Fit AI

Fit AI is a World ID-verified small app that uses AI-powered food identification to transform calorie monitoring. After confirming their identity using World ID to guarantee genuine involvement, users only need to take pictures of their meals to obtain comprehensive calorie data quickly, which is facilitated by Google's Gemini API.

Users get "Calorie XP" points for every meal they log, the app gamifies healthy eating and creates an entertaining leaderboard system where users can compete with friends and the community. This method uses World ID's proof-of-personhood authentication to turn boring calorie tracking into an engaging social activity while protecting privacy.

Instant food detection from images, thorough nutritional analyses, an XP-based progression system, community leaderboards, and smooth connection with the World App ecosystem are some of the main features. By simplifying food tracking to a straightforward snapshot capture, the software solves the widespread issue of manual food logging requiring more than fifteen minutes every day.

By confirming that they are actual people, users protect their privacy, stop leaderboard manipulation by bots, and guarantee genuine interaction. Through social incentive and accomplishment rewards, the gamification component promotes regular tracking and aids users in maintaining long-term healthy behaviours.

Tech Stack & Architecture:

Frontend: Next.js with TypeScript for robust, scalable development UI Framework: Tailwind CSS for responsive design + shadcn/ui components for polished interface Icons: Lucide React for consistent iconography AI Integration: Google Gemini API for multimodal food recognition and calorie analysis Identity: World ID integration via MiniKit SDK for human verification Deployment: Vercel for seamless CI/CD and global edge deployment Version Control: Git for collaborative development

Implementation Details Utilising Gemini's vision skills, the primary feature analyses food photographs using structured prompts to determine nutritional content, estimate servings, and identify components. By employing iterative refinement, in which preliminary analysis is sent back to improve accuracy, the integration adheres to Gemini's best practices for food recognition.

Only confirmed people are allowed to compete on leaderboards thanks to World ID integration, which verifies users during onboarding using MiniKit SDK instructions. Zero-knowledge proofs protect user privacy while preventing bot manipulation.

Points are allocated via the XP system, which is recorded locally and synchronised throughout the World App ecosystem, according to nutritional objectives and tracking consistency. Leaderboards use World ID's anonymous verification technology to compile community statistics while protecting user privacy.

Notable Hacks & Optimizations: Progressive image compression was used to reduce Gemini API expenses without sacrificing accuracy.

Tailwind's JIT compilation was utilised in the World App environment to achieve small bundle sizes.

created unique shadcn components that are tailored for World App interfaces that prioritise mobile devices.

used Vercel's edge functions to achieve quick API replies that met World App's performance requirements.

The application avoids troublesome characteristics like lengthy scrolling and adheres to World App's design criteria, which include tab navigation, snap-to text boxes, and mobile-optimized layouts. The integration shows how native-feeling experiences may be produced inside the World environment using conventional web tools.