dsh-linked-workspace 0.6.1 → 0.6.2

package/CHANGELOG.md

@@ -1,5 +1,11 @@
 # Changelog
 
+## 0.6.2 — 2026-11-05
+
+- Fixed Markdown/document JSON transport failing with a Windows GBK codec error for characters such as the superscript minus (U+207B); the Python bridge now emits UTF-8/ASCII-safe JSON and configures UTF-8 stdout.
+- PDF files now load lazily: only the current page is rendered; other pages are fetched on demand (`docs.pdfPage`), and a large PDF is downloaded/opened once and cached. This removes the long “加载中” hang on large (30 MB+) PDFs.
+- PDF page navigation, rotation, deletion and metadata remain available.
+
 ## 0.6.1 — 2026-11-05
 
 - Fixed the linked-workspace directory tab disappearing for a new or blank linked-workspace session: the client session store is now consulted for cwd/workspaceId before falling back to session.info, so the tab and the start-page option are registered immediately when the workspace session opens.

package/README.en.md

@@ -115,6 +115,10 @@ Old “Remote directories” tabs are restored after refresh through hidden lega
 
 See [CHANGELOG.md](CHANGELOG.md) for the full version history.
 
+### 0.6.2
+
+Fix Windows GBK JSON transport errors for Markdown characters such as the superscript minus; PDF pages now load lazily and the source PDF is cached, avoiding long loading hangs on large files.
+
 ### 0.6.1
 
 Fix the linked-workspace directory tab and start-page entry disappearing in a new blank linked-workspace session by consulting the client session store for cwd and workspaceId first.

package/lib/client.js

@@ -4430,7 +4430,7 @@ function fileIconSpec(name, isDir) {
       var stateRef = useState({
         loading: true, error: '', data: null, text: '', dirty: false, saving: false,
         preview: false, html: '', previewLoading: false, page: 0, sheet: 0, slide: 0,
-        updates: {}, pptxUpdates: {}, rotations: {}, deleted: {}, notice: '',
+        updates: {}, pptxUpdates: {}, rotations: {}, deleted: {}, pdfLoading: false, notice: '',
       });
       var state = stateRef[0];
       var setState = stateRef[1];
@@ -4451,7 +4451,7 @@ function fileIconSpec(name, isDir) {
             loading: false, error: '', data: result,
             text: result.kind === 'markdown' ? String(result.text || '') : '',
             preview: false, html: '', page: 0, sheet: 0, slide: 0,
-            updates: {}, pptxUpdates: {}, rotations: {}, deleted: {}, dirty: false,
+            updates: {}, pptxUpdates: {}, rotations: {}, deleted: {}, pdfLoading: false, dirty: false,
           });
         }).catch(function (error) {
           patch({ loading: false, error: messageOf(error, t, locale) });
@@ -4538,6 +4538,35 @@ function fileIconSpec(name, isDir) {
         patch({ pptxUpdates: updates, dirty: true });
       }
 
+      function goPdfPage(index) {
+        if (index < 0 || index >= (state.data.pageCount || 0)) return;
+        var cached = (state.data.pages || []).filter(function (page) { return page.index === index; })[0];
+        if (cached) {
+          patch({ page: index });
+          return;
+        }
+        patch({ pdfLoading: true, error: '' });
+        apiCall('docs.pdfPage', params({ page: index })).then(function (result) {
+          var pages = (state.data.pages || []).filter(function (page) { return page.index !== result.pages[0].index; }).concat(result.pages);
+          pages.sort(function (left, right) { return left.index - right.index; });
+          patch({ data: Object.assign({}, state.data, { pages: pages }), page: index, pdfLoading: false });
+        }).catch(function (error) {
+          patch({ pdfLoading: false, error: messageOf(error, t, locale) });
+        });
+      }
+
+      function stepPdfPage(delta) {
+        var total = state.data.pageCount || 0;
+        var next = state.page + delta;
+        while (next >= 0 && next < total) {
+          if (!state.deleted[next]) {
+            goPdfPage(next);
+            return;
+          }
+          next += delta;
+        }
+      }
+
       var toolbar = h('div', { className: 'rw-doc-toolbar' },
         h('span', { className: 'rw-chip' }, kind.toUpperCase()),
         h('span', { className: 'rw-hint', style: { flex: '1 1 auto', overflow: 'hidden', textOverflow: 'ellipsis', whiteSpace: 'nowrap' }, title: path }, path),
@@ -4567,9 +4596,8 @@ function fileIconSpec(name, isDir) {
               onChange: function (event) { patch({ text: event.currentTarget.value, dirty: true, html: '' }); },
             }));
       } else if (kind === 'pdf' && state.data) {
-        var pages = state.data.pages || [];
-        var visible = pages.filter(function (page) { return !state.deleted[page.index]; });
-        var current = visible[Math.min(state.page, Math.max(visible.length - 1, 0))] || null;
+        var total = state.data.pageCount || 0;
+        var current = (state.data.pages || []).filter(function (page) { return page.index === state.page; })[0] || null;
         body = h('div', { className: 'rw-doc-body' },
           h('div', { className: 'rw-doc-field' },
             h('label', null, t('docPdfTitle')),
@@ -4578,13 +4606,14 @@ function fileIconSpec(name, isDir) {
             h('input', { className: 'rw-doc-input', style: { flex: '1 1 auto' }, value: (state.data.metadata && state.data.metadata.author) || '', onChange: function (event) { patch({ data: Object.assign({}, state.data, { metadata: Object.assign({}, state.data.metadata, { author: event.currentTarget.value }) }), dirty: true }); } })),
           h('div', { className: 'rw-inline', style: { justifyContent: 'space-between', margin: '6px 0' } },
             h('div', { className: 'rw-inline' },
-              h('button', { type: 'button', className: 'rw-btn rw-btn-sm', disabled: state.page <= 0, onClick: function () { patch({ page: state.page - 1 }); } }, '‹'),
-              h('span', { className: 'rw-hint' }, current ? t('docPageOf', { page: current.index + 1, total: state.data.pageCount }) : ''),
-              h('button', { type: 'button', className: 'rw-btn rw-btn-sm', disabled: state.page >= visible.length - 1, onClick: function () { patch({ page: state.page + 1 }); } }, '›')),
+              h('button', { type: 'button', className: 'rw-btn rw-btn-sm', disabled: state.page <= 0, onClick: function () { stepPdfPage(-1); } }, '‹'),
+              h('span', { className: 'rw-hint' }, t('docPageOf', { page: state.page + 1, total: total })),
+              h('button', { type: 'button', className: 'rw-btn rw-btn-sm', disabled: state.page >= total - 1, onClick: function () { stepPdfPage(1); } }, '›')),
             h('div', { className: 'rw-inline' },
-              h('button', { type: 'button', className: 'rw-btn rw-btn-sm', disabled: !current, onClick: function () { var rotations = Object.assign({}, state.rotations); rotations[current.index] = ((rotations[current.index] || current.rotation || 0) + 90) % 360; patch({ rotations: rotations, dirty: true }); } }, t('docRotate')),
-              h('button', { type: 'button', className: 'rw-btn rw-btn-sm rw-btn-danger', disabled: !current || visible.length <= 1, onClick: function () { var deleted = Object.assign({}, state.deleted); deleted[current.index] = true; patch({ deleted: deleted, page: Math.max(0, state.page - 1), dirty: true }); } }, t('docDeletePage')))),
-          current ? h('img', { className: 'rw-doc-pdf-page', src: current.png, style: { transform: 'rotate(' + ((state.rotations[current.index] || current.rotation || 0) % 360) + 'deg)' } }) : h('div', { className: 'rw-empty' }, t('noEntries')));
+              h('button', { type: 'button', className: 'rw-btn rw-btn-sm', disabled: !current, onClick: function () { var rotations = Object.assign({}, state.rotations); rotations[state.page] = ((rotations[state.page] || (current && current.rotation) || 0) + 90) % 360; patch({ rotations: rotations, dirty: true }); } }, t('docRotate')),
+              h('button', { type: 'button', className: 'rw-btn rw-btn-sm rw-btn-danger', disabled: total <= 1, onClick: function () { var deleted = Object.assign({}, state.deleted); deleted[state.page] = true; var next = state.page + 1; while (next < total && deleted[next]) next += 1; if (next >= total) { next = state.page - 1; while (next >= 0 && deleted[next]) next -= 1; } var nextPage = next >= 0 ? next : state.page; patch({ deleted: deleted, dirty: true, page: nextPage }); if (next >= 0 && !((state.data.pages || []).some(function (page) { return page.index === next; }))) goPdfPage(next); } }, t('docDeletePage')))),
+          state.pdfLoading ? h('div', { className: 'rw-editor-message' }, Spinner(14), t('loading')) : null,
+          current ? h('img', { className: 'rw-doc-pdf-page', src: current.png, style: { transform: 'rotate(' + ((state.rotations[current.index] || current.rotation || 0) % 360) + 'deg)' } }) : null);
       } else if (kind === 'docx' && state.data) {
         var paragraphs = state.data.paragraphs || [];
         var tables = state.data.tables || [];

package/lib/index.js

@@ -1505,7 +1505,8 @@ export function apply(ctx) {
       return { data: readFileSync(current.path), mtime: info.mtimeMs, size: info.size }
     }
     const stat = await current.scope.remote.stat(current.path)
-    if (stat.type !== 'file') throw new ApiError('FS_NOT_DIR', '只能打开文件')
+    const isFile = target.remote ? stat.type === 'file' : stat.isFile()
+    if (!isFile) throw new ApiError('FS_NOT_DIR', '只能打开文件')
     if (stat.size > DOC_MAX_BYTES) throw new ApiError('DOC_TOO_LARGE', '文档超过 80 MB')
     const read = await current.scope.remote.readFile(current.path, { maxBytes: DOC_MAX_BYTES })
     return { data: read.data, mtime: stat.mtime, size: stat.size }
@@ -1524,6 +1525,44 @@ export function apply(ctx) {
     return { path: target.path, size: result.size, mtime: result.mtime }
   }
 
+  const docFileCache = new Map()
+
+  async function ensureDocCache(target) {
+    let stat
+    try {
+      stat = target.remote ? await target.scope.remote.stat(target.path) : await fsStat(target.path)
+    } catch (error) {
+      if (target.remote) throw error
+      throw mapLocalFsError(error)
+    }
+    const isFile = target.remote ? stat.type === 'file' : stat.isFile()
+    if (!isFile) throw new ApiError('FS_NOT_DIR', '只能打开文件')
+    if (stat.size > DOC_MAX_BYTES) throw new ApiError('DOC_TOO_LARGE', '文档超过 80 MB')
+    const mtime = target.remote ? stat.mtime : stat.mtimeMs
+    const key = (target.remote ? String(target.scope.id) : 'local') + '|' + target.path + '|' + String(mtime)
+    const existing = docFileCache.get(key)
+    if (existing && existsSync(existing)) return { file: existing, size: stat.size, mtime }
+    let data
+    if (target.remote) {
+      const read = await target.scope.remote.readFile(target.path, { maxBytes: DOC_MAX_BYTES })
+      data = read.data
+    } else {
+      data = readFileSync(target.path)
+    }
+    const directory = join(tmpdir(), 'dsh-linked-workspace-docs')
+    mkdirSync(directory, { recursive: true })
+    const file = join(directory, randomUUID() + (extname(target.path) || '.bin'))
+    writeFileSync(file, data)
+    docFileCache.set(key, file)
+    if (docFileCache.size > 20) {
+      const oldestKey = docFileCache.keys().next().value
+      const oldestFile = docFileCache.get(oldestKey)
+      docFileCache.delete(oldestKey)
+      try { rmSync(oldestFile, { force: true }) } catch { /* best effort */ }
+    }
+    return { file, size: stat.size, mtime }
+  }
+
   async function makeDocTemp() {
     const directory = mkdtempSync(join(tmpdir(), 'dsh-doc-'))
     return {
@@ -1733,6 +1772,11 @@ export function apply(ctx) {
     async 'docs.read'(params = {}) {
       const target = await docTarget(params)
       const kind = assertDocKind(target.path)
+      if (kind === 'pdf') {
+        const cached = await ensureDocCache(target)
+        const data = await runDocTool(['read', '--kind', 'pdf', '--file', cached.file])
+        return { ...data, path: target.path, size: cached.size, mtime: cached.mtime, cached: true }
+      }
       const current = await readDocBytes(target, target)
       if (kind === 'markdown') {
         if (looksBinary(current.data)) throw new ApiError('DOC_BINARY', 'Markdown 文件包含二进制内容')
@@ -1749,6 +1793,15 @@ export function apply(ctx) {
       }
     },
 
+    async 'docs.pdfPage'(params = {}) {
+      const target = await docTarget(params)
+      assertDocKind(target.path, 'pdf')
+      const page = Number(params.page)
+      if (!Number.isInteger(page) || page < 0) throw new ApiError('BAD_REQUEST', '页码无效')
+      const cached = await ensureDocCache(target)
+      return await runDocTool(['read', '--kind', 'pdf', '--file', cached.file, '--page', String(page)])
+    },
+
     async 'docs.write'(params = {}) {
       const target = await docTarget(params)
       const kind = assertDocKind(target.path)
@@ -1765,13 +1818,18 @@ export function apply(ctx) {
         const written = await writeDocBytes(target, Buffer.from(data.text, 'utf8'))
         return { kind, ...written }
       }
-      const current = await readDocBytes(target, target)
+      let sourceFile
+      if (kind === 'pdf') {
+        sourceFile = (await ensureDocCache(target)).file
+      } else {
+        sourceFile = (await readDocBytes(target, target)).data
+      }
       const temp = await makeDocTemp()
       try {
         const source = temp.file('source.' + kind)
         const spec = temp.file('data.json')
         const output = temp.file('output.' + kind)
-        writeFileSync(source, current.data)
+        writeFileSync(source, Buffer.isBuffer(sourceFile) ? sourceFile : readFileSync(sourceFile))
         writeFileSync(spec, JSON.stringify(data))
         await runDocTool(['write', '--kind', kind, '--file', source, '--data', spec, '--out', output])
         const written = await writeDocBytes(target, readFileSync(output))

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-linked-workspace",
-  "version": "0.6.1",
+  "version": "0.6.2",
   "description": "Linked local ⇄ remote workspaces and SSH-backed remote directories for DeepSeek Harness: theme-aware split editor, file-type icons, and an agent tool that edits remote files directly.",
   "type": "module",
   "main": "lib/index.js",

package/scripts/doc_tool.py

@@ -13,6 +13,12 @@ import os
 import re
 import sys
 
+try:
+    sys.stdout.reconfigure(encoding="utf-8")
+    sys.stderr.reconfigure(encoding="utf-8")
+except Exception:
+    pass
+
 MAX_PDF_PAGES = 40
 MAX_PDF_WIDTH = 1100
 MAX_SHEET_ROWS = 200
@@ -21,12 +27,12 @@ MAX_PPTX_SLIDES = 80
 
 
 def fail(message):
-    print(json.dumps({"ok": False, "error": str(message)}, ensure_ascii=False))
+    print(json.dumps({"ok": False, "error": str(message)}, ensure_ascii=True))
     sys.exit(1)
 
 
 def emit(data):
-    print(json.dumps({"ok": True, "data": data}, ensure_ascii=False))
+    print(json.dumps({"ok": True, "data": data}, ensure_ascii=True))
 
 
 def json_value(value):
@@ -198,14 +204,16 @@ def write_pptx(source, data, out):
 
 # ── PDF ───────────────────────────────────────────────────────────────────
 
-def read_pdf(path):
+def read_pdf(path, page_index=None):
     import fitz
 
     doc = fitz.open(path)
     pages = []
-    for index, page in enumerate(doc):
-        if index >= MAX_PDF_PAGES:
-            break
+    indexes = [page_index] if isinstance(page_index, int) else [0]
+    for index in indexes:
+        if index < 0 or index >= doc.page_count:
+            continue
+        page = doc[index]
         rect = page.rect
         scale = min(1.5, MAX_PDF_WIDTH / max(rect.width, 1))
         pixmap = page.get_pixmap(matrix=fitz.Matrix(scale, scale), alpha=False)
@@ -226,7 +234,7 @@ def read_pdf(path):
             "subject": metadata.get("subject", ""),
         },
         "pages": pages,
-        "truncated": doc.page_count > MAX_PDF_PAGES,
+        "truncated": doc.page_count > 1,
     })
 
 
@@ -347,6 +355,7 @@ def main():
     read = sub.add_parser("read")
     read.add_argument("--kind", required=True)
     read.add_argument("--file", required=True)
+    read.add_argument("--page", type=int, default=None)
 
     write = sub.add_parser("write")
     write.add_argument("--kind", required=True)
@@ -375,7 +384,7 @@ def main():
             elif args.kind == "pptx":
                 read_pptx(args.file)
             elif args.kind == "pdf":
-                read_pdf(args.file)
+                read_pdf(args.file, args.page)
             else:
                 raise ValueError(f"unsupported kind: {args.kind}")
         elif args.command == "write":