We ship every week. Here's the receipt.

Every printer setup — LAN, USB, serial, Bluetooth — now uses the same V5 Print Relay pattern. A new Setup Wizard asks two questions and tells you exactly which flavour (Desktop, Pi, or Mobile companion) fits your shop. Ask V5 has been updated to explain why cloud apps can never talk to a printer directly, and what to do about it.

• newPrinter Setup Wizard at Settings → Printers → Setup — pick where the printer lives and what your operators use, get one concrete recommendation with cost, setup steps and a Continue button.
• newAsk V5 now explains the three Print Relay flavours: Desktop Relay for Windows/Mac/Linux, Pi Relay (~$40 pre-imaged Raspberry Pi), and Mobile Relay for Bluetooth stations (roadmap). Ask 'how do I print from an iPad?' or 'my printer has an IP, why can't V5 see it?' to see the full guidance.
• improvedPositioning is honest: cloud apps physically cannot reach private LAN IPs, so every setup needs one small on-site Relay. Same constraint QuickBooks / Xero / Salesforce / Shopify POS all have — we just don't dress it up.
• improvedThe Setup Wizard steers Mac and iPad shops toward the Pi Relay by default (better than dedicating a Mac as a print server) and Windows shops toward the free Desktop Relay on an existing office PC.

The Print Helper ZIP now downloads with a dated filename and no browser caching, so Windows users do not accidentally rerun the older VBS-based print agent package.

• fixedPrint Helper downloads now use a dated ZIP and folder name so customers can tell they are running the latest installer, not the older VBS launcher.
• fixedThe installer bundle no longer includes or invokes any Windows Script Host / VBS launcher path.
• improvedThe Add Printer wizard now tells users to extract the latest dated ZIP into a fresh folder and remove any older v5-print-agent folder before retrying.

The Windows Print Helper now installs with a simple per-user Startup shortcut instead of a scheduled task, avoiding the access-denied failures seen on locked-down PCs while keeping USB printer discovery automatic.

• fixedReplaced Windows scheduled-task registration with a per-user Startup launcher so the installer can finish without UAC or administrator approval.
• improvedThe helper starts immediately after install and again when the Windows user signs in, then reports installed USB printers back to the Add Printer wizard automatically.
• improvedDownload filenames, installer screens and fallback instructions now consistently say Print Helper and avoid technical agent language.

The V5 print agent (v1.2) now enumerates every printer installed on its Windows PC and reports the list to V5. The Add Printer dialog turns 'Windows printer name' from a free-text field into a dropdown — plug in a Zebra or GoDex over USB, refresh, pick it, save.

• newPrint agent v1.2 auto-detects local printers via Get-Printer (Windows) or lpstat (mac/Linux) and reports them every 60s to /api/public/print-agent/printers.
• newAdd-printer dialog shows a live dropdown of printers detected on the selected Agent ID, with driver names and default marker. Falls back to text input if the agent hasn't reported yet.
• improvedRolling out a new site with 20 printers no longer means typing 20 printer names exactly right — plug them in, click Refresh, pick from the list.

Any thermal printer plugged into a Windows PC over USB — Zebra ZPL and GoDex EZPL — now prints through the V5 print agent without needing a network port. Ask V5 now walks customers through the exact driver and setup steps.

• newPrint agent gains a 'USB (Windows)' transport that sends raw ZPL or EZPL straight to the Windows spooler via winspool.drv — no LAN, no port 9100.
• newAdd-printer dialog exposes Connection = 'USB (Windows)' plus a 'Windows printer name' field and Printer type picker (Zebra ZPL / GoDex EZPL).
• improvedAsk V5 now answers 'how do I connect my USB Zebra / GoDex printer?' with the full driver, printer-name and language-picker walkthrough.

The welcome dashboard gains a compliance-and-reliability batch of predictive tiles: unsigned e-signature aging, downstream lot exposure from any current quarantine, instruments coming out of calibration in 30 days, a Poisson-based complaint spike detector, and assets whose mean-time-between-failures is collapsing vs. baseline.

• newSignature backlog risk — open e-sig requests older than 24h (and a critical count over 72h) so release-blocking reviews get chased before they breach SLA.
• newRecall exposure — walks lot_genealogy from every quarantined/hold lot to count distinct downstream child lots, showing recall blast radius under 21 CFR 7.42 / 211.204.
• newCalibration cliff 30d — instruments with next_due_at inside 30 days, plus any already overdue flagged critical, aligned with 21 CFR 211.68 / 820.72 / ISO 17025.
• newComplaint velocity — Poisson z-score comparing the last 14 days of complaints against a 12-week biweekly baseline, catching field-issue spikes early (21 CFR 820.198 / 211.198).
• newMTBF collapse risk — assets whose 30-day mean-time-between-failures has dropped more than 30% vs. the prior 90-day baseline, an early sign of wear or drift before a line goes down.
• improvedAll five tiles carry the ✨ sparkle chip and a plain-English formula in the customise drawer, so operators see a forecast and auditors can trace the math.

The dashboard gains ten forward-looking KPIs — yield forecast, cycle-time anomalies, changeover drag, CAPA effectiveness risk, deviation cluster alerts, vendor OTIF drift, expedite spend, training-cliff, shift OEE trend and PM overrun risk — each marked with a sparkle chip and a one-tap explanation of the formula behind it.

• newBatch yield forecast — open work orders flagged when the last 90 days on the same formula run ≥5pp below expected yield.
• newCycle-time anomaly — in-progress work orders past 1.5× the median cycle time for their formula.
• newChangeover drag 7d — total hours lost to changeover state on all lines over the last week.
• newCAPA effectiveness risk — closed CAPAs with an effectiveness check due in 30 days but not yet verified, or verified ineffective.
• newDeviation cluster alert — Poisson spike detector comparing this week's deviations against the prior 12-week baseline.
• newVendor OTIF drift — change in on-time delivery over the last 30 days vs. the prior 60 days, so the direction of travel is obvious.
• newExpedite spend 30d — purchase-order lines flagged as expedite in the last 30 days, an early warning of planning stress.
• newTraining cliff 30d — active operator competencies expiring inside the next 30 days, so re-qualification can beat the kiosk lockout.
• newShift OEE trend — percentage-point delta of average OEE over the last 14 days vs. the prior 14 days.
• newPM overrun risk — preventive maintenance jobs past their scheduled end and still open.
• improvedEvery predictive KPI shows a sparkle chip and — in the customise drawer — a plain-English formula, so operators know a value is a forecast and auditors can trace the math.

Inventory → By item now predicts when each SKU will run out and when to reorder, using 90 days of ledger usage plus scheduled work orders and expected PO arrivals. An AI Briefing panel turns the numbers into a short, planner-actionable to-do list — with a deterministic reasoning trace per item so every bullet is auditable.

• newPer-item run-out prediction — days-of-cover, run-out date and reorder-by date computed from 90 days of consume / ship / scrap plus each item's lead time.
• newCroston's SBA forecasting for intermittent items (spare parts, low-volume SKUs) so daily-average usage isn't overstated by rare bursts.
• new180-day balance-forward simulation folds in scheduled work-order draws and expected PO arrivals for a true run-out date, not straight-line extrapolation.
• newAI Briefing panel writes 3–6 planner-actionable bullets grouped by urgency (reorder now → within 7d → obsolescence), scoped to the filtered view.
• improvedEvery AI bullet has a one-line reasoning trace per item so recommendations remain auditable — deterministic fallback runs if AI is unavailable.

Production lines now roll up per shift with target-vs-actual variance, best/worst call-outs and a printable Monday-meeting report. A live banner and dashboard KPI expose any non-running line time that has no reason code so every minute is explained.

• newShift output report — target vs actual per shift × line, unaccounted-downtime called out, per-line best/worst shift, printable for the Monday meeting.
• newUnaccounted downtime (24h) KPI tile on the dashboard, deep-links straight to the shift report.
• newLive red banner on the Lines page whenever non-running time in the last 24 hours has no reason code attached — one tap to fix from the same screen.

Draw sample now leaves the loading skeleton and opens the selected request with its item, lot and label context.

• fixedSampler Draw sample no longer hangs on loading blocks when lot context uses the current supplier-name field.

Open sample requests now resolve reliably when an operator taps Draw sample from the sampler queue.

• fixedSampler queue cards now open the matching draw screen without falling into the Request not found state caused by request relationship lookups.

The lab now scales to thousands of SKUs: each product has its own QC tree across Receiving, IPC, FG, Stability, Release and Retain, powered by reusable Test panels bound to items by stage. The flat sample-plans register has been replaced by a Product QC register plus a Panels library — customers with hundreds of products no longer wade through one giant list.

• newProduct QC register — per-product coverage across every stage, with a missing-QC filter to spot untested SKUs fast.
• newTest panels library — build reusable bundles once, bind to any product by stage.
• improvedSample plans are now pure sampling rules (qty, AQL, frequency) rather than a mixed test-and-schedule list.

Lab Test methods and Sample plans now load their initial registers directly through the workspace-scoped browser client, avoiding the slow server-function handshake that could leave those setup pages spinning for tens of seconds.

• fixedLaboratory setup pages for Test methods and Sample plans now populate without the long spinner seen on larger workspaces.

Fresh pharma, supplements, cannabis, vet-pharma and radiopharma workspaces now come pre-wired for assay-adjusted dispensing: every active-ingredient formula component ships with a potency contract (basis, reference %, assay method, salt factor) and the largest excipient is flagged as QS. The seed engine goes through a dedicated SECURITY DEFINER routine so approved sample formulas stay immutable to normal edits while the contract fields still stamp cleanly at seed time.

• improvedNew workspaces in regulated industries open with potency contracts and QS diluents already configured — the kiosk shows assay-adjusted targets on day one, no manual setup required.
• fixedApproved sample formulas no longer block potency stamping during seed, without loosening the approved-formula immutability guard for real users.

Customers who choose on-prem can now download signed installers, manage license tokens, and see live activations from Workspace Settings. SG Systems staff manage releases, channels (Stable / Validated / Beta) and revocations from the Platform console. Installers verify a signed Ed25519 license offline and phone home for updates and revocation.

• newSelf-install portal for tenant owners: view active licenses, copy a single-use license token, and download the latest signed build for your channel.
• newPlatform console for SG Systems staff: issue and revoke licenses, publish releases, and manage Stable / Validated / Beta channels.
• newPublic installer endpoints — activate, heartbeat, latest and CRL — for on-prem hosts to phone home for updates and revocation.
• improvedEvery on-prem license is a signed Ed25519 token: works offline, tamper-evident, hardware-bindable, and revocable via the published CRL.

Tests (what you measure — analyte, unit, spec, pharmacopoeia ref) are now a first-class catalog separate from Methods (how you measure — SOP, technique, instrument). One test can be run by multiple methods (e.g. Assay by HPLC or UV) and the acceptance criteria live once, on the test. When an OOS investigation reaches Phase II with disposition Confirmed, a linked CAPA is opened automatically (source = OOS, severity carried over) and both records back-reference each other — no more forgetting to escalate.

• newLaboratory · new Tests catalog at /app/lab/tests with categories, spec type (range / min / max / pass-fail / report) and pharmacopoeia reference.
• improvedMethods are now purely procedural — SOP, technique, instrument, duration. Each method links to its parent test.
• newOOS · Phase II Confirmed auto-opens a linked CAPA tagged source=oos, with severity and description carried across.

The four modals an operator sees when stopping a batch (Stop, Abort with reason, Reconcile weighed material, Remake) are now heavy-duty kiosk-sized: bigger tap targets, icons twice the size, minimal wording, and no long descriptive paragraphs — the operator sees the choice, not a policy document. The source-lot chip on the dispense header now counts down in real time from typed weight as well as live-scale weight, so a manually entered pour immediately reduces "kg left" instead of waiting for the bag to be accepted.

• improvedKiosk · Stop / Abort / Reconcile / Remake dialogs redesigned for gloved operators — larger buttons, larger icons, one-word actions.
• fixedDispense · source-lot "kg left" countdown now updates from manual weight entry as well as live-scale readings.

The lab loop from bundle #1/#2/#3 now has a UI. The Item editor gets an "Assay required on every lot" toggle plus a default-methods multi-select — flick it on and every future goods-receipt line auto-opens a lab sample request tagged trigger=receive. The kiosk run-complete screen now shows Reagent prep and Instrument dropdowns with inline warnings (expired / QC-fail / cal-overdue / out of service) so the operator sees the Part 11 gate before hitting submit; the selection populates lab_test_run_consumables and lab_test_run_assets on completion. The Lot 360 drawer gets a new Assay tab — every lab run against samples taken from any lot of this material, with method, result, pass/fail/OOS, and date.

• newItem editor · new "Assay required on every lot" switch with a scrollable checkbox picker for default assay methods. Auto-hooks receiving so the lab request opens without operator intervention.
• newItem · Inventory & Handling tab now shows Assay required (Yes — N methods / No) so it's visible without opening the edit dialog.
• newKiosk lab run · Reagent prep and Instrument dropdowns on the Complete step, with red inline warnings when the selected prep or instrument will trip the Part 11 gate — no more mystery block on submit.
• newLot 360 · new Assay tab lists every lab run against samples from this lot (or every lot of the material when opened material-scoped), with method, result, unit, and Pass / Fail (OOS) status.

Dispense bag labels used to render the red diamond with just the code number inside (e.g. "07") and a coloured DANGER/WARNING banner beside it. The banner was redundant with the ingredient-select warnings the operator already acknowledges, and the number wasn't a proper hazard symbol. Diamonds now draw the actual GHS glyph (exclamation for GHS07, flame for GHS02, skull for GHS06, etc.) and the signal-word box is gone.

• improvedBag labels: red GHS diamonds now contain a real pictogram (exclamation, flame, skull, corrosion, health hazard, environment, gas cylinder, exploding bomb, flame-over-circle) instead of the two-digit code.
• improvedBag labels: removed the coloured signal-word banner (DANGER/WARNING) — that information is already surfaced at ingredient select.

The Lab → Potency → Formula loop no longer relies on someone remembering to raise a sample request when an API arrives. Items now carry an assay-required flag plus a list of default assay methods; the moment a receipt line posts, a lab_sample_request is auto-opened tagged trigger=receive. Run completion now enforces a Part 11 hard gate — analyst certification on the method, reagent-prep expiry, and instrument calibration due date — and writes every reagent lot / instrument used into normalized junction tables so every reviewed result is defensible on audit.

• newItems · new Assay required flag and Default assay methods list. When flagged, every goods-receipt line auto-opens a lab sample request against the received lot.
• newLab · run completion is now blocked with a clear reason if the analyst isn't certified on the method, a declared reagent prep is expired or failed QC, or the instrument's calibration is overdue.
• newLab · every completed run now records its consumed reagents (reagent, prep, source lot, qty) and instrument + calibration used into audit-grade junction tables — the basis for future reagent inventory decrement and lot-level traceability views.

The kiosk top-right no longer stacks Home, user pill and hamburger side-by-side. It's a single operator chip — tap it to open a right-hand sheet with Go to · Session · Preferences, so every action lives in one predictable place across every kiosk screen. Also removed a redundant membership-lookup round-trip from every Laboratory server call and debounced the Control Tower's realtime refetches, so the lab section keeps up with the rest of the product.

• improvedKiosk header collapses Home + user + settings into one operator chip → opens a sheet with Go to / Session / Preferences. Fewer buttons, more room for the screen title on narrow devices.
• improvedLaboratory section is noticeably faster — every lab server call skipped a redundant memberships lookup that was doubling latency; RLS still enforces workspace isolation.
• improvedLab Control Tower coalesces bursts of realtime events into one refetch every 2s and drops the safety-net poll from 30s to 60s, so busy dispense/QC activity no longer stampedes it.

Previously only pharma workspaces shipped a full LIMS test-method library. Dietary supplements, cannabis, veterinary pharma, cosmetics, radiopharma and blood/tissue workspaces now all get a baseline analytical panel (assay HPLC, ID, LOD, appearance, microbial, heavy metals, plus industry extras like THC/CBD potency and residual solvents for cannabis) at seed time — so the assay → lot → dispense potency loop is demoable end-to-end in every regulated sector, not just pharma.

• newSeed engine ships a default lab_test_methods panel for supplements, cannabis, vet-pharma, cosmetics, radiopharma and blood/tissue when the pack doesn't define its own — cannabis adds cannabinoid potency HPLC, residual-solvents GC-MS and pesticide LC-MS/MS.
• newSeed engine auto-derives a FG-release lab_sample_plan for every finished good with release specs when the pack didn't ship one — mapping each release test to its baseline method key so QC has samples-to-draw on day one.
• improvedseedRawLotPotency now covers pharma, dietary supplements, cannabis, veterinary pharma, cosmetics and radiopharma. Active-ingredient prefixes are industry-scoped (API-, DI-, CB-DIST-, VET-*, COS-actives, ISO-/RP-API) and the QS diluent picker no longer requires an EXC-/SOL- prefix.

Potency is now visible where you work: set the assay contract (basis / reference % / method / salt factor) directly on any formula component row, and see the latest reviewed lot assay results on the item detail page. Also patched a seed-data bug where a pharma formula's own finished good and WIP intermediate were being inserted into the bill of materials as consumed components.

• newFormula component rows show a Potency chip (pharma / supplements / vet-pharma / cannabis / radiopharma only) — opens a popover to set potency basis, reference %, LIMS assay method and salt factor without leaving the row.
• newItem detail (raw materials only) shows a Potency & assay card listing recent lots with the latest reviewed assay %, tested date, source (LIMS run vs vendor CoA) and retest history.
• improvedKiosk dispense: when a potency contract is set but no reviewed assay is on file for the selected lot, a muted "using nominal target — no assay on file" chip appears instead of silently falling back.
• fixedSeed data: pharma multi-stage formulas were emitting their own finished good and WIP intermediate into formula_components (showing a $1.7M FG line inside its own BOM). Backfill removes offending rows on existing tenants and the industry-seed engine now skips produced outputs during component aggregation.

A freshly seeded pharmaceutical (or dietary-supplements) workspace is now demoable end-to-end for LIMS-driven potency adjustment without any manual setup. Every active ingredient gets a potency contract, the largest excipient is flagged as the QS/counter-balance, and every incoming raw-material lot ships with a reviewed passing HPLC assay result — so the kiosk auto-adjusts the target and reduces the QS on lot selection out of the box.

• newPharma / dietary-supplements seed: active-ingredient formula components (API-*) are stamped with potency_basis, reference_potency_pct, assay_method_id and salt_factor at seed time.
• newPharma / dietary-supplements seed: the largest excipient by qty per formula is marked is_qs_component so counter-adjust has a partner.
• newPharma / dietary-supplements seed: each active raw-material lot receives an incoming release lab sample and a reviewed passing HPLC assay run (96–102%) with a QC reviewer e-signature — the exact shape resolveLotPotency() reads.

Dispense kiosk now preserves total batch mass automatically: when a component's target is potency-adjusted from its LIMS assay, the flagged QS/counter-balance component reduces its own target by the realised delta on the primary bags. Work-order component overrides accept a new `adjust` action so a potency correction becomes a first-class, signable audit record alongside skip/delay/substitute.

• newQS/diluent components (is_qs_component) automatically absorb the mass delta from any potency-adjusted primary they partner with, so the total batch weight stays on spec even after LIMS-driven target changes.
• newWork-order component overrides support an `adjust` action with a structured adjustment_reason (potency / substitution / manual), giving auditors a single row per component to review and sign.
• improvedAdjust-action overrides are constrained at the database level to always carry an adjusted quantity and reason — no silent adjustments.

Formulas can now carry an assay contract. When a component names a LIMS test method, the kiosk automatically adjusts the weigh-out target using the latest reviewed passing lot result, shows the operator the full factor formula, and snapshots the assay values used onto every bag.

• newFormula components can specify a potency basis, reference potency, LIMS assay method, and salt factor — the kiosk uses these to compute an adjusted target per lot.
• newDispense kiosk shows an 'adjusted target' strip when potency correction applies, with the exact assay/LOD/factor and a note when the lot has been re-assayed.
• newEvery dispensed bag now captures the assay %, LOD, basis, adjustment factor and source LIMS result id — immutable per Part 11.
• improvedAssay-adjusted charge glossary page updated to describe the kiosk implementation as on the roadmap while the multi-phase rollout ships.

Kiosk idle handling now lives in one guarded path, so admin users can leave a kiosk tab open without being unexpectedly signed out of the main app.

• fixedKiosk idle handling no longer runs duplicate timers that could unexpectedly sign out shared admin browser sessions.
• fixedOperator-only kiosk sessions still use the protected idle warning flow before returning to kiosk login.

New Bakery and Medical Device demo workspaces now seed data that matches the updated production flows, including whole-bag bakery steps and completed serialized device batches for UDI label reprints.

• improvedBakery demo formulas now include whole-bag issue steps so operators can test the whole-bag popout in realistic production data.
• improvedMedical Device demo batches now seed serialized finished units for completed DHR work orders, so UDI labels can be reprinted from the work order.
• fixedPurchase order receiving screens can now reliably resolve line items instead of showing a missing item relationship error.

The kiosk Pick lists tile now counts open pick lists instead of open sales orders, and pick-list-only operators stay on the launcher until they choose the tile.

• fixedPick-list badges now reflect actual open pick lists, including Material issue and Shipping-specific queues, instead of showing unrelated sales-order counts.
• fixedOperators with only pick-list access no longer skip the kiosk launcher automatically; they still choose the Pick lists tile before opening a job.

Bakery demo workspaces now include explicit make-to-order sales demand that exceeds finished-goods stock and open work orders, so the AI suggest work orders drawer returns real proposals instead of an empty covered-demand state.

• fixedAI suggest work orders now has seeded bakery shortages to analyse, so planners can test proposal creation immediately in demo workspaces.
• improvedThe MRP showcase seed adds marked make-to-order demand without changing existing inventory, pick, or work-order workflows.