feat(scrape): 任意目录就地刮削,按路径识别作品,每次刮削一条记录(SCR-2)

- POST /files/scrape 改为对目录 / 视频文件刮削:已登记用作品,否则按路径识别(本地 + TMDB),
  认不准返回候选,选定后再提交;未登记直接用 TMDB 字段,addToLibrary 顺手建作品
- 作品详情重新刮削、入库后自动刮削改为对作品目录刮削,同一实现
- 迁移 0058_scrape_record 替代 media_scrape;GET /files/scrape-records
- 与 MP ScrapingChain 的文件布局差分(7 个场景,3 处登记偏离)

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

apps/server/test/file-manager/harness.ts

@@ -22,6 +22,7 @@ import { EventBus } from "../../src/http/sse.ts";
 import { createLogger } from "../../src/logger.ts";
 import { fakeEngine } from "../../src/modules/acquisition-organize/__tests__/harness.ts";
 import { executeRecord, type FrozenPlan } from "../../src/modules/acquisition-organize/services/organize.ts";
+import { runScrapeRecord } from "../../src/modules/acquisition-organize/services/scrape.ts";
 import { createCapabilityRegistry } from "../../src/modules/agent/capability-registry.ts";
 import { createFileManagerModule, type FmOverrides } from "../../src/modules/file-manager/index.ts";
 import type { ServerContext } from "../../src/modules/types.ts";
@@ -57,6 +58,17 @@ export async function pgAvailable() {
 // biome-ignore lint/suspicious/noExplicitAny: 测试断言直接读响应 JSON
 type Loose = any;
 
+/** catalog.lookup 远端候选(测试用) */
+export interface RemoteHit {
+  provider: string;
+  entityType: "movie" | "tv";
+  externalId: string;
+  title: string;
+  originalTitle?: string | null;
+  year: number | null;
+  posterPath?: string | null;
+}
+
 export interface FmHarness {
   db: Kysely<DB>;
   engine: ReturnType<typeof fakeEngine>;
@@ -80,7 +92,11 @@ export async function createFmHarness(
   name: string,
   overrides: FmOverrides = {},
   /** runTransfers(默认 true):假引擎收到 /files 移动/复制的 verifyAndImport 时就地执行(模拟 worker) */
-  opts: { catalog?: boolean; runTransfers?: boolean } = {},
+  opts: {
+    /** true = 内置假 catalog.lookup("Remote Film");函数 = 自定义远端候选(按查询词返回) */
+    catalog?: boolean | ((query: string) => RemoteHit[]);
+    runTransfers?: boolean;
+  } = {},
 ): Promise<FmHarness> {
   const schema = `fm_test_${name}_${process.pid.toString(36)}${Math.random().toString(36).slice(2, 7)}`;
   const admin = new pg.Client({ connectionString: DB_URL });
@@ -124,9 +140,12 @@ export async function createFmHarness(
   if (opts.catalog)
     capabilities.register(capabilitySpec("catalog.lookup"), async (input) => ({
       local: [],
-      remote: String(input.query).includes("Remote")
-        ? [{ provider: "tmdb", entityType: "movie" as const, externalId: "999", title: "Remote Film", year: 2019 }]
-        : [],
+      remote:
+        typeof opts.catalog === "function"
+          ? opts.catalog(String(input.query))
+          : String(input.query).includes("Remote")
+            ? [{ provider: "tmdb", entityType: "movie" as const, externalId: "999", title: "Remote Film", year: 2019 }]
+            : [],
     }));
   const ctx: ServerContext = {
     role: "all",
@@ -173,6 +192,9 @@ export async function createFmHarness(
     engine.enqueue = async (task, input, o) => {
       const r = await enqueue(task, input, o);
       const recordId = (input as { recordId?: string }).recordId;
+      // SCR-2 刮削:模拟 worker 就地执行 scrapeMedia
+      if (task === "scrapeMedia" && recordId)
+        setTimeout(() => void runScrapeRecord(module.deps().ao, recordId).catch(() => {}), 0);
       if (task === "verifyAndImport" && recordId)
         setTimeout(async () => {
           try {

apps/server/test/file-manager/identify-organize.test.ts

@@ -14,7 +14,7 @@ import {
 } from "./harness.ts";
 
 // identify(只读,本地 + 远端召回前 5,不落库)、organize(按作品分组走 planImport,dryRun→execute 一致与源变更)、
-// probe(已登记写回 media_info)、scrape(仅已登记作品,入队 refreshMetadata;NFO 待 W5)。
+// probe(已登记写回 media_info)、刷新元数据(scrape-batch,入队 refreshMetadata)。
 
 const skip = !(await pgAvailable());
 
@@ -278,7 +278,7 @@ describe.skipIf(skip)("files 识别 / 整理 / 探测 / 刮削", () => {
     expect(r.status).toBeLessThan(500);
   });
 
-  it("probe:已登记资产写回 media_info;scrape 仅已登记作品", async () => {
+  it("probe:已登记资产写回 media_info;刷新元数据按作品入队", async () => {
     await mkdir(join(h.lib, "Blue Film probe"));
     await writeFile(join(h.lib, "Blue Film probe", "Blue.Film.2021.mkv"), "x");
     const a = await seedAsset(h.db, h, { entityId: film, rel: "Blue Film probe/Blue.Film.2021.mkv" });
@@ -303,21 +303,10 @@ describe.skipIf(skip)("files 识别 / 整理 / 探测 / 刮削", () => {
     const u2 = await h.api("POST", "/files/probe", { items: [{ root: h.dlRoot, path: "Remote.Film.2019.1080p.mkv" }] });
     expect(u2.body.data.items[0].cached).toBe(true);
 
-    const s = await h.api("POST", "/files/scrape", { idempotencyKey: key(), mediaEntityId: film, writeNfo: true });
-    expect(s.status).toBe(200);
-    // SCR-1:writeNfo 另排 scrapeMedia(写 NFO 与图片到媒体库目录)
-    expect(s.body.data).toMatchObject({ mediaEntityId: film, nfo: "queued", images: "not_requested" });
-    expect(s.body.data.scrapeTaskId).toBeTruthy();
-    expect(h.engine.enqueued.some((e) => e.task === "refreshMetadata")).toBe(true);
-    expect(h.engine.enqueued.some((e) => e.task === "scrapeMedia")).toBe(true);
-    const b = await h.api("POST", "/files/scrape-batch", {
-      idempotencyKey: key(),
-      mediaEntityIds: [film],
-      write: true,
-    });
+    // 刷新元数据(scrape-batch):按作品入队 refreshMetadata;刮削(写 NFO)见 scrape.test.ts
+    const b = await h.api("POST", "/files/scrape-batch", { idempotencyKey: key(), mediaEntityIds: [film] });
     expect(b.status).toBe(200);
     expect(b.body.data.submitted).toHaveLength(1);
-    const none = await h.api("POST", "/files/scrape", { idempotencyKey: key(), mediaEntityId: show });
-    expect(none.status).toBe(400);
+    expect(h.engine.enqueued.some((e) => e.task === "refreshMetadata")).toBe(true);
   });
 });

apps/server/test/file-manager/scrape.test.ts

@@ -0,0 +1,443 @@
+import { mkdir, readdir, readFile, stat, writeFile } from "node:fs/promises";
+import { join, relative } from "node:path";
+import type { TmdbRichDetail, TmdbRichSeasonDetail } from "@umt/adapters";
+import { afterAll, beforeAll, describe, expect, it } from "vitest";
+import { latestEntityScrape, scrapeEntity } from "../../src/modules/acquisition-organize/services/scrape.ts";
+import {
+  createFmHarness,
+  EXEC_KEY,
+  type FmHarness,
+  key,
+  pgAvailable,
+  type RemoteHit,
+  seedAsset,
+  seedEntity,
+} from "./harness.ts";
+
+// SCR-2 刮削照 MP:文件管理任意目录 / 视频文件 → 按路径识别作品(未登记走 TMDB,认不准返回候选)→ 就地写 NFO 与图片。
+// 场景:未登记的《人生切割术》S01 两集 + 《星际穿越》目录一次刮削;歧义候选(Solaris 1972 / 2002)选定后刮削并加入媒体库;
+// 只补缺失 / 全部覆盖;季目录与单集文件的范围;已登记目录直接用作品;只登记目录与非视频文件跳过。
+
+const skip = !(await pgAvailable());
+const JPEG = new Uint8Array([0xff, 0xd8, 0xff, 0xe0, 0x00, 0x10, 0x4a, 0x46, 0x49, 0x46, 0x00]);
+
+function detail(
+  p: Partial<TmdbRichDetail> & Pick<TmdbRichDetail, "entityType" | "externalId" | "title">,
+): TmdbRichDetail {
+  return {
+    titleLanguage: "zh-CN",
+    titleZh: p.title,
+    titleZhLanguage: "zh-CN",
+    originalTitle: null,
+    originalLanguage: "en",
+    overview: null,
+    overviewLanguage: null,
+    tagline: null,
+    year: null,
+    releaseDate: null,
+    lastAirDate: null,
+    genres: [],
+    voteAverage: null,
+    voteCount: null,
+    status: null,
+    runtime: null,
+    seasonCount: null,
+    episodeCount: null,
+    networks: [],
+    originCountry: [],
+    posterPath: null,
+    backdropPath: null,
+    cast: [],
+    crew: [],
+    externalIds: { imdb: null, tvdb: null, wikidata: null },
+    translations: [],
+    alternativeTitles: [],
+    seasons: [],
+    languages: ["zh-CN"],
+    ...p,
+  };
+}
+
+const DETAILS: Record<string, TmdbRichDetail> = {
+  "tv:95396": detail({
+    entityType: "tv",
+    externalId: "95396",
+    title: "人生切割术",
+    originalTitle: "Severance",
+    year: 2022,
+    releaseDate: "2022-02-17",
+    overview: "马克领导着一个办公室团队。",
+    overviewLanguage: "zh-CN",
+    genres: ["剧情", "悬疑"],
+    voteAverage: 8.4,
+    networks: ["Apple TV+"],
+    originCountry: ["US"],
+    posterPath: "/sev.jpg",
+    backdropPath: "/sev-bg.jpg",
+    externalIds: { imdb: "tt11280740", tvdb: "371980", wikidata: null },
+  }),
+  "movie:157336": detail({
+    entityType: "movie",
+    externalId: "157336",
+    title: "星际穿越",
+    originalTitle: "Interstellar",
+    year: 2014,
+    releaseDate: "2014-11-05",
+    runtime: 169,
+    posterPath: "/inter.jpg",
+    backdropPath: "/inter-bg.jpg",
+    crew: [{ id: 525, name: "克里斯托弗·诺兰", originalName: null, role: "Director", profilePath: null }],
+  }),
+  "movie:593": detail({
+    entityType: "movie",
+    externalId: "593",
+    title: "飞向太空",
+    originalTitle: "Solaris",
+    year: 1972,
+  }),
+  "movie:2103": detail({
+    entityType: "movie",
+    externalId: "2103",
+    title: "索拉里斯星",
+    originalTitle: "Solaris",
+    year: 2002,
+    posterPath: "/solaris.jpg",
+  }),
+};
+
+const SEASON1: TmdbRichSeasonDetail = {
+  tvId: "95396",
+  seasonNumber: 1,
+  name: "第 1 季",
+  nameLanguage: "zh-CN",
+  overview: null,
+  overviewLanguage: null,
+  airDate: "2022-02-17",
+  posterPath: "/s1.jpg",
+  fetchedLanguages: ["zh-CN"],
+  episodes: [
+    [1, "有关地狱的好消息", "/e1.jpg"],
+    [2, "半循环", "/e2.jpg"],
+  ].map(([n, name, still]) => ({
+    seasonNumber: 1,
+    episodeNumber: n as number,
+    name: name as string,
+    nameLanguage: "zh-CN",
+    overview: `第 ${n} 集简介`,
+    overviewLanguage: "zh-CN",
+    airDate: "2022-02-17",
+    runtime: 55,
+    stillPath: still as string,
+    voteAverage: 8,
+    voteCount: 10,
+  })),
+};
+
+const REMOTE: Record<string, RemoteHit[]> = {
+  人生切割术: [
+    {
+      provider: "tmdb",
+      entityType: "tv",
+      externalId: "95396",
+      title: "人生切割术",
+      originalTitle: "Severance",
+      year: 2022,
+    },
+  ],
+  星际穿越: [
+    {
+      provider: "tmdb",
+      entityType: "movie",
+      externalId: "157336",
+      title: "星际穿越",
+      originalTitle: "Interstellar",
+      year: 2014,
+    },
+  ],
+  Solaris: [
+    {
+      provider: "tmdb",
+      entityType: "movie",
+      externalId: "593",
+      title: "飞向太空",
+      originalTitle: "Solaris",
+      year: 1972,
+    },
+    {
+      provider: "tmdb",
+      entityType: "movie",
+      externalId: "2103",
+      title: "索拉里斯星",
+      originalTitle: "Solaris",
+      year: 2002,
+    },
+  ],
+};
+
+async function tree(root: string, under = ""): Promise<string[]> {
+  const out: string[] = [];
+  const walk = async (d: string) => {
+    for (const n of (await readdir(d)).sort()) {
+      if (n.startsWith(".")) continue;
+      const p = join(d, n);
+      if ((await stat(p)).isDirectory()) await walk(p);
+      else out.push(relative(root, p));
+    }
+  };
+  await walk(join(root, under));
+  return out.sort();
+}
+
+const SHOW = "人生切割术 (2022)";
+const EP1 = `${SHOW}/Season 1/人生切割术 - S01E01 - 有关地狱的好消息.mkv`;
+const EP2 = `${SHOW}/Season 1/人生切割术 - S01E02 - 半循环.mkv`;
+const MOVIE = "星际穿越 (2014)";
+const MOVIE_FILE = `${MOVIE}/星际穿越 (2014) - 1080p.mkv`;
+
+describe.skipIf(skip)("files 刮削(SCR-2:任意目录就地刮削)", () => {
+  let h: FmHarness;
+  const images: string[] = [];
+  const tmdbCalls: string[] = [];
+
+  beforeAll(async () => {
+    h = await createFmHarness(
+      "scrape2",
+      {
+        ao: {
+          async scrapeImage(p) {
+            images.push(p);
+            return { content: JPEG, mime: "image/jpeg" };
+          },
+          async scrapeTmdbDetail(type, id) {
+            tmdbCalls.push(`${type}:${id}`);
+            const d = DETAILS[`${type}:${id}`];
+            if (!d) throw new Error("TMDB 404");
+            return d;
+          },
+          async scrapeTmdbSeason(tvId, n) {
+            tmdbCalls.push(`season:${tvId}:${n}`);
+            return tvId === "95396" && n === 1 ? SEASON1 : null;
+          },
+          async entityFromTmdb(ref) {
+            const d = DETAILS[`${ref.type}:${ref.id}`]!;
+            const e = await h.db
+              .insertInto("media_entity")
+              .values({
+                kind: ref.type === "movie" ? "movie" : "series",
+                title: d.title,
+                year: d.year,
+                status: "active",
+              })
+              .returning("id")
+              .executeTakeFirstOrThrow();
+            await h.db
+              .insertInto("external_identity")
+              .values({
+                media_entity_id: e.id,
+                provider: "tmdb",
+                entity_type: ref.type,
+                external_id: ref.id,
+                status: "accepted",
+              })
+              .execute();
+            return e.id;
+          },
+        },
+      },
+      {
+        catalog: (q) =>
+          Object.entries(REMOTE)
+            .filter(([k]) => q.includes(k) || k.includes(q))
+            .flatMap(([, v]) => v),
+      },
+    );
+    await mkdir(join(h.lib, SHOW, "Season 1"), { recursive: true });
+    await mkdir(join(h.lib, MOVIE));
+    await mkdir(join(h.lib, "Solaris"));
+    for (const f of [EP1, EP2, MOVIE_FILE, "Solaris/Solaris.1080p.BluRay.x264-GRP.mkv"])
+      await writeFile(join(h.lib, f), "video-bytes");
+    await writeFile(join(h.lib, MOVIE, "readme.txt"), "x");
+  });
+  afterAll(async () => h?.close());
+
+  const scrape = (items: unknown[], extra: Record<string, unknown> = {}, k = EXEC_KEY) =>
+    h.api("POST", "/files/scrape", { idempotencyKey: key(), items, ...extra }, k);
+
+  it("未登记的剧集目录 + 电影目录:按路径识别 TMDB,就地写 NFO 与图片,不建作品", async () => {
+    const r = await scrape([
+      { root: h.libRoot, path: SHOW },
+      { root: h.libRoot, path: MOVIE },
+    ]);
+    expect(r.status).toBe(200);
+    const data = r.body.data;
+    expect(data.needsChoice).toEqual([]);
+    expect(data.pending).toBe(false);
+    expect(data.records).toHaveLength(2);
+    const [show, movie] = data.records;
+    expect(show).toMatchObject({
+      status: "succeeded",
+      path: SHOW,
+      mode: "missing",
+      trigger: "manual",
+      match: {
+        mediaEntityId: null,
+        provider: "tmdb",
+        externalId: "95396",
+        kind: "tv",
+        source: "auto",
+        title: "人生切割术",
+      },
+    });
+    expect(movie).toMatchObject({ status: "succeeded", match: { externalId: "157336", kind: "movie" } });
+
+    const files = await tree(h.lib);
+    expect(files).toEqual(
+      [
+        `${SHOW}/Season 1/poster.jpg`,
+        `${SHOW}/Season 1/season.nfo`,
+        EP1,
+        EP1.replace(/\.mkv$/, "-thumb.jpg"),
+        EP1.replace(/\.mkv$/, ".nfo"),
+        EP2,
+        EP2.replace(/\.mkv$/, "-thumb.jpg"),
+        EP2.replace(/\.mkv$/, ".nfo"),
+        `${SHOW}/fanart.jpg`,
+        `${SHOW}/poster.jpg`,
+        `${SHOW}/season01-poster.jpg`,
+        `${SHOW}/tvshow.nfo`,
+        `${MOVIE}/fanart.jpg`,
+        `${MOVIE}/poster.jpg`,
+        `${MOVIE}/readme.txt`,
+        MOVIE_FILE,
+        MOVIE_FILE.replace(/\.mkv$/, ".nfo"),
+        "Solaris/Solaris.1080p.BluRay.x264-GRP.mkv",
+      ].sort(),
+    );
+    // 未登记:不建作品,字段直接取 TMDB
+    expect(await h.db.selectFrom("media_entity").select("id").execute()).toHaveLength(0);
+    const tvshow = await readFile(join(h.lib, SHOW, "tvshow.nfo"), "utf8");
+    expect(tvshow).toContain("<title>人生切割术</title>");
+    expect(tvshow).toContain('<uniqueid type="imdb" default="true">tt11280740</uniqueid>');
+    const ep = await readFile(join(h.lib, EP1.replace(/\.mkv$/, ".nfo")), "utf8");
+    expect(ep).toContain("<title>有关地狱的好消息</title>");
+    expect(ep).toMatch(/<season>1<\/season>\n {2}<episode>1<\/episode>/);
+    const nfo = await readFile(join(h.lib, MOVIE_FILE.replace(/\.mkv$/, ".nfo")), "utf8");
+    expect(nfo).toContain('<director tmdbid="525">克里斯托弗·诺兰</director>');
+    expect(tmdbCalls).toEqual(expect.arrayContaining(["tv:95396", "season:95396:1", "movie:157336"]));
+    // 记录可按 id 回查;审计按记录
+    const back = await h.api("GET", `/files/scrape-records?ids=${show.id},${movie.id}`);
+    expect(back.body.data.records.map((x: { id: string }) => x.id)).toEqual([show.id, movie.id]);
+    expect(back.body.data.records[0].files.filter((f: { action: string }) => f.action === "written")).toHaveLength(10);
+    expect(h.events.some((e) => e.type === "files.changed" && e.payload.rootId === h.libRoot)).toBe(true);
+  });
+
+  it("歧义:返回候选;选定 TMDB 条目后刮削,并按需加入媒体库", async () => {
+    const first = await scrape([{ root: h.libRoot, path: "Solaris" }]);
+    expect(first.status).toBe(200);
+    expect(first.body.data.records).toEqual([]);
+    expect(first.body.data.needsChoice).toHaveLength(1);
+    const nc = first.body.data.needsChoice[0];
+    expect(nc).toMatchObject({ root: h.libRoot, path: "Solaris", title: "Solaris" });
+    expect(nc.candidates.map((c: { externalId: string }) => c.externalId).sort()).toEqual(["2103", "593"]);
+    expect(await tree(h.lib, "Solaris")).toEqual(["Solaris/Solaris.1080p.BluRay.x264-GRP.mkv"]);
+
+    const second = await scrape(
+      [{ root: h.libRoot, path: "Solaris", external: { provider: "tmdb", externalId: "2103", kind: "movie" } }],
+      { addToLibrary: true },
+    );
+    expect(second.status).toBe(200);
+    const rec = second.body.data.records[0];
+    expect(rec).toMatchObject({
+      status: "succeeded",
+      match: { provider: "tmdb", externalId: "2103", source: "chosen", title: "索拉里斯星", year: 2002 },
+    });
+    expect(second.body.data.entities).toMatchObject([{ title: "索拉里斯星", created: true }]);
+    expect(rec.match.mediaEntityId).toBe(second.body.data.entities[0].mediaEntityId);
+    expect(await tree(h.lib, "Solaris")).toEqual([
+      "Solaris/Solaris.1080p.BluRay.x264-GRP.mkv",
+      "Solaris/Solaris.1080p.BluRay.x264-GRP.nfo",
+      "Solaris/poster.jpg",
+    ]);
+    expect(await readFile(join(h.lib, "Solaris/Solaris.1080p.BluRay.x264-GRP.nfo"), "utf8")).toContain(
+      "<title>索拉里斯星</title>",
+    );
+    // 作品详情取的是该作品最近一条刮削记录
+    expect((await latestEntityScrape(h.module.deps().ao, rec.match.mediaEntityId))?.id).toBe(rec.id);
+  });
+
+  it("只补缺失跳过已有文件(包括别人写的);全部覆盖改写", async () => {
+    const show = join(h.lib, SHOW, "tvshow.nfo");
+    await writeFile(show, "<tvshow>别的工具写的</tvshow>");
+    const miss = await scrape([{ root: h.libRoot, path: SHOW }]);
+    const rec = miss.body.data.records[0];
+    expect(rec).toMatchObject({ status: "succeeded", written: 0 });
+    expect(rec.skipped).toBeGreaterThan(0);
+    expect(await readFile(show, "utf8")).toBe("<tvshow>别的工具写的</tvshow>");
+    const over = await scrape([{ root: h.libRoot, path: SHOW }], { mode: "overwrite" });
+    expect(over.body.data.records[0]).toMatchObject({ status: "succeeded", mode: "overwrite", skipped: 0 });
+    expect(await readFile(show, "utf8")).toContain("<title>人生切割术</title>");
+  });
+
+  it("季目录只写季与各集;单集文件只写该集", async () => {
+    const season = await scrape([{ root: h.libRoot, path: `${SHOW}/Season 1` }], { mode: "overwrite" });
+    const roles = season.body.data.records[0].files.map((f: { path: string; role: string }) => `${f.role}:${f.path}`);
+    expect(roles.sort()).toEqual(
+      [
+        `season_nfo:${SHOW}/Season 1/season.nfo`,
+        `season_poster:${SHOW}/season01-poster.jpg`,
+        `season_poster:${SHOW}/Season 1/poster.jpg`,
+        `episode_nfo:${EP1.replace(/\.mkv$/, ".nfo")}`,
+        `episode_nfo:${EP2.replace(/\.mkv$/, ".nfo")}`,
+        `episode_thumb:${EP1.replace(/\.mkv$/, "-thumb.jpg")}`,
+        `episode_thumb:${EP2.replace(/\.mkv$/, "-thumb.jpg")}`,
+      ].sort(),
+    );
+    const one = await scrape([{ root: h.libRoot, path: EP2 }], { mode: "overwrite" });
+    expect(one.body.data.records[0].files.map((f: { role: string }) => f.role).sort()).toEqual([
+      "episode_nfo",
+      "episode_thumb",
+    ]);
+  });
+
+  it("已登记的目录直接用作品;作品详情「重新刮削」对作品目录刮削", async () => {
+    const film = await seedEntity(h.db, { kind: "movie", title: "星际穿越", year: 2014 });
+    await h.db
+      .insertInto("external_identity")
+      .values({
+        media_entity_id: film,
+        provider: "tmdb",
+        entity_type: "movie",
+        external_id: "157336",
+        status: "accepted",
+      })
+      .execute();
+    await seedAsset(h.db, h, { entityId: film, rel: MOVIE_FILE });
+    const r = await scrape([{ root: h.libRoot, path: MOVIE }]);
+    expect(r.body.data.records[0].match).toMatchObject({
+      mediaEntityId: film,
+      source: "library",
+      externalId: "157336",
+    });
+    // 作品详情「重新刮削」(POST /library/{id}/scrape)的服务端做法
+    const again = await scrapeEntity(h.module.deps().ao, film, { trigger: "manual", mode: "overwrite" });
+    expect(again).toHaveLength(1);
+    const recs = await h.api("GET", `/files/scrape-records?ids=${again[0]!.recordId}`);
+    expect(recs.body.data.records[0]).toMatchObject({ path: MOVIE, root: h.libRoot, mode: "overwrite" });
+  });
+
+  it("只登记目录与非视频文件跳过;没有视频的目录跳过", async () => {
+    await mkdir(join(h.lib, "空目录"));
+    const r = await scrape([
+      { root: h.libRoot, path: `${MOVIE}/readme.txt` },
+      { root: h.libRoot, path: "空目录" },
+    ]);
+    expect(r.body.data.skipped).toEqual([
+      { root: h.libRoot, path: `${MOVIE}/readme.txt`, reason: "scrape_not_video" },
+      { root: h.libRoot, path: "空目录", reason: "scrape_no_videos" },
+    ]);
+    await h.db.updateTable("library_root").set({ management_mode: "register_only" }).execute();
+    const ro = await scrape([{ root: h.libRoot, path: MOVIE }]);
+    expect(ro.body.data.skipped).toEqual([{ root: h.libRoot, path: MOVIE, reason: "scrape_root_read_only" }]);
+    await h.db.updateTable("library_root").set({ management_mode: "managed" }).execute();
+  });
+});

apps/server/test/file-manager/v2.test.ts

@@ -277,7 +277,7 @@ describe.skipIf(skip)("files v2(W9-H)", () => {
     expect(src.body.data.mediaEntityId).toBe(film);
   });
 
-  it("G5:批量刷新元数据按作品去重,未登记作品跳过", async () => {
+  it("G5:批量刷新元数据按作品去重(不要求有已入库文件)", async () => {
     const reg = await seedEntity(h.db, { kind: "movie", title: "Registered Film", year: 2018 });
     await mkdir(join(h.lib, "Registered Film (2018)"), { recursive: true });
     await writeFile(join(h.lib, "Registered Film (2018)", "a.mkv"), big);
@@ -299,8 +299,8 @@ describe.skipIf(skip)("files v2(W9-H)", () => {
       EXEC_KEY,
     );
     expect(r.status).toBe(200);
-    expect(r.body.data.submitted.map((s: Loose) => s.mediaEntityId)).toEqual([reg]);
-    expect(r.body.data.skipped).toEqual([{ ref: bare, reason: "not_registered" }]);
+    expect(r.body.data.submitted.map((s: Loose) => s.mediaEntityId)).toEqual([reg, bare]);
+    expect(r.body.data.skipped).toEqual([]);
   });
 
   it("G8:目录摘要 — 上一级、子目录是否有下级、视频/已登记计数", async () => {

apps/server/test/scrape/gen-scrape-layout-mp.py

@@ -0,0 +1,144 @@
+"""SCR-2 刮削差分(文件布局):用 MP 上游 ScrapingChain.scrape_metadata 的真实决策代码,在临时目录里对同一组目录结构
+"就地刮削",记录 MP 写了哪些文件,写 scrape-layout-mp.json。与 apps/server/test/file-manager/scrape-layout.test.ts 对照。
+
+用法(MP 仓库根目录,用其 .venv):
+  cd /opt/moviepilot-plus/upstream/MoviePilot && PYTHONPATH=. \
+  .venv/bin/python /opt/uhd_media_tool/apps/server/test/scrape/gen-scrape-layout-mp.py
+
+只记录行为输出,不复制 MP 源码。替身:
+- 存储:本地目录的最小实现(list_files / get_parent_item / get_file_item / upload_file / contains_bluray_subdirectories);
+- NFO 内容:固定占位(内容差分见 gen-scrape-mp.py);
+- 图片清单:按 MP themoviedb 模块 get_metadata_img 的规则给(作品 poster + backdrop;季 seasonNN-poster;集 episode-thumb);
+- 识别:recognize_media 原样返回传入的作品(路径识别不在此差分范围)。
+刮削策略用 MP 缺省(ScrapingConfig 默认全部 missingonly);手动刮削 overwrite=True(同 MP /media/scrape 接口)。
+"""
+
+import json
+import pathlib
+import shutil
+import tempfile
+from types import SimpleNamespace
+
+import app.chain.scraping as scraping_mod
+from app.chain.scraping import ScrapingChain, ScrapingConfig
+from app.domain.context import MediaInfo
+from app.schemas.types import MediaSource, MediaType
+from app.schemas.workflow import FileItem
+
+HERE = pathlib.Path(__file__).parent
+VIDEO_EXT = (".mkv", ".mp4", ".ts", ".m2ts", ".avi", ".iso")
+
+
+def item(p: pathlib.Path) -> FileItem:
+    return FileItem(
+        storage="local",
+        type="dir" if p.is_dir() else "file",
+        path=str(p) + ("/" if p.is_dir() else ""),
+        name=p.name,
+        basename=p.stem if p.is_file() else p.name,
+        extension=p.suffix[1:] if p.is_file() else None,
+    )
+
+
+class LocalStorage:
+    def list_files(self, fileitem, recursion=False):
+        return [item(c) for c in sorted(pathlib.Path(fileitem.path).iterdir())]
+
+    def get_parent_item(self, fileitem):
+        return item(pathlib.Path(fileitem.path.rstrip("/")).parent)
+
+    def get_file_item(self, storage, path):
+        p = pathlib.Path(path)
+        return item(p) if p.exists() else None
+
+    def upload_file(self, fileitem, path, new_name=None):
+        target = pathlib.Path(fileitem.path) / (new_name or path.name)
+        shutil.copyfile(path, target)
+        return item(target)
+
+    def contains_bluray_subdirectories(self, files):
+        return any(f.type == "dir" and f.name.upper() in ("BDMV", "VIDEO_TS") for f in files)
+
+
+def make_chain() -> ScrapingChain:
+    c = object.__new__(ScrapingChain)
+    c.storagechain = LocalStorage()
+    c.scraping_policies = ScrapingConfig()
+    c.runtime_config = SimpleNamespace(
+        video_extensions=VIDEO_EXT, audio_extensions=(".flac", ".mp3"), season_zero_names=("Specials", "SPs")
+    )
+    c.metadata_nfo = lambda **kw: "<nfo/>"
+
+    def metadata_img(mediainfo, season=None, episode=None):
+        if season is not None:
+            if episode:
+                return {"episode-thumb.jpg": "https://img/still.jpg"}
+            return {f"season{str(season).rjust(2, '0')}-poster.jpg": "https://img/season.jpg"}
+        return {"poster.jpg": "https://img/poster.jpg", "backdrop.jpg": "https://img/backdrop.jpg"}
+
+    c.metadata_img = metadata_img
+
+    def save_image(fileitem, path, url):
+        pathlib.Path(path).write_bytes(b"jpg")
+
+    c._download_and_save_image = save_image
+    return c
+
+
+class FakeMediaChain:
+    def recognize_media(self, meta=None, **kw):
+        return CURRENT["media"]
+
+    def recognize_by_meta(self, meta):
+        return CURRENT["media"]
+
+
+CURRENT: dict = {}
+scraping_mod.MediaChain = FakeMediaChain  # 集文件重新识别季集后仍是同一作品
+
+
+def media(kind: MediaType, tmdb_id: int, title: str) -> MediaInfo:
+    mi = MediaInfo()
+    mi.type = kind
+    mi.tmdb_id = tmdb_id
+    mi.media_source = MediaSource.TMDB
+    mi.media_id = str(tmdb_id)
+    mi.title = title
+    return mi
+
+
+SHOW = "人生切割术 (2022)"
+EP1 = f"{SHOW}/Season 1/人生切割术 - S01E01 - 有关地狱的好消息.mkv"
+EP2 = f"{SHOW}/Season 1/人生切割术 - S01E02 - 半循环.mkv"
+MOVIE = "星际穿越 (2014)"
+MOVIE_FILE = f"{MOVIE}/星际穿越 (2014) - 1080p.mkv"
+LOOSE = "Interstellar.2014.1080p.BluRay.mkv"
+TREE = [EP1, EP2, MOVIE_FILE, LOOSE, "Heat (1995)/BDMV/index.bdmv"]
+
+SCENARIOS = {
+    "show_dir": (SHOW, MediaType.TV),
+    "season_dir": (f"{SHOW}/Season 1", MediaType.TV),
+    "episode_file": (EP2, MediaType.TV),
+    "movie_dir": (MOVIE, MediaType.MOVIE),
+    "movie_file": (MOVIE_FILE, MediaType.MOVIE),
+    "movie_file_in_root": (LOOSE, MediaType.MOVIE),
+    "bluray_dir": ("Heat (1995)", MediaType.MOVIE),
+}
+
+out = {}
+for name, (target, kind) in SCENARIOS.items():
+    root = pathlib.Path(tempfile.mkdtemp(prefix="mp-scr-"))
+    for f in TREE:
+        (root / f).parent.mkdir(parents=True, exist_ok=True)
+        (root / f).write_bytes(b"video")
+    before = {str(p.relative_to(root)) for p in root.rglob("*") if p.is_file()}
+    CURRENT["media"] = media(kind, 95396 if kind == MediaType.TV else 157336, "x")
+    chain = make_chain()
+    ok, msg = chain.scrape_metadata(fileitem=item(root / target), mediainfo=CURRENT["media"], overwrite=True)
+    after = {str(p.relative_to(root)) for p in root.rglob("*") if p.is_file()}
+    out[name] = {"target": target, "ok": ok, "written": sorted(after - before)}
+    shutil.rmtree(root)
+
+(HERE / "scrape-layout-mp.json").write_text(json.dumps(out, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
+print(json.dumps(out, ensure_ascii=False, indent=2))
+# 生成后在本仓库跑一次格式化(lint 要求):deno run -A npm:@biomejs/biome check --write apps/server/test/scrape/scrape-layout-mp.json

apps/server/test/scrape/scrape-layout-mp.json

@@ -0,0 +1,75 @@
+{
+  "show_dir": {
+    "target": "人生切割术 (2022)",
+    "ok": true,
+    "written": [
+      "人生切割术 (2022)/Season 1/poster.jpg",
+      "人生切割术 (2022)/Season 1/season.nfo",
+      "人生切割术 (2022)/Season 1/人生切割术 - S01E01 - 有关地狱的好消息.jpg",
+      "人生切割术 (2022)/Season 1/人生切割术 - S01E01 - 有关地狱的好消息.nfo",
+      "人生切割术 (2022)/Season 1/人生切割术 - S01E02 - 半循环.jpg",
+      "人生切割术 (2022)/Season 1/人生切割术 - S01E02 - 半循环.nfo",
+      "人生切割术 (2022)/backdrop.jpg",
+      "人生切割术 (2022)/fanart.jpg",
+      "人生切割术 (2022)/poster.jpg",
+      "人生切割术 (2022)/season01-poster.jpg",
+      "人生切割术 (2022)/tvshow.nfo"
+    ]
+  },
+  "season_dir": {
+    "target": "人生切割术 (2022)/Season 1",
+    "ok": true,
+    "written": [
+      "人生切割术 (2022)/Season 1/poster.jpg",
+      "人生切割术 (2022)/Season 1/season.nfo",
+      "人生切割术 (2022)/Season 1/人生切割术 - S01E01 - 有关地狱的好消息.jpg",
+      "人生切割术 (2022)/Season 1/人生切割术 - S01E01 - 有关地狱的好消息.nfo",
+      "人生切割术 (2022)/Season 1/人生切割术 - S01E02 - 半循环.jpg",
+      "人生切割术 (2022)/Season 1/人生切割术 - S01E02 - 半循环.nfo",
+      "人生切割术 (2022)/season01-poster.jpg"
+    ]
+  },
+  "episode_file": {
+    "target": "人生切割术 (2022)/Season 1/人生切割术 - S01E02 - 半循环.mkv",
+    "ok": true,
+    "written": [
+      "人生切割术 (2022)/Season 1/人生切割术 - S01E02 - 半循环.jpg",
+      "人生切割术 (2022)/Season 1/人生切割术 - S01E02 - 半循环.nfo"
+    ]
+  },
+  "movie_dir": {
+    "target": "星际穿越 (2014)",
+    "ok": true,
+    "written": [
+      "星际穿越 (2014)/backdrop.jpg",
+      "星际穿越 (2014)/fanart.jpg",
+      "星际穿越 (2014)/poster.jpg",
+      "星际穿越 (2014)/星际穿越 (2014) - 1080p.nfo"
+    ]
+  },
+  "movie_file": {
+    "target": "星际穿越 (2014)/星际穿越 (2014) - 1080p.mkv",
+    "ok": true,
+    "written": [
+      "星际穿越 (2014)/backdrop.jpg",
+      "星际穿越 (2014)/fanart.jpg",
+      "星际穿越 (2014)/poster.jpg",
+      "星际穿越 (2014)/星际穿越 (2014) - 1080p.nfo"
+    ]
+  },
+  "movie_file_in_root": {
+    "target": "Interstellar.2014.1080p.BluRay.mkv",
+    "ok": true,
+    "written": ["Interstellar.2014.1080p.BluRay.nfo", "backdrop.jpg", "fanart.jpg", "poster.jpg"]
+  },
+  "bluray_dir": {
+    "target": "Heat (1995)",
+    "ok": true,
+    "written": [
+      "Heat (1995)/Heat (1995).nfo",
+      "Heat (1995)/backdrop.jpg",
+      "Heat (1995)/fanart.jpg",
+      "Heat (1995)/poster.jpg"
+    ]
+  }
+}

apps/server/test/scrape/scrape-layout-mp.test.ts

@@ -0,0 +1,174 @@
+import { mkdir, readdir, readFile, stat, writeFile } from "node:fs/promises";
+import { dirname, join, relative } from "node:path";
+import type { TmdbRichDetail, TmdbRichSeasonDetail } from "@umt/adapters";
+import type { ScrapeMatch } from "@umt/contracts";
+import { afterAll, beforeAll, describe, expect, it } from "vitest";
+import { queueScrape, runScrapeRecord } from "../../src/modules/acquisition-organize/services/scrape.ts";
+import { createFmHarness, type FmHarness, pgAvailable } from "../file-manager/harness.ts";
+
+// SCR-2 刮削文件布局差分:同一组目录结构,MP ScrapingChain.scrape_metadata(gen-scrape-layout-mp.py 生成
+// scrape-layout-mp.json)写哪些文件 ⇔ 我们就地刮削写哪些文件。偏离只允许下面 DEVIATIONS 里登记的几类
+// (理由见 docs/runbook/scrape.md「与 MP 的差分」)。
+
+const skip = !(await pgAvailable());
+const MP = JSON.parse(await readFile(new URL("./scrape-layout-mp.json", import.meta.url), "utf8")) as Record<
+  string,
+  { target: string; ok: boolean; written: string[] }
+>;
+
+const SHOW = "人生切割术 (2022)";
+const EP1 = `${SHOW}/Season 1/人生切割术 - S01E01 - 有关地狱的好消息.mkv`;
+const EP2 = `${SHOW}/Season 1/人生切割术 - S01E02 - 半循环.mkv`;
+const MOVIE_FILE = "星际穿越 (2014)/星际穿越 (2014) - 1080p.mkv";
+const LOOSE = "Interstellar.2014.1080p.BluRay.mkv";
+const TREE = [EP1, EP2, MOVIE_FILE, LOOSE, "Heat (1995)/BDMV/index.bdmv"];
+
+/** 登记的偏离:把 MP 的输出换算成我们应有的输出 */
+const DEVIATIONS: { id: string; apply: (files: string[], scenario: string) => string[] }[] = [
+  // 1. 不写 backdrop.jpg 副本,只写 fanart.jpg(Emby / Jellyfin / Kodi 都认 fanart)
+  { id: "no_backdrop_copy", apply: (fs) => fs.filter((f) => !/(^|\/)backdrop\.jpg$/.test(f)) },
+  // 2. 剧照 <视频名>-thumb.jpg(Kodi 约定;MP 写 <视频名>.jpg)
+  {
+    id: "episode_thumb_suffix",
+    apply: (fs) => fs.map((f) => (/ - S\d+E\d+ - [^/]+\.jpg$/.test(f) ? f.replace(/\.jpg$/, "-thumb.jpg") : f)),
+  },
+  // 3. 视频直接在登记目录根:图片写 <视频名>-poster.jpg / -fanart.jpg,不往根里放 poster.jpg(MP 会)
+  {
+    id: "root_movie_images",
+    apply: (fs, s) =>
+      s !== "movie_file_in_root"
+        ? fs
+        : fs.map((f) => (f === "poster.jpg" || f === "fanart.jpg" ? `${LOOSE.replace(/\.mkv$/, "")}-${f}` : f)),
+  },
+];
+
+const detail = (entityType: "movie" | "tv", externalId: string): TmdbRichDetail =>
+  ({
+    entityType,
+    externalId,
+    title: "x",
+    titleLanguage: "zh-CN",
+    titleZh: "x",
+    titleZhLanguage: "zh-CN",
+    originalTitle: null,
+    originalLanguage: "en",
+    overview: null,
+    overviewLanguage: null,
+    tagline: null,
+    year: 2022,
+    releaseDate: null,
+    lastAirDate: null,
+    genres: [],
+    voteAverage: null,
+    voteCount: null,
+    status: null,
+    runtime: null,
+    seasonCount: null,
+    episodeCount: null,
+    networks: [],
+    originCountry: [],
+    posterPath: "/poster.jpg",
+    backdropPath: "/backdrop.jpg",
+    cast: [],
+    crew: [],
+    externalIds: { imdb: null, tvdb: null, wikidata: null },
+    translations: [],
+    alternativeTitles: [],
+    seasons: [],
+    languages: ["zh-CN"],
+  }) satisfies TmdbRichDetail;
+
+const season1: TmdbRichSeasonDetail = {
+  tvId: "95396",
+  seasonNumber: 1,
+  name: null,
+  nameLanguage: null,
+  overview: null,
+  overviewLanguage: null,
+  airDate: null,
+  posterPath: "/season.jpg",
+  fetchedLanguages: ["zh-CN"],
+  episodes: [1, 2].map((n) => ({
+    seasonNumber: 1,
+    episodeNumber: n,
+    name: null,
+    nameLanguage: null,
+    overview: null,
+    overviewLanguage: null,
+    airDate: null,
+    runtime: null,
+    stillPath: "/still.jpg",
+    voteAverage: null,
+    voteCount: null,
+  })),
+};
+
+async function files(root: string): Promise<Set<string>> {
+  const out = new Set<string>();
+  const walk = async (d: string) => {
+    for (const n of await readdir(d).catch(() => [])) {
+      if (n.startsWith(".")) continue;
+      const p = join(d, n);
+      if ((await stat(p)).isDirectory()) await walk(p);
+      else out.add(relative(root, p));
+    }
+  };
+  await walk(root);
+  return out;
+}
+
+describe.skipIf(skip)("刮削文件布局与 MP 差分", () => {
+  let h: FmHarness;
+  beforeAll(async () => {
+    h = await createFmHarness(
+      "scrlayout",
+      {
+        ao: {
+          scrapeImage: async () => ({ content: new Uint8Array([0xff, 0xd8, 0xff]), mime: "image/jpeg" }),
+          scrapeTmdbDetail: async (type, id) => detail(type, id),
+          scrapeTmdbSeason: async (_tv, n) => (n === 1 ? season1 : null),
+        },
+      },
+      { runTransfers: false },
+    );
+  });
+  afterAll(async () => h?.close());
+
+  for (const [scenario, mp] of Object.entries(MP)) {
+    it(`${scenario}:${mp.target}`, async () => {
+      // 每个场景一份独立目录树;"视频直接在根"的场景用下载根(整个根就是这棵树)
+      const inRoot = scenario === "movie_file_in_root";
+      const base = inRoot ? h.dl : join(h.lib, scenario);
+      for (const f of TREE) {
+        await mkdir(dirname(join(base, f)), { recursive: true });
+        await writeFile(join(base, f), "video");
+      }
+      const before = await files(base);
+      const tv = mp.target.startsWith(SHOW);
+      const match: ScrapeMatch = {
+        mediaEntityId: null,
+        provider: "tmdb",
+        externalId: tv ? "95396" : "157336",
+        kind: tv ? "tv" : "movie",
+        title: "x",
+        year: null,
+        source: "chosen",
+      };
+      const q = await queueScrape(h.module.deps().ao, {
+        ...(inRoot
+          ? { rootId: h.dlRoot, rootPath: h.dl, libraryRootId: null }
+          : { rootId: h.libRoot, rootPath: h.lib, libraryRootId: h.libraryRootId }),
+        path: inRoot ? mp.target : `${scenario}/${mp.target}`,
+        match,
+        mode: "overwrite",
+        trigger: "manual",
+      });
+      const rec = await runScrapeRecord(h.module.deps().ao, q.recordId);
+      expect(rec.status).toBe("succeeded");
+      const after = await files(base);
+      const ours = [...after].filter((f) => !before.has(f)).sort();
+      const expected = DEVIATIONS.reduce((fs, d) => d.apply(fs, scenario), [...mp.written]).sort();
+      expect(ours).toEqual(expected);
+    });
+  }
+});