Bite Theory | Devpost

key metrics on user health & preferences
use mobile camera or upload file for text extraction
user config and insight report export
user survived #n near-allergy events this week
mobile view (bottom navbar)
alt logo banner
color theme

Team ID 046577420FC3245E

Inspiration

Living in New York or really any major city, it is a big struggle to figure out what to eat at a restaurant. We thought of everyday difficulties people with certain dietary restrictions to receive accommodations. (Especially when trying out the new cuisines in a group setting -- that's super stressful.) We made Bite Theory to streamline the decision making process using agentic AI and mcp.

What it does

Unlike existing food and nutrition apps that rely on manually logging macros after meals, Bite Theory works in the moment: it detects ingredients, and delivers personalized recommendations based on each user’s dietary preferences, allergies, and evolving taste profile from initial user signup and previous scans. It will also describe whether the choice is a little adventurous or not so you can make fun choices depending on your mood. Additionally, a big part was adding allergy information so that Bite Theory can warn you about possible aggravating food/ingredients. Byte Theory also researches the ingredients in dishes from the most popular menus using mcp.

How we built it

We used react native expo for the front end. We used Yelp mcp and Open nutrition mcp to research food choices. We used gemini for transcribing the menu and for compiling all of our gathered data. We used mongo db to for accounts and to store the taste profile data which is allergies, food preferences, and previous food ratings.

Challenges we ran into

Expo caused a lot of problems with components in the frontend. We struggled with keeping the layout between web and mobile consistent and rendering certain visualizations for the most part. Also we had a crazy branching issue where our working branch was behind and ahead of the main branch at the same time which took a long time to resolve. We constantly communicated and checked in on each other's progress and the pieces started fitting

Accomplishments that we're proud of

We had a solid product at the end. I think this is a great result considering this was our first online/hybrid hackathon. Additionally, this was one of our teammates first hackathons so we are glad he had a great time during this experience. Additionally, getting the MCPs to work inside of our repository was something we had never done before. Getting fully functionally features in the MVP and also a clear statement of purpose makes us believe that we had a good idea that's highly relevant now and scalable in the future :)

What we learned

We had never used Mongodb as a relational database but that worked out by modeling our documents after the relational table structure. We learned about running MCP servers. For designing the frontend we read up on good UI/UX designs and learned that reducing user friction is more impactful than adding features. Users do not want to fill out forms manually multiple times a day per meal so we landed on implementing many visuals, minimal onboarding, and passive learning over time. Also for our sanity and code quality hackers are allowed to go to sleep.

What's next for Bite Theory

We would like to package our react native web app into a mobile app continuing MVP development. We’d love to talk with classmates and launch a small closed beta to refine the experience based on real feedback. This is kind of a reach but we hope that Beli will like us