Skip to content

Naughtian

Your operational pain is real, and a lone operator deserves a real answer. Tools small enough to audit in a day, with no consensus to babysit, that leave your infrastructure intact the day you walk away.

One operator cannot baby-sit a Vault cluster, guard a fragile tfstate, and pray each night that Consul avoids a split-brain. Each tool here takes one of those burdens and settles it for good. Each does one job, stands on its own, and attaches to assets you already own. Compose them the way you chain grep and awk, or use one and ignore the rest.

Kuberina

Offline scheduling for heterogeneous Kubernetes. Reformulates pod placement as multi-dimensional bin packing, borrowed from how mega-vessel shipping lines stow containers. Produces a reviewable blueprint instead of an opaque runtime decision.

Overview · Tutorial · Paper

Helvilette

Pull-based GitOps delivery for Ansible. Agents pull playbooks from Git and reconcile at the systemd layer. One last SSH session to install it, then port 22 closes forever.

Overview · Tutorial

Kallisto

A read-only secrets resolver on localhost. It speaks the Vault KV-v2 API over one encrypted file on an S3-compatible bucket. Your request path keeps working while Vault is sealed, failing over, or mid-upgrade, and while the bucket is down. Adopting it is one environment variable. So is removing it.

Overview · Tutorial

The capabilities that keep infrastructure calm, such as secret rotation, reconciliation loops and reviewed placement, usually arrive bundled inside an enterprise suite, a paid tier, or a platform team you do not have. Small organisations patch around the gap with duct tape that looks fine from outside and rots from within, and learn to leave running things untouched.

Naughtian brings those capabilities to a single operator. The approach can be blunt and the pace deliberate. The goal is to end the pain permanently, with no extra payment asked to make it stop.

  • You can read all of it. The codebase is small enough to audit in a day. An operator needs no certification to run it, and an ordinary AI agent can follow it.
  • It holds no state worth fearing. At rest the software is inert, so backup and restore are yours to do at will. It leans on distributed consensus as little as it can, because the most resilient consensus protocol is none.
  • It hides nothing and decides nothing for you. Every action is observable, and the final call stays in human hands.
  • It takes no hostages. Remove any tool and your playbooks, manifests and Vault-speaking applications keep working. Your assets come back to you.

Vault manages your secrets, and someone still has to manage Vault. Anything that needs a quorum to keep running cannot be the bottom of the stack, because losing a vote stops everything built on it.

Kuberina decides where workloads go before anything runs. Helvilette keeps the machines underneath converged on their desired state. Kallisto serves the secrets those workloads need from the same machine, without a round trip to the root of trust. When any of them fails, the layers around it keep working.

Every design is argued in public before it is built, so its reasoning can be questioned and broken by anyone. The papers carry the mathematics, the benchmarks and the argument, including what has not yet been measured.

These tools are one answer to pain that someone faced directly. It may not fit yours. The source is open for that reason: bring the operational stories these projects cannot yet resolve, fork a repository and reshape it, or build what is missing.

I have recognized my pain. Where is my solution? If nobody built it yet, I will create it myself.

Kalena, Kaeliir and Ginnungagap are the next answers in progress, and will appear here as they become real. Their names come from the same saga as the rest. See names and mythology.