tt-help-cli-ycl 1.4.166 → 1.4.168

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "tt-help-cli-ycl",
-  "version": "1.4.166",
+  "version": "1.4.168",
   "description": "TikTok user & video data scraper - extract ttSeller, verified, locationCreated from HTML source",
   "type": "module",
   "bin": {

package/scripts/deploy-node.sh

@@ -0,0 +1,83 @@
+#!/bin/bash
+# 将本地 tiktok-cli 打包部署到 Windows 采集节点（试运行，不经 npm registry）
+# 用法:
+#   ./scripts/deploy-node.sh tiktok-210                  # 打包+传输+远程安装+校验
+#   ./scripts/deploy-node.sh tiktok-210 --bump           # 先把 package.json 版本号 +1 再打包
+#   ./scripts/deploy-node.sh tiktok-210 --bump --restart # 安装后重启走 CDP 的命令进程
+#                                                        #（watchdog 检测到退出会用新代码自动拉起）
+# 流程: (可选)升版本号 → npm pack → scp → 远程 npm install -g → 校验 → (可选)重启命令
+# 前提: ~/.ssh/config 已配置节点别名（如 tiktok-210，经跳板机），BatchMode 免密可用
+# 说明:
+#   - 无需停进程再安装：采集客户端不加载 better-sqlite3（只有 videostats/watch 服务端用），无文件锁
+#   - watchdog 自升级：npm 覆盖安装后守护进程经 upgrader.js 自动重启，受管命令随后重建
+#   - 节点上保留 tgz 安装包，回滚时 npm install -g 旧版 tgz 即可
+set -euo pipefail
+
+ALIAS="${1:?用法: deploy-node.sh <ssh别名> [--bump] [--restart]}"
+shift || true
+DO_BUMP=0
+DO_RESTART=0
+for a in "$@"; do
+  case "$a" in
+    --bump) DO_BUMP=1 ;;
+    --restart) DO_RESTART=1 ;;
+    *) echo "未知参数: $a"; exit 1 ;;
+  esac
+done
+
+cd "$(dirname "$0")/.."
+
+if [ "$DO_BUMP" = 1 ]; then
+  node -e "
+    const fs = require('fs');
+    const p = JSON.parse(fs.readFileSync('package.json', 'utf8'));
+    const m = p.version.split('.');
+    m[2] = Number(m[2]) + 1;
+    p.version = m.join('.');
+    fs.writeFileSync('package.json', JSON.stringify(p, null, 2) + '\n');
+    console.log('版本号 ->', p.version);
+  "
+fi
+
+NAME=$(node -p "require('./package.json').name")
+VER=$(node -p "require('./package.json').version")
+TGZ="${NAME}-${VER}.tgz"
+
+echo "==> npm pack"
+rm -f "$TGZ"
+TGZ=$(npm pack 2>/dev/null | tail -1)
+echo "包: $TGZ ($(du -h "$TGZ" | cut -f1))"
+
+echo "==> scp → $ALIAS"
+scp -o ConnectTimeout=25 -o BatchMode=yes "$TGZ" "$ALIAS:$TGZ" 2>&1 |
+  grep -vE 'post-quantum|openssh.com/pq|store now|upgraded|CLIXML|^Warning|Permanently added' >/dev/null
+echo "已传输到 %USERPROFILE%\\$TGZ"
+
+echo "==> 远程 npm install -g"
+ssh -o ConnectTimeout=25 -o BatchMode=yes "$ALIAS" \
+  "npm install -g %USERPROFILE%\\$TGZ 2>&1" |
+  grep -vE 'post-quantum|openssh.com/pq|store now|upgraded|CLIXML|^Warning|Permanently added' |
+  tail -3
+
+echo "==> 校验远程版本"
+REMOTE_VER=$(
+  ssh -o ConnectTimeout=25 -o BatchMode=yes "$ALIAS" "tt-help --version 2>nul" 2>/dev/null |
+    grep -vE 'post-quantum|openssh.com/pq|store now|upgraded|CLIXML|^Warning|Permanently added' |
+    tail -1 | tr -d '\r'
+)
+if [ "$REMOTE_VER" = "$VER" ]; then
+  echo "✓ 远程版本 $REMOTE_VER"
+else
+  echo "✗ 版本不一致: 期望 $VER, 实际 '${REMOTE_VER:-空}'"
+  exit 1
+fi
+
+if [ "$DO_RESTART" = 1 ]; then
+  echo "==> 重启走 CDP 的命令进程（排除 watchdog/webserver，浏览器进程保留、新客户端直接重连）"
+  ssh -o ConnectTimeout=25 -o BatchMode=yes "$ALIAS" 'powershell -NoProfile -Command "Get-CimInstance Win32_Process -Filter \"Name='node.exe'\" | Where-Object { $_.CommandLine -match \"cli\.js\" -and $_.CommandLine -notmatch \"watchdog|webserver\" } | ForEach-Object { taskkill /F /PID $_.ProcessId }"' 2>&1 |
+    grep -vE 'post-quantum|openssh.com/pq|store now|upgraded|CLIXML|^Warning|Permanently added' |
+    head -10 || true
+  echo "（watchdog 检测到异常退出后会在 5-30s 内用新代码自动拉起，Edge 进程保留、客户端直连不重启浏览器）"
+fi
+
+echo "==> 完成"

package/scripts/diag-claim-branches.js

@@ -0,0 +1,72 @@
+// 只读诊断：201 本机实测 claimNextJob 各分支 SQL 耗时（readonly，不动生产数据）
+// 用完即删
+import path from "path";
+import Database from "file://D:/tiktok/tiktok-cli/node_modules/better-sqlite3/lib/index.js";
+const db = new Database(path.join("D:", "tiktok", "data", "result.db"), {
+  readonly: true,
+  timeout: 600000,
+});
+
+function t(name, sql, args = []) {
+  const stmt = db.prepare(sql);
+  const t0 = Date.now();
+  const r = stmt.get(...args);
+  console.log(
+    `${String(Date.now() - t0).padStart(6)}ms  ${r ? "HIT " : "MISS"}  ${name}`,
+  );
+}
+
+const ORDER = "ORDER BY follower_count DESC, created_at ASC, unique_id ASC LIMIT 1";
+const SC = "(status_code IS NULL OR status_code = 0 OR status_code = 209002)";
+
+console.log("=== A. pinned 分支（池当前为空）===");
+t(
+  "pinned+video",
+  `SELECT unique_id FROM task_queue WHERE status='pending' AND pinned=1 AND video_count>0 AND ${SC} ${ORDER}`,
+);
+t(
+  "pinned",
+  `SELECT unique_id FROM task_queue WHERE status='pending' AND pinned=1 AND ${SC} ${ORDER}`,
+);
+
+console.log("=== B. seed 级首国分支（MX）：OR 空位置 原版形态 ===");
+t(
+  "seed MX OR-empty (原版)",
+  `SELECT unique_id FROM task_queue WHERE status='pending' AND pinned=0 AND instr(COALESCE(sources,''), '"seed"') > 0 AND video_count > 0 AND (UPPER(COALESCE(guessed_location,'')) IN ('MX') OR COALESCE(guessed_location,'') = '') AND ${SC} ${ORDER}`,
+);
+
+console.log("=== C. 拆分形态：单值 IN 点查 ===");
+t(
+  "seed MX only",
+  `SELECT unique_id FROM task_queue WHERE status='pending' AND pinned=0 AND instr(COALESCE(sources,''), '"seed"') > 0 AND video_count > 0 AND UPPER(COALESCE(guessed_location,'')) IN ('MX') AND ${SC} ${ORDER}`,
+);
+t(
+  "seed empty-loc only",
+  `SELECT unique_id FROM task_queue WHERE status='pending' AND pinned=0 AND instr(COALESCE(sources,''), '"seed"') > 0 AND video_count > 0 AND UPPER(COALESCE(guessed_location,'')) = '' AND ${SC} ${ORDER}`,
+);
+t(
+  "seed ID only (最大池 10.8万)",
+  `SELECT unique_id FROM task_queue WHERE status='pending' AND pinned=0 AND instr(COALESCE(sources,''), '"seed"') > 0 AND video_count > 0 AND UPPER(COALESCE(guessed_location,'')) IN ('ID') AND ${SC} ${ORDER}`,
+);
+
+console.log("=== D. seller/follow/other 级抽测（MX 单值）===");
+t(
+  "seller MX hasVideo",
+  `SELECT unique_id FROM task_queue WHERE status='pending' AND pinned=0 AND tt_seller=1 AND verified=0 AND video_count>0 AND UPPER(COALESCE(guessed_location,'')) IN ('MX') AND ${SC} ${ORDER}`,
+);
+t(
+  "follow MX hasVideo",
+  `SELECT unique_id FROM task_queue WHERE status='pending' AND pinned=0 AND (instr(COALESCE(sources,''), '"following"') > 0 OR instr(COALESCE(sources,''), '"follower"') > 0 OR instr(COALESCE(sources,''), '"comment"') > 0) AND video_count>0 AND UPPER(COALESCE(guessed_location,'')) IN ('MX') AND ${SC} ${ORDER}`,
+);
+t(
+  "other MX hasVideo",
+  `SELECT unique_id FROM task_queue WHERE status='pending' AND pinned=0 AND video_count>0 AND UPPER(COALESCE(guessed_location,'')) IN ('MX') AND ${SC} ${ORDER}`,
+);
+console.log("=== E. NOT IN 24国兜底分支 ===");
+t(
+  "other NOT-IN fallback",
+  `SELECT unique_id FROM task_queue WHERE status='pending' AND pinned=0 AND video_count>0 AND UPPER(COALESCE(guessed_location,'')) NOT IN ('MX','JP','BR','SG','ID','VN','PH','PL','NL','BE','US','TH','MY','AT','CZ','HU','GR','PT','DE','FR','IT','IE','ES','GB') AND ${SC} ${ORDER}`,
+);
+db.close();
+console.log("done");
+

package/scripts/test-claim-fix.mjs

@@ -0,0 +1,216 @@
+// claimNextJob 修复验证：功能语义 + 性能（合成 32 万行模拟生产行数）
+// 用完即删
+import fs from "fs";
+import { createStore } from "../src/watch/data-store.js";
+import { getDb } from "../src/watch/db-schema.js";
+
+const TMP = "/tmp/tiktok-claim-test";
+fs.rmSync(TMP, { recursive: true, force: true });
+fs.mkdirSync(TMP, { recursive: true });
+
+function freshStore(name) {
+  const p = `${TMP}/${name}.db`;
+  const store = createStore(p, {});
+  return { store, db: getDb(), path: p };
+}
+
+function insertRows(db, rows) {
+  const stmt = db.prepare(`
+    INSERT OR REPLACE INTO task_queue
+      (unique_id, status, guessed_location, video_count, follower_count,
+       tt_seller, verified, status_code, pinned, sources, created_at, claimed_at, claimed_by)
+    VALUES
+      (@unique_id, @status, @guessed_location, @video_count, @follower_count,
+       @tt_seller, @verified, @status_code, @pinned, @sources, @created_at, @claimed_at, @claimed_by)
+  `);
+  const tx = db.transaction((list) => {
+    for (const r of list) stmt.run(r);
+  });
+  tx(rows);
+}
+
+const baseRow = {
+  status: "pending",
+  guessed_location: "",
+  video_count: 5,
+  follower_count: 1000,
+  tt_seller: 0,
+  verified: 0,
+  status_code: null,
+  pinned: 0,
+  sources: '["explore"]',
+  created_at: "2026-01-01T00:00:00Z",
+  claimed_at: null,
+  claimed_by: null,
+};
+const row = (over) => ({ ...baseRow, unique_id: `u_${Math.random().toString(36).slice(2, 9)}`, ...over });
+
+let pass = 0, fail = 0;
+function check(name, cond, detail = "") {
+  if (cond) { pass++; console.log(`  ✅ ${name}${detail ? "  " + detail : ""}`); }
+  else { fail++; console.log(`  ❌ ${name}${detail ? "  " + detail : ""}`); }
+}
+
+// ───────── B. 功能语义（小数据集，各用独立库） ─────────
+console.log("\n== B1. 首国等值严格优先于空位置（即使空位置粉丝更高）==");
+{
+  const { store, db } = freshStore("b1");
+  insertRows(db, [
+    row({ unique_id: "empty-high", guessed_location: "", follower_count: 9999 }),
+    row({ unique_id: "mx-low", guessed_location: "mx", follower_count: 100 }),
+  ]);
+  const j = await store.claimNextJob("c1", 5 * 60 * 1000, ["MX", "BR"], true);
+  check("领到 mx-low", j?.uniqueId === "mx-low", `got=${j?.uniqueId}`);
+}
+
+console.log("== B2. 首国无候选时空位置可领 ==");
+{
+  const { store, db } = freshStore("b2");
+  insertRows(db, [row({ unique_id: "only-empty", guessed_location: "" })]);
+  const j = await store.claimNextJob("c1", 5 * 60 * 1000, ["MX", "BR"], true);
+  check("领到 only-empty", j?.uniqueId === "only-empty", `got=${j?.uniqueId}`);
+}
+
+console.log("== B3. hasVideo 层整体优先于 noVideoReq 层（含空位置）==");
+{
+  const { store, db } = freshStore("b3");
+  insertRows(db, [
+    row({ unique_id: "mx-novideo", guessed_location: "mx", video_count: 0, follower_count: 5000 }),
+    row({ unique_id: "br-video", guessed_location: "br", video_count: 3, follower_count: 100 }),
+  ]);
+  const j = await store.claimNextJob("c1", 5 * 60 * 1000, ["MX", "BR"], true);
+  check("mx-noVideo 优先于 br-video（与原版层级一致）", j?.uniqueId === "mx-novideo", `got=${j?.uniqueId}`);
+}
+
+console.log("== B4. 国家顺序优先：MX 行优先于 BR 行 ==");
+{
+  const { store, db } = freshStore("b4");
+  insertRows(db, [
+    row({ unique_id: "br-high", guessed_location: "br", follower_count: 9999 }),
+    row({ unique_id: "mx-low", guessed_location: "mx", follower_count: 10 }),
+  ]);
+  const j = await store.claimNextJob("c1", 5 * 60 * 1000, ["MX", "BR"], true);
+  check("领到 mx-low", j?.uniqueId === "mx-low", `got=${j?.uniqueId}`);
+}
+
+console.log("== B5. pinned：空池记忆后 pin 立即可领（缓存失效生效）==");
+{
+  const { store, db } = freshStore("b5");
+  insertRows(db, [
+    row({ unique_id: "plain", guessed_location: "mx", follower_count: 100 }),
+    row({ unique_id: "pinned-later", guessed_location: "br", follower_count: 50 }),
+  ]);
+  // 第一次领取：pinned 分支执行且 miss → 写入空池记忆
+  const j1 = await store.claimNextJob("c1", 5 * 60 * 1000, ["MX", "BR"], true);
+  check("第一次领到 plain", j1?.uniqueId === "plain", `got=${j1?.uniqueId}`);
+  // pin 一个低优先级任务
+  const pinRet = store.togglePin("pinned-later");
+  check("togglePin 成功", pinRet?.saved === true);
+  // 立即换客户端领取：应直接领到 pinned 任务（不受空池记忆影响）
+  const j2 = await store.claimNextJob("c2", 5 * 60 * 1000, ["MX", "BR"], true);
+  check("pin 后立即领到 pinned-later", j2?.uniqueId === "pinned-later", `got=${j2?.uniqueId}`);
+}
+
+console.log("== B6. 未登录客户端不领商家 ==");
+{
+  const { store, db } = freshStore("b6");
+  insertRows(db, [row({ unique_id: "seller-only", guessed_location: "mx", tt_seller: 1 })]);
+  const jOut = await store.claimNextJob("c1", 5 * 60 * 1000, ["MX"], false);
+  check("未登录领不到", jOut === null, `got=${jOut?.uniqueId}`);
+  const jIn = await store.claimNextJob("c2", 5 * 60 * 1000, ["MX"], true);
+  check("登录可领", jIn?.uniqueId === "seller-only", `got=${jIn?.uniqueId}`);
+}
+
+console.log("== B7. status_code=209002 仅登录客户端可领 ==");
+{
+  const { store, db } = freshStore("b7");
+  insertRows(db, [row({ unique_id: "code-209002", guessed_location: "mx", status_code: 209002 })]);
+  const jOut = await store.claimNextJob("c1", 5 * 60 * 1000, ["MX"], false);
+  check("未登录领不到", jOut === null, `got=${jOut?.uniqueId}`);
+  const jIn = await store.claimNextJob("c2", 5 * 60 * 1000, ["MX"], true);
+  check("登录可领", jIn?.uniqueId === "code-209002", `got=${jIn?.uniqueId}`);
+}
+
+console.log("== B8. 超时 processing 回收优先 ==");
+{
+  const { store, db } = freshStore("b8");
+  const now = Date.now();
+  insertRows(db, [
+    row({ unique_id: "stale", guessed_location: "mx", status: "processing", claimed_at: now - 10 * 60 * 1000, claimed_by: "dead-client" }),
+    row({ unique_id: "fresh", guessed_location: "mx", follower_count: 99999 }),
+  ]);
+  const j = await store.claimNextJob("c1", 5 * 60 * 1000, ["MX"], true);
+  check("回收的 stale 优先于高粉丝 fresh", j?.uniqueId === "stale", `got=${j?.uniqueId}`);
+}
+
+console.log("== B9. 商家任意国家兜底（sellerAnyCountry）==");
+{
+  const { store, db } = freshStore("b9");
+  insertRows(db, [
+    row({ unique_id: "zz-seller", guessed_location: "zz", tt_seller: 1, verified: 0, follower_count: 10 }),
+    row({ unique_id: "mx-normal", guessed_location: "mx", follower_count: 9999 }),
+  ]);
+  const j = await store.claimNextJob("c1", 5 * 60 * 1000, ["MX"], true);
+  check("商家级高于 other 级：zz-seller 先领", j?.uniqueId === "zz-seller", `got=${j?.uniqueId}`);
+}
+
+console.log("== B10. seed 级优先于 follow/other ==");
+{
+  const { store, db } = freshStore("b10");
+  insertRows(db, [
+    row({ unique_id: "mx-seed", guessed_location: "mx", sources: '["explore"]', follower_count: 10 }),
+    row({ unique_id: "br-seed", guessed_location: "br", sources: '["seed"]', follower_count: 20 }),
+  ]);
+  const j = await store.claimNextJob("c1", 5 * 60 * 1000, ["MX", "BR"], true);
+  check("seed 级 br-seed 胜过 other 级 mx（国家级让位于来源级）", j?.uniqueId === "br-seed", `got=${j?.uniqueId}`);
+}
+
+// ───────── A. 性能（32 万行）─────────
+console.log("\n== A. 性能：320,600 行 pending，候选国家池全空，模拟生产全 miss 场景 ==");
+{
+  const { store, db } = freshStore("perf");
+  // 32 万行：位置全 ZZ（配置外），含 3000 个高粉丝商家（seller any-country 分支秒命中用）
+  db.exec(`
+    WITH RECURSIVE seq(i) AS (SELECT 1 UNION ALL SELECT i+1 FROM seq WHERE i < 320600)
+    INSERT INTO task_queue
+      (unique_id, status, guessed_location, video_count, follower_count, tt_seller, verified, status_code, pinned, sources, created_at)
+    SELECT
+      'bulk_' || i,
+      'pending',
+      'zz',
+      CASE WHEN i % 3 = 0 THEN 5 ELSE 0 END,
+      (i * 7919) % 1000000,
+      CASE WHEN i <= 3000 THEN 1 ELSE 0 END,
+      0,
+      NULL,
+      0,
+      '["explore"]',
+      '2026-01-01T00:00:00Z'
+    FROM seq
+  `);
+  // 保证 3 个最高粉丝都是商家 → seller any-country 命中不扫表
+  db.exec(`UPDATE task_queue SET follower_count = 100000000 + CAST(unique_id AS INTEGER) % 1000 WHERE tt_seller = 1 AND CAST(REPLACE(unique_id,'bulk_','') AS INTEGER) <= 3`);
+
+  const t1 = Date.now();
+  const j1 = await store.claimNextJob("perf-c1", 5 * 60 * 1000, ["MX", "BR"], true);
+  const t1ms = Date.now() - t1;
+  check("首领成功（命中商家兜底）", j1?.uniqueId?.startsWith("bulk_"), `got=${j1?.uniqueId} ${t1ms}ms`);
+
+  const t2 = Date.now();
+  const j2 = await store.claimNextJob("perf-c2", 5 * 60 * 1000, ["MX", "BR"], true);
+  const t2ms = Date.now() - t2;
+  check("次领成功", j2?.uniqueId?.startsWith("bulk_"), `got=${j2?.uniqueId} ${t2ms}ms`);
+  check("次领显著快于首领（pinned/seed 空池缓存生效）", t2ms < t1ms, `t1=${t1ms}ms t2=${t2ms}ms`);
+  check("次领 < 150ms", t2ms < 150, `${t2ms}ms`);
+
+  // EXPLAIN：确认单国等值分支走 fast 索引
+  const plan = db
+    .prepare(`EXPLAIN QUERY PLAN SELECT unique_id FROM task_queue WHERE status='pending' AND pinned=0 AND video_count > 0 AND UPPER(COALESCE(guessed_location, '')) IN ('MX') AND (status_code IS NULL OR status_code = 0 OR status_code = 209002) ORDER BY follower_count DESC, created_at ASC, unique_id ASC LIMIT 1`)
+    .all();
+  const detail = plan.map((r) => r.detail).join(" | ");
+  check("单国分支走 idx_tq_claim_fast", detail.includes("idx_tq_claim_fast"), detail);
+}
+
+console.log(`\n===== 结果: ${pass} passed, ${fail} failed =====`);
+fs.rmSync(TMP, { recursive: true, force: true });
+process.exit(fail > 0 ? 1 : 0);

package/scripts/test-precise-index.mjs

@@ -0,0 +1,114 @@
+// 验证 idx_jobs_precise_merchants 部分覆盖索引：EXPLAIN 命中 + A/B 耗时
+// 用完即删
+import fs from "fs";
+import { createStore } from "../src/watch/data-store.js";
+import { getDb } from "../src/watch/db-schema.js";
+
+const TMP = "/tmp/tiktok-precise-idx-test";
+fs.rmSync(TMP, { recursive: true, force: true });
+fs.mkdirSync(TMP, { recursive: true });
+
+const store = createStore(`${TMP}/t.db`, {});
+const db = getDb();
+
+const IDX_SQL =
+  "CREATE INDEX IF NOT EXISTS idx_jobs_precise_merchants ON jobs(location_created, product_sold_count DESC, unique_id) WHERE tt_seller = 1 AND verified = 0 AND top_video_seo_url IS NOT NULL AND top_video_seo_url != '' AND product_review_count = 0 AND product_sold_count > 10 AND recent_ec_videos IS NOT NULL AND recent_ec_videos != ''";
+
+// 与 data-store.js getPreciseMerchantsFromDb 逐词一致的查询（loc 参数化）
+const COUNT_SQL = (ph) =>
+  `SELECT COUNT(*) AS n FROM jobs WHERE tt_seller = 1 AND verified = 0 AND location_created IN (${ph}) AND top_video_seo_url IS NOT NULL AND top_video_seo_url != '' AND product_review_count = 0 AND product_sold_count > 10 AND recent_ec_videos IS NOT NULL AND recent_ec_videos != '' AND EXISTS (SELECT 1 FROM json_each(recent_ec_videos) WHERE CAST(json_extract(value, '$.playCount') AS INTEGER) > 1000)`;
+
+const LOCS = ["MX", "BR", "DE", "US", "TH"];
+const ph = LOCS.map(() => "?").join(", ");
+
+// ── 造数：30 万商家（大部分不满足便宜条件）+ 500 行满足 + 1 万非商家 ──
+console.log("造数 31 万行...");
+db.exec(`
+  WITH RECURSIVE seq(i) AS (SELECT 1 UNION ALL SELECT i+1 FROM seq WHERE i < 300000)
+  INSERT INTO jobs
+    (unique_id, status, tt_seller, verified, location_created,
+     top_video_seo_url, product_review_count, product_sold_count,
+     recent_ec_videos, follower_count)
+  SELECT
+    's_' || i, 'done', 1, 0,
+    CASE (i % 5) WHEN 0 THEN 'MX' WHEN 1 THEN 'BR' WHEN 2 THEN 'DE' WHEN 3 THEN 'US' ELSE 'TH' END,
+    NULL, 5, 3, NULL, 1000
+  FROM seq
+`);
+db.exec(`
+  WITH RECURSIVE seq(i) AS (SELECT 1 UNION ALL SELECT i+1 FROM seq WHERE i < 500)
+  INSERT INTO jobs
+    (unique_id, status, tt_seller, verified, location_created,
+     top_video_seo_url, product_review_count, product_sold_count,
+     recent_ec_videos, follower_count)
+  SELECT
+    'hit_' || i, 'done', 1, 0,
+    CASE (i % 5) WHEN 0 THEN 'MX' WHEN 1 THEN 'BR' WHEN 2 THEN 'DE' WHEN 3 THEN 'US' ELSE 'TH' END,
+    '/shop/xxx-' || i, 0, 20 + (i % 100),
+    '[{"playCount":' || (2000 + i) || '}]', 5000
+  FROM seq
+`);
+db.exec(`
+  INSERT INTO jobs
+    (unique_id, status, tt_seller, verified, location_created,
+     top_video_seo_url, product_review_count, product_sold_count, recent_ec_videos, follower_count)
+  SELECT 'ns_' || i, 'done', 0, 0, 'MX', '/x', 0, 20, '[{"playCount":3000}]', 100
+  FROM (WITH RECURSIVE seq(i) AS (SELECT 1 UNION ALL SELECT i+1 FROM seq WHERE i < 10000) SELECT i FROM seq)
+`);
+
+const args = [...LOCS];
+
+// ── A/B：无新索引 vs 有新索引 ──
+function bench(label) {
+  const t0 = Date.now();
+  const n = db.prepare(COUNT_SQL(ph)).get(...args).n;
+  const ms = Date.now() - t0;
+  console.log(`  ${label}: COUNT=${n} ${ms}ms`);
+  return { n, ms };
+}
+
+console.log("\n== A/B 耗时（30.6 万行）==");
+const before = bench("无新索引");
+db.exec(IDX_SQL);
+const after = bench("有新索引");
+
+console.log("\n== EXPLAIN：COUNT 与 LIST 查询计划 ==");
+const planCount = db.prepare(`EXPLAIN QUERY PLAN ${COUNT_SQL(ph)}`).all(...args);
+console.log("COUNT :", planCount.map((r) => r.detail).join(" | "));
+const LIST_SQL = `SELECT unique_id, product_sold_count FROM jobs WHERE tt_seller = 1 AND verified = 0 AND location_created IN (${ph}) AND top_video_seo_url IS NOT NULL AND top_video_seo_url != '' AND product_review_count = 0 AND product_sold_count > 10 AND recent_ec_videos IS NOT NULL AND recent_ec_videos != '' AND EXISTS (SELECT 1 FROM json_each(recent_ec_videos) WHERE CAST(json_extract(value, '$.playCount') AS INTEGER) > 1000) ORDER BY product_sold_count DESC, unique_id ASC LIMIT 100 OFFSET 0`;
+const planList = db.prepare(`EXPLAIN QUERY PLAN ${LIST_SQL}`).all(...args);
+console.log("LIST  :", planList.map((r) => r.detail).join(" | "));
+
+const COUNT_IDX = `SELECT COUNT(*) AS n FROM jobs INDEXED BY idx_jobs_precise_merchants WHERE tt_seller = 1 AND verified = 0 AND location_created IN (${ph}) AND top_video_seo_url IS NOT NULL AND top_video_seo_url != '' AND product_review_count = 0 AND product_sold_count > 10 AND recent_ec_videos IS NOT NULL AND recent_ec_videos != '' AND EXISTS (SELECT 1 FROM json_each(recent_ec_videos) WHERE CAST(json_extract(value, '$.playCount') AS INTEGER) > 1000)`;
+const planIdx = db.prepare(`EXPLAIN QUERY PLAN ${COUNT_IDX}`).all(...args);
+const nIdx = db.prepare(COUNT_IDX).get(...args).n;
+
+let pass = 0, fail = 0;
+const check = (name, cond, detail = "") => {
+  if (cond) { pass++; console.log(`  ✅ ${name}${detail ? "  " + detail : ""}`); }
+  else { fail++; console.log(`  ❌ ${name}${detail ? "  " + detail : ""}`); }
+};
+
+console.log("\n== 断言 ==");
+check("COUNT 结果一致（500）", after.n === 500 && before.n === 500, `n=${after.n}`);
+check("INDEXED BY 结果一致", nIdx === 500, `n=${nIdx}`);
+check("优化器自发命中 partial index（小库可自发选中）",
+  planCount.some((r) => r.detail.includes("idx_jobs_precise_merchants")),
+  planCount.map((r) => r.detail).join(" | ").slice(0, 120));
+check("INDEXED BY 强制命中 partial index", planIdx.some((r) => r.detail.includes("idx_jobs_precise_merchants")));
+
+// 回退：drop 索引后 INDEXED BY 应报 no such index（业务层 catch 后走普通查询）
+db.exec("DROP INDEX idx_jobs_precise_merchants");
+let fellBack = false;
+try {
+  db.prepare(COUNT_IDX).get(...args);
+} catch (e) {
+  fellBack = /no such index/i.test(e.message);
+}
+check("索引缺失时抛 no such index（业务层回退依赖此错误）", fellBack);
+const nPlainAfterDrop = db.prepare(COUNT_SQL(ph)).get(...args).n;
+check("普通查询不受影响", nPlainAfterDrop === 500, `n=${nPlainAfterDrop}`);
+
+console.log(`\n===== 结果: ${pass} passed, ${fail} failed =====`);
+fs.rmSync(TMP, { recursive: true, force: true });
+process.exit(fail > 0 ? 1 : 0);

package/scripts/test-stats-smoke.mjs

@@ -0,0 +1,43 @@
+// /api/stats 冒烟：fast 组直接含 targetUsers/preciseMerchants（烟花 30s 级）
+// 用完即删
+import fs from "fs";
+import { createStore } from "../src/watch/data-store.js";
+import { getDb } from "../src/watch/db-schema.js";
+import { startWatchServer } from "../src/watch/server.js";
+
+const TMP = "/tmp/tk-stats-smoke";
+fs.rmSync(TMP, { recursive: true, force: true });
+fs.mkdirSync(TMP, { recursive: true });
+const DB = `${TMP}/t.db`;
+const PORT = 3199;
+
+const store = createStore(DB, {});
+const db = getDb();
+db.exec("ALTER TABLE jobs ADD COLUMN product_skus TEXT");
+db.exec(`INSERT INTO jobs (unique_id, status, tt_seller, verified, location_created,
+  top_video_seo_url, product_review_count, product_sold_count, recent_ec_videos, follower_count)
+  VALUES ('m1','done',1,0,'DE','/x',0,20,'[{"playCount":3000}]',100)`);
+db.exec(`INSERT INTO jobs (unique_id, status, tt_seller, verified, location_created, follower_count)
+  VALUES ('plain','done',1,0,'BR',50)`);
+
+await startWatchServer(DB, PORT, store, { llmDiscover: false }, null);
+
+let ok = false;
+for (let i = 0; i < 20 && !ok; i++) {
+  await new Promise((r) => setTimeout(r, 500));
+  try {
+    const res = await fetch(`http://127.0.0.1:${PORT}/api/stats`);
+    const s = await res.json();
+    if (s && typeof s.targetUsers !== "undefined") {
+      ok = true;
+      console.log("targetUsers     =", s.targetUsers, s.targetUsers === 1 ? "✅" : "❌ 应为 1");
+      console.log("preciseMerchants=", s.preciseMerchants, s.preciseMerchants === 1 ? "✅" : "❌ 应为 1");
+      console.log("processedUsers  =", s.processedUsers, "(slow 组 jobsAgg 未就绪时可为 0，5 分钟内后台补全)");
+      const pass = s.targetUsers === 1 && s.preciseMerchants === 1;
+      console.log(pass ? "SMOKE PASS" : "SMOKE FAIL");
+      process.exit(pass ? 0 : 1);
+    }
+  } catch {}
+}
+console.log("SMOKE FAIL: /api/stats 未响应");
+process.exit(1);