feat: import CSV and XLSX into Library with Office previews

.github/workflows/ci.yml

@@ -144,6 +144,9 @@ jobs:
         run: pnpm exec playwright install --with-deps chromium
       - working-directory: apps/web
         run: pnpm test:integration
+      - name: Library import, search and task reference tests
+        working-directory: apps/web
+        run: node --test ../../packages/plugins/library/tests/*.test.mjs
       - name: Enterprise desktop identity and collaboration tests
         working-directory: apps/web
         run: node --test ../../packages/providers/identity-enterprise/tests/*.test.mjs ../../packages/plugins/enterprise-collaboration/tests/*.test.mjs

apps/web/scripts/build-library.mjs

@@ -1,6 +1,19 @@
 import { build } from 'esbuild';
 import { writeFile } from 'node:fs/promises';
 import { fileURLToPath } from 'node:url';
+import { createRequire } from 'node:module';
 const entry = new URL('../../../packages/plugins/library/', import.meta.url);
 const result = await build({ entryPoints: [fileURLToPath(new URL('src/client.tsx', entry))], bundle: true, write: false, format: 'cjs', platform: 'browser', target: 'es2022', external: ['@deepseek-ai/dsh-client-ui-primitives', 'react', 'react/jsx-runtime'] });
 await writeFile(new URL('dist/client.browser.js', entry), `window.__ModuleLoader__.load({id: "workdsh-plugin-library", factory: function(require) { const module = {exports:{}};\n${result.outputFiles[0].text}\nreturn module.exports; }});\n`);
+
+// Reuse Office's public extraction module at build time. Library must not require
+// an unpublished Office plugin at runtime or activate Office in a personal Profile.
+const requireOffice = createRequire(new URL('package.json', entry));
+await build({
+  entryPoints: [fileURLToPath(new URL('src/index.ts', entry))],
+  bundle: true, format: 'esm', platform: 'node', target: 'node22', packages: 'external',
+  plugins: [{ name: 'office-tabular-only', setup(builder) {
+    builder.onResolve({ filter: /^workdsh-plugin-office\/tabular$/ }, () => ({ path: requireOffice.resolve('workdsh-plugin-office/tabular') }));
+  } }],
+  outfile: fileURLToPath(new URL('dist/index.js', entry)),
+});

apps/web/scripts/build-office.mjs

@@ -8,6 +8,8 @@ import { fileURLToPath } from "node:url";
 const root = new URL("../../../packages/plugins/office/", import.meta.url);
 const wordOnly = process.argv.includes("--word-only");
 await mkdir(new URL("dist/", root), { recursive: true });
+await build({ entryPoints: [fileURLToPath(new URL("src/tabular.js", root))], bundle: true, platform: "node", format: "esm", target: "node22", external: ["exceljs"], outfile: fileURLToPath(new URL("dist/tabular.js", root)) });
+await writeFile(new URL("dist/tabular.d.ts", root), await readFile(new URL("src/tabular.d.ts", root)));
 const font=await readFile(new URL("src/pdf/fonts/NotoSansSC.ttf",root));
 const fontSource=JSON.parse(await readFile(new URL("src/pdf/fonts/source.json",root),"utf8"));
 if(font.length!==fontSource.bytes||createHash("sha256").update(font).digest("hex")!==fontSource.sha256)throw Error("Bundled PDF font integrity mismatch");

apps/web/scripts/pack-library-release.mjs

@@ -17,7 +17,7 @@ delete manifest.devDependencies; delete manifest.scripts;
 manifest.workdshRelease = {
   scope: 'local-library-alpha',
   storage: '$DSH_HOME/library',
-  formats: ['md', 'markdown', 'txt', 'html', 'htm', 'pdf', 'docx', 'pptx'],
+  formats: ['md', 'markdown', 'txt', 'html', 'htm', 'pdf', 'docx', 'pptx', 'csv', 'xlsx'],
   preservesOriginals: true,
 };
 await writeFile(join(stage, 'package.json'), `${JSON.stringify(manifest, null, 2)}\n`);

apps/web/scripts/pack-office-release.mjs

@@ -21,7 +21,7 @@ const manifest = JSON.parse(await readFile(join(plugin, "package.json"), "utf8")
 delete manifest.devDependencies;
 delete manifest.scripts;
 manifest.description = "Independent Harness plugin: browser Office working copies and native PPTX templates (alpha preview)";
-manifest.workdshRelease = {scope: "office-browser-preview", wordOnly: false, formats: ["docx", "pptx", "xlsx", "pdf", "html"]};
+manifest.workdshRelease = {scope: "office-browser-preview", wordOnly: false, formats: ["docx", "pptx", "xlsx", "csv", "pdf", "html"]};
 await writeFile(join(stage, "package.json"), JSON.stringify(manifest, null, 2) + "\n");
 for (const file of manifest.files) await cp(join(plugin, file), join(stage, file), {recursive: true});
 execFileSync("corepack", ["pnpm", "pack", "--pack-destination", destination], {cwd: stage, stdio: "inherit"});

packages/contracts/src/library.ts

@@ -1,6 +1,6 @@
 import type { ActorContext, ResourceOwner } from './governance.js';
 
-export type LibraryAssetKind = 'markdown' | 'text' | 'pdf' | 'docx' | 'pptx' | 'html';
+export type LibraryAssetKind = 'markdown' | 'text' | 'pdf' | 'docx' | 'pptx' | 'html' | 'csv' | 'xlsx';
 export type LibraryConversionStatus = 'ready' | 'pending' | 'failed';
 export type LibraryAssetStatus = 'active' | 'disabled';
 
@@ -114,7 +114,7 @@ export interface LibraryDraft {
 
 export interface LibraryOriginalPreviewInput {
   readonly name: string;
-  readonly kind: Extract<LibraryAssetKind, 'docx' | 'pptx'>;
+  readonly kind: Extract<LibraryAssetKind, 'docx' | 'pptx' | 'csv' | 'xlsx'>;
   readonly bytes: Uint8Array;
 }
 

packages/plugins/library/README.md

@@ -17,7 +17,7 @@
 
 ## 验收与下一步
 
-完成对应 PLAN 任务及 [验收矩阵](../../../apps/web/docs/ACCEPTANCE.md) 场景，记录真实测试证据后才更新为完成。当前 Host 与 Client 入口可加载，数据保存到 `$DSH_HOME/library`。支持 MD、TXT、HTML、PDF、DOCX、PPTX；所有格式保留原件并生成独立检索文本。HTML 在无 Host 权限、无网络访问的沙箱中按原始样式和内联交互展示；PDF 可直接预览；安装 Office 插件时，Word 使用 `docx-preview`、PowerPoint 使用 `pptx-react-viewer` 打开原件，未安装时回退为派生检索文本和原件下载。Library 与 Office 通过公开可选预览注册表协作，不互相导入运行时内部实现。
+完成对应 PLAN 任务及 [验收矩阵](../../../apps/web/docs/ACCEPTANCE.md) 场景，记录真实测试证据后才更新为完成。当前 Host 与 Client 入口可加载，数据保存到 `$DSH_HOME/library`。支持 MD、TXT、HTML、PDF、DOCX、PPTX、CSV、XLSX；所有格式保留原件并生成独立检索文本。HTML 在无 Host 权限、无网络访问的沙箱中按原始样式和内联交互展示；PDF 可直接预览；安装 Office 插件时，Word 使用 `docx-preview`、PowerPoint 使用 `pptx-react-viewer` 打开原件，未安装时回退为派生检索文本和原件下载。Library 与 Office 通过公开可选预览注册表协作，不互相导入运行时内部实现。
 
 发布包可通过 `corepack pnpm probe:library` 验证独立安装、两次冷启动、卸载保留 `$DSH_HOME/library`、重新安装和资料恢复。页面支持搜索、最近、本地产物、完整目录树、目录内新建与导入、移动、添加到当前任务、停用与重新启用；停用会立即阻断历史任务引用的正文读取。搜索结果显示目录、类型、来源、修订、转换状态和页/段/幻灯片位置。
 
@@ -36,3 +36,9 @@
 ### 可选 Host 成员服务选择
 
 公开 registerLibraryConnection、registerLibraryTools、registerLibraryContextInjection 接受可选可信选择器，返回服务端 actor、LibraryService 和 verify。工具与上下文使用实际 Agent 选择，不从模型参数读取主体；操作前后核验，拒绝不匹配的会话上下文。默认 apply 不传选择器，保持个人行为。企业装配必须提供隔离存储和可撤权身份；这些入口本身不是文件执行沙箱。
+
+## 表格资料
+
+CSV/XLSX 可上传、按类型检索并引用到任务；原始文件按字节保留。Host 转换在构建时复用 Office 的公开 `workdsh-plugin-office/tabular` 入口的 CSV 解码/解析；表格提取模块随 Library 构建，运行时复用 ExcelJS，无需安装或激活 Office 插件即可导入和检索。CSV 支持 UTF-8、GB18030、UTF-16，以及逗号、分号和制表符；XLSX 提取工作表名称和单元格已有值，公式不重新计算，图表、图片和格式不进入检索文本。检索文本最多 1500 行/120 列/24000 单元格，XLSX 最多 20 个工作表；截取会显示警告，完整原件仍可下载。
+
+原件预览通过公开可选注册表复用 Office：CSV 为只读表格，XLSX 使用现有隔离编辑器，导出为副本，不覆盖资料库修订。未启用 Office 预览时显示检索文本。个人与企业 Desktop、Web 复用相同导入和预览实现，实际资料授权保持不变。

packages/plugins/library/package.json

@@ -76,12 +76,14 @@
     "react": "19.2.4",
     "workdsh-contracts": "workspace:*",
     "workdsh-ui": "workspace:*",
-    "@deepseek-ai/dsh-client-ui-primitives": "0.2.0-rc.2"
+    "@deepseek-ai/dsh-client-ui-primitives": "0.2.0-rc.2",
+    "workdsh-plugin-office": "workspace:*"
   },
   "dependencies": {
     "jszip": "3.10.1",
     "pdfjs-dist": "5.4.624",
-    "zod": "4.4.3"
+    "zod": "4.4.3",
+    "exceljs": "4.4.0"
   },
   "dsh": {
     "bundle": {