Skip to content

Latest commit

 

History

History
38 lines (26 loc) · 1.65 KB

architecture.md

File metadata and controls

38 lines (26 loc) · 1.65 KB

MedTagger Architecture

MedTagger's architecture was designed to be as much scalable as possible. It means that each of the system's component should allow to increase its performance easily.

Diagram

MedTagger Architecture

Components

Above diagram consists of several components. Each of them has its own responsibilities:

  • GUI - Graphics User Interface written in TypeScript. It's using Angular 4, SCSS and many other frontend technologies, libraries and tools. It works on user's web browser. This project is stored in /frontend directory.

  • Application - main backend application written in Python 3.6+. It's using Flask as a framework for REST API and many extentions that enable WebSocket support, easy input validation and output marshaling. It also use SQLAlchemy as SQL ORM engine, Alembic for databases management and logic for our data storage in Cassandra. It's part of a project in /backend directory.

  • Celery Workers - part of a backend application for long running and performance impacting tasks. It uses RabbitMQ as a broker that passes tasks for running. Workers may be running on multiple other nodes. Tasks are defined in the /backend/workers directory.

  • RabbitMQ - message broker used for communication between main application instance and Celery Workers' nodes.

  • Cassandra - NoSQL database storage used for aggregation of all Scans and its Slices. It stores data using high-scalable tables with key-value access.

  • PostgreSQL - SQL database that stores all of our business models and metadata related to medical data.