fix: anima-extract 支援 wiki URL 與自然語言參數，並依序逐章處理

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

hooks/extract.ts

@@ -228,3 +228,50 @@ export const parseRemote = (url: string): { host: string; repo: string } | undef
   const m = url.trim().match(/^(?:https?:\/\/(?:[^@/]+@)?|ssh:\/\/(?:[^@/]+@)?|[^@/]+@)([^/:]+)(?::\d+)?[/:](.+?)(?:\.git)?$/)
   return m ? { host: m[1]!, repo: m[2]! } : undefined
 }
+
+export type WikiSource = { kind: 'list' | 'page'; url: string; origin: string; base: string; page?: string }
+
+// 從自然語言參數中取出第一個 URL；Gitea wiki 頁面清單（?action=_pages 或 /wiki 首頁）視為整本逐章，單頁網址視為一章。
+export const parseWikiSource = (args: string): WikiSource | undefined => {
+  const raw = /https?:\/\/[^\s<>"'「」『』（）()]+/.exec(args)?.[0]?.replace(/[，。、；,.;:!?！？]+$/, '')
+  if (!raw) return undefined
+  let u: URL
+  try {
+    u = new URL(raw)
+  } catch {
+    return undefined
+  }
+  const m = /^(.*\/wiki)(?:\/(.*))?$/.exec(u.pathname.replace(/\/+$/, ''))
+  if (!m) return { kind: 'page', url: raw, origin: u.origin, base: raw }
+  const base = `${u.origin}${m[1]}`
+  const page = m[2] && !m[2].startsWith('raw/') && m[2] !== '_pages' ? decodeURIComponent(m[2]) : undefined
+  if (u.searchParams.get('action') === '_pages' || !page) return { kind: 'list', url: raw, origin: u.origin, base }
+  return { kind: 'page', url: raw, origin: u.origin, base, page }
+}
+
+const SKIP_PAGES = new Set(['home', 'sidebar', 'footer', '_sidebar', '_footer', '_pages', 'new'])
+
+// 解析 Gitea wiki 頁面清單 HTML，依章節名稱自然排序（第2章 < 第10章）。
+export const parseWikiPages = (html: string, base: string): string[] => {
+  const path = new URL(base).pathname
+  const names = new Set<string>()
+  for (const m of html.matchAll(/href="([^"#]+)"/g)) {
+    let p: string
+    try {
+      p = new URL(m[1]!.replace(/&amp;/g, '&'), base).pathname
+    } catch {
+      continue
+    }
+    if (!p.startsWith(`${path}/`)) continue
+    const rest = p.slice(path.length + 1)
+    if (!rest || rest.includes('/')) continue
+    let name: string
+    try {
+      name = decodeURIComponent(rest)
+    } catch {
+      continue
+    }
+    if (!SKIP_PAGES.has(name.toLowerCase())) names.add(name)
+  }
+  return [...names].sort((a, b) => a.localeCompare(b, 'zh-Hant', { numeric: true }))
+}