all writing

writing / june 4, 2025

open source finance app + open source AI

how & why i integrated local ai into open source finance app

maybe finance with a local ai assistant

what did I do?

since early this year, i was using a open source finance app called actual. it was great in budgeting and help me stay on track with my finances. then I stumbled into this new open source finance app called maybe finance. it's a personal finance app built in rust. it’s designed for people who want to take control of their financial data without relying on closed platforms. it has a modern ui, cool visualizations and yeah, ai.

what's cool about maybe finance?

  • spending visualization: tracks your income and expenses with clean, insightful charts and graphs.

  • budgeting: helps you set spending limits and monitor financial goals.

  • ai assistant: uses a third-party ai - gpt - to answer questions about your finances and provide insights.

these features make maybe a powerful and privacy conscious alternative to mainstream finance apps. but while the app itself is open, the ai assistant relies on a third-party provider. that's where I had resistant in using it. raising questions about data control and long-term trust. thus, my journey toward local ai began.

why I made this effort?

i love maybe finance. it’s open-source, modern, clean. and the built-in third party ai? I ain't gonna use it because I have trust issues.

here’s why:

  • my data is gold — financial info is personal. it says more about you than your search history ever could.

  • policies don’t equal trust, especially from big tech — companies say “we don’t store your data.” but how would you ever know? and even if they don’t today, what about tomorrow?

  • local = control, with ollama and open models like qwen and gemma, i can run the same kind of assistant locally. no data leaves my machine. no silent logging. no terms to agree to. it’s mine.

local ai isn’t just a nerdy experiment — it’s an opportunity.

for privacy. for ownership. for trust.

quick detour into history

there's always a lesson to be observed on any subject when we back into the history.

back in the day…

search engines rolled out in the early 2000s. totally "free". answered everything.

we loved it. we used it. a lot. and over time? they started collecting more. and more. and more.

I dont remember when I felt like I was dropped into an ocean of ads because it didn't happen overtime, it happened gradually. the influence of ads in a human is extremely impactful which is exactly why everyone's spending significant amount of resources into it.

and naturally ai is next.

ai is awesome. no doubt!

my worry is we repeating the same mistake again - UNAWARE of the consequences.

thus the effort of integrating local ai into a modern finance app.

how did I do it?

took lot of help from cursor. I underestimated that it would 1 weekend thing, ended up being 3 weekends. main challenge was with getting the local models to perform effective function callings. finally was able to make it work with qwen2.5 and gemma3.

I dont think I'll open a pull request to the original maybe finance project because I think I changed lot of their code to make it work with ollama. if any of you want to my version of maybe finance with your ollama models, dm me and I ll push my code to github and share it with you.

closing thoughts…

the point of me sharing all this is because I want to start a discussion on what I discussed here - privacy, search & ai, opportunities provided by local ai. I hope to connect with like minded ppl and ultimately we build something that reflects our values.