UHD_Media_Tool/apps/server/src/modules/sources-identity/tasks.ts

import { defineTask, type OutboxRouteFn, type TaskRegistry } from "@umt/tasks";
import { z } from "zod";
import type { SourcesRuntime } from "./runtime.ts";
import { refreshSiteAccount } from "./service/account.ts";
import { AUTO_ENRICH_TASK, runAutoEnrich } from "./service/enrich.ts";
import { ingestFeed } from "./service/feeds.ts";
import { refreshMetadata } from "./service/metadata.ts";
import { resolveRelease } from "./service/resolve.ts";
import { snapshotSite } from "./service/snapshot.ts";
import { liveSearch } from "./service/workbench.ts";

// 持久任务:输入只放业务 ID,处理器读业务表。外部请求失败的退避/熔断由适配器状态与业务表承担,
// 凭据失效返回正常结果(不重试);只有意外错误(DB、TMDB 网络)抛出交给引擎重试。

const NETWORK_RETRY = {
  maximumAttempts: 5,
  initialInterval: "1s",
  backoffCoefficient: 2,
  maximumInterval: "60s",
} as const;

const id = z.string().uuid();

/** Feed 疑似缺口时的补漏页数预算 */
export const BACKFILL_MAX_PAGES = 3;

export function registerSourcesTasks(registry: TaskRegistry, rt: () => SourcesRuntime): void {
  registry.register(
    defineTask({
      name: "ingestFeed",
      version: 1,
      lane: "network",
      inputSchema: z.object({ feedSourceId: id, manual: z.boolean().optional() }),
      retryPolicy: NETWORK_RETRY,
      handler: (t) => t.step.run("ingest", () => ingestFeed(rt(), t.input.feedSourceId, { manual: !!t.input.manual })),
    }),
    defineTask({
      name: "resolveObservation",
      version: 1,
      lane: "network",
      inputSchema: z
        .object({
          observationId: id.nullish(),
          sourceReleaseId: id.nullish(),
          force: z.boolean().optional(),
          // 惰性识别(DEC-1):manual / search 总是查 TMDB;缺省按 force 视为 review / observed
          trigger: z.enum(["observed", "review", "manual", "search"]).optional(),
        })
        .refine((v) => !!v.observationId || !!v.sourceReleaseId, { message: "需要 observationId 或 sourceReleaseId" }),
      retryPolicy: NETWORK_RETRY,
      handler: (t) =>
        t.step.run("resolve", () =>
          resolveRelease(rt(), {
            observationId: t.input.observationId ?? undefined,
            sourceReleaseId: t.input.sourceReleaseId ?? undefined,
            force: t.input.force,
            trigger: t.input.trigger,
          }),
        ),
    }),
    // META-1 自动元数据补全(每个池条目一次,只出建议):经能力注册中心调用 metadata.enrich
    defineTask({
      name: AUTO_ENRICH_TASK,
      version: 1,
      lane: "network",
      inputSchema: z.object({ poolEntryId: id }),
      retryPolicy: { maximumAttempts: 1 },
      handler: (t) => t.step.run("enrich", () => runAutoEnrich(rt().ctx, t.input.poolEntryId, t.runId, t.abortSignal)),
    }),
    defineTask({
      name: "refreshMetadata",
      version: 1,
      lane: "network",
      inputSchema: z.object({ mediaEntityId: id, full: z.boolean().optional() }),
      retryPolicy: NETWORK_RETRY,
      handler: (t) =>
        t.step.run("refresh", () => refreshMetadata(rt(), t.input.mediaEntityId, { full: !!t.input.full })),
    }),
    defineTask({
      name: "searchSubscription",
      version: 1,
      lane: "network",
      inputSchema: z.object({ subscriptionId: id, manual: z.boolean().optional() }),
      retryPolicy: NETWORK_RETRY,
      handler: (t) =>
        t.step.run("search", () => searchSubscription(rt(), t.input.subscriptionId, { manual: !!t.input.manual })),
    }),
    defineTask({
      name: "snapshotSite",
      version: 1,
      lane: "network",
      inputSchema: z.object({
        siteId: id,
        kind: z.enum(["periodic", "backfill", "manual"]).default("periodic"),
        maxPages: z.number().int().min(1).max(20).optional(),
      }),
      retryPolicy: NETWORK_RETRY,
      handler: (t) =>
        t.step.run("snapshot", () =>
          snapshotSite(rt(), { siteId: t.input.siteId, kind: t.input.kind, maxPages: t.input.maxPages }),
        ),
    }),
    defineTask({
      name: "refreshSiteAccount",
      version: 1,
      lane: "network",
      inputSchema: z.object({ accountId: id, mode: z.enum(["login", "userData"]).default("login") }),
      // 网络失败只记 last_error 不抛错;不自动重试(站点登录频繁重试有封号风险)
      retryPolicy: { maximumAttempts: 1 },
      handler: (t) =>
        t.step.run("refresh", async () => {
          const exists = await rt()
            .ctx.db.selectFrom("site_account")
            .select("id")
            .where("id", "=", t.input.accountId)
            .executeTakeFirst();
          if (!exists) return { skipped: "not_found" };
          return refreshSiteAccount(rt(), t.input.accountId, { mode: t.input.mode });
        }),
    }),
  );
}

// ── 订阅主动搜索(W5-C,对照 MP SubscribeSearchOwner):按作品标题/原名/别名在已开启站点搜索 ──

/** 主动搜索周期缺省 24h;subscription.params.searchIntervalHours 覆盖(0 = 关闭周期搜索) */
export const DEFAULT_SEARCH_INTERVAL_HOURS = 24;
/** 每次最多用几个检索词(标题、原名、别名),控制站点请求量 */
export const SEARCH_MAX_KEYWORDS = 3;

export interface SubscriptionSearchResult {
  status: "ok" | "skipped";
  reason?: string;
  keywords: string[];
  sites: string[];
  searches: number;
  errors: { siteId: string; keyword: string; error: string }[];
  nextSearchAt: string | null;
}

/**
 * 主动搜索:对订阅站点范围内已开启且健康的站点,按标题/原名/别名经站点适配器 search(共享限流器、实时补查合并与失败退避,
 * 记 site_snapshot kind=live_search);新结果 upsert 为 source_release 并发 release.observed → 识别 → release.resolved →
 * evaluateAcquisition(与 RSS 同一路径);已识别的旧结果由本任务末尾入队的订阅重评覆盖。
 */
export async function searchSubscription(
  rt: SourcesRuntime,
  subscriptionId: string,
  opts: { manual?: boolean } = {},
): Promise<SubscriptionSearchResult> {
  const { db } = rt.ctx;
  const none = { keywords: [], sites: [], searches: 0, errors: [], nextSearchAt: null };
  const sub = await db.selectFrom("subscription").selectAll().where("id", "=", subscriptionId).executeTakeFirst();
  if (!sub) return { status: "skipped", reason: "not_found", ...none };
  if (!opts.manual && sub.status !== "active" && sub.status !== "pending")
    return { status: "skipped", reason: `subscription_${sub.status}`, ...none };
  if (!sub.media_entity_id) return { status: "skipped", reason: "no_entity", ...none };
  const automation = await rt.ctx.settings.get("automation");
  if (!opts.manual && (!automation.enabled || automation.readOnlyReconcile))
    return { status: "skipped", reason: "automation_blocked", ...none };
  const entity = await db
    .selectFrom("media_entity")
    .select(["id", "title", "original_title"])
    .where("id", "=", sub.media_entity_id)
    .executeTakeFirst();
  if (!entity) return { status: "skipped", reason: "entity_not_found", ...none };
  const aliases = await db
    .selectFrom("title_alias")
    .select(["alias", "kind"])
    .where("media_entity_id", "=", entity.id)
    .where("content_unit_id", "is", null)
    .orderBy("created_at")
    .limit(20)
    .execute();
  const seen = new Set<string>();
  const keywords: string[] = [];
  for (const k of [entity.title, entity.original_title, ...aliases.map((a) => a.alias)]) {
    const v = (k ?? "").trim();
    const norm = v.toLowerCase();
    if (!v || seen.has(norm)) continue;
    seen.add(norm);
    keywords.push(v);
    if (keywords.length >= SEARCH_MAX_KEYWORDS) break;
  }
  const params = (sub.params ?? {}) as { siteIds?: string[]; searchIntervalHours?: number; keywords?: string[] };
  // 订阅的搜索关键词(MP keyword)追加在标题/别名之后,不占 SEARCH_MAX_KEYWORDS;结果同样经识别 → 评估
  // (订阅私有规则组的包含 / 排除在评估时过滤,与 RSS 来的资源同一口径)
  for (const k of params.keywords ?? []) {
    const v = typeof k === "string" ? k.trim() : "";
    if (!v || seen.has(v.toLowerCase())) continue;
    seen.add(v.toLowerCase());
    keywords.push(v);
  }
  let sq = db.selectFrom("site").selectAll().where("enabled", "=", true).where("health_status", "!=", "failing");
  if (params.siteIds?.length) sq = sq.where("id", "in", params.siteIds);
  const sites = await sq.orderBy("name").execute();
  const errors: SubscriptionSearchResult["errors"] = [];
  let searches = 0;
  for (const site of sites) {
    for (const kw of keywords) {
      searches++;
      const r = await liveSearch(rt, site, kw).catch((e) => ({ error: (e as Error).message, coalesced: false }));
      if (r.error) errors.push({ siteId: site.id, keyword: kw, error: r.error });
    }
  }
  const hours = params.searchIntervalHours ?? DEFAULT_SEARCH_INTERVAL_HOURS;
  const now = rt.now();
  const next = hours > 0 ? new Date(now.getTime() + hours * 3600_000) : null;
  await db
    .updateTable("subscription")
    .set({ last_searched_at: now, next_search_at: next })
    .where("id", "=", sub.id)
    .execute();
  // 已识别的旧结果:按订阅重评(新结果走 release.observed → resolved → evaluate)
  await rt.ctx.engine
    .enqueue(
      "evaluateAcquisition",
      { subscriptionId: sub.id },
      { idempotencyKey: `search-eval:${sub.id}:${now.getTime()}`, traceId: sub.id },
    )
    .catch(() => {});
  return {
    status: "ok",
    keywords,
    sites: sites.map((s) => s.id),
    searches,
    errors,
    nextSearchAt: next ? next.toISOString() : null,
  };
}

type P = Record<string, unknown>;

export const sourcesOutboxRoutes: Record<string, OutboxRouteFn> = {
  "feed.due": (ev) => {
    const p = ev.payload as P;
    return { task: "ingestFeed", input: { feedSourceId: p.feedId, manual: p.reason === "manual" } };
  },
  "release.observed": (ev) => {
    const p = ev.payload as P;
    // 标题更新(isNew=false):按新标题重新解析并重新评估,已识别的也不短路(ID-2);触发来源仍是 observed(DEC-1 不变)
    const retitled = p.isNew === false;
    return {
      task: "resolveObservation",
      input: {
        sourceReleaseId: p.sourceReleaseId,
        observationId: p.observationId,
        ...(retitled ? { force: true, trigger: "observed" } : {}),
      },
    };
  },
  "feed.gap_suspected": (ev) => {
    const p = ev.payload as P;
    if (!p.siteId) return null;
    return { task: "snapshotSite", input: { siteId: p.siteId, kind: "backfill", maxPages: BACKFILL_MAX_PAGES } };
  },
};