QuoteDeclinedByCustomer was used for both approve and decline responses,
so approval notifications showed the wrong type in the log. Added a distinct
QuoteApprovedByCustomer = 16 enum value, wired up the correct type in
NotificationService, added default templates in both the service fallback
dictionary and SeedData, and updated placeholder hints in CompanySettings.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Previously the quick quote omitted the oven charge entirely, so saved quotes
were under-priced relative to full quotes from the same items.
Pricing: CalculatePricing now calculates ovenBatchCost = (cycleMin/60) × OvenOperatingCostPerHour
using DefaultOvenCycleMinutes (fallback 50 min), then adds it to the total as a quote-level
charge matching how PricingCalculationService handles oven costs.
Save path: SaveQuickQuoteRequest gains OvenBatchCost + OvenCycleMinutes; the Quote record
now stores OvenBatchCost, OvenCycleMinutes, and Total = ItemsSubtotal + OvenBatchCost.
Display: results card shows a sub-line under the estimate price:
"incl. oven 1 batch 50 min: $12.00"
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Anthropic returns overloaded_error (HTTP 529) during high-demand periods.
Previously this failed immediately with a generic error. Now the service
retries Sonnet once after 5s, then falls back to Haiku (a separate
capacity pool) after another 3s before giving up. If all three attempts
are overloaded the user sees a clear "high demand" message rather than a
generic error. Non-overload errors still log at Error level.
Also consolidated AI wizard error display in item-wizard.js: photo upload
failures were using browser alert() while analyze failures used the inline
red alert bar. All errors now go through aiShowError() so they always
appear consistently as the red bar below the Analyze button. Removed the
alert() fallback from aiShowError() itself.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Invoice-basis report showing taxable vs non-taxable sales, tax billed
by GL account, monthly trend table/chart, and full invoice detail grid.
Non-taxable invoice rows shaded grey for easy scanning. Quick-preset
date buttons (This Month, Last Month, YTD, Last Year) for common filing
periods. CSV export formatted for accountants and tax-filing software.
Gated behind AllowAccounting() like other financial reports.
- SalesTaxReportDto + 3 supporting DTOs in FinancialReportDtos.cs
- GetSalesTaxReportAsync on IFinancialReportService + implementation
- GenerateSalesTaxReportPdfAsync on IPdfService + QuestPDF implementation
- SalesTax / SalesTaxPdf / SalesTaxCsv actions in ReportsController
- Views/Reports/SalesTax.cshtml with Chart.js monthly trend chart
- Landing page card added to Finance section
- HelpKnowledgeBase and Help/Reports.cshtml updated with full docs
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- PowderCatalogController: Create, Edit, ToggleDiscontinued actions; searchable/filterable/sortable Index with pagination; AiLookup and AiAugmentFromUrl endpoints backed by IInventoryAiLookupService
- New views: Create, Edit, _Form partial (with AI-assisted field population), overhauled Index grid with completeness quality badges and responsive mobile cards
- New ViewModels: PowderCatalogIndexViewModel, PowderCatalogFormViewModel, PowderCatalogListItemViewModel
- AI lookup improvements: SpecificGravity field added to InventoryAiLookupResult; ApplyPowderFallbacks derives CoverageSqFtPerLb from specific gravity when docs omit it; DefaultTransferEfficiency (65%) applied everywhere transfer efficiency is null
- powder-catalog-ai-lookup.js: client-side AI lookup and URL augment wiring for the catalog form
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- PWA: manifest.json + minimal service worker so iOS/Android persist camera
permission after "Add to Home Screen"; theme-color and apple meta tags in layout
- PWA icons: 192x192 and 512x512 from transparent PCL logo; updated pcl-logo.png
- AI pricing: apply AdditionalCoatLaborPercent per extra coat on AI items,
matching the calculated-item path (was ignoring extra coats entirely)
- AI wizard: live price recalc when coats are added/removed; session-expiry
errors now show a clear "refresh and sign in" message instead of raw HTTP status;
smooth-scroll to follow-up/results sections on AI response
- Catalog lookup: exclude SKUs already in company inventory from results;
pass currentId on edit so own entry still appears; vendor-scoped search
with cross-vendor fallback; result count shown in multi-match modal
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
After the main AI lookup and catalog search, if CureTemperatureF or
CureTimeMinutes is still null but a TDS URL was found, fetch that page
and ask Claude to extract just the cure schedule.
- IInventoryAiLookupService.FetchTdsCureSpecsAsync: new interface method
- InventoryAiLookupService.FetchTdsCureSpecsAsync: fetches the TDS URL via
the existing FetchPageAsync pipeline (JSON-LD + doc-link extraction, HTML
stripping). If the page is a PDF or unreachable, returns Success=false
silently so no error surfaces in the UI. Otherwise sends a small targeted
prompt that asks only for cureTemperatureF and cureTimeMinutes and uses
MaxTokens=256 so the call is fast and cheap.
- InventoryController.ScanLabel: after catalog lookup, computes the resolved
cure values (catalog preferred over AI result). If either is null and a
TDS URL exists, calls FetchTdsCureSpecsAsync and merges any newly found
values back into aiResult before building the JSON response.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
QR scanning:
- Run BarcodeDetector and jsQR in parallel — jsQR starts after JSQR_DELAY_MS
(1.5 s) so both decode simultaneously. BarcodeDetector silently returns empty
arrays for some QR variants; running jsQR in parallel via a separate rAF loop
(rafId2) and its own off-screen canvas catches those cases. First decoder to
find anything calls handleQrResult and sets qrFound = true; the other stops.
Price extraction (two bugs):
- ScanLabel: unitPrice was catalogMatch?.UnitPrice ?? 0m, ignoring aiResult
.UnitCostPerLb entirely when no catalog match — changed to fall through to AI result
- AppendOffer: only read JSON-LD "price" field; Shopify AggregateOffer uses
"lowPrice" instead — now checked as fallback so Prismatic Powders prices are found
Camera pre-warm:
- Reverted localStorage approach (caused getUserMedia to fire on every page load,
showing Chrome's "Ask" prompt immediately before user clicked anything)
- Restored Permissions API gate: preWarmCamera only calls getUserMedia when
navigator.permissions.query returns 'granted', never risks a page-load prompt
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- LookupByUrlAsync now maps all identity + spec fields from Claude response
(manufacturer, SKU, colorName, description, sdsUrl, tdsUrl, unitCostPerLb, etc.)
Previously only augmenting fields were mapped; Columbia QR path left 80% blank
- Vision scan follow-up: after ScanLabelAsync reads label text, automatically run
LookupAsync using the extracted manufacturer + color/SKU to fill SDS/TDS URLs,
product page, image, description, and any specs not printed on the bag;
label values (cure schedule, SKU) remain authoritative and are never overwritten
- SDS/TDS URL extraction: added ExtractDocumentLinks() that scans anchor tags in
raw HTML before tag-stripping, injects found URLs as [Structured Data] lines so
Claude can read and echo them back in the JSON response; previously all hrefs
were lost with the HTML stripping
- Added SdsUrl/TdsUrl to InventoryAiLookupResult, Claude system prompt JSON schema,
LookupAsync mapping, and ScanLabel response (catalog match ?? aiResult fallback)
- SDS/TDS now also stored on auto-contributed catalog entries
- jsQR inversionAttempts: 'dontInvert' → 'attemptBoth' for better QR detection
under varying label contrast and lighting conditions
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Platform PowderCatalogItem table (IPlainRepository, no tenant filter) with
full spec fields: cure temp/time, finish, color families, clear coat flag,
coverage sq ft/lb, transfer efficiency, IsUserContributed
- Two EF migrations: AddPowderCatalogItem + AddPowderCatalogSpecFields
- PowderCatalogController (SuperAdminOnly): import from Prismatic JSON scrape,
Lookup AJAX endpoint (catalog-first, ranked by SKU exact match), stats view
with Tenant Contributed card
- Unified smart Lookup button on inventory Create/Edit: catalog hit fills all
fields via catalogSnapshot pattern; AI augments cure/finish data from product
URL if subscription enabled; catalog miss falls through to AI lookup
- In-browser label scanner (_LabelScanModal): getUserMedia live camera feed,
jsQR auto-detects QR codes in rAF loop; "Scan Label Text" fallback sends
captured frame to Claude vision via /Inventory/ScanLabel
- ScanLabel endpoint handles both QR URL path (LookupByUrlAsync) and vision
path (ScanLabelAsync); auto-inserts unrecognized products as
IsUserContributed=true; returns wasInCatalog/addedToCatalog flags
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds a customizable QuoteSent SMS template to seed data and
DefaultTemplates so companies can edit the quote approval message
from Notification Templates. Wires NotifyQuoteSentSmsAsync to use
the template system instead of a hardcoded string. Updates
SmsConsentConfirmation wording to mention quote approvals alongside
job updates. Help docs and AI knowledge base updated to match.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add 'Send Quote via SMS' button on quote details page that sends the approval
link to the customer via SMS (respects NotifyBySms, handles prospects via ProspectPhone)
- Reuses existing valid approval token rather than regenerating, so a previously
emailed link stays valid when SMS is also sent
- Fix Twilio appsettings.json placeholders (real credentials moved to gitignored
appsettings.Development.json)
- Fix passkey login ignoring ReturnUrl: biometric login on the login page now
respects the form's ReturnUrl hidden field so QR-code and deep-link flows
redirect correctly after authentication instead of always going to the dashboard
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Three-tier SMS gate: platform kill-switch → admin force-disable → plan AllowSms → company opt-in
- CompanySmsAgreement entity records admin acceptance of TCPA terms with IP, user agent, and terms version
- SMS terms of service modal on Company Settings with versioned re-agreement (AppConstants.SmsTermsVersion)
- Dev redirect: non-production SMS routed to Twilio:DevRedirectPhone to protect real customer numbers
- Removed redundant Ready for Pickup SMS (Job Completed covers it)
- Role-based compose modal on job completion: Admin/Manager reviews and edits before send; ShopFloor auto-sends
- Send SMS button on job details for ad-hoc messages (Admin/Manager only)
- SendJobSmsAsync auto-appends STOP opt-out language if missing
- Migrations: AddSmsGating, AddCompanySmsAgreement
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
DashboardReadService no longer loads full entity lists and filters in memory.
All job panels (today/overdue/in-progress) now execute targeted COUNT + capped
SELECT queries in SQL. AR aging buckets, powder order lines, bill totals, and
active-customer counts are all aggregated at the DB level. The SuperAdmin action
previously loaded every company row to compute plan distribution and alert lists;
it now delegates to a new GetSuperAdminDashboardDataAsync() that uses SQL GROUP BY
and projections instead.
DashboardIndexData record updated to carry pre-sliced counts and capped lists so
the controller only does lightweight DTO projection. DashboardPowderOrderLineData
replaces the deep Job→JobItem→Coat Include chains with a single flat coat query
projected in SQL. OnlineUserMiddleware switches its per-user throttle from a
static ConcurrentDictionary (grows forever) to IMemoryCache with a 60-second
sliding expiry.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
When AI Lookup fetches a manufacturer product page, it now extracts the
og:image (Open Graph) meta tag before stripping HTML tags. The image URL
is returned in InventoryAiLookupResult.ImageUrl and automatically shown
as a preview on the Create/Edit form alongside the other filled fields.
The preview includes a Remove button to clear the image, and the Wrong
Match? button clears it along with the other AI-filled fields.
On the inventory Details page a product image card is rendered above the
Stock & Pricing card whenever ImageUrl is set. The field is nullable so
existing records and powders without an image are unaffected.
New field: InventoryItem.ImageUrl (nvarchar, nullable).
Migration: AddInventoryItemImageUrl.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Phase 3 — eliminated ApplicationDbContext from all non-exempt controllers,
routing all data access through IUnitOfWork. Added IPlainRepository<T> for
the four platform entities (Announcement, BannedIp, DashboardTip, ReleaseNote)
that intentionally don't extend BaseEntity and therefore can't use the
constrained IRepository<T>. Added permanent-exception comments to the 18
controllers that legitimately retain direct DbContext access (Identity infra,
cross-tenant platform ops, bulk streaming exports).
Phase 4 — added EnforceDataAccessArchitecture() to Program.cs, a startup
gate that reflects over every Controller subclass and throws at boot if any
non-exempt controller injects ApplicationDbContext. The app cannot start with
a violation.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Six IUnitOfWork properties upgraded from generic IRepository<T> to domain-specific
typed interfaces (IJobRepository, IQuoteRepository, IInvoiceRepository,
ICustomerRepository, IBillRepository, IPurchaseOrderRepository). Each backed by a
concrete typed repository that encapsulates complex include chains previously
inlined in controllers.
Also adds IFinancialReportService and IOperationalReportService stub implementations
(NotImplementedException placeholders) to Application.Interfaces and Infrastructure.Services,
registered in Program.cs. These are the migration targets for ReportsController's
aggregate query methods in Phase 2.
No controller behaviour changed in this commit — all callers still compile because
typed interfaces extend IRepository<T>.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Company-level toggle now grants access regardless of plan tier, checked
before the plan gate. Useful for enabling the feature on individual
Pro/Basic companies without upgrading their plan.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Rather than relying on reactive 65s retries, each semaphore slot is held
for at least MinBatchIntervalSeconds (20s). With 2 concurrent slots that
limits throughput to ~3 batches/min × ~2k tokens = ~6k output TPM,
safely under the 8k/min limit.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Haiku has generous rate limits so parallelism is safe again. Retry
logic catches any 429s. Progress estimate updated to ~8s per wave.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Testing Haiku 4.5 for catalog price analysis — structured JSON output
with explicit rules is well within its capabilities. Revert to Sonnet
if result quality is insufficient.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Tier 1 Anthropic accounts are capped at 8,000 output tokens/minute on
Sonnet. 3 concurrent batches burst well past that, causing 429s.
- MaxConcurrentBatches: 3 → 1 (sequential prevents burst)
- Add retry: on rate_limit_error, wait 65s then retry up to 3 times
so the per-minute window resets before the next attempt
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Item name + category path give Claude sufficient context for surface area
estimation. Descriptions add input tokens without meaningfully improving
verdict quality.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Skip $0-priced items (placeholders/category headers) in RunAiPriceCheck
- Build full category path (e.g. "Cerakote > Firearms") via BuildCategoryPath
so Claude receives coating-type context — Cerakote pricing differs significantly
from standard powder coat
- Update AI system prompt to instruct Claude to use the category path when
determining process type, equipment, cure times, and market rates
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
500-item catalog was making 50 sequential API calls, causing progressive rate-limit
throttling (explains "super slow towards the end") and ~$3 in credits.
- BatchSize: 10 → 25 (word limits are in place; 25 items × ~80 tokens ≈ 2000
output tokens, well within MaxTokens=8192 — the original truncation cause)
- Run up to 3 batches concurrently via SemaphoreSlim(3) — independent API calls
with no shared state, so no growing context issue
- For a 500-item catalog: 50 sequential calls → 20 calls in ~7 parallel waves,
roughly 4× faster and 60% cheaper
- Dropped unused `costs` param from AnalyzeBatchAsync (system prompt has all costs)
- JS progress timing updated to reflect parallel waves
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Still seeing stubs despite MaxTokens=8192 — smaller batches and explicit
word limits in the prompt eliminate any remaining truncation risk.
- BatchSize: 15 → 10 (~1200 output tokens per batch vs. potential 3000+)
- Prompt: added 20-word cap on assumptions, 25-word cap on reasoning
- Prompt: strengthened "nothing before or after the '['" instruction
- Error log: now includes item IDs and first 300 chars of raw response
so the next failure tells us exactly what Claude returned
- JS timing: updated batch divisor from 25 → 10 to match actual batch size
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Root cause: MaxTokens=4096 was too low — 25 items at ~250 tokens each hit the
limit mid-array (logged error showed Path: $[17]).
- MaxTokens: 4096 → 8192
- BatchSize: 25 → 15 items (keeps each response well under the limit)
- StripJsonFences → ExtractJsonArray: now also handles prose before/after the
JSON array, and recovers truncated responses by finding the last complete
object and closing the array — so partial batches return whatever Claude
finished rather than nothing
- GET action: added try-catch around ResultsJson deserialization so a bad DB
row shows a friendly "re-run" warning instead of a raw error page
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude reviews every active catalog item against the shop's own operating costs
and returns a per-item verdict (below-cost / thin-margin / high / ok) with a
suggested price range, cost floor, and assumptions.
- New entity: CatalogPriceCheckReport (JSON blob, archived per company)
- New service: IAiCatalogPriceCheckService / AiCatalogPriceCheckService
batches items 25 at a time to stay within model context limits
- Two new controller actions: GET AiPriceCheck (view report) + POST RunAiPriceCheck
- AiPriceCheck view: summary cards (counts by verdict), color-coded item cards
with Edit Price link, assumptions detail, and loading spinner on submit
- AI Price Check button added to catalog Index header
- Migration AddCatalogPriceCheckReport applied
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- New AI Quick Quote floating button: staff type a verbal description to
get an instant price estimate for phone/walk-in customers; detected
color names are fuzzy-matched against inventory for stock status;
saves draft quote under a Walk-In / Phone customer with one click
- Inline customer change on Quote Details and Job Details: always-visible
native select with inline confirmation banner (no TomSelect dependency);
ChangeCustomer AJAX endpoints on QuotesController and JobsController
- Quote Edit page: customer dropdown is now editable (lock removed)
- Fix AutoMapper missing CatalogCategory -> UpdateCategoryDto mapping
that caused a crash on the catalog category Edit page
- Help docs and AI knowledge base updated for all three features
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
spouliot