revert(#170): drop the info-deleting compression, keep the measurement (round 2 start)

按 FIX-token-card-1：
- references/product_runtime_contract.md 恢复原样
- 删除自拟任务书（以 Claude 的 TASK-token-card-20261109 为准）
- 三处被弱化的断言恢复原样
- consultation-evidence-card.ts 回到 staging：删信息换来的压缩全部作废，重新按
  「不删信息」做
- 新增测量脚本 consult-card-size-170.mts（真实工具链，state.evidenceCardChars）

基线 c032823f 实测（3 盘 × 10 领域）：卡片 min 41,963 / 中位 46,017 / max 53,930 /
平均 46,587；modelVisibleChars 平均 91,559。块大小表见 PROGRESS。

Co-Authored-By: Grok (agent loop)

docs/tasks/PROGRESS-token-card-compress-20261109.md

@@ -51,3 +51,53 @@
 
 - 无模型凭据：盲测与「准确度不下降」留待 staging 真机。
 - 本机 Node 22 / Windows；未跑前端全量 `npm test` 与 Python 快速门（按执行方分工由 Claude 另跑）。
+
+
+## 第 1 次打回后的重做（2026-11-09，按 FIX-token-card-1 + 正式任务书）
+
+### 已撤回
+
+- `references/product_runtime_contract.md` 已恢复原样（不再改生产合同文件）。
+- 删除自拟任务书 `docs/tasks/TASK-token-card-compress-20261109.md`；以 Claude 的 `TASK-token-card-20261109.md` + `FIX-token-card-1.md` 为准。
+- 上一轮改动过的三处断言（`consult-card-calendar-narayana-20261108`、`consult-downstream-labels-r3-20261107`、`consult-evidence-card-full-data-20261107`）已恢复原样。
+- `frontend/src/lib/consultation-evidence-card.ts` 回到 staging 版本：上一轮靠删信息换来的压缩全部作废，重新按「不删信息」来做。
+
+### 改前基线（用真实工具链量，`state.evidenceCardChars`）
+
+用 `frontend/scripts/research/consult-card-size-170.mts`，3 张公开盘 × 10 个引擎领域 = 30 例；
+基线提交 `c032823f9`（staging，已含 #169）：
+
+| 指标 | 模型可见卡片字符数 |
+| --- | ---: |
+| 最小 | 41,963 |
+| 中位 | 46,017 |
+| 最大 | 53,930 |
+| 平均 | 46,587 |
+
+同一脚本量出的 `modelVisibleChars`（含契约与 methodology）平均 91,559。
+
+### 基线里各块有多大（30 例平均）
+
+| 块 | 平均字符数 |
+| --- | ---: |
+| `base`（合计） | 35,892 |
+| ├ `base.timing` | 17,360 |
+| │ ├ `dasha_systems`（26 个系统） | 7,525 |
+| │ ├ `sade_sati_timeline` | 2,622 |
+| │ ├ `transits_detail` | 2,478 |
+| │ ├ `annual_brief` | 2,105 |
+| │ └ `narayana` | 1,624 |
+| ├ `base.natal` | 15,539 |
+| │ ├ `yogas_detail` | 8,298 |
+| │ ├ `affliction_evidence` | 2,721 |
+| │ └ `planets` | 1,503 |
+| └ `base.parameters` | 1,719 |
+| `domain_sections` | 8,882 |
+| `meta` | 1,553 |
+
+### 一个重要约束（下一步的前提）
+
+查卡段名（`EVIDENCE_LOOKUP_SECTIONS`）里只有 `vimshottari_mahadashas`、`chara_dasha` 这类，
+**没有**其余 24 个大运系统，也**没有**年运（Tajika）块。按 FIX 的要求「某层没有对应段名时，这层不许下卡」，
+这两块必须留在卡上。所以 30% 只能靠结构压缩（同事实引用、恒定说明只写一次、去空值、表格改列名 + 行数组），
+不能靠下卡。下一步从 `base.timing.dasha_systems` 与 `base.natal.yogas_detail` 两块开始。

docs/tasks/TASK-token-card-compress-20261109.md

@@ -1,92 +0,0 @@
-# TASK · 省 token：数据卡按本题路由裁层（#170 / BUG-1813）
-
-- 基线：`origin/staging` `c032823f9`（2026-11-09，round7 整合重封之后）
-- 执行分支：`codex/token-card-compress-20261109`
-- 关联：issue #170（`agent/排队` → 本单领取后 `agent/执行中`）、#171（查卡多段，另一分支）、BUG-1813
-
-## 1. 事实与测量
-
-写答案的模型每轮读「答题契约 + 数据卡」。2026-11-07 占星师方案把「完整数据」放上卡（`TASK-consult-evidence-card-v2-20260927` 之后的第四轮）：全部 26 个大运系统的当前层、全部 Yoga 行连同每行的行星元数据、年运精简块、受冲账本，全部随每张卡送达。
-
-在 `c032823f9` 上用真实金样（`frontend/tests/fixtures/consult-evidence-card-golden.json`，3 张公开盘 × 12 个领域）测「答题契约 + 数据卡」的模型可见字符数：
-
-| 指标 | 基线 |
-| --- | --- |
-| 平均 | 46,664 |
-| 最小 | 41,705 |
-| 最大 | 54,532（`barack_obama` / `general`） |
-
-单张卡里最大的块（`barack_obama` / `general`）：
-
-| 块 | 字符 | 说明 |
-| --- | --- | --- |
-| `base.natal.yogas_detail` | 7,918 | 全部 Yoga 行 + 每行的 `owns` / `functional` / `dignity` 元数据 |
-| `base.timing.dasha_systems` | 7,794 | 26 个大运系统，其中 20+ 个是 `applicable:false` 的说明行 |
-| `base.timing.sade_sati_timeline` | 2,695 | 近 7 年 Sade Sati 行 + 方法说明 |
-| `base.natal.affliction_evidence` | 2,765 | 受冲账本 |
-| `base.timing.transits_detail` | 2,476 | 行运 |
-| `base.timing.annual_brief` | 2,270 | 年运块，每个领域都带 |
-| `domain.argala` | 3,199 | 领域 Argala |
-| `domain.kp` | 2,007 | KP Placidus |
-
-`EVIDENCE_CARD_SPECS[domain].route` 已经在每张卡上写明本题要读的技法（`D1 / D9 / D10 / … / Vimshottari / Narayana / transits / KP`）。基础段却把路由没点名的层也塞进去：非路由大运系统、年运块、yoga 元数据。
-
-## 2. 根因
-
-`baseExtras()` 是「整个盘一份基础段」的实现，不知道当前领域的路由；只要引擎算出了某一层，就无条件上卡。2026-11-07 的「完整数据」决定把可上卡的层扩到全部大运与年运，于是每张卡都带着本题不读的层。
-
-## 3. 决策记录（产品 #170，2026-11-09）
-
-1. **按本题路由裁层**：基础段只保留 `EVIDENCE_CARD_SPECS[domain].route` 点名的层。
-   - 大运系统：只留路由点名的 Vimshottari / Narayana；其余 24 个系统改为查卡工具可取。
-   - Yoga：只在 `spec.yogas === true` 的领域上卡（career / wealth / parents / children / general）；其余领域查卡。
-   - 年运块：只在 `annual` 路由上卡；其余领域查卡。
-2. **同一事实只算一次**：Yoga 行的 `planets` 从 `{ 星: { owns, functional, dignity } }` 压成星名数组；行星事实仍在 `base.natal.planets`、`functional_benefic_malefic` 与领域宫位行里。
-3. 这是一次**推翻 2026-11-07「回答模型拿到完整数据」的有限让步**：只让出「本题路由不读的层」，路由点名的层一张不少；完整数据仍在服务端回执与查卡工具里。
-4. 不改引擎、不改提示词规则、不改数据库；金样不重采。
-
-## 4. 硬红线
-
-1. 路由点名的层不得缺失：每条路由的 Vimshottari / Narayana / 本领域正式分盘 / 本领域宫 / KP Placidus / 行运。
-2. 不得删事实：被裁的层必须仍能从 `read-consultation-evidence` 取到，或在卡上另一处存在。
-3. 不改既有断言的语义；必须改的断言写「原值 / 新值 / 原因」三栏。
-4. 金样 `consult-evidence-card-golden.json` 不动。
-5. 不改 `scripts/**`、不改数据库、不改 `.gitea/**`。
-
-## 5. 任务分解
-
-| # | 任务 | 验收 |
-| --- | --- | --- |
-| T1 | `baseExtras()` 接收领域；大运系统按 `spec.route` 过滤；Yoga 按 `spec.yogas` 上卡；年运块按 `spec.annualHouses` 上卡 | 36 张卡测量通过；每条路由的双大运轨仍在 |
-| T2 | Yoga 行 `planets` 压成星名数组 | 行的星名与引擎 `yoga_conditions.rows[].planets` 的键一致 |
-| T3 | 体积回归测试 `consult-card-size-170.test.ts` | 平均 ≤ 33,000、最大 ≤ 44,000；路由层逐项存在 |
-| T4 | 改既有断言（两处大运、一处 yoga、一处年运） | 三栏说明齐全；`consult*.test.ts` 全量通过 |
-| T5 | 文案：运行合同里 `Jaimini karaka` 改为 `Jaimini karakas`（清单词扫描要求是卡上键名） | `consult-affliction-reading-20261001` 通过 |
-
-## 6. 让步顺序
-
-1. 先保「路由点名的层一张不少」与「事实可查」。
-2. 再保 30% 体积目标。
-3. 最后才考虑改断言措辞；不得为过测试删字段。
-
-## 7. 开工前置
-
-```powershell
-git fetch origin --prune
-git worktree add -b codex/token-card-compress-20261109 .worktrees/token-card-compress-20261109 origin/staging
-```
-
-前端自测（Windows，工作树内）：
-
-```powershell
-frontend\node_modules\.bin\tsc --noEmit -p frontend
-frontend\node_modules\eslint\bin\eslint.js <改动文件>
-node frontend\node_modules\tsx\dist\cli.mjs --test frontend/tests/consult-card-size-170.test.ts
-node frontend\node_modules\tsx\dist\cli.mjs --test (全部 frontend/tests/consult*.test.ts)
-```
-
-盲测（同一批题、改前 / 改后各跑一次、人工评分）需要模型凭据；本机没有，按 `BLOCKED.md` 记为环境缺口，部署后由产品负责人在 staging 走真机。
-
-## 8. BUG 编号
-
-开工时 `docs/BUG_HISTORY.md` 最大号是 BUG-1809（本分支）；#170 标题预留 BUG-1813，本单写 BUG-1813。

frontend/scripts/research/consult-card-size-170.mts

@@ -0,0 +1,70 @@
+/**
+ * #170 测量：模型可见数据卡字符数，逐盘逐领域（走真实工具链，读 runtime state）。
+ * 运行：frontend > node node_modules/tsx/dist/cli.mjs scripts/research/consult-card-size-170.mts
+ */
+import { readFileSync } from "node:fs";
+
+import {
+  createConsultationRuntimeState,
+  createConsultationTools,
+} from "../../src/mastra/consultation-tools.ts";
+import { consultationEngineDomainIds } from "../../src/lib/consultation-domain-registry.ts";
+
+type Json = Record<string, unknown>;
+
+const golden = JSON.parse(readFileSync(
+  new URL("../../tests/fixtures/consult-evidence-card-golden.json", import.meta.url),
+  "utf8",
+)) as { charts: Array<{ id: string; workflow: Json }> };
+
+const rows: Array<{ chart: string; domain: string; card: number; visible: number }> = [];
+
+for (const chart of golden.charts) {
+  for (const domain of consultationEngineDomainIds) {
+    const state = createConsultationRuntimeState();
+    const tools = createConsultationTools({
+      userId: "u",
+      sessionId: "s",
+      requestId: `size-${chart.id}-${domain}`,
+      consultationMode: "verified_chart",
+      theme: domain,
+      serverChart: {
+        name: "public",
+        toolInput: {
+          year: 1955, month: 2, day: 24, hour: 19, minute: 15, city: "San Francisco",
+          lat: 37.77, lon: -122.42, tz: -8, ayanamsa: "raman",
+        },
+        truth: { birthTimeSource: "reported" },
+      } as never,
+      state,
+      runWorkflow: async () => structuredClone(chart.workflow) as never,
+    });
+    await tools["run-jyotish-consultation"].execute!(
+      { question: "这一块怎么样", domains: [domain] } as never,
+      { writer: { custom: async () => {} } } as never,
+    );
+    rows.push({
+      chart: chart.id,
+      domain,
+      card: state.evidenceCardChars ?? 0,
+      visible: state.modelVisibleChars ?? 0,
+    });
+  }
+}
+
+const summarize = (values: number[]) => {
+  const sorted = [...values].sort((a, b) => a - b);
+  return {
+    cases: sorted.length,
+    min: sorted[0],
+    median: sorted[Math.floor(sorted.length / 2)],
+    max: sorted[sorted.length - 1],
+    mean: Math.round(sorted.reduce((sum, value) => sum + value, 0) / sorted.length),
+  };
+};
+
+console.log(JSON.stringify({
+  card: summarize(rows.map((row) => row.card)),
+  visible: summarize(rows.map((row) => row.visible)),
+  rows,
+}, null, 2));

frontend/src/lib/consultation-evidence-card.ts

@@ -1904,18 +1904,7 @@ const DASHA_ROW_KEYS = [
   "continuity_issue",
 ] as const;
 
-function baseExtras(
-  context: unknown,
-  base: EvidenceCard["base"],
-  gaps: string[],
-  domain: ConsultationDomain,
-) {
-  // #170: the base section carries only the layers this domain's route names.
-  // A route does not name a dasha system, a yoga family or the annual Tajika
-  // block, so the card leaves it to the one-shot lookup instead of shipping it
-  // on every question.
-  const spec = EVIDENCE_CARD_SPECS[domain];
-  const routeSystems = new Set(spec.route.map((item) => item.toLowerCase()));
+function baseExtras(context: unknown, base: EvidenceCard["base"], gaps: string[]) {
   const layers = nativeLayers(context);
   const localLayers = record(record(context).local_layers);
   const parameters = readyLayer(layers, "card_parameters");
@@ -1932,9 +1921,6 @@ function baseExtras(
     for (const raw of dashas.rows) {
       const row = record(raw);
       if (typeof row.system !== "string") continue;
-      // #170: other systems stay in the lookup; the card keeps the ones this
-      // route reads (every route names Vimshottari and Narayana).
-      if (!routeSystems.has(row.system.toLowerCase())) continue;
       if (row.status === NOT_AVAILABLE_YET && row.applicable === undefined) {
         gaps.push(`base.dasha_systems.${row.system}:${NOT_AVAILABLE_YET}`);
         continue;
@@ -1998,49 +1984,37 @@ function baseExtras(
   } else {
     gaps.push("base.sade_sati_timeline");
   }
-  // #170: the annual Tajika block rides only on the annual route; other
-  // questions read it through the lookup when they ask about this year.
-  if (spec.annualHouses) {
-    const annualPack = record(layers.annual_tajika);
-    if (annualPack.status === "parameter_sensitive" || annualPack.status === "partial_verified") {
-      const reference = String(record(layers.transit_detail).reference_date ?? "").slice(0, 10);
-      const mudda = Array.isArray(annualPack.mudda_dasha) ? annualPack.mudda_dasha.map(record) : [];
-      const at = reference ? mudda.findIndex((row) => String(row.start) <= reference && reference < String(row.end)) : -1;
-      const muddaNow = at >= 0 ? mudda.slice(at, at + 2) : mudda.slice(0, 2);
-      base.timing.annual_brief = {
-        ...pickKeys(annualPack, ["status", "varsha_year", "annual_lagna", "varshesha", "varshesha_basis", "muntha", "tajika_aspects", "tajika_aspects_status", "boundary",
-          // BUG-1722: the Mudda rows are a view clipped to this return -> next return (her Q2).
-          "mudda_view",
-          // BUG-1731 (her Q2): 「今年」 is the calendar year, which spans two Tajika years split at the
-          // return; both years' coverage and the split date travel (the engine computed them, the card dropped them).
-          "calendar_year"]),
-        mudda_running_and_next: muddaNow.map((row) => copy(row)).filter((row): row is Json => row !== undefined),
-      };
-    } else {
-      gaps.push(`base.annual_brief${typeof annualPack.reason === "string" ? `:${annualPack.reason}` : ""}`);
-    }
+  const annualPack = record(layers.annual_tajika);
+  if (annualPack.status === "parameter_sensitive" || annualPack.status === "partial_verified") {
+    const reference = String(record(layers.transit_detail).reference_date ?? "").slice(0, 10);
+    const mudda = Array.isArray(annualPack.mudda_dasha) ? annualPack.mudda_dasha.map(record) : [];
+    const at = reference ? mudda.findIndex((row) => String(row.start) <= reference && reference < String(row.end)) : -1;
+    const muddaNow = at >= 0 ? mudda.slice(at, at + 2) : mudda.slice(0, 2);
+    base.timing.annual_brief = {
+      ...pickKeys(annualPack, ["status", "varsha_year", "annual_lagna", "varshesha", "varshesha_basis", "muntha", "tajika_aspects", "tajika_aspects_status", "boundary",
+        // BUG-1722: the Mudda rows are a view clipped to this return -> next return (her Q2).
+        "mudda_view",
+        // BUG-1731 (her Q2): 「今年」 is the calendar year, which spans two Tajika years split at the
+        // return; both years' coverage and the split date travel (the engine computed them, the card dropped them).
+        "calendar_year"]),
+      mudda_running_and_next: muddaNow.map((row) => copy(row)).filter((row): row is Json => row !== undefined),
+    };
+  } else {
+    gaps.push(`base.annual_brief${typeof annualPack.reason === "string" ? `:${annualPack.reason}` : ""}`);
   }
   const bookRules = readyLayer(layers, "book_rules");
   if (bookRules) base.book_rules = pickKeys(bookRules, ["status", "routes", "runtime_packets", "route_reference_rules"]) as JsonRecord;
-  const yogas = spec.yogas ? readyLayer(layers, "yoga_conditions") : null;
+  const yogas = readyLayer(layers, "yoga_conditions");
   if (yogas && Array.isArray(yogas.rows)) {
     // 产品 2026-11-07 同步占星师 Yoga 规则批次: claim_kind (structural_observation
     // is not a named yoga) and the executed formation checks, copied verbatim.
-    // #170: the per-row planet block (owns / functional / dignity) repeated the
-    // base planets table; the row keeps the names, the facts stay in base.natal.
-    const rows = yogas.rows.map((row) => {
-      const picked = pickKeys(row, [
-        "name", "category", "status", "claim_kind", "claim_status", "condition", "conditions_met", "reductions",
-        "formation", "formation_checks", "planets", "running_dasha_levels",
-        // 占星师 2026-11-08 第六轮 R3: Neecha Bhanga witnesses with reference / relative_house /
-        // self_reference; `self_reference_only` rows hold by the literal source branch only.
-        "formation_evidence",
-      ]);
-      const names = Object.keys(record(picked.planets));
-      if (names.length) picked.planets = names;
-      else delete picked.planets;
-      return picked;
-    });
+    const rows = yogas.rows.map((row) => pickKeys(row, [
+      "name", "category", "status", "claim_kind", "claim_status", "condition", "conditions_met", "reductions",
+      "formation", "formation_checks", "planets", "running_dasha_levels",
+      // 占星师 2026-11-08 第六轮 R3: Neecha Bhanga witnesses with reference / relative_house /
+      // self_reference; `self_reference_only` rows hold by the literal source branch only.
+      "formation_evidence",
+    ]));
     if (rows.some((row) => row.condition === NOT_AVAILABLE_YET)) gaps.push(`base.yogas_detail.condition:${NOT_AVAILABLE_YET}`);
     // 占星师第三轮审查 2026-11-07: rules the engine could not evaluate are listed
     // apart (never as "not formed"); absent when every rule was evaluated.
@@ -2052,7 +2026,7 @@ function baseExtras(
       rows,
       ...(notEvaluated.length ? { not_evaluated: notEvaluated, ...pickKeys(yogas, ["not_evaluated_rule", "rule_engine_status"]) } : {}),
     };
-  } else if (spec.yogas) {
+  } else {
     gaps.push("base.yogas_detail");
   }
   const strength = readyLayer(layers, "strength_detail");
@@ -2793,7 +2767,7 @@ export function buildEvidenceCard(
     natal: baseNatal(claimEvidence(first.packet, "natal_foundation"), gaps),
     timing: baseTiming(claimEvidence(first.packet, "timing"), gaps, horizon),
   };
-  baseExtras(first.context, base, gaps, first.domain);
+  baseExtras(first.context, base, gaps);
   const domainSections: EvidenceCard["domain_sections"] = {};
   const domains: ConsultationDomain[] = [];
   const planFor = (domain: ConsultationDomain): EvidenceCardVargaPlan | undefined => {
@@ -2819,7 +2793,7 @@ export function buildEvidenceCard(
         natal: baseNatal(claimEvidence(source.packet, "natal_foundation"), ownGaps),
         timing: baseTiming(claimEvidence(source.packet, "timing"), ownGaps, horizon),
       };
-      baseExtras(source.context, candidate, ownGaps, source.domain);
+      baseExtras(source.context, candidate, ownGaps);
       if (JSON.stringify(candidate) !== JSON.stringify(base)) {
         own = candidate;
         gaps.push(`${source.domain}.base_differs`, ...ownGaps.map((gap) => `${source.domain}.${gap}`));

frontend/tests/consult-card-calendar-narayana-20261108.test.ts

@@ -11,7 +11,7 @@ import type { ConsultationDomain } from "../src/lib/consultation-domain-registry
 
 type Json = Record<string, unknown>;
 const golden = JSON.parse(readFileSync(new URL("./fixtures/consult-evidence-card-golden.json", import.meta.url), "utf8")) as {
-  charts: Array<{ id: string; workflow: Json; annual_workflow: Json; timing_workflow: Json }>;
+  charts: Array<{ id: string; workflow: Json; timing_workflow: Json }>;
 };
 
 function card(workflow: Json, domain: ConsultationDomain): EvidenceCard {
@@ -24,11 +24,8 @@ const nativeAnnual = (workflow: Json) => ((workflow.chart as Json).consultation_
 const narayanaModule = (workflow: Json) => (((workflow.chart as Json).modules as Json).narayana_dasha as Json);
 
 test("BUG-1731: the card carries the calendar year the engine computed: both Tajika years and the return that splits them", () => {
-  // 原值：timing / wealth 卡也带 base.timing.annual_brief.calendar_year
-  // 新值：年运块只在 annual 路由上卡（EVIDENCE_CARD_SPECS.annual.annualHouses），其余题目从查卡工具取
-  // 原因：#170 省 token：数据卡按本题路由裁层
   for (const chart of golden.charts) {
-    for (const [workflow, domain] of [[chart.annual_workflow, "annual"]] as const) {
+    for (const [workflow, domain] of [[chart.timing_workflow, "timing"], [chart.workflow, "wealth"]] as const) {
       const engine = nativeAnnual(workflow).calendar_year as Json;
       const onCard = (card(workflow, domain).base.timing.annual_brief as Json).calendar_year as Json;
       assert.deepEqual(onCard, engine, `${chart.id} ${domain}`);

frontend/tests/consult-downstream-labels-r3-20261107.test.ts

@@ -63,27 +63,24 @@ test("Shadbala and Argala sections carry their reading labels", () => {
   }
 });
 
-test("dasha systems carry the boundary rule and the route's own systems", () => {
+test("dasha systems carry the boundary rule and full instants", () => {
   const built = card(golden.charts[0].workflow, "career");
   const timing = record(record(built.base).timing);
   const dashas = record(timing.dasha_systems);
   assert.equal(dashas.boundary_rule, "start_inclusive_end_exclusive");
-  // 原值：卡上 26 个大运系统，逐行检查 md.start_at 的时刻格式
-  // 新值：卡上只留本题路由点名的 Vimshottari / Narayana；其余系统交给查卡工具
-  // 原因：#170 按本题路由裁层（省 token），大运系统不再每张卡全量送达
-  const systems = record(dashas.systems);
-  assert.deepEqual(Object.keys(systems).sort(), ["narayana", "vimshottari"]);
-  assert.match(String(record(record(systems.narayana).md).start), /^\d{4}-\d{2}-\d{2}$/);
+  // Producers with a time of day carry start_at; date-only producers keep the date alone.
+  const stamps = Object.values(record(dashas.systems))
+    .map((row) => record(record(row).md).start_at)
+    .filter((value): value is string => typeof value === "string");
+  assert.ok(stamps.length > 0);
+  for (const stamp of stamps) assert.match(stamp, /^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}([+-]\d{2}:\d{2})$/);
 });
 
 test("yoga rules the engine could not evaluate are listed apart from formed yogas", () => {
-  // 原值：family 卡也带 yogas_detail
-  // 新值：yoga 只在 EVIDENCE_CARD_SPECS[domain].yogas 为 true 的领域上卡，改用 career
-  // 原因：#170 按本题路由裁层
-  const built = card(golden.charts[0].workflow, "career");
+  const built = card(golden.charts[0].workflow, "family");
   const detail = record(record(record(built.base).natal).yogas_detail);
   const notEvaluated = Array.isArray(detail.not_evaluated) ? detail.not_evaluated as Json[] : [];
-  assert.ok(notEvaluated.length > 0, "the career packet has three rules without an evaluator");
+  assert.ok(notEvaluated.length > 0, "the family packet has three rules without an evaluator");
   const formed = new Set((Array.isArray(detail.rows) ? detail.rows as Json[] : []).map((row) => row.name));
   for (const row of notEvaluated) {
     assert.ok(["not_implemented", "evaluation_failed"].includes(String(row.evaluation_status)));

frontend/tests/consult-evidence-card-full-data-20261107.test.ts

@@ -68,23 +68,13 @@ test("base parameters are the engine's own (ayanamsa, nodes, house systems, kara
   }
 });
 
-test("only the route's dasha systems ride the base card, the rest stay a lookup", () => {
-  // 原值：每个 applicable 的系统都在卡上（2026-11-07「回答模型拿到完整数据」），
-  //       not_available 的系统写 base.dasha_systems.<name> gap
-  // 新值：只保留本题路由（EVIDENCE_CARD_SPECS[domain].route）点名的大运系统；
-  //       其余系统交给查卡工具，不再上卡、不再记 gap
-  // 原因：#170 省 token：数据卡按本题路由裁层，同一事实不在每个问题上重复送达
+test("every dasha system's running layer is on the base card, not-yet-computed systems are gaps", () => {
   for (const chart of golden.charts) {
     const card = cardFor(chart.workflow, "career");
     const systems = dig(card.base.timing, "dasha_systems", "systems") as Json;
     const rows = dig(native(chart.workflow), "dasha_systems", "rows") as Json[];
-    const routeSystems = new Set(["vimshottari", "narayana"]);
     for (const row of rows) {
       const name = row.system as string;
-      if (!routeSystems.has(name)) {
-        assert.equal(systems[name], undefined, `${chart.id} ${name} not on the card`);
-        continue;
-      }
       if (row.applicable === undefined) {
         assert.equal(systems[name], undefined, `${chart.id} ${name}`);
         assert.ok(card.meta.gaps.includes(`base.dasha_systems.${name}:not_available_yet`), `${chart.id} ${name} gap`);
@@ -92,7 +82,7 @@ test("only the route's dasha systems ride the base card, the rest stay a lookup"
       }
       for (const level of ["md", "ad", "pd"]) assert.deepEqual(dig(systems, name, level), row[level], `${chart.id} ${name} ${level}`);
     }
-    for (const name of ["vimshottari", "narayana"]) {
+    for (const name of ["vimshottari", "narayana", "chara", "sthira", "drig", "shoola", "niryaana_shoola", "navamsha", "lagna_kendradi", "yogini", "kalachakra", "ashtottari", "tribhagi"]) {
       assert.ok(systems[name], `${chart.id} ${name}`);
     }
     // Vimshottari agrees with the card's own running periods; Narayana carries dates.
@@ -100,17 +90,13 @@ test("only the route's dasha systems ride the base card, the rest stay a lookup"
     assert.equal(dig(systems, "narayana", "md", "sign"), dig(card.base.timing.narayana, "md", "sign"));
     assert.match(String(dig(systems, "narayana", "md", "start")), /^\d{4}-\d{2}-\d{2}$/);
     assert.equal(typeof dig(systems, "narayana", "date_basis"), "string");
+    assert.equal(typeof dig(systems, "ashtottari", "applicable"), "boolean");
   }
 });
 
 test("transit detail, yogas with conditions, strength and Jaimini sit on the base card", () => {
   for (const chart of golden.charts) {
     const card = cardFor(chart.workflow, "timing");
-    // 原值：timing 卡也带 yogas_detail
-    // 新值：yoga 只在 EVIDENCE_CARD_SPECS[domain].yogas 为 true 的领域（career / general / …）上卡，
-    //       timing / annual 等路由不含 yoga 的题目从查卡工具取
-    // 原因：#170 按本题路由裁层
-    const yogaCard = cardFor(chart.workflow, "general");
     const layers = native(chart.workflow);
     const transits = card.base.timing.transits_detail as Json;
     assert.deepEqual(transits.bodies, dig(layers, "transit_detail", "bodies"), chart.id);
@@ -122,7 +108,7 @@ test("transit detail, yogas with conditions, strength and Jaimini sit on the bas
     assert.notEqual(dig(transits, "bodies", "Moon", "chandra_bala"), undefined);
     assert.equal(typeof dig(transits, "sade_sati", "active"), "boolean");
     assert.deepEqual(transits.key_dates, dig(layers, "slow_transits", "key_dates"));
-    const yogas = yogaCard.base.natal.yogas_detail as Json;
+    const yogas = card.base.natal.yogas_detail as Json;
     assert.equal((yogas.rows as Json[]).length, (dig(layers, "yoga_conditions", "rows") as Json[]).length);
     // 原值：status 只有 candidate / computed
     // 新值：结构观察行写 structural_observation，名字带 (Observation)
@@ -131,13 +117,6 @@ test("transit detail, yogas with conditions, strength and Jaimini sit on the bas
       assert.ok(["candidate", "computed", "structural_observation"].includes(String(row.status)), `${chart.id} ${row.name}`);
       assert.equal(row.status === "structural_observation", String(row.name).endsWith("(Observation)"), `${chart.id} ${row.name}`);
       if (row.status === "structural_observation") assert.equal(row.claim_kind, "structural_observation");
-      // 原值：row.planets 是 { 星: { owns, functional, dignity } }
-      // 新值：row.planets 是星名数组；owns / functional / dignity 仍在 base.natal 的行星表与功能吉凶表
-      // 原因：#170 同一事实只算一次，去掉逐行重复的行星元数据
-      if (row.planets !== undefined) {
-        assert.ok(Array.isArray(row.planets), `${chart.id} ${row.name} planets are names`);
-        for (const name of row.planets as string[]) assert.equal(typeof name, "string");
-      }
     }
     const strength = card.base.natal.strength as Json;
     assert.deepEqual(strength.bhava_bala_rupas, dig(layers, "strength_detail", "bhava_bala", "houses"));

references/product_runtime_contract.md

@@ -203,7 +203,7 @@ evidence_card.backstage records, for the backend only, that global web evidence
 <!-- 产品规则，不是占星师的方法：S-12 讲时间时两套运程都读到当前层。 -->
 12. 讲时间（最近、今年、接下来几年）时，两套运程都读到当前那一层：Vimshottari 的大运、小运、细运，Narayana 的主运和子运，各落在哪个宫、管什么。主运和子运方向不同时两面都说，不只取主运。
 <!-- 产品规则，不是占星师的方法：S-13 跨层同源只算一个来源（BUG-1773）。落实占星师 2026-11-08 验收说明（rulings-round6-reply-20261108「产品正在做的四件事」第 1 条：「不能把同源规则条数当独立证据票数」）。BUG-1758 只在受冲来源账内按事实去重；r6 盲测人工复核 3/10、staging 2/9 的回答把跨层复用（同一颗星既是 AmK 又是 AL 主星；同一宫主在 D1、Arudha、Jaimini 各算一次；同一月亮事实既进瑜伽又进力量）当成独立支持。卡上新增各领域 `independence`（fact_id → 复用它的层；support / pressure 按不同 fact_id 列）。 -->
-13. 卡上各领域的 `independence` 是同源索引：`facts` 的每个键是一个盘面事实（一颗星的 D1 落座落宫、一个同宫、一个相位、一颗星的功能吉凶），值是复用这个事实的层（宫主、宫内星、Arudha 主星、Jaimini karakas、瑜伽、事业分析、Shadbala、大运主星、受冲来源）；`support` / `pressure` 列出支持 / 压力一面的不同事实。几层共用同一个事实时，写成一个来源，并点名复用它的层（例「土星落天秤 2 宫这一个事实，同时是 AmK 和 A2 主星」），不当成几条独立证据；独立的支持或压力只按不同的事实计。几层共用一个事实时，不写「多层一致支持」「多体系一致」「N 条独立线索」。
+13. 卡上各领域的 `independence` 是同源索引：`facts` 的每个键是一个盘面事实（一颗星的 D1 落座落宫、一个同宫、一个相位、一颗星的功能吉凶），值是复用这个事实的层（宫主、宫内星、Arudha 主星、Jaimini karaka、瑜伽、事业分析、Shadbala、大运主星、受冲来源）；`support` / `pressure` 列出支持 / 压力一面的不同事实。几层共用同一个事实时，写成一个来源，并点名复用它的层（例「土星落天秤 2 宫这一个事实，同时是 AmK 和 A2 主星」），不当成几条独立证据；独立的支持或压力只按不同的事实计。几层共用一个事实时，不写「多层一致支持」「多体系一致」「N 条独立线索」。
 <!-- /product-block:shared-reading -->
 
 ## F. 可能违反占星师 2026-11-07 写作底线的现有产品规则（第 1–6 条已按她 2026-11-08 第五轮裁定 6 修改）