fix(settling): 审查修复——IPv6 SSRF 绕过/英文断行粘连/重试重复安放/响应体内存/完成计数

- validateFetchUrl 增加 IPv6 ULA(fc00::/7) 与 link-local(fe80::/10) 拦截，
  封堵 [fd00::1] 等内网地址绕过（SSRF 防护补全）
- contentSanitizer 断行合并时 ASCII 词边界补空格，修复 broken\\nline →
  brokenline 英文单词粘连（测试断言同步纠正，新增中英混排用例）
- SettlingPage 部分失败重试时移除已成功安放项，消除重复笔记/卡片风险
- import:fetch-url 响应体流式截断 + Content-Length 预检，防超大响应
  拖垮主进程内存
- done 页展示实际安放数（过滤空名称后），修正数字误导

Author: Aparencia

client/electron/importHandlers.ts

@@ -57,6 +57,9 @@ function toSettlingRecord(row: ImportRow): SettlingRecord {
 
 /**
  * SSRF 防护：仅允许公网 HTTP(S)，拦截内网/回环/链路本地地址。
+ * 覆盖 WHATWG URL 规范化的十进制/八进制/十六进制 IPv4（统一变成点分
+ * 十进制后可被 /^127\./ 等捕获）与 IPv6 内网前缀（ULA fc00::/7、
+ * link-local fe80::/10、IPv4-mapped ::ffff:）。
  * @returns 违规原因；null 表示放行 / Return block reason or null
  */
 export function validateFetchUrl(rawUrl: string): string | null {
@@ -82,7 +85,10 @@ export function validateFetchUrl(rawUrl: string): string | null {
     /^172\.(1[6-9]|2\d|3[01])\./.test(host) ||
     /^169\.254\./.test(host) ||
     host === '::1' ||
-    host === '0:0:0:0:0:0:0:1';
+    host === '0:0:0:0:0:0:0:1' ||
+    // IPv6 内网前缀：ULA fc00::/7（fc00-fdff）与 link-local fe80::/10（fe80-febf）
+    /^f[cd]/.test(host) ||
+    /^fe[89ab]/.test(host);
   return isPrivate ? '不允许访问内网或本机地址' : null;
 }
 
@@ -113,6 +119,32 @@ async function fetchWithRedirectGuard(initialUrl: string): Promise<Response> {
   }
 }
 
+/**
+ * 流式读取响应体并截断到上限字符数（防超大响应内存风险）
+ * Stream the response body, bounded to maxChars characters.
+ */
+async function readBodyBounded(res: Response, maxChars: number): Promise<string> {
+  if (!res.body) return (await res.text()).slice(0, maxChars);
+  const reader = res.body.getReader();
+  const decoder = new TextDecoder();
+  let body = '';
+  try {
+    for (;;) {
+      const { done, value } = await reader.read();
+      if (done) break;
+      body += decoder.decode(value, { stream: true });
+      if (body.length >= maxChars) {
+        body = body.slice(0, maxChars);
+        await reader.cancel(); // 提前终止下载
+        break;
+      }
+    }
+  } finally {
+    reader.releaseLock();
+  }
+  return body;
+}
+
 /**
  * 轻量 HTML 正文提取（无第三方依赖）：去噪声节点后取 <article>/<main>
  * 或 body 文本，折叠空白。非完整渲染，仅服务知识入籍的正文抓取。
@@ -228,8 +260,15 @@ export function registerImportHandlers(): void {
       if (!res.ok) {
         return { success: false, error: `网页访问失败（HTTP ${res.status}），可手动粘贴内容` };
       }
-      const raw = await res.text();
-      const { title, text } = extractHtmlText(raw.slice(0, MAX_BODY_CHARS));
+      // Content-Length 预检：超大响应直接拒绝，避免无谓下载
+      const declared = Number(res.headers.get('content-length') ?? 0);
+      if (declared > MAX_BODY_CHARS) {
+        return { success: false, error: '网页内容过大，可手动粘贴正文内容' };
+      }
+      // 流式截断：即使服务器不声明 Content-Length 或分块传输，
+      // 也只累计前 MAX_BODY_CHARS 字符，防止超大响应拖垮主进程内存
+      const raw = await readBodyBounded(res, MAX_BODY_CHARS);
+      const { title, text } = extractHtmlText(raw);
       if (!text) {
         return { success: false, error: '未能从该网页提取到正文，可手动粘贴内容' };
       }

client/src/features/settling/lib/contentSanitizer.test.ts

@@ -14,15 +14,26 @@ describe('sanitizeExtractedText', () => {
     expect(result).toBe('');
   });
 
-  it('should merge broken English lines into one sentence', () => {
-    // Arrange：PDF 提取常见的断行
+  it('should merge broken English lines with a space to avoid word sticking', () => {
+    // Arrange：PDF 提取常见的英文断行——直接拼接会产生 brokenline 粘连
     const raw = 'This is a broken\nline that should\nbe merged.';
 
     // Act
     const result = sanitizeExtractedText(raw);
 
+    // Assert：词边界保留空格
+    expect(result).toBe('This is a broken line that should be merged.');
+  });
+
+  it('should not insert a space between CJK and CJK, or CJK and ASCII edges', () => {
+    // Arrange：中文字符间直接连接；中英边界也不插空格（保持原文排布）
+    const raw = '这是被换行打断的\n中文句子，应当合并。\n学习费曼\nLearning 方法。';
+
+    // Act
+    const result = sanitizeExtractedText(raw);
+
     // Assert
-    expect(result).toBe('This is a brokenline that shouldbe merged.');
+    expect(result).toBe('这是被换行打断的中文句子，应当合并。\n学习费曼Learning 方法。');
   });
 
   it('should merge broken Chinese lines without inserting spaces', () => {

client/src/features/settling/lib/contentSanitizer.ts

@@ -24,6 +24,24 @@ const SENTENCE_END_RE = /[。！？.!?…]$/;
 /** 疑似页码/装饰行（纯数字、纯符号、空格夹杂） / Page-number-like lines */
 const JUNK_LINE_RE = /^[\d\s\-—–·•*|/\\]+$/;
 
+/** ASCII 字母数字（英文断行合并时需补空格，避免单词粘连） / ASCII word chars */
+const ASCII_WORD_RE = /[A-Za-z0-9]/;
+
+/**
+ * 断行合并：中英文混排场景的换行还原。
+ * 中文字符间直接连接；两侧均为 ASCII 字母/数字时插入空格
+ * （PDF 断行常把 "broken line" 切成 "broken\nline"，直接拼接会变
+ * "brokenline"，破坏英文语义）。
+ * Join a wrapped line to the pending one; insert a space when both
+ * edges are ASCII word characters (English word-wrap restoration).
+ */
+function joinPending(pending: string, collapsed: string): string {
+  const prevChar = pending[pending.length - 1] ?? '';
+  const nextChar = collapsed[0] ?? '';
+  const needsSpace = ASCII_WORD_RE.test(prevChar) && ASCII_WORD_RE.test(nextChar);
+  return needsSpace ? `${pending} ${collapsed}` : `${pending}${collapsed}`;
+}
+
 /**
  * 单行清理：压缩内部连续空白 / Collapse inner whitespace of a line
  */
@@ -73,7 +91,7 @@ export function sanitizeExtractedText(raw: string): string {
 
     // 上行末尾无句读 → 与当前行合并（断行还原）
     if (pending && !SENTENCE_END_RE.test(pending)) {
-      pending = `${pending}${collapsed}`;
+      pending = joinPending(pending, collapsed);
       continue;
     }
 

client/src/features/settling/pages/SettlingPage.tsx

@@ -54,6 +54,8 @@ export default function SettlingPage() {
   const [concepts, setConcepts] = useState<ConceptCandidate[]>([]);
   const [error, setError] = useState<string | null>(null);
   const [isConceptualizing, setIsConceptualizing] = useState(false);
+  /** 实际安放的概念数（过滤空名称后，done 页展示用） / Actually settled count */
+  const [settledCount, setSettledCount] = useState(0);
 
   /** 进入预览：统一过内容清理（断行合并/页眉丢弃），提升 AI 概念化输入质量 */
   const enterPreview = (input: ParsedInput) => {
@@ -117,11 +119,20 @@ export default function SettlingPage() {
     if (valid.length === 0) { setError('请至少填写一个概念名称'); return; }
     setStep('settling'); setError(null);
     const res = await settleConcepts({ title: parsed.title, source: parsed.source, rawName: parsed.rawName, concepts: valid });
-    if (res.ok) setStep('done');
-    else { setError(res.error ?? '安放失败，可稍后重试'); setStep('preview'); }
+    if (res.ok) {
+      setSettledCount(valid.length);
+      setStep('done');
+    } else {
+      // 部分失败：从概念列表移除已成功安放项，重试不会重复创建笔记/卡片
+      // （settleConcepts 按序安放，noteIds 即前 N 个概念的成功结果）
+      const remaining = valid.slice(res.noteIds.length);
+      setConcepts(remaining.length > 0 ? remaining : [emptyConcept()]);
+      setError(res.error ?? '安放失败，可稍后重试');
+      setStep('preview');
+    }
   };
 
-  const resetAll = () => { setStep('source'); setParsed(null); setConcepts([]); setText(''); setUrl(''); setTitle(''); setError(null); };
+  const resetAll = () => { setStep('source'); setParsed(null); setConcepts([]); setText(''); setUrl(''); setTitle(''); setError(null); setSettledCount(0); };
   const updateConcept = (i: number, patch: Partial<ConceptCandidate>) =>
     setConcepts((cs) => cs.map((c, j) => (j === i ? { ...c, ...patch } : c)));
 
@@ -226,7 +237,7 @@ export default function SettlingPage() {
   const renderDone = () => (
     <div className="flex flex-col items-center gap-4 py-16 text-center">
       <div className="text-4xl">🌌</div>
-      <div className="text-lg font-medium text-slate-200">已安放 {concepts.length} 个概念</div>
+      <div className="text-lg font-medium text-slate-200">已安放 {settledCount} 个概念</div>
       <p className="text-sm text-slate-400">它们已在你的世界里亮起，呈雾中轮廓 · 可以从容地慢慢复习</p>
       <div className="mt-4 flex gap-3">
         <button onClick={() => navigate('/')} className={primaryBtn}>回到世界</button>