Skip to content

AI-native knowledge base & Evernote/Obsidian alternative with instant sync & native MCP. Free on Cloudflare or self-hosted via Docker.

fix extension web page clipping

tianma-ife66be31b8d1a
diff --git a/apps/extension/public/_locales/en/messages.json b/apps/extension/public/_locales/en/messages.json
index 795356e7..4e156981 100644
--- a/apps/extension/public/_locales/en/messages.json
+++ b/apps/extension/public/_locales/en/messages.json
@@ -106,7 +106,32 @@
"message": "Could not find the current page."
},
"captureTimeout": {
- "message": "Page extraction timed out. Please try again."
+ "message": "Page extraction did not finish within 15 seconds. The page may be large or still loading; refresh it and try again."
+ },
+ "pageAccessDenied": {
+ "message": "The extension cannot read this page. Browser internal pages, extension stores, and restricted pages cannot be captured."
+ },
+ "captureScriptFailed": {
+ "message": "The page extraction script failed: $DETAIL$",
+ "placeholders": {
+ "detail": {
+ "content": "$1"
+ }
+ }
+ },
+ "captureUnknownError": {
+ "message": "No further details were provided"
+ },
+ "instanceNetworkFailed": {
+ "message": "Could not connect to the EdgeEver instance. Check the instance URL, network connection, and CORS settings."
+ },
+ "saveFailedWithReason": {
+ "message": "Could not save to EdgeEver: $DETAIL$",
+ "placeholders": {
+ "detail": {
+ "content": "$1"
+ }
+ }
},
"emptyPageContent": {
"message": "(No readable content was found on this page)"
diff --git a/apps/extension/public/_locales/zh_CN/messages.json b/apps/extension/public/_locales/zh_CN/messages.json
index d2963d11..3dcee0fb 100644
--- a/apps/extension/public/_locales/zh_CN/messages.json
+++ b/apps/extension/public/_locales/zh_CN/messages.json
@@ -106,7 +106,32 @@
"message": "找不到当前页面。"
},
"captureTimeout": {
- "message": "网页内容提取超时,请重试。"
+ "message": "网页内容提取在 15 秒内未完成,页面可能过大或仍在加载。请刷新页面后重试。"
+ },
+ "pageAccessDenied": {
+ "message": "插件无法读取此页面。请确认当前不是浏览器内置页、扩展商店页或受限页面。"
+ },
+ "captureScriptFailed": {
+ "message": "网页提取脚本执行失败:$DETAIL$",
+ "placeholders": {
+ "detail": {
+ "content": "$1"
+ }
+ }
+ },
+ "captureUnknownError": {
+ "message": "未返回具体错误"
+ },
+ "instanceNetworkFailed": {
+ "message": "无法连接 EdgeEver 实例。请检查实例地址、网络连接和跨域设置。"
+ },
+ "saveFailedWithReason": {
+ "message": "保存到 EdgeEver 失败:$DETAIL$",
+ "placeholders": {
+ "detail": {
+ "content": "$1"
+ }
+ }
},
"emptyPageContent": {
"message": "(网页没有可提取的正文内容)"
diff --git a/apps/extension/src/background.ts b/apps/extension/src/background.ts
index 5663b36c..5626cb92 100644
--- a/apps/extension/src/background.ts
+++ b/apps/extension/src/background.ts
@@ -17,6 +17,30 @@ const toMarkdown = (page: CapturedPage) => {
return `# ${page.title.replace(/\n/g, " ")}\n\n${t("sourceLabel")}: [${page.url}](${page.url})\n\n${t("capturedAtLabel")}: ${capturedAt}\n\n---\n\n${page.markdown}`;
};
+const compactError = (error: unknown) => {
+ const message = error instanceof Error ? error.message : String(error);
+ return message.replace(/\s+/g, " ").trim().slice(0, 180);
+};
+
+const describeCaptureError = (error: unknown) => {
+ const message = compactError(error);
+ if (message === t("captureTimeout")) {
+ return message;
+ }
+ if (/cannot access (contents|a page)|extensions gallery|chrome:\/\/|edge:\/\/|about:\/\//i.test(message)) {
+ return t("pageAccessDenied");
+ }
+ return t("captureScriptFailed", message || t("captureUnknownError"));
+};
+
+const describeSaveError = (error: unknown) => {
+ const message = compactError(error);
+ if (/failed to fetch|networkerror|load failed|network request/i.test(message)) {
+ return t("instanceNetworkFailed");
+ }
+ return t("saveFailedWithReason", message || t("saveFailed"));
+};
+
const createMemo = async (settings: ExtensionSettings, page: CapturedPage) => {
const notebooks = await listNotebooks(settings);
const notebookId = settings.notebookId || notebooks.notebooks[0]?.id;
@@ -83,13 +107,19 @@ chrome.runtime.onMessage.addListener((message: { type?: string; page?: CapturedP
}).catch((error: unknown) => {
clearTimeout(timeout);
pendingCapture = null;
- reject(error);
+ reject(new Error(describeCaptureError(error)));
});
});
await createMemo(settings, page);
sendResponse({ ok: true });
} catch (error) {
- sendResponse({ ok: false, message: error instanceof Error ? error.message : t("saveFailed") });
+ const message = error instanceof Error ? error.message : "";
+ sendResponse({
+ ok: false,
+ message: message === t("captureTimeout") || message.startsWith(t("captureScriptFailed", ""))
+ ? describeCaptureError(error)
+ : describeSaveError(error),
+ });
}
})();
return true;
diff --git a/apps/extension/src/capture.ts b/apps/extension/src/capture.ts
index 0766decb..d4c71926 100644
--- a/apps/extension/src/capture.ts
+++ b/apps/extension/src/capture.ts
@@ -1,6 +1,10 @@
import { Readability } from "@mozilla/readability";
import TurndownService from "turndown";
-import { t } from "./i18n";
+
+// This file is injected with chrome.scripting.executeScript({ files }). It
+// must remain self-contained because injected files are not loaded as ES
+// modules and cannot resolve imports from the extension bundle.
+const t = (key: string) => chrome.i18n.getMessage(key) || key;
type CapturedPage = {
title: string;
@@ -24,28 +28,73 @@ const getSelectionHtml = () => {
const normalizeMarkdown = (value: string) => value.replace(/\n{3,}/g, "\n\n").trim();
+// Dynamic sites can contain very large inline SVGs, scripts, and embedded
+// documents. They are not article content, but cloning and parsing them makes
+// Readability disproportionately expensive (知乎 pages are a common example).
+const prepareDocumentForExtraction = () => {
+ const candidate = document.querySelector("article, main, [role='main']");
+ const clone = candidate
+ ? document.implementation.createHTMLDocument(document.title)
+ : (document.cloneNode(true) as Document);
+ if (candidate) {
+ clone.body.appendChild(candidate.cloneNode(true));
+ }
+ clone
+ .querySelectorAll("script, style, noscript, template, iframe, object, embed, canvas, svg, video, audio, source, link, meta")
+ .forEach((element) => element.remove());
+ return clone;
+};
+
+const textFallback = () => {
+ const text = document.body?.innerText?.replace(/\s+/g, " ").trim() || "";
+ return text ? text.replace(/([\\`*_{}[\]()#+.!|])/g, "\\$1") : t("emptyPageContent");
+};
+
const capturePage = (): CapturedPage => {
const selectionHtml = getSelectionHtml();
- const documentClone = document.cloneNode(true) as Document;
- const article = new Readability(documentClone, {
- charThreshold: 80,
- keepClasses: false,
- }).parse();
- const sourceHtml = selectionHtml || article?.content || document.body?.innerHTML || "";
+ let article: ReturnType<Readability["parse"]> = null;
+ let sourceHtml = selectionHtml;
+
+ if (!sourceHtml) {
+ const documentClone = prepareDocumentForExtraction();
+ article = new Readability(documentClone, {
+ charThreshold: 80,
+ keepClasses: false,
+ }).parse();
+ sourceHtml = article?.content || document.querySelector("article, main, [role='main']")?.innerHTML || "";
+ }
+
const title = article?.title?.trim() || document.title.trim() || location.hostname;
- const turndown = new TurndownService({
- bulletListMarker: "-",
- codeBlockStyle: "fenced",
- headingStyle: "atx",
- linkStyle: "inlined",
- });
- const markdown = normalizeMarkdown(turndown.turndown(sourceHtml));
+ let markdown = "";
+ if (sourceHtml) {
+ const turndown = new TurndownService({
+ bulletListMarker: "-",
+ codeBlockStyle: "fenced",
+ headingStyle: "atx",
+ linkStyle: "inlined",
+ });
+ markdown = normalizeMarkdown(turndown.turndown(sourceHtml));
+ }
return {
title,
url: location.href,
- markdown: markdown || t("emptyPageContent"),
+ markdown: markdown || textFallback(),
};
};
-void chrome.runtime.sendMessage({ type: "capturedPage", page: capturePage() });
+try {
+ void chrome.runtime.sendMessage({ type: "capturedPage", page: capturePage() });
+} catch {
+ // Keep the injected script from failing silently if a page exposes an
+ // unusual DOM implementation. The background page will show its timeout
+ // message, while ordinary pages still use the normal extraction path.
+ void chrome.runtime.sendMessage({
+ type: "capturedPage",
+ page: {
+ title: document.title.trim() || location.hostname,
+ url: location.href,
+ markdown: textFallback(),
+ },
+ });
+}