dsh-data-cleaning-agent 0.8.2 → 0.8.4

package/CHANGELOG.md

@@ -2,6 +2,21 @@
 
 本文件记录 `dsh-data-cleaning-agent` 的版本变更。格式遵循 [Keep a Changelog](https://keepachangelog.com/zh-CN/1.1.0/)，版本号遵循 [SemVer](https://semver.org/lang/zh-CN/)。
 
+## [0.8.4] - 2026-09-06
+
+### Fixed
+- 向导数据来源页在文本模式也接管图片粘贴，避免 Modlens 图片路径被当成企业名称；文本名单明确拒绝本地图片路径。
+- 右侧上传允许 PNG/JPEG/WebP，并通过同会话事件桥转交图片向导；接收失败保留工作台并说明原因。
+- 回填前检查原生输入机提交状态，回填后核对草稿；输入机未接收时保留向导并提示，不强制启用发送按钮。
+- 移除通过全局捕获点击和按钮文案推测导航的逻辑，改为订阅真实会话选择，目标打开失败时保留清洗首页、工作台和草稿。
+- 兼容无回调参数的 Store 通知；成功退出时只清理插件默认文案，保留用户自写草稿。
+- 新清洗会话在准备和选择成功后才激活，避免提前撤销旧会话归属。
+
+## [0.8.3] - 2026-09-06
+
+### Fixed
+- 「新会话」不再复用数据清洗入口的空白会话：入口从 `sessions.create({ workspaceId, sessionId })` 改为 `sessions.create({ cwd, sessionId })`，仅以工作区路径创建、不再挂进工作区 `sessionIds`，DSH「新会话」的空白会话复用因此跳过它，恢复默认 DSH 标准页；只有点击「数据清洗补全」才进入带标题的工作台初始页。
+
 ## [0.8.2] - 2026-09-06
 
 ### Fixed

package/README.en.md

@@ -2,7 +2,7 @@
 
 > A data cleaning & completion agent plugin for DeepSeek Harness: local CSV/XLSX/JSON engine plus optional Qichacha (QCC) MCP enterprise-data enrichment. Initiated and maintained by the Qichacha (QCC) team.
 >
-> Current source version / 当前源码版本: **0.8.2** (stable release)
+> Current source version / 当前源码版本: **0.8.4** (stable release)
 
 [![CI](https://github.com/duhu2000/dsh-data-cleaning-agent/actions/workflows/ci.yml/badge.svg)](https://github.com/duhu2000/dsh-data-cleaning-agent/actions/workflows/ci.yml)
 [![npm](https://img.shields.io/npm/v/dsh-data-cleaning-agent)](https://www.npmjs.com/package/dsh-data-cleaning-agent)

package/README.md

@@ -2,7 +2,7 @@
 
 > 在 DeepSeek Harness 中清洗、补全、画像企业名单数据的智能体插件：本地 CSV/XLSX/JSON 引擎 + 可选企查查（Qichacha/QCC）MCP 企业数据补全，由企查查（Qichacha/QCC）团队发起并维护。
 >
-> 当前源码版本 / Current source version: **0.8.2**（正式版本）
+> 当前源码版本 / Current source version: **0.8.4**（正式版本）
 
 [![CI](https://github.com/duhu2000/dsh-data-cleaning-agent/actions/workflows/ci.yml/badge.svg)](https://github.com/duhu2000/dsh-data-cleaning-agent/actions/workflows/ci.yml)
 [![npm](https://img.shields.io/npm/v/dsh-data-cleaning-agent)](https://www.npmjs.com/package/dsh-data-cleaning-agent)

package/lib/client.js

@@ -1105,9 +1105,8 @@ window.__ModuleLoader__.load({
     }
 
     /**
-     * 清洗补全是一个显式进入的业务子系统。切换会话、点击新会话或点击侧栏其它业务
-     * 入口时立即撤销其激活态，即使 DSH 复用了同一个空白 sessionId 也不会残留首页、
-     * 提示词、能力栏或工作台。点击本插件自己的顶部入口不会提前撤销。
+     * 以实际会话选择提交作为退出条件；点击其它入口但创建失败时，保留清洗首页、
+     * 面板和用户草稿。新版本入口始终创建独立会话，不再依赖 DOM 文案推测导航。
      */
     function installSessionOwnershipBridge(ctx) {
       let staleDraftObserver = null;
@@ -1124,26 +1123,9 @@ window.__ModuleLoader__.load({
       };
       const leaveSubsystem = () => {
         const sessionId = activeCleaningSessionId;
-        if (sessionId) clearCleaningDraft(ctx, sessionId);
+        if (sessionId) clearCleaningDraft(ctx, sessionId, true);
         deactivateCleaningSession(sessionId);
       };
-      const handleSidebarNavigation = (event) => {
-        if (!activeCleaningSessionId) return;
-        const target = event?.target;
-        if (!target || typeof target.closest !== 'function') return;
-        if (target.closest('.dcAgentLauncher, .dcAgentExperience, .dcAgentPromptPopover, .dcAgentWorkbench, .dcAgentHeaderAction')) return;
-
-        const actionable = target.closest('button, a, [role="button"]');
-        if (!actionable) return;
-        const workspaceSlot = document.querySelector?.(SIDEBAR_WORKSPACES_SELECTOR);
-        const sidebarRoot = workspaceSlot?.closest?.('aside, nav, [data-slot="sidebar"], [class*="sidebar"]')
-          ?? workspaceSlot?.parentElement;
-        const label = `${actionable.getAttribute?.('aria-label') ?? ''} ${actionable.textContent ?? ''}`.trim();
-        const sessionOrAgentEntry = /新建?会话|新会话|new\s+session|智能体|agent|尽调/i.test(label);
-        if (sidebarRoot?.contains?.(actionable) || sessionOrAgentEntry) leaveSubsystem();
-      };
-
-      document.addEventListener?.('click', handleSidebarNavigation, true);
       // 升级 / 刷新后不恢复子系统所有权；仅清理由旧版本遗留且仍未被用户编辑的默认文案。
       clearStaleDefaultDraft();
       // DSH 会在插件 apply() 之后异步恢复 composer draft。在首屏稳定期内
@@ -1163,15 +1145,17 @@ window.__ModuleLoader__.load({
         }, 10_000);
       }
       const unsubscribe = ctx.sessions?.list?.subscribe?.((snapshot) => {
-        const nextSessionId = snapshot?.current;
-        if (activeCleaningSessionId && nextSessionId && nextSessionId !== activeCleaningSessionId) {
+        // Store subscriptions may signal without a payload; read authoritative state.
+        const selected = ctx.sessions?.list?.getSnapshot?.() ?? snapshot;
+        if (!selected) return;
+        const nextSessionId = selected.current;
+        if (activeCleaningSessionId && nextSessionId !== activeCleaningSessionId) {
           leaveSubsystem();
         } else if (!activeCleaningSessionId) {
           clearStaleDefaultDraft();
         }
       });
       return () => {
-        document.removeEventListener?.('click', handleSidebarNavigation, true);
         staleDraftObserver?.disconnect();
         if (staleDraftObserverTimer !== null) clearTimeout(staleDraftObserverTimer);
         if (typeof unsubscribe === 'function') unsubscribe();
@@ -1579,29 +1563,32 @@ window.__ModuleLoader__.load({
       const workspace = ctx.workspaces.list.getSnapshot();
       const current = ctx.sessions.list.getSnapshot().current;
       const items = Array.isArray(workspace.items) ? workspace.items : [];
-      const currentWorkspaceId = current === undefined
+      const currentWorkspace = current === undefined
         ? undefined
-        : items.find((item) => Array.isArray(item.sessionIds) && item.sessionIds.includes(current))?.workspaceId;
-      const targetWorkspaceId = currentWorkspaceId ?? workspace.recentWorkspaceId ?? items[0]?.workspaceId;
-      if (targetWorkspaceId === undefined) {
+        : items.find((item) => Array.isArray(item.sessionIds) && item.sessionIds.includes(current));
+      const recentWorkspace = items.find((item) => item.workspaceId === workspace.recentWorkspaceId);
+      const cwd = currentWorkspace?.path ?? recentWorkspace?.path ?? items[0]?.path;
+      if (cwd === undefined) {
         throw new Error('请先选择一个工作空间，再打开数据清洗补全智能体');
       }
       // 单一会话所有权：与招投标入口一致，显式创建带前缀的独立原生会话，绝不复用空白会话。
       // connectWorkspace 会把「招投标」尚为空白的工作台会话复用为清洗会话，
       // 使同一会话同时被招投标与清洗两个插件认领，进而争夺 Hero 标题与右侧工作台面板。
+      // 只用 cwd 创建（不传 workspaceId）：会话不挂进工作区，
+      // 「新会话」的 blank 复用逻辑就不会把它当成空白会话复用。
       if (typeof ctx.sessions.create !== 'function') {
         throw new Error('当前 DSH 版本没有可用的独立会话创建能力');
       }
       const requestedId = createCleaningSessionId();
-      const sessionId = await ctx.sessions.create({ workspaceId: targetWorkspaceId, sessionId: requestedId });
+      const sessionId = await ctx.sessions.create({ cwd, sessionId: requestedId });
       if (sessionId !== requestedId) {
         throw new Error('当前 DSH 版本不支持独立会话创建，请升级 DSH 后再打开数据清洗补全智能体');
       }
       const conversation = typeof ctx.get === 'function' ? ctx.get('conversation') : ctx.conversation;
       if (!conversation) throw new Error('DSH 对话服务尚未就绪，请稍后重试');
-      markCleaningSession(sessionId);
       conversation.input.shell(sessionId).setDraft(DEFAULT_SESSION_PROMPT);
       ctx.sessions.open(sessionId);
+      markCleaningSession(sessionId);
       return sessionId;
     }
 
@@ -1990,6 +1977,18 @@ window.__ModuleLoader__.load({
         .filter(Boolean);
     }
 
+    function isImagePathEntry(value) {
+      return /^(?:file:\/\/|\/|[a-z]:[\\/])/i.test(String(value).trim().replace(/^["']/, ''))
+        && /\.(?:png|jpe?g|webp|gif|heic|bmp)(?:["']|\s)*$/i.test(String(value));
+    }
+
+    function requestPromptImage(sessionId, file = null) {
+      if (!sessionId || typeof document.createEvent !== 'function') return false;
+      const event = document.createEvent('CustomEvent');
+      event.initCustomEvent('dsh:data-cleaning-prompt-image', false, true, { sessionId, file });
+      return !document.dispatchEvent(event);
+    }
+
     async function imageFileBase64(file) {
       if (!file || typeof file.arrayBuffer !== 'function') throw new Error('图片文件不可读。');
       const bytes = new Uint8Array(await file.arrayBuffer());
@@ -2013,7 +2012,7 @@ window.__ModuleLoader__.load({
     }
 
     function PromptGenerator(props) {
-      const { sessionId, inputActions, attachImages, removeImage } = props;
+      const { sessionId, inputActions, attachImages, removeImage, getInputSnapshot } = props;
       const enabled = useCleaningSession(sessionId);
       const [open, setOpen] = react.useState(false);
       const [wizardStep, setWizardStep] = react.useState(1);
@@ -2174,6 +2173,7 @@ window.__ModuleLoader__.load({
       const validateSource = () => {
         const values = selectedEntries();
         if (mode === 'text' && !values.length) return '请先录入企业名称或统一社会信用代码，每行一条。';
+        if (mode === 'text' && values.some(isImagePathEntry)) return '检测到图片文件路径，不是企业名单。请切换「上传图片识别」，重新选择或粘贴图片。';
         if (mode === 'excel' && !values.length) return '请先上传并解析 Excel / CSV / JSON 文件。';
         if (mode === 'image' && !imageCommand) return '请先选择、粘贴或拖入一张包含企业名单的图片。';
         if (mode === 'image' && imageCommand?.state === 'failed') return imageCommand.error?.message || '图片识别失败，请重新选择图片。';
@@ -2202,13 +2202,26 @@ window.__ModuleLoader__.load({
           setError('当前会话输入机尚未就绪，请稍后重试。');
           return;
         }
+        const inputSnapshot = getInputSnapshot?.(sessionId);
+        if (['adjudicating', 'submitting'].includes(inputSnapshot?.phase)) {
+          setError('当前对话正在提交，请等待完成后再回填。图片与字段选择已保留。');
+          return;
+        }
         const config = promptConfig();
         if (mode === 'image' && imageCommand?.state === 'prepared' && imageAttachment && typeof removeImage === 'function') {
           // qcc-document-mcp 只读取 Host 暂存文件；发送前释放原生附件，文本模型不会收到图片。
           removeImage(sessionId, imageAttachment);
           setImageAttachment(null);
         }
-        inputActions.setDraft(buildTaskPrompt(config));
+        const prompt = buildTaskPrompt(config);
+        try {
+          inputActions.setDraft(prompt);
+          const updatedInput = getInputSnapshot?.(sessionId);
+          if (updatedInput && updatedInput.draft !== prompt) throw new Error('当前对话尚未接收任务说明，请稍后重试回填。');
+        } catch (draftError) {
+          setError(draftError instanceof Error ? draftError.message : String(draftError));
+          return;
+        }
         if (mode === 'text') {
           requestWorkbenchDataset(entriesToDataset(config.entries), sessionId, { type: 'text', fileName: '' });
         }
@@ -2257,22 +2270,38 @@ window.__ModuleLoader__.load({
           const hasImage = files.length > 0;
           if (!hasImage) return;
           const target = event.target;
-          const inComposer = !open && Boolean(target?.closest?.('textarea, [contenteditable="true"], [role="textbox"]'));
-          const inImageStep = open && wizardStep === 1 && mode === 'image';
-          if (!inComposer && !inImageStep) return;
+          const inComposer = !open && !target?.closest?.('.dcAgentWorkbench')
+            && Boolean(target?.closest?.('textarea, [contenteditable="true"], [role="textbox"]'));
+          const inSourceStep = open && wizardStep === 1 && Boolean(target?.closest?.('.dcAgentPromptPanel'));
+          if (!inComposer && !inSourceStep) return;
           event.preventDefault();
           event.stopImmediatePropagation?.();
-          if (!open) {
-            setMode('image');
-            setWizardStep(1);
-            setOpen(true);
-            setError(null);
-          }
+          if (busy || imageCommand?.state === 'running') return;
+          setMode('image');
+          setWizardStep(1);
+          setOpen(true);
+          setError(null);
           acceptImageFiles(files);
         };
         window.addEventListener?.('paste', handleWindowPaste, true);
         return () => window.removeEventListener?.('paste', handleWindowPaste, true);
-      }, [enabled, open, wizardStep, mode, busy, sessionId]);
+      }, [enabled, open, wizardStep, mode, busy, sessionId, imageCommand, imageAttachment]);
+
+      // 右侧工作台共用此图片入口；仅由当前会话确认接收，不把图片交给表格解析器。
+      react.useEffect(() => {
+        if (!enabled) return undefined;
+        const receiveImage = (event) => {
+          if (event.detail?.sessionId !== sessionId || busy || imageCommand?.state === 'running') return;
+          event.preventDefault();
+          setMode('image');
+          setWizardStep(1);
+          setOpen(true);
+          setError(null);
+          if (event.detail.file) acceptImageFiles([event.detail.file]);
+        };
+        document.addEventListener('dsh:data-cleaning-prompt-image', receiveImage);
+        return () => document.removeEventListener('dsh:data-cleaning-prompt-image', receiveImage);
+      }, [enabled, sessionId, busy, imageCommand, imageAttachment]);
 
       if (!enabled) return null;
 
@@ -2855,9 +2884,22 @@ window.__ModuleLoader__.load({
         }
       };
 
+      const openImageIntake = (file = null) => {
+        if (!requestPromptImage(activeSessionId, file)) {
+          actions.setError('图片入口尚未就绪或正在处理图片，请回到当前清洗会话稍后重试。');
+          return false;
+        }
+        actions.close();
+        return true;
+      };
       const handleFile = async (event) => {
         const file = event.target && event.target.files && event.target.files[0];
         if (!file) return;
+        if (String(file.type ?? '').startsWith('image/')) {
+          openImageIntake(file);
+          event.target.value = '';
+          return;
+        }
         actions.setBusy(true);
         actions.setError(null);
         try {
@@ -3603,19 +3645,28 @@ window.__ModuleLoader__.load({
         );
       } else {
         pane = h('div', { className: 'dcAgentPane' },
-          h('p', { className: 'dcAgentHint' }, '上传 CSV / XLSX / XLS / JSON，或粘贴数据。原始明细仅在本机 Host 与同源工作台处理，不进入模型上下文。'),
+          h('p', { className: 'dcAgentHint' }, '上传 CSV / XLSX / XLS / JSON，或粘贴数据。也可选择图片，通过企查查智能文档解析提取名单后继续清洗补全。'),
           h('input', {
             type: 'file',
-            accept: '.csv,.json,.xlsx,.xls',
+            accept: '.csv,.json,.xlsx,.xls,image/png,image/jpeg,image/webp',
             className: 'dcAgentField',
             'aria-label': '选择数据文件',
             onChange: handleFile,
           }),
+          h('button', { type: 'button', className: 'dcAgentButton', disabled: busy,
+            onClick: () => openImageIntake() }, '图片识别（粘贴 / 选择）'),
           h('textarea', {
             className: 'dcAgentTextarea',
             placeholder: '粘贴 CSV 文本，或 JSON 数组（例如 [{"name":"某公司","amount":"100"}]）…',
             'aria-label': '粘贴数据',
             value: input,
+            onPaste: (event) => {
+              const files = imageFilesFromTransfer(event.clipboardData);
+              if (files.length && openImageIntake(files[0])) {
+                event.preventDefault();
+                event.stopPropagation?.();
+              }
+            },
             onInput: (event) => actions.setInput(event.target.value),
           }),
           dataset ? h('div', { className: 'dcAgentChips' },
@@ -3767,6 +3818,10 @@ window.__ModuleLoader__.load({
           inject: () => ({
             attachImages: (sessionId, files) => attachPromptImages(ctx, sessionId, files),
             removeImage: (sessionId, attachment) => removePromptImage(ctx, sessionId, attachment),
+            getInputSnapshot: (sessionId) => {
+              const conversation = typeof ctx.get === 'function' ? ctx.get('conversation') : ctx.conversation;
+              return conversation?.input?.shell?.(sessionId)?.snapshot;
+            },
           }),
         }, PromptGenerator));
 
@@ -3805,6 +3860,8 @@ window.__ModuleLoader__.load({
       entriesToDataset,
       extractPromptEntries,
       imageFilesFromTransfer,
+      isImagePathEntry,
+      requestPromptImage,
       guessMappings,
       deactivateCleaningSession,
       installSessionOwnershipBridge,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-data-cleaning-agent",
-  "version": "0.8.2",
+  "version": "0.8.4",
   "description": "Clean, complete, and profile enterprise name lists in DeepSeek Harness — a data cleaning & completion agent plugin with local CSV/XLSX/JSON engine and optional Qichacha (QCC) MCP enrichment. Maintained by Qichacha/QCC.",
   "type": "module",
   "main": "lib/index.js",