clearai-dsh 0.4.0 → 0.5.0

package/CHANGELOG.md

@@ -2,6 +2,46 @@
 
 All notable changes to this project are recorded here. The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [Unreleased]
+
+## [0.5.0] — 2026-10-08
+
+这一版是 0.5 升级的阶段一与阶段二:预期与未解释项、评估者查数据可信、不可逆命令拦在执行前、三档等级、经验与本体分类,宿主下界提到 DSH 0.2。
+
+**开发集上的初步结果**(盲评,不当作有效性证据,见 [docs/optimization/sim-runs/](docs/optimization/sim-runs/)):反应器题漂移纠正 ClearAI 3/4、裸模型 1/4;同一现场 3 题序列里第 3 题全部用上了前两题的经验,第 3 题均分 1.5 对 1.17,差距来自漂移这一项;配方与探索深度两边持平。正式验证和探索深度([设计草稿](docs/optimization/0.5-plan/exploration-depth.zh-CN.md))留到之后的版本。
+
+### Added
+
+- **预期与未解释项(反常)**:步骤可以带预期 `expect`(预计看到什么、从哪条关系或经验来),动手前也能用 `RevisePlan(action="expect")` 补写。结果和预期或本体对不上的地方,交付时写进 `AdvancePlan.anomalies`,或随时用新工具 `Anomaly` 登记;它们挂在卡上,只有三个去处:被解释、写明理由排除、交给人。不阻塞结案。`STATE_VERSION` 17(投影多了 `anomalies`)。
+- **评估者查还有什么没解释**:任务书列出未解释项,纪律要求查原始数据里没登记的异常(参考读数、重复点漂移、装置之间的差)和排除理由站不站得住;裁决多一个 `anomalies` 字段,折成「评估者发现」的未解释项退回给模型。
+- **不可逆动作拦在命令上**:`Frame` 的 `irreversible` 声明动作和命令特征,匹配的 bash 命令执行前当场问人;不放行或没人能答就拒。
+- **经验(跨任务的「下次怎么做」)**:`Conclude(outcome="achieved")` 多一个可选的 `lessons`(每条一句,分坑 / 先核 / 会骗人的捷径 / 先验四类,带涉及的装置或量、凭据、不适用的条件)。结案评估者对照记录逐条判(裁决多一个 `lessons` 字段),支持的由系统写进 `clear/knowledge/lessons/<id>.json`(做的人写不了),其余在回执里列出。别的会话经工作区同步读到它们;卡上在立题、定计划、写预期前摆出最多 6 条,和眼下目标、下一步提到同一装置或量的排前面。人把文件里的 `status` 改成 `retracted` 就撤回。`STATE_VERSION` 18(投影多了 `lessons`,派生多了 `lessonRows`)。
+- **本体只留能产出预期的东西**:概念分类别 / 度量 / 现象(`kind`),度量必须写 `unit`;关系分影响 / 定义 / 测量 / 表现为(`kind`),影响关系可写大致形状 `shape`(单调升、单调降、有峰、阈值、耦合),测量关系写读数怎么核(`check`);实体可带带日期的 `history`。卡上「用到的关系」把形状和核法递出来;词面命中也看下一步要做什么和它的预期。都是可选格,旧文件照旧有效,指纹不变。
+- 0.5 的设计与升级计划:[meta 设计](docs/optimization/0.5-plan/meta-design.zh-CN.md)、[升级计划](docs/optimization/0.5-plan/upgrade-plan.zh-CN.md)。
+
+### Changed
+
+- 本体与认识论状态在决策点递到模型眼前:运行态卡给出判断用到的概念释义和相关事实的原话与边界;目标还没立或已结案时,直接列出以前留下的事实。
+- 本体的提示改为只写推理要用的东西:度量写成概念、释义写口径;因果关系写进关系;实体只在结论要指认它时才建。
+- 独立评估者可以复跑(`auditRerun`,预设打开):评估者在工作区副本里跑产物自带的脚本核数字,原工作区只读;评估期间收下的产物被改动过,这份裁决不认。
+- **等级只给三档**:L2 自己检验、L3 独立评估、L4 人放行。L0–L2 的细分从没改变过谁来判,合并;旧账里的 L0、L1 读作自己检验,新写的收成 L2。
+
+- **宿主下界提到 DSH `0.2.0-rc.2`(`engines.dsh` 为 `>=0.2.0-rc.2`),更早的宿主不再支持。** ClearAI 用到的事件、服务方法、客户端插座和 `dsh plugin` 命令行在 0.2.x 上都没有破坏性变化,运行时代码不用改;改的是只在旧的 root 目录名册上才有意义的工具和文档。
+- **`install.sh`(开发形态)改走产品那条路**:构建出包,再 `dsh plugin add` 装进 profile。旧做法把预设拷进 `~/.dsh/.agent-presets`,名册不读它,装得上却选不到 ClearAI。
+- **`verify-deploy` 只验装出来的那个包**,预期的工具数与段数从部署的目录现算;`verify-clean-install` / `verify-lifecycle` 只认声明行;`e2e-run` 只在宿主还带着 `@deepseek-ai/dsh-invariants` 时才插那一行(`0.2.1-alpha.1` 已移除该包)。
+
+### Removed
+
+- 对手判断(`rival` / `split_by`)与「检验可能太弱」的提示:两道难题的 8 组运行里对手 0 次使用,提示连续出现也没改变做法。
+- 运行态卡上的本体提纲与计数:它从没改变过下一步。卡上只出现被判断引用到的本体项。
+- `clearai-dsh root-yaml` / `seed` / `unseed`:它们服务的是旧名册的目录扫描,0.2.x 宿主里没有这个名册。`doctor` 改为检查组合里有没有 `preset-clearai` 声明行。
+
+### Fixed
+
+- 判断几乎从不升格成事实:升格只看模型自己给步骤标的等级,而 L0–L2 允许自判,五个长程任务的 9 次运行里模型一律标 L1/L2,一条事实都没有。现在结案那次独立评估顺带逐条判还活着的判断,判为支持的升格(证据记为独立、等级按目标门槛),判为推翻的不升格并照常复核已有事实;不多派评估者。
+- 别的会话留下的事实被新证据推翻时不会问人:这一问只认本目标的判断编号。现在判断可带 `retests: "<事实 id>"`,复检被推翻就当场问人撤回还是维持,结论写回事实文件。
+- 复检旧事实时模型把判断写成「旧事实被推翻」,评估判支持后它升格成一条与旧事实相反的新事实,旧事实照旧「已确立」,也没人被问。`retests` 的说明和提示词现在要求 claim 照抄旧事实原来的说法。
+
 ## [0.4.0] — 2026-10-02
 
 这一版包括「少即是多」改造的全部六个阶段(见下面各段)、本体改成文件树,以及中英两种语言的完整支持。

package/README.md

@@ -22,8 +22,8 @@ ClearAI is an **ontology discovery and exploration platform**, built on two core
 
 ```bash
 # Install (npm package, prebuilt — no build step, no allowBuilds prompt)
-dsh plugin --profile web add clearai-dsh@0.4.0
-# or in the app: Plugins → Add plugin → clearai-dsh@0.4.0
+dsh plugin --profile web add clearai-dsh@0.5.0
+# or in the app: Plugins → Add plugin → clearai-dsh@0.5.0
 ```
 
 Restart `dsh web`, then pick **ClearAI** in the preset picker at the top of a new session. That is the whole setup. [Full install notes ↓](#install-and-use)
@@ -74,16 +74,16 @@ ClearAI does **not** claim recursive self-improvement. It provides the epistemic
 
 ## Install and use
 
-**Requirements:** DSH ≥ `0.1.7-alpha.1` — that generation introduced the composition declaration line this preset rides on. Verified against the host's `0.1.7-rc.2` and `0.2.0-rc.1`.
+**Requirements:** DSH ≥ `0.2.0-rc.2` (the current `latest`). The preset is registered through a composition declaration line, which is the only way the host's preset registry finds presets. Verified against the host's `0.2.0-rc.2`, and also checked against `0.2.1-alpha.1`. Hosts older than `0.2.0` are no longer supported.
 
 **Recommended — install it in the app, with the version pinned:**
 
-In the sidebar open **Plugins → Add plugin**, enter `clearai-dsh@0.4.0`, and install. That is DSH's own plugin manager: it hands what you type to pnpm, checks that the package declares a bundle and is compatible with this host, and applies it live. (The Settings page **插件列表 / Plugins** is the read-only inventory — installing happens on the sidebar's Plugins page.)
+In the sidebar open **Plugins → Add plugin**, enter `clearai-dsh@0.5.0`, and install. That is DSH's own plugin manager: it hands what you type to pnpm, checks that the package declares a bundle and is compatible with this host, and applies it live. (The Settings page **插件列表 / Plugins** is the read-only inventory — installing happens on the sidebar's Plugins page.)
 
 **Or from a terminal — the same install:**
 
 ```bash
-dsh plugin --profile web add clearai-dsh@0.4.0
+dsh plugin --profile web add clearai-dsh@0.5.0
 ```
 
 This installs the prebuilt package from the npm registry. Nothing is compiled on your machine, so there is no `allowBuilds` grant to approve — the plugin is ready the moment the command returns.
@@ -116,7 +116,7 @@ dsh plugin --profile web add github:Clearailhc/clearai-dsh
 
 Git fetches source rather than build artifacts, so pnpm ≥10 will refuse to run the `prepare` script until you add an `allowBuilds` entry to the profile's `pnpm-workspace.yaml`. That grant means *permission for this package's code to execute on your machine at install time* — grant it only if you have read the source, and pin a commit. If you just want to use ClearAI, use the npm install above.
 
-The installer's output follows your system language (`--lang zh|en` overrides it, `doctor` / `seed` / `unseed` take the same flag). Its only runtime dependency is `zod`; the graph stack is bundled into the client half at build time.
+The installer's output follows your system language (`--lang zh|en` overrides it, and `doctor` takes the same flag). Its only runtime dependency is `zod`; the graph stack is bundled into the client half at build time.
 
 Restart `dsh web` afterwards (`npx @deepseek-ai/dsh web`), then **create a session and switch to the `ClearAI` mode in the picker at the top**:
 

package/README.zh-CN.md

@@ -22,8 +22,8 @@ ClearAI 是一个**本体发现与探索平台**，核心由两个概念支撑
 
 ```bash
 # 安装（npm 包，预构建——无需构建步骤，不会触发 allowBuilds 授权）
-dsh plugin --profile web add clearai-dsh@0.4.0
-# 或在应用里：侧栏「插件」→ 添加插件 → clearai-dsh@0.4.0
+dsh plugin --profile web add clearai-dsh@0.5.0
+# 或在应用里：侧栏「插件」→ 添加插件 → clearai-dsh@0.5.0
 ```
 
 重启 `dsh web`，在新建会话顶部的模式选择器里选 **ClearAI** 即可。这就是全部步骤。[完整安装说明 ↓](#安装与使用)
@@ -74,16 +74,16 @@ ClearAI **不**声称递归自我改进。它提供的是自我改进系统所
 
 ## 安装与使用
 
-**要求：** DSH ≥ `0.1.7-alpha.1`——预设靠的是那一代引入的组合声明行。已在宿主的 `0.1.7-rc.2` 与 `0.2.0-rc.1` 上验过。
+**要求：** DSH ≥ `0.2.0-rc.2`（当前的 `latest`）。预设靠组合里的声明行注册，宿主的预设名册只认这一种。已在宿主的 `0.2.0-rc.2` 上验过，`0.2.1-alpha.1` 也核过。`0.2.0` 以前的宿主不再支持。
 
 **推荐——在应用里装，并把版本钉住：**
 
-侧栏打开**「插件」→ 添加插件**，填 `clearai-dsh@0.4.0`，安装。这就是 DSH 自己的插件管理器：它把你填的东西交给 pnpm，校验这个包声明了组合包、与当前宿主兼容，然后当场生效。（设置里的**插件列表**是**只读清单**；安装入口在侧栏那个「插件」页。）
+侧栏打开**「插件」→ 添加插件**，填 `clearai-dsh@0.5.0`，安装。这就是 DSH 自己的插件管理器：它把你填的东西交给 pnpm，校验这个包声明了组合包、与当前宿主兼容，然后当场生效。（设置里的**插件列表**是**只读清单**；安装入口在侧栏那个「插件」页。）
 
 **或者开终端——同一次安装：**
 
 ```bash
-dsh plugin --profile web add clearai-dsh@0.4.0
+dsh plugin --profile web add clearai-dsh@0.5.0
 ```
 
 从 npm registry 装预构建产物。本机不跑任何编译，因此不需要批准 `allowBuilds` 授权——命令返回时插件就已经可用。
@@ -116,7 +116,7 @@ dsh plugin --profile web add github:Clearailhc/clearai-dsh
 
 Git 拉的是源码而不是构建产物，所以 pnpm ≥10 会拒绝运行 `prepare` 脚本，直到你在该 profile 的 `pnpm-workspace.yaml` 里加上 `allowBuilds` 条目。那条授权的含义是**允许该包代码在安装时于你机器上执行**——只在你读过源码后再授权，并且固定 commit。如果你只是想用 ClearAI，请用上面的 npm 安装。
 
-安装侧的输出**跟系统语言走**（`--lang zh|en` 可覆盖；`doctor` / `seed` / `unseed` 同样认这个开关）。运行时依赖只有 `zod`——图谱那套栈在构建期就打进客户端了。
+安装侧的输出**跟系统语言走**（`--lang zh|en` 可覆盖；`doctor` 同样认这个开关）。运行时依赖只有 `zod`——图谱那套栈在构建期就打进客户端了。
 
 
 装完重启 `dsh web`（`npx @deepseek-ai/dsh web`），然后**新建会话，在顶部的模式选择器里切换到 `ClearAI`**：

package/bin/clearai.mjs

@@ -5,27 +5,23 @@
  * 分包的三样东西落在三个平面:
  *   · 宿主半(`lib/host.js`)—— 由包自带的 `cordis.patch.yml` 在 profile 层插一行,`dsh plugin add` 自动生效;
  *   · 浏览器半(`lib/client.js`)—— 同一份清单里的 `dsh.client`,浏览器 Loader 自己扫;
- *   · agent 预设(`presets/clearai/`)—— **名册(roster)只从 root 目录扫**,包没法直接声明,
- *     所以要么把 root 指进包里,要么把它播种到用户根。这一件就是本工具存在的理由。
+ *   · agent 预设(`presets/clearai/clearai.patch.yml`)—— 一条 `- id: preset-clearai` 声明行,
+ *     经清单的 `dsh.bundle.patch` 挂进组合;名册(`agent-preset-registry`)读的就是它。
  *
- * 处理方式(前四个默认只**打印**不改;默认动词仍是 `doctor` —— 一个安装侧工具不该在
- * 你没说要装的时候动你的部署):
- *   doctor      看现状:包在哪、预设在哪、名册能不能看见它、探测到的 dsh / profile 是什么
+ * 宿主 ≥0.2.0 的名册**不扫目录、不收预设路径**(`@deepseek-ai/dsh-agent-preset-registry` 的 README:
+ * "the registry neither scans directories nor accepts preset paths"),所以旧的 root 目录那几个动词
+ * (`root-yaml` / `seed` / `unseed`)已经删掉:它们写出去的东西在这些宿主上没人读。
+ *
+ * 两个动词(默认是 `doctor` —— 一个安装侧工具不该在你没说要装的时候动你的部署):
+ *   doctor      看现状:包在哪、预设在哪、组合里有没有宿主行与预设声明行、探测到的 dsh / profile 是什么
  *   install     把包装进 profile —— `dsh plugin --profile <p> add <spec>` 的一层**前置解析**
  *               (读者不必知道 profile 叫什么、CLI 从哪来、包名怎么写),装完给读数与下一步
- *   root-yaml   打印**可以直接粘进 profile 的 `cordis.patch.yml`** 的那一行(路径已算成绝对路径)
- *   seed        把预设**播种**到用户根 `~/.dsh/.agent-presets/<id>`(带哈希记账:改过的不覆盖)
- *   unseed      撤掉播种:只删「我们播的、且没被改过」的文件;用户 fork 的 id 一律不碰
- *
- * 纪律(与内核播种模板技能同一条):**改过的东西不覆盖,只如实报漂移**;删东西之前先看哈希。
- * 不改用户的 `cordis.patch.yml` —— 那一行的内容由人自己粘(工具只负责把绝对路径算对)。
  */
 
-import { createHash } from 'node:crypto'
 import { spawnSync } from 'node:child_process'
-import { cpSync, existsSync, mkdirSync, readFileSync, readdirSync, rmSync, statSync, writeFileSync } from 'node:fs'
+import { existsSync, readFileSync, statSync } from 'node:fs'
 import { homedir } from 'node:os'
-import { delimiter, dirname, join, relative, resolve } from 'node:path'
+import { delimiter, dirname, join, resolve } from 'node:path'
 import { fileURLToPath } from 'node:url'
 
 const HERE = dirname(fileURLToPath(import.meta.url))
@@ -33,7 +29,6 @@ const PKG_DIR = resolve(HERE, '..')
 const PRESET_ID = 'clearai'
 const PRESET_SRC = join(PKG_DIR, 'presets', PRESET_ID)
 const DSH_HOME = process.env.DSH_HOME ?? join(homedir(), '.dsh')
-const LEDGER = join(DSH_HOME, '.clearai-dsh.json')
 
 const argv = process.argv.slice(2)
 const command = argv[0] ?? 'doctor'
@@ -67,13 +62,9 @@ const TEXT = {
 	zh: {
 		pkgDir: '包目录        ',
 		presetSrc: '预设源        ',
-		presetOk: '(agent.cordis.yml ✓)',
-		presetMissing: '(缺 agent.cordis.yml ✗)',
+		presetOk: '(clearai.patch.yml ✓)',
+		presetMissing: '(缺 clearai.patch.yml ✗)',
 		dshHome: 'DSH_HOME      ',
-		userRoot: '用户根        ',
-		rosterVisible: '✓ 名册看得见',
-		rosterAbsent: '✗ 还没有(用 seed 播种,或把 root-yaml 那一行粘进 profile)',
-		shadowCopy: '  ⚠️ 影子副本     {path} 与包里的那份**不一致**,而它被包的 root 遮住、永远不会被读到(自检却会优先读它)。删掉它,或用它来承载你自己的改动并换一个 id。',
 		profileRow: 'profile       ',
 		profileAbsent: '(不存在:先跑一次 dsh --profile {profile})',
 		composeRow: '组合          ',
@@ -81,13 +72,10 @@ const TEXT = {
 		hostRow: '宿主行        ',
 		hostIn: '在组合里 ✓',
 		hostOut: '不在组合里(先 dsh plugin --profile {profile} add <本包>)',
-		rosterRootRow: '名册 root     ',
-		rosterNoRoots: '组合里没看到 roots(这个部署可能不挂名册)',
-		rosterExpr: '{expr}(dump 打的是表达式原文,求值在装载时)',
-		seedLedgerRow: '播种记账      ',
-		ledgerEmpty: '(空)',
-		ledgerCount: '{count} 个条目 · {path}',
-		doctorHint: '\n提示:doctor 只读;它不会替你改 profile,也不会替你播种。',
+		presetRow: '预设声明行    ',
+		presetIn: '在组合里 ✓(preset-clearai)',
+		presetOut: '不在组合里:名册里不会有 ClearAI(这个 profile 可能没挂 agent-preset-registry,或宿主早于 0.2.0)',
+		doctorHint: '\n提示:doctor 只读;它不会替你改 profile。',
 		installHeader: '【安装】{name}@{version}',
 		installWhat: '装什么      ',
 		installWhere: '装到哪      ',
@@ -104,30 +92,15 @@ const TEXT = {
 		hostAbsent: '**不在组合里** —— 装是装上了,但组合里没看到它(用 doctor 查)',
 		nextStep: '\n  下一步      重启 dsh web(两半都在进程里按模块 URL 缓存,只刷新浏览器不够),然后在预设选择器里选 ClearAI。',
 		uninstall: '  卸载        dsh plugin --profile {profile} remove {name}(同样可以冠 npx)',
-		rootYamlComment1: '# ClearAI 预设的 root(由 clearai-dsh 的 bin 打印,路径已算成绝对路径)',
-		rootYamlComment2: '    # ⚠️ 补丁层会**替换整份 config**:下面这些键必须与你部署里那份一致,否则会丢。',
-		rootYamlComment3: '    #    先 `dsh --profile <p> --dump-config | grep -A 20 agent-presets` 看一眼当前值再粘。',
-		noPresetInPackage: '✗ 包里没有预设:{path}',
-		seededTo: '  播种到        {dest}',
-		seedCounts: '  新增 {seeded} · 刷新 {refreshed}',
-		seedDrifted: ' · **你改过、没覆盖** {count}({list})',
-		seedNext: '  下一步:重启 dsh(或刷新页面)后,预设选择器里应出现「ClearAI」。',
-		noSeedLedger: '  没有播种记账:什么都不做(不知道哪些是我们播的,就不删)。',
-		unseedRemoved: '  删掉 {count} 个「我们播的、没被改过」的文件',
-		unseedKept: '  **保留** {count} 个你改过的文件(它们现在是你的):{list}',
 		unknownCommand: 'unknown command: {command}',
-		usage: '用法:clearai-dsh [doctor|install|root-yaml|seed|unseed] [--profile web] [--home <dir>] [--lang zh|en]\n     install 还可以:--dist <dir> | --tarball <tgz> | --spec <spec>',
+		usage: '用法:clearai-dsh [doctor|install|version] [--profile web] [--home <dir>] [--lang zh|en]\n     install 还可以:--dist <dir> | --tarball <tgz> | --spec <spec>',
 	},
 	en: {
 		pkgDir: 'package       ',
 		presetSrc: 'preset src    ',
-		presetOk: '(agent.cordis.yml ✓)',
-		presetMissing: '(no agent.cordis.yml ✗)',
+		presetOk: '(clearai.patch.yml ✓)',
+		presetMissing: '(no clearai.patch.yml ✗)',
 		dshHome: 'DSH_HOME      ',
-		userRoot: 'user root     ',
-		rosterVisible: '✓ visible to the roster',
-		rosterAbsent: '✗ not there yet (seed it, or paste the root-yaml line into the profile)',
-		shadowCopy: '  ⚠️ shadow copy  {path} differs from the copy in the package, and the package root shadows it — it will never be read (while the self-check reads it first). Delete it, or keep your changes there under a different id.',
 		profileRow: 'profile       ',
 		profileAbsent: '(does not exist yet: run dsh --profile {profile} once)',
 		composeRow: 'composition   ',
@@ -135,13 +108,10 @@ const TEXT = {
 		hostRow: 'host row      ',
 		hostIn: 'is in the composition ✓',
 		hostOut: 'is NOT in the composition (run dsh plugin --profile {profile} add <this package>)',
-		rosterRootRow: 'roster root   ',
-		rosterNoRoots: 'no roots in the composition (this deployment may not mount the roster)',
-		rosterExpr: '{expr}(the dump prints the expression verbatim; it is evaluated at load time)',
-		seedLedgerRow: 'seed ledger   ',
-		ledgerEmpty: '(empty)',
-		ledgerCount: '{count} entries · {path}',
-		doctorHint: '\ndoctor only reads. It will not change your profile and will not seed anything for you.',
+		presetRow: 'preset row    ',
+		presetIn: 'is in the composition ✓ (preset-clearai)',
+		presetOut: 'is NOT in the composition: the roster will not offer ClearAI (this profile may not mount agent-preset-registry, or the host predates 0.2.0)',
+		doctorHint: '\ndoctor only reads. It will not change your profile.',
 		installHeader: '[install] {name}@{version}',
 		installWhat: 'what        ',
 		installWhere: 'where       ',
@@ -158,73 +128,18 @@ const TEXT = {
 		hostAbsent: '**NOT in the composition** — it installed, but the composition does not show it (run doctor)',
 		nextStep: '\n  next        restart dsh web (both halves are cached in the running process by module URL, so a browser refresh is not enough), then pick ClearAI in the preset picker.',
 		uninstall: '  uninstall   dsh plugin --profile {profile} remove {name}(npx works too)',
-		rootYamlComment1: '# ClearAI preset root (printed by clearai-dsh; the path is already absolute)',
-		rootYamlComment2: '    # ⚠️ a patch layer **replaces the whole config**: these keys must match the ones in your deployment, or they are lost.',
-		rootYamlComment3: '    #    Run `dsh --profile <p> --dump-config | grep -A 20 agent-presets` first and paste over the current values.',
-		noPresetInPackage: '✗ no preset in the package: {path}',
-		seededTo: '  seeded to   {dest}',
-		seedCounts: '  added {seeded} · refreshed {refreshed}',
-		seedDrifted: ' · **kept your edits** {count}({list})',
-		seedNext: '  next: restart dsh (or refresh the page) — ClearAI should now appear in the preset picker.',
-		noSeedLedger: '  nothing was seeded: doing nothing (without the ledger there is no way to tell which files are ours, so nothing is deleted).',
-		unseedRemoved: '  removed {count} file(s) we seeded and you did not edit',
-		unseedKept: '  **kept** {count} file(s) you edited (they are yours now): {list}',
 		unknownCommand: 'unknown command: {command}',
-		usage: 'usage: clearai-dsh [doctor|install|root-yaml|seed|unseed] [--profile web] [--home <dir>] [--lang zh|en]\n       install also takes: --dist <dir> | --tarball <tgz> | --spec <spec>',
+		usage: 'usage: clearai-dsh [doctor|install|version] [--profile web] [--home <dir>] [--lang zh|en]\n       install also takes: --dist <dir> | --tarball <tgz> | --spec <spec>',
 	},
 }
 
-/** `{name}` 是占位符;需要拼计数与列表的句子写成函数(见 doctor / seed / unseed)。 */
+/** `{name}` 是占位符。 */
 function t(key, params) {
 	const text = TEXT[lang][key] ?? TEXT.zh[key]
 	if (text === undefined) throw new Error(`missing text: ${key}`)
 	return typeof text === 'function' ? text(params ?? {}) : text.replace(/\{(\w+)\}/g, (_, name) => String(params?.[name] ?? ''))
 }
 
-/** 递归列出文件(相对路径 → sha256 前 16 位)。只为记账,不为校验完整性。 */
-function hashTree(root) {
-	const out = {}
-	const walk = (dir) => {
-		let entries = []
-		try {
-			entries = readdirSync(dir, { withFileTypes: true })
-		} catch {
-			return
-		}
-		for (const entry of entries) {
-			const full = join(dir, entry.name)
-			if (entry.isDirectory()) walk(full)
-			else if (entry.isFile()) out[relative(root, full)] = createHash('sha256').update(readFileSync(full)).digest('hex').slice(0, 16)
-		}
-	}
-	walk(root)
-	return out
-}
-
-function readLedger() {
-	try {
-		const parsed = JSON.parse(readFileSync(LEDGER, 'utf8'))
-		return parsed !== null && typeof parsed === 'object' ? parsed : {}
-	} catch {
-		return {}
-	}
-}
-
-function writeLedger(next) {
-	try {
-		mkdirSync(dirname(LEDGER), { recursive: true })
-		writeFileSync(LEDGER, `${JSON.stringify(next, null, 2)}\n`, 'utf8')
-		return true
-	} catch {
-		return false
-	}
-}
-
-/** 名册会扫哪些 root —— 与 `@deepseek-ai/dsh-agent-presets` 的规则同形(自带 → 配置 → 用户根)。 */
-function rosterRoots() {
-	return [join(DSH_HOME, '.agent-presets')]
-}
-
 /**
  * 在 PATH 上找一个可执行文件,找到就返回全路径。
  *
@@ -266,35 +181,12 @@ function composeQuery(profileName, env) {
 function doctor() {
 	const rows = []
 	rows.push(`${t('pkgDir')}${PKG_DIR}`)
-	rows.push(`${t('presetSrc')}${PRESET_SRC}${existsSync(join(PRESET_SRC, 'agent.cordis.yml')) ? t('presetOk') : t('presetMissing')}`)
+	rows.push(`${t('presetSrc')}${PRESET_SRC}${existsSync(join(PRESET_SRC, 'clearai.patch.yml')) ? t('presetOk') : t('presetMissing')}`)
 	rows.push(`${t('dshHome')}${DSH_HOME}`)
-	const roots = rosterRoots()
-	for (const root of roots) {
-		const dir = join(root, PRESET_ID)
-		const present = existsSync(join(dir, 'agent.cordis.yml'))
-		rows.push(`${t('userRoot')}${dir} ${present ? t('rosterVisible') : t('rosterAbsent')}`)
-		/**
-		 * 用户根里那份**会不会被包的 root 遮住**。
-		 *
-		 * 名册按 root 顺序先到先得(自带 root → 配置 root → 用户根),所以包一装,用户根里
-		 * 同 id 的那份副本就**再也读不到**——它会安静地烂在那里,还会把「部署出去的那份」的
-		 * 自检引到你手改过的旧副本上。判据只取事实:两份 `preset.yml` 的字节是否一致。
-		 */
-		const mine = join(PRESET_SRC, 'preset.yml')
-		const theirs = join(dir, 'preset.yml')
-		if (!present || !existsSync(mine) || !existsSync(theirs)) continue
-		let same = false
-		try {
-			same = readFileSync(mine).equals(readFileSync(theirs))
-		} catch {
-			same = false
-		}
-		if (!same) rows.push(t('shadowCopy', { path: theirs }))
-	}
 	const profileDir = join(DSH_HOME, 'profiles', profile)
 	rows.push(`${t('profileRow')}${profileDir}${existsSync(profileDir) ? '' : t('profileAbsent', { profile })}`)
 	/**
-	 * 宿主行与名册 root **问组合**,不问某个文件(2026-09-11 修:装出来的形态下这两行来自
+	 * 宿主行与预设声明行**问组合**,不问某个文件(2026-09-11 修:装出来的形态下这两行来自
 	 * **包的补丁层**,不在 profile 的 cordis.patch.yml 里 —— 读文件的写法会误报「还没挂」)。
 	 * 事实在组合里;问不到就如实说问不到,不猜。
 	 */
@@ -304,12 +196,9 @@ function doctor() {
 	} else {
 		const rowIds = [...composed.matchAll(/^- id: (\S+)$/gm)].map((match) => match[1])
 		rows.push(`${t('hostRow')}${rowIds.includes('clearai-host') ? t('hostIn') : t('hostOut', { profile })}`)
-		const roster = /- id: agent-presets[\s\S]{0,600}?\n\s*- path:[^\n]*/.exec(composed)
-		const expr = roster === null ? null : roster[0].split('\n').pop().trim()
-		rows.push(`${t('rosterRootRow')}${expr === null ? t('rosterNoRoots') : t('rosterExpr', { expr })}`)
+		/** 名册读的是组合里的 `preset-clearai` 声明行(由包的 `dsh.bundle.patch` 挂进来),不是某个目录。 */
+		rows.push(`${t('presetRow')}${rowIds.includes('preset-clearai') ? t('presetIn') : t('presetOut')}`)
 	}
-	const ledger = readLedger()
-	rows.push(`${t('seedLedgerRow')}${Object.keys(ledger).length === 0 ? t('ledgerEmpty') : t('ledgerCount', { count: Object.keys(ledger).length, path: LEDGER })}`)
 	for (const row of rows) console.log(`  ${row}`)
 	console.log(t('doctorHint'))
 }
@@ -386,99 +275,8 @@ function install() {
 	console.log(t('uninstall', { profile, name: manifest.name }))
 }
 
-function rootYaml() {
-	const line = [
-		t('rootYamlComment1'),
-		'- id: \'@deepseek-ai/dsh-agent-presets\'',
-		'  config:',
-		t('rootYamlComment2'),
-		t('rootYamlComment3'),
-		'    default: standard',
-		'    includeShippedRoot: true',
-		'    includeUserRoot: true',
-		'    roots:',
-		`      - path: '${join(PKG_DIR, 'presets')}'`,
-		'        trust: system',
-	].join('\n')
-	console.log(line)
-}
-
-function seed() {
-	if (!existsSync(join(PRESET_SRC, 'agent.cordis.yml'))) {
-		console.error(t('noPresetInPackage', { path: PRESET_SRC }))
-		process.exit(1)
-	}
-	const root = value('root', rosterRoots()[0])
-	const dest = join(root, PRESET_ID)
-	const next = hashTree(PRESET_SRC)
-	const before = readLedger()
-	const previous = before[PRESET_ID]?.files ?? {}
-	const seeded = []
-	const refreshed = []
-	const drifted = []
-	for (const [rel, hash] of Object.entries(next)) {
-		const target = join(dest, rel)
-		const onDisk = existsSync(target) ? createHash('sha256').update(readFileSync(target)).digest('hex').slice(0, 16) : null
-		if (onDisk === null) {
-			mkdirSync(dirname(target), { recursive: true })
-			cpSync(join(PRESET_SRC, rel), target)
-			seeded.push(rel)
-			continue
-		}
-		if (onDisk === hash) continue
-		// 改过的不覆盖:与内核播种模板技能同一条纪律
-		if (previous[rel] !== undefined && onDisk !== previous[rel]) {
-			drifted.push(rel)
-			continue
-		}
-		cpSync(join(PRESET_SRC, rel), target)
-		refreshed.push(rel)
-	}
-	before[PRESET_ID] = { files: next, at: new Date().toISOString(), dest }
-	writeLedger(before)
-	console.log(t('seededTo', { dest }))
-	console.log(`${t('seedCounts', { seeded: seeded.length, refreshed: refreshed.length })}${drifted.length === 0 ? '' : t('seedDrifted', { count: drifted.length, list: drifted.slice(0, 5).join(lang === 'zh' ? '、' : ', ') })}`)
-	console.log(t('seedNext'))
-}
-
-function unseed() {
-	const ledger = readLedger()
-	const entry = ledger[PRESET_ID]
-	if (entry === undefined) {
-		console.log(t('noSeedLedger'))
-		return
-	}
-	const dest = entry.dest ?? join(rosterRoots()[0], PRESET_ID)
-	const removed = []
-	const kept = []
-	for (const [rel, hash] of Object.entries(entry.files ?? {})) {
-		const target = join(dest, rel)
-		if (!existsSync(target)) continue
-		const onDisk = createHash('sha256').update(readFileSync(target)).digest('hex').slice(0, 16)
-		if (onDisk === hash) {
-			rmSync(target)
-			removed.push(rel)
-		} else {
-			kept.push(rel)
-		}
-	}
-	delete ledger[PRESET_ID]
-	writeLedger(ledger)
-	console.log(t('unseedRemoved', { count: removed.length }))
-	if (kept.length > 0) console.log(t('unseedKept', { count: kept.length, list: kept.slice(0, 5).join(lang === 'zh' ? '、' : ', ') }))
-	// 空目录收掉(只收我们知道的那些;非空绝不动)
-	try {
-		if (existsSync(dest) && readdirSync(dest).length === 0) rmSync(dest, { recursive: true })
-	} catch {
-		/* 收不掉就算了,不影响判定 */
-	}
-}
-
 if (command === 'doctor') doctor()
 else if (command === 'install') install()
-else if (command === 'root-yaml') rootYaml()
-else if (command === 'seed') seed()
-else if (command === 'unseed') unseed()
 else if (command === 'version' || flag('version')) console.log(JSON.parse(readFileSync(join(PKG_DIR, 'package.json'), 'utf8')).version)
 else {
 	console.error(`${t('unknownCommand', { command })}\n${t('usage')}`)

package/cordis.patch.yml

@@ -12,7 +12,7 @@
     - id: clearai-host
       name: 'clearai-dsh'
 
-# ── 预设不在这里声明 root:宿主 ≥0.1.7-alpha.1 起名册不再扫目录 ────────────────
+# ── 预设不在这里声明 root:宿主的名册不扫目录(本包要求宿主 ≥0.2.0-rc.2)──────────
 #
 # 0.2.3 及以前走的是「root 目录」机制:宿主自带 `- id: agent-presets` 行,本文件覆盖它的
 # config,把 root 指到包内 presets/,归属落在 system 信任层。
@@ -28,5 +28,5 @@
 # 依据与实测:`lab/release/0.2.3-release-blocker.md`、`lab/adapter/0.2.4-design.md`。
 #
 # 预设本体现在由 `presets/clearai/clearai.patch.yml` 承载(构建期由 preset/agent.cordis.yml
-# 生成,经 package.json 的 `dsh.bundle.patch` 数组挂上)。想改预设的人仍有两条路:
-# `clearai-dsh seed`(播种到用户根,可改可回滚)或复制成新 id。
+# 生成,经 package.json 的 `dsh.bundle.patch` 数组挂上)。想改预设的人:把声明行复制一份、
+# 换一个 id,作为自己的 bundle patch 装(`clearai-dsh seed` 那条播种到用户根的路已随旧名册删掉)。

package/lib/client.js

@@ -185,8 +185,8 @@ window.__ModuleLoader__.load({
 
 		/** 语言命名空间。表按**源文**索引:键就是中文原文,所以漏翻译一条只会退回中文,不会把 key 显示给人。 */
 		const LOCALE_NS = 'clearai'
-		const LOCALE_ZH = {":":":"," · 改过 ":" · 改过 "," · 最近一次修订的独立裁决:":" · 最近一次修订的独立裁决:"," 次":" 次"," 次评估者":" 次评估者"," 步":" 步"," 步)":" 步)"," 看核验记录":" 看核验记录"," 看核验过程":" 看核验过程"," 轮":" 轮","(点一下开右栏「世界树」)":"(点一下开右栏「世界树」)","(点一下开右栏「世界树」看拓扑)":"(点一下开右栏「世界树」看拓扑)","(缺)":"(缺)","clearai 面板:右栏页签类型注册失败 ":"clearai 面板:右栏页签类型注册失败 ","clearai-loop: 席位跟着会话预设进出":"clearai-loop: 席位跟着会话预设进出","」相关的":"」相关的","一句话":"一句话","一步检验":"一步检验","上位":"上位","下一步:找一步去检验它":"下一步:找一步去检验它","下一步:换个办法再检验":"下一步:换个办法再检验","下一步:独立核验通过 → 已验证":"下一步:独立核验通过 → 已验证","下一步:结案时写进长期知识":"下一步:结案时写进长期知识","下位":"下位","不确定":"不确定","世界树":"世界树","主语":"主语","事实":"事实","产物":"产物","人撤回":"人撤回","人放行":"人放行","代码":"代码","依据":"依据","值的形态":"值的形态","做什么":"做什么","入本体":"入本体","全屏":"全屏","全文在 ":"全文在 ","公式":"公式","关系":"关系","关闭":"关闭","写进长期知识":"写进长期知识","出自":"出自","切回历史的世界树(计划都还在,文档也归档在 clear/goals/plans/)":"切回历史的世界树(计划都还在,文档也归档在 clear/goals/plans/)","判据":"判据","判据:":"判据:","判断":"判断","原生预览打开它":"原生预览打开它","原生预览打开它(计划声明的产物)":"原生预览打开它(计划声明的产物)","取不到":"取不到","取值":"取值","只看与「":"只看与「","只看相关":"只看相关","可信度怎么变的":"可信度怎么变的","可复算":"可复算","命题":"命题","图组件不可用":"图组件不可用","在图上找它":"在图上找它","在检验这一站被推翻,不再往下走。":"在检验这一站被推翻,不再往下走。","实体图":"实体图","实体图还空着。":"实体图还空着。","实例":"实例","审批记录":"审批记录","已交付":"已交付","已作废":"已作废","已推翻":"已推翻","已提出":"已提出","已收尾 · 存档可看":"已收尾 · 存档可看","已替换":"已替换","已被替换,不再往下走。":"已被替换,不再往下走。","已验证":"已验证","引用":"引用","引用已有材料":"引用已有材料","待推进":"待推进","待核验":"待核验","打开「本体」那一格并展开这条结论":"打开「本体」那一格并展开这条结论","打开目标文档(原生预览)":"打开目标文档(原生预览)","打开计划文档(原生预览)":"打开计划文档(原生预览)","推翻":"推翻","提出":"提出","支持":"支持","收起":"收起","收起详情":"收起详情","放行":"放行","数值":"数值","暂无计划。建立后此处显示计划的步骤与闸门。":"暂无计划。建立后此处显示计划的步骤与闸门。","有 ":"有 ","有矛盾":"有矛盾","未声明":"未声明","本体":"本体","本体图":"本体图","查看此步骤的证据":"查看此步骤的证据","查看步骤详情":"查看步骤详情","查看评估者":"查看评估者","检验":"检验","检验结果:":"检验结果:","概念":"概念","正在取…":"正在取…","正在裁决":"正在裁决","步 ":"步 ","清除":"清除","点击查看详情":"点击查看详情","点开看进度和来历":"点开看进度和来历","状态":"状态","独立核验":"独立核验","独立核验:":"独立核验:","用到它的结论":"用到它的结论","盘上没有这个文件":"盘上没有这个文件","目标":"目标","相关":"相关","看核验":"看核验","看记录":"看记录","矛盾":"矛盾","第 n 步":"第 n 步","算错":"算错","类型":"类型","结论":"结论","结论照常可读。":"结论照常可读。","背景(不参与判定):":"背景(不参与判定):","自己推了一遍":"自己推了一遍","自己检验":"自己检验","范围":"范围","范围:":"范围:","补充":"补充","观测":"观测","计划":"计划","计划 ":"计划 ","计划受阻,等人处置":"计划受阻,等人处置","计划已交付 ":"计划已交付 ","计划已收尾(":"计划已收尾(","计划文档":"计划文档","计划的步骤与闸门":"计划的步骤与闸门","证据":"证据","评 ":"评 ","评估":"评估","评估卡":"评估卡","起过 ":"起过 ","还没有内容。发第一条消息后,这里显示本体图和结论。":"还没有内容。发第一条消息后,这里显示本体图和结论。","还没有立目标":"还没有立目标","还没有结论。模型立下判断之后,这里一行一条。":"还没有结论。模型立下判断之后,这里一行一条。","这一项已经不在当前的词里了(可能刚被废止)。":"这一项已经不在当前的词里了(可能刚被废止)。","这个词已废止,旧结论里的用法照样可读。":"这个词已废止,旧结论里的用法照样可读。","进度":"进度","退出全屏":"退出全屏","适配":"适配","释义":"释义","待处理":"待处理","待处理 ":"待处理 ","验证":"验证","验证中":"验证中","展开下一层":"展开下一层","收起下一层":"收起下一层","本体文件有 ":"本体文件有 "," 处问题(这些不进图)":" 处问题(这些不进图)"," 个断言的主语还没写成实体文件。":" 个断言的主语还没写成实体文件。","clear/ontology/entities/ 下的实体文件、写进长期知识的断言会出现在这里。":"clear/ontology/entities/ 下的实体文件、写进长期知识的断言会出现在这里。","本体图还空着。模型在 clear/ontology/concepts/ 和 relations/ 下写概念与关系文件之后,它们会长在这里。":"本体图还空着。模型在 clear/ontology/concepts/ 和 relations/ 下写概念与关系文件之后,它们会长在这里。"}
-		const LOCALE_EN = {":":": "," · 改过 ":" · changed "," · 最近一次修订的独立裁决:":" · latest revision decided by an independent verdict: "," 次":" times"," 次评估者":" evaluators"," 步":" steps"," 步)":" steps)"," 看核验记录":" Review record"," 看核验过程":" Watch the review"," 轮":" rounds","(点一下开右栏「世界树」)":" (click to open the World Tree)","(点一下开右栏「世界树」看拓扑)":" (click to see the steps in the World Tree)","(缺)":" (missing)","clearai 面板:右栏页签类型注册失败 ":"clearai panel: failed to register the right-sidebar tab type ","clearai-loop: 席位跟着会话预设进出":"clearai-loop: seats come and go with the session's preset","」相关的":"\"","一句话":"Statement","一步检验":"A test","上位":"Broader","下一步:找一步去检验它":"Next: test it in a step","下一步:换个办法再检验":"Next: test it another way","下一步:独立核验通过 → 已验证":"Next: independent check passes → Verified","下一步:结案时写进长期知识":"Next: written to long-term knowledge at close","下位":"Narrower","不确定":"Uncertain","世界树":"World Tree","主语":"Subject","事实":"Facts","产物":"Deliverables","人撤回":"Withdrawn by a person","人放行":"Released by a person","代码":"Code","依据":"Basis","值的形态":"Value form","做什么":"What it does","入本体":"In ontology","全屏":"Full screen","全文在 ":"Full text at ","公式":"Formula","关系":"Relation","关闭":"Close","写进长期知识":"Written to long-term knowledge","出自":"From","切回历史的世界树(计划都还在,文档也归档在 clear/goals/plans/)":"Switch back to an earlier World Tree (the plans are all still here, and their documents are archived under clear/goals/plans/)","判据":"Criterion","判据:":"Criterion: ","判断":"Judgment","原生预览打开它":"Open it in the native preview","原生预览打开它(计划声明的产物)":"Open it in the native preview (an artifact declared by the plan)","取不到":"Unavailable","取值":"Value","只看与「":"Only related to \"","只看相关":"Only related","可信度怎么变的":"How trust changed","可复算":"Reproducible","命题":"Propositions","图组件不可用":"The graph component is unavailable","在图上找它":"Find it on the graph","在检验这一站被推翻,不再往下走。":"Refuted at the test stage; it goes no further.","实体图":"Entity graph","实体图还空着。":"The entity graph is empty. ","实例":"Instances","审批记录":"approval record","已交付":"delivered","已作废":"voided","已推翻":"Refuted","已提出":"proposed","已收尾 · 存档可看":"closed · archived and readable","已替换":"Replaced","已被替换,不再往下走。":"Replaced; it goes no further.","已验证":"Verified","引用":"Reference","引用已有材料":"Cites existing material","待推进":"to advance","待核验":"Awaiting check","打开「本体」那一格并展开这条结论":"Open the Ontology pane and expand this conclusion","打开目标文档(原生预览)":"Open the goal document (native preview)","打开计划文档(原生预览)":"Open the plan document (native preview)","推翻":"Refute","提出":"Proposed","支持":"Support","收起":"Collapse","收起详情":"Collapse details","放行":"release","数值":"Quantity","暂无计划。建立后此处显示计划的步骤与闸门。":"No plan yet. Once created, its steps show here.","有 ":"","有矛盾":"Conflicting","未声明":"not declared","本体":"Ontology","本体图":"Ontology graph","查看此步骤的证据":"see the evidence for this step","查看步骤详情":"view step details","查看评估者":"view evaluator","检验":"Test","检验结果:":"Result: ","概念":"concepts","正在取…":"Loading…","正在裁决":"Deciding","步 ":"step ","清除":"Clear","点击查看详情":"click for details","点开看进度和来历":"Open to see progress and history","状态":"Status","独立核验":"Independent check","独立核验:":"Independent check: ","用到它的结论":"Conclusions using it","盘上没有这个文件":"this file is not on disk","目标":"Goal","相关":"Related","看核验":"See check","看记录":"See record","矛盾":"Conflicts","第 n 步":"Step n","算错":"Wrong if","类型":"Type","结论":"CONCLUSIONS","结论照常可读。":"conclusions are still readable.","背景(不参与判定):":"Background (not part of the verdict): ","自己推了一遍":"Reasoned through alone","自己检验":"Self-tested","范围":"Scope","范围:":"Scope: ","补充":"More","观测":"Observation","计划":"Plan","计划 ":"Plan ","计划受阻,等人处置":"plan blocked, waiting for a person","计划已交付 ":"Plan delivered ","计划已收尾(":"Plan closed (","计划文档":"Plan document","计划的步骤与闸门":"Plan steps","证据":"Evidence","评 ":"E","评估":"Evaluation","评估卡":"evaluation card","起过 ":"ran ","还没有内容。发第一条消息后,这里显示本体图和结论。":"Nothing yet. After the first message, the graph and conclusions show here.","还没有立目标":"No goal yet","还没有结论。模型立下判断之后,这里一行一条。":"No conclusions yet. Each judgment gets one line here.","这一项已经不在当前的词里了(可能刚被废止)。":"This item is no longer in the vocabulary (it may have just been deprecated).","这个词已废止,旧结论里的用法照样可读。":"This term is deprecated; older conclusions still read fine.","进度":"Progress","退出全屏":"Exit full screen","适配":"Fit","释义":"Gloss","待处理":"To handle","待处理 ":"to handle ","验证":"Verification","验证中":"Testing","展开下一层":"Expand the next level","收起下一层":"Collapse the next level","本体文件有 ":"Ontology files have "," 处问题(这些不进图)":" problems (these stay off the graph)"," 个断言的主语还没写成实体文件。":" assertion subjects have no entity file yet. ","clear/ontology/entities/ 下的实体文件、写进长期知识的断言会出现在这里。":"Entity files under clear/ontology/entities/ and assertions in long-term knowledge appear here.","本体图还空着。模型在 clear/ontology/concepts/ 和 relations/ 下写概念与关系文件之后,它们会长在这里。":"The ontology graph is empty. Concepts and relations appear once the model writes files under clear/ontology/concepts/ and relations/."}
+		const LOCALE_ZH = {":":":"," · 改过 ":" · 改过 "," · 最近一次修订的独立裁决:":" · 最近一次修订的独立裁决:"," 次":" 次"," 次评估者":" 次评估者"," 步":" 步"," 步)":" 步)"," 看核验记录":" 看核验记录"," 看核验过程":" 看核验过程"," 轮":" 轮","(点一下开右栏「世界树」)":"(点一下开右栏「世界树」)","(点一下开右栏「世界树」看拓扑)":"(点一下开右栏「世界树」看拓扑)","(缺)":"(缺)","clearai 面板:右栏页签类型注册失败 ":"clearai 面板:右栏页签类型注册失败 ","clearai-loop: 席位跟着会话预设进出":"clearai-loop: 席位跟着会话预设进出","」相关的":"」相关的","一句话":"一句话","一步检验":"一步检验","上位":"上位","下一步:找一步去检验它":"下一步:找一步去检验它","下一步:换个办法再检验":"下一步:换个办法再检验","下一步:独立核验通过 → 已验证":"下一步:独立核验通过 → 已验证","下一步:结案时写进长期知识":"下一步:结案时写进长期知识","下位":"下位","不确定":"不确定","世界树":"世界树","主语":"主语","事实":"事实","产物":"产物","人撤回":"人撤回","人放行":"人放行","代码":"代码","依据":"依据","值的形态":"值的形态","做什么":"做什么","入本体":"入本体","全屏":"全屏","全文在 ":"全文在 ","公式":"公式","关系":"关系","关闭":"关闭","写进长期知识":"写进长期知识","出自":"出自","切回历史的世界树(计划都还在,文档也归档在 clear/goals/plans/)":"切回历史的世界树(计划都还在,文档也归档在 clear/goals/plans/)","判据":"判据","判据:":"判据:","判断":"判断","原生预览打开它":"原生预览打开它","原生预览打开它(计划声明的产物)":"原生预览打开它(计划声明的产物)","取不到":"取不到","取值":"取值","只看与「":"只看与「","只看相关":"只看相关","可信度怎么变的":"可信度怎么变的","可复算":"可复算","命题":"命题","图组件不可用":"图组件不可用","在图上找它":"在图上找它","在检验这一站被推翻,不再往下走。":"在检验这一站被推翻,不再往下走。","实体图":"实体图","实体图还空着。":"实体图还空着。","实例":"实例","审批记录":"审批记录","已交付":"已交付","已作废":"已作废","已推翻":"已推翻","已提出":"已提出","已收尾 · 存档可看":"已收尾 · 存档可看","已替换":"已替换","已被替换,不再往下走。":"已被替换,不再往下走。","已验证":"已验证","引用":"引用","引用已有材料":"引用已有材料","待推进":"待推进","待核验":"待核验","打开「本体」那一格并展开这条结论":"打开「本体」那一格并展开这条结论","打开目标文档(原生预览)":"打开目标文档(原生预览)","打开计划文档(原生预览)":"打开计划文档(原生预览)","推翻":"推翻","提出":"提出","支持":"支持","收起":"收起","收起详情":"收起详情","放行":"放行","未解释":"未解释","数值":"数值","暂无计划。建立后此处显示计划的步骤与闸门。":"暂无计划。建立后此处显示计划的步骤与闸门。","有 ":"有 ","有矛盾":"有矛盾","未声明":"未声明","本体":"本体","本体图":"本体图","查看此步骤的证据":"查看此步骤的证据","查看步骤详情":"查看步骤详情","查看评估者":"查看评估者","检验":"检验","检验结果:":"检验结果:","概念":"概念","正在取…":"正在取…","正在裁决":"正在裁决","步 ":"步 ","清除":"清除","点击查看详情":"点击查看详情","点开看进度和来历":"点开看进度和来历","状态":"状态","独立核验":"独立核验","独立核验:":"独立核验:","用到它的结论":"用到它的结论","盘上没有这个文件":"盘上没有这个文件","目标":"目标","相关":"相关","看核验":"看核验","看记录":"看记录","矛盾":"矛盾","第 n 步":"第 n 步","算错":"算错","类型":"类型","结论":"结论","结论照常可读。":"结论照常可读。","背景(不参与判定):":"背景(不参与判定):","自己推了一遍":"自己推了一遍","自己检验":"自己检验","范围":"范围","范围:":"范围:","补充":"补充","观测":"观测","计划":"计划","计划 ":"计划 ","计划受阻,等人处置":"计划受阻,等人处置","计划已交付 ":"计划已交付 ","计划已收尾(":"计划已收尾(","计划文档":"计划文档","计划的步骤与闸门":"计划的步骤与闸门","证据":"证据","评 ":"评 ","评估":"评估","评估卡":"评估卡","起过 ":"起过 ","还没有内容。发第一条消息后,这里显示本体图和结论。":"还没有内容。发第一条消息后,这里显示本体图和结论。","还没有立目标":"还没有立目标","还没有结论。模型立下判断之后,这里一行一条。":"还没有结论。模型立下判断之后,这里一行一条。","这一项已经不在当前的词里了(可能刚被废止)。":"这一项已经不在当前的词里了(可能刚被废止)。","这个词已废止,旧结论里的用法照样可读。":"这个词已废止,旧结论里的用法照样可读。","进度":"进度","退出全屏":"退出全屏","适配":"适配","释义":"释义","待处理":"待处理","待处理 ":"待处理 ","验证":"验证","验证中":"验证中","展开下一层":"展开下一层","收起下一层":"收起下一层","本体文件有 ":"本体文件有 "," 处问题(这些不进图)":" 处问题(这些不进图)"," 个断言的主语还没写成实体文件。":" 个断言的主语还没写成实体文件。","clear/ontology/entities/ 下的实体文件、写进长期知识的断言会出现在这里。":"clear/ontology/entities/ 下的实体文件、写进长期知识的断言会出现在这里。","本体图还空着。模型在 clear/ontology/concepts/ 和 relations/ 下写概念与关系文件之后,它们会长在这里。":"本体图还空着。模型在 clear/ontology/concepts/ 和 relations/ 下写概念与关系文件之后,它们会长在这里。"}
+		const LOCALE_EN = {":":": "," · 改过 ":" · changed "," · 最近一次修订的独立裁决:":" · latest revision decided by an independent verdict: "," 次":" times"," 次评估者":" evaluators"," 步":" steps"," 步)":" steps)"," 看核验记录":" Review record"," 看核验过程":" Watch the review"," 轮":" rounds","(点一下开右栏「世界树」)":" (click to open the World Tree)","(点一下开右栏「世界树」看拓扑)":" (click to see the steps in the World Tree)","(缺)":" (missing)","clearai 面板:右栏页签类型注册失败 ":"clearai panel: failed to register the right-sidebar tab type ","clearai-loop: 席位跟着会话预设进出":"clearai-loop: seats come and go with the session's preset","」相关的":"\"","一句话":"Statement","一步检验":"A test","上位":"Broader","下一步:找一步去检验它":"Next: test it in a step","下一步:换个办法再检验":"Next: test it another way","下一步:独立核验通过 → 已验证":"Next: independent check passes → Verified","下一步:结案时写进长期知识":"Next: written to long-term knowledge at close","下位":"Narrower","不确定":"Uncertain","世界树":"World Tree","主语":"Subject","事实":"Facts","产物":"Deliverables","人撤回":"Withdrawn by a person","人放行":"Released by a person","代码":"Code","依据":"Basis","值的形态":"Value form","做什么":"What it does","入本体":"In ontology","全屏":"Full screen","全文在 ":"Full text at ","公式":"Formula","关系":"Relation","关闭":"Close","写进长期知识":"Written to long-term knowledge","出自":"From","切回历史的世界树(计划都还在,文档也归档在 clear/goals/plans/)":"Switch back to an earlier World Tree (the plans are all still here, and their documents are archived under clear/goals/plans/)","判据":"Criterion","判据:":"Criterion: ","判断":"Judgment","原生预览打开它":"Open it in the native preview","原生预览打开它(计划声明的产物)":"Open it in the native preview (an artifact declared by the plan)","取不到":"Unavailable","取值":"Value","只看与「":"Only related to \"","只看相关":"Only related","可信度怎么变的":"How trust changed","可复算":"Reproducible","命题":"Propositions","图组件不可用":"The graph component is unavailable","在图上找它":"Find it on the graph","在检验这一站被推翻,不再往下走。":"Refuted at the test stage; it goes no further.","实体图":"Entity graph","实体图还空着。":"The entity graph is empty. ","实例":"Instances","审批记录":"approval record","已交付":"delivered","已作废":"voided","已推翻":"Refuted","已提出":"proposed","已收尾 · 存档可看":"closed · archived and readable","已替换":"Replaced","已被替换,不再往下走。":"Replaced; it goes no further.","已验证":"Verified","引用":"Reference","引用已有材料":"Cites existing material","待推进":"to advance","待核验":"Awaiting check","打开「本体」那一格并展开这条结论":"Open the Ontology pane and expand this conclusion","打开目标文档(原生预览)":"Open the goal document (native preview)","打开计划文档(原生预览)":"Open the plan document (native preview)","推翻":"Refute","提出":"Proposed","支持":"Support","收起":"Collapse","收起详情":"Collapse details","放行":"release","未解释":"Unexplained","数值":"Quantity","暂无计划。建立后此处显示计划的步骤与闸门。":"No plan yet. Once created, its steps show here.","有 ":"","有矛盾":"Conflicting","未声明":"not declared","本体":"Ontology","本体图":"Ontology graph","查看此步骤的证据":"see the evidence for this step","查看步骤详情":"view step details","查看评估者":"view evaluator","检验":"Test","检验结果:":"Result: ","概念":"concepts","正在取…":"Loading…","正在裁决":"Deciding","步 ":"step ","清除":"Clear","点击查看详情":"click for details","点开看进度和来历":"Open to see progress and history","状态":"Status","独立核验":"Independent check","独立核验:":"Independent check: ","用到它的结论":"Conclusions using it","盘上没有这个文件":"this file is not on disk","目标":"Goal","相关":"Related","看核验":"See check","看记录":"See record","矛盾":"Conflicts","第 n 步":"Step n","算错":"Wrong if","类型":"Type","结论":"CONCLUSIONS","结论照常可读。":"conclusions are still readable.","背景(不参与判定):":"Background (not part of the verdict): ","自己推了一遍":"Reasoned through alone","自己检验":"Self-tested","范围":"Scope","范围:":"Scope: ","补充":"More","观测":"Observation","计划":"Plan","计划 ":"Plan ","计划受阻,等人处置":"plan blocked, waiting for a person","计划已交付 ":"Plan delivered ","计划已收尾(":"Plan closed (","计划文档":"Plan document","计划的步骤与闸门":"Plan steps","证据":"Evidence","评 ":"E","评估":"Evaluation","评估卡":"evaluation card","起过 ":"ran ","还没有内容。发第一条消息后,这里显示本体图和结论。":"Nothing yet. After the first message, the graph and conclusions show here.","还没有立目标":"No goal yet","还没有结论。模型立下判断之后,这里一行一条。":"No conclusions yet. Each judgment gets one line here.","这一项已经不在当前的词里了(可能刚被废止)。":"This item is no longer in the vocabulary (it may have just been deprecated).","这个词已废止,旧结论里的用法照样可读。":"This term is deprecated; older conclusions still read fine.","进度":"Progress","退出全屏":"Exit full screen","适配":"Fit","释义":"Gloss","待处理":"To handle","待处理 ":"to handle ","验证":"Verification","验证中":"Testing","展开下一层":"Expand the next level","收起下一层":"Collapse the next level","本体文件有 ":"Ontology files have "," 处问题(这些不进图)":" problems (these stay off the graph)"," 个断言的主语还没写成实体文件。":" assertion subjects have no entity file yet. ","clear/ontology/entities/ 下的实体文件、写进长期知识的断言会出现在这里。":"Entity files under clear/ontology/entities/ and assertions in long-term knowledge appear here.","本体图还空着。模型在 clear/ontology/concepts/ 和 relations/ 下写概念与关系文件之后,它们会长在这里。":"The ontology graph is empty. Concepts and relations appear once the model writes files under clear/ontology/concepts/ and relations/."}
 
 		/**
 		 * 翻译函数:**由原生 locale 座位绑定**(`ctx.locale.bind`),不是我们自建的一套 i18n。
@@ -607,6 +607,7 @@ window.__ModuleLoader__.load({
 			evidence: t('证据'),
 			fact: t('事实'),
 			release: t('放行'),
+			anomaly: t('未解释'),
 		}))
 
 		/**
@@ -747,7 +748,7 @@ window.__ModuleLoader__.load({
 		 */
 		const TRUST_ORDER = ['credible', 'pending', 'testing', 'unclear', 'refuted', 'replaced']
 		const TRUST_WORD = lazyTable(() => ({ credible: t('已验证'), pending: t('待核验'), testing: t('验证中'), unclear: t('不确定'), refuted: t('已推翻'), replaced: t('已替换') }))
-		const LEVEL_WORD = lazyTable(() => ({ L0: t('自己推了一遍'), L1: t('引用已有材料'), L2: t('可复算'), L3: t('独立核验'), L4: t('人放行') }))
+		const LEVEL_WORD = lazyTable(() => ({ L0: t('自己检验'), L1: t('自己检验'), L2: t('自己检验'), L3: t('独立核验'), L4: t('人放行') }))
 		/** 进度轨的四站。顺序即一条结论要走的路。 */
 		const STATIONS = () => [t('判断'), t('检验'), t('已验证'), t('入本体')]
 		/** 还没走完时,下一步是什么(时间线末尾那条虚点)。 */