Receipt-first budget tracker with OCR-driven parsing. Built with Next.js (React) and a local OCR model.

• Node.js 18+ and npm
• Local OCR model running at http://127.0.0.1:1234

npm install
npm run dev

Create .env.local if needed:

OCR_BASE_URL=http://127.0.0.1:1234
OCR_MODEL=nanonets-ocr2-3b

Data is stored in data/db.json (single-user).

• OCR-powered receipt scanning using a local model (image upload, preview, item parsing).
• Receipt editor: update store/date/currency/payment method, edit line items and totals, recalculate totals, and save.
• Edit mode preserves the scanned image and OCR data; you can re-run OCR without losing draft changes and decide whether to apply the new result.
• Dashboard with filters (month, category, store), KPI cards, trend chart, category breakdown, and savings rate.
• Receipts list with store filter, totals overview, inline items preview, edit, and delete.
• Items catalog with search and store filter, plus per-item price history tracking.
• Item detail view with price trend sparkline and price history list.
• REST API endpoints for OCR, receipts CRUD, items list/detail, stores list, and stats summary.
• Localization (EN/UK).

• Improve overall design and visuals (charts should respect currency formatting; dashboard charts should scale to average values instead of max-only scaling).
• Add Google authentication to persist user data and progress.
• Enable sharing receipts with other users (family mode) so invited users can add/edit receipts and affect the dashboard.
• Add shopping lists with multiple lists (e.g., “weekly groceries”, “missing at home”) and autocomplete from previously purchased items; if multiple store variants exist, use a shared name and average price; show total list cost with “+” when some items lack prices. Maybe add where is the lowest price.
• Improve OCR for electronic and paper receipts: strengthen text parsing, item classification, and JSON schema; handle extra promotions/discounts; add tests with sample receipts; validate on large receipts; support multi-image receipts; allow user-pasted receipt text (image uses OCR model; text uses LLM or VLM with a different prompt).
• MOBILE VERSION SUPPORT AND MOBILE APPLICATION!!!!!