Files
shopdb-flask/docs/proposals/printedparts-plugin.md
cproudlock f9d6b57f0f Move the documentation to the line that ships, and correct it against the code
The docs lived on one lineage and the fleet code on another. That is why six
pages described a shared fleet key as how an agent authenticates, why
COLLECTOR-AUTH had to warn that the per-PC credential code was not in the
repository it sat in, and why ADR-019 recorded a decision whose implementation
was somewhere else. Documentation that cannot see the code it documents goes
stale by construction, and no amount of editing fixes that.

So the corpus moves here: 84 files, 56 pages, 20 ADRs, the proposals, INDEX.md,
llms.txt and both gates. Filenames are unchanged, because a wiki page name is
derived from the filename and a rename breaks every link anybody saved.

THE POINT OF THE MOVE IS THE CORRECTION IT ENABLES. Every caveat that said "not
in this tree" was false the moment the files landed beside the code, so each was
rewritten from what is actually here:

- COLLECTOR-AUTH gained the real enroll request body, the two bounds on the
  unauthenticated route (per-IP rate limit, pending cap with stale rows dropped
  before a real PC is refused), and the rejected pending reason that binds to
  the endpoint identity rather than the nonce.
- ADR-019 now describes an implementation that is present, including its tests,
  and says plainly that what remains is the FLEET migration, not the code.
- proposals/device-identity claimed phase 1 was unbuilt. plugins/computers
  declares deviceid on the collector schema and prefers it over the older
  fallbacks. Corrected in the page, its summary and the index.

WINDOWS/IIS IS NOW DOCUMENTED AS THE LEGACY PATH, per the product decision, and
carefully: it is what production runs today. Every Windows page opens by saying
it is supported, not for a new site, and where a new site should go instead. No
procedure was cut - an operator on an IIS box at 2am needs these pages to work
exactly as well tomorrow. MIGRATE-TOPOLOGY stops being a footnote and becomes
the off-ramp, which is why it nearly doubled: what was there did not match how
the Docker stack really works here.

The Docker material that already lived on this line was kept and given
ownership rather than buried under the ported Windows-centric prose: DEPLOY now
describes the four services, the migrate service that runs the upgrade chain
before api starts, TLS terminated by nginx, and the plugin count corrected from
thirteen to seventeen - a number both lineages had wrong.

ROADMAP is rewritten from evidence rather than memory: what has shipped, read
out of the changelog, the ADRs and the manifests, and ahead of it exactly two
items - finalize a distributed release and polish the setup wizard.

PROJECT-MAP is generated and was stale; the GENERATOR was what was wrong.
--check passes now.

API-REFERENCE is retired to a pointer. /api/docs serves 573 operations with
derived auth, query parameters and harvested examples, and a hand-written page
competing with that is a liability. The file stays so its wiki page and its
inbound links survive.

AND THE MOVE TURNED ON A SCRUB THAT HAD NEVER RUN HERE. test_docs_publishable
keys on docs/*.md, which this line did not have, so it skipped - its own comment
predicted what would surface. Porting the corpus activated it and it found 48
lines that must not reach a published wiki: branch names abbreviating the site,
assistant-tooling paths, the site name in two deployment pages, and an employee
ID in a Windows path that the scrub does not even look for.
2026-09-19 09:10:15 -04:00

11 KiB

Proposal: printedparts plugin (3D-printed parts storefront + kiosk)

Status: BUILT. The plugin ships as plugins/printedparts/, kiosk route included. This page is the design behind it, kept as the record of why, and it is also the reference design for the plugin-development lab in docs/PLUGIN-LAB-PRINTEDPARTS.md. It is NOT current guidance: where it and the shipped plugin disagree, the code wins.

1. Problem

The 3D-printer engineers stock bins of printed parts (fixtures, clips, covers, spacers). Anyone on the floor can take parts, so stock silently runs out and nobody knows who took what or how fast items burn down. They need:

  • a catalog ("storefront") of printable items: photo, description, quantity on hand;
  • a barcode label per item (1in x 0.5in) stuck on each bin;
  • a touch-screen kiosk: scan the bin barcode, scan your badge, enter how many you took, submit;
  • restock and correction flows for the engineers;
  • stock monitoring plus consumption metrics.

2. Shape: standalone model plugin, NOT an asset type

These are quantity-based consumables: one row represents a kind of part with a count, not an individually tracked machine. ADR-001 assets are one-row-per- physical-thing (a PC, a printer). So printedparts follows the knowledgebase/usb shape - own tables, own blueprint, no AssetType row - and does NOT join the asset-label TYPE_CONFIG; it ships its own print view the way USB labels do.

Item identity: itemcode, generated 3DP-<zero-padded id> (prefix configurable via setting printedparts_code_prefix). Short, CODE128-friendly, human-readable. This is what the bin label encodes.

3. Data model (2 tables, LOCKED naming, per-plugin Alembic)

printeditems

column type notes
printeditemid int PK autoincrement
itemcode varchar(20) unique, indexed generated on create
itemname varchar(120) NOT NULL
itemdescription varchar(500) brief description
imageurl varchar(255) served upload, models.py pattern
quantityonhand int NOT NULL default 0 cached; ledger is truth
lowstockthreshold int NOT NULL default 5 per-item, seeds from setting
binlocation varchar(100) where the bin lives
printnotes mediumtext material, print time, slicer file path
isactive tinyint(1) soft retire
createddate / modifieddate datetime AuditMixin/BaseModel

printeditemtransactions (the ledger - source of truth)

column type notes
transactionid int PK
printeditemid int FK -> printeditems CASCADE, indexed
transactiontype varchar(10) NOT NULL take / restock / adjust
quantitychange int NOT NULL negative for take, signed for adjust
employeesso varchar(20) NOT NULL, indexed who (badge-resolved)
employeename varchar(120) resolved at write time (USB pattern)
reason varchar(255) required for adjust
transactiondate datetime NOT NULL default naive-UTC, indexed

Invariants: quantityonhand = sum of quantitychange (enforced by writing both in one session/commit; an adjust can never drive it below 0 - reject). Every write records WHO via badge scan; there is no anonymous mutation.

Both tables registered in PLUGIN_TABLE_OWNERS (shopdb/plugins/alembic_template.py); migration 0001 is a REAL baseline (measuringtools pattern - hand-written op.create_table, no cross-schema FK so create_plugin_tables would also work, but write the ops explicitly for the exercise).

4. Badge resolution (reuse the USB contract exactly)

Same input shapes as plugins/usb/api/routes.py:

  • all digits -> SSO;
  • 0<digits>BZ (case-insensitive) -> physical badge wrapping a PayNo;
  • resolution to a display name via the employees plugin directory (DirectoryEmployee, selfhosted mode) with graceful "" fallback.

Extract-or-copy decision for the lab: copy the small _PAYNO_BADGE regex + lookup into the plugin (contract-pure, no cross-plugin import of usb). Resolution happens SERVER-side on the kiosk endpoint - the kiosk client never supplies a name, only the raw badge string.

Manifest dependencies: ["employees"] (name lookup). Badge that resolves to no employee: configurable policy setting printedparts_unknown_badge (allow = record SSO with empty name, deny = 422). Default deny.

5. API surface (blueprint at /api/printedparts)

Authenticated management (JWT + permission):

route method permission
/items GET list (search, paginate, lowstock filter) open read (jwt optional)
/items/<id> GET detail + recent transactions open read
/items POST create (mints itemcode) printedparts.create
/items/<id> PUT update printedparts.edit
/items/<id> DELETE soft-retire printedparts.delete
/items/<id>/image POST/DELETE upload/remove printedparts.edit
/image/<filename> GET serve public (models.py pattern)
/items/<id>/restock POST {quantity, badge} printedparts.restock
/items/<id>/adjust POST {quantitychange, reason, badge} printedparts.restock
/items/<id>/transactions GET history, ?format=csv open read

Kiosk (unauthenticated, notifications/employees open-endpoint precedent):

route method body
/kiosk/item/<itemcode> GET item summary by scanned code
/kiosk/take POST {itemcode, badge, quantity}

/kiosk/take validation: item exists + active; quantity 1..quantityonhand (clamp/reject configurable? no - reject with clear message, kiosk shows it); badge resolves per policy. Writes ledger row (negative) + decrements cached quantity in one commit. Rate of abuse is low (plant floor), but the endpoint only ever DECREMENTS stock with a recorded badge - it cannot edit the catalog.

Permissions declared via get_permissions(): printedparts.view/create/edit/ delete/restock (category printedparts), seeded on install/enable.

6. Frontend

Management pages (scaffold output, standard layout, master templates PrintersList/PrinterDetail):

  • PrintedItemsList.vue - table: image thumb, code, name, qty (red badge when <= threshold), bin; filters: search, low-stock-only; row click -> detail.
  • PrintedItemDetail.vue - hero image + fields, transaction history table, restock/adjust buttons (modal w/ quantity + badge + reason).
  • PrintedItemForm.vue - create/edit incl. image upload, threshold, bin.
  • Router file router/routes/printedparts.js, list/detail plugin-gated only, new/edit + requiresAuth (ADR-009, usb.js precedent).
  • Nav via get_navigation_items() -> "3D Parts".

Kiosk (net-new, top-level route /parts-kiosk, NO requiresAuth, outside AppLayout - shopfloor precedent):

  • Full-screen, 3-step flow: (1) SCAN ITEM - a focused invisible input catches the keyboard-wedge scan of the bin barcode, shows item card w/ photo + qty; (2) SCAN BADGE - same wedge input pattern for the badge; (3) QUANTITY - big touch keypad (0-9, clear, backspace - net-new component TouchKeypad.vue) + TAKE button. Success screen w/ remaining count, auto reset after a few seconds. All state client-side; one POST at the end.
  • Scanner UX rule: keyboard-wedge scanners type the code + Enter. A hidden always-focused input with @keydown.enter handles both scans; on-screen prompt tells the user what to scan. Touch fallback: item search + manual badge entry (small link, for damaged labels).

Labels (own print view, USBLabelBatch precedent):

  • /print/printedparts-labels public print route.
  • NEW physical size: 1in x 0.5in stock -> @page { size: 1in 0.5in; margin: 0 } one label per page (label printers feed roll stock; per-page = per-label). Layout: CODE128 barcode (JsBarcode, ~0.9in x 0.28in, displayValue false) + itemcode text under it (~7pt) + optional item name truncated. QR variant offered but barcode is default at this size (a 0.4in QR is at the edge of scanner tolerance; CODE128 of 3DP-0042 is comfortable).
  • Batch mode: pick items -> one label per page sequence for roll printers; also a ULINE mini-grid fallback for sheet printers (reuse mini72 pattern).

7. Metrics / reports (get_reports hook)

  • printedparts-stock - current stock levels w/ threshold flags (CSV).
  • printedparts-consumption - takes per item over a date range (CSV).
  • printedparts-by-person - takes grouped by employee (CSV).
  • Dashboard widget via get_dashboard_widgets(): low-stock item count.
  • Nice-to-have later: burn-rate (avg takes/week per item + weeks-to-empty projection) - plain SQL over the ledger, add once basics work.

8. Settings (get_settings_cards, category printedparts)

key default purpose
printedparts_code_prefix 3DP itemcode prefix
printedparts_default_threshold 5 seed for new items
printedparts_unknown_badge deny kiosk policy for unresolvable badges

9. Manifest

name printedparts, version 0.1.0, api_prefix /api/printedparts, core_version ">=0.12.0,<1.0.0", dependencies ["employees"], default_enabled false (site opts in - USB precedent).

10. Explicitly out of scope (v1)

  • Reservations/approvals, per-item cost, print-queue integration, multi-bin per item, email low-stock alerts (the reports + dashboard widget cover monitoring; alerting can ride the existing report-email endpoint later).

11. Risks / decisions taken

  • Cached quantity vs ledger drift: single-commit writes + a reconcile query in the stock report (flags items where cache != ledger sum).

Decision: the kiosk take endpoint is an unauthenticated WRITE

This is the first open mutation in the product - every existing kiosk endpoint (notifications, employees, shopfloor) is a read, and the closest write (USB checkout) is JWT + permission gated. Accepted deliberately, on these grounds, and any future open-write endpoint must meet the same bar:

  1. Decrement-only: it can reduce stock of an active item, nothing else - no catalog edits, no restocks, no reads it does not already expose.
  2. Fully attributed: it refuses to act without a badge that resolves per the site policy; every action lands in the ledger with SSO + name + time.
  3. Bounded blast radius: worst case is stock counts driven low, which the ledger makes visible and reversible (adjust with reason).
  4. Physically rate-limited: it exists for a touch screen on the shop floor; there is nothing to enumerate and nothing returned worth scraping.
  • 1x0.5in QR marginal: default to CODE128 barcode.
  • Not an Asset: no floor-map plotting or warranty for items. If a site later wants bins on the floor map, revisit via get_map_overlays (ADR-010).