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]>
This commit is contained in:
rootandClaude Opus 5.5 committed 2026-09-27 20:12:01 +08:00
1 parent 829ca4f7ac
commit a24bda952b
39 files changed
+6861 -15

No files matched your search

+145 -1
View File
@@ -4,7 +4,8 @@ import { EntityBase, Id, JsonObject, NonNegInt, SERVER_MANAGED, Timestamp } from
// 来源域:站点、站点账号、Feed、站点资源、观察(REQ §4)
// 含 passkey 的 Feed URL 只以 urlSecretRef 保存,对外只给脱敏的 urlDisplay。
export const SiteAdapter = z.enum(["nexusphp", "mteam_api", "generic_rss"]);
/** api = MP 专用 API spider(Haidan/HDDolby/RousiPro/SunnyPT/TNode/TorrentLeech/Yema) */
export const SiteAdapter = z.enum(["nexusphp", "mteam_api", "api", "generic_rss"]);
export type SiteAdapter = z.infer<typeof SiteAdapter>;
export const SiteHealth = z.enum(["unknown", "healthy", "degraded", "failing", "disabled"]);
export type SiteHealth = z.infer<typeof SiteHealth>;
@@ -180,6 +181,8 @@ export const SourceRelease = EntityBase.extend({
resolutionEvidence: JsonObject,
availability: ReleaseAvailability,
raw: JsonObject,
/** 人工确认的身份关联:自动识别/复查不得改写 mediaEntityId/contentUnitId */
resolutionLocked: z.boolean(),
});
export type SourceRelease = z.infer<typeof SourceRelease>;
/** 由适配器/任务写入(按 siteId+siteReleaseId upsert) */
@@ -209,6 +212,7 @@ export const SourceReleaseCreate = SourceRelease.omit({
resolutionEvidence: true,
availability: true,
raw: true,
resolutionLocked: true,
});
export type SourceReleaseCreate = z.infer<typeof SourceReleaseCreate>;
export const SourceReleaseUpdate = SourceReleaseCreate.omit({ siteId: true, siteReleaseId: true }).partial();
@@ -259,3 +263,143 @@ export const ObservationUpdate = Observation.pick({
lastError: true,
}).partial();
export type ObservationUpdate = z.infer<typeof ObservationUpdate>;
// ── W2-B:站点快照、调度事件与站点工作台 ──
export const SiteSnapshotKind = z.enum(["periodic", "backfill", "live_search", "manual"]);
export type SiteSnapshotKind = z.infer<typeof SiteSnapshotKind>;
export const SiteSnapshotStatus = z.enum(["running", "ok", "partial", "failed"]);
export type SiteSnapshotStatus = z.infer<typeof SiteSnapshotStatus>;
/** 一次列表快照 / 实时补查 / 补漏的覆盖范围与结果(不把一次分页伪称全站快照,REQ §4.6) */
export const SiteSnapshot = EntityBase.extend({
siteId: Id,
kind: SiteSnapshotKind,
query: z.string().nullable(),
filters: JsonObject,
/** 实际抓取的页码(1 起) */
pages: z.array(z.number().int().positive()),
itemCount: NonNegInt,
newCount: NonNegInt,
status: SiteSnapshotStatus,
error: JsonObject.nullable(),
definitionVersion: z.string().nullable(),
startedAt: Timestamp,
finishedAt: Timestamp.nullable(),
});
export type SiteSnapshot = z.infer<typeof SiteSnapshot>;
/** 站点注册表条目(随产品发布的适配器定义)+ 是否已在本系统登记 */
export const SiteDefinitionView = z.object({
key: z.string(),
name: z.string(),
adapter: SiteAdapter,
baseUrl: z.string(),
/** 定义是否随产品启用(未实测站点为 false) */
shippedEnabled: z.boolean(),
definitionVersion: z.string(),
capabilities: z.object({
list: z.boolean(),
search: z.boolean(),
detail: z.boolean(),
download: z.boolean(),
rss: z.boolean(),
}),
defaultRateLimit: z.object({ minIntervalMs: NonNegInt, burst: z.number().int().positive() }),
authSupported: z.array(z.string()),
/** 已登记的 site 行 ID;未登记为 null */
siteId: Id.nullable(),
});
export type SiteDefinitionView = z.infer<typeof SiteDefinitionView>;
/** Feed 到期(调度器发出,outbox 路由到 ingestFeed) */
export const FeedDuePayload = z.object({ feedId: Id, reason: z.enum(["schedule", "manual"]) });
/** 新的 source_release 被观察到(或同一资源标题更新),outbox 路由到 resolveObservation */
export const ReleaseObservedPayload = z.object({
sourceReleaseId: Id,
observationId: Id.nullable(),
siteId: Id,
isNew: z.boolean(),
});
/** Feed 窗口与上次零重叠:疑似缺口,路由到有预算的列表补漏(snapshotSite backfill) */
export const FeedGapSuspectedPayload = z.object({
feedId: Id,
siteId: Id.nullable(),
previousFetchedAt: Timestamp.nullable(),
detectedAt: Timestamp,
});
/** 站点工作台列表查询(整理视图):快照优先,必要时实时补查 */
export const WorkbenchReleaseQuery = z.object({
q: z.string().trim().min(1).max(200).optional(),
category: z.string().max(100).optional(),
/** 只看某类促销(free/2xfree/50%…) */
promotion: z.string().max(20).optional(),
resolution: ResolutionStatus.optional(),
/** true = 强制实时补查;false = 只看本地快照;缺省 = 本地无结果或快照过期时补查 */
live: z.enum(["true", "false"]).optional(),
cursor: z.string().min(1).optional(),
limit: z.coerce.number().int().min(1).max(200).default(50),
});
export type WorkbenchReleaseQuery = z.infer<typeof WorkbenchReleaseQuery>;
/** 数据来源:快照 / Feed 观察 / 实时补查 */
export const ReleaseOrigin = z.enum(["snapshot", "feed", "live", "unknown"]);
export type ReleaseOrigin = z.infer<typeof ReleaseOrigin>;
/**
* 工作台条目:source_release + 来源标注 + 本地实体关联。
* 库存叠加(已有资产/Slot/下载状态)由前端按 mediaEntityId 二次查询 W2-C 端点,不在这里猜。
*/
export const WorkbenchRelease = SourceRelease.extend({
origin: ReleaseOrigin,
/** 本条数据最近一次被站点确认的时间(快照/Feed/实时) */
dataUpdatedAt: Timestamp,
observedByFeeds: NonNegInt,
entity: z.object({ id: Id, title: z.string(), year: z.number().int().nullable(), kind: z.string() }).nullable(),
});
export type WorkbenchRelease = z.infer<typeof WorkbenchRelease>;
export const WorkbenchCoverage = z.object({
siteEnabled: z.boolean(),
lastSnapshotAt: Timestamp.nullable(),
/** 快照过期(超过 2 个周期或从未快照) */
stale: z.boolean(),
fromSnapshot: z.boolean(),
live: z.boolean(),
liveError: z.string().nullable(),
/** 同一搜索正在进行时合并到已有请求 */
coalesced: z.boolean(),
});
export type WorkbenchCoverage = z.infer<typeof WorkbenchCoverage>;
export const WorkbenchReleasePage = z.object({
items: z.array(WorkbenchRelease),
nextCursor: z.string().nullable(),
coverage: WorkbenchCoverage,
});
export type WorkbenchReleasePage = z.infer<typeof WorkbenchReleasePage>;
export const WorkbenchReleaseDetail = z.object({
release: WorkbenchRelease,
observations: z.array(
z.object({
observationId: Id,
feedId: Id,
feedName: z.string(),
firstSeenAt: Timestamp,
lastSeenAt: Timestamp,
}),
),
/** 易变字段的复核时间:缓存命中不能替代下载前复核(REQ §4.6) */
volatile: z.object({
promotionObservedAt: Timestamp.nullable(),
promotionExpiresAt: Timestamp.nullable(),
peersObservedAt: Timestamp,
detailFetchedAt: Timestamp.nullable(),
recheckBeforeDownload: z.literal(true),
}),
/** 刷新详情失败时保留既有数据并给出原因 */
refreshError: z.string().nullable(),
});
export type WorkbenchReleaseDetail = z.infer<typeof WorkbenchReleaseDetail>;