vh/esh-pfi-infrastructure
1
0
Fork 0
Code Issues 7 Pull Requests Actions Packages Projects Releases Wiki Activity

feat(homepage): recategorise on "do I open this?", collapse the API groups

Browse Source 
The board mixed tools with endpoints. A vLLM seat whose href is a /docs page
sat in the same band as ComfyUI; the MQTT broker and the RustDesk relay, which
have no page at all, sat in Apps; and `Service Networking` was thirteen members
spanning three AdGuards, five Dockges, two Traefiks and four headless agents.

Every group is now one of two kinds and they never mix. TOOLS are expanded and
sit at the top of their tab. ENDPOINTS — an API, a broker, a background agent,
an href that is /docs or /ping or nothing — carry `initiallyCollapsed: true`
and sit at the bottom. Collapsed is not hidden: the eyebrow and its rule still
render, so the tab still says the thing exists and one click expands it.

A second rule fell out of the same pass and now shapes the group boundaries: a
group's members should all carry a widget or none should. A stat strip makes a
card ~50px taller, so one widget card in a row of plain ones opens a void under
the plain ones. That is why AdGuard and Traefik get their own groups rather
than sharing one with Dockge, and it is most of why the old Service Networking
band looked broken. AdGuard (ANA) was the last short card in its row and now
carries the same query/blocked/latency strip as its two siblings — one
infra-ops AdGuard login authenticates against all three instances, verified
against each; it lives in that stack's .env on the host and is vaulted.

The sixteen GPU-backed model seats were deliberately NOT relabelled.
`homepage.group` is read at container creation, so clearer names for
`AI - Inference` and friends would have cost a recreate on six vLLM seats, four
eval seats and four TTS engines — multi-minute model reloads on endpoints peers
reach through the gateway. Order plus `initiallyCollapsed` buys the same
separation for nothing, so those names stay as they are on purpose.

28 containers that ARE cheap to bounce were relabelled, across five hosts, via
rerunnable elway playbooks. Their label steps are gated on the old value still
being present, so a second run reports skipped rather than churning. Two verify
steps were wrong on first contact and are fixed with the reason recorded: the
traefik check raced its own recreate, and asserting a model seat is "running"
cannot answer "did I bounce it" when a seat may be legitimately stopped —
container age can, and now does.

The canonical stacks/ tree was synced to the deployed labels afterwards, so
intent and reality agree again on all fourteen tracked stacks.

Also documents the real nature of the post-recreate blank dashboard, which cost
~25 minutes here and an hour on 2026-08-19. `initialSettings":{}` in the served
HTML is the catch branch of the page's data loader, not a warm-up and not a
cache — and the error can vanish entirely, because the logger is assigned inside
the same try and the catch only logs if the logger exists. Ruled out by
measurement this time: all four API routes return 200 with correct content while
the page serves {}, and the previous known-good settings.yaml reproduces it
identically. The README now carries the one-command test and the next lead.

Before/after, all four tabs: http://10.100.10.50:8090/b/homepage-relayout/
This commit is contained in:
vh
2026-08-24 08:54:06 -07:00
parent f6f2f69649
commit 39da1d4a97
25 changed files with 1022 additions and 111 deletions
Download Patch File Download Diff File Expand all files Collapse all files
stacks/homepage/conf/services.yaml
+25 -19
View File
@@ -27,24 +27,37 @@
icon: mdi-filmstrip
siteMonitor: http://10.100.10.50:8090/healthz
description: Media drop + upload-for-pickup + the standing agent link board — nh3-dev, 24h TTL except kept boards
- Voice Design Studio:
href: http://10.100.79.3:8216/
icon: mdi-microphone
siteMonitor: http://10.100.79.3:8216/health
description: Mint, audition and keeper-mark synthetic fleet voices — irv-ml1, CPU-only
- The Henge:
href: http://park.phasefinal.com:8420/
icon: mdi-clipboard-check
siteMonitor: http://park.phasefinal.com:8420/healthz
description: Durable needs-attention / idea parking (stonehenge-park) — ana-docker
# Was its own one-card `Games` group, which burned a full 4-wide row on a
# single panel. It is an app you open; this is where apps you open live.
- Pterodactyl:
href: http://10.250.50.55/
icon: mdi-gamepad-square
siteMonitor: http://10.250.50.55
description: Game server panel
# The AI tab is fully Docker-auto-discovered. Each inference service carries
# a homepage.group=AI - <role> label on its compose file (AI - Inference,
# AI - Eval & Retrieval, AI - Gateways & Chat, AI - Speech (TTS),
# AI - Audio Tools, AI - Image & Media). Tab assignment, group order, and
# column counts live in settings.yaml. Do not add entries here or they'll
# double up. To move a service between AI groups, change the label on its
# compose file and recreate the container (labels only apply on recreate).
# The AI tab is otherwise fully Docker-auto-discovered. Each service carries a
# homepage.group=AI - <role> label on its compose file (AI - Gateways & Chat,
# AI - Studios, AI - Inference, AI - Eval & Retrieval, AI - Speech (TTS),
# AI - Audio Tools, AI - Dormant). Tab assignment, group order, columns and
# collapse state live in settings.yaml. Do not add a labelled container here as
# well or it renders twice. To move a service between AI groups, change the
# label on its compose file and recreate the container — labels are read at
# creation, so `restart` will not do it.
- AI - Studios:
# Manual entry — Voice Design Studio is a user-level systemd service on
# irv-ml1, not a Docker-labeled stack, so it cannot auto-discover. It sits
# with the other studios rather than in Apps: it is a workspace you open and
# produce something in, which is exactly what that group is for.
- Voice Design Studio:
href: http://10.100.79.3:8216/
icon: mdi-microphone
siteMonitor: http://10.100.79.3:8216/health
description: Mint, audition and keeper-mark synthetic fleet voices — irv-ml1, CPU-only
- Media:
- Plex:
@@ -67,13 +80,6 @@
key: '{{HOMEPAGE_VAR_JELLYFIN_KEY}}'
enableBlocks: true
- Games:
- Pterodactyl:
href: http://10.250.50.55/
icon: mdi-gamepad-square
siteMonitor: http://10.250.50.55
description: Game server panel
- Infra - ANA:
- ANA-Firewall:
href: https://10.250.250.1
stacks/homepage/conf/settings.yaml
+124 -56
View File
@@ -46,34 +46,55 @@ statusStyle: ""
# reads as a rendering fault.
useEqualHeights: false
# Function-first layout, four-tab split:
# Main - daily-use apps, media, bookmarks, monitoring
# AI - the inference fleet, grouped by role (see below)
# Infrastructure - hardware, hypervisors, BMCs (per site)
# Toolchain - backend services running but rarely clicked
# ===========================================================================
# THE ORGANISING QUESTION IS "DO I OPEN THIS?" — NOT "WHAT IS IT?"
# (operator, 2026-08-24: "most of the issues are that tools I use and have a
# UI are interspersed with API endpoints which are largely informational
# only. They might even go in their own cards or start collapsed.")
#
# The AI tab splits the fleet by function so a 20+ service list reads as
# sorted groups instead of one endless column. Group membership is set by
# the homepage.group=AI - <role> label on each service's compose file:
# AI - Inference LLM seats you call (gen, char-rp, char-rp-reasoning, summarizer)
# AI - Eval & Retrieval judges, reward, rerank, embed, image-quality
# AI - Gateways & Chat routing gateway, control plane, chat frontends
# AI - Speech (TTS) text-to-speech engines
# AI - Audio Tools speech-to-text + audio dataset tooling
# AI - Image & Media image/video generation + pipelines
# AI - Dormant stopped stacks (rollback seats, retired auditions)
# Every group on this board is one of two kinds, and they never mix:
#
# AI TAB ORDER IS BY CLICKABILITY, NOT BY IMPORTANCE (operator, 2026-08-18).
# Groups render in the order they appear in this block, so the top of the tab
# is prime real estate and it should hold the things you actually open in a
# browser — chat frontends, ComfyUI, the control plane. Most of the model
# seats below them are vLLM API endpoints whose href is a `/docs` page: they
# are worth SEEING (status at a glance) but not worth reaching for, so they
# sink. Order is therefore:
# interactive UIs -> mixed -> API-only seats -> dormant
# If you add an AI group, place it by asking "would I click this?", not by
# how central the service is to the fleet.
# TOOLS — you click the card and do something in the thing it opens.
# Expanded, and placed at the TOP of its tab.
# ENDPOINTS — an API, a broker, a background agent. Its href is a `/docs`
# page, a `/ping`, or nothing at all. The only thing you want from
# the card is "is it alive". `initiallyCollapsed: true`, and placed
# at the BOTTOM of its tab.
#
# A collapsed group is not hidden — the eyebrow and its rule still render, so
# the tab still tells you the thing exists, and one click expands it. That is
# the whole point: presence without cost.
#
# When you add a service, ask "would I open this in a browser to get work
# done?" If no, it belongs in a collapsed endpoint group, no matter how
# central it is to the fleet. `AI - Inference` holds the seats the entire
# fleet runs on and it is collapsed, because you consume them through the
# gateway rather than by clicking them.
#
# Four tabs:
# Main - what you actually open day to day
# AI - AI tools up top, model/API seats collapsed below
# Infrastructure - hardware, hypervisors, BMCs (per site) — all consoles
# Toolchain - the plumbing, split by kind of plumbing
#
# ---------------------------------------------------------------------------
# GROUP MEMBERSHIP LIVES ON THE CONTAINER, NOT HERE.
#
# This block controls tab, order, columns and collapse. WHICH services are in
# a group is set by `homepage.group=` on each container's compose file, and
# labels only apply at container CREATION — moving a service between groups
# means editing the label and running `docker compose up -d <service>`, not
# `restart`. The 2026-08-24 pass did 28 of those; the playbooks that did it
# are `playbooks/homepage-regroup-<host>.yaml` and they are rerunnable.
#
# ⚠ THE MODEL SEATS ARE DELIBERATELY STILL NAMED `AI - Inference`,
# `AI - Eval & Retrieval`, `AI - Speech (TTS)` AND `AI - Audio Tools`.
# Renaming them to something like "AI API - …" would be clearer, and it would
# cost a recreate on sixteen GPU-backed seats — multi-minute model reloads on
# endpoints peers reach through the gateway. Order and `initiallyCollapsed`
# buy the same separation for free. Do not spend that recreate on a label.
#
# ---------------------------------------------------------------------------
# COLUMNS ARE 4 EVERYWHERE. DO NOT TUNE THEM PER GROUP.
#
# `columns: N` is not a density dial — it sets `lg:grid-cols-N` on that one
@@ -97,24 +118,36 @@ useEqualHeights: false
# still collapses this to 2-up and 1-up on narrow viewports, so 4 is a desktop
# maximum, not a hard floor.
#
# ---------------------------------------------------------------------------
# GROUPS ARE ALSO KEPT UNIFORM IN CARD HEIGHT, WHICH IS WHY ADGUARD AND
# TRAEFIK GOT THEIR OWN GROUPS.
#
# A service with a widget (AdGuard's query counts, Traefik's router counts,
# Uptime Kuma's uptime) renders a stat strip that makes its card ~50px taller
# than a plain link card. Put one of those in a row of three plain cards and
# you get a void under the plain ones — which is what made the old 13-member
# `Service Networking` group look broken. Split so that a group's members all
# have widgets or all do not, and every row comes out flush. That is the real
# reason `DNS & Filtering` (3 widget cards) and `Reverse Proxies` (2 widget
# cards) are separate from `Compose Consoles` (5 plain cards).
#
# ---------------------------------------------------------------------------
# EVERY GROUP NEEDS A `tab:` — including bookmark groups, and including groups
# that arrive from a `homepage.group=` container label rather than from this
# file. A group with no tab assignment renders on ALL FOUR TABS. That is how
# UltraSeedbox ended up repeated at the bottom of every tab (fixed 2026-08-18)
# and how Scriberr's `AI Systems` label did the same from 2026-08-23 (fixed
# 2026-08-24 by relabelling it into `AI - Audio Tools`). It is Homepage
# behaviour, not a bug, and it will happen again to the next container labelled
# with a group name that does not appear below.
# 2026-08-24). It is Homepage behaviour, not a bug, and it will happen again to
# the next container labelled with a group name that does not appear below.
# `GET /api/services` prints the live group list — anything in it that is not a
# key here is currently leaking onto all four tabs.
# ===========================================================================
layout:
Notes:
icon: mdi-note-text-outline
tab: Main
style: row
columns: 4
News:
icon: mdi-rss
# ---- Main: what you actually open -------------------------------------
# Replaces the old Notes (1 member) and News (2) bands, which each burned a
# full 4-wide row on a single card.
Daily:
icon: mdi-coffee-outline
tab: Main
style: row
columns: 4
@@ -123,6 +156,9 @@ layout:
tab: Main
style: row
columns: 4
# Absorbed the old one-card Games band (Pterodactyl). Lost SearXNG to Daily,
# the two chat frontends to the AI tab, and Mosquitto + the RustDesk relay to
# Agents (no UI) — neither of those has a page to open.
Apps:
icon: mdi-apps
tab: Main
@@ -133,11 +169,6 @@ layout:
tab: Main
style: row
columns: 4
Games:
icon: mdi-gamepad-square
tab: Main
style: row
columns: 4
# Bookmarks. Listed here for the tab pin above all else — without it this
# group appears on every tab. `style: row` also turns the eight entries
# from full-width stacked bars into a compact grid.
@@ -146,44 +177,53 @@ layout:
tab: Main
style: row
columns: 4
# --- AI tab: ordered interactive -> API-only -> dormant (see note above) ---
# Things you open: chat frontends, the control plane, the LiteLLM UI.
# ---- AI: tools, then collapsed endpoints ------------------------------
# Chat frontends and the control plane — Lobe Chat and the ESH Open WebUI
# joined from Main on 2026-08-24; they are chat frontends and belong with the
# other chat frontends.
AI - Gateways & Chat:
icon: mdi-router-network
tab: AI
style: row
columns: 4
# ComfyUI is a full node editor and Arbo has a real UI — both get clicked.
AI - Image & Media:
icon: mdi-image-multiple
# Replaces `AI - Image & Media`. Everything here is a workspace you open and
# produce something in: ComfyUI's node editor, Arbo, Waterland, the YT
# Voice Clipper audition console, Scriberr's transcription UI.
AI - Studios:
icon: mdi-palette-outline
tab: AI
style: row
columns: 4
# Mixed: YT Voice Clipper has an audition console, Parakeet is an API.
# Scriberr joined here 2026-08-24 (it was the stray `AI Systems` group).
AI - Audio Tools:
icon: mdi-waveform
tab: AI
style: row
columns: 4
# Below here: model seats whose href is a vLLM `/docs` page. Status at a
# glance is the whole value; you consume these through the gateway, not by
# clicking them.
# ---- collapsed from here down: seats you call, not pages you open ------
# Named `AI - Inference` rather than something clearer on purpose — see the
# warning above about what renaming these costs.
AI - Inference:
icon: mdi-brain
tab: AI
style: row
columns: 4
initiallyCollapsed: true
AI - Eval & Retrieval:
icon: mdi-scale-balance
tab: AI
style: row
columns: 4
initiallyCollapsed: true
AI - Speech (TTS):
icon: mdi-account-voice
tab: AI
style: row
columns: 4
initiallyCollapsed: true
# What is left of the old Audio Tools group after Scriberr and the YT Voice
# Clipper moved to Studios: the two ASR API seats.
AI - Audio Tools:
icon: mdi-waveform
tab: AI
style: row
columns: 4
initiallyCollapsed: true
# Stopped stacks kept for rollback / superseded seats / retired auditions.
# They stay 'created' (not running) via `docker compose up --no-start`, so
# they show here as offline cards and revive with `docker compose start`.
@@ -192,6 +232,9 @@ layout:
tab: AI
style: row
columns: 4
initiallyCollapsed: true
# ---- Infrastructure: every card is a console --------------------------
Infra - ANA:
icon: si-proxmox
tab: Infrastructure
@@ -212,13 +255,38 @@ layout:
tab: Infrastructure
style: row
columns: 4
Service Networking:
# ---- Toolchain: the plumbing, split by kind ---------------------------
# The old `Service Networking` group was thirteen members mixing three
# AdGuards, five Dockges, two Traefiks and four headless agents — widget
# cards next to plain ones next to things with no href at all. Split four
# ways on 2026-08-24.
DNS & Filtering:
icon: mdi-dns
tab: Toolchain
style: row
columns: 4
Reverse Proxies:
icon: mdi-transit-connection-variant
tab: Toolchain
style: row
columns: 4
Compose Consoles:
icon: mdi-docker
tab: Toolchain
style: row
columns: 4
Toolchain:
icon: mdi-toolbox
tab: Toolchain
style: row
columns: 4
# No href, or an href that is an API. CrowdSec, Mailrise, the restic
# rest-server, the Gitea Actions runner, the MQTT broker, the RustDesk relay.
# You never open these; you only ever want to know they are up.
Agents (no UI):
icon: mdi-cog-transfer-outline
tab: Toolchain
style: row
columns: 4
initiallyCollapsed: true