# ProdDash

A modular **production dashboard** for Waters Church. Any browser on the
building network opens one page and composes its own dashboard out of
**modules** — independent tiles that each show one aspect of a live service.
Ships with the live **ProdCom transcript**, the **ProPresenter Now / Next**
display, and a booth **clock**; new modules drop into `modules/` without
touching the shell (see [docs/MODULE-GUIDE.md](docs/MODULE-GUIDE.md)).

## How it works

`server.js` (zero dependencies, just Node.js) does four things:

1. Serves the dashboard shell (`public/`) and the admin page (`/admin`).
2. Discovers modules in `modules/` and serves each one's client at
   `/modules/<id>/`.
3. Mounts each enabled module's server routes under `/api/modules/<id>/` —
   that's where module servers follow ProdCom, poll ProPresenter, and stream
   live updates over SSE — one upstream connection per module, however many
   dashboards are open — so browsers never talk to production gear directly
   (no CORS, and API keys never leave the server).
4. Stores server-wide module config and named layouts in a per-machine data
   directory outside the app folder, so updating ProdDash never loses them.

Each browser arranges its own tiles (add, drag, resize, remove, per-tile
settings); the layout autosaves to that browser's `localStorage`. Named
layouts can also be saved to the server and loaded from any other browser.
Every module survives its upstream disappearing: tiles show a red status dot
and a clear degraded state, then recover on their own.

## Run it

Ready-made builds are on the
[Releases page](http://chronicles-ab.waterschurch.org:3000/michael/ProdDash/releases);
the files themselves live in
[ProdDash-releases](http://chronicles-ab.waterschurch.org:3000/michael/ProdDash-releases).
There is a disk image for Apple silicon Macs, one for Intel Macs, and a
bundle for 64-bit Windows. Every build updates itself afterwards from the
admin page, so the download only has to be recent, not current.

### macOS — the app

Download the disk image for your Mac — `ProdDash-<version>-arm64.dmg` for
Apple silicon, `ProdDash-<version>-x86_64.dmg` for Intel — open it and drag
**ProdDash** to Applications. That's the whole
installation: the app carries the dashboard and its own Node runtime, so
nothing else has to be installed and nothing can be missing. It lives in the
menu bar, starts and stops the server, can open at login, and holds the
permissions the modules need — the microphone the LTC listener reads, and
local network access for ProdCom and ProPresenter. macOS grants those to the
app that owns the process, so a server started any other way (over SSH, from
a launchd job) simply doesn't have them, and an LTC input goes silently to
zeroes.

Build the disk image from a checkout with
[launcher/macos](launcher/macos)/`build.sh` (`ARCH=x86_64` in front of it
for Intel):

```bash
cd launcher/macos && ./build.sh --dmg
```

### Windows — the bundle

Download `ProdDash-<version>-win-x64.zip`, unzip it somewhere you can write
to — `C:\ProdDash`, say, not Program Files, because updates write into the
folder — and double-click **ProdDash.bat**. The window it opens is the
dashboard running: it stays open, and it starts the server again when the
admin page applies an update. Allow `node.exe` on private networks when the
firewall asks, so other screens can reach it. Settings and layouts live in
`%APPDATA%\ProdDash`, so a new bundle never loses them. To start with
Windows, put a shortcut to `ProdDash.bat` in `shell:startup`. There is no
menu-bar-style launcher for Windows yet; the bundle is built with
[launcher/windows](launcher/windows)/`build.sh`.

### Anywhere else, or from a checkout

Requires [Node.js](https://nodejs.org) 18 or newer.

```bash
node server.js
```

Or double-click **Start ProdDash.command** (macOS) / **start.bat** (Windows).

Then open the URL it prints — `http://localhost:24500` on this machine, or
the `http://<this-machine's-IP>:24500` address from any other device on the
network.

## Configure

Server-wide module settings (ProdCom URL, ProPresenter host, API keys,
enable/disable) live in the **admin page** — open **Admin** from the
dashboard's top bar, or go to `/admin`. Saving re-initialises the module
immediately; open tiles reconnect on their own.

### Where settings are kept

Everything you enter in the admin page (module connection settings, API keys,
enabled/disabled flags) and every named layout is written to a **data
directory on the machine running the server** — not into the app folder:

| OS | Data directory |
| --- | --- |
| Windows | `%APPDATA%\ProdDash` (e.g. `C:\Users\<you>\AppData\Roaming\ProdDash`) |
| macOS | `~/Library/Application Support/ProdDash` |
| Linux | `$XDG_CONFIG_HOME/proddash`, else `~/.config/proddash` |

Because it lives outside the checkout, you can `git pull`, re-clone, or reset
the app folder from `main` and start the new version with everything still
configured. The admin page shows the exact path in use, and the server prints
it at startup (`Settings : …`).

The directory holds `modules.json` (module config), `layouts/` (named
layouts) and `modules/` (modules installed from the admin page), all written
by the app itself. Installed from the macOS app, it also holds `app/` — the
copy of ProdDash that actually runs, laid down from inside the app bundle on
first launch and updated in place from then on (nothing may write inside a
signed .app, and ProdDash updates itself). It may also hold a `proddash.json`
with this machine's shell settings — the same keys as the checked-in
[config/proddash.json](config/proddash.json), which supplies the defaults:

| Key | Meaning |
| --- | --- |
| `port` | Port ProdDash serves on (default `24500`). The macOS launcher writes this one, so the port it is set to holds however ProdDash is started. |
| `adminPin` | The admin PIN, as a scrypt hash. **Set it in the admin page's Admin PIN section**, not by hand — 6 digits, and only the hash is stored. Absent means admin is open to anyone on the network. |
| `adminPasscode` | An older free-text passcode. Still honoured, so an existing install keeps working; new installs should use `adminPin`. |
| `forge` | Where updates and module installs are fetched from. Absent or `"github"` uses GitHub; a Gitea or Forgejo base URL such as `"http://10.3.8.90:3000"` uses a server of your own. Set it per machine in the data directory, alongside `repo`. |
| `theme` | The dashboard palette: `booth` (default), `harbor`, `graphite`, `ember`, `daylight`, `midnight`, `moss`, `plum`, `sepia`, `blackout`, `paper` or `iron`. The admin page's **Theme** section writes this one — see [Themes](#themes). |

**Locked out?** Delete `adminPin` from the data directory's `proddash.json`
and restart, or start the server with `PRODDASH_PASSCODE` set — that passcode
overrides the PIN and always gets you in.

Only `/admin` and the admin API are behind the PIN. Dashboards stay open, so a
wall display needs no PIN, and it is unrelated to a module's own PIN (Slack's
4–8 digit send PIN, say), which guards that module's actions only.

**Updating from your own git server.** ProdDash fetches one `tar.gz` of the
branch — nothing else — so any Gitea or Forgejo works. Point `forge` at its
base URL and `repo` at the owner/repo there:

```json
{ "forge": "http://10.3.8.90:3000", "repo": "michael/ProdDash", "branch": "main" }
```

Gitea spells an archive URL differently from GitHub, which is the only thing
that changes; the tarball itself is identical, commit hash included, so update
detection and module installs behave exactly as before. Make the repository
readable without signing in, or ProdDash cannot fetch it.

Environment variables override both files: `PORT`, `PRODDASH_PASSCODE`, and
`PRODDASH_FORGE`, and `PRODDASH_DATA_DIR` to put the data directory somewhere else (a relative path
is taken from the app folder; `PRODDASH_DATA_DIR=config` restores the old
in-repo location). If the default directory can't be created, the server says
so and falls back to the app's `config/` folder.

**Upgrading from an earlier version:** the first start copies an existing
`config/modules.json` and `config/layouts/` from the app folder into the data
directory, so nothing has to be re-entered. The old files are left in place and
ignored from then on.

### Themes

The whole dashboard — every tile, on every browser — draws with one palette,
chosen in the admin page's **Theme** section and switched live: pick a swatch
and open dashboards change without a reload. The section starts folded —
a building picks its palette once — and holds twelve presets:

| | |
|---|---|
| **Booth** | the default dark, mint accent |
| **Harbor** | deep navy, cyan |
| **Graphite** | neutral charcoal, electric blue |
| **Ember** | warm broadcast black, orange |
| **Daylight** | a cool light palette, for a front-of-house desk in daytime |
| **Midnight** | indigo night, violet |
| **Moss** | forest floor, lime |
| **Plum** | aubergine, purple |
| **Sepia** | warm brown room, cool teal — Ember's arrangement reversed |
| **Blackout** | true black, cyan, for a dark room where any glow carries |
| **Paper** | the warm light one, indigo accent |
| **Iron** | no colour of its own: a near-white accent, so the only hues on screen are a module's own and the red and amber that mean something |

Each keeps the status colours apart — the accent is never near the danger red
or the warning amber — and muted text readable in a dark booth. The choice is
stored as `theme` in the data directory's `proddash.json` (table above), so
it holds across restarts and updates; a browser remembers the last theme it
saw and applies it before the server answers, so nothing flashes. Modules
style with the shell's colour variables, which is why they follow; a colour
set on a tile itself (a PCO Plan header colour, say) is deliberately kept.

## Installing modules and updating

Everything below happens in the **admin page**; ProdDash talks to its git
repository — `michael/ProdDash` on the church's own Gitea at
`chronicles-ab.waterschurch.org:3000`, branch `main`. Change it with `forge`,
`repo` and `branch` in `proddash.json`, or the `PRODDASH_FORGE` and
`PRODDASH_REPO` variables (`owner/repo#branch`). It downloads one small archive of the branch, at most
every few minutes, only while the admin page is open or for the update check
below. Nothing is installed without someone clicking.

- **Available modules** lists modules in the repo that aren't installed
  here. **Install** downloads the module into the data directory's `modules/`
  and starts it right away — nothing is written into the app folder, and no
  restart is needed. Modules that ship with your copy of ProdDash are
  "bundled"; ones you installed are "installed"; if both exist, the newer
  version runs.
- **Uninstall** on a module card removes an installed module's files. For a
  bundled module (its folder belongs to the app) it hides the module from
  every dashboard instead; it reappears under Available modules and can be
  brought back without downloading. Settings are kept either way.
- **Module updates**: when the repo has a newer version of an installed
  module, its card shows **Update to vX**.
- **Version requirements**: every module declares the ProdDash versions it
  needs (`"proddash": ">=1.1.0"` in its manifest). A module that needs a
  newer shell is listed but not loaded, and cannot be installed until
  ProdDash is updated.
- **ProdDash updates**: the admin page checks the repo shortly after start
  and every six hours, and shows a banner when `main` has a newer version
  (or new commits). **Update now** brings the app folder up to date and
  restarts the server; dashboards reconnect on their own. A git checkout is
  updated with `git pull --ff-only` (only when it is on the repo branch with
  no local changes — otherwise the banner says why and you update with git).
  A plain folder (no `.git`) has the repo archive unpacked over it; files
  removed upstream are left behind but harmless. Settings and layouts are
  never touched — nothing inside the data directory is written, even on an
  install where that directory *is* the app's own `config/`. **What changed**
  opens the commit comparison on the forge.
- **If an update goes wrong**: the download has to parse before anything is
  swapped in, so a truncated or mangled `server.js` is refused with the
  reason and the running copy is left exactly as it was. Past that, the copy
  being replaced is set aside as `app.previous` next to the app folder, and
  dropped once the new version has stayed up for ninety seconds — so that
  folder existing means "the last update has not proven itself yet". The
  macOS app watches for a version that will not stay up (four stops in two
  minutes) and puts the old one back by itself, once, with a notification.
  Running some other way, the way back is a rename: `mv app.previous app`.
- **Restarting**: after an update the server exits with code 75 and starts
  its own replacement, which waits for the port to free up. If you supervise
  ProdDash with your own launcher, set `PRODDASH_LAUNCHER=1` so the server
  just exits and your launcher restarts it on exit code 75.
  `PRODDASH_NO_REMOTE_CHECK=1` turns off the periodic check (the admin page
  still checks when opened).

## Using the dashboard

- **＋** — add any enabled module; multiple tiles of the same module are
  fine (e.g. two transcripts filtered to different channels).
- Drag a tile by its **title bar** to move it; **✕** removes it.
- **Resize from any edge or bottom corner**, the way a window resizes:
  nothing is drawn on a tile to mark the spots — the cursor over an edge is
  what says it can be dragged. An edge moves only itself, so the tile grows
  or shrinks into the space beside it while the opposite edge stays put,
  stopping at a neighbour or at the module's own minimum size. Two
  exceptions, both of which move a *shared* edge instead: with another tile
  directly below, the bottom edge trades rows between the two so the column
  keeps its total height; and where two tiles meet side by side, the seam
  between them trades width (macOS split-view style), showing itself only
  while you drag it.
- **Fullscreen** (the ⛶ button) is a viewing mode: the menu bar hides (its
  green lip at the top-left brings it back, minus the Admin button), the
  occupied columns stretch to the full width, and every column runs to the
  bottom of the screen — the last tile in each column is stretched to end
  there. A short tile like the clock is never stretched into a ribbon: under
  another tile it slides to the bottom edge and the tile above grows into the
  gap; alone in its columns it stays as it is. Nothing is saved — the layout
  reverts when you leave fullscreen. A layout deeper than the screen keeps
  its normal row height and scrolls.
- A tile's actions (menus, text size, clear…) live in its **title bar**.
- **Auto-hide bars** (the button left of ⛶) folds every title bar away, for
  a windowed screen that should read like a wall display. A folded bar is
  not a gone bar: point at the tile's top edge and it slides back in over
  the content, where its buttons click and its width drags the tile exactly
  as before — so a dashboard can be arranged with every bar out of the way.
  The switch is off on a fresh dashboard and saves with the layout —
  including a named layout on the server, so a booth loads its screen with
  the bars already tucked away. Fullscreen folds every bar regardless, and
  the same hover brings one back there too.
- The **pin** beside a tile's title keeps *that* tile's bar showing while
  auto-hide is on — for the one or two tiles whose controls are wanted at
  a glance. It turns the accent colour when it is on and saves with the
  layout. Fullscreen still folds a pinned bar with the rest.
- The menu bar's own handle is the **ProdDash
  name itself**, wrapped in a translucent green pill at the left — click it
  to hide the bar; the slim green lip it leaves at the top-left edge brings
  the bar back. It stays at the left so it never sits on top of a tile's
  own tab.
- The **gear** on a tile holds per-tile settings (filters, text size,
  display options) — these belong to this browser's layout, not the server.
- The **grid icon** — save the current arrangement to the server under a
  name, load a named layout (it becomes a local copy — browsers are never
  linked), or reset this browser's dashboard. Loading says what it did —
  *Loaded "FOH": 9 tiles, 2 not shown — Service, #ops aren't offered right
  now* — with **Undo** in the message to get back the arrangement you had.
  A tile whose module or picker entry isn't offered at the moment keeps its
  place and shows nothing, then appears when it is; it is never replaced by
  the module's default tile. Saving over an existing name asks first, and
  the server keeps the copy it replaced: **Restore previous** in Admin →
  Saved layouts swaps it back (and swapping again undoes that). The menu's
  **Grid Size** stepper sets how many columns this browser cuts the board into —
  8 to 24, 12 by default. Finer lets tiles be smaller so more fit across (a
  4K wall), coarser suits a small screen read close up; row height follows,
  so tiles keep their shape and simply change size. Going coarser, a tile
  that no longer fits where it was moves to the first free spot rather than
  onto another tile. It saves with the layout, so a named layout comes back
  at the density it was arranged on.
- Every tile has a **status dot**: green = live, amber = connecting,
  red = upstream unreachable (hover it for details). Tiles reconnect
  automatically.

## Included modules

- **ProdCom** — live transcript stream with in-place updates of in-progress
  speech, channel colors, per-tile channel visibility toggles and group
  presets, timestamps, text size, a top-down / bottom-up direction toggle
  and jump-to-latest in the title bar.
  Admin config: ProdCom server (host : port), API key.
  Tile settings: channel icons, direction and timestamps.
- **ProPresenter** — what's live and what's next, with slide progress,
  driven by the shared `propresenter-core` client (ProPresenter 7/20).
  The ＋ picker offers two tiles: **Now / Next**, the full card, and **Now**,
  the live item alone with the gradient filling the tile edge to edge — a
  wall panel that says one thing.
  Admin config: ProPresenter endpoint (host : port), password.
  Tile settings: a **Show** group of switches for the playlist name, live
  status, NOW and NEXT labels, slide count, progress bar, the next list and
  its item tags, plus tidy titles.
- **Timers** — live ProPresenter timers (countdown, countdown-to-time,
  elapsed) and LTC timecode as auto-sized cards. The ＋ picker lists every
  discovered timer (and the LTC card) as its own tile, or add "All timers"
  and pick a selection via the tile's "Timers ▾" menu; overruns are
  unmistakably red. Read-only — no start/stop controls. Admin config is in
  three groups: **ProPresenter** (endpoint host : port, API password),
  **LTC Timecode** (the listener switch, and — when it's on — the audio
  device and channel), and **Display** (Timer font — Default or Monospace,
  which keeps the digits from shifting as the numbers change).

  ProPresenter exposes no timecode over any API, so LTC is decoded from the
  audio signal, behind the admin **LTC listener** switch (off ⇒ no LTC tile
  in the picker at all). On, the module captures the selected audio input
  device/channel on the ProdDash machine — a native `ltc-capture` helper,
  compiled on first use — and decodes the SMPTE frames in-process:
  running/stopped/no-signal, the timecode counting frame-accurately in real
  time, and the frame rate (incl. 29.97 drop-frame). macOS note: the
  ProdDash server needs microphone permission — launch it from a logged-in
  session (Start ProdDash.command) and approve the one-time prompt; a server
  started over SSH or by bare launchd silently captures zeros. (Engine
  self-test: `node modules/propresenter-timers/ltc.selftest.js`.)
- **PCO Plan** — Planning Center Services, as two tiles in the ＋ picker:
  **Plan** (the order of service) and **Teams** (who is scheduled).

  **Plan** shows every
  plan item with its scheduled start time, length, type, song key,
  description and item notes. When the ProPresenter module is running it
  follows that module's live item, highlights the matching plan item, dims
  what's done and shows the elapsed vs planned runtime and how far
  ahead/behind the service is. Items are matched by name — a document named
  like the Planning Center item or its song, tags and all. The message is
  the exception: a sermon's document is named after the sermon, so a live
  document that reads like the plan itself — its date code, series title
  and part, as in `260927 Where is Your Faith? - Part 6` — is taken as the
  message and lands on the plan's message item (an item titled Message,
  Sermon or Teaching, else the first item under such a header). Admin
  config is the connection: Planning
  Center credentials (Personal Access Token app ID + secret, or an OAuth
  token), the service type (a list grouped by your Planning Center folders)
  and an optional pinned plan. Everything about the display is per tile
  (gear menu): the top-bar text (placeholders such as
  `{series} • {part} | {serviceType} • {date}`), colors, text size, and a
  show/hide checklist for every detail.

  **Teams** shows the same plan's team and position assignments — each team
  as a band of its people, with their photo (fetched through ProdDash, never
  by the browser; initials when Planning Center has none) and a dot for
  confirmed, unconfirmed or declined. Its gear menu picks which teams and
  which positions appear, groups people under their position or lists them
  plainly, and sets the colours: one for the team headers, or **Tint teams**
  to give each team its own, optionally carried into the rows. Teams nobody
  has signed up for are hidden until you ask for them — that list comes from
  the service type, since Planning Center only returns a team once somebody
  is on it.
- **Slack** — live transcripts of this location's Slack channels, plus
  PIN-protected sending as the location. The ＋ picker lists one transcript
  tile per configured channel, titled `#channel`, and two send surfaces (one
  of each per dashboard): **Send message**, a free-text box with the admin's
  quick emojis one press away, and **Quick replies**, a grid of preset
  buttons. Both post to the channel of whichever transcript was clicked last
  — its bottom edge lights up in the accent color — and to the first
  configured channel until one is clicked. Messages arrive with avatars,
  names and times, Slack's inline formatting and the workspace's custom
  emoji; a right-click on one opens the reaction menu, and text size and
  jump-to-latest sit in the title bar. Nothing leaves the machine until the
  PIN is entered: one PIN unlocks every Slack tile on that dashboard page
  until the auto-lock idles out, a reload asks again, and the lock button on
  a send tile's title bar ends it early.
  Admin config is in four groups: **Slack account** (the location account's
  User OAuth Token — `xoxp-…`, so messages are literally from the campus —
  or a bot token plus "Send as"), **Channels** (the joined channels to
  follow, one transcript tile each), **Transcript** (refresh interval,
  messages kept, thread replies) and **Sending** (the switch on the group's
  title row, a 4–8 digit PIN, auto-lock minutes, quick replies one per line
  as `Label | Message` — add `| #channel` to bind one to a channel — and up
  to eight quick emojis).
  Tile settings: avatars, timestamps, names and channel notices.
- **Resi Cues** — how long until the next Resi cue point, for a campus
  running video from another site. The tile is one number, as large as it
  will go: the countdown to the next cue, its name above it, and the playhead
  and the cue just passed underneath. It also offers that countdown and the
  playhead to the **Timers** module, so they can sit among the ProPresenter
  timers instead of in a tile of their own.
  Both the cue list and the decoder's position come from Resi's cloud rather
  than the decoder — the decoder reports upstream and central.resi.io hands it
  back — so nothing is installed at the decoder end. Resi republishes a
  position only every few seconds and each reading is already stale, so the
  playhead is estimated between readings rather than repeated; a countdown
  built straight from Resi's number sits seconds behind and moves in lurches.
  Admin config: Resi login and password, which decoder to follow (the list is
  read from the account), and a folded **Timing** group for the poll
  intervals and a position trim.
  Tile settings: cue name, previous cue, playhead, cue list.
- **Band Lineup** — the backstage stage board: everyone on stage in their mic
  slot with photo, name and mic/pack label, on one shared board that every
  screen sees the same way — whoever edits it changes it everywhere, over SSE.
  A photo library fills it: point the module at a Google Drive folder shared
  "anyone with the link" (file name → person's name) or at a folder on the
  ProdDash machine, and photos added or removed there follow. **Edit** in the
  title bar opens that library — drag a photo onto a slot, drag between slots
  to swap, rename or clear. Planning Center is optional auto-fill on top, not
  the source of truth: **Sync** reads the upcoming plan's team members and
  drops each into the slot matching their position, filling only slots nobody
  placed by hand, so a guest pinned on Thursday survives Sunday's sync.
  Birthdays in the service week (Wednesday→Tuesday) bring a gold glow and
  raining confetti, and a Choir button adds the extra spot for the weeks there
  is one. Admin config: Planning Center credentials, service type and plan, the
  teams to include, the stage slots (one mic/pack label per line, in stage
  order), default columns, and the photo folder. Tile settings: parallax,
  names, mic labels, and birthday confetti with its intensity and a test
  picker — columns are − / + in the title bar.
- **Service Times** — every campus's Sunday service times, read from the
  church website and re-read on a timer, so the board follows the site when
  the times change through the year. Any service can be marked for
  production — issue, replay, or do not use — with a note, and a campus
  taking resi records which of the origin's services it is playing. Marking
  a broadcast service "do not use" flags every campus riding on it, which is
  the question the board exists to answer. Marks are shared by every tile
  and every browser, survive a restart, and expire on their own so last
  week's never greet a new Sunday. The ＋ picker offers the whole board or a
  single campus.
  Admin config is in four groups: **Website** (address, which campuses to
  follow, refresh interval), **Broadcast** (the campuses the broadcast comes
  from, and each other campus's resi source), **Manual campuses** (a stopgap
  for a campus the site has not published yet — the website takes over as
  soon as it does) and **Marks** (how long a mark lives).
  Tile settings: campuses, card or list layout, and a **Show** group for the
  day, role, resi source, campuses fed, notes and the updated line.
- **Clock** — a big booth clock; also the smallest possible module and the
  reference for module authors.

To develop modules against mocks instead of real gear:
`node tools/prodcom-mock.js` (ProdCom on `127.0.0.1:24480`),
`node band-lineup-display/tmp/pp-mock.js 1599` (ProPresenter slides on
`127.0.0.1:1599`), `node tools/pp-timers-mock.js 1600` (ProPresenter
timers + timecode on `127.0.0.1:1600`), `node tools/pco-mock.js`
(Planning Center Services on `127.0.0.1:24700` — run ProdDash with
`PCO_PLAN_API_BASE=http://127.0.0.1:24700`) and
`node tools/slack-mock.js 24716` (the Slack Web API on `127.0.0.1:24716` —
run ProdDash with `SLACK_API_BASE=http://127.0.0.1:24716/api`) and
`node tools/pco-band-mock.js 24710` (Planning Center for the Band Lineup
board — a plan, its team members, birthdays and photos — run ProdDash with
`BAND_LINEUP_API_BASE=http://127.0.0.1:24710`). See the testing section of
the [module guide](docs/MODULE-GUIDE.md).

## Install as a desktop web app

Open the page in a browser first, then:

- **macOS — Safari**: File → **Add to Dock…**
- **macOS/Windows — Chrome**: click the install icon in the address bar, or
  ⋮ menu → Cast, save and share → **Install page as app…**
- **Windows — Edge**: ⋯ menu → Apps → **Install this site as an app**

Note: Chrome only offers the one-click PWA install on `http://localhost` (a
"secure context"). When opening from *another* machine over plain HTTP, use
Edge's "Install this site as an app" or Chrome's ⋮ → Cast, save and share →
"Create shortcut…" (check "Open as window") — same result: a dock/taskbar
app in its own window.

## Repository layout

```
ProdDash/
├── server.js                  shell server: static files, module discovery/mounting,
│                              admin + layouts APIs
├── public/                    dashboard shell client (index.html, shell.js, style.css)
│   └── admin/                 admin page client
├── modules/                   installed modules, one folder each
│   ├── clock/
│   ├── prodcom-transcript/
│   └── propresenter-now-next/ (carries its own copy of propresenter-core)
├── package.json               name + version (what the update check compares)
├── config/                    proddash.json defaults (runtime state lives in the
│                              per-machine data directory — see "Configure")
├── docs/MODULE-GUIDE.md       how to build and install a module
├── tools/prodcom-mock.js      mock ProdCom API for development
├── prodcom-listener/          reference fork — read-only, not served
└── band-lineup-display/       reference fork — read-only, not served
```

The two reference folders are snapshots of the apps ProdDash grew out of;
they are not wired into anything (but `band-lineup-display/tmp/pp-mock.js`
is handy for testing).