Hi HN! I've tried many budgeting apps in the past, but I've never been able to build the habit to rigorously manage and update my expenses as they came in. Typically I would try something out, would fail miserably, give up and then live in guilt that I do not know where my money was going.

I built OpenArgentum for people like me - who want to check in on their finances once every few weeks - and see where their money is going, if they're trending correctly, and so on.

With OpenArgentum, you upload whatever statements or transaction exports you have (it supports pdfs, csvs, zip files). It then uses an (configurable) LLM model to correctly classify and tag your expenses as well as identify duplicates, potential transfers (you don't want to double count your credit card payment in both your chequing and your credit card statements inflating your category/income/expense numbers) and much more.

Since we're dealing with transactions in bulk (once every few weeks), Aurelia, the chat assistant, allows you to mine through them, categorize and manage them, and even report on them with charts and graphs generated on the fly. So this allows you workflows such as "I travelled to Japan in the first week of may. Identify all transactions during that period that seem travel related and add it to my japan project. Also, how much did I spend on food there?".

I wanted something that was self-hosted and that I controlled. So OpenArgentum is completely self-hosted. It currently uses the Gemini model for it's AI features - local model support is a top roadmap item.

To explore, just clone the repo and run `./start.sh --demo` to boot into a synthetic dataset. Add a Google AI studio API key if you want to try out the AI features (there's an example statement you could upload).

I'll be around for the next few hours - happy to get torn apart :)