Skip to main content

AstroPulse Platform

AstroPulse runs your applications and AI models on Kubernetes clusters in your own cloud account. It creates and operates the clusters, installs and maintains what they need, deploys and serves your workloads, and shows what they cost and what is vulnerable. Every change is checked with a dry run before it happens, recorded, and can be rolled back. Your code, data and models stay in your accounts.

You work in the console, and the same operations are in astroctl, the API and SDKs, and Nova, the AI platform engineer that answers questions about your platform and makes changes with your approval.

AstroPulse, the technical tour: from a cluster to a model answering, and running it

What you get​

  • Clusters in your cloud. Amazon EKS, Google GKE or Azure AKS created in your account from one wizard, or a cluster you already run, registered. Create a cluster
  • What a cluster needs, managed. Ingress, DNS, certificates, node and workload autoscaling, model serving, cost and vulnerability scanning, installed as add-ons AstroPulse keeps healthy and upgrades. Add-ons
  • Applications from any source. A container image, a Helm chart, a Git repository or YAML, on your domain with HTTPS. Deploy an application
  • AI models behind an API. Pick a model from Hugging Face or your own storage; AstroPulse sizes the GPU, serves it and puts it behind an API key. Serve a model
  • Cost and security you can see. Spend per application, live metrics, and known vulnerabilities in what you run. Cost, Vulnerability findings
  • Control built in. Roles, approvals, audit logs and short-lived, scoped access to your cloud instead of stored keys. Security

Start here​

  1. Set up AstroPulse: sign in to the console.
  2. Core concepts: the few pieces and how they connect.
  3. Deploy a Kubernetes cluster: create one in your cloud, or register yours.
  4. Deploy an application, or serve a model.

Prefer to ask? Open Nova and say what you want, for example "help me deploy a container image to my dev cluster". Nova plans it, shows you the change and makes it after you approve.

How it fits together​

Surfaces
ConsoleGuided UI for onboarding, clusters, profiles, and deployments
NovaConversational workflows with guided approvals
astroctl CLIRepeatable, scriptable command-line workflow
Platform · control plane
Platform APIEndpoints the console, Nova, and CLI call
Controllers & agentsReconcile desired state, connect clusters, keep deployments healthy
Infrastructure
Kubernetestoday
ManagedEKS · GKE · AKS
Self-hostedAWS · GCP
Registeredany provider / on-prem
  • Surfaces. The console, Nova, astroctl, and the API and SDKs make the same requests, so every way in is checked, refused and recorded the same way.
  • Control plane. The API, and the controllers and agents that turn what you ask for into running clusters and workloads, check each change first, and keep them healthy.
  • Your infrastructure. Managed Kubernetes on AWS, Google Cloud and Azure, self-hosted Kubernetes on AWS or Google Cloud, and registered clusters from any provider or on-premises.

Everything belongs to an organization: its clusters, applications, members, keys and plan.

Next​