main
3 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
|
03e030d6d2 |
feat: tier 3 polish — M12/13/14/15 deep cleanup
Closes the remaining ~50h of items from CONFORMITY-12-15.md across all
four modules. Single umbrella migration (2026_06_05_000004) lands four
tables + 5 column additions, no downtime risk.
== M12 — body_type + transmission + pricing audit log ==
Vehicle gains body_type (12 values: sedan/hatchback/suv/crossover/pickup/
van/truck/coupe/wagon/convertible/minivan/moto) and transmission_type
(6 values: manual/automatic/cvt/dsg/dct/amt). These are separate from
vehicle_class so admin can configure DSG-only coefficients without
contaminating the SUV detection.
PricingCoefficient.matches() now also tests:
- conditions.body_types[] against ctx.body_type
- conditions.transmissions[] against ctx.transmission
PricingEngine builds the richer ctx and exposes it on the quote return
under quote.context.
New pricing_application_logs table (append-only) — call
PricingEngine::logApplication($quote, $subject, $vehicle, $client, $part)
after applying a price to a WO line. Stores base, final, full
applied[] array, and the ctx snapshot so the question "why was this
priced at 218 lei in March?" stays answerable forever.
PricingCoefficientResource form gains CheckboxList for body_types and
transmissions (3-column layouts, full-width). Both are optional —
empty list = applies to anything.
== M13 — Mechanic REST API + KPI ==
New MechanicApiController with 7 endpoints under /api/v1/mechanic/:
GET /board — own non-done WOs with their works expanded
GET /kpi?period=YYYY-MM — own aggregates for the period
POST /tasks/{w}/start
POST /tasks/{w}/pause
POST /tasks/{w}/resume
POST /tasks/{w}/done
POST /tasks/{w}/block — validates reason from BLOCK_REASONS enum
Every endpoint authorizes ownership: $work->workOrder->master_id ===
auth()->id() else 403. board() returns null pending_works so native
apps don't make round-trips. workPayload() emits efficiency_pct and
efficiency_class on every response.
New MechanicKpi Filament page at /app/mechanic-kpi (Service group). Same
aggregation logic but tenant-wide: groups WorkOrderWork rows by
master_id for the selected period, computes totals + efficiency_pct +
revenue. Period navigation via ◀/▶ buttons, default = current month.
Color-coded efficiency badges (green ≤100%, amber ≤130%, red >130%).
Rows sort by revenue descending — easy "top earners this month" view.
== M14 — OCR async via Laravel queue ==
New ocr_jobs table (id, supplier_id?, source_type, file_path, status,
result JSON, error_message, ai_provider, tokens_used, purchase_id?,
processed_at). Idempotent migration.
New OcrJob model + ProcessOcrJob queueable job. Job re-establishes
tenant context inside the worker (Company::find + TenantManager::setCurrent)
since queue workers don't inherit middleware-resolved tenants.
handle() walks: status=pending → processing, calls OcrInvoiceService::extract,
on success → status=done + result + ai_provider; on throw → status=failed
+ error_message. Failed jobs auto-retry once (tries=2) with 120s timeout.
The existing synchronous OcrInvoiceService stays for inline use cases
(tests, quick imports). The job is now the canonical path for the
admin UI to keep requests sub-100ms.
== M15 — eta_promised + JSON tracking + notifications log ==
Three new wo columns: eta_promised (initial commitment, never changes),
eta_change_reason (text for "așteptăm piesă"), eta_updated_at (when
the current eta was last touched). Existing eta_at remains as "current"
ETA so the UI can render both side-by-side.
New /api/track/{token} JSON endpoint (public, tenant-scoped via subdomain):
number, status, status_label, progress %, client, vehicle, plate, master,
eta_promised, eta_current, eta_change_reason, total, pay_status,
pending_approvals[] (each with kind/id/name/amount/approve_url —
signed URLs ready for native app webview),
timeline[] (from activity_log, last 20 events).
NotificationDispatcher::dispatch() gains optional workOrderId param.
Every send call (success or failure) now writes one row to the new
client_notifications_log table with channel/template_key/status (sent
or failed)/error_detail/sent_at. Failures of logging are swallowed
so a missing activity_log never breaks notifications. workOrderReady
and paymentReceived pass the WO id through; others can be wired in
future commits without schema change.
New tables tracked:
client_notifications_log — every push to client, append-only
pricing_application_logs — every pricing decision, append-only
ocr_jobs — async OCR job queue
== Tests ==
PolishTier3Test (11):
- M12: body_type condition match/no-match; transmission DSG match;
pricing_log row persists base/final/applied/ctx
- M13: mechanic API board scoped to own WOs; start task on foreign
work returns 403; KPI endpoint computes 2.5/3 = 83% efficiency
across 2 done works in period
- M14: ocr_job queueable + Queue::fake assertion
- M15: tracking JSON returns ETA promised/current/reason + pending
approvals with correctly-signed approve_url; dispatcher writes
ClientNotificationLog row on workOrderReady
- M12: vehicle body_type + transmission_type round-trip through save
Suite: 269 passed (761 assertions). Was 258.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
||
|
|
0e3f9e8bca |
feat: AI model selector + i18n nav labels (RU/EN) on new modules
AI model selector:
- AiAssistantService::MODEL_DEFAULTS and MODEL_OPTIONS const tables (3 picks per
provider: Claude Opus 4.7 / Sonnet 4.6 / Haiku 4.5, OpenAI 4o / 4o-mini,
Gemini 1.5 Pro / Flash). Default upgraded from Sonnet 4.5 → Sonnet 4.6.
- modelFor(provider, company?) resolves tenant override > global default.
- All 8 hardcoded model strings replaced with modelFor() across callClaude
(chat with tool-use), callOpenAI, callGemini (chat), postClaude/postOpenAI/
postGemini (single-shot), and OcrInvoiceService.
- Settings page adds 3 model selectors per provider with persistence at
settings.ai.models.{claude,gpt,gemini}.
i18n nav labels:
- TireSet / Bodyshop / Subcontractor / SubcontractJob / PricingCoefficient /
ShopCustomer resources: getNavigationLabel / getNavigationGroup /
getModelLabel / getPluralModelLabel return __()-wrapped strings.
- 20 keys added to lang/ru.json and lang/en.json.
Tests (4 new): default model, tenant override wins, unknown provider falls
back to claude default, options dictionary contains each default key.
Full suite: 134 passed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
||
|
|
c90c35d930 |
Stage 8 — Smart Pricing Engine: contextual coefficients
Contextual multipliers layered on top of base MarkupRule pricing, applied
per work-order line based on vehicle, client and urgency.
Schema:
- pricing_coefficients (multiplier, conditions JSON, priority, stackable)
- vehicles.vehicle_class (sedan/suv/commercial/hybrid/ev/premium)
- clients.is_vip
- work_orders.urgency (normal/urgent/express)
PricingEngine::quote(Part, Vehicle?, Client?, urgency):
- base = MarkupRule on buy_price (fallback sell_price or buy×1.30)
- context: class (explicit or inferred hybrid/ev from fuel), age, vip, urgency
- stackable coefficients all multiply; non-stackable take only the highest
- returns {base, final, applied[]} breakdown
PricingCoefficient::matches(ctx) — classes/age range/vip/urgency conditions
(empty = always applies).
Filament:
- PricingCoefficientResource with condition builder (classes, age, vip, urgency)
- vehicle_class select, client is_vip toggle, WO urgency select
- "Preț inteligent" action on WO parts shows breakdown + applies sell_price
Tests (6 new):
- base-only without coefficients; age coefficient gating; VIP; express urgency;
stackable multiply vs non-stackable highest-wins; hybrid inferred from fuel
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|