What's For Dinner — AI meal suggestions from your pantry
Snap a photo of your pantry, let on-device ML label the ingredients, and get personalised meal suggestions back from an LLM. Includes a grocery list, weekly planner, suggestion history, and a Google-Maps powered nearby-shop finder.
The problem to solve
People stare into a full pantry and still order takeout. The friction isn't lack of ingredients — it's the mental load of picking a meal that uses what's already on the shelf. We needed an app that turned a pantry photo into a ready-to-cook plan.
What we built
A Flutter app that captures a pantry photo, runs Google ML Kit image labeling on device to extract ingredients, and posts the labels to an OpenRouter-hosted Claude 3 Sonnet model that returns structured recipe suggestions. Each user has a pantry, grocery list, weekly planner, and history stored in Firestore. A nearby-shop screen uses Google Places + Geolocator to pinpoint grocery stores within walking distance of the missing ingredients.
The impact
Pantry-to-meal in under 10 seconds · On-device ML for offline ingredient detection · LLM recipes tailored to what's actually on the shelf · Built end-to-end in 5 weeks.
