Split the Bill Calculator

Snap a photo of a restaurant receipt, let an AI vision model read the line items, tick off who had what, and get a clean per-person breakdown - including a fair share of the tip and tax.

How it works

1. Upload a receipt photo (JPEG or PNG).
2. The Flask backend sends the image to a vision LLM (qwen-omni-turbo via an OpenAI-compatible API), which extracts the store, date, line items, subtotal, tax, and total. (we chose this api because it has a free tier)
3. Add the people at the table and check who shared each item. Items checked by multiple people are split evenly between them.
4. Pick a tip percentage and how to split the tip (proportional to what each person ordered, or evenly), and everyone gets their own receipt-style total.

Demo

Parsed from the sample receipt in reciepts/aandklas_test.jpeg:

Assign items with checkboxes — shared items (like the nachos below) are split evenly between everyone who ticked them. Missing items can be added manually, and wrong ones deleted:

Each person gets their own mini receipt with their items, tip share, tax share, and total:

Features

• Receipt image upload with live preview and AI-powered parsing
• Item assignment via checkboxes, with even splitting of shared items
• Tip calculation — split proportionally to item cost or evenly per person
• Tax read from the receipt and included in the split
• Manual item entry and deletion for parser misses
• Per-person receipt-style breakdown
• South African Rand (R) currency display
• Rate-limited parsing endpoint (10 requests/minute) with image validation

Project structure

File	Purpose
index.html	UI layout and styling
webapp.js	Frontend logic: upload, assignment, and split calculation
rest_server.py	Flask API, rate limiting, and static file serving
image_extraction.py	Receipt parsing via the OCR/LLM backend
reciepts/	Sample receipt image for testing

Getting started

Requirements

• Python 3.10+
• An API key for the vision model backend

Setup

Install dependencies:

pip install -r requirements.txt

Create a .env file in the project root with your API key:

API_KEY=your-api-key-here

Run

Start the Flask server:

python rest_server.py

Then open http://127.0.0.1:5000 in your browser. Try it with the sample receipt at reciepts/aandklas_test.jpeg.

API

POST /parse-receipt

Accepts a multipart form upload with a receipt image field and returns the extracted data:

{
  "store": "AANDKLAS",
  "date": "07/06/2026",
  "items": [
    { "name": "Slow Boat", "price": 42.0 },
    { "name": "Nachos Half", "price": 80.0 },
    ...
  ],
  "subtotal": 278.27,
  "tax": 41.73,
  "total": 320.0
}

Uploads are validated as real images before parsing, and the endpoint is rate-limited to 10 requests per minute.

Notes

• Add at least one person before assigning items — new items default to the first person.
• Items checked by multiple people are split evenly between them.
• Tax is read from the receipt and split using the same method as the tip.

Roadmap for the future

• Optional alternative OCR backend
• Better receipt parsing accuracy (especially taxes)
• Add support for other currencies and adding automatically adding tax to the total if applicable
• Export or print individual receipts