SupletAI

Your AI assistant for nutritional supplements and vitamins

Comment

Inspiration

As a fitness enthusiast, I was always confused by supplements. It felt like a world of hype and guesswork. I wanted to build something for people like me—a simple tool that uses AI to give clear, personalized advice without the marketing fluff.

What it does

SupletAI acts as an intelligent, personal supplement advisor. Its key feature is a scanner that uses Google's multimodal Gemini model; users can simply take a photo of a supplement bottle, and the app instantly identifies the product, its ingredients, and quantity.

Beyond the scanner, the app asks users about their health goals to provide a list of recommended supplements. For premium users, it also features a conversational voice agent (powered by ElevenLabs) to ask questions and receive advice in a natural, spoken dialogue.

How we built it

Thanks to Bolt.new, was built this project with a modern tech stack that is both powerful and efficient:

Frontend: React & TypeScript with Vite, styled using Tailwind CSS.

Backend: Netlify Functions for serverless logic and Supabase for the database and user accounts.

The AI Core: This project leverages multiple AI models. The supplement scanner is driven by Google's multimodal Gemini model, text-based recommendations are powered by the OpenRouter.ai API, and the premium voice agent uses the ElevenLabs API.

Challenges we ran into

My biggest challenge was implementing the multimodal Gemini model for the scanner. Getting the AI to accurately parse information from diverse product labels with different layouts and lighting conditions required a lot of experimentation. Additionally, crafting the text-based AI prompts to deliver safe and relevant advice took significant fine-tuning.

Accomplishments that we're proud of

I'm particularly proud of the supplement scanner feature. Building a tool that uses multimodal AI to "see" and understand a real-world object feels like a piece of the future, and it makes the app incredibly easy to use. Integrating three different AI models for scanning, text, and voice into one cohesive application is an accomplishment that makes me really proud of this project.

What we learned

This project expanded my skills into the exciting field of multimodal AI. I learned how to process images on the client-side and prompt a vision model to get back structured data. It also taught me a ton about the complexities of integrating multiple, distinct AI services into a single, seamless user experience.

What's next for SupletAI

The next steps are to build out user profiles with Supabase to save recommendations and track progress. I also plan to fully integrate Stripe to manage subscriptions for premium features. For the scanner, I want to improve its accuracy across a wider range of products and add barcode scanning for even faster identification.

Built With

Share this project:

Updates