Fridge to List

Fridge to List turns photos of your fridge into a grocery list, so you don't have to stand there with the door open writing things down. Upload one to four photos of shelves, the door or the pantry; a local vision model reports only what it can actually see, each item with a rough quantity and a confidence level, plus a separate “could not tell” list for anything blurry. The app compares that against your own list of usual staples and builds an editable checklist you can copy or download as a text file.

The privacy story is the point: Gemma 3 (4B) runs on your own machine through Ollama, photos are resized in the browser and held in memory only, and nothing is sent to a third party — no account, no API key, no cloud bill. A “what the model actually saw” panel shows every item with its confidence so you can verify and fix the list before shopping, and if Ollama isn't running you get a clear error with the setup commands instead of sample data dressed up as AI output.

Vibe-coded evidence, in the author's own words on DEV (25 October 2026): “I built the app with an AI coding agent (Arena) from a single prompt; I did not write the code by hand. What I did myself was set it up and test it: installed Ollama, pulled gemma3:4b, merged the agent's branch into main, ran the app, and tried it on real photos.” GitHub stars: 0 at the 28 October 2026 check (github.com/anshraj3060/fridge-to-list; 4 commits, created 22 October 2026).

Honest limits: Fridge to List is not a hosted app and has no online demo — it runs locally (Node.js server on localhost plus Ollama), so the repository page is the demo link and there is nothing to try in the browser here. “Missing” is an inference, not an observation: a photo can never show what is gone, so the list is your staples minus what the model saw, and the UI says so out loud. A 4B model misidentifies things, quantities are impressions not measurements, and on CPU-only machines a run can take tens of seconds to a couple of minutes.