AIoD - RAIL (Research and Innovation AI Lab)

RAIL service is a tool that allows AI practitioners to explore and use AI assets directly in AIoD. RAIL is developed within the AI4Europe project as one of the core services of the AI on Demand platform.

More extensively, RAIL:

• Is a web application
• … that enables AI practitioners
• … to work with AIoD AI Assets (explore, search, compare, …)
• … and create experiments that are reproducible and reusable
• … that are executable directly in the AIoD platform supported by its infrastructure
• … and that make use of AIoD AI Assets.

IMPORTANT: RAIL is a service built on top of AIoD API and relies heavily on its contents and functionalities.

Installation

This repository consists of two main components - frontend and backend applications, served by nginx configured as a reverse proxy (for more details see nginx/default.conf). Each of these components can be built and deployed individually (see the corresponding sub-folders frontend and backend). However, it is recommended to deploy RAIL as a whole using Docker Compose from the root directory.

Running command

docker compose up -d

starts the backend REST API application (FastAPI) together with its underlying database (MongoDB) and the frontend web application (Angular).

Note: Some additional changes to file docker-compose.yml might be needed in order to change the default ports, etc.

Configuration

• Backend - It is mandatory to configure environment variables in file backend/.env. See the section "Setup->Using docker compose" in the corresponding README file for more details.
• Frontend - There are some environments created and can be selected in the docker-compose.yml file by a build argument PROFILE of the frontend-app. Update one of the existing ones to suit your specific needs.

Authentication

Some of the functionality is accessible only for authenticated users. Authentication server needs to be correctly setup and referenced in the frontend and backend configuration files. The current setup relies on Keycloak deployed as part of AIoD API.

Usage

Following the installation instructions above, the web application may be reached at 127.0.0.1:80. Documentation of the underlying backend REST API is automatically generated and can be viewed at 127.0.0.1:80/api/docs.

Development and debugging

For development purposes, we recommend running RAIL's development docker containers (docker-compose) and then attach to the applications running inside. This helps to ensure that all developers develop in the same environment.

To start development and debugging, follow these steps:

1. Create development .env-dev file in the Backend root folder. It must define all variables the .env file defines.
2. Start the images either by running run-dev-docker-env.sh or by running the following command from the root of the project:

   • docker-compose -f docker-compose.yml -f docker-compose-dev.yml up -d --build

3. Attach to the application.
   • For Visual Studio Code, there are a .vscode/launch.json configuration file in the root directory. In the Run and Debug menu, you should see:
     • 

This gif shows how to start development containers and attach a debugger in Visual Studio Code:

Under the hood:

• Frontend starts the development server (ng serve ...).
• Backend utilizes the debugpy utility (debugpy ... -m uvicorn ...). The debug server listens on 0.0.0.0:5678.

Notes

When you install a new package (Python lib or an NPM package), you need to restart the dev containers.

Current setup doesn't support Frontend hot reload (--watch). You need to refresh your browser after every change.

For better development experience, we recommend installing the local dev environments and libraries. Depending on your IDE, this will allow for features like autocomplete.

• Frontend
  • Install node and npm package manager.
  • Install all the dependencies: npm ci
• Backend
  • [optional] Create a Python virtual environment, conda environment or equivalent.
  • Install all the dependencies: pip install --no-cache-dir --upgrade -r requirements.txt

Contact

For any questions please contact the maintainers:

• Martin Tamajka ([email protected])
• Andrej Ridzik ([email protected])