Skip to content
RepWise

RepWise

RepWise is an advanced fitness app integrating real-time form detection with voice feedback, macro tracking with barcode scanning, and AI-powered workout programs.

28
Built at HackByte 3.0

Created on 5th April 2025

•

RepWise

RepWise

RepWise is an advanced fitness app integrating real-time form detection with voice feedback, macro tracking with barcode scanning, and AI-powered workout programs.

Carousel Gallery Item: 1
Carousel Gallery Item: 2
Carousel Gallery Item: 3
Carousel Gallery Item: 4
Carousel Gallery Item: 5

The problem RepWise solves

RepWise is an advanced AI-powered fitness application that helps users achieve their fitness goals through personalized workout plans, real-time form correction, and comprehensive health tracking.

Features

  • AI-Powered Workout Generation: Personalized plans based on user goals, equipment, and fitness level
  • Real-time Pose Detection: Live form correction with voice feedback
  • Macro & Nutrition Tracking: Food logging with barcode scanning capabilities
  • Gamification: PVP fitness challenges and achievement tracking
  • WearOS & Health Connect: Seamless data synchronization across devices
  • WebRTC Integration: Real-time form scoring and interactive workout sessions
  • Personalized User Onboarding: Comprehensive fitness profile creation

Tech Stack

  • Frontend: SvelteKit with TypeScript
  • UI Components: Custom component library with responsive design
  • Database: Prisma ORM
  • Authentication: User authentication and profile management
  • Real-time Communication: WebRTC for live workout feedback
  • AI Integration: For workout plan generation and form correction
  • PWA Support: Progressive Web App capabilities

Getting Started

Prerequisites

  • Node.js (v16 or later)
  • npm or pnpm
  • Database (see Prisma schema for requirements)

Challenges we ran into

Difficulties Encountered

  • WebRTC Issues: Unstable network and lack of familiarity forced us to switch to Agora for a more robust solution.
  • OCR Limitations: Finding an OCR model capable of accurately scanning nutrition tables was challenging; Mindee was barely sufficient.
  • Model Constraints: Smaller Gemini models struggled with function/tool calling, while higher-powered models were slower and had compatibility issues with the OpenAI toolspec.

Tracks Applied (2)

Gemini API

GitHub Issues: Managed tasks with GitHub Issues. Conventional Commits: Enforced consistent commit messaging. Pull Reques...Read More
Major League Hacking

Major League Hacking

Track: GitHub

Gemini Toolspec: Implemented Gemini’s toolspec to develop an agentic AI. Fixed-Type Responses: Produced deterministic ou...Read More

GitHub

Technologies used

OCR

TypeScript

Svelte

Mediapipe

agora

Prisma

SvelteKit

gemini

Barcode

Cheer Project

Cheering for a project means supporting a project you like with as little as 0.0025 ETH. Right now, you can Cheer using ETH on Arbitrum, Optimism and Base.

Discussion

Builders also viewed

See more projects on Devfolio