The public front door, run entirely by Cloudflare.
Keeping DNS, caching, the firewall and a zero-trust tunnel at the edge means attacks are filtered far from the server and no inbound port is ever opened — the origin stays invisible.
Everything behind anton.social runs on a single server I own — no rented cloud. This is the full map: how a visitor's request travels from the edge of the internet, through a private cloud and a stack of containerised apps, and back again.
Every request travels top-to-bottom through seven layers — from the public edge of the internet down to stored data. Here's that path at a glance; explore it live on the interactive map below.
Click the map to zoom and drag around it like a canvas; press Esc to scroll the page again. Click any service for a plain-English explanation, or run Trace request.
The same journey the Trace request button animates above — in plain words. Each step lights up as the trace runs.
Seven layers carry a request from the public internet down to stored data, with five supporting tracks running alongside. Here's what each one is and why it's there. Click a technology to open its detail on the map.
The public front door, run entirely by Cloudflare.
Keeping DNS, caching, the firewall and a zero-trust tunnel at the edge means attacks are filtered far from the server and no inbound port is ever opened — the origin stays invisible.
The gate and wiring inside the building.
A hardware firewall and managed switch zone the network and decide what may move where, with NGINX serving and proxying at the edge of the estate.
One physical server turned into a private cloud.
OpenStack on bare-metal Debian gives an AWS-like control plane — virtual machines, networks and identity — so workloads are provisioned on my own hardware instead of rented cloud.
Apps packed into portable, replaceable boxes.
Docker images give identical runs everywhere; Compose and Kubernetes start and heal them; Harbor stores them — so deploys are repeatable and rollbacks are instant.
The website and everything you interact with.
Next.js and React in TypeScript build fast, SEO-friendly interfaces; SASS keeps styling maintainable; Payload manages content without touching code.
The brains the website talks to.
NestJS and .NET expose clean, documented APIs built spec-first with Swagger — structured, typed and testable business logic.
Where information lives and is cached.
PostgreSQL is the source of truth, MongoDB holds flexible documents, and Redis caches hot data so the site stays fast under load.
How code and infrastructure ship automatically.
GitLab and Jenkins build, test and deploy on every change; Terraform and Ansible define the whole estate as code — reproducible and version-controlled.
Where credentials are kept safe.
Vault issues and rotates secrets so nothing sensitive is hard-coded; Vaultwarden holds operator logins.
Keeping the estate healthy and the code clean.
Prometheus measures, Grafana visualises and alerts, and SonarQube gates code quality before anything ships.
Proving it works before users do.
A test pyramid — fast unit tests with Jest and the Node runner, full user-flow checks with Cypress.
Where it's designed and written down.
Figma and Photoshop shape the interface; Obsidian documents every decision, runbook and the estate map itself.
The conventions that keep a one-person estate maintainable.
Backends follow EBI / clean-architecture boundaries and SOLID principles, so logic stays testable and swappable.
Contracts are designed in Swagger before code — the spec is the single source of truth for every endpoint.
Feature, release and hotfix branches with reviewed merges; every change runs the pipeline before it lands.
Terraform's root-and-leaf model plus Ansible mean the whole estate is reproducible from version-controlled text.
No open inbound ports; a tunnel at the edge and Vault-issued, short-lived secrets everywhere inside.
One Dockerfile per service makes local, staging and production behave identically and rollbacks trivial.