About

How this project was built, the technology behind it, and our commitment to a low-carbon web.

About This Project

FitThali Recipes is an experimental project exploring the intersection of AI content generation and static site architecture. It is built with:

Eleventy (11ty) — static site generator
Tailwind CSS — utility-first styling
AI — gemini-3-flash-preview for recipe text and gemini-2.5-flash-image for image generation
Vercel — hosting and deployment

This project is inspired by recipe.polente.de and built for educational and demonstration purposes. It is not intended as a commercial recipe platform or a substitute for professional culinary or nutritional guidance.

Carbon Footprint

New recipe generation has been paused

To reduce the CO₂ footprint of this project, the automated daily pipeline that generated new recipes and AI images has been switched off. Every recipe currently on the site remains published and will continue to be served as-is.

Even though each individual generation is small, running it on a schedule adds up: large language and image models consume meaningful energy per call, and a static archive of existing recipes is more than enough to serve readers. Pausing generation is the simplest, most honest step in line with the rest of this site's low-carbon design.

We care about the environmental impact of the web. This site is designed to be as lightweight as possible.

Static-first architecture. FitThali is a fully static site — no server-side rendering on each request, no database queries, no always-on application server. Pages are pre-built at deploy time and served as plain HTML files from a CDN edge node closest to you. This means near-zero compute energy per page view.

Minimal page weight. Each page is optimised for performance: small HTML documents, utility CSS with unused styles purged at build time, and no client-side JavaScript frameworks. A lighter page means fewer bytes transferred, which translates directly to less energy used by networks and devices.

Green hosting. The site is deployed on Vercel, which runs on renewable energy-backed infrastructure. Serving from edge locations also reduces the physical distance data travels, lowering network energy consumption.

AI generation trade-off. While the content was generated using AI models — which do consume energy during the generation process — that cost was incurred once at build time, not on every visitor request. The resulting static pages have an extremely low ongoing carbon cost compared to dynamic, server-rendered applications. To take that further, automated generation of new recipes has now been paused, so no additional model energy is being spent.

You can test any page on Website Carbon Calculator to see its estimated carbon footprint per visit.

Last updated: June 22, 2026