FitLife - AI Fitness and Nutrition Coach

Scan food labels, review workout form, and ask FitLife Coach for meal and recovery guidance.

---

Live Demo

• Frontend: https://fit-life-ai-1eqx.vercel.app
• Backend API: https://shinzobolte-fitlife-ai.hf.space

---

Overview

FitLife is a full-stack fitness product with three core surfaces:

• Fuel Scan - upload a nutrition label photo and get extracted nutrition data, scoring, and profile-aware recommendations.
• Form Coach - upload a workout clip and receive rep counting, movement scoring, and coaching cues powered by YOLO pose checkpoints.
• FitLife Coach - ask nutrition and recovery questions through a profile-aware AI assistant.

This repo is deployed with:

• Vercel for the Next.js frontend
• Hugging Face Spaces (Docker) for the Flask backend

---

Tech Stack

Layer	Technology
Frontend	Next.js 15, React 19, Tailwind CSS 4, TypeScript
Backend API	Flask, SQLAlchemy, PyJWT, Flask-Login
AI / ML	YOLOv8 pose, Groq vision/text models, RAG retrieval
Async Tasks	Celery + Redis (optional)
Database	SQLite (dev), PostgreSQL / Neon (prod)
Deployment	Vercel (frontend) + Hugging Face Spaces (backend)

---

Project Structure

fitlife/
|-- frontend/                # Next.js frontend
|-- gateway/                 # Flask API gateway
|-- services/
|   |-- muscle_ai_service/   # Workout analysis service
|   |-- nutri_ai_service/    # Nutrition / RAG service
|   `-- shared/database/     # Shared SQLAlchemy models
|-- data/
|   `-- nutri-ai/            # Knowledge base JSON files
|-- model/
|   `-- models/              # Tracked YOLO workout checkpoints
|-- docs/                    # Resume + deployment docs
`-- web/                     # Legacy Jinja templates / static output

---

Quick Start

Prerequisites

• Python 3.11+
• Node.js 18+
• Groq API key for nutrition OCR / chat features

1. Clone the repo

git clone https://github.com/shinzoxD/FitLife-AI.git
cd FitLife-AI

2. Backend setup

python -m venv .venv

# Windows
.\.venv\Scripts\Activate.ps1

# macOS / Linux
source .venv/bin/activate

pip install -r requirements.txt
copy env.example .env

Edit .env and set at least:

• SECRET_KEY
• JWT_SECRET_KEY
• GROQ_API_KEY
• DATABASE_URL for PostgreSQL in production

3. Frontend setup

cd frontend
npm install
copy .env.example .env.local
cd ..

Set this in frontend/.env.local:

NEXT_PUBLIC_API_URL=http://localhost:5000

4. Run locally

Open two terminals:

# Terminal 1
python gateway/app.py

# Terminal 2
cd frontend && npm run dev

Open http://localhost:3000.

---

Workout Models

The workout checkpoints are tracked in:

• model/models

Current checkpoints:

• best.pt
• sumo_best.pt
• squats_best.pt
• best_romanian.pt
• zercher_best.pt
• front_squats_best.pt

The workout service auto-resolves these from:

• services/muscle_ai_service/config/settings.py

To inspect embedded validation metrics:

G:\fitlife\.venv\Scripts\python.exe scripts\evaluate_muscle_models.py

To run a fresh validation pass once you have the dataset YAML:

G:\fitlife\.venv\Scripts\python.exe scripts\evaluate_muscle_models.py --data path\to\config.yaml

---

API Endpoints

All API routes are prefixed with /api/v1/.

Method	Endpoint	Auth	Description
POST	/auth/register	-	Create account
POST	/auth/login	-	Get JWT tokens
POST	/auth/refresh	-	Refresh access token
GET	/user	JWT	Get user profile
PUT	/user/settings	JWT	Update profile
GET	/user/scans	JWT	Scan history
GET	/user/workouts	JWT	Workout history
GET	/dashboard/stats	JWT	Dashboard statistics
POST	/nutri-ai/upload	Optional	Upload nutrition label
POST	/nutri-ai/analyze	Optional	Analyze extracted nutrition data
POST	/muscle-ai/upload	Optional	Upload workout video
GET	/muscle-ai/task/:id	-	Poll async task status
POST	/ana/chat	Optional	Chat with FitLife Coach

---

Environment Variables

Use env.example as the base. The important ones are:

Variable	Required	Description
SECRET_KEY	Yes	Flask session secret
JWT_SECRET_KEY	Yes	JWT signing secret
DATABASE_URL	Production	PostgreSQL connection string (Neon works well)
GROQ_API_KEY	For nutrition/chat	Groq API key
FRONTEND_URL	In production	Public Vercel frontend URL
NEXT_PUBLIC_API_URL	Frontend	Public backend URL

---

Deployment

Current Deployment: Vercel + Hugging Face Space

1. Deploy the backend on Hugging Face Spaces

Use a Docker Space and point it at this repo.

The root Dockerfile is now prepared for Spaces:

• installs the full backend + workout-model dependencies
• copies the tracked YOLO checkpoints from model/models
• serves the Flask gateway on port 7860

Space variables to set:

• SECRET_KEY
• JWT_SECRET_KEY
• DATABASE_URL=postgresql://...
• GROQ_API_KEY
• FRONTEND_URL=https://your-vercel-app.vercel.app
• GOOGLE_CLIENT_ID (optional)
• GOOGLE_CLIENT_SECRET (optional)

Use the template in docs/HUGGINGFACE_SPACE_README.md for the Space repo README.md metadata.

2. Deploy the frontend on Vercel

• Import the GitHub repo into Vercel
• Keep the existing root directory from vercel.json:
  • frontend
• Set:
  • NEXT_PUBLIC_API_URL=https://shinzobolte-fitlife-ai.hf.space

3. Final wiring

• After Vercel gives you the frontend URL, set FRONTEND_URL in the Hugging Face Space to that exact URL
• Redeploy the Space once after updating FRONTEND_URL

Notes:

• Hugging Face free Spaces can sleep when idle
• The backend is public at the .hf.space URL, which is required for the Vercel frontend to call it directly
• For durable user accounts and settings, use PostgreSQL instead of the default SQLite fallback

---

GitHub Actions CI/CD

This repo now includes:

• ci.yml for frontend build, gateway smoke tests, and workout model loading
• cd.yml for production deployment from main

Required GitHub Secrets

For the Vercel deployment job:

• VERCEL_TOKEN
• VERCEL_ORG_ID
• VERCEL_PROJECT_ID

For the Hugging Face Space deployment job:

• HF_TOKEN
• HF_SPACE_REPO

Example HF_SPACE_REPO value:

shinzobolte/fitlife-ai

How the pipeline works

• Pull requests and pushes run CI
• A successful CI run on main triggers CD
• CD deploys the frontend to Vercel
• CD syncs a Space-ready backend bundle to your Hugging Face Space repo

If you use the GitHub Actions Vercel deploy, disable duplicate auto-deploy behavior in Vercel if you do not want both Git integration and Actions creating separate production deployments.

---

Resume Positioning

Use docs/RESUME_PROJECT.md for resume bullets, a portfolio summary, and interview talking points.

---

Acknowledgments

• Ultralytics YOLOv8 for pose estimation
• Groq for nutrition OCR and text inference
• Harvard Medical School nutrition research used in the knowledge base

---

License

This project is licensed under the MIT License.