HomechefAI

HomechefAI user interface - summarized recipe information (ingredients and instructions) and link to recipe for more detailed instructions
HomechefAI user interface - pre-recipe generation, after scanning ingredients
HomechefAI user interface - post-recipe generation; list of potential recipes with the scanned ingredients

Inspiration

Being a team that enjoys cooking in their free-time, a big challenge we face is determining what we can cook with the ingredients that we have. Finding recipes is also difficult, when you have so many ingredients that could or could not work together. We wanted to find a solution to this, by creating an app that could identify every ingredient you have, and select the right items to generate delicious recipes, to always know that you can cook an amazing meal.

What it does

Using this app, users are able to either manually input ingredients they have, or much more conveniently, with the snap of a photo, scan their fridge, cupboard, or counter-top full of ingredients to generate real and popular recipes.

How we built it

We wrote most of our backend in python, and our front-end was developed using C# and Unity. We connected our backend with our front end using the Python package Flask. We scraped our recipes from websites found on the DuckDuckGo search engine, and used Google's Gemini API for object recognition and recipe ingredient/instruction summaries.

Challenges we ran into

We struggled a lot with deciding how we wanted to securely store our API keys. First we went to use Google's Firebase, but we ran into more complications there. After more trial and error, we found the best solution was to use a secrets manager. Alongside this challenge, we also did not know exactly how we were going to deploy our application. Fortunately, Google had not only their own management platform that allowed us to deploy our application in a secure and stable container, but also their own secrets manager. Because of this discovery, we were committed to using Google's platform, Cloud Run, to deploy our project.

Accomplishments that we're proud of

As freshmen, most of us are new to app development. Much of our projects were simpler and smaller, such as those required of us in class projects. However, we wanted to challenge ourselves and explore our passion in computer science and app development. This being our first hackathon, we are not only proud to have a completed and functioning application, but also of our determination and ability to work under stressful situations and unfamiliar time restraints to make our imaginations become reality.

What we learned

In our project, we learned about how to connect different APIs to create a secure and well-rounded application. We also learned how to use Docker to deploy our application, since we are new to app development. Outside of what we learned in our programming, we learned much about how to cooperate as a team in a serious programming project. Having to coordinate our GitHub commits and dividing and conquering our work load, ensuring we did not have conflicts in our both our mental implementations/interpretations and actual developments in our codebase, was an extremely important learning experience.

What's next for HomechefAI

We plan to continue developing HomechefAI into a fully fledged app that we can potentially distribute across multiple mobile platforms. We also want to improve our user interface to be more user-friendly and visually appealing. We believe that this app solves a problem that many people experience on a day-to-day basis, and this app could potentially help people make healthier and more cost-effective meal decisions. Our next step is also to include restrictions on certain recipes depending on allergens or general food preferences/diets, and give user's more personalized recipes through the use of better-optimized LLMs.