feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
'use client';
|
|
|
|
|
|
|
|
|
|
import { useRef, useState, useEffect } from 'react';
|
2026-05-04 22:55:42 +02:00
|
|
|
import Image from 'next/image';
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
import { useRouter } from 'next/navigation';
|
|
|
|
|
import { Camera, Loader2, RotateCcw, AlertTriangle, CheckCircle2, Save } from 'lucide-react';
|
|
|
|
|
|
2026-05-04 22:55:42 +02:00
|
|
|
const LOGO_URL =
|
|
|
|
|
'https://s3.portnimara.com/images/Port%20Nimara%20New%20Logo-Circular%20Frame_250px.png';
|
|
|
|
|
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
import { Button } from '@/components/ui/button';
|
|
|
|
|
import { Input } from '@/components/ui/input';
|
|
|
|
|
import { Label } from '@/components/ui/label';
|
|
|
|
|
import { Textarea } from '@/components/ui/textarea';
|
|
|
|
|
import {
|
|
|
|
|
Select,
|
|
|
|
|
SelectContent,
|
|
|
|
|
SelectItem,
|
|
|
|
|
SelectTrigger,
|
|
|
|
|
SelectValue,
|
|
|
|
|
} from '@/components/ui/select';
|
|
|
|
|
import { useUIStore } from '@/stores/ui-store';
|
|
|
|
|
import { apiFetch } from '@/lib/api/client';
|
|
|
|
|
import { cn } from '@/lib/utils';
|
|
|
|
|
import { EXPENSE_CATEGORIES, PAYMENT_METHODS } from '@/lib/constants';
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
import { runTesseract } from '@/lib/ocr/tesseract-client';
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
|
2026-05-12 21:21:37 +02:00
|
|
|
// Lazy-loaded compression — the worker bundle isn't on the critical path,
|
|
|
|
|
// and most users won't reach this code without first granting camera/file
|
|
|
|
|
// access, by which point the module is already paged in.
|
|
|
|
|
async function compressReceiptIfHeavy(file: File): Promise<File> {
|
|
|
|
|
// Only compress raster images > ~1 MB. PDFs, SVGs, and small files pass
|
|
|
|
|
// through untouched. Magic-byte check via mime type — the caller is the
|
|
|
|
|
// file picker which trusts the picker output already.
|
|
|
|
|
if (!file.type.startsWith('image/') || file.type === 'image/svg+xml') return file;
|
|
|
|
|
if (file.size < 1 * 1024 * 1024) return file;
|
|
|
|
|
const { default: imageCompression } = await import('browser-image-compression');
|
|
|
|
|
try {
|
|
|
|
|
const compressed = await imageCompression(file, {
|
|
|
|
|
maxSizeMB: 0.5, // ~500 KB target — plenty of resolution for OCR
|
|
|
|
|
maxWidthOrHeight: 2000, // tesseract.js's sweet spot for receipt text
|
|
|
|
|
useWebWorker: true, // off the main thread; UI stays responsive
|
|
|
|
|
// Auto-rotate to EXIF orientation, strip metadata. Phones often
|
|
|
|
|
// store the rotation as EXIF rather than rotating pixels; without
|
|
|
|
|
// this the receipt comes out sideways and OCR confidence tanks.
|
|
|
|
|
preserveExif: false,
|
|
|
|
|
fileType: file.type === 'image/png' ? 'image/jpeg' : file.type,
|
|
|
|
|
initialQuality: 0.85,
|
|
|
|
|
});
|
|
|
|
|
// Browser-image-compression typings always say `File`, but in some
|
|
|
|
|
// runtimes the value comes through as a plain Blob. Belt-and-suspenders:
|
|
|
|
|
// wrap in a File so downstream FormData uses the original filename.
|
|
|
|
|
const blob = compressed as unknown as Blob;
|
|
|
|
|
if (typeof File !== 'undefined' && blob instanceof File) return blob;
|
|
|
|
|
return new File([blob], file.name, { type: blob.type });
|
|
|
|
|
} catch {
|
|
|
|
|
// Fall back to the original — we don't want a corner-case compression
|
|
|
|
|
// bug to block the user from saving an expense.
|
|
|
|
|
return file;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
// ─── Types ────────────────────────────────────────────────────────────────────
|
|
|
|
|
|
|
|
|
|
interface ParsedReceipt {
|
|
|
|
|
establishment: string | null;
|
|
|
|
|
date: string | null;
|
|
|
|
|
amount: number | null;
|
|
|
|
|
currency: string | null;
|
|
|
|
|
lineItems: Array<{ description: string; amount: number }>;
|
|
|
|
|
confidence: number;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
type ScanState =
|
|
|
|
|
| { kind: 'idle' }
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
| { kind: 'processing'; engine: 'tesseract' | 'ai' }
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
| {
|
|
|
|
|
kind: 'verify';
|
|
|
|
|
parsed: ParsedReceipt;
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
source: 'ai' | 'tesseract' | 'manual';
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
reason?: string;
|
|
|
|
|
providerError?: string;
|
|
|
|
|
}
|
|
|
|
|
| { kind: 'saving' }
|
|
|
|
|
| { kind: 'saved'; expenseId: string }
|
|
|
|
|
| { kind: 'error'; message: string };
|
|
|
|
|
|
|
|
|
|
interface ScanResp {
|
|
|
|
|
data: {
|
|
|
|
|
parsed: ParsedReceipt;
|
|
|
|
|
source: 'ai' | 'manual';
|
|
|
|
|
reason?: string;
|
|
|
|
|
provider?: string;
|
|
|
|
|
model?: string;
|
|
|
|
|
providerError?: string;
|
|
|
|
|
};
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// ─── Form ─────────────────────────────────────────────────────────────────────
|
|
|
|
|
|
|
|
|
|
interface VerifyFormProps {
|
|
|
|
|
parsed: ParsedReceipt;
|
|
|
|
|
imagePreview: string;
|
|
|
|
|
imageFile: File;
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
source: 'ai' | 'tesseract' | 'manual';
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
reason?: string;
|
|
|
|
|
providerError?: string;
|
|
|
|
|
onSubmit: (input: {
|
|
|
|
|
establishmentName: string;
|
|
|
|
|
amount: string;
|
|
|
|
|
currency: string;
|
|
|
|
|
expenseDate: string;
|
|
|
|
|
category: string;
|
|
|
|
|
paymentMethod: string;
|
|
|
|
|
description: string;
|
|
|
|
|
file: File;
|
|
|
|
|
}) => void;
|
|
|
|
|
onRetake: () => void;
|
|
|
|
|
saving: boolean;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
const TODAY = () => new Date().toISOString().slice(0, 10);
|
|
|
|
|
|
|
|
|
|
function VerifyForm({
|
|
|
|
|
parsed,
|
|
|
|
|
imagePreview,
|
|
|
|
|
imageFile,
|
|
|
|
|
source,
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
reason: _reason,
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
providerError,
|
|
|
|
|
onSubmit,
|
|
|
|
|
onRetake,
|
|
|
|
|
saving,
|
|
|
|
|
}: VerifyFormProps) {
|
|
|
|
|
const [establishmentName, setEstablishment] = useState(parsed.establishment ?? '');
|
|
|
|
|
const [amount, setAmount] = useState(parsed.amount != null ? String(parsed.amount) : '');
|
|
|
|
|
const [currency, setCurrency] = useState((parsed.currency ?? 'USD').toUpperCase());
|
|
|
|
|
const [expenseDate, setExpenseDate] = useState(parsed.date ?? TODAY());
|
|
|
|
|
const [category, setCategory] = useState<string>('other');
|
|
|
|
|
const [paymentMethod, setPaymentMethod] = useState<string>('credit_card');
|
|
|
|
|
const [description, setDescription] = useState('');
|
|
|
|
|
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
const lowConfidence = source !== 'manual' && parsed.confidence < 0.6;
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
const noOcr = source === 'manual';
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
const engineLabel = source === 'ai' ? 'AI' : source === 'tesseract' ? 'on-device OCR' : 'manual';
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
|
|
|
|
|
const banner = noOcr ? (
|
|
|
|
|
<div className="flex items-start gap-2 rounded-lg border border-amber-300 bg-amber-50 px-3 py-2 text-sm text-amber-900">
|
|
|
|
|
<AlertTriangle className="mt-0.5 h-4 w-4 shrink-0" />
|
|
|
|
|
<div>
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
<p className="font-medium">Manual entry mode</p>
|
|
|
|
|
<p className="text-xs mt-0.5">
|
|
|
|
|
{providerError
|
|
|
|
|
? `We couldn't read the receipt automatically: ${providerError}.`
|
|
|
|
|
: "We couldn't read the receipt automatically."}{' '}
|
|
|
|
|
Fill in the details below to save the expense with the photo attached.
|
|
|
|
|
</p>
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
</div>
|
|
|
|
|
</div>
|
|
|
|
|
) : lowConfidence ? (
|
|
|
|
|
<div className="flex items-start gap-2 rounded-lg border border-amber-300 bg-amber-50 px-3 py-2 text-sm text-amber-900">
|
|
|
|
|
<AlertTriangle className="mt-0.5 h-4 w-4 shrink-0" />
|
|
|
|
|
<div>
|
2026-05-04 22:55:42 +02:00
|
|
|
<p className="font-medium">Low-confidence read - please double-check the fields</p>
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
<p className="text-xs mt-0.5">
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
{engineLabel} returned {Math.round(parsed.confidence * 100)}% confidence.
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
</p>
|
|
|
|
|
</div>
|
|
|
|
|
</div>
|
|
|
|
|
) : (
|
|
|
|
|
<div className="flex items-start gap-2 rounded-lg border border-emerald-300 bg-emerald-50 px-3 py-2 text-sm text-emerald-900">
|
|
|
|
|
<CheckCircle2 className="mt-0.5 h-4 w-4 shrink-0" />
|
|
|
|
|
<div>
|
2026-05-04 22:55:42 +02:00
|
|
|
<p className="font-medium">Receipt parsed - confirm the fields and save</p>
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
<p className="text-xs mt-0.5">
|
|
|
|
|
{engineLabel} · {Math.round(parsed.confidence * 100)}% confidence.
|
|
|
|
|
</p>
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
</div>
|
|
|
|
|
</div>
|
|
|
|
|
);
|
|
|
|
|
|
|
|
|
|
return (
|
|
|
|
|
<form
|
|
|
|
|
className="flex flex-col gap-4"
|
|
|
|
|
onSubmit={(e) => {
|
|
|
|
|
e.preventDefault();
|
|
|
|
|
onSubmit({
|
|
|
|
|
establishmentName,
|
|
|
|
|
amount,
|
|
|
|
|
currency,
|
|
|
|
|
expenseDate,
|
|
|
|
|
category,
|
|
|
|
|
paymentMethod,
|
|
|
|
|
description,
|
|
|
|
|
file: imageFile,
|
|
|
|
|
});
|
|
|
|
|
}}
|
|
|
|
|
>
|
|
|
|
|
{banner}
|
|
|
|
|
|
|
|
|
|
<div className="overflow-hidden rounded-lg border border-border">
|
|
|
|
|
{/* eslint-disable-next-line @next/next/no-img-element */}
|
|
|
|
|
<img
|
|
|
|
|
src={imagePreview}
|
|
|
|
|
alt="Receipt preview"
|
|
|
|
|
className="block w-full max-h-[40vh] object-contain bg-muted"
|
|
|
|
|
/>
|
|
|
|
|
</div>
|
|
|
|
|
|
|
|
|
|
<div className="grid grid-cols-1 gap-3 sm:grid-cols-2">
|
|
|
|
|
<div className="space-y-1.5 sm:col-span-2">
|
|
|
|
|
<Label htmlFor="establishmentName">Vendor / establishment</Label>
|
|
|
|
|
<Input
|
|
|
|
|
id="establishmentName"
|
|
|
|
|
value={establishmentName}
|
|
|
|
|
onChange={(e) => setEstablishment(e.target.value)}
|
|
|
|
|
placeholder="e.g. Marina Fuel Station"
|
|
|
|
|
/>
|
|
|
|
|
</div>
|
|
|
|
|
<div className="space-y-1.5">
|
|
|
|
|
<Label htmlFor="amount">Amount</Label>
|
|
|
|
|
<Input
|
|
|
|
|
id="amount"
|
|
|
|
|
type="number"
|
|
|
|
|
step="0.01"
|
|
|
|
|
min="0"
|
|
|
|
|
inputMode="decimal"
|
|
|
|
|
value={amount}
|
|
|
|
|
onChange={(e) => setAmount(e.target.value)}
|
|
|
|
|
required
|
|
|
|
|
/>
|
|
|
|
|
</div>
|
|
|
|
|
<div className="space-y-1.5">
|
|
|
|
|
<Label htmlFor="currency">Currency</Label>
|
|
|
|
|
<Input
|
|
|
|
|
id="currency"
|
|
|
|
|
value={currency}
|
|
|
|
|
onChange={(e) => setCurrency(e.target.value.toUpperCase())}
|
|
|
|
|
maxLength={3}
|
|
|
|
|
required
|
|
|
|
|
/>
|
|
|
|
|
</div>
|
|
|
|
|
<div className="space-y-1.5">
|
|
|
|
|
<Label htmlFor="expenseDate">Date</Label>
|
|
|
|
|
<Input
|
|
|
|
|
id="expenseDate"
|
|
|
|
|
type="date"
|
|
|
|
|
value={expenseDate}
|
|
|
|
|
onChange={(e) => setExpenseDate(e.target.value)}
|
|
|
|
|
required
|
|
|
|
|
/>
|
|
|
|
|
</div>
|
|
|
|
|
<div className="space-y-1.5">
|
|
|
|
|
<Label htmlFor="category">Category</Label>
|
|
|
|
|
<Select value={category} onValueChange={setCategory}>
|
|
|
|
|
<SelectTrigger id="category">
|
|
|
|
|
<SelectValue />
|
|
|
|
|
</SelectTrigger>
|
|
|
|
|
<SelectContent>
|
|
|
|
|
{EXPENSE_CATEGORIES.map((c) => (
|
|
|
|
|
<SelectItem key={c} value={c} className="capitalize">
|
|
|
|
|
{c.replace(/_/g, ' ')}
|
|
|
|
|
</SelectItem>
|
|
|
|
|
))}
|
|
|
|
|
</SelectContent>
|
|
|
|
|
</Select>
|
|
|
|
|
</div>
|
|
|
|
|
<div className="space-y-1.5">
|
|
|
|
|
<Label htmlFor="paymentMethod">Payment method</Label>
|
|
|
|
|
<Select value={paymentMethod} onValueChange={setPaymentMethod}>
|
|
|
|
|
<SelectTrigger id="paymentMethod">
|
|
|
|
|
<SelectValue />
|
|
|
|
|
</SelectTrigger>
|
|
|
|
|
<SelectContent>
|
|
|
|
|
{PAYMENT_METHODS.map((p) => (
|
|
|
|
|
<SelectItem key={p} value={p} className="capitalize">
|
|
|
|
|
{p.replace(/_/g, ' ')}
|
|
|
|
|
</SelectItem>
|
|
|
|
|
))}
|
|
|
|
|
</SelectContent>
|
|
|
|
|
</Select>
|
|
|
|
|
</div>
|
|
|
|
|
<div className="space-y-1.5 sm:col-span-2">
|
|
|
|
|
<Label htmlFor="description">Notes (optional)</Label>
|
|
|
|
|
<Textarea
|
|
|
|
|
id="description"
|
|
|
|
|
value={description}
|
|
|
|
|
onChange={(e) => setDescription(e.target.value)}
|
|
|
|
|
rows={2}
|
|
|
|
|
/>
|
|
|
|
|
</div>
|
|
|
|
|
</div>
|
|
|
|
|
|
|
|
|
|
<div className="flex flex-col gap-2 sm:flex-row">
|
|
|
|
|
<Button
|
|
|
|
|
type="submit"
|
|
|
|
|
disabled={saving || !amount}
|
|
|
|
|
className="h-12 text-base sm:flex-1"
|
|
|
|
|
data-testid="scan-save"
|
|
|
|
|
>
|
|
|
|
|
{saving ? (
|
|
|
|
|
<Loader2 className="mr-2 h-4 w-4 animate-spin" />
|
|
|
|
|
) : (
|
|
|
|
|
<Save className="mr-2 h-4 w-4" />
|
|
|
|
|
)}
|
|
|
|
|
Save expense
|
|
|
|
|
</Button>
|
|
|
|
|
<Button
|
|
|
|
|
type="button"
|
|
|
|
|
variant="outline"
|
|
|
|
|
onClick={onRetake}
|
|
|
|
|
disabled={saving}
|
|
|
|
|
className="h-12 text-base"
|
|
|
|
|
>
|
|
|
|
|
<RotateCcw className="mr-2 h-4 w-4" />
|
|
|
|
|
Retake
|
|
|
|
|
</Button>
|
|
|
|
|
</div>
|
|
|
|
|
</form>
|
|
|
|
|
);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// ─── Shell ────────────────────────────────────────────────────────────────────
|
|
|
|
|
|
|
|
|
|
export function ScanShell() {
|
|
|
|
|
const router = useRouter();
|
|
|
|
|
const portSlug = useUIStore((s) => s.currentPortSlug);
|
|
|
|
|
const fileRef = useRef<HTMLInputElement>(null);
|
|
|
|
|
const [state, setState] = useState<ScanState>({ kind: 'idle' });
|
|
|
|
|
const [imagePreview, setImagePreview] = useState<string | null>(null);
|
|
|
|
|
|
|
|
|
|
// Revoke blob URL on unmount.
|
|
|
|
|
useEffect(() => {
|
|
|
|
|
return () => {
|
|
|
|
|
if (imagePreview) URL.revokeObjectURL(imagePreview);
|
|
|
|
|
};
|
|
|
|
|
}, [imagePreview]);
|
|
|
|
|
|
2026-05-12 21:21:37 +02:00
|
|
|
async function handleFile(rawFile: File) {
|
|
|
|
|
// Compress oversized phone photos to ~500 KB in a WebWorker BEFORE
|
|
|
|
|
// we hand the bytes to tesseract or the server. Receipts from 12MP
|
|
|
|
|
// cameras are usually 4-8 MB; this drops them to ~250-500 KB without
|
|
|
|
|
// visible quality loss for text OCR. Mobile bandwidth + the server's
|
|
|
|
|
// sharp pipeline both benefit.
|
|
|
|
|
const file = await compressReceiptIfHeavy(rawFile);
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
if (imagePreview) URL.revokeObjectURL(imagePreview);
|
|
|
|
|
setImagePreview(URL.createObjectURL(file));
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
setState({ kind: 'processing', engine: 'tesseract' });
|
|
|
|
|
|
2026-05-04 22:55:42 +02:00
|
|
|
// Always run Tesseract first - it's free, on-device, and gives us a
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
// baseline parse we can fall back to if the optional AI pass is off
|
|
|
|
|
// or fails. The WASM bundle dynamic-imports inside `runTesseract`.
|
|
|
|
|
let tesseract: Awaited<ReturnType<typeof runTesseract>> | null = null;
|
|
|
|
|
try {
|
|
|
|
|
tesseract = await runTesseract(file);
|
|
|
|
|
} catch (err) {
|
2026-05-04 22:55:42 +02:00
|
|
|
// Tesseract.js itself failed (corrupt image, OOM, etc). Don't bail -
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
// give the user the manual form so they can still save the expense.
|
|
|
|
|
setState({
|
|
|
|
|
kind: 'verify',
|
|
|
|
|
parsed: {
|
|
|
|
|
establishment: null,
|
|
|
|
|
date: null,
|
|
|
|
|
amount: null,
|
|
|
|
|
currency: null,
|
|
|
|
|
lineItems: [],
|
|
|
|
|
confidence: 0,
|
|
|
|
|
},
|
|
|
|
|
source: 'manual',
|
|
|
|
|
reason: 'tesseract-error',
|
|
|
|
|
providerError: err instanceof Error ? err.message : 'On-device OCR failed',
|
|
|
|
|
});
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Now ask the server whether AI is enabled for this port. If it is,
|
|
|
|
|
// the server runs the configured provider and returns a richer parse;
|
|
|
|
|
// otherwise we keep the Tesseract result.
|
|
|
|
|
setState({ kind: 'processing', engine: 'ai' });
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
try {
|
|
|
|
|
const fd = new FormData();
|
|
|
|
|
fd.append('file', file);
|
|
|
|
|
const portId = useUIStore.getState().currentPortId;
|
|
|
|
|
const headers = new Headers();
|
|
|
|
|
if (portId) headers.set('X-Port-Id', portId);
|
|
|
|
|
const res = await fetch('/api/v1/expenses/scan-receipt', {
|
|
|
|
|
method: 'POST',
|
|
|
|
|
body: fd,
|
|
|
|
|
credentials: 'include',
|
|
|
|
|
headers,
|
|
|
|
|
});
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
if (!res.ok) throw new Error(`Server returned ${res.status}`);
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
const body = (await res.json()) as ScanResp;
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
|
|
|
|
|
if (body.data.source === 'ai' && body.data.parsed.confidence >= tesseract.parsed.confidence) {
|
2026-05-04 22:55:42 +02:00
|
|
|
// AI did at least as well as Tesseract - prefer its result.
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
setState({
|
|
|
|
|
kind: 'verify',
|
|
|
|
|
parsed: body.data.parsed,
|
|
|
|
|
source: 'ai',
|
|
|
|
|
reason: body.data.reason,
|
|
|
|
|
providerError: body.data.providerError,
|
|
|
|
|
});
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Either AI is disabled (`source: 'manual', reason: 'ai-disabled'`),
|
2026-05-04 22:55:42 +02:00
|
|
|
// not configured, or it underperformed - fall back to Tesseract.
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
setState({
|
|
|
|
|
kind: 'verify',
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
parsed: tesseract.parsed,
|
|
|
|
|
source: 'tesseract',
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
reason: body.data.reason,
|
|
|
|
|
providerError: body.data.providerError,
|
|
|
|
|
});
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
} catch {
|
2026-05-04 22:55:42 +02:00
|
|
|
// Server unreachable - still let the user verify with the Tesseract
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
// result and save the expense. We don't surface the network error
|
|
|
|
|
// because the local parse is usable.
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
setState({
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
kind: 'verify',
|
|
|
|
|
parsed: tesseract.parsed,
|
|
|
|
|
source: 'tesseract',
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
});
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
async function handleSubmit(input: {
|
|
|
|
|
establishmentName: string;
|
|
|
|
|
amount: string;
|
|
|
|
|
currency: string;
|
|
|
|
|
expenseDate: string;
|
|
|
|
|
category: string;
|
|
|
|
|
paymentMethod: string;
|
|
|
|
|
description: string;
|
|
|
|
|
file: File;
|
|
|
|
|
}) {
|
|
|
|
|
setState({ kind: 'saving' });
|
|
|
|
|
try {
|
2026-05-04 22:55:42 +02:00
|
|
|
// Upload the image (multipart - apiFetch wraps JSON, so use raw fetch).
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
const fd = new FormData();
|
|
|
|
|
fd.append('file', input.file);
|
|
|
|
|
fd.append('category', 'receipt');
|
|
|
|
|
const portId = useUIStore.getState().currentPortId;
|
|
|
|
|
const headers = new Headers();
|
|
|
|
|
if (portId) headers.set('X-Port-Id', portId);
|
|
|
|
|
const upRes = await fetch('/api/v1/files/upload', {
|
|
|
|
|
method: 'POST',
|
|
|
|
|
body: fd,
|
|
|
|
|
credentials: 'include',
|
|
|
|
|
headers,
|
|
|
|
|
});
|
|
|
|
|
if (!upRes.ok) throw new Error(`Upload failed: ${upRes.status}`);
|
|
|
|
|
const upJson = (await upRes.json()) as { data: { id: string } };
|
|
|
|
|
|
|
|
|
|
const expense = await apiFetch<{ data: { id: string } }>(`/api/v1/expenses`, {
|
|
|
|
|
method: 'POST',
|
|
|
|
|
body: {
|
|
|
|
|
establishmentName: input.establishmentName || undefined,
|
|
|
|
|
amount: input.amount,
|
|
|
|
|
currency: input.currency,
|
|
|
|
|
expenseDate: input.expenseDate,
|
|
|
|
|
category: input.category,
|
|
|
|
|
paymentMethod: input.paymentMethod,
|
|
|
|
|
description: input.description || undefined,
|
|
|
|
|
receiptFileIds: [upJson.data.id],
|
|
|
|
|
paymentStatus: 'unpaid',
|
|
|
|
|
},
|
|
|
|
|
});
|
|
|
|
|
|
|
|
|
|
setState({ kind: 'saved', expenseId: expense.data.id });
|
|
|
|
|
} catch (err) {
|
|
|
|
|
setState({
|
|
|
|
|
kind: 'error',
|
|
|
|
|
message: err instanceof Error ? err.message : 'Save failed',
|
|
|
|
|
});
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
function reset() {
|
|
|
|
|
if (imagePreview) {
|
|
|
|
|
URL.revokeObjectURL(imagePreview);
|
|
|
|
|
setImagePreview(null);
|
|
|
|
|
}
|
|
|
|
|
setState({ kind: 'idle' });
|
|
|
|
|
if (fileRef.current) fileRef.current.value = '';
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (
|
audit: Tier 1/3/6/7 batch — PII redaction, mobile safe-area, perf, build hardening
Tier 1.4: error_events.request_body_excerpt sanitizer now redacts
GDPR-relevant fields (email, phone, dob, address, fullName, firstName,
lastName, postcode, nationalId, etc.) on top of the existing
credential list. A 5xx in /api/v1/clients no longer lands full client
PII in the super-admin inspector.
Tier 3.10: ScanShell <main> now adds pb-[max(1.5rem, env(safe-area-
inset-bottom))]. Mobile-pwa audit caught the Save expense button sitting
flush against the iPhone 14/15 home indicator in standalone PWA mode.
Tier 6.2: dashboard widget-registry now dynamic-imports every
recharts-backed chart widget (berth status, lead source, occupancy
timeline, pipeline funnel, revenue breakdown, source conversion).
~80-150KB initial-bundle savings when reps have charts disabled.
ssr:false because recharts needs window.
Tier 6.3: DataTable wraps the assembled columns in useMemo keyed on
(columns, hasBulkActions). TanStack docs explicitly warn that
rebuilding columns every render resets the table's internal state.
Tier 7.1: Added .dockerignore (was missing — 7.6 GB context with
.env reachable via COPY . .). Excludes git, env files, node_modules,
build artefacts, IDE config, test artefacts, audit docs.
Tier 7.4: Dockerfile.dev now runs as the node user (uid 1000) — was
root. Working dir moves to /home/node/app.
Tier 7.5: docker-compose.prod.yml adds memory limits (2g postgres,
512m redis, 1g crm-app, 1g crm-worker) and json-file log rotation
(max-size, max-file) to every service.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-12 17:18:35 +02:00
|
|
|
<main
|
|
|
|
|
// pb-[max(1.5rem,env(safe-area-inset-bottom))] — mobile-pwa-auditor
|
|
|
|
|
// caught that the "Save expense" button was sitting flush against
|
|
|
|
|
// the home indicator on iPhone 14/15 in standalone PWA mode
|
|
|
|
|
// (viewportFit:cover + statusBarStyle:default exposes the safe-
|
|
|
|
|
// area inset, but the original `py-6` ignored it).
|
2026-05-12 22:14:38 +02:00
|
|
|
className="mx-auto flex min-h-dvh w-full max-w-xl flex-col gap-4 px-4 py-6 pb-[max(1.5rem,env(safe-area-inset-bottom))] sm:py-10"
|
audit: Tier 1/3/6/7 batch — PII redaction, mobile safe-area, perf, build hardening
Tier 1.4: error_events.request_body_excerpt sanitizer now redacts
GDPR-relevant fields (email, phone, dob, address, fullName, firstName,
lastName, postcode, nationalId, etc.) on top of the existing
credential list. A 5xx in /api/v1/clients no longer lands full client
PII in the super-admin inspector.
Tier 3.10: ScanShell <main> now adds pb-[max(1.5rem, env(safe-area-
inset-bottom))]. Mobile-pwa audit caught the Save expense button sitting
flush against the iPhone 14/15 home indicator in standalone PWA mode.
Tier 6.2: dashboard widget-registry now dynamic-imports every
recharts-backed chart widget (berth status, lead source, occupancy
timeline, pipeline funnel, revenue breakdown, source conversion).
~80-150KB initial-bundle savings when reps have charts disabled.
ssr:false because recharts needs window.
Tier 6.3: DataTable wraps the assembled columns in useMemo keyed on
(columns, hasBulkActions). TanStack docs explicitly warn that
rebuilding columns every render resets the table's internal state.
Tier 7.1: Added .dockerignore (was missing — 7.6 GB context with
.env reachable via COPY . .). Excludes git, env files, node_modules,
build artefacts, IDE config, test artefacts, audit docs.
Tier 7.4: Dockerfile.dev now runs as the node user (uid 1000) — was
root. Working dir moves to /home/node/app.
Tier 7.5: docker-compose.prod.yml adds memory limits (2g postgres,
512m redis, 1g crm-app, 1g crm-worker) and json-file log rotation
(max-size, max-file) to every service.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-12 17:18:35 +02:00
|
|
|
>
|
2026-05-04 22:55:42 +02:00
|
|
|
{/* Brand header - logo centered, page title underneath. Establishes
|
|
|
|
|
the standalone identity (this is the PWA home for the scanner). */}
|
|
|
|
|
<header className="flex flex-col items-center gap-3">
|
|
|
|
|
<Image
|
|
|
|
|
src={LOGO_URL}
|
|
|
|
|
alt="Port Nimara"
|
|
|
|
|
width={64}
|
|
|
|
|
height={64}
|
|
|
|
|
className="rounded-full shadow-md"
|
|
|
|
|
priority
|
|
|
|
|
unoptimized
|
|
|
|
|
/>
|
|
|
|
|
<div className="flex w-full items-center justify-between">
|
|
|
|
|
<h1 className="text-xl font-semibold">Scan a receipt</h1>
|
|
|
|
|
{state.kind !== 'idle' ? (
|
|
|
|
|
<Button variant="ghost" size="sm" onClick={reset}>
|
|
|
|
|
Start over
|
|
|
|
|
</Button>
|
|
|
|
|
) : null}
|
|
|
|
|
</div>
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
</header>
|
|
|
|
|
|
|
|
|
|
{state.kind === 'idle' ? (
|
|
|
|
|
<section className="flex flex-1 flex-col items-center justify-center gap-4 rounded-2xl border-2 border-dashed border-border bg-muted/20 p-8 text-center">
|
|
|
|
|
<Camera className="h-12 w-12 text-muted-foreground/60" aria-hidden />
|
|
|
|
|
<div>
|
|
|
|
|
<p className="text-base font-medium">Tap to capture a receipt</p>
|
|
|
|
|
<p className="mt-1 text-xs text-muted-foreground">
|
|
|
|
|
Use your camera or pick an image from your library. We'll read it and pre-fill
|
|
|
|
|
the form for you to confirm.
|
|
|
|
|
</p>
|
|
|
|
|
</div>
|
|
|
|
|
<Button
|
|
|
|
|
type="button"
|
|
|
|
|
className="h-12 px-6 text-base"
|
|
|
|
|
onClick={() => fileRef.current?.click()}
|
|
|
|
|
data-testid="scan-capture"
|
|
|
|
|
>
|
|
|
|
|
<Camera className="mr-2 h-5 w-5" />
|
|
|
|
|
Capture receipt
|
|
|
|
|
</Button>
|
|
|
|
|
<input
|
|
|
|
|
ref={fileRef}
|
|
|
|
|
type="file"
|
|
|
|
|
accept="image/*"
|
|
|
|
|
capture="environment"
|
|
|
|
|
className="hidden"
|
|
|
|
|
onChange={(e) => {
|
|
|
|
|
const f = e.target.files?.[0];
|
|
|
|
|
if (f) void handleFile(f);
|
|
|
|
|
}}
|
|
|
|
|
/>
|
|
|
|
|
</section>
|
|
|
|
|
) : null}
|
|
|
|
|
|
|
|
|
|
{state.kind === 'processing' ? (
|
|
|
|
|
<section className="flex flex-1 flex-col items-center justify-center gap-3 py-12">
|
|
|
|
|
<Loader2 className="h-10 w-10 animate-spin text-brand" />
|
feat(ocr): Tesseract.js as default scanner, AI as opt-in per port
The mobile receipt scanner now runs Tesseract.js in-browser by default —
on-device, free, and image bytes never leave the device. AI providers
(OpenAI / Claude) become a per-port opt-in for higher accuracy on
hard-to-read receipts.
- Lazy-load Tesseract WASM in src/lib/ocr/tesseract-client.ts (5 MB
bundle dynamic-imports on first scan, not in main chunk)
- Heuristic parser src/lib/ocr/parse-receipt-text.ts extracts vendor,
date, amount, currency, and line items from raw OCR text
- New port-scoped aiEnabled flag on OcrConfig (defaults false). Resolved
flag never inherits from the global row — each port admin opts in
independently
- Scan endpoint short-circuits to manual-mode when aiEnabled=false so
the AI provider is never invoked unless the admin has flipped the
switch
- Scan UI runs Tesseract first, then asks the server whether AI is
enabled — uses the AI result only when its confidence beats Tesseract;
network failures degrade gracefully to the local parse
- Admin OCR-settings form gains the per-port aiEnabled checkbox
Tests: 756/756 vitest (was 747) — +7 parser unit tests, +2 aiEnabled
config tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:46:29 +02:00
|
|
|
<p className="text-sm text-muted-foreground">
|
|
|
|
|
{state.engine === 'tesseract' ? 'Reading on-device…' : 'Refining with AI…'}
|
|
|
|
|
</p>
|
feat(phase-b): ship analytics dashboard, alerts, scanner PWA, dedup, audit view
Phase B (Insights & Alerts) PR4-11 in one drop. Builds on the schema +
service skeletons committed in PRs 1-3.
PR4 Analytics dashboard — 4 chart types (funnel/timeline/breakdown/source),
date-range picker (today/7d/30d/90d), CSV+PNG export per card.
PR5 Alert rail UI + /alerts page — topbar bell w/ live count, dashboard
right-rail, three-tab page (active/dismissed/resolved), socket-driven
invalidation. Bell lazy-loads list on popover open to keep cold pages
fast in non-dashboard routes.
PR6 EOI queue tab on documents hub — filters to in-flight EOIs, count
surfaces in tab label.
PR7 Interests-by-berth tab on berth detail — replaces the stub.
PR8 Expense duplicate detection — BullMQ job runs scan on create, yellow
banner on detail w/ Merge / Not-a-duplicate, transactional merge
consolidates receipts and archives the source.
PR9 Receipt scanner PWA + multi-provider AI — port-scoped /scan route in
its own (scanner) group with no dashboard chrome, dynamic per-port
manifest, OpenAI + Claude provider abstraction, admin OCR settings
page (port-level + super-admin global default w/ opt-in fallback),
test-connection endpoint, manual-entry fallback when no key is
configured. Verify form always shown before save — no ghost rows.
PR10 Audit log read view — swap to tsvector full-text search on the
existing GIN index, cursor pagination, filters for entity/action/user
/date range, batched actor-email resolution.
PR11 Real-API tests — opt-in receipt-ocr.spec (admin save+test, optional
real-receipt parse via REALAPI_RECEIPT_FIXTURE) and alert-engine
socket-fanout spec gated behind RUN_ALERT_ENGINE_REALAPI. Both skip
cleanly without their gate envs so CI stays green.
Test totals: vitest 690 -> 713, smoke 130 -> 138, realapi +2 opt-in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 17:21:55 +02:00
|
|
|
</section>
|
|
|
|
|
) : null}
|
|
|
|
|
|
|
|
|
|
{state.kind === 'verify' && imagePreview ? (
|
|
|
|
|
<VerifyForm
|
|
|
|
|
parsed={state.parsed}
|
|
|
|
|
imagePreview={imagePreview}
|
|
|
|
|
imageFile={fileRef.current?.files?.[0] as File}
|
|
|
|
|
source={state.source}
|
|
|
|
|
reason={state.reason}
|
|
|
|
|
providerError={state.providerError}
|
|
|
|
|
onSubmit={handleSubmit}
|
|
|
|
|
onRetake={reset}
|
|
|
|
|
saving={false}
|
|
|
|
|
/>
|
|
|
|
|
) : null}
|
|
|
|
|
|
|
|
|
|
{state.kind === 'saving' ? (
|
|
|
|
|
<section className="flex flex-1 flex-col items-center justify-center gap-3 py-12">
|
|
|
|
|
<Loader2 className="h-10 w-10 animate-spin text-brand" />
|
|
|
|
|
<p className="text-sm text-muted-foreground">Saving expense…</p>
|
|
|
|
|
</section>
|
|
|
|
|
) : null}
|
|
|
|
|
|
|
|
|
|
{state.kind === 'saved' ? (
|
|
|
|
|
<section className="flex flex-1 flex-col items-center justify-center gap-3 rounded-2xl border border-emerald-200 bg-emerald-50 p-8 text-center">
|
|
|
|
|
<CheckCircle2 className="h-12 w-12 text-emerald-600" />
|
|
|
|
|
<p className="text-base font-semibold text-emerald-900">Expense saved</p>
|
|
|
|
|
<div className="flex gap-2">
|
|
|
|
|
<Button onClick={reset} variant="outline" data-testid="scan-another">
|
|
|
|
|
Scan another
|
|
|
|
|
</Button>
|
|
|
|
|
<Button
|
|
|
|
|
onClick={() => router.push(`/${portSlug}/expenses/${state.expenseId}` as never)}
|
|
|
|
|
>
|
|
|
|
|
View expense
|
|
|
|
|
</Button>
|
|
|
|
|
</div>
|
|
|
|
|
</section>
|
|
|
|
|
) : null}
|
|
|
|
|
|
|
|
|
|
{state.kind === 'error' ? (
|
|
|
|
|
<section
|
|
|
|
|
className={cn(
|
|
|
|
|
'flex flex-col items-center gap-3 rounded-2xl border border-destructive/30 bg-destructive/5 p-6 text-center',
|
|
|
|
|
)}
|
|
|
|
|
>
|
|
|
|
|
<AlertTriangle className="h-10 w-10 text-destructive" />
|
|
|
|
|
<p className="text-base font-medium text-destructive">{state.message}</p>
|
|
|
|
|
<Button onClick={reset} variant="outline">
|
|
|
|
|
Try again
|
|
|
|
|
</Button>
|
|
|
|
|
</section>
|
|
|
|
|
) : null}
|
|
|
|
|
</main>
|
|
|
|
|
);
|
|
|
|
|
}
|