ProdDash/CLAUDE.md

ProdDash

Modular production dashboard web app for Waters Church, developed in this repo. The app lives at the repo root and is built: zero-dependency Node server (server.js, Node 18+), dashboard shell client in public/, admin page at /admin, modules in modules/, and a native macOS menu-bar launcher in launcher/macos/ (Swift, no dependencies — the shipped macOS app: it carries the server and its own Node runtime, supervises them, and owns the macOS permissions modules declare; see its README). Runtime state (module config, named layouts) is written to a per-machine data directory outside the checkout (PRODDASH_DATA_DIR overrides; see README "Where settings are kept") — never add code that writes runtime state into the repo folder. SPEC.md was the authority for the initial build; README.md has run/configure/use instructions.

Building a new module — read docs/MODULE-GUIDE.md first

docs/MODULE-GUIDE.md is the authority on the module contract (manifest, client factory + moduleApi, optional server entry, title-bar controls). The rules that matter most:

• A module is one self-contained folder in modules/ — never modify the shell (server.js, public/) to add one. Installation = the admin page's Available modules (downloads from the repo into the data directory), or drop the folder and restart; configure in /admin.
• All settings in the admin page (configSchema / instanceSchema) — never a separate setup page, hand-edited file or env var. Dependencies ship inside the module folder (zero deps is the norm). Everything the module does runs inside its server entry — no helper scripts, companion programs or daemons the operator starts by hand; a listener or decoder is started in init() and stopped in stop(). Declare "proddash": ">=x.y.z" and bump the module's version on every change.
• Browsers never call upstream services directly; the module's server routes proxy everything (no CORS, secrets stay server-side).
• Style with the shell's CSS variables only, scoped under [data-module="<id>"]; survive upstream loss (reconnect + setStatus).
• Verify against the guide's testing section and resilience checklist. Mocks for the existing upstreams: node tools/prodcom-mock.js (ProdCom) and node band-lineup-display/tmp/pp-mock.js 1599 (ProPresenter). node tools/slack-mock.js 24716 (Slack; run ProdDash with SLACK_API_BASE=http://127.0.0.1:24716/api).

Reference code (read-only forks — don't develop these here)

• prodcom-listener/ — live ProdCom transcript web app. Zero-dependency Node server.js that serves public/ and proxies /prodcom/* (including the SSE transcript stream) to the ProdCom API at config.json → prodcomUrl (set per site; each campus runs its own). Good reference for: SSE handling, reconnect logic, channel colors/filtering, PWA install.
• band-lineup-display/ — backstage Electron app (stage lineup photos with mic/pack labels) plus ProPresenter Now/Next:
  • band-lineup-display/propresenter-core/ — shared ProPresenter 7/20 client, no deps, no Electron. All ProPresenter quirks live here (PP20 runtime route discovery, arrangement-expanded slide totals, uuid-then-name live-item matching). Reuse this rather than re-implementing ProPresenter access.
  • band-lineup-display/propresenter-app/ — standalone Now/Next web display.
  • band-lineup-display/tmp/pp-mock.js — mock ProPresenter API for testing without the real thing: node tmp/pp-mock.js 1599, then point at 127.0.0.1:1599.

Upstream sources: ProdCom Listener lives at github.com/michaelmcgary2008/prodcom-listener; BandLineupDisplay exists only on the production machine. These folders are snapshots, not submodules.