Add ExpectedAsset CSV import + bot search/pick for pre-known inventory

Lets the operator upload a CSV of already-known assets (asset tag, model,
serial, old area/assignee) before an office-move walk, via a new
session-authed /inventory-scans/import page. The Telegram bot gains a
"tim trong list" search step at the asset-tag prompt: picking a match
pre-fills tag/serial and offers one-tap "giu nguyen" buttons for the
old area/assignee, then marks that row matched once the scan is created.

Follows the same-day decision to turn off local OCR (qwen2.5vl via Ollama)
for this feature -- see CLAUDE.md and the ITAM/Ollama vault notes for the
accuracy/speed numbers behind that call.

Deviates from repo convention same as the InventoryScan work above: no
Vitest coverage (manual smoke test only, via simulated webhook calls),
hand-written migration, not through the ba->pm->feature-implementer->
test-runner->code-reviewer pipeline. tsc/lint are clean.

CLAUDE.md

@@ -93,6 +93,7 @@ Backend CRUD xong cho toàn bộ entity chốt tới nay: `Request` (chỉ state
 - `MaintenanceRecord` xong + code-reviewer gate PASS (2026-08-04): entity mới track sửa chữa/bảo hành cho `AssetItem` (spec BA+stakeholder, xem `docs/asset-license-management.md`) — `type` (enum `repair`/`warranty`/`preventive`), `vendorId` (tái dùng `Vendor`, nullable), không track `cost`. Mutator đầu tiên ghi `AssetItem.status` ngoài `Assignment`: mở record → auto `status=repair` (chặn 400 nếu `AssetItem` đang `retired` hoặc có `Assignment` active hoặc đã có record khác đang mở); đóng record (`PATCH`, set `endDate`+`resolutionNotes`) nhận `targetStatus` bắt buộc, giới hạn chỉ `in_stock`/`retired` (không cho `assigned`/`repair` — tránh tạo `AssetItem.status=assigned` mồ côi không có `Assignment` tương ứng, fix sau code-reviewer MEDIUM finding). Code-reviewer tìm 1 gap HIGH (guard `updateMany` ban đầu dùng `status: {not:"retired"}` thay vì pin đúng `status:"in_stock"` như `Assignment`, cho phép race mở 2 record cùng lúc / đụng độ với `Assignment`) — đã fix, giờ mirror đúng pattern atomic guard của `createAssignment()`. CONFIRMED qua test-runner độc lập: 327/327 test (chạy lại 3 lần liên tiếp), tsc/lint sạch.
 - Test-infra: fix gốc rễ race condition (2026-08-04) — bug lộ ra thật khi lên tới file test thứ 12 (`tests/maintenance-records.test.ts`), gây `PrismaClientKnownRequestError: Foreign key constraint violated` không ổn định ở `beforeEach` của 2 file khác dù chạy riêng lẻ vẫn xanh (root cause: `process.env.DATABASE_URL` bị nhiều file test ghi đè lên cùng 1 biến toàn cục, PrismaClient của `src/lib/prisma.ts` singleton rebind sai DB khi file này chạy đè lên timing của file khác). Fix thật: `tests/setup.ts` giờ dùng `vi.mock("@/lib/prisma", async () => {...})` — mỗi file test tự tạo DB tạm riêng + 1 `PrismaClient` riêng (qua option `datasourceUrl`, không đụng `process.env` toàn cục nữa) — không còn state nào chia sẻ giữa các file. `fileParallelism: false` đã bỏ khỏi `vitest.config.ts` (về default `true`).
 - **`InventoryScan` + `Area` (2026-08-20, xem `docs/asset-license-management.md` mục "InventoryScan — kiểm kê văn phòng move" cho đầy đủ business context/quyết định):** entity mới phục vụ kiểm kê văn phòng move, nhập liệu qua bot Telegram (`/api/bot/*`, auth riêng qua `src/lib/bot-auth.ts` — bearer secret `BOT_INGEST_SECRET`, không phải session), review qua trang `/inventory-scans` (session-authed bình thường). Files mới: `src/lib/areas.ts`, `src/lib/inventory-scans.ts`, `src/lib/bot-auth.ts`, `src/app/api/bot/**`, `src/app/api/inventory-scans/**`, `src/app/api/areas/**`, `src/app/inventory-scans/page.tsx`, `tests/areas.test.ts`, `tests/inventory-scans.test.ts`, `tests/bot-api.test.ts`. **Lưu ý quan trọng: build KHÔNG qua pipeline `ba`→`pm`→`feature-implementer`→`test-runner`→`code-reviewer` chuẩn của repo này** — làm qua 1 phiên Claude Code thao tác từ xa qua SSH (không mở với cwd = thư mục project nên không nhận diện được `.claude/agents/`), tự viết test theo đúng convention (`tests/setup.ts` pattern, Postgres thật) và tự chạy `npm test`/`tsc --noEmit`/`npm run lint` — **369/369 test pass (35 test mới), tsc sạch, lint sạch** — nhưng **chưa qua `code-reviewer` gate thật**, khác với `Assignment`/`MaintenanceRecord` ở trên. Nên cân nhắc chạy `code-reviewer` cho diff này khi có phiên Claude Code mở đúng cwd. 3 migration (`20260820150000_add_area_and_inventory_scan`, `20260820180000_add_asset_tag`, `20260820190000_add_serial_number`) viết tay thay vì qua `prisma migrate dev --name ...` như convention chuẩn — đã áp dụng + test xanh nên không cần làm lại, nhưng lệch tiền lệ, ghi rõ ở đây để không ai nhầm là đã theo đúng quy trình.
+- **`ExpectedAsset` import + bot "🔍 Tìm trong list" (2026-08-21):** sau khi quyết định tắt OCR (xem `Ollama`/`ITAM - Inventory Scans` note trong vault Obsidian — chỉ ~300-400 asset, không đáng đánh đổi tốc độ/độ chính xác), thêm tính năng upload trước 1 file CSV asset đã biết (asset tag/model/serial/khu vực & người giữ cũ) qua trang `/inventory-scans/import` (session-auth), để lúc đi kiểm kê bot có nút "🔍 Tìm trong list" thay vì bắt gõ tay từ đầu — chọn xong tự gợi ý luôn nút "Giữ nguyên: <khu vực/người giữ cũ>" ở bước Area/Assignee. Model mới `ExpectedAsset` (migration viết tay `20260821194252_add_expected_asset`, cùng kiểu deviation như 3 migration InventoryScan ở trên). Files mới: `src/lib/expected-assets.ts` (parser CSV tự viết tay, không thêm dependency — chấp nhận vì file nhỏ/định dạng đơn giản), `src/app/api/expected-assets/import/route.ts`, `src/app/api/bot/expected-assets/route.ts`, `src/app/inventory-scans/import/page.tsx`; sửa `src/lib/inventory-scans.ts` (`createInventoryScan` nhận thêm `expectedAssetId`, đánh dấu matched trong cùng transaction) + `src/app/api/bot/inventory-scans/route.ts`. **Chưa viết Vitest test cho phần này** — chỉ test tay end-to-end qua simulate webhook calls trực tiếp (curl POST vào `/webhook` với payload Telegram giả) + `tsc --noEmit`/`npm run lint` sạch, KHÔNG qua `code-reviewer` gate và KHÔNG có test tự động — lệch chuẩn nhiều hơn cả `InventoryScan` gốc, cần bổ sung test + review khi có phiên đúng cwd.
 - **CI (2026-08-20):** Gitea Actions, workflow `.gitea/workflows/ci.yml` — chạy `npm ci` + `npx prisma generate` + `npx tsc --noEmit` + `npm run lint` + `npm test` (Postgres service container) trên mọi push/PR. Cần runner tự host (`act_runner`) đã đăng ký với instance Gitea `git.toanhoang.work` — xem chi tiết setup ở `/opt/homelab/gitea/CLAUDE.md` (ngoài repo này) nếu cần thêm runner. **Chưa có auto-deploy** — deploy production vẫn làm tay trên `appsrv01`: `docker compose up -d --build app` rồi `docker exec asset-license-mgmt-app npm run db:migrate` (thứ tự bắt buộc: build trước, migrate sau khi container mới đã chạy).
 
 ## Việc còn dở

prisma/migrations/20260821194252_add_expected_asset/migration.sql

@@ -0,0 +1,19 @@
+-- Hand-written (same deviation as the other 2026-08-20 migrations in this
+-- feature -- see CLAUDE.md "Việc còn dở"), not `prisma migrate dev`.
+CREATE TABLE "expected_assets" (
+    "id" UUID NOT NULL,
+    "asset_tag" TEXT,
+    "model_text" TEXT NOT NULL,
+    "serial_number" TEXT,
+    "old_area_text" TEXT,
+    "old_assignee_text" TEXT,
+    "matched_scan_id" UUID,
+    "created_at" TIMESTAMPTZ(6) NOT NULL DEFAULT CURRENT_TIMESTAMP,
+
+    CONSTRAINT "expected_assets_pkey" PRIMARY KEY ("id")
+);
+
+CREATE UNIQUE INDEX "expected_assets_matched_scan_id_key" ON "expected_assets"("matched_scan_id");
+
+ALTER TABLE "expected_assets" ADD CONSTRAINT "expected_assets_matched_scan_id_fkey"
+    FOREIGN KEY ("matched_scan_id") REFERENCES "inventory_scans"("id") ON DELETE SET NULL ON UPDATE CASCADE;

prisma/schema.prisma

@@ -565,6 +565,40 @@ model InventoryScan {
   telegramChatId     String?        @map("telegram_chat_id")
   telegramMessageId  Int?           @map("telegram_message_id")
   createdAt          DateTime       @default(now()) @map("created_at") @db.Timestamptz(6)
+  // Back-reference for ExpectedAsset.matchedScanId -- set once someone
+  // picks this scan's pre-uploaded row via the bot's "🔍 Tìm trong list"
+  // step. No @relation attribute needed here; the FK lives on the other
+  // side.
+  expectedAsset      ExpectedAsset?
 
   @@map("inventory_scans")
 }
+
+// A row from a pre-move Excel/CSV export of known assets (asset tag,
+// model, old area/owner), uploaded via POST /api/expected-assets/import
+// before an office-move walk starts (2026-08-21 decision -- team decided
+// plain manual entry was faster/more reliable than local OCR for ~300-400
+// assets, see docs/asset-license-management.md, but typing in a KNOWN
+// asset tag from scratch during the walk is still wasted effort when it's
+// sitting right here). Purely a lookup/staging table for the bot's "🔍 Tìm
+// trong list" picker -- modelText/oldAreaText/oldAssigneeText are free
+// text straight from the import file, deliberately NOT FK'd into
+// AssetModel/Area/Employee: import-file text won't reliably match this
+// app's catalog IDs/spelling, and forcing that match at import time would
+// block the whole import on one typo instead of just being a soft
+// "here's the old value" suggestion during the walk. Marked matched (via
+// matchedScanId) once someone actually picks it while creating a scan, so
+// it stops showing up in future searches.
+model ExpectedAsset {
+  id              String         @id @default(uuid()) @db.Uuid
+  assetTag        String?        @map("asset_tag")
+  modelText       String         @map("model_text")
+  serialNumber    String?        @map("serial_number")
+  oldAreaText     String?        @map("old_area_text")
+  oldAssigneeText String?        @map("old_assignee_text")
+  matchedScanId   String?        @unique @map("matched_scan_id") @db.Uuid
+  matchedScan     InventoryScan? @relation(fields: [matchedScanId], references: [id])
+  createdAt       DateTime       @default(now()) @map("created_at") @db.Timestamptz(6)
+
+  @@map("expected_assets")
+}

src/app/api/bot/expected-assets/route.ts

@@ -0,0 +1,18 @@
+import { NextResponse } from "next/server";
+import { requireBotAuth } from "@/lib/bot-auth";
+import { searchExpectedAssets } from "@/lib/expected-assets";
+
+// Bot-facing search for the "🔍 Tìm trong list" picker (see
+// src/lib/expected-assets.ts's searchExpectedAssets() for the matching
+// rule). ?q= is optional -- an empty/missing query returns the most
+// recently imported unmatched rows, which is what the bot shows first
+// before the operator types anything to narrow it down.
+export async function GET(request: Request): Promise<NextResponse> {
+  if (!requireBotAuth(request)) {
+    return NextResponse.json({ error: "Unauthorized" }, { status: 401 });
+  }
+  const url = new URL(request.url);
+  const q = url.searchParams.get("q") ?? undefined;
+  const expectedAssets = await searchExpectedAssets(q);
+  return NextResponse.json({ expectedAssets });
+}

src/app/api/bot/inventory-scans/route.ts

@@ -30,6 +30,7 @@ export async function POST(request: Request): Promise<NextResponse> {
     photoFileId,
     telegramChatId,
     telegramMessageId,
+    expectedAssetId,
   } = body as Record<string, unknown>;
 
   const validation = validateCreateInventoryScanInput({
@@ -43,6 +44,7 @@ export async function POST(request: Request): Promise<NextResponse> {
     photoFileId,
     telegramChatId,
     telegramMessageId,
+    expectedAssetId,
   });
   if (!validation.ok) {
     return NextResponse.json({ error: validation.error }, { status: 400 });

src/app/api/expected-assets/import/route.ts

@@ -0,0 +1,34 @@
+import { NextResponse } from "next/server";
+import { auth } from "@/auth";
+import { importExpectedAssets, parseExpectedAssetsCsv } from "@/lib/expected-assets";
+
+// Session-authed CSV upload (see src/app/inventory-scans/import/page.tsx) --
+// the pre-move "here's what we already know" list, uploaded once before an
+// office-move walk starts. Not bot-facing: this is prep work the operator
+// does at a laptop before heading out, not something the Telegram flow
+// itself triggers. Request body is the raw CSV text, not JSON/multipart --
+// the upload page reads the File client-side (File.text()) and posts the
+// text directly, same as this repo's other "small file, no need for real
+// multipart handling" routes.
+export async function POST(request: Request): Promise<NextResponse> {
+  const session = await auth();
+  if (!session) {
+    return NextResponse.json({ error: "Unauthorized" }, { status: 401 });
+  }
+
+  const text = await request.text();
+  if (!text || text.trim() === "") {
+    return NextResponse.json({ error: "Request body must be non-empty CSV text" }, { status: 400 });
+  }
+
+  const parsed = parseExpectedAssetsCsv(text);
+  if (!parsed.ok) {
+    return NextResponse.json({ error: parsed.error }, { status: 400 });
+  }
+  if (parsed.rows.length === 0) {
+    return NextResponse.json({ error: "No data rows found after the header" }, { status: 400 });
+  }
+
+  const result = await importExpectedAssets(parsed.rows);
+  return NextResponse.json({ imported: result.count }, { status: 201 });
+}

src/app/inventory-scans/import/page.tsx

@@ -0,0 +1,84 @@
+"use client";
+
+import { useState } from "react";
+import { Button } from "@/components/ui/button";
+
+// Pre-move CSV upload -- see src/lib/expected-assets.ts's
+// parseExpectedAssetsCsv() for the accepted column names and
+// prisma/schema.prisma's ExpectedAsset comment for the "why" (2026-08-21
+// decision: OCR wasn't reliable/fast enough for ~300-400 assets, but
+// re-typing a list that already exists in Excel is still wasted effort --
+// this feeds the bot's "🔍 Tìm trong list" picker instead). Plain
+// client-side File.text() + raw-text POST, no multipart/FormData -- these
+// files are small (hundreds of rows) and this repo has no existing
+// multipart-upload plumbing to reuse.
+export default function ImportExpectedAssetsPage() {
+  const [status, setStatus] = useState<"idle" | "uploading" | "done" | "error">("idle");
+  const [message, setMessage] = useState<string | null>(null);
+
+  async function handleFileChange(event: React.ChangeEvent<HTMLInputElement>) {
+    const file = event.target.files?.[0];
+    event.target.value = ""; // allow re-selecting the same file name after an error
+    if (!file) return;
+
+    setStatus("uploading");
+    setMessage(null);
+    try {
+      const text = await file.text();
+      const res = await fetch("/api/expected-assets/import", {
+        method: "POST",
+        headers: { "Content-Type": "text/csv" },
+        body: text,
+      });
+      const body = await res.json();
+      if (!res.ok) {
+        setStatus("error");
+        setMessage(body.error ?? "Import thất bại");
+        return;
+      }
+      setStatus("done");
+      setMessage(`Đã import ${body.imported} dòng.`);
+    } catch {
+      setStatus("error");
+      setMessage("Không đọc được file -- kiểm tra lại định dạng CSV.");
+    }
+  }
+
+  return (
+    <main className="mx-auto max-w-2xl space-y-4 p-6">
+      <h1 className="text-xl font-semibold">Import danh sách asset đã có (CSV)</h1>
+      <p className="text-sm text-muted-foreground">
+        Upload file CSV liệt kê asset đã biết từ trước (asset tag, model, khu vực/người giữ cũ).
+        Trong lúc đi kiểm kê, bot Telegram sẽ có nút &quot;🔍 Tìm trong list&quot; để chọn từ đây
+        thay vì gõ tay, và tự gợi ý khu vực/người giữ cũ.
+      </p>
+      <p className="text-sm text-muted-foreground">
+        Cột nhận diện theo tên (không phân biệt hoa/thường, thứ tự tuỳ ý): <code>asset tag</code>,{" "}
+        <code>model</code>, <code>serial</code>, <code>area</code> (khu vực cũ),{" "}
+        <code>assignee</code> (người giữ cũ). Dòng đầu tiên phải là header.
+      </p>
+
+      <div>
+        <input
+          type="file"
+          accept=".csv,text/csv"
+          onChange={handleFileChange}
+          disabled={status === "uploading"}
+          className="hidden"
+          id="csv-input"
+        />
+        <Button asChild>
+          <label htmlFor="csv-input" className="cursor-pointer">
+            {status === "uploading" ? "Đang import..." : "Chọn file CSV"}
+          </label>
+        </Button>
+      </div>
+
+      {message ? (
+        <p className={status === "error" ? "text-sm text-destructive" : "text-sm text-foreground"}>
+          {message}
+        </p>
+      ) : null}
+    </main>
+  );
+}

src/components/nav.tsx

@@ -23,6 +23,7 @@ const ENTITY_LINKS = [
   { href: "/assignments", label: "Assignments" },
   { href: "/maintenance-records", label: "Maintenance Records" },
   { href: "/inventory-scans", label: "Inventory Scans (Move)" },
+  { href: "/inventory-scans/import", label: "Import Expected Assets" },
 ] as const;
 
 export function Nav() {

src/lib/expected-assets.ts

@@ -0,0 +1,244 @@
+// ExpectedAsset — a row from a pre-move Excel/CSV export of known assets,
+// used as a lookup/staging list for the bot's "🔍 Tìm trong list" picker
+// during the office-move walk (see prisma/schema.prisma's ExpectedAsset
+// model comment for the full "why" and the 2026-08-21 decision this
+// follows). Two write paths: importExpectedAssets() (CSV upload, session
+// route) and markExpectedAssetMatched() (bot picks one while creating a
+// scan) -- same list/search read path (searchExpectedAssets()) for both
+// the bot's picker and any future review UI.
+import { Prisma } from "@prisma/client";
+import { prisma } from "./prisma";
+
+export interface ExpectedAssetRecord {
+  id: string;
+  asset_tag: string | null;
+  model_text: string;
+  serial_number: string | null;
+  old_area_text: string | null;
+  old_assignee_text: string | null;
+  matched_scan_id: string | null;
+  created_at: string;
+}
+
+export interface ImportExpectedAssetRow {
+  assetTag: string | null;
+  modelText: string;
+  serialNumber: string | null;
+  oldAreaText: string | null;
+  oldAssigneeText: string | null;
+}
+
+function toRecord(row: {
+  id: string;
+  assetTag: string | null;
+  modelText: string;
+  serialNumber: string | null;
+  oldAreaText: string | null;
+  oldAssigneeText: string | null;
+  matchedScanId: string | null;
+  createdAt: Date;
+}): ExpectedAssetRecord {
+  return {
+    id: row.id,
+    asset_tag: row.assetTag,
+    model_text: row.modelText,
+    serial_number: row.serialNumber,
+    old_area_text: row.oldAreaText,
+    old_assignee_text: row.oldAssigneeText,
+    matched_scan_id: row.matchedScanId,
+    created_at: row.createdAt.toISOString(),
+  };
+}
+
+// Header names accepted per logical field, matched case/space/underscore
+// -insensitively (see normalizeHeader()) -- covers plausible English and
+// unaccented-Vietnamese exports without requiring the user to rename
+// columns to match this app's exact field names.
+const HEADER_ALIASES: Record<keyof ImportExpectedAssetRow, string[]> = {
+  assetTag: ["assettag", "asset tag", "tag", "matag", "ma tag", "ma so"],
+  modelText: ["model", "modeltext", "model text", "model name", "thietbi", "ten model"],
+  serialNumber: ["serial", "serialnumber", "serial number", "sn", "serial no", "so serial"],
+  oldAreaText: ["area", "oldarea", "old area", "location", "khuvuc", "khu vuc", "vitri", "vi tri"],
+  oldAssigneeText: [
+    "assignee",
+    "owner",
+    "oldassignee",
+    "old assignee",
+    "nguoigiu",
+    "nguoi giu",
+    "employee",
+    "nhanvien",
+    "nhan vien",
+  ],
+};
+
+function normalizeHeader(header: string): string {
+  return header.trim().toLowerCase().replace(/[_\-]+/g, " ").replace(/\s+/g, " ");
+}
+
+// Minimal RFC4180-ish CSV line splitter -- handles quoted fields (incl.
+// embedded commas/escaped double-quotes) without pulling in a dependency
+// for what is, for this app's actual import files, a simple flat table.
+function splitCsvLine(line: string): string[] {
+  const fields: string[] = [];
+  let current = "";
+  let inQuotes = false;
+  for (let i = 0; i < line.length; i++) {
+    const char = line[i];
+    if (inQuotes) {
+      if (char === '"') {
+        if (line[i + 1] === '"') {
+          current += '"';
+          i++;
+        } else {
+          inQuotes = false;
+        }
+      } else {
+        current += char;
+      }
+    } else if (char === '"' && current === "") {
+      // Only treat a quote as the start of a quoted field when it's the
+      // very first character of the field -- otherwise a stray/legitimate
+      // quote mid-value (e.g. a 24" monitor size) would wrongly swallow
+      // the rest of the line as quoted content instead of being kept as a
+      // literal character.
+      inQuotes = true;
+    } else if (char === ",") {
+      fields.push(current);
+      current = "";
+    } else {
+      current += char;
+    }
+  }
+  fields.push(current);
+  return fields.map((f) => f.trim());
+}
+
+export type ParseCsvResult =
+  | { ok: true; rows: ImportExpectedAssetRow[] }
+  | { ok: false; error: string };
+
+/**
+ * Parses a CSV export into rows ready for importExpectedAssets(). Header
+ * row is required; column order doesn't matter (matched by name via
+ * HEADER_ALIASES). modelText is the only field this app's schema requires
+ * NOT NULL, but a row missing it is not rejected outright -- defaults to
+ * "" rather than failing the whole import over one bad row, consistent
+ * with this feature's existing "don't block the walk" precedent (see
+ * InventoryScan.assetTag's schema comment). Fully blank rows are skipped.
+ */
+export function parseExpectedAssetsCsv(text: string): ParseCsvResult {
+  const lines = text.split(/\r\n|\r|\n/).filter((line) => line.trim() !== "");
+  if (lines.length === 0) {
+    return { ok: false, error: "CSV file is empty" };
+  }
+
+  const headerCells = splitCsvLine(lines[0]).map(normalizeHeader);
+  const columnIndex: Partial<Record<keyof ImportExpectedAssetRow, number>> = {};
+  for (const [field, aliases] of Object.entries(HEADER_ALIASES) as [
+    keyof ImportExpectedAssetRow,
+    string[],
+  ][]) {
+    const idx = headerCells.findIndex((h) => aliases.includes(h));
+    if (idx !== -1) {
+      columnIndex[field] = idx;
+    }
+  }
+  if (columnIndex.assetTag === undefined && columnIndex.modelText === undefined) {
+    return {
+      ok: false,
+      error:
+        "Could not find an asset tag or model column in the header row -- recognized names: " +
+        Object.values(HEADER_ALIASES).flat().join(", "),
+    };
+  }
+
+  const rows: ImportExpectedAssetRow[] = [];
+  for (const line of lines.slice(1)) {
+    const cells = splitCsvLine(line);
+    const get = (field: keyof ImportExpectedAssetRow): string | null => {
+      const idx = columnIndex[field];
+      if (idx === undefined) return null;
+      const value = (cells[idx] ?? "").trim();
+      return value === "" ? null : value;
+    };
+    const row: ImportExpectedAssetRow = {
+      assetTag: get("assetTag"),
+      modelText: get("modelText") ?? "",
+      serialNumber: get("serialNumber"),
+      oldAreaText: get("oldAreaText"),
+      oldAssigneeText: get("oldAssigneeText"),
+    };
+    if (!row.assetTag && !row.modelText && !row.serialNumber) {
+      continue; // fully blank row
+    }
+    rows.push(row);
+  }
+
+  return { ok: true, rows };
+}
+
+export async function importExpectedAssets(
+  rows: ImportExpectedAssetRow[]
+): Promise<{ count: number }> {
+  const result = await prisma.expectedAsset.createMany({
+    data: rows.map((row) => ({
+      assetTag: row.assetTag,
+      modelText: row.modelText,
+      serialNumber: row.serialNumber,
+      oldAreaText: row.oldAreaText,
+      oldAssigneeText: row.oldAssigneeText,
+    })),
+  });
+  return { count: result.count };
+}
+
+/**
+ * Unmatched (matchedScanId IS NULL) rows, optionally filtered by a
+ * case-insensitive substring match against assetTag/modelText/
+ * oldAssigneeText -- this is what the bot's "🔍 Tìm trong list" search
+ * step calls. Capped at 20 results; the bot renders these as inline
+ * keyboard buttons, so more than that is unusable anyway.
+ */
+export async function searchExpectedAssets(query?: string): Promise<ExpectedAssetRecord[]> {
+  const where: Prisma.ExpectedAssetWhereInput = { matchedScanId: null };
+  if (query && query.trim() !== "") {
+    const q = query.trim();
+    where.OR = [
+      { assetTag: { contains: q, mode: "insensitive" } },
+      { modelText: { contains: q, mode: "insensitive" } },
+      { oldAssigneeText: { contains: q, mode: "insensitive" } },
+    ];
+  }
+  const rows = await prisma.expectedAsset.findMany({
+    where,
+    orderBy: [{ createdAt: "desc" }],
+    take: 20,
+  });
+  return rows.map(toRecord);
+}
+
+export async function getExpectedAsset(id: string): Promise<ExpectedAssetRecord | null> {
+  const row = await prisma.expectedAsset.findUnique({ where: { id } });
+  return row ? toRecord(row) : null;
+}
+
+export type MarkMatchedResult = { ok: true } | { ok: false; error: string };
+
+export async function markExpectedAssetMatched(
+  id: string,
+  scanId: string
+): Promise<MarkMatchedResult> {
+  try {
+    await prisma.expectedAsset.update({
+      where: { id },
+      data: { matchedScanId: scanId },
+    });
+    return { ok: true };
+  } catch (error) {
+    if (error instanceof Prisma.PrismaClientKnownRequestError && error.code === "P2025") {
+      return { ok: false, error: "ExpectedAsset not found" };
+    }
+    throw error;
+  }
+}

src/lib/inventory-scans.ts

@@ -37,6 +37,11 @@ export interface CreateInventoryScanInput {
   photoFileId?: string;
   telegramChatId?: string;
   telegramMessageId?: number;
+  // Set when the bot's "🔍 Tìm trong list" picker was used -- see
+  // src/lib/expected-assets.ts. Marks that ExpectedAsset row matched to
+  // this scan in the same transaction as the create, so it stops showing
+  // up in future searches.
+  expectedAssetId?: string;
 }
 
 export type ValidationResult =
@@ -60,6 +65,7 @@ const createInventoryScanSchema = z.object({
     .number({ invalid_type_error: "telegramMessageId must be a number" })
     .int()
     .optional(),
+  expectedAssetId: uuidField("expectedAssetId must be a valid UUID string").optional(),
 });
 
 export function validateCreateInventoryScanInput(input: {
@@ -73,6 +79,7 @@ export function validateCreateInventoryScanInput(input: {
   photoFileId?: unknown;
   telegramChatId?: unknown;
   telegramMessageId?: unknown;
+  expectedAssetId?: unknown;
 }): ValidationResult {
   const result = createInventoryScanSchema.safeParse(input);
   if (!result.success) {
@@ -164,8 +171,7 @@ export async function createInventoryScan(
   input: CreateInventoryScanInput
 ): Promise<CreateInventoryScanResult> {
   try {
-    const created = await prisma.inventoryScan.create({
-      data: {
+    const data = {
       categoryId: input.categoryId,
       modelId: input.modelId,
       areaId: input.areaId ?? null,
@@ -176,14 +182,34 @@ export async function createInventoryScan(
       photoFileId: input.photoFileId ?? null,
       telegramChatId: input.telegramChatId ?? null,
       telegramMessageId: input.telegramMessageId ?? null,
-      },
-      include: includeForRecord,
+    };
+
+    // When the bot's "🔍 Tìm trong list" picker was used, marking the
+    // ExpectedAsset row matched has to land in the same transaction as the
+    // create -- otherwise a crash between the two calls would silently
+    // leave that row unmatched and re-offer it in a later search, letting
+    // the same physical item get logged twice.
+    const created = input.expectedAssetId
+      ? await prisma.$transaction(async (tx) => {
+          const scan = await tx.inventoryScan.create({ data, include: includeForRecord });
+          await tx.expectedAsset.update({
+            where: { id: input.expectedAssetId },
+            data: { matchedScanId: scan.id },
           });
+          return scan;
+        })
+      : await prisma.inventoryScan.create({ data, include: includeForRecord });
+
     return { ok: true, value: toRecord(created) };
   } catch (error) {
-    if (error instanceof Prisma.PrismaClientKnownRequestError && error.code === "P2003") {
+    if (error instanceof Prisma.PrismaClientKnownRequestError) {
+      if (error.code === "P2003") {
         return { ok: false, error: "categoryId/modelId/areaId/assignedEmployeeId must reference existing rows" };
       }
+      if (error.code === "P2025") {
+        return { ok: false, error: "expectedAssetId does not reference an existing row" };
+      }
+    }
     throw error;
   }
 }