Data Dishes

Don't know what to eat at home? Just enter in the ingredients you have at home and get inspiration!

Comment

Inspiration

As college students, we do not have a lot of time to cook fancy meals. This means that our culinary skills are below average. With the help of Data Dishes, we can enter in what ingredients we have and it will give us recipes to make.

What it does

Data Dishes takes in ingredients in the form of text and it generates recipes with all the details needed to make it including ingredients, cooking and prep time, and the general instructions.

How we built it

We used HTML and CSS for the webpage while JS for the overall functionality. The JS implements a local Ollama 3.2 model that we have in our devices. This means that we do not need to have API calls but in turn in only works for devices with Ollama. (We would use API keys but we do not want to use any credits).

Challenges we ran into

We have never used a local AI model for our projects. This was hard to call. The first version of Data Dishes uses images instead of text. The problem with this was that it would take too much RAM and was also very time consuming. Since this was so inefficient, we decided that text was better than images, although we will use images later on. We could use API keys as a work around which we plan to in the future.

Accomplishments that we're proud of

We are proud of implement a local AI into our project. This was really cool as we do not need to have API calls and technically, this could run without use of the Internet. The recipes that are generated using the Ollama model and prompt were also pretty accurate and I could actually make some of these recipes.

What we learned

We learned how to call local AIs as well as installing them onto our devices. I think it was really cool to use AIs that are running on our devices since it removes the dependance of the internet.

What's next for Data Dishes

Later on, we will use API keys to make the AI run more efficiently. We will also integrate imagery and text which will make it easier for the user to use. We could also make an app version of this which uses a better camera and is mobile.

Built With

Share this project:

Updates