C.O.O.K.I.E.

This image shows the results of printing a recipe as a PDF.
This image shows the basic UI for the program.
This image shows a example recipe.
This image shows examples of generated recipe ideas.

Inspiration

Approximately 60% of Americans report avoiding at least one type of nutritional component in their diets. This often means people are forced to give up some of their favorite meals due to dietary restrictions. Each member of our team either has one of these restrictions personally or has family members who do, so this project hits close to home for all of us.

What it does

Our website helps users rediscover their favorite meals by using Google’s Gemini to generate personalized recipes. The AI takes the user's dietary restrictions and substitutes ingredients accordingly. It outputs three unique recipes, each with step-by-step instructions, required tools, estimated prep time, and number of servings. Our application also can export the recipe to a PDF view, which the user can then print for future use or electronic distribution.

How we built it

To create the outputs, we utilized Google’s Gemini to generate a recipe based on a custom template we provided. We then fine-tuned the prompt to make sure the responses were as accurate and helpful as possible. After generating the recipe, we created an additional prompt to generate images that best matched the dish. We used Node.js, HTML, and CSS to build an interactive webpage designed to be user-friendly.

Challenges we ran into

This was the first hackathon for the majority of our team, so even getting GitHub set up to sync properly across devices was a challenge. On the technical side, we struggled to balance the quality of AI responses with reasonable loading times. It was a learning curve, but a valuable one.

Accomplishments that we're proud of

We didn’t just get the AI working--we got it to generate realistic food images to go with the recipes. We’re also proud of the stylish, intuitive UI we built and how seriously we took recipe accuracy, especially since food allergies can be serious or even life-threatening. We added a print-friendly format so users can easily bring the recipe into the kitchen.

What we learned

We learned how to integrate AI into both the functionality and design aspects of our app. Beyond the technical knowledge, we also grew a lot in teamwork, communication, and in our use of development tools like GitHub and Devpost.

What's next for test

We want to improve our design so it works seamlessly on all screen sizes, especially mobile devices. We’re also exploring ways to reduce AI token usage and cut down on generation time, while still maintaining the quality and reliability of the recipe results.