CoreX is an AI-powered sports and nutrition platform that combines cloud-based coaching agents with privacy-conscious edge inference. It helps users understand meals, plan nutrition, evaluate body composition, generate training programs, and receive real-time movement feedback.
CoreX is organized around five cooperating capabilities:
The repository contains three primary application surfaces:
CoreX/
|-- backend/ FastAPI service, domain modules, agents, and migrations
|-- mobile/ React Native and Expo mobile application
|-- preview/ Static web preview served by Nginx
|-- packages/ Shared TypeScript packages, including biomechanics
|-- docs/ Architecture, API, and vision system documentation
|-- scripts/ Data, training, benchmarking, and Docker utilities
|-- docker-compose.yml Local PostgreSQL, Redis, backend, and preview stack
The backend exposes versioned REST endpoints under /api/v1 and uses PostgreSQL for persistent application data and Redis for cache and agent memory. The application startup path initializes SQLAlchemy metadata and exposes health and OpenAPI endpoints. Model weights, datasets, image assets, and local database files are intentionally kept outside source control.
| Area | Technologies |
|---|---|
| Backend | Python, FastAPI, SQLAlchemy, Alembic, Pydantic |
| AI orchestration | Specialized nutrition, body-composition, workout, diet, and coaching agents |
| Computer vision | Vision-language inference, food segmentation, nutrition retrieval, and pose analysis |
| Mobile | React Native, Expo, TypeScript |
| Shared packages | TypeScript biomechanics engine |
| Data services | PostgreSQL and Redis |
| Local deployment | Docker Compose and Nginx |
Create a local .env.docker file with the required application and model configuration, then run:
docker compose up -d --build
The services are available at:
Inspect or stop the stack with:
docker compose ps
docker compose down
Windows and Unix helper scripts are also available in scripts/.
cd backend
python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt
uvicorn app.main:app --reload
The backend expects a configured database and the environment variables documented in the application configuration. Do not commit secrets or local database files.
cd mobile
npm install
npx expo start
Open preview/index.html directly in a browser, or use the Docker Compose web-preview service for the containerized version.
The main API domains are:
/api/v1/auth for account registration and authentication/api/v1/biometrics for body composition and metabolic calculations/api/v1/nutrition for food scanning, dish retrieval, macro recalculation, and diet plans/api/v1/workouts for multi-sport workout plans/api/v1/pose for exercise biomechanics rules/api/v1/coach for readiness, chat, and action-plan workflowsSee docs/API_SPECIFICATION.md for request and response details.
This repository contains source code, configuration, migrations, documentation, and reproducible scripts. Image files, local databases, model weights, generated binaries, and other large artifacts are excluded from Git to keep the repository portable and reviewable. Obtain private or externally hosted assets through the project setup process instead of committing them.
Before opening a pull request, verify the staged file list and confirm that no secrets, images, database files, datasets, or model artifacts are included.
CoreX is an AI-powered sports and nutrition platform that combines cloud-based coaching agents with privacy-conscious edge inference. It helps users understand meals, plan nutrition, evaluate body composition, generate training programs, and receive real-time movement feedback.
CoreX is organized around five cooperating capabilities:
The repository contains three primary application surfaces:
CoreX/
|-- backend/ FastAPI service, domain modules, agents, and migrations
|-- mobile/ React Native and Expo mobile application
|-- preview/ Static web preview served by Nginx
|-- packages/ Shared TypeScript packages, including biomechanics
|-- docs/ Architecture, API, and vision system documentation
|-- scripts/ Data, training, benchmarking, and Docker utilities
|-- docker-compose.yml Local PostgreSQL, Redis, backend, and preview stack
The backend exposes versioned REST endpoints under /api/v1 and uses PostgreSQL for persistent application data and Redis for cache and agent memory. The application startup path initializes SQLAlchemy metadata and exposes health and OpenAPI endpoints. Model weights, datasets, image assets, and local database files are intentionally kept outside source control.
| Area | Technologies |
|---|---|
| Backend | Python, FastAPI, SQLAlchemy, Alembic, Pydantic |
| AI orchestration | Specialized nutrition, body-composition, workout, diet, and coaching agents |
| Computer vision | Vision-language inference, food segmentation, nutrition retrieval, and pose analysis |
| Mobile | React Native, Expo, TypeScript |
| Shared packages | TypeScript biomechanics engine |
| Data services | PostgreSQL and Redis |
| Local deployment | Docker Compose and Nginx |
Create a local .env.docker file with the required application and model configuration, then run:
docker compose up -d --build
The services are available at:
Inspect or stop the stack with:
docker compose ps
docker compose down
Windows and Unix helper scripts are also available in scripts/.
cd backend
python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt
uvicorn app.main:app --reload
The backend expects a configured database and the environment variables documented in the application configuration. Do not commit secrets or local database files.
cd mobile
npm install
npx expo start
Open preview/index.html directly in a browser, or use the Docker Compose web-preview service for the containerized version.
The main API domains are:
/api/v1/auth for account registration and authentication/api/v1/biometrics for body composition and metabolic calculations/api/v1/nutrition for food scanning, dish retrieval, macro recalculation, and diet plans/api/v1/workouts for multi-sport workout plans/api/v1/pose for exercise biomechanics rules/api/v1/coach for readiness, chat, and action-plan workflowsSee docs/API_SPECIFICATION.md for request and response details.
This repository contains source code, configuration, migrations, documentation, and reproducible scripts. Image files, local databases, model weights, generated binaries, and other large artifacts are excluded from Git to keep the repository portable and reviewable. Obtain private or externally hosted assets through the project setup process instead of committing them.
Before opening a pull request, verify the staged file list and confirm that no secrets, images, database files, datasets, or model artifacts are included.