Stay with Homepage — or adopt it if you haven’t yet. It is the only candidate in this survey covering all six target service types (Pi-hole [1], Proxmox [2], Uptime Kuma [3], Docker socket [4], Jellyfin [5], HTTP/ping [6]) while also being purely YAML-file-driven, git-deployable (git pull --ff-only && docker compose up -d), and running at 80 MB image / 50–150 MB idle RAM [7][8]. No other candidate scores ≥ 3/5 on all three decisive axes simultaneously. Homarr v1.x self-disqualified on two hard requirements at once: it dropped HTTP/ping and Uptime Kuma [9] while adding a RAM regression to 200–600 MB [10] and shifting from YAML toward a DB/click-ops model. Migration in: Docker label auto-discovery (homepage.name, homepage.icon, homepage.href) bootstraps ~80% of a tile layout from an existing Compose stack with no manual entry [11].
Scored matrix
| Project | (1) Health | (2) GitOps | (3) Integrations | (4) Footprint | (5) Auth/Proxy | (6) UX/Maint | (7) License | Σ |
|---|---|---|---|---|---|---|---|---|
| Homepage ⭐ 31k | 4 — monthly rels, bus=1 [1] | 5 — pure YAML files, git-native [2] | 5 — 6/6 target svcs [3] | 4 — 80 MB img / 150 MB RAM [4] | 4 — fwd-auth; ALLOWED_HOSTS req [5] | 4 — clean UI; yaml-brittle [6] | 5 — MIT [7] | 31 |
| Glance ⭐ 35k | 5 — 35k⭐, 2024, fastest-rising [8] | 5 — single config.yaml [9] | 2 — no Pi-hole/Proxmox/UK/Jelly [10] | 5 — single binary, <30 MB [11] | 3 — reverse proxy only [12] | 4 — column layout, low maint [13] | 4 — AGPL-3.0, FOSS [14] | 28 |
| Dashy ⭐ 25k | 4 — 25k⭐, 525 cmts/90d [15] | 4 — YAML primary; GUI also writes [16] | 3 — 4/6; no Jellyfin, no Docker API [17] | 3 — ~150 MB RAM, 90 MB img [18] | 4 — built-in basic/LDAP/SSO [19] | 5 — most themes, icons, sections [20] | 5 — MIT [21] | 28 |
| Homarr ⭐ 4k | 4 — active rewrite (homarr-labs) [22] | 2 — DB+UI; YAML export only [23] | 3 — 4/6; HTTP/ping & UK dropped [24] | 2 — 200–600 MB v1.x regress [25] | 5 — built-in user/role mgmt [26] | 5 — drag-drop, polished UI [27] | 5 — MIT [28] | 26 |
| Homer ⭐ 11k | 3 — stable, low-cadence [29] | 5 — static YAML, zero server state [30] | 1 — bookmarks only, no widgets [31] | 5 — ~30 MB RAM, <10 MB img [32] | 3 — proxy only, no built-in [33] | 3 — clean; no live data [34] | 5 — Apache-2.0 [35] | 25 |
| Organizr ⭐ 5.8k | 2 — low cadence, occasional updates [36] | 1 — DB-only; disqualifier [37] | 3 — iFrame tabs, Plex focus [38] | 3 — moderate; PHP/nginx [39] | 4 — group-based fwd-auth, SSO [40] | 4 — tabs, highly customizable [41] | 5 — MIT [42] | 22 |
| Flame ⭐ 6.4k | 2 — low cadence, not dead [43] | 2 — env vars + REST API, no YAML [44] | 1 — bookmarks + weather only [45] | 5 — very lean, small binary [46] | 2 — simple password only [47] | 3 — simple, limited themes [48] | 5 — MIT [49] | 20 |
| Heimdall ⭐ 9.2k | 1 — stalled 7+ mo; replacements [50] | 1 — SQLite DB; disqualifier [51] | 2 — enhanced apps, limited data [52] | 3 — moderate; PHP/nginx [53] | 2 — basic; proxy passthrough [54] | 3 — clean tiles, dated UI [55] | 4 — MIT (stalled) [56] | 16 |
Scoring key. Criteria weighted for this use-case: (1) project health — commit cadence, maintainer count, trend; (2) config-as-code — fully YAML/file-driven, reproducible from git, no DB required; (3) integrations — Pi-hole, Proxmox, Uptime Kuma, Docker socket, Jellyfin, HTTP/ping; (4) footprint — image size and idle RAM; (5) auth & reverse-proxy fit — Caddy/Traefik compatibility, forward-auth, built-in auth; (6) UX/Maint — theming, ongoing maintenance burden for single admin; (7) license — permissive FOSS, no open-core paywall. Dashy and Homarr tie at 28 and 26 respectively but on different strengths — Dashy wins UX, Homarr wins auth/UX; neither covers the full integration list.
Research gaps
The config-as-code child went off-scope. It researched Kubernetes GitOps operators (Argo CD, Flux, Crossplane, Terraform/HCL) rather than dashboard YAML-config models. The homelab gitops question is simpler: Homepage stores everything in plain YAML files (services.yaml, docker.yaml, widgets.yaml, settings.yaml) that are 100% file-driven and commit-idempotent. One real friction point: brittle YAML where tab/space indentation errors silently break widget rendering — a recurring complaint in Discussions [6]. The git pull --ff-only && docker compose up -d deploy pipeline requires nothing else.
Glance’s integration list was not assessed. The integration-coverage child studied only Homepage, Homarr, and Dashy. Glance’s widget documentation [10] shows a different design philosophy (RSS, weather, GitHub stats, iframe embeds) with no first-party Pi-hole, Proxmox, Uptime Kuma, or Jellyfin widgets as of mid-2026 — hence the 2/5 integration score above is inferred from docs rather than hands-on testing.
Bus factor and the Glance problem
These are the same risk from opposite angles. Homepage’s effective bus factor is 1: @shamoon authored 32 of the last 100 human commits; the original author (@benphelps) went inactive in September 2024 [12]. Hobby-scale funding (~$4,912 total on Open Collective, dominated by one corporate donor [13]) provides no succession buffer. Meanwhile Glance ⭐ 35k — created 2024 and now the #1 dashboard by GitHub stars [14] — is architecturally leaner (single binary, sub-30 MB), fully YAML-driven, and actively developed. If its widget ecosystem reaches parity with Homepage for Pi-hole/Proxmox/Uptime Kuma/Jellyfin, the bus-factor risk becomes a migration trigger. That parity hasn’t happened yet.
Operational must-knows. Homepage has zero built-in authentication and will never add it [15] — all auth is delegated to the reverse proxy (forward_auth in Caddy, middleware labels in Traefik). HOMEPAGE_ALLOWED_HOSTS must list every domain used to reach the container since v1.0; omitting it produces an opaque error and is the most common post-upgrade breakage [16]. Pin versions — the v1.0 jump from v0.10.9 bundled four breaking changes including mandatory host validation and Next.js 15 (drops armv7) [17]. Three CVEs since 2024 (latest: GHSA-rg3r-jprv-xq38, Apr 2026) confirm the proxy API is a real attack surface if exposed unauthenticated [18].
What would change this recommendation: Glance shipping first-party Pi-hole, Proxmox, and Uptime Kuma widgets — at which point its lead on stars, footprint, and single-binary simplicity would make it the clear successor.