Finance App
From a receipt to a reviewed record.
Make everyday financial records easier to capture, organize and review, including a reviewable path from an image to suggested transactions.

Captured from the development build
How it works
- 01
Capture
The user selects or captures an image; the app prepares a bounded JPEG payload.
- 02
Extract a suggestion
The provider returns structured candidate transactions from the image.
- 03
Review the fields
The user inspects amount, date, category and merchant before committing entries.
- 04
Confirm & save
Selected candidates become local transaction records.
The product idea
Let AI suggest. Let the person confirm.
A receipt can reduce typing, but its extracted fields should remain reviewable before they become financial records.
- Automation
Turn the image into a draft
Prepare the receipt image and parse candidate transactions through a configured vision model.
- Human control
Put review before saving
Let the person inspect and change dates, amounts and categories before accepting a transaction.
- Data integrity
Keep confirmed records separate
Store accepted transactions in the app’s ledger, separating model output from the saved financial record.
Independent projectIndependent earlier finance app with a reviewable image-to-ledger flow.
Stack
SwiftUI · SwiftDataDomain/repository layersPhotosUI · UIKit image resizingVision-capable LLM APIStructured transaction JSONStoreKit 2Keychain · LocalAuthentication
Implementation details
An independent earlier finance app. Receipt recognition uses a vision-model route.
- Receipt image
- Start with an image or a manual transaction.
- Review interface
- Keep suggested fields editable before saving.
- Receipt service
- Prepare the image and parse candidate transactions.
- Human confirmation
- Review date, amount and category before accepting.
- Saved ledger
- Persist confirmed records in the app’s own store.
- Vision model
- Read the image into structured suggestions through configured access.