With getting bored by writing the same GeminiAPIClient more times than I can count on my fingers, I had a thought: it’s all just a wrapper. Why not make it flexible? So I built a generic, customizable wrapper powered by Gemini’s REST APIs—turning natural-language tasks into reusable AI-driven endpoints. It supports structured agents with clean JSON I/O, making automation predictable and scalable. Plus, with Gemini integrated as a service, adding AI features feels less like engineering and more like plugging in superpowers.

I always find that breaking a habit is easier when you just start, and tracking it keeps you consistent—there’s real motivation in not breaking a streak. So I built a full-stack habit tracker with Next.js, Prisma, and PostgreSQL. I started with another Gemini client, but now I’ve plugged in my Task-to-API custom agent that I can tweak on the go—powering AI insights, progress analysis, and a more dynamic, motivating experience.