Overview
Nimbus is a container platform for teams that want production containers without operating Kubernetes day to day. You describe services and how they depend on each other. Nimbus handles health checks, startup order, rollouts, secrets, and scaling.
Under the hood it is Kubernetes in your cloud account, running ordinary OCI images. You do not manage node pools, ingress controllers, or Helm charts to ship the next version of a service.
Concepts
Four nouns cover almost every screen in the product.
- Project — one product. Usually one Git repo, sometimes a small set.
- Service — a long-running container with an image, a port, a health check, and optional dependencies.
- Environment — an isolated copy of those services. Most teams use dev, staging, and production.
- Deploy — a rollout of one or more services. Startup order comes from the dependency graph.
Design a service
Add a service on the canvas, point it at an image, and connect what it depends on. Nimbus fills in the restart policy, the probe, and the startup order. You can edit any of those defaults.
Prefer reviewable changes? Keep the same spec in Git. The canvas and the file stay in sync.
service.yaml
service: api
image: ghcr.io/acme/api:1.8.2
port: 8080
health: /healthz
depends_on:
- postgres
scale:
min: 1
max: 8
metric: requests
Deploy
Connect the repo. Pushes to your staging branch can deploy automatically. Production stays a click, or a tag, so it does not surprise the person on call.
A rollout pulls the image, waits for dependencies, hits the health check, then shifts traffic. If the probe fails, the previous version keeps serving.
shell
nimbus link git@github.com:acme/api.git
nimbus deploy --env staging
nimbus promote staging production
Environments and secrets
Each environment has its own variables and its own running copies. Promoting a service copies configuration, not secret values. Dev cannot read production secrets. After you save a secret, the UI shows that it is set — not the value.
Scaling
Set a minimum and a maximum. Scale on CPU, memory, or request rate. A minimum of zero scales the service down when it is idle. That fits workers and preview apps. Leave a floor of one on anything whose cold start your users would notice.
Git
Nimbus watches the branch you choose. Every deploy points at a commit, so rollback is redeploying an earlier SHA. The generated spec stays Kubernetes-shaped if you ever want to run it somewhere else.