I built WaterShortcut as a minimalist web app to help homeowners and renters analyze their water bills, reduce usage, and save money. Behind the scenes, it uses a combination of Google Document AI for PDF parsing and GPT-4o mini to generate concise recommendations on potential water savings. The site’s front end runs serverless (with Cloudflare Workers) and relies on a pipeline that processes user-uploaded PDF bills, extracts text via Document AI, and then feeds that text into GPT-4o mini to produce actionable water-saving tips.

I tried to make this process as frictionless as possible—users just upload a recent water bill, get an instant breakdown, and see targeted suggestions for fixture upgrades, leak detection, or usage cutbacks. The system addresses real-world concerns (like typical daily shower durations or the cost per gallon in various regions) while respecting file-size constraints and letting the AI handle nuance.

I’m eager for feedback on how to improve, refine, or expand. Thanks to ChatGPT and Cloudflare, I'm coding and building web apps for the first time in ten years.

Please poke holes. Let me know if the GPT-4o mini–driven analysis is clear or if there are additional features you’d like to see. All thoughts, critiques, and suggestions are most welcome. Thank you for taking a look!

Full disclosure: I am experimenting to see if Google Adsense will cover the cost of using the API.