Spending Tracker

What I Built

I built a Smart Spending Tracker that completely eliminates the dreaded chore of manual budgeting. Let’s face it: manually entering expenses into spreadsheets—picking categories, inputting dates, and typing out exact amounts—is tedious. It’s exactly the kind of task we procrastinate on.

To channel my inner sloth, I created a tracker that rewards laziness. Instead of navigating menus, you just brain-dump your expenses in plain text (e.g., "Spent 15 bucks on a burrito for lunch and 40 on gas"). The underlying open-source AI instantly parses your messy input, extracts the amount, assigns the correct category, dates it, and logs it for you. You get to be incredibly lazy, but your finances stay perfectly organized.

Code

Mahasvan / SpendingTracker

A personal macOS spending tracker with an AI chatbot that reads and modifies your finances using natural language.

Features

Dashboard - income/expense summary, spending trends, top categories, budget progress, payment method distribution, recent transactions, largest expense, and month-over-month comparisons
Transactions — full CRUD with list view, inline rows, and editor sheet.
Advanced Search — filter by text, date range, type, amount, and category/payment-method chips
Budget Tracking - per-category monthly budgets
AI Chatbot - Currently Supports Ollama and Apple Intelligence. Add, update, delete transactions; search data; generate summaries - all in natural language (e.g., "Add ₹250 lunch paid by UPI"). Supports visual inputs, using LLM Vision capabilities, or using the macOS Vision API
Category & Payment Methods - customize names, colors, icons; archive unused items
Backup & Restore - JSON backup with merge or replace modes

Tech Stack

Swift 6 · SwiftUI · SwiftData · Swift Charts · swift-markdown-ui…

View on GitHub

How I Built It

My project is built around open-weight models and local inference.

Instead of writing complex Regex to parse human text, I integrated an open-source AI model [Note: e.g., Llama 3.2 / Mistral running locally via Ollama / Transformers.js] as the core engine. When the user submits a messy string of text, the prompt is passed to the local model which is instructed to extract the item_name, amount, and category, returning it purely as structured JSON.

The frontend/backend [Note: mention your framework, e.g., Next.js, Python/Flask, etc.] captures this JSON and instantly logs it to the database. The AI does all the heavy lifting of categorizing and structuring the data, so the user doesn't have to.

Why Does Open Innovation Matter?

When it comes to personal finance, privacy is everything. Open innovation is the only reason a project like this can exist securely.

Because this project uses open-weight models and local inference, it can run entirely offline on a laptop. Users don't have to send their daily habits, sensitive financial data, or location history to a closed-source, third-party API server (like OpenAI) that they don't control. Their financial data stays strictly on their own machine. Furthermore, because it relies on open-source AI, there are zero subscription fees to keep the automated tracking running. It’s private, free, and completely under the user's control.

The Lazy Payoff

I usually put off budgeting until the end of the month, at which point I'm forced to sift through dozens of bank statements and receipts—a chore that usually takes over an hour.

By building this tracker, I've turned a dreaded monthly chore into a 5-second brain dump whenever I buy something. It saves me hours of tedious spreadsheet management every month. More importantly, because it requires almost zero friction or mental energy, I actually stick to tracking my money now. I let the AI do the work so I can get back to relaxing.

DEV Community

Spending Tracker - Hacktoberfest Week 3

What I Built

Demo

Code

Mahasvan / SpendingTracker