Skip to main content

Getting Started

What is CAIP Workflows?

Workflow API abstracts the complexity of managing AI infrastructure resources in a distributed cloud environment and allows users to perform Machine Learning tasks using the Ray framework.

Features

  • Define versioned workflows from reusable pipeline templates
  • Promote a version per environment with stages (test, int, e2e, prod)
  • Parameterize a single version per stage with configurations (${config.KEY})
  • Trigger user-defined workflows and inspect run logs
  • Start runs automatically with webhook & schedule triggers
  • Integrate with GitHub to fetch source code directly
  • Submit tasks to be executed on Ray in a distributed manner
  • Use the UV package manager to distribute code dependencies in Ray
  • Use custom container images as the subject of Ray executions

Quickstart

Get started quickly with these hands-on guides:

GuideDescription
API Usage →Run an end-to-end workflow — create a version, configure it per stage, and trigger a run.
Portal Integration →Manage and monitor workflows through the CAIP portal interface.

Reference