Dashboard Decision · Proxmox / Docker / GitOps Homelab · 2026
STAY WITH HOMEPAGE
Only candidate scoring ≥ 3/5 on all three decisive axes simultaneously.
6/6 target integrations · pure-YAML config-as-code · 80 MB lean footprint.
No other surveyed tool matches all three hard requirements at once.
Full Scoreboard — 8 candidates · 7 criteria · max 35 pts
| Project | (1) Health | (2) GitOps | (3) Integrations | (4) Footprint | (5) Auth/Proxy | (6) UX/Maint | (7) License | Σ |
|---|---|---|---|---|---|---|---|---|
| Homepage ⭐ 31k | 4 | 5 | 5 | 4 | 4 | 4 | 5 | 31 |
| Glance ⭐ 35k | 5 | 5 | 2 | 5 | 3 | 4 | 4 | 28 |
| Dashy ⭐ 25k | 4 | 4 | 3 | 3 | 4 | 5 | 5 | 28 |
| Homarr ⭐ 4k | 4 | 2DB | 3 | 2 | 5 | 5 | 5 | 26 |
| Homer ⭐ 11k | 3 | 5 | 1 | 5 | 3 | 3 | 5 | 25 |
| Organizr ⭐ 5.8k | 2 | 1DB | 3 | 3 | 4 | 4 | 5 | 22 |
| Flame ⭐ 6.4k | 2 | 2 | 1 | 5 | 2 | 3 | 5 | 20 |
| Heimdall ⭐ 9.2k | 1 | 1DB | 2 | 3 | 2 | 3 | 4 | 16 |
DB = uses a database for config storage — disqualifier for a pure-YAML gitops homelab. Dashy and Homarr both score 28/26 but on opposite strengths (UX/auth vs integrations/footprint).
5 — Best in class
4 — Strong
3 — Adequate
2 — Weak
1 — Disqualifying gap
Three Decisive Axes — where the decision was actually made
⚡ Integrations (6 targets)
Homepage
5
Dashy / Homarr
3
4/6 each. Homarr v1.x dropped Uptime Kuma & HTTP/ping [src];
Dashy lacks Jellyfin + Docker API.
Glance
2
RSS/weather/GitHub design; no Pi-hole, Proxmox, Uptime Kuma, or Jellyfin widgets as of mid-2026.
Homer / Flame
1
Bookmark launchers only — no live widget data.
📄 Config-as-Code (GitOps fit)
Homepage
5
Four YAML files: services.yaml, docker.yaml, widgets.yaml, settings.yaml.
100% file-driven, commit-idempotent, zero server state.
Deploy:
git pull --ff-only && docker compose up -d
Glance / Homer
5
Single config.yaml / static YAML. Equally git-native, but both lose on integrations (see left).
Dashy
4
YAML primary; GUI editor can also rewrite the file — acceptable but adds a drift risk in practice.
Homarr DQ
2
DB + drag-drop UI is the primary model; YAML export only. Breaks the gitops contract.
Organizr / Heimdall DQ
1
SQLite-only configuration. Hard disqualifier for this use-case.
🪶 Footprint (image + idle RAM)
Glance / Homer / Flame
5
Single binary, <30 MB RAM. Leanest possible architecture.
Homepage
4
80.6 MB compressed image,
50–150 MB idle RAM [~118 MiB measured on TrueNAS].
Acceptable for most homelabs.
Dashy / Organizr
3
~90–150 MB image; ~150 MB steady-state RAM.
Homarr
2
200–600 MB RAM in v1.x [confirmed regression].
Worst footprint of any candidate surveyed.
Challenger Watch — what changes this recommendation
- No first-party Pi-hole, Proxmox, Uptime Kuma, or Jellyfin widgets — integration score 2/5 is the blocker, not a caveat.
- Widget philosophy is RSS / weather / GitHub stats / iframe embeds — fundamentally different from Homepage's service-polling model. Parity would require a design pivot, not just feature additions.
- Migration trigger: Glance shipping first-party Pi-hole + Proxmox + Uptime Kuma widgets would make its star lead (35k vs 31k), footprint edge (<30 MB vs 80 MB), and single-binary simplicity decisive. That parity hasn't happened yet as of 2026-06-07.
Operational Must-Knows — Homepage deployment gotchas
⚠ HOMEPAGE_ALLOWED_HOSTS
Mandatory since v1.0 for all non-localhost access — must list every FQDN used to reach the container. Omitting it produces an opaque error; the most common post-upgrade breakage. → docs
🔐 No Built-In Auth — By Design
Zero built-in authentication and maintainers have permanently declined to add it. Decision #529 — all auth delegates to Caddy / Traefik forward-auth or VPN. Single shared read-only view; no per-user ACLs.
🚌 Bus Factor = 1
@shamoon authored 32 of the last 100 human commits; original creator @benphelps inactive since Sep 2024. ~$4,912 total Open Collective funding provides no succession buffer. Same bus-factor risk, opposite angle, as Glance (also single-maintainer).
📦 Version Pinning Critical
v1.0 jump from v0.10.9 bundled 4 breaking changes: mandatory host validation, Next.js 15 (drops armv7), Tailwind v4, Kubernetes Gateway API. Pin versions — never use :latest in production. Test upgrades before pulling.