An AI-powered companion app for Thermomix users that generates recipes based on whatever ingredients you already have in your pantry. Solved the classic 'fridge full, nothing to cook' problem by wiring OpenAI to a curated Thermomix recipe engine.
What Is it?
Thermomix owners face a recurring frustration: they have a powerful machine and a stocked pantry, but no idea what to make. Cookbooks and official apps require you to already know the recipe — they don't meet you where you are. ThermoRecipes flips that model. You input whatever ingredients you have on hand, and the app uses OpenAI to generate step-by-step Thermomix-native recipes tailored to exactly what's available. No substitutions, no missing ingredients, no wasted food.
The Problem
Existing Thermomix recipe resources are browse-first — you find a recipe you like, then go shop for its ingredients. ThermoRecipes inverts this entirely. The real user problem is: 'I have these 8 things in my kitchen, what can I actually make right now?' No existing tool answered that question specifically for Thermomix cooking.
Tech Decisions
Vue.js and Tailwind CSS were chosen for rapid UI iteration and a clean, food-focused visual design. Supabase handles user authentication and stores saved recipes with minimal backend overhead. OpenAI API calls are proxied through Supabase Edge Functions — keeping the API key server-side, enabling rate limiting, and giving full control over the prompt before it ever reaches the model. The prompts themselves are carefully engineered to constrain outputs to valid Thermomix parameters (speed, temperature, bowl size), ensuring the generated recipes are actually usable on the device.
Results
ThermoRecipes delivers usable, realistic recipes from arbitrary ingredient lists in seconds. The prompt engineering to constrain OpenAI outputs to valid Thermomix parameters (speed, temperature, bowl size) was the hardest technical challenge — and getting that right made the difference between a novelty and a genuinely useful kitchen tool.