Angalia: AI Nutrition Lookup - Showcase Project

Angalia tells you what is in your food. Snap a photo and a vision AI model identifies the dish, or just type or say what you ate, like "2 slices of pepperoni pizza". Angalia then returns a full nutrition-facts panel from Wolfram Alpha. Photo, text or voice input: voice is transcribed on the server with Whisper, in any of 99 languages, and you confirm the text before looking it up Accounts: email/password on the web, or no-password guest sessions in the installable phone app (PWA). Guests can upgrade to a full account later without losing anything Built-in usage model: everyone gets 5 free nutrition lookups a day; users can add their own free Wolfram Alpha App ID to remove the cap Pluggable food detection: a hosted vision model by default, the Gemini API as a cloud fallback, or a Food-101 classifier that runs entirely on the server React + Tailwind front end with dark mode, Express API, Supabase auth and Postgres, and server, frontend and live-model test suites All API keys stay on the server. Buyer note: the live deployment uses a university-hosted vision model and Wolfram Alpha's non-commercial free tier. To run it commercially you would switch to the Gemini or on-server detector and buy a commercial Wolfram key; both are configuration changes.