Files
Arcrun/kbdb/tests/triplet-library-batch-backfill.test.ts
T
uncle6me-web 179dd60571 fix(kbdb): 三元組版標庫批次補存量+對外通道(Arcrun#87 二次收尾)
藏書地圖(library-map.ts)讀的是三元組 record 自己的 'library' slot(entry_values),
不是 entries.metadata_json.library——Arcrun#85 建的 POST /entries/backfill-library
補的是後者,補了地圖依然是 0。既有能補三元組 library 值的通道只有單筆
PATCH /kbdb/records/:id(b6ef0f0,已在 main),母體上千筆時逐筆呼叫不現實,也沒有
節流/冪等保護。

新增:
- kbdb/src/actions/library-backfill.ts:backfillTripletLibraryTags +
  tripletLibraryBackfillStatus,安全原則與既有 entries 版逐條對齊——呼叫端決定
  library 值+source_prefix 篩選(base 不猜語意)、冪等(NOT EXISTS 找缺 library 的
  那半)、與 entries 版/embed reconcile 共用同一顆 D69 每日 D1 寫入額度、owner_id 必填、
  寫入沿用既有 updateRecord(不手刻第二套 entry_values UPSERT SQL)。
- kbdb/src/routes/records.ts:POST /records/backfill-library + GET .../status。
- cypher-executor/src/routes/kbdb-proxy.ts:對外通道,owner_id 強制用租戶身份(同本檔
  既有 POST /kbdb/records 慣例,不信任 caller 自帶 owner_id)。

測試:kbdb 新增 5 案(含跨租戶隔離/冪等/D69 共用額度截斷/status 統計),全套
218/218 綠燈;cypher-executor 新增 5 案(租戶閘/參數驗證/轉發/owner_id 覆蓋),
兩份既有失敗(auth.test.ts ExecutionContext 型別缺陷、portal-data/portal-admin/
console-library-map-page/executor.test.ts 訊息措辭)複驗與本次改動無關(stash 前後
一致,pre-existing)。

紅線:未併 main、未部署、未對 leo21c 執行任何實際 backfill(缺安全的寫入通道——
MCP 工具集無 update/patch record 能力,CLI 全域設定指著 leo21c 不可裸跑)。
批次執行是下一步,需總管/leo 決定執行方式與時機。

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-13 11:12:56 +08:00

165 lines
9.9 KiB
TypeScript
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
// 三元組版標庫「批次」backfillArcrun#87 二次收尾,2026-08-13)測試。
//
// 背景:kbdb/tests/triplet-library-backfill.test.ts2026-08-11)已經證明「逐筆 updateRecord
// 補標庫」這條寫入路徑語意正確(源頭順序/存量補標/冪等三案)。但那份測試操作的是逐筆手刻的
// backfillPass() helper,不是可對外呼叫的批次能力——通道只有單筆 PATCH /kbdb/records/:id
// (b6ef0f0),母體上千筆時逐筆 PATCH 給呼叫端自己刻節流/冪等不現實。本檔驗證新加的
// backfillTripletLibraryTagskbdb/src/actions/library-backfill.ts):把「呼叫端給 library
// + source_prefix,一次處理一批」包成一個安全、節流的函式,語意與 entries 版
// library-backfill.test.ts)逐條對齊。
//
// 覆蓋:
// 1. 只補「符合 source_prefix、目前未標記 library」的候選三元組;已標記的不動(冪等)
// 2. owner_id 必填(缺了要拋錯,同 2026-08-11 leo 直令)
// 3. 與 entries 版 backfillEntryLibraryTags 共用同一顆每日 D1 寫入額度(D69:兩者都要受管,
// 任一邊獨立跑就是把另一邊的閘繞過去)
// 4. status 欄位存在的 record 也一起處理(backfill 不管 active/superseded,那是地圖聚合層的
// 篩選責任,不是補標層的責任——見 library-map.ts withLib 的 COALESCE 判準)
//
// 測試手法沿 triplet-library-backfill.test.tslibrary-backfill.test.ts 慣例:真 node:sqlite
// 套 migrations/0001_base.sql 原檔(本檔只用得到 templates/entries/entry_values 三張表,
// 不涉 library_map,不需 0003),kbdb-sql-ok 行尾標記同既有慣例(測試治具本身,非牆外業務邏輯繞過 API)。
import { describe, it, expect } from 'vitest';
import { DatabaseSync } from 'node:sqlite';
import { readFileSync } from 'node:fs';
import { backfillTripletLibraryTags, tripletLibraryBackfillStatus } from '../src/actions/library-backfill';
import { backfillEntryLibraryTags } from '../src/actions/library-backfill';
import { createTemplate, createRecord, getRecord } from '../src/actions/record-crud';
import type { Bindings } from '../src/types';
function makeSqliteD1(): D1Database {
const raw = new DatabaseSync(':memory:');
raw.exec(readFileSync(new URL('../migrations/0001_base.sql', import.meta.url), 'utf8')); // kbdb-sql-ok:測試治具(node:sqlite→D1 shim)套 migration 原檔
function stmt(sql: string, params: unknown[]) {
const s = {
bind(...args: unknown[]) { return stmt(sql, args); },
async all<T>() { return { results: raw.prepare(sql).all(...params) as T[] }; }, // kbdb-sql-ok:測試治具(node:sqlite→D1 shim
async first<T>() { return (raw.prepare(sql).get(...params) ?? null) as T | null; }, // kbdb-sql-ok:測試治具(node:sqlite→D1 shim
async run() {
const r = raw.prepare(sql).run(...params); // kbdb-sql-ok:測試治具(node:sqlite→D1 shim),非牆外業務邏輯繞過 API
return { success: true, meta: { changes: r.changes } };
},
};
return s;
}
return { prepare: (sql: string) => stmt(sql, []) } as unknown as D1Database;
}
function makeEnv(db: D1Database, opts: { maintenanceLimit?: string } = {}): Bindings {
return {
DB: db,
ENVIRONMENT: 'test',
KBDB_MAINTENANCE_DAILY_WRITE_LIMIT: opts.maintenanceLimit,
} as unknown as Bindings;
}
// prod 實際 triplet template 的 slotslibrary-map.test.tstriplet-library-backfill.test.ts 同款常數)。
const PROD_TRIPLET_SLOTS = [
'subject', 'predicate', 'object', 'source_block_id', 'confidence', 'clusters_json',
'bridge_score', 'subject_entity_type', 'object_entity_type', 'status', 'superseded_by',
'source_uri', 'content_hash', 'source_anchor', 'predicate_embed',
];
async function seedTripletTemplate(db: D1Database): Promise<void> {
await createTemplate(db, { id: 'tpl-triplet-test', name: 'triplet', slots: PROD_TRIPLET_SLOTS, created_by: 'kbdb-graph' });
}
describe('backfillTripletLibraryTags — 批次補三元組的 library slot(藏書地圖真正讀的那一半)', () => {
it('只補符合 source_prefix、目前缺 library 的三元組;owner 不符/已標記過的不動', async () => {
const db = makeSqliteD1();
await seedTripletTemplate(db);
const env = makeEnv(db);
const r1 = await createRecord(db, { template: 'triplet', values: { subject: 'A', predicate: 'r', object: 'B', source_uri: 'gitea:Leo/Arcrun@a.md' }, owner_id: 'bfezv28v' });
const r2 = await createRecord(db, { template: 'triplet', values: { subject: 'C', predicate: 'r', object: 'D', source_uri: 'gitea:Leo/Arcrun@b.md' }, owner_id: 'bfezv28v' });
// 別的 owner,同前綴——不該被補到(跨租戶隔離)。
const r3 = await createRecord(db, { template: 'triplet', values: { subject: 'E', predicate: 'r', object: 'F', source_uri: 'gitea:Leo/Arcrun@c.md' }, owner_id: 'someone-else' });
// 不符 source_prefix——不該被補到。
const r4 = await createRecord(db, { template: 'triplet', values: { subject: 'G', predicate: 'r', object: 'H', source_uri: 'gitea:Leo/mira@d.md' }, owner_id: 'bfezv28v' });
const result = await backfillTripletLibraryTags(db, env, {
library: 'arcrun',
owner_id: 'bfezv28v',
source_prefix: 'gitea:Leo/Arcrun@',
});
expect(result.scanned).toBe(2);
expect(result.tagged).toBe(2);
expect(result.remaining).toBe(0);
const rec1 = await getRecord(db, r1.record_id);
const rec2 = await getRecord(db, r2.record_id);
const rec3 = await getRecord(db, r3.record_id);
const rec4 = await getRecord(db, r4.record_id);
expect(rec1!.values.library).toBe('arcrun');
expect(rec2!.values.library).toBe('arcrun');
expect(rec3!.values.library).toBeUndefined(); // 別的 owner,沒被動到
expect(rec4!.values.library).toBeUndefined(); // 不符前綴,沒被動到
});
it('冪等:已標記過 library 的三元組不會被第二輪 touch(不會覆蓋成別的值)', async () => {
const db = makeSqliteD1();
await seedTripletTemplate(db);
const env = makeEnv(db);
const r1 = await createRecord(db, { template: 'triplet', values: { subject: 'A', predicate: 'r', object: 'B', source_uri: 'gitea:Leo/kb@a.md' }, owner_id: 'bfezv28v' });
const first = await backfillTripletLibraryTags(db, env, { library: 'kb', owner_id: 'bfezv28v', source_prefix: 'gitea:Leo/kb@' });
expect(first.tagged).toBe(1);
// 第二輪同條件再跑一次:這筆已經有 library 值了,不該再被選中,即使再給一個不同的 library 值。
const second = await backfillTripletLibraryTags(db, env, { library: 'something-else', owner_id: 'bfezv28v', source_prefix: 'gitea:Leo/kb@' });
expect(second.scanned).toBe(0);
expect(second.tagged).toBe(0);
const rec1 = await getRecord(db, r1.record_id);
expect(rec1!.values.library).toBe('kb'); // 沒被第二輪的 'something-else' 洗掉
});
it('owner_id 必填 — 缺了直接拋錯,不准無租戶範圍地掃全庫', async () => {
const db = makeSqliteD1();
await seedTripletTemplate(db);
const env = makeEnv(db);
await expect(backfillTripletLibraryTags(db, env, { library: 'arcrun', owner_id: '' })).rejects.toThrow(/owner_id required/);
});
it('與 entries 版 backfillEntryLibraryTags 共用同一顆每日 D1 寫入額度(D69:任一邊都不能繞過去)', async () => {
const db = makeSqliteD1();
await seedTripletTemplate(db);
const env = makeEnv(db, { maintenanceLimit: '3' }); // 每日只剩 3 次背景維護寫入額度
// entries 版先用掉 2 筆額度。
await db.prepare(`INSERT INTO entries (id, content, entry_type, owner_id, created_at, updated_at) VALUES (?, 'x', 'block', 'bfezv28v', 1000, 1000)`).bind('e1').run(); // kbdb-sql-ok:測試治具灌資料
await db.prepare(`INSERT INTO entries (id, content, entry_type, owner_id, created_at, updated_at) VALUES (?, 'x', 'block', 'bfezv28v', 1000, 1000)`).bind('e2').run(); // kbdb-sql-ok:測試治具灌資料
const entriesResult = await backfillEntryLibraryTags(db, env, { library: 'notes', owner_id: 'bfezv28v' });
expect(entriesResult.tagged).toBe(2);
// 三元組版此時只剩 1 次額度可用,即使候選有 2 筆,也只能補 1 筆、如實回報 quota_exceeded。
await createRecord(db, { template: 'triplet', values: { subject: 'A', predicate: 'r', object: 'B', source_uri: 'gitea:Leo/Arcrun@a.md' }, owner_id: 'bfezv28v' });
await createRecord(db, { template: 'triplet', values: { subject: 'C', predicate: 'r', object: 'D', source_uri: 'gitea:Leo/Arcrun@b.md' }, owner_id: 'bfezv28v' });
const tripletResult = await backfillTripletLibraryTags(db, env, { library: 'arcrun', owner_id: 'bfezv28v', source_prefix: 'gitea:Leo/Arcrun@' });
expect(tripletResult.scanned).toBe(2);
expect(tripletResult.tagged).toBe(1); // 只剩 1 額度,被截斷
expect(tripletResult.quota_exceeded).toBe(true);
expect(tripletResult.remaining).toBe(1); // 還有 1 筆沒補到
});
it('tripletLibraryBackfillStatus 回報還剩多少待補(backfill 前後都能查)', async () => {
const db = makeSqliteD1();
await seedTripletTemplate(db);
const env = makeEnv(db);
await createRecord(db, { template: 'triplet', values: { subject: 'A', predicate: 'r', object: 'B', source_uri: 'gitea:Leo/notes@a.md' }, owner_id: 'bfezv28v' });
await createRecord(db, { template: 'triplet', values: { subject: 'C', predicate: 'r', object: 'D', source_uri: 'gitea:Leo/notes@b.md' }, owner_id: 'bfezv28v' });
const before = await tripletLibraryBackfillStatus(db, { owner_id: 'bfezv28v', source_prefix: 'gitea:Leo/notes@' });
expect(before.pending).toBe(2);
await backfillTripletLibraryTags(db, env, { library: 'notes', owner_id: 'bfezv28v', source_prefix: 'gitea:Leo/notes@' });
const after = await tripletLibraryBackfillStatus(db, { owner_id: 'bfezv28v', source_prefix: 'gitea:Leo/notes@' });
expect(after.pending).toBe(0);
});
});