Sisyphus / status

Hosted Sites

Architecture

Host OS

Flatcar Container Linux

Runtime

Docker

Orchestration

docker-compose

Reverse Proxy

Caddy v2 (auto-HTTPS)

Networking

Headscale (zero-trust mesh)

Security

CrowdSec + Headscale ACLs

Monitoring

Python SRE Agent + Claude API

Backups

restic → Hetzner Storage Box

Secrets

.env files (not in git)

Provisioning

Bash + Butane/Ignition

SRE Agent →

Lightweight self-healing daemon. Polls every 60s, detects anomalies, remediates automatically. Claude Haiku provides root-cause analysis for correlated failures, capped at €5/month.

Operating Modes

Automated Remediations

Service Level Objectives

How to Host a Site in Sisyphus

You don't need to know anything about infrastructure. You work on your site, Sisyphus handles the rest.

What you need

• Claude Code installed on your laptop
• A GitHub repository for your site (e.g. martasmaldone.com, Apapacho)

Your workflow

1. Clone your repo and open it with Claude Code
2. Work on your site — edit HTML, CSS, images, whatever you need. Ask Claude to help you build pages, fix styles, add content.
3. Preview locally — open your index.html in a browser to check it looks right
4. Commit & push — when you're happy, commit your changes and push to GitHub. Claude can do this for you.
5. Done. Sisyphus picks up the changes automatically, runs security and sanitation checks, and pushes your site live.

That's it

You never touch servers, config files, or infrastructure. You just work on your site and push. If something goes wrong, the SRE Agent catches it before it reaches production.

App projects (Next.js, etc.)

Projects with a backend run as their own container behind Caddy. Deploys are fully automated: push to main → CI builds image → VPS pulls and restarts.

Onboarding a new app (one-time, done by infra operator)

1. Dockerfile — must exist in the project repo (multi-stage, alpine, non-root user)
2. Sisyphus config — add the service to docker-compose.yml with build: context, a Caddy reverse proxy block, and a Consul project entry
3. Secrets — generate a deploy secret (openssl rand -hex 32), set it on the VPS as DEPLOY_SECRET_<PROJECT>
4. Env vars — place .env.local on the VPS (API keys, database URLs — never in git)

Your workflow (after onboarding)

1. Work on your code — develop, test locally
2. Push to main — that's it
3. Deploy webhook triggers: VPS pulls source, builds Docker image, restarts container
4. Health check confirms it's up. If unhealthy, auto-rollback to previous version.

You never SSH into the VPS, run docker commands, or touch infrastructure. You just push code.

Roadmap

Design Priorities

1. Don't distract from the projects. Infrastructure must be boring once running.
2. Serve the sites reliably. Real people depend on these sites.
3. Testbed for future career. Relevant tech: AI ops, containers, immutable OS, zero-trust.
4. Test cutting-edge technology. Be novel in layers touched daily, boring in plumbing.

Be cutting-edge where it teaches you something transferable. Be boring where it just needs to work.

Monthly Cost