Yes. You read that correctly. Gaia is a multiplexer for cloud infrastructure providers, implemented as a Terraform provider. It does this using Archetypal Resources, currently implemented as an extension of the Terrafom SDK's schema package. An instance of Archetypal Resource encapsulates a multicloud representation of an archetype of IaaS(virtual machines, load balancers, etc). The Archetypal Resource, a multicloud and product-agnostic representation of a particular paradigm of service, allows resources implemented by Gaia to represent a spectrum of all supported solutions as opposed to just one.
At this point you're probably wondering why go through the trouble of using what's essentially an abstract resource, as opposed to just giving you the tools to declare against an actual set of infrastructure that's concrete. The intended application of providing a set of archetypal resources to cloud engineers is to allow the delegation of inferencing optimized multicloud service configurations from the vast and ever expanding selection of services that are available, to artificial intelligence. As much as I love the experience of stumbling through all the service offerings of various cloud providers while dumping service specifications into a spreadsheet and assembling a portfolio of resources that's bound to be suboptimally organized, I figured : why not let an algorithm broker the optimal infrastructure for me while I get back to working on the software that I actually just trying to host in the first place. The future is obviously multicloud, so let's democratize it.
Gaia is not yet capable of supporting the machine learning inferences that the project aspires to implement. What Gaia does currently perform is rudimentary provider multiplexing. This repository contains only the small start to a big vision, but feel free to read and play around with it, and maybe even contribute if you happen to feel sufficiently inspired.