Foody.ai

Eating local, seasonal, sustainable foods is good for the earth, our health and local communities. Foody.ai is an AI nutritionist creating personalized meal plans based on your goals and stats.

Comment

Inspiration: We were inspired by [the food at the black-woman-owned farmers market] and my journey finding healthy meals for my family, many of who suffer from food allergies.

I found that moving to a diet of fresh, local, seasonal ingredients made with their individual, point in time health needs in mind led to step change improvements in health and ultimately freedom from food allergies!

Meal plans need to be tailored to each person depending on age, gender, health goals but the principle of eating local, seasonal, traditional foods helps the earth and our health.

What it does

This webapp asks some basic questions about the users' demographics and dietary preferences and creates a mealplan for the user designed from local, seasonal ingredients, keeping in mind their dietary preferences. The meals in the plan are available for order as ingredients in Instacart

Future work: integrate Doordash, local farm box options and celebrity nutritionist coaches

How we built it

The front end is a flask app written in python. It calls GPT4 with a tailored prompt that takes location, timing and personal preferences into account in generating recipes for 3 meals. The output is a personalized meal plan. The user can chat back and forth to optimize it further. And then choose to buy the ingredients on Instacart using the GPT4 plugin.

We input the produce in season from the USDA website. GPT4 uses that as input to create recipes. In the future we will also have a store of celebrity (nutritionists, doctors etc). inspired recipes.

Challenges we ran into

GPT4 would give us a great answer but the format was inconsistent just slightly. Sometimes it would add a word like 'answer' or 'labels' at the beginning of the JSON throwing off our downstream integration with Instacart. Massaging and displaying the output meals and instacart was difficult and inconsistent due to this issue.

Accomplishments that we're proud of

We are proud of the design and the idea. We are happy to see how well GPT4 works with the constraints real users provide.

What we learned

We learnt a lot about prompt engineering and prompt chaining. How to make GPT4 output better and constrain it.

What's next for Foody.ai

We will be adding more authoritative data sources for input (recipes, meal timings, portions, and local seasonal ingredients). We will add additional place to buy the ingredients (local markets, family businesses) and add doordash as an option.

We will roll this app out to trusted testers and grow from there.

Built With

Share this project:

Updates