Deep Dive
What Proxmox VE and Docker actually are, what each self-hosted service does and why it exists, and why running your own software is a fundamentally different relationship with your data.
The Foundation
A single physical computer can only run one operating system at a time — unless you use a hypervisor. Proxmox Virtual Environment (PVE) is an open-source hypervisor that runs directly on the hardware and lets one physical machine host multiple completely independent virtual machines and containers simultaneously.
In this lab, a single ASUS NUC with an Intel i5 and 64GB of RAM runs a Home Assistant VM, an Ubuntu Desktop VM for Minecraft servers, a Docker host VM running over a dozen containerized services, and several LXC containers for network services — all at the same time, all isolated from each other, all manageable from a single web interface.
The practical result: the lab runs the equivalent of several servers on a machine that draws less power than a light bulb, takes up the space of a hardcover book, and costs a fraction of what individual server hardware would.
Containerization
Inside the Docker Host VM, individual applications run in containers — lightweight, isolated environments that package an application and everything it needs to run into a single portable unit. Containers are not full virtual machines; they share the host operating system's kernel, which makes them fast to start, small in memory footprint, and simple to manage.
The practical advantage is isolation and reproducibility. Each service — the recipe database, the document scanner, the AI interface — runs in its own container. If one service crashes, it doesn't affect any other. If a service needs updating, only that container is touched. If the whole host needs to be rebuilt, every service can be brought back up from configuration files in minutes, not days.
All containers in this lab are defined in a single docker-compose.yaml file. The entire service configuration lives as code — version-controlled, portable, and reproducible.
What's Running
Each service below replaces something a tech company would otherwise own — the data, the interface, the subscription, and the decision about when to shut it down or change the terms.
A full-featured recipe manager with meal planning, shopping lists, and automatic import from recipe URLs. Every recipe in the family collection lives here — privately, without ads, without the 4,000-word blog post before the ingredients. Accessible from any device on the network.
Saves bookmarks along with a full archived copy of each page. When websites go down, change their URLs, or disappear entirely, the content is still accessible locally. Feeds directly into an AI pipeline via n8n for automated summarization and categorization into Obsidian.
Ingests scanned documents, runs OCR to make them searchable, and organizes them with automatic tagging and categorization. Every important document — tax records, insurance, warranties, medical records — is searchable, backed up, and accessible from anywhere on the network. No paper filing system, no lost documents.
A full-featured chat interface for large language models that run entirely on local hardware — no data sent to OpenAI, no API costs, no conversation logs on someone else's servers. qwen3:8b handles everyday tasks; phi4:14b handles heavier reasoning. The same capability as commercial AI assistants, running on a NUC in a home office.
A self-hosted automation platform that connects services together with visual workflows. In this lab it powers a knowledge pipeline: saved links in Linkwarden are automatically summarized by Ollama and written to Obsidian as structured notes. The automation equivalent of Zapier or Make — running locally, for free, with no data leaving the network.
Keeps files in sync across multiple devices — phones, laptops, the Proxmox Docker host — without any cloud intermediary. The Obsidian vault syncs silently between all devices this way. Changes made on a laptop appear on the NUC, and vice versa, with no Dropbox or iCloud involved.
Monitors every self-hosted service and sends alerts when something goes down. A dashboard that shows at a glance whether everything is healthy — response times, uptime percentages, and instant notifications if any service stops responding. The lab's own status page.
A self-hosted search engine that queries multiple search providers simultaneously and returns combined results — without sending any identifying information to those providers. Search Google's index without Google knowing who searched. Runs as an LXC container, accessible from any device on the network.
The Case For It
Every cloud service is a relationship built on terms that can change without notice. Prices go up. Free tiers disappear. Companies get acquired. Services shut down. Your data — your recipes, your documents, your bookmarks, your files — lives on infrastructure you don't control, under policies you can't negotiate, accessible to employees you'll never meet.
Self-hosting is the alternative. It's more work upfront. But the benefits compound over time in ways that cloud services simply cannot offer.
Your recipe collection, documents, and bookmarks live on hardware in your home. No company can sell it, lose it in a breach, or delete it when they sunset the product.
The services here cost the electricity to run them. No monthly fees for recipe apps, document storage, file sync, or AI access. The hardware pays for itself quickly.
When Google Drive has an outage, you can't access your files. When your self-hosted services are running, access doesn't depend on anyone else's infrastructure.
Every service configured, every problem debugged, every integration built is a skill added. The lab is as much an education platform as it is an infrastructure platform.