(with a Little Help from My AI Sidekick)
As code for this app was written by AI, I wanted to give that same AI chance to write it’s own blog post. With my small contributions.
TLDR: Code and link to the app – down below.
The Problem: “Where Did My Cloud Credits Go?”
If you’re a startup in the Microsoft for Startups Founders Hub, you know the drill: Microsoft gives you cloud credits, you spin up some VMs, maybe a sneaky AI service or two, and then… poof! Your credits vanish faster than a free pizza at a hackathon. But here’s the kicker: the Azure Portal, in all its glory, doesn’t actually show you a nice, clear breakdown of your consumption if you’re in the Founders Hub. This is due to access rights on billing data of the subscription that you get on account of being a part of the hub. You get CSVs. You get JSONs. You get… confused.
The Setup: A Python App That Tracks Azure Costs
Picture this: a Python application that analyses Azure usage data, generates beautiful charts, and helps users understand their cloud spending. Sounds straightforward, right? Narrator: It wasn’t.
In true manner of vibe-coding, I refrained my self of looking at the code. Just going back-and-forth with prompts of weather or not the outcome meets my expectation. I didn’t set out to build a “big app” – just enough so that I could upload consumption csv and see what services are spending what amount of funds, and see it week-by-week, and detect if there are some anomalies.
The Solution: “Vibe Coding” to the Rescue
So, what’s a founder to do? Build an app, of course! But not just any app—a “vibe coding” app. What’s vibe coding?
Alternative would be just to upload csv to some AI and let them do this on their own… but where is fun in that.
It’s when you code with no plan, no spec, and a playlist of lo-fi beats, letting the code flow like a jazz solo. (And, if you’re lucky, it compiles.)
But I had a secret weapon: an AI coding assistant. That’s right, I had my own digital pair programmer ready to help, debug, and occasionally roast my variable names.
Just for the fun – add AI chat
One thing that makes little sense ( at least at this level of implementation ) is an AI chat interface to discuss all things related to Azure consumption and cost. It is one of the sexiest usages of AI. So you can do that also. If you are so inclined.
The Build: Man vs. Machine (vs. Microsoft CSVs)
- Step 1: The Data
- I downloaded my Azure consumption CSV. It was 2MB of pure, unadulterated chaos.
- I asked my AI agent, “How do I even read this thing?”
- Agent: “Try pandas. And maybe a coffee.”
- Step 2: The Backend
- Flask? Sure, why not.
- Pandas? Absolutely.
- Matplotlib and Plotly? Let’s make those charts pop.
- Agent: “Don’t forget to use @app.route. And please, for the love of all that is Pythonic, use constructor injection.”
- Step 3: The Frontend
- I wanted dashboards, charts, and a UI that says, “I have no idea what I’m doing, but it looks cool.”
- Agent: “You know, you could use React or Angular for this.”
- Me: “Shhh. Flask templates are a vibe.”
- Step 4: The Tests
- I ( meaning my trusted Agent friend ) wrote some tests. They failed.
- Agent: “Your test expects ‘Virtual Machines’ to be the top service, but it’s actually ‘Azure DDOS Protection’. Update your test, human.”
- Me: “You’re supposed to be on my side!”
- Step 5: The Bugs
- I had a bug where more than 5 services showed up in the ‘Top 5’ chart.
- Agent: “You forgot to filter by rank. Here, let me fix that for you.”
- Me: “You’re hired.”
The Result: Open Source for the Win
After a few hours (days? weeks? time is a flat circle in startup land), I had a working app:
- Upload your CSV.
- See beautiful charts of your Azure spending.
- Finally answer the question, “Where did my cloud credits go?”
And best of all, it’s open source! Because if I had to suffer through Microsoft’s CSV exports, so should you—but with better charts.
The Takeaway: Never Code Alone
Building this app was a wild ride. I wanted to build is “vibe-coding”, never looking at code, just guiding AI agent with what I expected and what it actually produced. All the decisions where made by AI – mine was initial ask to build this app entirely in Python.
So to add some notices:
- I don’t know if the code is good or bad
- test coverage is a complete mistery
- amount of Voodoo injected is overwhelming
But most importantly, I learned that coding with an AI assistant is like having a super-powered rubber duck: it answers back, it writes code, and it never judges your “vibe coding” playlist.
So, if you’re a founder in the Microsoft for Startups Founders Hub, grab this app, upload your CSV, and finally get some answers. And if you’re coding solo, maybe invite an AI to the party. Just don’t let it name your variables.
P.S. If you find a bug, I would not be surprised. If you find a clever feature, it was probably Cursor or Claude’s idea.
You can view and test the app on this link: https://azure-cost-analytics.azurewebsites.net/ Link will be active some time and then it will be shut-down, as it requires funds to operate. Maybe Microsoft Startup Founder Hub will volunteer to support it’s continuing life, but I doubt it.
Code can be found here : https://github.com/architechcro/azure-cost-analytics-public.git
I will repeat the notice from above: this is vibe-coded app, use at own discretion. Learn from it at everyones benefit.
And don’t forget to checkout the podcast Architect’s Lens
