feat(server): scrapeMedia 刮削任务,入库后自动刮削(SCR-1)

- 文件道任务 scrapeMedia:NFO 原子写入、图片经图片缓存下载(并发 4,失败不阻塞 NFO)、只补缺失 / 全部覆盖
- 设置 organize.scrapeAfterImport / scrapeMode / scrapeImages / scrapeEpisodeThumbs
- POST /library/{id}/scrape、详情 scrape 状态;files/scrape writeNfo 与 scrape-batch write 走同一任务
- 迁移 0057_media_scrape、原因码、审计 organize.scrape
- 与 MP 输出的 NFO 差分测试(apps/server/test/scrape)

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

apps/server/src/modules/file-manager/probe.ts

@@ -13,6 +13,7 @@ import type {
 } from "@umt/contracts";
 import { sql } from "@umt/db";
 import type { Principal } from "../../http/auth.ts";
+import { enqueueScrape } from "../acquisition-organize/services/scrape.ts";
 import { FilesError, type FmDeps, jsonb, resolveIn } from "./common.ts";
 import { audit } from "./ledger.ts";
 import { kindOfName, mediaSummary } from "./list.ts";
@@ -20,7 +21,8 @@ import { entitiesUnder } from "./records.ts";
 
 // probe:ffprobe 摘要;已登记资产写回 media_asset.media_info.probe,未登记只缓存 24h(DATA_DIR/cache/probe)。
 // scrape:只对已登记作品,等价 POST /media-entities/{id}/refresh(入队 refreshMetadata,字段锁优先、不动路径);
-// NFO/图片投影 W5 实现(TODO),请求时返回 not_implemented。
+// writeNfo / images(SCR-1)另排 scrapeMedia,把 NFO 与图片写进媒体库目录(与作品详情「重新刮削」同一任务)。
+// scrape-batch write=true:文件管理「刮削」,按作品去重各排一次 scrapeMedia(不刷新元数据)。
 
 const CACHE_TTL_MS = 24 * 3600_000;
 
@@ -133,13 +135,23 @@ export async function scrape(
     .where("locked", "=", true)
     .execute();
   const lockedFields = [...new Set([...(e.field_locks ?? []), ...overrides.map((o) => o.field)])].sort();
-  // TODO(W5):writeNfo/images → 把最终字段投影成 NFO/海报写到绑定目录(临时文件 + 原子提交;非本系统 NFO 不覆盖)
+  const scrapeTaskId =
+    body.writeNfo || body.images
+      ? (
+          await enqueueScrape(d.ao, {
+            mediaEntityId: e.id,
+            trigger: "manual",
+            actor: { type: principal?.kind === "api_key" ? "api_key" : "user", id: principal?.id ?? null },
+          })
+        ).taskId
+      : null;
   const result: FilesScrapeResult = {
     mediaEntityId: e.id,
     taskId: runId,
     lockedFields,
-    nfo: body.writeNfo ? "not_implemented" : "not_requested",
-    images: body.images ? "not_implemented" : "not_requested",
+    nfo: body.writeNfo ? "queued" : "not_requested",
+    images: body.images ? "queued" : "not_requested",
+    scrapeTaskId,
   };
   await audit(d, principal, "scrape", { ...result });
   return result;
@@ -202,17 +214,28 @@ export async function scrapeBatch(
       skipped.push({ ref: id, reason: "not_registered" });
       continue;
     }
-    const { runId } = await d.ctx.engine.enqueue(
-      "refreshMetadata",
-      { mediaEntityId: id },
-      { idempotencyKey: `refreshMetadata:files-batch:${body.idempotencyKey}:${id}`, traceId: randomUUID() },
-    );
+    const runId = body.write
+      ? (
+          await enqueueScrape(d.ao, {
+            mediaEntityId: id,
+            trigger: "manual",
+            actor: { type: principal?.kind === "api_key" ? "api_key" : "user", id: principal?.id ?? null },
+          })
+        ).taskId
+      : (
+          await d.ctx.engine.enqueue(
+            "refreshMetadata",
+            { mediaEntityId: id },
+            { idempotencyKey: `refreshMetadata:files-batch:${body.idempotencyKey}:${id}`, traceId: randomUUID() },
+          )
+        ).runId;
     const lockedFields = [
       ...new Set([...(e.field_locks ?? []), ...locks.filter((l) => l.media_entity_id === id).map((l) => l.field)]),
     ].sort();
     submitted.push({ mediaEntityId: id, title: e.title, taskId: runId, lockedFields });
   }
   await audit(d, principal, "scrape_batch", {
+    write: body.write,
     submitted: submitted.map((s) => s.mediaEntityId),
     skipped,
   });