The ultimate fitness sidekick that knows your form, your diet, and your goals — powered by real-time Computer Vision and Voice AI.
GymBro isn't just a tracker; it’s a living context engine that acts as your personal elite-level coach. By merging high-frequency computer vision analysis with persistent memory, GymBro provides guidance that is specifically tailored to YOUR body and YOUR history.
- Precision Pose Estimation: Analyzes 17 key COCO keypoints at 30 FPS using YOLO11n-Pose to ensure your squat depth is perfect and your back is straight.
- Audio Coaching (Voice AI): Integrated with ElevenLabs, giving you instant, natural-sounding voice corrections (e.g., "Keep your chest up, Shriram!") while you are mid-set.
- WebRTC/WebSocket Streaming: Ultra-low latency streaming between your mobile camera and the AI backend for zero-lag feedback.
- Metabolic Calculation: Uses Mifflin-St Jeor formulas to precisely calculate your BMR and TDEE based on your profile.
- Context-Aware Dieting: AI-generated meal plans that adapt to your training volume.
- Supplementation Intelligence: Specific advice on micronutrients and performance boosters based on your goal (Bulking, Cutting, or Maintenance).
- Imbalance Detection: Identifies anterior pelvic tilt, kyphosis, and other common lifting posture issues.
- Physical Evolution Tracking: Visualized progress reports that show how your posture and body composition improve over months of training.
- RPG-Style XP System: Earn experience for consistent logs, perfect form scores, and distance traveled.
- Tiered Leaderboard: Rise from Beginner to Beast status and compete on the global "Bros" leaderboard.
GymBro is built on a Central Context Engine that ensures the AI doesn't forget who you are. This engine aggregates:
- Bio-Data: Weight, height, age, and gender.
- Training History: Last 50 workout summaries including average form scores and common faults.
- Voice Insights: Remembers previous questions you've asked the AI coach.
- Posture History: Maintains a baseline of your skeletal alignment to track improvements.
| Category | Technology |
|---|---|
| Mobile App | React Native, Expo, TypeScript |
| Design Language | Glassmorphic Dark UI, Ionicons, Linear Gradients |
| Backend | FastAPI (Python 3.11), Uvicorn (High-Concurrency) |
| Compute Vision | VisionAgents SDK, YOLO11, NumPy, OpenCV |
| Large Language Models | Google Gemini 1.5 Flash (Context Framing) |
| Real-time Protocol | WebRTC (Stream), WebSocket (Full-Duplex) |
| Persistence | MongoDB (Motor Async), Atlas Geo-Spatial Indexing |
- Velkey Integration: We are moving towards Velkey for ultra-optimized user context storage. This will allow the AI to retrieve years of training context in milliseconds, providing an even more personalized and predictive coaching experience.
- Wearable Sync: Direct integration with Apple Health and Google Fit for heart-rate-aware sessions.
- Multi-Camera Form Check: Simultaneous front/side analysis for 3D form verification.
- Recipe Scanner: Photo-to-Macro conversion for instant nutrition logging.
Ensure you have Python 3.11+ installed.
cd backend
python -m venv venv
# Activate & Install
source venv/bin/activate # Mac/Linux
pip install -r requirements.txt
# Start the engine
uvicorn main:app --reload --host 0.0.0.0 --port 8000cd frontend
npm install
npx expo startDeployed to Azure App Service for Containers. We provide two guides depending on your preference:
- Azure Deployment Guide (Portal Method) — Recommended for a GUI-based setup.
- Azure Deployment Guide (CLI Method) — For automated terminal-based setup.
All sensitive configurations are managed via .env in the backend/ directory. Never hardcode keys. Reference the .env.example for the required keys: GEMINI_API_KEY, STREAM_API_KEY, ELEVENLABS_API_KEY, and DEEPGRAM_API_KEY.
Project Owner: BEASTSHRIRAM
Engineered with ❤️ for the pursuit of gains.