• 

  FRIENDS

• 

  FRIEND OWING

• 

  SCAN

• 

  RECEIPT HISTORY

• 

  PROFILE 2

• 

  ASSIGN ITEM

• 

  HOME

• 

  BREAKDOWN

• 

  PROFILE

• 

  RYT INSIGHT 1

• 

  RECEIPT

• 

  WRAP 4

• 

  WRAP 1

• 

  MAP

• 

  WRAP 3

• 

  WRAP 2

• 

  WRAP 5

• 

  RYT INSIGHT 2

• 

  MAP

RytSplit — AI-Powered Receipt Scanning & Group Expense Splits

Inspiration

Splitting group bills is a chore. People forget who paid, receipts get lost, and travel expenses pile up across different currencies. We wanted a single, mobile-first experience that handles scanning, splitting, and settling—all without manual data entry.

---

What It Does

• AI Receipt Scanning using OpenAI Vision to extract:

  • Line items
  • Totals
  • Tax
  • Tips

• Flexible Splitting

  • Split evenly or by individual items
  • Assign shares to friends
  • Track who owes whom

• Running Ledger

  • Transaction history
  • Payment requests
  • Friend management

• Smart Mapping

  • Maps receipts/venues using Leaflet + OpenStreetMap

• Multi-language Support

  • English / 中文 / Bahasa Melayu

---

How We Built It

Frontend

• React + Vite + Tailwind (mobile-first)
• Capacitor for Android/iOS builds
• Smooth UX with Vite HMR

Backend

• FastAPI (Python)
• SQLite database
• JWT authentication

• REST endpoints for:

  • Auth
  • Receipts
  • Friends
  • Splits
  • Transactions

AI & OCR

• OpenAI GPT-4 Vision for parsing messy receipts
• OpenStreetMap + Leaflet for venue mapping

Tooling

• .env for API keys/secrets
• Batching API calls for responsive UI

---

Challenges

• Emulator networking headaches (e.g., localhost vs 10.0.2.2).
• Prompting Vision reliably to parse inconsistent receipts.
• Maintaining shared state across auth, receipts, friends, and splits.
• Handling multi-language UI and currency/tax rounding edge cases.

---

Accomplishments

• Fully working flow: Scan → Parse → Split → Assign → Ledger on both web and emulator.
• Clean, mobile-first UI aligned with the Ryt Bank design language.
• AI-driven item-level splitting (not just manual entry).
• Built-in multi-language support and location tagging.

---

What We Learned

• Prompt engineering significantly boosts OCR accuracy.
• Clear backend API boundaries simplify React state management.
• Emulator networking can waste hours—document IP rules early.
• UX polish (prefilled splits, toasts, defaults) removes friction.

---

What’s Next

• Integrate real payment rails: Stripe, PayPal, Venmo, etc.
• Push notifications for pending splits/requests.
• Budgeting insights by merchant/category with export to CSV/PDF.
• Enhanced multi-currency logic + offline caching for travelers.
• More automated testing around OCR edge cases and rounding rules.

Built With

• android
• capacitor
• capacitor-(android/ios-packaging).-backend:-python-(fastapi)
• cli;
• frontend:-react-+-vite
• jwt-auth
• sqlite-database.-apis/services:-openai-gpt?4-vision-for-receipt-ocr;-leaflet-+-openstreetmap-for-mapping;-optional-google-maps-api-key-for-locations.-tooling/build:-npm
• studio
• tailwind-css
• vite-hmr-dev-server
• xcode