feat(subtitles): 字幕站(SubHD、字幕库)在线搜索与下载;rar / 7z 解压;验证码与页面解析在沙箱里

Co-Authored-By: Claude Opus 5.5 <[email protected]>

apps/server/src/modules/sources-identity/routes/sources.ts

@@ -25,6 +25,7 @@ import {
   SiteUserDataPoint,
   SiteUserDataQuery,
   type SourceRelease,
+  SUBTITLE_SITES,
 } from "@umt/contracts";
 import { type FeedSourceRow, sql } from "@umt/db";
 import { z } from "zod";
@@ -92,19 +93,35 @@ export function sourcesRoutes(app: ModuleApp, rt: () => SourcesRuntime) {
       const rows = await db().selectFrom("site").select(["id", "key"]).where("deleted_at", "is", null).execute();
       const byKey = new Map(rows.map((r) => [r.key, r.id]));
       return {
-        items: listSiteDefinitions().map((d) => ({
-          key: d.id,
-          name: d.umt.kind === "nexusphp" ? (d.umt.displayName ?? d.name) : d.umt.displayName,
-          kind: definitionKind(d),
-          adapter: adapterOf(d),
-          baseUrl: d.domain,
-          shippedEnabled: d.enabled !== false,
-          definitionVersion: d.umt.version,
-          capabilities: d.umt.capabilities,
-          defaultRateLimit: d.umt.rateLimit,
-          authSupported: ("authSupported" in d.umt ? d.umt.authSupported : undefined) ?? [],
-          siteId: byKey.get(d.id) ?? null,
-        })),
+        items: [
+          ...listSiteDefinitions().map((d) => ({
+            key: d.id,
+            name: d.umt.kind === "nexusphp" ? (d.umt.displayName ?? d.name) : d.umt.displayName,
+            kind: definitionKind(d),
+            adapter: adapterOf(d),
+            baseUrl: d.domain,
+            shippedEnabled: d.enabled !== false,
+            definitionVersion: d.umt.version,
+            capabilities: d.umt.capabilities,
+            defaultRateLimit: d.umt.rateLimit,
+            authSupported: ("authSupported" in d.umt ? d.umt.authSupported : undefined) ?? [],
+            siteId: byKey.get(d.id) ?? null,
+          })),
+          // 内置字幕站:不是资源适配器,只用于补字幕(subtitles 模块)
+          ...SUBTITLE_SITES.map((d) => ({
+            key: d.key,
+            name: d.name,
+            kind: "subtitle" as const,
+            adapter: "subtitle" as const,
+            baseUrl: d.baseUrl,
+            shippedEnabled: true,
+            definitionVersion: "builtin",
+            capabilities: { list: false, search: false, detail: false, download: true, rss: false },
+            defaultRateLimit: { minIntervalMs: 2000, burst: 2 },
+            authSupported: [],
+            siteId: byKey.get(d.key) ?? null,
+          })),
+        ],
       };
     },
   );
@@ -159,6 +176,9 @@ export function sourcesRoutes(app: ModuleApp, rt: () => SourcesRuntime) {
         throw new ApiException("bad_request", `注册表中没有站点 ${body.key}；未适配站点只能以 generic_rss 登记`);
       }
       if (def && definitionKind(def) !== "pt") throw new ApiException("bad_request", `${body.key} 不是 PT 站`);
+    } else if (kind === "subtitle") {
+      if (!SUBTITLE_SITES.some((d) => d.key === body.key) || body.adapter !== "subtitle")
+        throw new ApiException("bad_request", "字幕站要从内置字幕站里选（SubHD、字幕库）");
     } else if (publicDef) {
       if (body.adapter !== adapterOf(publicDef))
         throw new ApiException("bad_request", `站点 ${body.key} 的适配器类型应为 ${adapterOf(publicDef)}`);

apps/server/src/modules/sources-identity/service/account.ts

@@ -810,7 +810,20 @@ export async function checkSiteHealth(
   let message: string | null = null;
   let detail: SiteHealthDetail | null = null;
   const def = getSiteDefinition(site.key);
-  if (!def) {
+  if (site.kind === "subtitle") {
+    // 字幕站:用站点出口(代理 / UA / Cloudflare 处理)打开首页,2xx 即可用
+    const t0 = Date.now();
+    try {
+      const res = await rt.siteFetch(site)(site.base_url, { signal: AbortSignal.timeout(20_000) });
+      await res.body?.cancel();
+      ok = res.ok;
+      status = res.ok ? "ok" : "unreachable";
+      latencyMs = Date.now() - t0;
+      message = res.ok ? null : `首页返回 HTTP ${res.status}`;
+    } catch (e) {
+      message = (e as Error).message.slice(0, 300);
+    }
+  } else if (!def) {
     message = "站点没有适配器定义，无法检测";
     status = "unsupported";
   } else {

apps/server/src/modules/subtitles/index.ts

@@ -1,6 +1,9 @@
-import { Api, type SubtitleApplyRequest } from "@umt/contracts";
+import { Api, type SubtitleApplyRequest, type SubtitleFetchRequest, type SubtitleSource } from "@umt/contracts";
+import { ApiException } from "../../http/errors.ts";
 import { route } from "../agent/http.ts";
+import { sourcesIdentityModule } from "../sources-identity/index.ts";
 import type { ServerModule } from "../types.ts";
+import { downloadOnline, SubtitleSiteError, searchOnline, subtitleSearchQuery } from "./online/index.ts";
 import { applySubtitles, parseSubtitles, subtitleTargets, subtitleWrites, undoSubtitleWrite } from "./service.ts";
 
 /** 补字幕(docs/plan/subtitles.md):上传 / 在线下载的字幕解析、配对、写入、历史与撤销 */
@@ -9,8 +12,39 @@ export function createSubtitlesModule(): ServerModule {
     name: "subtitles",
     registerRoutes(app, ctx) {
       const tag = { tag: "subtitles" };
+      const rt = () => sourcesIdentityModule.runtime(ctx);
+      const siteErr = async <T>(fn: () => Promise<T>): Promise<T> => {
+        try {
+          return await fn();
+        } catch (e) {
+          if (e instanceof SubtitleSiteError) throw new ApiException("upstream_unavailable", e.message);
+          throw e;
+        }
+      };
       route(app, Api.subtitles.targets, ({ query }) => subtitleTargets(ctx, query.entityId), tag);
       route(app, Api.subtitles.parse, ({ body }) => parseSubtitles(ctx, body.files, body.entityId), tag);
+      route(
+        app,
+        Api.subtitles.search,
+        ({ body }) => siteErr(async () => searchOnline(ctx, rt(), await subtitleSearchQuery(ctx, body.entityId, body))),
+        tag,
+      );
+      route(
+        app,
+        Api.subtitles.fetch,
+        ({ body }) =>
+          siteErr(async () => {
+            const b = body as SubtitleFetchRequest;
+            const f = await downloadOnline(ctx, rt(), b.siteId, b.ref);
+            const parsed = await parseSubtitles(ctx, [{ name: f.name, base64: f.base64 }], b.entityId);
+            // 在线来源:每项带上字幕站与条目地址,写入时记进历史
+            return {
+              ...parsed,
+              items: parsed.items.map((it) => ({ ...it, source: f.siteKey as SubtitleSource, sourceRef: f.pageUrl })),
+            };
+          }),
+        tag,
+      );
       route(
         app,
         Api.subtitles.apply,

apps/server/src/modules/subtitles/online/index.ts

@@ -0,0 +1,359 @@
+// 在线字幕(docs/plan/subtitles.md 第 2 步):在已启用的字幕站(site.kind = subtitle)搜索、下载。
+// 请求走站点出口 rt.siteFetch(站点代理、UA、Cloudflare 挑战走浏览器);页面解析与验证码识别在沙箱里(ADR-005)。
+// 流程对照 MP 插件 online_subtitles/providers/{subhd,zimuku}.py,页面结构 2026-10-03 实测。
+import type { SubtitleSearchResponse, SubtitleSearchResult } from "@umt/contracts";
+import type { SiteRow } from "@umt/db";
+import { subtitleEpisode } from "@umt/domain";
+import { runSandbox } from "../../../sandbox/client.ts";
+import { b64encode } from "../../../sandbox/jobs.ts";
+import type { SourcesRuntime } from "../../sources-identity/runtime.ts";
+import type { ServerContext } from "../../types.ts";
+import type { PageEntry, PageKind, PageResult } from "./parse.ts";
+
+const MAX_DOWNLOAD_BYTES = 30 * 1024 * 1024;
+
+/** 站内操作失败(给用户看的原因) */
+export class SubtitleSiteError extends Error {
+  override name = "SubtitleSiteError";
+}
+
+// ── 请求:站点出口 + 每站一个 Cookie 罐(字幕库的验证 Cookie、SubHD 的会话) ──
+
+const jars = new Map<string, Map<string, string>>();
+
+interface Env {
+  rt: SourcesRuntime;
+  site: SiteRow;
+}
+
+async function request(
+  env: Env,
+  url: string,
+  init: { referer?: string; json?: unknown; signal?: AbortSignal } = {},
+): Promise<Response> {
+  const jar = jars.get(env.site.id) ?? new Map<string, string>();
+  jars.set(env.site.id, jar);
+  const headers: Record<string, string> = { Accept: "text/html,application/json;q=0.9,*/*;q=0.8" };
+  if (jar.size) headers.cookie = [...jar].map(([k, v]) => `${k}=${v}`).join("; ");
+  if (init.referer) headers.referer = init.referer;
+  if (init.json !== undefined) headers["content-type"] = "application/json";
+  let res: Response;
+  try {
+    res = await env.rt.siteFetch(env.site)(url, {
+      method: init.json === undefined ? "GET" : "POST",
+      headers,
+      ...(init.json !== undefined ? { body: JSON.stringify(init.json) } : {}),
+      signal: init.signal ?? AbortSignal.timeout(30_000),
+    });
+  } catch (e) {
+    const n = (e as Error).name;
+    if (n === "TimeoutError" || n === "AbortError")
+      throw new SubtitleSiteError(`连接${env.site.name}超时，稍后再试（站点可能在限流或网络不通）`);
+    throw new SubtitleSiteError(`连接${env.site.name}失败：${(e as Error).message.slice(0, 200)}`);
+  }
+  for (const c of res.headers.getSetCookie()) {
+    const kv = c.split(";")[0] ?? "";
+    const i = kv.indexOf("=");
+    if (i > 0) jar.set(kv.slice(0, i).trim(), kv.slice(i + 1).trim());
+  }
+  return res;
+}
+
+async function getText(env: Env, url: string, referer?: string): Promise<string> {
+  const res = await request(env, url, referer ? { referer } : {});
+  if (!res.ok) throw new SubtitleSiteError(`${env.site.name} 返回 HTTP ${res.status}`);
+  return await res.text();
+}
+
+const page = <K extends PageKind>(kind: K, html: string) =>
+  runSandbox("subtitle.page", { kind, html }, { timeoutMs: 15_000 }) as Promise<PageResult<K>>;
+
+const abs = (env: Env, path: string) => new URL(path, env.site.base_url).href;
+
+/** content-disposition 里的文件名:常见是 UTF-8 字节被当成 latin1 发出,按字节还原 */
+function fileNameOf(res: Response, url: string): string {
+  const cd = res.headers.get("content-disposition") ?? "";
+  const star = /filename\*=UTF-8''([^;]+)/i.exec(cd)?.[1];
+  if (star) return decodeURIComponent(star);
+  const plain = /filename="?([^";]+)"?/i.exec(cd)?.[1];
+  if (plain) {
+    const bytes = Uint8Array.from(plain, (c) => c.charCodeAt(0) & 0xff);
+    const utf8 = new TextDecoder("utf-8", { fatal: false }).decode(bytes);
+    return utf8.includes("�") ? plain : utf8;
+  }
+  return decodeURIComponent(new URL(url).pathname.split("/").pop() || "subtitle.zip");
+}
+
+async function getFile(env: Env, url: string, referer: string): Promise<{ name: string; bytes: Uint8Array }> {
+  const res = await request(env, url, { referer });
+  if (!res.ok) throw new SubtitleSiteError(`下载返回 HTTP ${res.status}`);
+  const len = Number(res.headers.get("content-length") ?? 0);
+  if (len > MAX_DOWNLOAD_BYTES) throw new SubtitleSiteError("字幕包超过 30 MB");
+  const bytes = new Uint8Array(await res.arrayBuffer());
+  if (bytes.length > MAX_DOWNLOAD_BYTES) throw new SubtitleSiteError("字幕包超过 30 MB");
+  const head = new TextDecoder().decode(bytes.subarray(0, 200)).toLowerCase();
+  if (/<!doctype html|<html/.test(head)) throw new SubtitleSiteError("下载地址返回的是网页，不是字幕文件");
+  return { name: fileNameOf(res, res.url || url), bytes };
+}
+
+// ── 搜索词与匹配 ──
+
+export interface SearchQuery {
+  keywords: string[];
+  doubanId: string | null;
+  season: number | null;
+  episode: number | null;
+}
+
+const norm = (s: string) => s.toLowerCase().replace(/[\s\p{P}\p{S}]/gu, "");
+
+/** 标题里含任一搜索词(去掉空白与标点比较) */
+function relevant(title: string, keywords: string[]): boolean {
+  const t = norm(title);
+  return keywords.some((k) => {
+    const n = norm(k);
+    return n.length >= 2 && t.includes(n);
+  });
+}
+
+/** 季集不符的丢掉;合集 / 季包保留 */
+function episodeOk(title: string, q: SearchQuery): { ok: boolean; season: number | null; episode: number | null } {
+  const ep = subtitleEpisode(`${title}.srt`);
+  if (q.season !== null && !ep.seasonInferred && ep.season !== null && ep.season !== q.season)
+    return { ok: false, season: ep.season, episode: ep.episode };
+  if (q.episode !== null && ep.episode !== null && ep.episode !== q.episode)
+    return { ok: false, season: ep.season, episode: ep.episode };
+  return { ok: true, season: ep.seasonInferred ? null : ep.season, episode: ep.episode };
+}
+
+const LANG_SCORE: Record<string, number> = { 双语: 3, 简体: 2, 繁体: 1 };
+
+function toResult(env: Env, e: PageEntry, q: SearchQuery): SubtitleSearchResult | null {
+  const ep = episodeOk(e.title, q);
+  if (!ep.ok) return null;
+  const lang = Math.max(0, ...e.languages.map((l) => LANG_SCORE[l] ?? 0));
+  const exact = q.episode !== null && ep.episode === q.episode ? 2 : 0;
+  return {
+    siteId: env.site.id,
+    siteKey: env.site.key,
+    siteName: env.site.name,
+    ref: e.ref,
+    title: e.title,
+    pageUrl: abs(env, e.ref),
+    languages: e.languages,
+    format: e.format,
+    downloads: e.downloads,
+    uploader: e.uploader,
+    publishedAt: e.publishedAt,
+    season: ep.season,
+    episode: ep.episode,
+    score: lang * 10 + exact * 5 + Math.log10((e.downloads ?? 0) + 1),
+  };
+}
+
+// ── SubHD ──
+
+async function subhdSearch(env: Env, q: SearchQuery): Promise<PageEntry[]> {
+  // SubHD 的条目 ID 就是豆瓣条目 ID:有豆瓣 ID 直接打开条目页
+  if (q.doubanId) {
+    const d = await page("subhd.list", await getText(env, abs(env, `/d/${q.doubanId}`), env.site.base_url));
+    if (d.entries.length) return d.entries;
+  }
+  for (const kw of q.keywords.slice(0, 3)) {
+    const html = await getText(env, abs(env, `/search/${encodeURIComponent(kw)}`), env.site.base_url);
+    const r = await page("subhd.list", html);
+    const hits = r.entries.filter((e) => relevant(e.title, q.keywords));
+    if (hits.length) return hits;
+    for (const d of r.details.filter((x) => relevant(x.title, q.keywords)).slice(0, 2)) {
+      const dd = await page("subhd.list", await getText(env, abs(env, `/d/${d.id}`), env.site.base_url));
+      if (dd.entries.length) return dd.entries;
+    }
+  }
+  return [];
+}
+
+async function subhdDownload(env: Env, ref: string): Promise<{ name: string; bytes: Uint8Array }> {
+  const subUrl = abs(env, ref);
+  const sub = await page("subhd.sub", await getText(env, subUrl, env.site.base_url));
+  let down = sub.down;
+  if (!down && sub.dataSid) {
+    const prep = (await (
+      await request(env, abs(env, "/api/sub/prepare-download"), { referer: subUrl, json: { sid: sub.dataSid } })
+    ).json()) as { success?: boolean; url?: string };
+    if (prep.success && prep.url?.includes("/down/")) down = prep.url;
+  }
+  if (!down) throw new SubtitleSiteError("SubHD 字幕页没有下载按钮");
+  const downUrl = abs(env, down);
+  const sid = new URL(downUrl).pathname.split("/").pop() ?? "";
+  await (await request(env, downUrl, { referer: subUrl })).body?.cancel();
+  const api = async (cap: string) =>
+    (await (await request(env, abs(env, "/api/sub/down"), { referer: downUrl, json: { sid, cap } })).json()) as {
+      success?: boolean;
+      pass?: boolean;
+      msg?: string;
+      url?: string;
+    };
+  let r = await api("");
+  if (r.pass === false) {
+    // SVG 文字验证码:沙箱里识别后带上再提交一次
+    const { code } = await runSandbox("captcha.solve", { kind: "subhd", svg: String(r.msg ?? "") });
+    r = await api(code ?? "");
+  }
+  if (!r.success || !r.url)
+    throw new SubtitleSiteError("SubHD 下载验证没通过：可在站点的原站视图里打开这条字幕手动下载后上传");
+  return await getFile(env, abs(env, r.url), downUrl);
+}
+
+// ── 字幕库 ──
+
+/** 打开页面;遇到云锁验证码页就在沙箱里识别、提交验证后重开(最多 3 次) */
+async function zimukuPage(env: Env, url: string, referer?: string): Promise<string> {
+  for (let i = 0; i < 3; i++) {
+    const res = await request(env, url, referer ? { referer } : {});
+    const html = await res.text();
+    const cap = await page("zimuku.captcha", html);
+    if (!cap.bmpBase64) {
+      if (!res.ok) throw new SubtitleSiteError(`${env.site.name} 返回 HTTP ${res.status}`);
+      return html;
+    }
+    const { code } = await runSandbox("captcha.solve", { kind: "zimuku", bmpBase64: cap.bmpBase64 });
+    if (!code) break;
+    const hex = [...code].map((c) => c.charCodeAt(0).toString(16)).join("");
+    await (await request(env, `${url}${url.includes("?") ? "&" : "?"}security_verify_img=${hex}`)).body?.cancel();
+  }
+  throw new SubtitleSiteError("字幕库的验证码没能自动通过：可在站点的原站视图里打开一次再试");
+}
+
+async function zimukuSearch(env: Env, q: SearchQuery): Promise<PageEntry[]> {
+  for (const kw of q.keywords.slice(0, 3)) {
+    const s = await page("zimuku.search", await zimukuPage(env, abs(env, `/search?q=${encodeURIComponent(kw)}`)));
+    const subs = s.subs.filter((x) => !x.title || relevant(x.title, q.keywords)).slice(0, 3);
+    const out: PageEntry[] = [];
+    for (const x of subs) {
+      const p = await page("zimuku.subs", await zimukuPage(env, abs(env, x.path), env.site.base_url));
+      out.push(...p.entries);
+    }
+    if (out.length) return out;
+  }
+  return [];
+}
+
+async function zimukuDownload(env: Env, ref: string): Promise<{ name: string; bytes: Uint8Array }> {
+  const detailUrl = abs(env, ref);
+  const d = await page("zimuku.detail", await zimukuPage(env, detailUrl, env.site.base_url));
+  if (!d.dl) throw new SubtitleSiteError("字幕库字幕页没有下载入口");
+  const dlUrl = new URL(d.dl, detailUrl).href;
+  const links = (await page("zimuku.dl", await zimukuPage(env, dlUrl, detailUrl))).links;
+  let last = "字幕库没有可用的下载线路";
+  for (const l of links) {
+    try {
+      const f = await getFile(env, new URL(l, dlUrl).href, dlUrl);
+      if (f.bytes.length > 64) return f;
+    } catch (e) {
+      last = (e as Error).message;
+    }
+  }
+  throw new SubtitleSiteError(last);
+}
+
+const PROVIDERS: Record<
+  string,
+  {
+    search(env: Env, q: SearchQuery): Promise<PageEntry[]>;
+    download(env: Env, ref: string): Promise<{ name: string; bytes: Uint8Array }>;
+  }
+> = {
+  subhd: { search: subhdSearch, download: subhdDownload },
+  zimuku: { search: zimukuSearch, download: zimukuDownload },
+};
+
+// ── 对外 ──
+
+export async function subtitleSearchQuery(
+  ctx: ServerContext,
+  entityId: string,
+  opts: { season?: number | undefined; episode?: number | undefined; keyword?: string | undefined },
+): Promise<SearchQuery> {
+  const e = await ctx.db
+    .selectFrom("media_entity")
+    .select(["title", "original_title"])
+    .where("id", "=", entityId)
+    .executeTakeFirst();
+  if (!e) throw new SubtitleSiteError("作品不存在");
+  const aliases = await ctx.db
+    .selectFrom("title_alias")
+    .select("alias")
+    .where("media_entity_id", "=", entityId)
+    .where("content_unit_id", "is", null)
+    .limit(5)
+    .execute();
+  const douban = await ctx.db
+    .selectFrom("external_identity")
+    .select("external_id")
+    .where("media_entity_id", "=", entityId)
+    .where("provider", "=", "douban")
+    .where("status", "=", "accepted")
+    .executeTakeFirst();
+  const keywords = [opts.keyword, e.title, e.original_title, ...aliases.map((a) => a.alias)]
+    .map((x) => x?.trim())
+    .filter((x): x is string => !!x && x.length >= 2);
+  return {
+    keywords: [...new Set(keywords)],
+    doubanId: douban?.external_id ?? null,
+    season: opts.season ?? null,
+    episode: opts.episode ?? null,
+  };
+}
+
+async function subtitleSites(ctx: ServerContext): Promise<SiteRow[]> {
+  return await ctx.db
+    .selectFrom("site")
+    .selectAll()
+    .where("kind", "=", "subtitle")
+    .where("enabled", "=", true)
+    .where("deleted_at", "is", null)
+    .orderBy("name")
+    .execute();
+}
+
+export async function searchOnline(
+  ctx: ServerContext,
+  rt: SourcesRuntime,
+  q: SearchQuery,
+): Promise<SubtitleSearchResponse> {
+  const sites = (await subtitleSites(ctx)).filter((s) => PROVIDERS[s.key]);
+  const results: SubtitleSearchResult[] = [];
+  const errors: SubtitleSearchResponse["errors"] = [];
+  await Promise.all(
+    sites.map(async (site) => {
+      const env = { rt, site };
+      try {
+        const entries = await PROVIDERS[site.key]!.search(env, q);
+        const seen = new Set<string>();
+        for (const e of entries) {
+          if (seen.has(e.ref)) continue;
+          seen.add(e.ref);
+          const r = toResult(env, e, q);
+          if (r) results.push(r);
+        }
+      } catch (e) {
+        errors.push({ siteName: site.name, reason: (e as Error).message.slice(0, 300) });
+      }
+    }),
+  );
+  results.sort((a, b) => b.score - a.score);
+  return { results: results.slice(0, 60), keywords: q.keywords, errors, noSites: sites.length === 0 };
+}
+
+export async function downloadOnline(
+  ctx: ServerContext,
+  rt: SourcesRuntime,
+  siteId: string,
+  ref: string,
+): Promise<{ name: string; base64: string; siteKey: string; pageUrl: string }> {
+  const site = (await subtitleSites(ctx)).find((s) => s.id === siteId);
+  if (!site || !PROVIDERS[site.key]) throw new SubtitleSiteError("字幕站不存在或已停用");
+  const env = { rt, site };
+  const f = await PROVIDERS[site.key]!.download(env, ref);
+  return { name: f.name, base64: b64encode(f.bytes), siteKey: site.key, pageUrl: abs(env, ref) };
+}

apps/server/src/modules/subtitles/online/parse.ts

@@ -0,0 +1,169 @@
+// 字幕站页面解析(纯函数,在沙箱任务 subtitle.page 里跑):SubHD / 字幕库的搜索页、条目页、字幕页、下载页 → 结构化数据。
+// 页面结构 2026-10-03 实测(样本在 .local/site-samples/subs/),对照 MP 插件 online_subtitles/providers/{subhd,zimuku}.py。
+
+export interface PageEntry {
+  /** 站内引用:SubHD 为 /a/<sid>,字幕库为 /detail/<id>.html */
+  ref: string;
+  title: string;
+  /** 简体 / 繁体 / 双语 / 英文 …(站点标注或从标题、说明里猜) */
+  languages: string[];
+  format: string | null;
+  downloads: number | null;
+  uploader: string | null;
+  publishedAt: string | null;
+}
+
+const strip = (s: string) =>
+  s
+    .replace(/<[^>]+>/g, " ")
+    .replace(/&nbsp;/g, " ")
+    .replace(/&amp;/g, "&")
+    .replace(/&lt;/g, "<")
+    .replace(/&gt;/g, ">")
+    .replace(/&#39;/g, "'")
+    .replace(/&quot;/g, '"')
+    .replace(/\s+/g, " ")
+    .trim();
+const href = (q: string) => `href=["']${q}["']`;
+
+/** 从标题 / 说明里猜语言 */
+export function guessLanguages(text: string): string[] {
+  const t = text.toLowerCase();
+  const out: string[] = [];
+  if (/双语|中英|简英|繁英|chs&eng|cht&eng|bilingual/.test(t)) out.push("双语");
+  if (/简体|简中|chs|gb|简繁|简日/.test(t)) out.push("简体");
+  if (/繁体|繁體|繁中|cht|big5|简繁/.test(t)) out.push("繁体");
+  if (/英文|英语|\beng?\b|english/.test(t) && !out.includes("双语")) out.push("英文");
+  return [...new Set(out)];
+}
+
+const formatOf = (text: string) => /\b(ass|ssa|srt|sup|vtt|sub|idx)\b/i.exec(text)?.[1]?.toUpperCase() ?? null;
+const num = (s: string | undefined) => {
+  if (!s) return null;
+  const m = /([\d.]+)\s*(万)?/.exec(s.replace(/,/g, ""));
+  return m ? Math.round(Number(m[1]) * (m[2] ? 10_000 : 1)) : null;
+};
+
+// ── SubHD ──
+
+/** 搜索页 / 条目页(/d/<豆瓣 ID>):结果区里的字幕(/a/<sid>),不含侧栏热门 */
+function subhdList(html: string): { entries: PageEntry[]; details: { id: string; title: string }[] } {
+  const hot = html.search(/<h\d[^>]*>\s*热门|class="[^"]*hot[^"]*"/);
+  const body = hot > 0 ? html.slice(0, hot) : html;
+  const entries: PageEntry[] = [];
+  const seen = new Set<string>();
+  for (const m of body.matchAll(new RegExp(`<a[^>]+${href("/a/([A-Za-z0-9_-]+)")}[^>]*>([\\s\\S]*?)</a>`, "g"))) {
+    const sid = m[1]!;
+    const title = strip(m[2]!);
+    if (seen.has(sid) || !title) continue;
+    seen.add(sid);
+    const around = strip(body.slice(Math.max(0, m.index! - 600), m.index! + 1200));
+    entries.push({
+      ref: `/a/${sid}`,
+      title,
+      languages: guessLanguages(`${title} ${around}`),
+      format: formatOf(`${title} ${around}`),
+      downloads: null,
+      uploader: /发布人\s*(\S+)/.exec(around)?.[1] ?? null,
+      publishedAt: /<time datetime="([^"]+)"/.exec(body.slice(m.index!, m.index! + 2500))?.[1] ?? null,
+    });
+  }
+  const details: { id: string; title: string }[] = [];
+  for (const m of body.matchAll(new RegExp(`<a[^>]+${href("/d/(\\d+)")}[^>]*>([\\s\\S]*?)</a>`, "g"))) {
+    const alt = /alt=["']([^"']+)["']/.exec(m[2]!)?.[1];
+    const title = strip(m[2]!) || alt || "";
+    if (title && !details.some((d) => d.id === m[1])) details.push({ id: m[1]!, title });
+  }
+  return { entries, details };
+}
+
+/** 字幕页(/a/<sid>):下载按钮 /down/<sid> 或 data-sid(要先 prepare-download) */
+function subhdSub(html: string): { down: string | null; dataSid: string | null } {
+  return {
+    down: new RegExp(`${href("([^\"']*/down/[A-Za-z0-9_-]+[^\"']*)")}`).exec(html)?.[1] ?? null,
+    dataSid: /\bdata-sid=["']([A-Za-z0-9_-]+)["']/.exec(html)?.[1] ?? null,
+  };
+}
+
+// ── 字幕库 ──
+
+/** 字幕库语言旗帜的说明(「简体中文字幕」「双语字幕」…)→ 统一叫法 */
+function langLabel(t: string): string {
+  if (/双语/.test(t)) return "双语";
+  if (/繁/.test(t)) return "繁体";
+  if (/简/.test(t)) return "简体";
+  if (/英/.test(t)) return "英文";
+  if (/日/.test(t)) return "日文";
+  if (/韩|韓/.test(t)) return "韩文";
+  return t.replace(/字幕$/, "");
+}
+
+function zimukuSearch(html: string): { subs: { path: string; title: string }[] } {
+  const subs: { path: string; title: string }[] = [];
+  for (const m of html.matchAll(/<a[^>]+href=["']([^"']*\/subs\/\d+\.html)["'][^>]*>([\s\S]*?)<\/a>/g)) {
+    // 链接可能写成 //zimuku.org/subs/…(别的域名):只取路径,拼到站点地址上
+    const path = /\/subs\/\d+\.html/.exec(m[1]!)![0];
+    const title =
+      strip(m[2]!) || /title=["']([^"']+)["']/.exec(m[0])?.[1] || /alt=["']([^"']+)["']/.exec(m[2]!)?.[1] || "";
+    const prev = subs.find((s) => s.path === path);
+    if (!prev) subs.push({ path, title });
+    else if (!prev.title && title) prev.title = title;
+  }
+  return { subs };
+}
+
+/** 条目页(/subs/<id>.html)的字幕表 */
+function zimukuSubs(html: string): { entries: PageEntry[] } {
+  const entries: PageEntry[] = [];
+  for (const row of html.matchAll(/<tr[^>]*>([\s\S]*?)<\/tr>/g)) {
+    const r = row[1]!;
+    const a = /<a href=["'](\/detail\/\d+\.html)["'][^>]*title=["']([^"']+)["']/.exec(r);
+    if (!a) continue;
+    const langs = [...r.matchAll(/<img[^>]+title=["']([^"']+字幕)["']/g)].map((m) => langLabel(m[1]!));
+    const cells = [...r.matchAll(/<td[^>]*>([\s\S]*?)<\/td>/g)].map((m) => m[1]!);
+    entries.push({
+      ref: a[1]!,
+      title: strip(a[2]!),
+      languages: langs.length ? [...new Set(langs)] : guessLanguages(a[2]!),
+      format: /<span class="label label-info">([^<]+)<\/span>/.exec(r)?.[1]?.trim() ?? formatOf(a[2]!),
+      downloads: num(strip(cells[4] ?? "")),
+      uploader: /<a href=["']\/u\/[^"']+["']>([^<]+)<\/a>/.exec(r)?.[1] ?? null,
+      publishedAt: strip(/glyphicon-time"><\/span>([^<]+)</.exec(r)?.[1] ?? "") || null,
+    });
+  }
+  return { entries };
+}
+
+/** 字幕页:下载入口(li.dlsub 里的链接) */
+function zimukuDetail(html: string): { dl: string | null } {
+  return {
+    dl: /<li[^>]+class=["'][^"']*dlsub[^"']*["'][^>]*>[\s\S]*?<a[^>]+href=["']([^"']+)["']/.exec(html)?.[1] ?? null,
+  };
+}
+
+/** 下载跳转页:各线路的 /download/ 链接 */
+function zimukuDl(html: string): { links: string[] } {
+  return { links: [...new Set([...html.matchAll(/href=["']([^"']*\/download\/[^"']+)["']/g)].map((m) => m[1]!))] };
+}
+
+/** 是否是字幕库(云锁)的验证码页;是则带回 BMP */
+function zimukuCaptcha(html: string): { bmpBase64: string | null } {
+  return {
+    bmpBase64:
+      /class=["']verifyimg["'][^>]*src=["']data:image\/bmp;base64,([^"']+)["']/.exec(html)?.[1] ??
+      /data:image\/bmp;base64,([^"']+)/.exec(html)?.[1] ??
+      null,
+  };
+}
+
+export const PAGE_PARSERS = {
+  "subhd.list": subhdList,
+  "subhd.sub": subhdSub,
+  "zimuku.search": zimukuSearch,
+  "zimuku.subs": zimukuSubs,
+  "zimuku.detail": zimukuDetail,
+  "zimuku.dl": zimukuDl,
+  "zimuku.captcha": zimukuCaptcha,
+} as const;
+export type PageKind = keyof typeof PAGE_PARSERS;
+export type PageResult<K extends PageKind> = ReturnType<(typeof PAGE_PARSERS)[K]>;

apps/server/src/modules/subtitles/service.ts

@@ -269,7 +269,7 @@ export async function applySubtitles(
             library_root_id: t.libraryRootId,
             relative_path: rel,
             language: it.language,
-            source: body.source ?? "upload",
+            source: it.source ?? body.source ?? "upload",
             source_name: it.name,
             source_ref: it.sourceRef ?? null,
             backup_path: backup,