or measure, and the one
+that answers data questions aggregates over records — and leaves the platform
+side undecided rather than asserting it in either direction.
+
+The same retired vocabulary is gone from eight further pages: `whats-new` now
+lists the nine datasets it actually shipped, `getting-started/for-developers`
+draws `datasets/` instead of the `src/cubes/` directory removed in #492, and
+`reference/performance-and-limits` plus `reference/faq` state the refresh cadence
+the app really declares — each dashboard's own `refreshInterval`, 60 s on
+Customer Service, 180 s on Sales Performance, 300 s on the other three — in place
+of an "every 5 min incremental / nightly full" figure nothing in `src/`
+configures, and of a manual refresh button no dashboard declares.
+`sales/pipeline-management`, `getting-started/introduction`,
+`marketplace/fork-hotcrm` and `administration/sandbox-and-releases` were renames
+only. The glossary's definition of the *cube* concept is deliberately untouched:
+the concept is real, and only its misapplication to this app was not.
+
+Documentation only in all three locales; no metadata changed.
+`test/docs-analytics-vocabulary.test.ts` now derives every count and product name
+on these pages from `src/` and fails the build when the app and the page disagree.
+Refs #976 #977.

content/docs/administration/sandbox-and-releases.mdx

@@ -90,7 +90,7 @@ A **change package** captures a set of metadata changes (objects, fields, valida
 | Flows | ✅ |
 | Sharing rules, profiles, permission sets | ✅ |
 | Pages, list views, dashboards | ✅ |
-| Reports, cubes | ✅ |
+| Reports, datasets | ✅ |
 | AI skills, knowledge base configs | ✅ |
 | Approval processes | ✅ |
 | Email templates | ✅ |

content/docs/administration/sandbox-and-releases.zh-Hans.mdx

@@ -90,7 +90,7 @@ description: 安全地进行配置变更 —— 沙箱租户、变更包、提
 | 流程 | ✅ |
 | 共享规则、权限档案、权限集 | ✅ |
 | 页面、列表视图、仪表板 | ✅ |
-| 报表、立方体 | ✅ |
+| 报表、dataset | ✅ |
 | AI 技能、知识库配置 | ✅ |
 | 审批流程 | ✅ |
 | 电子邮件模板 | ✅ |

content/docs/administration/sandbox-and-releases.zh-Hant.mdx

@@ -90,7 +90,7 @@ description: 安全地進行設定變更 —— 沙箱租戶、變更包、提
 | 流程 | ✅ |
 | 共用規則、權限設定檔、權限集 | ✅ |
 | 頁面、清單檢視、儀表板 | ✅ |
-| 報表、立方體 | ✅ |
+| 報表、dataset | ✅ |
 | AI 技能、知識庫設定 | ✅ |
 | 審批流程 | ✅ |
 | 電子郵件範本 | ✅ |

content/docs/analytics/index.mdx

@@ -1,6 +1,6 @@
 ---
 title: Analytics
-description: Dashboards, reports, and the data cubes that power them — how HotCRM turns daily activity into insight.
+description: Dashboards, reports, and the semantic layer under them — how HotCRM turns daily activity into insight.
 ---
 
 # Analytics
@@ -9,15 +9,17 @@ HotCRM Analytics turns your CRM activity into **decisions**. Three layers work t
 
 - **Dashboards** — visual at-a-glance views for execs, managers, and reps.
 - **Reports** — detailed list-and-summary reports anyone can filter and export.
-- **Cubes** — the data layer underneath, ready for self-service slice-and-dice.
+- **Cubes** — the semantic layer underneath. In this app that layer is a set of **datasets**: HotCRM declares no cube of its own, and the analytics service compiles each dataset into its cube internally (ADR-0021).
 
 ## What's included
 
 | Layer | What it does |
 | --- | --- |
-| [**Dashboards**](/docs/analytics/dashboards) | 4 ready-made dashboards (CRM, Sales, Service, Executive) |
-| [**Reports**](/docs/analytics/reports) | 10+ standard reports across leads, deals, cases, contracts |
-| [**Cubes**](/docs/analytics/cubes) | 4 cubes (sales, pipeline, service, marketing) for self-service analysis |
+| [**Dashboards**](/docs/analytics/dashboards) | **Five** ready-made dashboards — CRM Overview, Sales Performance, Sales Activity, Customer Service, Executive Overview |
+| [**Reports**](/docs/analytics/reports) | **Ten** standard reports — four on deals, three on cases, one each on leads, accounts and churn |
+| [**Cubes**](/docs/analytics/cubes) | **Nine** datasets in `src/datasets/`, the layer both of the above bind by name |
+
+Two counts on that table used to read differently, and the difference is worth knowing before you go looking. **There is no contract report** — and none can be built, because `src/datasets/` reads nine objects and `crm_contract` is not one of them, so nothing in analytics counts, sums or groups a contract. And **no cube is named** *Sales*, *Pipeline*, *Service* or *Marketing* — those four names exist nowhere in the app. [Cubes](/docs/analytics/cubes) says which datasets answer the first three and why the fourth has no data source at all.
 
 ## What you get out of the box
 
@@ -33,32 +35,34 @@ The moment you have data in HotCRM, you have analytics. No setup required.
 
 The same sharing rules that govern records also govern dashboards and reports:
 
-- A rep running *Pipeline by Stage* sees only their pipeline.
+- A rep running **Opportunities by Stage** sees only their own deals.
 - A manager sees the rollup for their team.
 - An exec sees company-wide.
 
 This means **one dashboard** works for the whole org — what each person sees depends on who they are.
 
+*Pipeline by Stage* is not a report you can run, though the name is real twice over: it is the open-pipeline funnel **tile** that CRM Overview, Sales Performance and Executive Overview share (`src/dashboards/shared-widgets.ts`), and it is the title of the bar chart on the **Opportunities by Stage** report. Search the report library for it and you find nothing — see [Reports](/docs/analytics/reports).
+
 ## Where to find things
 
-In the Enterprise CRM app, the **Analytics** group contains:
+In the Enterprise CRM app the group is called **Insights**, and it holds five entries:
+
+- **CRM Overview** — the cross-functional dashboard.
+- **Forecasts** — the forecast records themselves.
+- **Pipeline Coverage**, **Lead Inflow**, **SLA Performance** — three reports pinned to the sidebar by name.
 
-- **Dashboards** — the visual hub.
-- **Reports** — the report library.
-- **Cubes** — the data-modeling layer (admins).
+There is no *Analytics* group, and nothing in the sidebar is called *Dashboards*, *Reports* or *Cubes*. The other dashboards are pinned to the group that owns them — **Sales Performance** under Sales, **Sales Activity** under Activity, **Service Overview** under Service — and the rest of the report library is reached from the reports screen rather than from the nav. [Quick tour](/docs/getting-started/quick-tour) lists the whole sidebar, group by group.
 
 ## AI-powered insights
 
-The AI Copilot can read directly from cubes — ask questions like:
+**On this app's side, nothing connects the Copilot to a dataset.** The six skills under `src/skills/` name no dataset, no cube and no measure. The one that answers data questions, **Live Data Access**, is wired to the platform's object tools and is instructed to re-read the object's schema and aggregate over records — so a Copilot answer is computed from the objects, not read off a compiled cube.
 
-- *"What's driving the gap between commit and forecast?"*
-- *"Why did East close 3 fewer deals than West last quarter?"*
-- *"Show me which products correlate with longer sales cycles."*
+That also sets what it can be asked. *"What's driving the gap between commit and forecast?"* has data behind it: **Commit**, **Quota**, **Closed Won** and **Pipeline** are all real figures on the forecast record, published as **Commit**, **Quota**, **Closed** and **Pipeline** on the **Forecast Metrics** dataset. *"Why did East close 3 fewer deals than West last quarter?"* and *"Show me which products correlate with longer sales cycles"* do not — no region is stored on a deal, no dataset reads line items, and the one duration the opportunity carries (**Days in Current Stage**) is a formula evaluated after the query, so nothing can aggregate it. [Cubes](/docs/analytics/cubes) lists the gaps one by one.
 
-See [Sales Skills](/docs/ai-copilot/sales-copilot) for more.
+Whether the platform's own agent additionally reads the compiled cubes is a question about the runtime rather than about this app, so this page makes no claim about it in either direction; check it against your deployment. See [Sales Skills](/docs/ai-copilot/sales-copilot) for what the sales-side skills do.
 
 ## Start here
 
-- [Dashboards](/docs/analytics/dashboards) — the four built-in dashboards.
-- [Reports](/docs/analytics/reports) — the standard report library.
-- [Cubes](/docs/analytics/cubes) — for analysts who want to build their own.
+- [Dashboards](/docs/analytics/dashboards) — the five built-in dashboards.
+- [Reports](/docs/analytics/reports) — the ten standard reports.
+- [Cubes](/docs/analytics/cubes) — the nine datasets underneath, and what each can and cannot answer.

content/docs/analytics/index.zh-Hans.mdx

@@ -1,6 +1,6 @@
 ---
 title: 分析
-description: 仪表盘、报表，以及为它们提供支持的多维数据集——HotCRM 如何将日常活动转化为洞察。
+description: 仪表盘、报表，以及它们底下的语义层——HotCRM 如何将日常活动转化为洞察。
 ---
 
 # 分析
@@ -9,15 +9,17 @@ HotCRM 分析将你的 CRM 活动转化为**决策**。三个层次协同工作
 
 - **仪表盘** — 面向高管、经理和销售代表的可视化一览视图。
 - **报表** — 任何人都可以筛选和导出的详细列表与汇总报表。
-- **多维数据集** — 底层数据层，可用于自助式切片分析。
+- **多维数据集** — 底下的语义层。在本应用里，这一层是一组 **dataset**：HotCRM 自己不声明任何 cube，由分析服务在内部把每个 dataset 编译成它的 cube（ADR-0021）。
 
 ## 包含内容
 
 | 层次 | 作用 |
 | --- | --- |
-| [**仪表盘**](/zh-Hans/docs/analytics/dashboards) | 4 个开箱即用的仪表盘（CRM、销售、服务、高管） |
-| [**报表**](/zh-Hans/docs/analytics/reports) | 涵盖线索、交易、工单、合同的 10+ 个标准报表 |
-| [**多维数据集**](/zh-Hans/docs/analytics/cubes) | 4 个多维数据集（销售、管道、服务、市场营销），用于自助分析 |
+| [**仪表盘**](/zh-Hans/docs/analytics/dashboards) | **五个**开箱即用的仪表盘 —— CRM Overview、Sales Performance、Sales Activity、Customer Service、Executive Overview |
+| [**报表**](/zh-Hans/docs/analytics/reports) | **十份**标准报表 —— 交易 4 份、工单 3 份，线索、客户、流失各 1 份 |
+| [**多维数据集**](/zh-Hans/docs/analytics/cubes) | `src/datasets/` 下的**九个** dataset，也就是上面两层按名字绑定的那一层 |
+
+上表有两个数字过去写的是别的样子，去找东西之前值得先知道。**这里没有合同报表** —— 也建不出来：`src/datasets/` 只读九个对象，`crm_contract` 不在其中，所以分析里没有任何东西能对合同做计数、求和或分组。另外，**没有任何 cube 叫这四个名字** —— *Sales*、*Pipeline*、*Service*、*Marketing* 在应用里零命中。前三个由哪些 dataset 回答、第四个为什么连数据源都没有，见 [多维数据集](/zh-Hans/docs/analytics/cubes)。
 
 ## 开箱即用的能力
 
@@ -33,32 +35,34 @@ HotCRM 分析将你的 CRM 活动转化为**决策**。三个层次协同工作
 
 管理记录的共享规则同样管理仪表盘和报表：
 
-- 运行 *Pipeline by Stage* 的销售代表只能看到自己的管道。
+- 运行 **Opportunities by Stage** 的销售代表只能看到自己的交易。
 - 经理看到其团队的汇总。
 - 高管看到全公司范围。
 
 这意味着**一个仪表盘**适用于整个组织——每个人看到的内容取决于他们的身份。
 
+*Pipeline by Stage* 不是一张可以运行的报表，但这个名字有两个真身：它是 CRM Overview、Sales Performance 与 Executive Overview 共用的那块未结管道漏斗**磁贴**（`src/dashboards/shared-widgets.ts`），也是 **Opportunities by Stage** 报表上那张柱状图的标题。在报表库里按这个名字搜索，什么也搜不到 —— 见 [报表](/zh-Hans/docs/analytics/reports)。
+
 ## 在哪里可以找到
 
-在 Enterprise CRM 应用中，**分析**组包含：
+在 Enterprise CRM 应用里，这个分组叫 **Insights**，下面有五个条目：
+
+- **CRM Overview** —— 跨职能的那块仪表盘。
+- **Forecasts** —— 预测记录本身。
+- **Pipeline Coverage**、**Lead Inflow**、**SLA Performance** —— 三张按名字钉在侧边栏上的报表。
 
-- **仪表盘** — 可视化中枢。
-- **报表** — 报表库。
-- **多维数据集** — 数据建模层（管理员）。
+没有 *Analytics* 分组，侧边栏上也没有任何条目叫 *Dashboards*、*Reports* 或 *Cubes*。其余仪表盘各自钉在负责它的分组下 —— **Sales Performance** 在 Sales、**Sales Activity** 在 Activity、**Service Overview** 在 Service —— 报表库里剩下的报表则从报表界面进入，而不是从导航进入。整条侧边栏逐组的列法，见 [快速导览](/zh-Hans/docs/getting-started/quick-tour)。
 
 ## AI 驱动的洞察
 
-AI Copilot 可以直接从多维数据集读取数据——可以提出如下问题：
+**在本应用这一侧，没有任何东西把 Copilot 连到某个 dataset 上。** `src/skills/` 下的六个技能，没有一个点名任何 dataset、cube 或度量。其中负责回答数据问题的那个 —— **Live Data Access** —— 接的是平台的对象工具，而它的指令要求先重新读一遍对象的 schema，再在记录上做聚合。所以 Copilot 的回答是从对象上算出来的，而不是从编译好的 cube 上读出来的。
 
-- *"是什么造成了承诺额与预测额之间的差距？"*
-- *"为什么上季度东区比西区少成交了 3 笔交易？"*
-- *"告诉我哪些产品与更长的销售周期相关。"*
+这也决定了能问它什么。*"是什么造成了承诺额与预测额之间的差距？"* 背后是有数据的：预测记录上真实带着 **Commit**、**Quota**、**Closed Won**、**Pipeline** 四个数字，在 **Forecast Metrics** dataset 上发布为 **Commit**、**Quota**、**Closed**、**Pipeline** 四个度量。*"为什么上季度东区比西区少成交了 3 笔交易？"* 与 *"告诉我哪些产品与更长的销售周期相关。"* 则没有 —— 交易上不存区域，没有任何 dataset 读行项目，而商机唯一带着的那个时长（**Days in Current Stage**）是查询之后才算的公式，什么都聚合不了。这些缺口逐条列在 [多维数据集](/zh-Hans/docs/analytics/cubes)。
 
-更多内容请参阅 [销售技能](/zh-Hans/docs/ai-copilot/sales-copilot)。
+平台自带的 agent 是否另外也读编译后的 cube，是关于运行时而不是关于本应用的问题 —— 所以本页对它不做任何方向上的断言，请对着你自己的部署去核。销售侧的技能各自做什么，见 [销售技能](/zh-Hans/docs/ai-copilot/sales-copilot)。
 
 ## 从这里开始
 
-- [仪表盘](/zh-Hans/docs/analytics/dashboards) — 四个内置仪表盘。
-- [报表](/zh-Hans/docs/analytics/reports) — 标准报表库。
-- [多维数据集](/zh-Hans/docs/analytics/cubes) — 适合想要自行构建的分析师。
+- [仪表盘](/zh-Hans/docs/analytics/dashboards) —— 五个内置仪表盘。
+- [报表](/zh-Hans/docs/analytics/reports) —— 十份标准报表。
+- [多维数据集](/zh-Hans/docs/analytics/cubes) —— 底下的九个 dataset，以及每个能回答什么、不能回答什么。