Start with what you have
Add foods manually or open Scan a shelf. Choose a storage location and take or select a photo. Clear views help; partly hidden foods can be missed.
Choose your scanning method
Fast on-device scanning uses local detection and image matching. Download its embedding model in Settings first. No photo is sent to a server.
System AI is available only on supported devices. If it is unavailable, use fast scanning.
Local LLM uses a separately downloaded vision model on compatible devices with enough memory. Speed depends on your hardware.
Server connects to your own OpenAI-compatible endpoint. Configure its address, model and optional API key in Settings. This sends your selected photo to that provider. Use HTTPS for encrypted connections.
Review one crop at a time
Select a suggestion or type the actual product. Add the crop to your final review and continue with the next crop. Confirmed examples can be remembered locally to help match very similar photos.
Finish with the details
At the end, review product names, quantities and fill levels. Repeated selected products are grouped into one entry. Overlapping crops do not prove that there are multiple physical items: correct quantities before saving.
Keep your shopping list useful
Update quantities and fill levels as you use foods. Set desired amounts to plan what you need to buy.
Need a hand?
Contact sergio.alejandro.mz@gmail.com or report an issue on GitHub.