From cd77340d92bb6f89c5ed74e598129b683c0abff8 Mon Sep 17 00:00:00 2001 From: weichen Date: Wed, 30 Sep 2026 17:51:14 +0000 Subject: [PATCH 1/2] Sniff docx ZIP magic and fall back to text rendering Why: agent-generated ".docx" artifacts were historically plain text bytes written via string-only write tools; the docx preview loader fed them straight to mammoth/JSZip, which threw "Can't find end of central directory" on the meeting-notes docx from the review session. Evidence: docxLoader now checks the PK\x03\x04 (or empty-archive PK\x05\x06) magic before mammoth; non-ZIP bytes render as markdown with a readable note instead of an error. vitest 33/33 pass, tsc --noEmit clean. Impact: stock fake-docx artifacts become readable immediately; real docx previews unchanged. Rendering only, no API/contract change. Refs: .task/feature-preview-sniff-fallback.md --- src/previewsetup.ts | 31 +++++++++++++++++++++++++++++-- 1 file changed, 29 insertions(+), 2 deletions(-) diff --git a/src/previewsetup.ts b/src/previewsetup.ts index b236bc3..294bb05 100644 --- a/src/previewsetup.ts +++ b/src/previewsetup.ts @@ -2,6 +2,7 @@ // 全部 dynamic import——对应格式首次预览时才下载 chunk，主 bundle 不含解析器。 // 链接预览 fetcher 走服务端 /api/url-preview（SSRF 防护在服务端执行端）。 import { configureFilePreviewLoaders, configureLinkPreviewFetcher } from '@ui' +import { marked } from 'marked' import { auth } from './auth' import type { LinkPreviewData } from '@ui' @@ -11,9 +12,35 @@ const fetchBuf = async (url: string | undefined): Promise => { return r.arrayBuffer() } -const docxLoader = async (el: HTMLElement, source: { contentUrl?: string }) => { +// docx/OOXML 是 ZIP 容器，真文件必以 PK 魔数开头（PK\x03\x04 常规，PK\x05\x06 空包）。 +// 智能体生成侧可能写出"内容是文本的伪 docx"，先嗅探再喂 mammoth： +// 非 ZIP 直接按文本/markdown 回退渲染，不再抛 JSZip 底层错误。 +const isZipBytes = (buf: ArrayBuffer): boolean => { + if (buf.byteLength < 4) return false + const h = new Uint8Array(buf, 0, 4) + return h[0] === 0x50 && h[1] === 0x4b && + ((h[2] === 0x03 && h[3] === 0x04) || (h[2] === 0x05 && h[3] === 0x06)) +} + +const renderTextFallback = (el: HTMLElement, buf: ArrayBuffer, name?: string) => { + const note = document.createElement('div') + note.className = 'ui-fp-note' + note.textContent = `「${name ?? '该文件'}」不是有效的 docx 文件（缺少 ZIP 结构），已按文本内容展示；需要正式 Word 格式请让智能体以 docx 格式重新生成。` + const body = document.createElement('div') + body.className = 'doc-viewer' + body.innerHTML = marked.parse(new TextDecoder().decode(buf)) as string + el.innerHTML = '' + el.append(note, body) +} + +const docxLoader = async (el: HTMLElement, source: { contentUrl?: string; name?: string }) => { + const buf = await fetchBuf(source.contentUrl) + if (!isZipBytes(buf)) { + renderTextFallback(el, buf, source.name) + return + } const mammoth = await import('mammoth') - const { value } = await mammoth.convertToHtml({ arrayBuffer: await fetchBuf(source.contentUrl) }) + const { value } = await mammoth.convertToHtml({ arrayBuffer: buf }) el.innerHTML = value } From bfa634d2a991523376302eb4f9479b0fa1df1c1c Mon Sep 17 00:00:00 2001 From: weichen Date: Thu, 1 Oct 2026 04:07:05 +0000 Subject: [PATCH 2/2] Fall back to text when xlsx and pptx are not ZIP files Why: docx preview already sniffs ZIP magic so a text file renamed to .docx stays readable. xlsx and pptx loaders still fed those bytes to the parser and surfaced a raw library error. Evidence: npx tsc --noEmit and npx vitest run (40 tests) on this branch after merging origin/main. Legacy .xls/.ppt skip the sniff because they are not OOXML. Impact: pseudo xlsx/pptx render as text with a short note. Real ZIP workbooks and decks still use the existing parsers. Refs: Dev PR #96, task feature/preview-sniff-fallback Co-authored-by: Cursor --- src/previewsetup.ts | 26 +++++++++++++++++++------- 1 file changed, 19 insertions(+), 7 deletions(-) diff --git a/src/previewsetup.ts b/src/previewsetup.ts index 294bb05..2ee393b 100644 --- a/src/previewsetup.ts +++ b/src/previewsetup.ts @@ -22,10 +22,10 @@ const isZipBytes = (buf: ArrayBuffer): boolean => { ((h[2] === 0x03 && h[3] === 0x04) || (h[2] === 0x05 && h[3] === 0x06)) } -const renderTextFallback = (el: HTMLElement, buf: ArrayBuffer, name?: string) => { +const renderTextFallback = (el: HTMLElement, buf: ArrayBuffer, name: string | undefined, kindLabel: string) => { const note = document.createElement('div') note.className = 'ui-fp-note' - note.textContent = `「${name ?? '该文件'}」不是有效的 docx 文件（缺少 ZIP 结构），已按文本内容展示；需要正式 Word 格式请让智能体以 docx 格式重新生成。` + note.textContent = `「${name ?? '该文件'}」不是有效的 ${kindLabel} 文件（缺少 ZIP 结构），已按文本内容展示；需要正式格式请让智能体重新生成。` const body = document.createElement('div') body.className = 'doc-viewer' body.innerHTML = marked.parse(new TextDecoder().decode(buf)) as string @@ -36,7 +36,7 @@ const renderTextFallback = (el: HTMLElement, buf: ArrayBuffer, name?: string) => const docxLoader = async (el: HTMLElement, source: { contentUrl?: string; name?: string }) => { const buf = await fetchBuf(source.contentUrl) if (!isZipBytes(buf)) { - renderTextFallback(el, buf, source.name) + renderTextFallback(el, buf, source.name, 'DOCX') return } const mammoth = await import('mammoth') @@ -44,9 +44,15 @@ const docxLoader = async (el: HTMLElement, source: { contentUrl?: string; name?: el.innerHTML = value } -const xlsxLoader = async (el: HTMLElement, source: { contentUrl?: string }) => { +const xlsxLoader = async (el: HTMLElement, source: { contentUrl?: string; name?: string }) => { + const buf = await fetchBuf(source.contentUrl) + // .xls 是 OLE 复合文档，不是 ZIP；只对 OOXML（xlsx/xlsm，或未给文件名）做魔数嗅探。 + if (!/\.xls$/i.test(source.name ?? '') && !isZipBytes(buf)) { + renderTextFallback(el, buf, source.name, 'XLSX') + return + } const XLSX = await import('xlsx') - const wb = XLSX.read(await fetchBuf(source.contentUrl), { type: 'array' }) + const wb = XLSX.read(buf, { type: 'array' }) el.innerHTML = '' const tabs = document.createElement('div') tabs.className = 'ui-fp-sheet-tabs' @@ -90,7 +96,13 @@ function renderSheet(wrap: HTMLElement, rows: unknown[][]) { wrap.append(table) } -const pptxLoader = async (el: HTMLElement, source: { contentUrl?: string }) => { +const pptxLoader = async (el: HTMLElement, source: { contentUrl?: string; name?: string }) => { + const buf = await fetchBuf(source.contentUrl) + // .ppt 是老版二进制；OOXML（pptx/ppsx）才要求 ZIP 魔数。 + if (!/\.ppt$/i.test(source.name ?? '') && !isZipBytes(buf)) { + renderTextFallback(el, buf, source.name, 'PPTX') + return + } // pptx2html@0.3.4 的 module 入口引用了大小写错配的 ./txml（Linux 打包必炸）， // 按 D2 备选条款改用 pptx-preview（init(dom).preview(arrayBuffer) 全保真渲染）。 const { init } = await import('pptx-preview') @@ -99,7 +111,7 @@ const pptxLoader = async (el: HTMLElement, source: { contentUrl?: string }) => { stage.className = 'ui-fp-slide-stage' el.append(stage) const previewer = init(stage, { width: stage.clientWidth || 900 }) - await previewer.preview(await fetchBuf(source.contentUrl)) + await previewer.preview(buf) } export function wirePreviewCapabilities() {