feat(identity): escalate ambiguous/no_match via site detail-page IDs (ID-3)
Fetch the detail page (cached, rate-limited, per-site hourly cap identity.detailIdsPerSiteHour) for search results, subscription-related candidates, pool enrich and manual re-identify; resolve IMDb/TMDB via TMDB, douban via cached douban subjects (site detail info block / META-2 snippets). No identity_unresolved decisions for releases with site IDs. Co-Authored-By: Claude Opus 5.5 <[email protected]>
This commit is contained in:
rootandClaude Opus 5.5 committed
1 parent
cba214e93e
commit
3d5a50b30a
9 files changed
+867
-10
No files matched your search
@@ -0,0 +1,356 @@
|
||||
/**
|
||||
* ID-3 识别升级链(样本驱动,零真实请求):标题识别不确定 → 抓站点详情页(观众 DET-2 脱敏样本 / 憨憨 HH-2 样本,经注入的 fetch)
|
||||
* → IMDb / TMDB / 豆瓣 ID → TMDB(mock)→ 确定;每站每小时上限;RSS 入库不抓;事项不开 / 自动关闭;豆瓣条目缓存路径。
|
||||
*/
|
||||
import { readFileSync } from "node:fs";
|
||||
import { sql } from "@umt/db";
|
||||
import { normalizeTitle } from "@umt/domain";
|
||||
import {
|
||||
createMockTmdb,
|
||||
HHANCLUB_DETAIL_PAGE_2026,
|
||||
HHANCLUB_FILELIST_FRAGMENT,
|
||||
type MockTmdb,
|
||||
type TmdbFixture,
|
||||
} from "@umt/testkit";
|
||||
import { afterAll, beforeAll, beforeEach, describe, expect, it } from "vitest";
|
||||
import {
|
||||
DOUBAN_SUBJECT_REGION,
|
||||
doubanSubjectFromSnippet,
|
||||
loadDoubanSubject,
|
||||
saveDoubanSubject,
|
||||
} from "../../src/modules/sources-identity/service/douban-subject.ts";
|
||||
import {
|
||||
ESCALATION_COUNTER_REGION,
|
||||
escalationCount,
|
||||
} from "../../src/modules/sources-identity/service/id-escalation.ts";
|
||||
import { resolveRelease } from "../../src/modules/sources-identity/service/resolve.ts";
|
||||
import { sharedCache } from "../../src/shared-cache.ts";
|
||||
import { createHarness, type Harness, pgAvailable, seedSite } from "./harness.ts";
|
||||
|
||||
const pg = await pgAvailable();
|
||||
|
||||
const SAMPLES = new URL("../../../../packages/adapters/test/differential/inputs/detail/", import.meta.url);
|
||||
const audiencesDetail = (id: string) => readFileSync(new URL(`rl-audiences-detail-${id}.html.txt`, SAMPLES), "utf8");
|
||||
|
||||
const fx = (
|
||||
type: "movie" | "tv",
|
||||
id: number,
|
||||
en: string,
|
||||
zh: string,
|
||||
date: string,
|
||||
imdb: string | null = null,
|
||||
): TmdbFixture => ({
|
||||
type,
|
||||
id,
|
||||
original_title: en,
|
||||
original_language: "zh",
|
||||
origin_country: ["CN"],
|
||||
date,
|
||||
titles: { "zh-CN": zh, "en-US": en },
|
||||
popularity: 10,
|
||||
alternative_titles: [],
|
||||
external_ids: { imdb_id: imdb },
|
||||
...(type === "tv"
|
||||
? {
|
||||
seasons: [
|
||||
{ season_number: 1, air_date: date, names: { "zh-CN": "第 1 季", "en-US": "Season 1" }, episodes: [] },
|
||||
],
|
||||
}
|
||||
: {}),
|
||||
});
|
||||
|
||||
describe.skipIf(!pg)("ID-3 识别升级链:详情页 ID", () => {
|
||||
let h: Harness;
|
||||
let tmdb: MockTmdb;
|
||||
let audiences: string;
|
||||
let hhan: string;
|
||||
const siteRequests: string[] = [];
|
||||
|
||||
beforeAll(async () => {
|
||||
tmdb = await createMockTmdb({
|
||||
fixtures: [
|
||||
// 观众 727077《我不是大师》:站点英文名 A Prophet 与 TMDB 英文名对不上
|
||||
fx("tv", 990002, "The Fortune Teller", "我不是大师", "2026-09-01", "tt36758489"),
|
||||
// 观众 725858《Rose》:同名同年两部 → 标题识别有歧义
|
||||
fx("movie", 990001, "Rose", "罗斯", "2026-03-01", "tt25399154"),
|
||||
fx("movie", 990011, "Rose", "玫瑰", "2026-05-01", "tt0000011"),
|
||||
// 憨憨详情样本:TMDB tv/290001(站点英文名 Example Show 搜不到)
|
||||
fx("tv", 290001, "Sample Series", "示例剧集", "2026-08-01", "tt45950001"),
|
||||
// 豆瓣路径:只凭豆瓣条目(标题 + 年份)匹配
|
||||
fx("tv", 990021, "Lanterns of Qing", "青灯记", "2025-04-01"),
|
||||
],
|
||||
});
|
||||
h = await createHarness("id3", {
|
||||
tmdbUrl: tmdb.url,
|
||||
// 站点请求全部走样本;没有样本的地址直接报错(零真实请求)
|
||||
fetch: async (input) => {
|
||||
const u = new URL(typeof input === "string" ? input : input instanceof URL ? input.href : input.url);
|
||||
siteRequests.push(`${u.host}${u.pathname}${u.search}`);
|
||||
const id = u.searchParams.get("id") ?? "";
|
||||
const html = (body: string) =>
|
||||
new Response(body, { status: 200, headers: { "content-type": "text/html; charset=utf-8" } });
|
||||
if (u.host === "audiences.me" && u.pathname === "/details.php") return html(audiencesDetail(id));
|
||||
if (u.host === "hhanclub.net" && u.pathname === "/details.php") return html(HHANCLUB_DETAIL_PAGE_2026);
|
||||
if (u.host === "hhanclub.net" && u.pathname === "/viewfilelist.php") return html(HHANCLUB_FILELIST_FRAGMENT);
|
||||
throw new Error(`测试不允许的站点请求 ${u.href}`);
|
||||
},
|
||||
});
|
||||
audiences = (await seedSite(h, { key: "audiences", baseUrl: "https://audiences.me/", cookie: "c_secure_uid=1" }))
|
||||
.site.id;
|
||||
hhan = (await seedSite(h, { key: "hhanclub", baseUrl: "https://hhanclub.net/", cookie: "c_secure_uid=2" })).site.id;
|
||||
});
|
||||
afterAll(async () => {
|
||||
await h?.close();
|
||||
await tmdb?.close();
|
||||
});
|
||||
beforeEach(async () => {
|
||||
await setCap(10);
|
||||
await sql`delete from kv_cache where region = ${ESCALATION_COUNTER_REGION}`.execute(h.db);
|
||||
});
|
||||
|
||||
const setCap = async (cap: number) => {
|
||||
await sql`insert into settings (key, value) values ('identity', ${JSON.stringify({ detailIdsPerSiteHour: cap })}::jsonb)
|
||||
on conflict (key) do update set value = excluded.value`.execute(h.db);
|
||||
};
|
||||
const release = (siteId: string, siteReleaseId: string, title: string, raw: unknown = {}) =>
|
||||
h.db
|
||||
.insertInto("source_release")
|
||||
.values({ site_id: siteId, site_release_id: siteReleaseId, title, raw: JSON.stringify(raw) } as never)
|
||||
.returningAll()
|
||||
.executeTakeFirstOrThrow();
|
||||
const row = (id: string) =>
|
||||
h.db.selectFrom("source_release").selectAll().where("id", "=", id).executeTakeFirstOrThrow();
|
||||
const detailRequests = (id: string) =>
|
||||
siteRequests.filter((r) => r.includes("/details.php") && r.includes(`id=${id}&`));
|
||||
const tmdbIdOf = async (entityId: string) =>
|
||||
(
|
||||
await h.db
|
||||
.selectFrom("external_identity")
|
||||
.select("external_id")
|
||||
.where("media_entity_id", "=", entityId)
|
||||
.where("provider", "=", "tmdb")
|
||||
.executeTakeFirstOrThrow()
|
||||
).external_id;
|
||||
const openDecisions = () =>
|
||||
h.db
|
||||
.selectFrom("decision")
|
||||
.select(["id", "kind", "status"])
|
||||
.where("kind", "in", ["identity_ambiguous", "identity_unresolved"])
|
||||
.where("status", "in", ["open", "investigating", "awaiting_user"])
|
||||
.execute();
|
||||
|
||||
it("观众 727077:列表只有豆瓣 ID,搜索结果标题认不出 → 抓详情取 IMDb → TMDB find → 自动确认;不登记对不上的站点名", async () => {
|
||||
// 列表页只给了豆瓣 ID(样本 rl-audiences-list 里这一行没有 IMDb 徽标)
|
||||
const r = await release(audiences, "727077", "A Prophet S01E16 2026 2160p iQIYI WEB-DL H265 HDR DDP5.1-ADWeb", {
|
||||
_site: { doubanId: "37297001" },
|
||||
});
|
||||
const find0 = tmdb.callsTo(/\/3\/find\/tt36758489/).length;
|
||||
const res = await resolveRelease(h.rt, { sourceReleaseId: r.id, trigger: "search" });
|
||||
expect(res.outcome).toBe("auto");
|
||||
expect(detailRequests("727077")).toHaveLength(1);
|
||||
expect(tmdb.callsTo(/\/3\/find\/tt36758489/).length).toBe(find0 + 1);
|
||||
expect(await tmdbIdOf(res.mediaEntityId!)).toBe("990002");
|
||||
const after = await row(r.id);
|
||||
expect(after.media_entity_id).toBe(res.mediaEntityId);
|
||||
expect((after.raw as { _detail?: unknown })._detail).toMatchObject({ imdbId: "tt36758489", doubanId: "37297001" });
|
||||
expect((after.raw as { _escalation?: { status: string } })._escalation?.status).toBe("fetched");
|
||||
expect(after.resolution_evidence).toMatchObject({
|
||||
escalation: { detail: "fetched", via: "site_id", pending: false },
|
||||
});
|
||||
// 站点英文名 A Prophet 与作品名对不上:只凭 ID 确认,不登记为别名(ID-2 守卫照旧)
|
||||
const aliases = await h.db
|
||||
.selectFrom("title_alias")
|
||||
.select("alias")
|
||||
.where("media_entity_id", "=", res.mediaEntityId!)
|
||||
.where("source", "=", "site")
|
||||
.execute();
|
||||
expect(aliases.map((a) => a.alias)).not.toContain("A Prophet");
|
||||
expect(await escalationCount(h.rt, audiences)).toBe(1);
|
||||
|
||||
// 再识别一次:详情已抓过,不再请求站点;IDs 在 raw._detail 里,第一轮就按 ID 认出
|
||||
const again = await resolveRelease(h.rt, { sourceReleaseId: r.id, force: true, trigger: "manual" });
|
||||
expect(["auto", "skipped"]).toContain(again.outcome);
|
||||
expect(detailRequests("727077")).toHaveLength(1);
|
||||
});
|
||||
|
||||
it("观众 725858:同名同年两部 → 有歧义的事项;手动重新识别抓详情 → IMDb 确认 → 事项自动关闭;详情页豆瓣信息块进豆瓣条目缓存", async () => {
|
||||
const r = await release(audiences, "725858", "Rose 2026 1080p BluRay AVC DTS-HD MA 5.1-DIY@Audies");
|
||||
// 不允许升级时(旧行为):歧义 → 开 identity_ambiguous
|
||||
const first = await resolveRelease(h.rt, { sourceReleaseId: r.id, trigger: "manual", escalate: false });
|
||||
expect(first.outcome).toBe("ambiguous");
|
||||
expect(first.decisionId).not.toBeNull();
|
||||
expect(detailRequests("725858")).toHaveLength(0);
|
||||
|
||||
const res = await resolveRelease(h.rt, { sourceReleaseId: r.id, force: true, trigger: "manual" });
|
||||
expect(res.outcome).toBe("auto");
|
||||
expect(await tmdbIdOf(res.mediaEntityId!)).toBe("990001");
|
||||
expect(detailRequests("725858")).toHaveLength(1);
|
||||
const d = await h.db
|
||||
.selectFrom("decision")
|
||||
.select(["status"])
|
||||
.where("id", "=", first.decisionId!)
|
||||
.executeTakeFirst();
|
||||
expect(d?.status).toBe("resolved");
|
||||
const pool = await h.db
|
||||
.selectFrom("identity_candidate")
|
||||
.select(["status", "resolved_media_entity_id"])
|
||||
.where("source_release_id", "=", r.id)
|
||||
.executeTakeFirstOrThrow();
|
||||
expect(pool).toMatchObject({ status: "resolved", resolved_media_entity_id: res.mediaEntityId });
|
||||
// 详情页信息块(◎片名 Rose / ◎译名 罗斯 / ◎年代 2026)+ 豆瓣 ID 35322132 → 豆瓣条目
|
||||
const subject = await loadDoubanSubject(h.ctx, "35322132");
|
||||
expect(subject).toMatchObject({ id: "35322132", title: "Rose", year: 2026, source: "site_detail" });
|
||||
expect(subject?.aliases).toContain("罗斯");
|
||||
});
|
||||
|
||||
it("憨憨详情样本:MediaInfo 里的 TMDB tv/290001 直取详情 → 自动确认", async () => {
|
||||
const r = await release(hhan, "91003", "Example Show S01 2026 2160p YK WEB-DL H.265 DV HQ AAC2.0-HHWEB");
|
||||
const res = await resolveRelease(h.rt, { sourceReleaseId: r.id, trigger: "search" });
|
||||
expect(res.outcome).toBe("auto");
|
||||
expect(await tmdbIdOf(res.mediaEntityId!)).toBe("290001");
|
||||
expect(detailRequests("91003")).toHaveLength(1);
|
||||
expect((await row(r.id)).resolution_evidence).toMatchObject({ escalation: { via: "site_id" } });
|
||||
});
|
||||
|
||||
it("每站每小时上限:达到后推迟(不开事项、15 分钟内复查),不再请求站点;0 = 不抓", async () => {
|
||||
await setCap(1);
|
||||
const a = await release(audiences, "990101", "Nowhere Title 2026 1080p WEB-DL");
|
||||
const b = await release(audiences, "990102", "Another Nowhere 2026 1080p WEB-DL");
|
||||
const before = siteRequests.length;
|
||||
await resolveRelease(h.rt, { sourceReleaseId: a.id, trigger: "search" });
|
||||
const rb = await resolveRelease(h.rt, { sourceReleaseId: b.id, trigger: "search" });
|
||||
// a 用掉名额(样本里没有这个 id 的页面,详情请求失败也算用掉;失败不阻塞,照常进池,6 小时内不再为识别重抓)
|
||||
expect(siteRequests.length - before).toBe(1);
|
||||
expect((await row(a.id)).resolution_evidence).toMatchObject({
|
||||
escalation: { detail: "failed:fetch_failed", pending: false },
|
||||
});
|
||||
await resolveRelease(h.rt, { sourceReleaseId: a.id, force: true, trigger: "manual" });
|
||||
expect((await row(a.id)).resolution_evidence).toMatchObject({ escalation: { detail: "skipped:recent_failure" } });
|
||||
expect(siteRequests.length - before).toBe(1);
|
||||
expect(rb.decisionId).toBeNull();
|
||||
const pool = await h.db
|
||||
.selectFrom("identity_candidate")
|
||||
.select(["query", "next_review_at"])
|
||||
.where("source_release_id", "=", b.id)
|
||||
.executeTakeFirstOrThrow();
|
||||
expect(pool.query).toMatchObject({ escalation: { detail: "deferred:cap_reached", pending: true } });
|
||||
expect(pool.next_review_at!.getTime() - Date.now()).toBeLessThanOrEqual(15 * 60_000 + 5_000);
|
||||
|
||||
await setCap(0);
|
||||
const c = await release(audiences, "990103", "Third Nowhere 2026 1080p WEB-DL");
|
||||
await resolveRelease(h.rt, { sourceReleaseId: c.id, trigger: "search" });
|
||||
expect((await row(c.id)).resolution_evidence).toMatchObject({ escalation: { detail: "skipped:disabled" } });
|
||||
expect(siteRequests.length - before).toBe(1);
|
||||
});
|
||||
|
||||
it("订阅相关的资源:RSS 入库时不抓详情、不开「没识别出作品」;复查时抓详情认出;旧事项自动关闭", async () => {
|
||||
// 本地作品(如订阅)登记了别名「Example Show」,但没有 TMDB 身份、年份也对不上:资源与它相关(tracked_alias)却认不出
|
||||
const e = await h.db
|
||||
.insertInto("media_entity")
|
||||
.values({ kind: "series", title: "Example Show", year: 1990 } as never)
|
||||
.returning("id")
|
||||
.executeTakeFirstOrThrow();
|
||||
await sql`insert into title_alias (media_entity_id, alias, normalized, kind, source)
|
||||
values (${e.id}, 'Example Show', ${normalizeTitle("Example Show")}, 'other', 'user')`.execute(h.db);
|
||||
const r = await release(hhan, "91004", "Example Show S01E18 2027 2160p YK WEB-DL H.265 DV HQ AAC2.0-HHWEB");
|
||||
|
||||
// 旧行为(不升级):相关却认不出 → identity_unresolved
|
||||
const old = await resolveRelease(h.rt, { sourceReleaseId: r.id, trigger: "review", force: true, escalate: false });
|
||||
expect(old.decisionId).not.toBeNull();
|
||||
const oldKind = (
|
||||
await h.db.selectFrom("decision").select("kind").where("id", "=", old.decisionId!).executeTakeFirstOrThrow()
|
||||
).kind;
|
||||
|
||||
// RSS 入库(observed):不抓详情,推迟到复查;不再保留事项
|
||||
const obs = await resolveRelease(h.rt, { sourceReleaseId: r.id, force: true, trigger: "observed" });
|
||||
expect(obs.decisionId).toBeNull();
|
||||
expect(detailRequests("91004")).toHaveLength(0);
|
||||
expect((await row(r.id)).resolution_evidence).toMatchObject({
|
||||
escalation: { detail: "not_now:batch_ingest", pending: true },
|
||||
});
|
||||
const d1 = await h.db.selectFrom("decision").select("status").where("id", "=", old.decisionId!).executeTakeFirst();
|
||||
expect(d1?.status).not.toBe("open");
|
||||
|
||||
// 复查(订阅相关):抓详情 → TMDB tv/290001(年份 2027 与 2026 差 1 年,在容差内)→ 自动确认
|
||||
const rev = await resolveRelease(h.rt, { sourceReleaseId: r.id, force: true, trigger: "review" });
|
||||
expect(detailRequests("91004")).toHaveLength(1);
|
||||
expect(rev.outcome).toBe("auto");
|
||||
expect(oldKind).toBe("identity_unresolved");
|
||||
expect(await openDecisions()).toHaveLength(0);
|
||||
});
|
||||
|
||||
it("豆瓣路径:详情只有豆瓣 ID → 缓存的豆瓣条目(搜索摘要抽取)标题 + 年份 → TMDB 匹配(ID-2 打分)", async () => {
|
||||
// META-2 搜索摘要里抽到的豆瓣条目(不请求豆瓣)
|
||||
const s = doubanSubjectFromSnippet(
|
||||
{
|
||||
url: "https://movie.douban.com/subject/36000099/",
|
||||
title: "青灯记 (豆瓣)",
|
||||
snippet: "导演: 某导演 主演: 甲 / 乙 类型: 剧情 / 古装 制片国家/地区: 中国大陆 年份: 2025 集数: 30",
|
||||
},
|
||||
new Date(),
|
||||
);
|
||||
expect(s).toMatchObject({ id: "36000099", title: "青灯记", year: 2025, source: "web_snippet" });
|
||||
await saveDoubanSubject(h.ctx, s!);
|
||||
// 详情已抓过,只给了豆瓣 ID
|
||||
const r = await release(hhan, "91005", "Qing Deng Ji S01E03 2025 1080p WEB-DL H264 AAC-HHWEB", {
|
||||
_detail: { doubanId: "36000099" },
|
||||
});
|
||||
await h.db.updateTable("source_release").set({ detail_fetched_at: new Date() }).where("id", "=", r.id).execute();
|
||||
const res = await resolveRelease(h.rt, { sourceReleaseId: r.id, trigger: "manual" });
|
||||
expect(res.outcome).toBe("auto");
|
||||
expect(await tmdbIdOf(res.mediaEntityId!)).toBe("990021");
|
||||
expect(detailRequests("91005")).toHaveLength(0);
|
||||
expect((await row(r.id)).resolution_evidence).toMatchObject({
|
||||
escalation: { via: "douban", douban: { id: "36000099", title: "青灯记", source: "web_snippet" } },
|
||||
});
|
||||
|
||||
// 缓存里没有这个豆瓣条目:不请求豆瓣,只进池;有 ID 的资源不开「没识别出作品」
|
||||
await sharedCache(h.ctx).delete(DOUBAN_SUBJECT_REGION, "36000098");
|
||||
const r2 = await release(hhan, "91006", "Unknown Drama S01E01 2025 1080p WEB-DL", {
|
||||
_detail: { doubanId: "36000098" },
|
||||
});
|
||||
await h.db.updateTable("source_release").set({ detail_fetched_at: new Date() }).where("id", "=", r2.id).execute();
|
||||
const res2 = await resolveRelease(h.rt, { sourceReleaseId: r2.id, trigger: "manual" });
|
||||
expect(res2.outcome).toBe("no_match");
|
||||
expect(res2.decisionId).toBeNull();
|
||||
expect(siteRequests.some((x) => x.includes("douban"))).toBe(false);
|
||||
});
|
||||
|
||||
it("池条目补全前先走升级链:详情页给了 ID 就直接认出,返回 409,不做网页补全", async () => {
|
||||
const r = await release(hhan, "91007", "Example Show S01 2026 2160p YK WEB-DL H.265 DV HQ AAC2.0-HHWEB");
|
||||
const first = await resolveRelease(h.rt, { sourceReleaseId: r.id, trigger: "manual", escalate: false });
|
||||
expect(first.outcome).toBe("no_match");
|
||||
const pool = await h.db
|
||||
.selectFrom("identity_candidate")
|
||||
.select("id")
|
||||
.where("source_release_id", "=", r.id)
|
||||
.executeTakeFirstOrThrow();
|
||||
const res = await h.req(`/identity-pool/${pool.id}/enrich`, { body: {} });
|
||||
expect(res.status).toBe(409);
|
||||
expect(((await res.json()) as { error: { message: string } }).error.message).toContain("《示例剧集》");
|
||||
expect(detailRequests("91007")).toHaveLength(1);
|
||||
expect((await row(r.id)).resolution_status).toBe("resolved");
|
||||
});
|
||||
|
||||
it("站点详情来源的豆瓣条目不会被搜索摘要覆盖", async () => {
|
||||
await saveDoubanSubject(h.ctx, {
|
||||
id: "1",
|
||||
title: "甲",
|
||||
aliases: [],
|
||||
year: 2020,
|
||||
source: "site_detail",
|
||||
ref: "x",
|
||||
at: new Date().toISOString(),
|
||||
});
|
||||
await saveDoubanSubject(h.ctx, {
|
||||
id: "1",
|
||||
title: "乙",
|
||||
aliases: [],
|
||||
year: 2021,
|
||||
source: "web_snippet",
|
||||
ref: "y",
|
||||
at: new Date().toISOString(),
|
||||
});
|
||||
expect((await loadDoubanSubject(h.ctx, "1"))?.title).toBe("甲");
|
||||
});
|
||||
});
|
||||
Reference in new issue
Block a user