Experimental API

Experimental API (EAPI) is the graduation project for my Bachelor degree.

Goals

EAPI attempts to create a easy-to-use and uniform interface to support the task of adding, removing and interacting with Machine Learning models over the Internet.

Architecture

In order to achieve its goals, EAPI makes use of the microservices architecture; its main components are shown in the following diagram:

It's important to note that, as of now, there's a name mismatch between some components shown in the diagram and their counter parts in the codebase; with respect to the docker-compose.yml config file:

• Entrypoint is called backend; its source code resides in the be/ directory.
• Reverse Proxy is called nginx gateway; its configuration files reside in the nginx/ directory.
• Register is called consul server;
• User Interface is called frontend; its source code resides in the fe/ directory.

Installation

Github Oauth App

This system leverages GitHub for authentication; thus you need to create a GitHub Oauth App from your account's settings.

During the process, you will be assigned a client ID and a secret key; this data is needed for the next step.

Environment Variables

Assuming you have just cloned the repository, you have to manually create a file .env to configure the basic environment variables needed to run the system; use .env.sample as a template.

Start by filling in GITHUB_CLIENT_ID and GITHUB_CLIENT_SECRET variables using the information obtained during the previous step.

The other variables can be left as they are or deleted altogether, since the system will use defaults.

Run

To run the system, in the root directory run the following command

docker compose up

and wait for all sub-systems to start up.

The first run will take longer than the next ones, since several dependencies will need to be installed (no input required).

Stopping

To stop the system, press CTRL+C inside the terminal the docker compose command was run from and wait for all subsystems to shut down gracefully.

Clean up

To clean up run (in the root directory):

docker compose down

Usage

Once the system is running:

• To use the UI, connect to http://localhost:3000 (assuming FRONTEND_PORT=3000).
• To use the register's UI (Consul UI), connect to http://localhost:8500.

Licence

This project is currently licenced under the MIT Licence.