Skip to content

Concepts

The Cubonacci Console

Much of the work you will be doing on the Cubonacci Platform will be done through the Cubonacci Console. It is the main interface through which you can manage all of your projects.

Project

A Cubonacci project comprises all resources that concern a single business case. This includes experimentation, model training, data snapshots, deployments, and so on. Every project must be attached to at least one Git repository.

Repository

Throughout the Cubonacci platform, Git repositories are used as the versioning system for project code and settings. Every time a new version of the project code is added to Git, this version is mirrored automatically in the Cubonacci Console.

Commit

Every version of the code that is in your repository is called a commit. A commit in Cubonacci contains at least one algorithm, plus all the code and configuration required to load data snapshots, fit the model to the data, and some other peripheral functions.

Experiment

Experiments can be used to find the best hyperparameter settings to train your model. They can be started from the detail page of any successfully validated commit.

Model

Models can be trained from the detail page of any successfully validated commit. Trained models will be stored in a hybernated state. Once a model is trained, it is ready to be deployed to an API using a deployment.

Deployment

Trained models can be deployed using a deployment. When creating a deployment, the required infrastructure is provisioned to run the model and the model code is loaded into memory. The model can now be invoked through the deployment.

Endpoint

An endpoint is a URL that exposes a deployment to the outside world. Each deployment has a default endpoint that always points to the deployment, but additional endpoints can be created and connected to any deployment at any time.