Sous

Inspiration

As students who are complete ass at cooking and rarely even know what to cook with the things we have, we wanted to make a tool, that could help us, and others (particularly students), who struggle to figure out what to cook with the things they have, and reduce decision fatigue and the overall time things take when it comes to cooking.

What it does

Sous uses AI (the Gemini API) to scan the items in your fridge, pantry and cupboards to then figure out what meals you could make, you can even sort recipes/meals by budget, time to cook, protein amount (to fit your macros and fitness goals), as well as many other filters...

How we built it

We built Sous as a web-app (primarily focusing the mobile user experience), so we decided to go with NextJS (using TypeScript as the main programming language), we used Gemini API in the backend to scan the items in the user's photos and return them to NextJS.

Challenges we ran into

We ran into challenges finding the right Gemini model that would get us both a reasonable scan time as well as accurate results. We also had to spend a while finding the right compression level and resolution to input into Gemini that would allow for a fast prompting and high quality item detection simultaneously.

Accomplishments that we're proud of

We're quite of the quality level and accuracy of our item detection system. As well as the demo / promo video we made for the project.

What we learned

The main thing we learned is to integrate the Gemini API for vision purposes into an app.

What's next for Sous

In future we'd likely add an account feature to Sous, allowing you to save your progress across multiple devices. We'd also like to add support for sharing data between roommates / people in your household.