SV
← Back to code
— Fullstack Development

Household Harmony

Private Beta

A budgeting app for couples and households, built to replace the fragile spreadsheet my wife and I outgrew. Plan a shared budget, forecast months ahead, and let an LLM sort credit-card statements into categories - with a Swedish tax engine and everything sensitive encrypted on your own device. In private beta.

FastAPI React PostgreSQL Web Crypto API TypeScript Supabase RLS Auth JWT Tailwind CSS shadcn/ui React Query Vite Lucide Icons AI Document Parsing pdfplumber Python
Live App ↗ GitHub ↗
~ when Nov 2025 → Ongoing
~ status Active
~ kind Personal Project

My wife and I ran our household on a spreadsheet: estimated income, the fixed costs we knew about, and category budgets we'd worked out by averaging a year of receipts (groceries, mostly). To look ahead, we'd duplicate the sheet, label it a future month, and fill in what we already knew was coming - a stretch with less income, a big one-off, that kind of thing.

It did the job, but it was fragile, awkward to share, and every change meant babysitting formulas. Household Harmony is what I built to replace it: a budgeting app for couples and households, in private beta - the same budget-and-forecast habit, but maintainable, shareable, and private by default, with everything sensitive encrypted on your own device.

Core Features

  • Client-side encryption - two-tier DEK/KEK system with AES-256-GCM + PBKDF2 (100k iterations). Keys derived from user passwords, held in memory only, never persisted to disk. Auto-lock after 30 min inactivity
  • Swedish tax engine - 2025 bracket calculations (32%/52% progressive), annual prognosis with refund/owing estimates
  • LLM invoice parsing - PDF credit card statements parsed via user-provided API keys (Groq, Gemini), with merchant category learning and hash-based caching
  • Financial month awareness - configurable billing cycles (default 25th-24th) as an alternative to calendar months
  • Encrypted demo mode - ephemeral accounts with pre-populated Swedish household data, rate-limited creation

Architecture

  • Frontend: React 18 + TypeScript, Vite, Tailwind CSS with shadcn/ui, React Context API, TanStack React Query
  • Backend: FastAPI (Python 3.11+) with Supabase Auth (JWT)
  • Database: PostgreSQL via Supabase with Row Level Security
  • Deployment: Vercel (frontend) + Railway (backend)
Core Focus AI Augmented Infrastructure & APIFrontend & UICore LogicAgentic Logic
About this chart
Each axis is a functional pillar; the orange area is where my focus went, and the purple how much of that work was AI-augmented. That AI layer is where tools sped up implementation - architecture, code review, and the quality bar stay mine. I treat AI as a precision tool with strict conventions, not auto-pilot.