feat(server): W2-B sources, identity resolution, metadata and site workbench

- Sites / accounts / feeds: registry view (/site-definitions), site CRUD with enable/disable (history stays queryable, marked site_disabled), accounts and feed URLs stored as secrets (only secret_ref / redacted urlDisplay exposed), feed enable/disable/reset/fetch; one shared rate limiter + adapter per site. - Tasks: ingestFeed (FeedFetchState/FeedCursor on feed_source, observation feed+guid upsert with digest-change updates, source_release site+id aggregation across feeds, auth failure -> account invalid + decision, no retry, gap -> feed.gap_suspected -> budgeted backfill snapshot), resolveObservation (parseRelease -> local recall via title_alias pg_trgm + accepted identities -> TMDB -> resolveIdentity; auto binds entity / provisional units / identities / aliases, otherwise identity_candidate pool with evidence version and bounded review backoff; manual lock respected), refreshMetadata (metadata_field + field_override, user locks win, paths untouched), snapshotSite (site_snapshot coverage records). - Events: feed.due, feed.gap_suspected, release.observed (self-consumed), release.resolved, decision.created/updated; scheduler in worker start(). - Identity API: entity detail with redirects, manual mapping correction with identity_revision, merge with identity_redirect, pool list/assign/dismiss/ retry, content unit numbering lock, aliases, external identities. - Decisions: list/summary (by entity and kind)/detail/resolve; identity options executed directly. - Workbench: /sites/{id}/releases (snapshot first, coalesced live search with backoff, origin + freshness) and /sites/{id}/releases/{rid} (volatile field recheck times, refresh keeps data on failure). - Capabilities: catalog.lookup, release.search, release.inspect, identity.resolve. - Migrations 0018-0020: site.adapter api, source_release.resolution_locked, site_snapshot, identity_candidate, resolution_attempt, identity_redirect, metadata_field, field_override. Co-Authored-By: Claude Opus 5.5 <[email protected]>
2026-09-27 20:12:01 +08:00
1 parent 829ca4f7ac
commit a24bda952b
@@ -119,3 +119,142 @@ export const IdentityRevisionCreate = IdentityRevision.pick({
  reason: true,
 }).partial({ fromStatus: true, toStatus: true, before: true, actorId: true, reason: true });
 export type IdentityRevisionCreate = z.infer<typeof IdentityRevisionCreate>;
+
+// ── W2-B:待识别池、识别尝试、合并跳转与人工纠错(REQ §7.2、§7.5) ──
+
+export const IdentityCandidateStatus = z.enum(["pending", "resolved", "dismissed", "archived"]);
+export type IdentityCandidateStatus = z.infer<typeof IdentityCandidateStatus>;
+/** proposed = 证据足够但自动化关闭/只读,待确认 */
+export const IdentityCandidateReason = z.enum(["ambiguous", "no_match", "proposed", "lookup_failed"]);
+export type IdentityCandidateReason = z.infer<typeof IdentityCandidateReason>;
+
+/** 待识别池里的一个候选(只存 JSON,不建正式作品) */
+export const PoolCandidate = z.object({
+  provider: z.string(),
+  entityType: z.enum(["movie", "tv"]),
+  externalId: z.string(),
+  title: z.string(),
+  year: z.number().int().nullable(),
+  /** 本地候选的作品 ID */
+  localEntityId: Id.nullable(),
+  level: z.enum(["rule_confirmed", "multi_evidence", "needs_confirm", "rejected"]),
+  evidence: z.array(z.string()),
+});
+export type PoolCandidate = z.infer<typeof PoolCandidate>;
+
+/** 待识别池条目:每条 source_release 至多一条;失败原因、证据版本与下一次复查时间 */
+export const IdentityCandidate = EntityBase.extend({
+  sourceReleaseId: Id,
+  status: IdentityCandidateStatus,
+  reason: IdentityCandidateReason,
+  /** 解析出的检索条件(标题/年份/季集) */
+  query: JsonObject,
+  candidates: z.array(PoolCandidate),
+  evidenceVersion: z.string(),
+  attempts: z.number().int().nonnegative(),
+  lastAttemptAt: Timestamp.nullable(),
+  nextReviewAt: Timestamp.nullable(),
+  decisionId: Id.nullable(),
+  resolvedMediaEntityId: Id.nullable(),
+  resolvedBy: z.string().nullable(),
+  resolvedAt: Timestamp.nullable(),
+});
+export type IdentityCandidate = z.infer<typeof IdentityCandidate>;
+
+export const ResolutionOutcome = z.enum([
+  "auto",
+  "proposed",
+  "ambiguous",
+  "no_match",
+  "lookup_failed",
+  "manual",
+  "skipped_locked",
+]);
+export type ResolutionOutcome = z.infer<typeof ResolutionOutcome>;
+
+/** 一次识别尝试(只追加) */
+export const ResolutionAttempt = EntityBase.extend({
+  sourceReleaseId: Id,
+  observationId: Id.nullable(),
+  outcome: ResolutionOutcome,
+  reason: z.string().nullable(),
+  evidenceVersion: z.string(),
+  candidateCount: z.number().int().nonnegative(),
+  mediaEntityId: Id.nullable(),
+  best: JsonObject.nullable(),
+  error: JsonObject.nullable(),
+  actorType: ActorType,
+  actorId: z.string().nullable(),
+});
+export type ResolutionAttempt = z.infer<typeof ResolutionAttempt>;
+
+/** 作品合并后的旧 ID 跳转 */
+export const IdentityRedirect = EntityBase.extend({
+  fromEntityId: Id,
+  toEntityId: Id,
+  reason: z.string().nullable(),
+  actorType: ActorType,
+  actorId: z.string().nullable(),
+});
+export type IdentityRedirect = z.infer<typeof IdentityRedirect>;
+
+/** 外部身份引用(provider + 类型 + ID) */
+export const ExternalRef = z.object({
+  provider: IdentityProvider,
+  entityType: z.enum(["movie", "tv"]),
+  externalId: z.string().min(1),
+});
+export type ExternalRef = z.infer<typeof ExternalRef>;
+
+/**
+ * 人工指定来源资源的身份:本地作品 / 外部身份(未登记时按外部详情建档)/ 池内候选序号,三选一。
+ * 指定后加人工锁,自动识别不再改写。
+ */
+export const ReleaseIdentityAssign = z
+  .object({
+    mediaEntityId: Id.optional(),
+    external: ExternalRef.optional(),
+    candidateIndex: z.number().int().min(0).optional(),
+    /** 同时关联到某个内容单元(季/集);缺省按解析出的季集建临时单元 */
+    contentUnitId: Id.nullable().optional(),
+    lock: z.boolean().default(true),
+    reason: z.string().max(2000).optional(),
+  })
+  .refine((v) => [v.mediaEntityId, v.external, v.candidateIndex].filter((x) => x !== undefined).length === 1, {
+    message: "mediaEntityId / external / candidateIndex 必须且只能给一个",
+  });
+export type ReleaseIdentityAssign = z.infer<typeof ReleaseIdentityAssign>;
+
+/** 改作品的外部映射:旧的同 provider+类型 accepted 映射退役(留修订),新映射 accepted + 人工锁 */
+export const EntityIdentitySet = ExternalRef.extend({
+  reason: z.string().max(2000).optional(),
+  /** 默认加人工锁 */
+  lock: z.boolean().default(true),
+});
+export type EntityIdentitySet = z.infer<typeof EntityIdentitySet>;
+
+export const IdentityPoolQuery = z.object({
+  status: IdentityCandidateStatus.optional(),
+  reason: IdentityCandidateReason.optional(),
+  /** 按原始标题/解析标题检索(已观察到的资源必须能从待识别入口找回) */
+  q: z.string().trim().min(1).max(200).optional(),
+  siteId: Id.optional(),
+  cursor: z.string().min(1).optional(),
+  limit: z.coerce.number().int().min(1).max(200).default(50),
+});
+export type IdentityPoolQuery = z.infer<typeof IdentityPoolQuery>;
+
+/** 待识别池列表条目:附来源资源摘要 */
+export const IdentityPoolItem = IdentityCandidate.extend({
+  release: z.object({
+    id: Id,
+    siteId: Id,
+    siteReleaseId: z.string(),
+    title: z.string(),
+    subtitle: z.string().nullable(),
+    categoryRaw: z.string().nullable(),
+    firstSeenAt: Timestamp,
+    resolutionStatus: z.string(),
+  }),
+});
+export type IdentityPoolItem = z.infer<typeof IdentityPoolItem>;