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.
self-hosted runner is pinned to homelab_proxy network (needed to resolve the gitea host for checkout), so job+service containers are separate network members, not sharing a netns like GitHub-hosted runners service containers do -- localhost inside the job container never reached the db service. Also drops the now-unneeded host port publish (avoids repeating the port-already-allocated failure an earlier orphaned run left behind).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
npm ci -> prisma generate -> tsc --noEmit -> lint -> test (Postgres service) -> build, on every push/PR. Requires a self-hosted act_runner registered to git.toanhoang.work (set up 2026-08-20, see /opt/homelab/gitea-runner/ on homelab01 -- outside this repo). No auto-deploy yet, still manual on appsrv01.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
369/369 tests pass (35 new), tsc --noEmit clean, next lint clean. Not yet run through the repos code-reviewer subagent gate (this session operated via SSH, cwd was not the project dir so .claude/agents/ subagents were not recognized) -- see CLAUDE.md note.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
List + inline edit (model within category, area, assignee, specs, asset tag, serial number) + delete. No create form -- creation only happens via the bot.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
GET /api/areas, GET /api/inventory-scans, PATCH+DELETE /api/inventory-scans/[id], GET /api/inventory-scans/[id]/photo (proxies Telegram file_id on demand, derives Content-Type from file_path extension since Telegram serves application/octet-stream regardless of actual type).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Auth via BOT_INGEST_SECRET bearer header (src/lib/bot-auth.ts), not Auth.js session -- caller is the inventory-capture Telegram bot service, not a browser.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Office-move inventory kiem ke feature. See docs/asset-license-management.md InventoryScan section for full business rationale (Area vs Seat, why not AssetItem, asset tag/serial nullable).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Per-row employee <Select> + Transfer button alongside the existing Return
button, shown for active (returned_at null) rows. PATCHes
/api/assignments/[id] with {employeeId}, which the route already dispatches
to transferAssignment() vs returnAssignment() based on body shape.
Mirrors assignments/page.tsx's per-row close mini-form pattern; targetStatus
select reuses the closed AssetItemStatus set client-side, same precedent as
licenses page's licenseType select.
Avoids leaking the plaintext password via shell history/`ps`/process
listings. --email/--name stay as CLI flags (not sensitive); password is now
read via a masked interactive prompt using only Node built-ins (readline +
raw-mode stdin), no new dependency.
Was implicitly SET NULL (Prisma default for nullable FK columns),
inconsistent with employeeId's RESTRICT. Once a delete endpoint for
AssetItem/License exists, SET NULL would silently null out historical
Assignment rows on delete, erasing what was assigned and re-violating
the assetItemId/licenseId XOR retroactively. No delete endpoint exists
yet, so no real data was ever affected by the old behavior.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Closes the coverage gap noted in CLAUDE.md's "Việc còn dở" — mirrors the
existing pattern in tests/licenses.test.ts and tests/assignments.test.ts for
POST /api/assets, POST /api/employees, POST /api/requests (the only body-
accepting methods on these three routes; none have a PATCH route yet).
Root-cause fix for cross-file DB pollution in the vitest suite, which
surfaced as intermittent FK-violation failures once the suite hit 12
test files. tests/setup.ts previously mutated the global
process.env.DATABASE_URL and relied on src/lib/prisma.ts's singleton
picking it up, which is racy across files sharing a worker process
(process.env isn't reset between files even though Vitest resets each
file's module registry). fileParallelism: false only masked this.
Now each test file's own tests/setup.ts creates its own throwaway
Postgres database and vi.mock()s "@/lib/prisma" with a dedicated
PrismaClient constructed via the explicit datasourceUrl option, so no
mutable state is shared between files and fileParallelism can go back
to its default (true).
Verified with 4 consecutive full `npm test` runs (327/327 passing each
time) and no leftover throwaway databases after each run.
createMaintenanceRecord()'s AssetItem status guard pinned status:{not:"retired"}
instead of status:"in_stock", so two concurrent opens (or an open racing a
concurrent createAssignment()) could both pass the guard, producing two open
MaintenanceRecords or an orphaned active Assignment. Guard now pins "in_stock"
exactly, mirroring createAssignment()'s own atomic guard.
closeMaintenanceRecord()'s targetStatus accepted any of the 4 AssetItemStatus
values, including "assigned" — which sets AssetItem.status without creating a
corresponding Assignment row, desyncing Assignment (source of truth for who
holds what) from AssetItem.status with no way back. Restricted to
in_stock|retired, the only two outcomes a closed repair can actually resolve
to.
createMaintenanceRecord() opens a record and flips AssetItem.status to
"repair" inside one transaction, rejecting on missing/retired AssetItem,
an active Assignment, or an already-open record for the same item.
closeMaintenanceRecord() sets endDate/resolutionNotes and applies a
caller-supplied targetStatus (never defaulted) to the AssetItem, guarding
double-close and endDate>=startDate. Mirrors src/lib/assignments.ts's
Zod-validate + $transaction + WHERE-guarded updateMany race-closing pattern.
New entity tracking repair/warranty/preventive maintenance jobs for an
AssetItem. Reuses Vendor (nullable) for the repair location, no cost
field (mirrors PurchaseOrder). "At most 1 open record per AssetItem"
invariant left to the application layer, mirroring Assignment.
Entity mới track sửa chữa/bảo hành cho AssetItem, hiện thực hoá case
"in repair" đã note từ trước. Field, type enum, đồng bộ AssetItem.status,
và quyết định không track cost đều đã chốt với stakeholder — sẵn sàng
giao pm chia task.
Condensed the "Trạng thái hiện tại"/"Việc còn dở" sections — most of the
detailed per-entity narrative was either resolved, superseded by a later
change, or derivable from git log, so it's trimmed to what isn't:
decisions, gotchas, and still-open risks. Added the UI (worktree-ui-basic
merge), the Assignment transfer fix, and the two dormant findings
(XOR defense-in-depth, Assignment FK onDelete) from the code-reviewer gate.
Vitest's default glob isn't excluded for .claude/, so a sibling worktree
under .claude/worktrees/ (e.g. worktree-ui-basic) got silently collected
and run against this repo's current src/ via the "@" alias, doubling the
test count and running a stale test file against current code under a
shared identity. Found by an independent test-runner pass while verifying
the Assignment transfer fix.
The only way to move an Assignment to a new employee was two separate calls
(PATCH return + POST create), which is not atomic and left a race window
where a third party could grab the now-in_stock AssetItem in between, plus
no rollback if the second call failed. Adds transferAssignment() which
closes the old Assignment and creates the new one inside a single
prisma.$transaction(), per docs/asset-license-management.md's "Assignment
(chốt 2026-08-02)" spec. Does not touch AssetItem.status or
License.seatsUsed — the resource never becomes free mid-transfer. Wired
into PATCH /api/assignments/[id]: an employeeId key in the body dispatches
to transfer, a returnedAt key to return, both together is rejected as
mutually exclusive.
48 tests covering the XOR validation, the AssetItem.status/active-assignment
preconditions, seat overbooking, seat/status sync across assign+return,
FK-not-found for all three FKs, double-return, 404s, auth-gating, malformed
bodies, and two explicit concurrency tests (Promise.all against a
1-remaining-seat License and against a single in_stock AssetItem) exercising
the transaction race-closing logic in src/lib/assignments.ts.
POST/GET /api/assignments mirror the existing entity route pattern
(auth-gated, thin handler delegating to src/lib/assignments.ts). PATCH
/api/assignments/[id] is the first dynamic-segment route in the codebase,
per the stakeholder-confirmed convention (docs/asset-license-management.md)
of update routes being PATCH /api/<entity>/[id] rather than an
action-based sub-route.
createAssignment() enforces the XOR (assetItemId/licenseId) at the Zod
layer and the AssetItem.status/License.seatsUsed business rules inside a
single prisma.$transaction(), using atomic WHERE-guarded updateMany() calls
(not plain read-then-write) to close the seat-overbooking and
double-assignment race conditions flagged for License by code-reviewer.
returnAssignment() reverses those side-effects the same way and closes the
double-return race with the same pattern.
Links Employee to exactly one of AssetItem/License (XOR enforced at the
Zod layer, not built yet) via nullable FKs, plus assignedAt/returnedAt/
signedAt timestamps. See docs/asset-license-management.md's "Assignment"
section for the full spec.
Retrofit AssetItem.model (string) -> modelId (FK to AssetModel) —
closes out Lane B (AssetCategory -> AssetCategoryAttributeDef +
AssetModel -> dynamic specs validation -> AssetItem.modelId).
Verified green by test-runner (225/225 tests, tsc/lint/build clean)
before merge.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
AssetItem.model was a free-text string with no referential integrity.
Now a required FK to AssetModel, mirroring the Request.employeeId and
License.vendorId retrofit precedent (no backfill needed, no real
AssetItem data existed).
Wire dynamic specs validation into AssetModel creation: builds a
runtime Zod schema from the category's AssetCategoryAttributeDef
rows, rejects unknown keys, enforces required keys. Verified green
by test-runner (223/223 tests, tsc/lint/build clean) before merge.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
AssetModel entity: Prisma model+migration, Zod+CRUD (categoryId FK
+ business rule requiring valueGroup=high_value, specs JSONB loosely
validated for now), API routes, tests. Verified green by test-runner
(217/217 tests, tsc/lint/build clean) before merge.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
toanhoang