Skip to content

Taiga Cloud docs

Taiga Cloud is a fully managed cloud platform for running GPU intensive compute at scale. These guides cover working with it programmatically — provisioning machines, connecting them, and organising who can do what.

What you can manage
Compute Create and control machines from a catalogue of images and flavours, across clusters. Reach them over a console, and read their metrics and activity.
Networking Define networks and assign IPv4/IPv6 addresses to the machines.
Access Register SSH keys, manage devices, and issue tokens for automation.
Organisation Group resources into customers and projects, and grant users roles within them.

Where things live
API reference Every endpoint, generated from the running service, with a client for trying calls. Also the place for conventions — status codes, errors, rate limits, versioning.
Guides These pages: the things a specification cannot express, like how to complete an OAuth flow or migrate between versions.
Portal The web interface, for anything you would rather click than script.

The reference is generated, the guides are written

Anything derived from the code — paths, parameters, response shapes, status codes — belongs in /reference and is always in step with what is deployed. If these guides ever disagree with it, believe the reference.

How resources are organised

Two levels of grouping determine what you can see and do:

  • Customers are the top-level tenant group. Most resources belong to one.
  • Projects subdivide a customer, and are usually what day-to-day resources — machines, networks, keys — actually belong to.

Your role is granted within one of these, not globally, so the same endpoint can return different results to different people. That is scope, not personalisation. A resource outside your scope is reported as though it does not exist, so a surprising 404 is more often an access question than a wrong identifier.

Start here