FeedRx

Are you a big back but money kinda lack?

Comment

Inspiration

FeedRx was inspired by the connection between food access, nutrition, and overall wellbeing. One of our team members works as a CNA and has seen how deeply health can be affected by diet, especially for people managing blood sugar, acidity, inflammation, and other nutrition-sensitive conditions. Another team member grew up with fruit trees in their backyard and saw firsthand how produce changes after harvest as it becomes softer, sweeter, and sometimes more acidic. Those visible changes reflect real shifts in nutritional composition, meaning the same food can affect the body differently depending on its ripeness. For example, a banana can go from being a slower-digesting, lower-glycemic food to a sweeter, faster-digesting one as it ripens. At the same time, many people living in food deserts are not only struggling to access fresh produce, but also lack the time, resources, and information to think about how that food affects their health. When the priority is simply getting food on the table, nutrition education and wellness guidance can feel out of reach. FeedRx was created to help close that gap by offering evidence-based dietary guidance that helps people make more informed decisions with the food they already have. It is not a replacement for healthcare professionals, but a tool that makes everyday health and wellness more manageable.

What it does

FeedRX is a comprehensive web application designed to combat food insecurity across Texas by connecting communities with fresh, affordable food through smart technology. The platform features an interactive map that visualizes food deserts across the state, showing 16+ counties with limited grocery access alongside 11+ food bank locations with detailed contact information and hours. Users can scan produce with their camera to get AI-powered freshness analysis, including percentage scores, estimated shelf life, storage tips, and recipe suggestions, helping reduce food waste at home. The marketplace allows community members to share surplus produce safely with public meetup locations, while the SNAP assistant educates users about benefits and provides budget-friendly meal planning. A persistent AI health chatbot offers nutrition advice tailored to Texas foods like green chile and pinto beans, and the nutrition hub lets users search foods, scan barcodes, and compare nutritional values. All of this is wrapped in a cohesive Wild West cowboy theme that makes the experience both functional and memorable, addressing the critical problem of food deserts affecting over 2.3 million Texans.

How we built it

We built FeedRX as a full-stack JavaScript application using React for the frontend and Node.js with Express for the backend. The frontend leverages Material-UI for responsive components, Framer Motion for smooth scroll animations and immersive effects like dust particles and floating emojis, and Leaflet with OpenStreetMap for the interactive Texas map. Camera integration uses react-webcam to capture produce images for the AI scanner. The backend integrates multiple external APIs: Google Gemini AI for freshness analysis and chat, USDA FoodData Central for nutrition information, and Open Food Facts for barcode scanning. We also implemented Ollama with Llama2 for local AI capabilities as a fallback. To handle API rate limits, we built a key rotation system that automatically cycles through four Gemini API keys and tracks failures. The project follows a modular structure with separate frontend and backend folders, uses node-cache for performance optimization, and maintains a consistent Western theme with custom color palettes, monospace fonts, and cowboy terminology throughout all pages.

Challenges we ran into

One of the biggest challenges was managing API rate limits; the free tier of Google Gemini API would frequently hit quotas during testing, causing the freshness scanner to fail. We solved this by implementing a sophisticated key rotation system with four API keys and a failure tracking mechanism that automatically switches keys when quotas are exceeded, with an intelligent fallback knowledge base that provides realistic produce data even when all keys are exhausted. Another major challenge was Git-related: we accidentally committed massive node_modules folders and a 173MB Ollama installer file to the repository, which prevented pushes to GitHub. We had to use git filter-repo to remove these files from history and create a comprehensive .gitignore. File case sensitivity also caused issues on macOS, with duplicate files like home.js and Home.js confusing the module resolver. We also struggled with Leaflet marker icons not loading properly, CORS configuration between frontend and backend running on different ports, and getting Ollama to work on older macOS versions before switching to a fallback solution. Each challenge taught us valuable lessons about production-grade development practices.

Accomplishments that we're proud of

We're most proud of building a complete, production-ready application that solves a real-world problem affecting millions of Texans. The AI-powered freshness scanner works reliably with intelligent fallbacks, and our key rotation system ensures the demo never fails due to API limits. The interactive Texas map visualizes 16 food desert counties and 11 food banks with accurate coordinates, contact information, and filtering capabilities, turning raw data into actionable community resources. We successfully integrated multiple AI services (Gemini, Ollama, Llama2) across different features and built a cohesive Western theme that runs consistently through every page, from the dust particle effects to the cowboy terminology like "Dry Gulches" and "The Medicine Man." The marketplace with commenting and reply functionality, the SNAP benefits quiz, and the nutrition comparison tool all work seamlessly together. Perhaps most satisfying is that the entire application remains functional even when external APIs fail, thanks to our fallback systems, ensuring a smooth user experience regardless of external dependencies.

What we learned

This project taught us invaluable lessons about API integration and rate limit management, specifically how to build robust fallback systems and key rotation mechanisms that keep applications running even when external services hit their limits. We gained deep experience with Git best practices, learning the hard way about the importance of proper .gitignore configuration and how to clean up repository history when large files are accidentally committed. We mastered React hooks and state management across complex components, as well as animation techniques with Framer Motion for creating immersive user experiences. On the backend, we learned to structure REST APIs effectively, implement caching with node-cache to reduce API calls, and configure CORS correctly. We also discovered the nuances of working with mapping libraries like Leaflet, handling camera integration with react-webcam, and prompting AI models to return structured JSON data. Beyond technical skills, we learned the importance of user experience design, like how a consistent theme and intuitive flows make complex functionality accessible. Most importantly, we learned that technology can meaningfully address social issues like food insecurity when thoughtfully applied.

What's next for FeedRx

Next, FeedRx will expand its scan feature to deliver more diverse and accurate food analysis, making guidance even more reliable across a wider range of produce and food items. The platform will also introduce user login and profile features to improve security, personalize the experience, and make everyday use more convenient. In addition, FeedRx plans to add a recipe feature that recommends meals based on each user’s budget and dietary needs, helping turn informed food choices into practical action.

Built With

Share this project:

Updates