This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle. Freelancers are constantly collecting receipts from email, camera rolls, and wallets. By the time tax season arrives, they have to remember what each purchase was, which client it belonged to, and whether it was already recorded.
I built PocketLedger, a private, local-first expense tracker for freelancers.
PocketLedger lets users upload a receipt image or paste receipt text. Local AI extracts the merchant, date, amount, currency, category, and description. Before anything is saved, the user can review and edit every field, assign a client or project, and mark the expense as business deductible.
The app also includes:
PocketLedger is designed for independent designers, developers, photographers, consultants, and other solo business owners who want a simpler alternative to heavyweight accounting software and cloud-based receipt tools.
The product idea is a one-time paid desktop setup with optional onboarding and support. My next step is to test the workflow with a real freelancer and learn which parts of receipt capture, client tracking, and export are most valuable.
PocketLedger runs locally at:
The demo flow:
Because PocketLedger is designed around privacy, it is intentionally a local application rather than a hosted demo.
GitHub repository:
https://github.com/SaikatDash/sideHustlers PocketLedger is built with Python's standard library, SQLite, HTML, CSS, and vanilla JavaScript.
For receipt understanding, I used Qwen2.5-VL 3B through Ollama. The application sends the receipt text or image to the local Ollama installation with a structured JSON schema. The model returns a draft containing fields such as: The draft is displayed in an editable form so that the user remains in control. An expense is not added to the ledger until the user explicitly saves it.
The app does not require a Python package installation. Once Python, Ollama, and the vision model are installed, PocketLedger can run locally and continue working without a hosted AI API.
Receipt images are processed in memory and discarded after extraction. Only the fields approved by the user are saved to SQLite.
Receipts contain sensitive information about spending habits, vendors, dates, and sometimes addresses or payment details. Sending every receipt to a cloud AI service creates privacy concerns and may introduce an ongoing cost for every scan.
Using an open-weight model locally makes PocketLedger possible without requiring users to upload their financial documents: Open innovation also gives users more control over the technology behind their workflow. They are not locked into one hosted provider, pricing model, or opaque extraction system.
The trade-off is that users need enough local storage and hardware to run the model, and AI-generated results still need to be reviewed. That trade-off is acceptable for a privacy-focused tool where the user approves every expense before it is recorded.
I have not yet completed a formal first-user trial. The next validation step is to watch a freelancer use PocketLedger with real receipts and measure:
Future improvements could include:
PocketLedger is a small experiment in making everyday bookkeeping more private, understandable, and manageable for solo freelancers.