Active Context: Corporate View
Full-Stack Financial Architecture / Next.js 15 PWA
DebTracker is a modern web application built for type safety, performance, and serverless scalability. I engineered this platform to compete directly with industry incumbents like Splitwise by offering a "Purist's Choice" - respecting user time and data by eliminating ad interruptions, artificial transaction limits, and "monetization by friction".
To ensure the absolute integrity of financial data, I deployed a strict Zero-Trust, Server-Authoritative write model. In this architecture, the client SDK is entirely read-only. Every financial mutation is routed through Next.js Server Actions, centralizing validation and enforcing complex business logic natively on the server. By implementing a snapshot strategy and denormalizing data at write-time, I eradicated the N+1 query problem, replacing iterative fetching with highly efficient collection group queries.
Code is useless if the business logic is flawed. DebTracker operates under three unbreakable systemic protocols:
Demo visualization uses deterministic iteration so you can track the extra cent. Production logic uses the Fisher-Yates randomization shown above.
The system adheres to the Law of Conservation of Money. Instead of standard division, it utilizes a randomized distribution (Fisher-Yates Shuffle) to handle algorithmic remainders, ensuring that not a single cent is ever lost or created out of thin air when splitting complex bills.
A zero-trust approach to debt clearance. When a debtor claims to have paid, the transaction is marked solely as pending. The debt remains visible in the system balances until the Creditor explicitly confirms receipt of the funds.
A systemic blockade preventing users from abandoning groups or deleting their accounts if their calculated liability is not exactly zero (with a 0.01 floating-point tolerance).
export function distributeAmount(totalAmount: number, numberOfSplits: number): number[] {
if (numberOfSplits <= 0) return [];
if (numberOfSplits === 1) return [Number(totalAmount.toFixed(2))];
// Work in cents to avoid precision loss
const cleanTotal = Math.round(totalAmount * 100);
// Integer division + remainder
const baseAmountInt = Math.floor(cleanTotal / numberOfSplits);
const remainderInt = cleanTotal % numberOfSplits;
// Base amount for all participants
const splits: number[] = new Array(numberOfSplits).fill(baseAmountInt);
// Participant indexes
const indices = Array.from({ length: numberOfSplits }, (_, i) => i);
// Fisher-Yates: randomize who gets the extra cents
for (let i = indices.length - 1; i > 0; i--) {
const j = Math.floor(Math.random() * (i + 1));
[indices[i], indices[j]] = [indices[j], indices[i]];
}
// Assign +1 cent to the first "remainderInt" participants from random order
for (let i = 0; i < remainderInt; i++) {
splits[indices[i]] += 1;
}
// Convert back to currency units
return splits.map((s) => s / 100);
}Beyond database architecture, DebTracker is built as a first-class citizen on mobile devices. By leveraging the Web Share Target API, I engineered a service worker interception protocol that allows users to seamlessly share receipt images directly from their native OS gallery into the application's file storage, bypassing traditional browser upload friction.
