My Entire Home Lab &
Automation Ecosystem

Over the past year I've built out a complete home lab and smart home ecosystem — not from a single kit or guided tutorial, but piece by piece, project by project. This post is the overview: what everything is, how it connects, and why I built it the way I did.

Each section below gets its own deep-dive post with real screenshots and config details. This is the map before the individual journeys.

🖥️ The Foundation — Proxmox on Two HP Mini PCs

Everything runs on two HP EliteDesk mini PCs, each with an Intel i5, 32GB RAM, and NVMe storage. They run Proxmox VE in a two-node cluster — giving me a combined 64GB RAM and 12 CPU cores in a setup that draws about 25W idle and fits on a small shelf.

Proxmox is the hypervisor layer. It runs both KVM virtual machines (for things that need their own kernel — Home Assistant OS, Windows VMs) and LXC containers (for lightweight services — Pi-hole, Unifi controller, monitoring stack). Everything in this post lives inside Proxmox.

• Two-node cluster — live VM migration between nodes, no downtime for maintenance
• Proxmox Backup Server running nightly snapshots of everything
• Isolated VLANs enforced at the VM/CT network bridge level
• All IoT devices on their own VLAN — no route to main LAN

📡 The Network — Unifi U7 Pro + OneTouch VPN

The home network runs on Ubiquiti Unifi U7 Pro access points — Wi-Fi 7, tri-band, PoE powered. They're managed by a self-hosted Unifi Network Application running as an LXC container on Proxmox.

The network is segmented into VLANs: main LAN, IoT, lab, and guest — each with its own SSID and firewall rules. IoT devices can reach the internet but not my computers. Lab VMs are fully isolated.

For remote access, I use Unifi Fabrics OneTouch VPN — WireGuard-based, set up with a QR code scan. When I'm working from a café, I'm on my home network. Pi-hole DNS follows me so I'm still ad-free everywhere.

🚫 DNS & Ad Blocking — Pi-hole

Every device on my network — phones, laptops, iPads, smart TVs — blocks ads at the DNS level. I didn't install anything on any of them. One Ubuntu LXC container on Proxmox runs Pi-hole, and every DNS query from every device passes through it first.

If the domain is in a blocklist (ads, trackers, malware C2 servers), Pi-hole returns nothing. The request never leaves. On an average day, about 28–32% of all DNS queries on my network get blocked.

I point my Unifi DHCP server to Pi-hole's IP — so every device that gets a lease automatically uses it. Zero per-device config. The blocklists include URLhaus malware domains, so it doubles as a lightweight security layer.

🏠 Smart Home — 'Tiasha Home' on Home Assistant

Tiasha Home is the name of my Home Assistant dashboard — named after my daughter. It runs as a full HAOS virtual machine on Proxmox, isolated on the IoT VLAN. An iPad mounted on the wall shows the dashboard 24/7 in kiosk mode.

The dashboard uses Home Assistant's sections layout with card-mod CSS theming. Warm amber tones for living areas, purple for media, green for kitchen and balcony. All three Google Cast devices — JBL Bar 500, Google TV Streamer, and the Nest Mini on my desk — controllable from the same screen.

Most of the YAML was written with Claude AI. I'd describe what I wanted, Claude generated the config, I pasted it in. The header card shows a dynamic greeting that changes based on time of day using Jinja2 templates.

• iPad kiosk: iPadOS 17.7.10, Guided Access triple-click lock, Auto-Lock disabled
• Google Cast: JBL Bar 500, Google TV Streamer, Nest Mini — one integration
• Kiosk Mode (HACS): hides header and sidebar for clean tablet UI
• Card-mod (HACS): custom CSS per card, colour-coded by room

📊 Monitoring — Grafana + InfluxDB

I can tell you exactly how many DNS queries Pi-hole blocked yesterday, what signal strength my phone had at 3am, and whether any of my Proxmox VMs spiked CPU overnight. That visibility comes from a Grafana + InfluxDB stack running in a Docker container on Proxmox.

The data pipeline looks like this: Unifi-Poller pulls metrics from the Unifi controller (AP stats, client counts, WAN throughput) and writes to InfluxDB every 30 seconds. Grafana reads from InfluxDB and renders the dashboards. Pi-hole has a built-in InfluxDB exporter too.

The numbers above are real — that's roughly what my dashboard shows on a typical evening. The visual is what matters: everything in one place, time-series so you can look back at trends, and alert rules you can set so Grafana notifies you if something goes wrong.

🤖 The Glue — Claude AI Throughout

If there's one thread through all of this, it's Claude AI. I used it for:

• Writing Home Assistant YAML and card-mod CSS from plain-English descriptions
• Debugging Jinja2 template errors in the dashboard
• Writing the Docker Compose stack for the Grafana/InfluxDB/Unifi-Poller setup
• Generating Pi-hole blocklist recommendations and whitelist fixes
• Writing these blog posts (yes, including this one — then I edit for accuracy and tone)

The projects that would have taken a weekend of documentation-reading and trial-and-error took a few hours with Claude in the conversation. That's not a shortcut — it's the new baseline for how solo builders work.