FinMan VS Socket for Python

FinMan's answer

Python with FastAPI and async SQLAlchemy on MySQL for the backend, Vue 3 with TypeScript for the web app (installable as a PWA), and aiogram plus the WhatsApp Business API for the two bots, which share one core. Statement parsing and photo recognition run on LLMs through OpenRouter, with model selection handled as data — a daily sync, health checks and automatic failover — rather than hardcoded model names.

FinMan's answer

It feeds itself. Instead of typing transactions, you drop in the statement file your bank already exports and AI writes a reusable parsing template for that bank's format — no bank login, no Plaid, no per-bank connector, duplicates skiThree reasons. It works where bank aggregators don't reach — any bank that can export a statement is supported, including EU banks and cards Plaid never covered, and you never hand over bank credentials. It covers the parts other trackers ignore: utility meters per address with per-service tariffs, fuel receipts, odometer readings and real car cost per km. And it is genuinely built for a household — one shared family group, everyone logging from their own phone, without bill-splitting mechanics. The free tier covers all of that; paying only raises AI limits.pped by a stable transaction ID. And instead of typing meter readings or fuel receipts, you photograph them and send the photo to a Telegram or WhatsApp bot, which extracts the values, books them against the right address and tariff, and flags implausible readings for confirmation.

FinMan's answer

Households that track money seriously and are tired of manual entry: couples and families sharing one budget, people whose banks aren't covered by aggregators (EU especially), drivers who want real cost per km instead of a guess, and small landlords who bill tenants for submetered utilities — tenants photograph their own meters, no tenant account needed. Ex-Mint users looking for a free replacement are a natural fit.

FinMan's answer

It started as a Perl script I wrote for my own household years ago, because every finance app expected me to type in what my bank and my meters already knew. Over the years it grew into what my family actually uses every day: statements parsed by AI, meter and receipt photos sent to a bot, car costs tracked automatically. I turned it into a product only after it had been running our own household for years — so the features exist because we needed them, not because a roadmap called for them.