Nourish.ai

Inspiration

My inspiration came from having friends and family with certain medical conditions that restrict what diet they can have, such as lactose intolerance, diabetes, etc

What it does

It acts as an LLM and AI agent for searching through the web, looking for food within a certain range from an address, where users must include dietary restrictions/medical conditions in their profile. Then the app, when gathering the data, will categorize the dishes from safest to riskiest through a "policy condition engine" and output the safest dishes.

How we built it

Frontend: Next.js/Typescript Backend: Python/FastAPI DataBase: SupaBase/pgvector, Redis Cache/Celery APIs: Browser Use, Google Gemini, Google Maps Autocomplete Playwright for automation

Challenges we ran into

Lots of issues with the latency of Browser use and high credit use by the Browser Use agents, which led to working into a pipeline with both Playwright and Browser Use

Accomplishments that we're proud of

Proud of getting and developing the playwright/Browser Use/Gemini pipeline for producing and providing accurate data for users

What we learned

Learned a lot about backend development and agentic AI for completing tasks

What's next for Nourish.ai

Creating a more efficient app in terms of latency and adding support for more medical conditions, as well as adding greater support for google maps api for location and addresses

Built With

• amazon-web-services
• browser-use
• celery
• gemini
• google-maps
• next.js
• pgvector
• playwright
• python
• redis
• supabase
• typescript