Skip to content

Homelab Evaluation

Comprehensive evaluation of the planned homelab stack. Reviewed 2026-01-15.

Summary

Factor Score Notes

| Utility | 9/10 | All bases covered, maybe over-engineered initially | | Learning | 10/10 | Exceptional breadth and depth | | Privacy | 9.5/10 | Near-perfect sovereignty | | Costs | 9/10 | ~$15/mo for enterprise-grade setup | | Geekiness | 11/10 | Off the charts |


Utility: 9/10

Strengths

Service Daily Use Value

| Headscale | Remote access anywhere, no vendor lock-in | | Vaultwarden | Password management across all devices | | Pi-hole (x3) | Ad-free browsing everywhere | | Jellyfin + *arr | Media consumption without subscriptions | | Frigate | Security monitoring with AI detection | | Home Assistant | Automation potential | | Syncthing | File sync without cloud | | Start9 | Bitcoin sovereignty |

What You'll Actually Use Daily

  • Vaultwarden (passwords)
  • Pi-hole (transparent)
  • Jellyfin (media)
  • Headscale (remote access)

Potentially Underutilized

  • Home Assistant (needs smart devices to shine)
  • Start9/Bitcoin (unless actively using Lightning)
  • Frigate (3 cameras may be overkill initially)

Suggestion

Consider if all 23 services are needed at launch. Start with core (Headscale, Pi-hole, Vaultwarden, Jellyfin) and add incrementally.


Learning Opportunities: 10/10

Coverage

Domain Technologies

| Virtualization | Proxmox VE, VM management | | Networking | OPNsense, VLANs, DNS, mesh networking | | Containers | Docker, docker-compose, multi-host | | Security | Firewalls, certificates, encrypted backups | | Storage | NFS, Samba, backup strategies | | Linux | Debian, systemd, CLI tools | | Bitcoin | Full node, Lightning, Electrum | | ML/AI | Frigate object detection, Coral TPU | | IaC | SOPS, age encryption, Ansible (planned) |

Unique Learning Paths

  • Running Headscale (not just using Tailscale)
  • Proxmox + OPNsense virtualized router
  • NFS for distributed storage
  • Hardware video decode (QuickSync)

Growth Areas (Future)

  • Kubernetes (if you want to level up from Docker)
  • Terraform (for VPS provisioning)
  • Monitoring stack (Prometheus/Grafana)

Data Privacy: 9.5/10

Implementation

Aspect Implementation Rating

| Mesh control | Headscale (self-hosted) | Excellent | | DNS | Pi-hole everywhere | Excellent | | Passwords | Vaultwarden (local) | Excellent | | Files | Syncthing P2P (no cloud) | Excellent | | Media | Jellyfin (no tracking) | Excellent | | Bitcoin | Start9 full node | Excellent | | Backups | rclone crypt to Google | Good |

Minor Concerns

  • Google Drive for offsite backup (encrypted, but Google sees metadata)
  • VPS on Vultr (US jurisdiction)

Privacy Hardening Options

  • Consider Backblaze B2 or Hetzner Storage Box instead of Google
  • VPS on privacy-focused provider (Njalla, 1984.is) if paranoid
  • Current setup is already excellent for 99% of threat models

The "carry your mesh in your backpack" philosophy is peak sovereignty


Costs: 9/10

Monthly Costs

Item Cost Notes

| VPS (Vultr) | $6/mo | Helper only, not critical | | Google One | $0 extra | Already have AI Pro subscription | | cronova.dev | Owned | No additional cost | | verava.ai | ~$5/mo | ~$60/yr to purchase | | Electricity | ~$5-10/mo | Estimate for ~200W | | Total|~$12-16/mo | |

Hardware (One-Time, Already Owned)

  • Most hardware repurposed (NAS from 2013, RPi 4)
  • Smart purchases (PoE switch for cameras, UPS)

Cost Optimizations Already Done

  • Skipped nanduti.io + verava.net ($42/yr saved)
  • No Portainer (free tier limits)
  • No Nextcloud (Syncthing is lighter)
  • VPS as helper only (could be $0 if removed)

Potential Savings

Could eliminate VPS entirely (~$72/yr) if you:

  • Run DERP on mobile kit when traveling
  • Use free uptime monitoring (Uptime Robot)
  • Skip changedetection

Verdict: Exceptional value. Commercial equivalents would cost $50-100+/mo.


Geekiness: 11/10

Highlights

Factor Geek Points

| Headscale on RPi 5 | "Carry your mesh in your backpack" - peak mobile sovereignty | | Start9 Bitcoin node | Full node + Lightning on dedicated hardware | | Mini-ITX NAS from 2013 | Repurposed hardware, sustainable | | CLI-first philosophy | lazydocker > Portainer, no GUIs needed | | Modern shell tools | eza, bat, fd, ripgrep, starship | | Guarani domain research | Cultural flex (nanduti.io research) | | 3D printable case | Local maker culture | | QuickSync video decode | Hardware optimization for Frigate | | SOPS + age | Encrypted secrets in git | | Three-tier redundancy | Mobile/Fixed/VPS independent operation |

Geek Credentials

  • Conventional commits
  • Bare git repo for dotfiles
  • GPG signed commits
  • Vim keybindings everywhere
  • Neovim + LazyVim

What Would Increase Geekiness Further

  • OpenWrt on Beryl AX (instead of stock)
  • Custom Frigate ML model training
  • Home Assistant automations with ESPHome
  • Lightning payments for services

Recommendations

Phased Deployment

Prevents overwhelm and lets you learn each layer properly:

Phase Services Focus

| 1 | Headscale, Pi-hole, Vaultwarden | Core infrastructure | | 2 | Jellyfin, *arr stack | Media consumption | | 3 | Home Assistant, Frigate, Mosquitto | Automation & security | | 4 | Start9 (Bitcoin, Lightning) | Financial sovereignty |

Quick Wins

  1. Deploy mobile kit first (RPi 5 + Pi-hole + Headscale)
  2. Get Vaultwarden running early (password access)
  3. Media stack can wait until core is stable

Long-Term Considerations

  • Monitor actual usage of each service
  • Prune services that don't get used
  • Consider Kubernetes when Docker feels limiting
  • Add Prometheus/Grafana for observability

Conclusion

A thoughtfully designed, privacy-respecting, cost-effective homelab with excellent learning potential and maximum geek factor.

The architecture prioritizes sovereignty (self-hosted everything), resilience (three-tier redundancy), and practicality (CLI-first, minimal dependencies).

This is not just a homelab. It's a philosophy