CheesyServe is a deployment platform on Talos Kubernetes, with self-hosting on the roadmap. Push an image to any registry, define a workload, and the operator handles HA, isolation, and routing — public or strictly internal.
No real multi-tenancy, limited high-availability guarantees, and no native path to Kubernetes-grade resilience once your team outgrows a single box.
Spread constraints, disruption budgets, probes, graceful draining — CheesyServe encodes those defaults so a hardware failure doesn't become dropped requests or lost data.
Build however you like, then push to any registry — Docker Hub, GHCR, or your own.
Image, scaling, exposure method. Grouped into projects and environments.
Desired state becomes the right K8s resources, with resilience defaults applied.
Publicly over HTTPS, or internally with AWS VPC and Tailnet bridges.
Run it on your own infrastructure and own your data end to end. Packaging for self-hosting hasn't started yet.
An immutable, minimal OS purpose-built for Kubernetes — not Docker Swarm.
CheesyServe never builds your app. Push to any registry — Docker Hub, GHCR, or your own — and deploy.
Several teams share one cluster with proper isolation and lower overhead.
Deploy workloads that are never exposed to the public internet.
Reach internal workloads from an AWS VPC or your Tailnet — no public endpoint. Planned for a later iteration.
Self-hosted packaging hasn't started yet. The cloud tier's core is what we're building first; the AWS and Tailnet bridges are planned for a later iteration.