VAULT LEDGER

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

VaultLedger AI is a local-first bookkeeping and invoice sorter for freelancers. It keeps their financial data on their own device.

Freelancers handle sensitive data: client names, payment amounts, tax details, bank references. Most AI bookkeeping tools ask them to upload all of it to someone else’s server. VaultLedger AI does the opposite. You drop in invoices and receipts, and an open-weight model reads them, sorts them into categories (income, expenses, tax-deductible items, client-wise), and builds a clean ledger without your documents leaving your machine.

Who would pay for it: solo freelancers, designers, developers, writers, and consultants who want AI help with their books but can’t or won’t hand client financials to a third-party cloud. It could be a one-time purchase or a small monthly plan.

The problem it solves:

Invoices pile up in inboxes and folders, and sorting them by hand costs billable hours.
Cloud tools create a privacy risk, and sometimes a contractual one when client NDAs are involved.
Spreadsheet-based tracking is tedious and easy to get wrong at tax time.

It also has a 3D interactive “vault” interface in a black-and-gold theme. The vault shows that your data is secured and makes the app feel premium rather than like another dull accounting tool.

Demo

🔗 Live: https://hacktoberfestweak2.vercel.app/

Code

https://github.com/mitrasayak623-crypto/Hacktoberfestweak2
How I Built It

The AI at the core

An open-weight model, [model name, and how you run it], handles invoice parsing, field extraction (vendor, date, amount, tax), and categorisation.
The app is local-first: documents and the ledger stay on the user’s device.
Gemini is an optional layer for users who want faster responses and accept the trade-off. The core sorting works on the open-weight model alone.

How the project was built

Frontend: I designed the interface in Google Stitch and built it out in Google AI Studio, including the black-and-gold theme and the 3D interactive vault.
Backend: I built the backend logic with Claude, including the invoice-processing pipeline and ledger handling.
Research: I used Perplexity to research the open-weight model options, freelancer bookkeeping needs, and invoice and tax categories.
Deployment: Vercel.

Flow: upload → extract text and fields → model classifies → ledger updates → vault view summarises totals.

Why Does Open Innovation Matter?

A bookkeeping tool is only as trustworthy as its handling of your data. With a closed API, every invoice you sort goes through someone else’s servers, under terms you can’t audit and can’t change.

Open-weight models made the central promise of this project possible: your financial data never has to leave your device. A closed API wouldn’t allow that. It also means:

1. Privacy you can verify: the model runs where you can inspect and control it.
2. No per-call costs eating freelancer margins: local inference doesn’t bill per invoice.
3. No lock-in: if a better open model arrives, you swap it in without rebuilding the product. 4.Trust for client-sensitive work: freelancers under NDAs can use it without breaking agreements.