An interactive, artist-made composition portfolio with robust PDF and playback support.

Praetorius is a small, disciplined toolchain for composers, artists, educators, festivals, and universities to publish “score-centric” works pages: synchronized audio, score PDFs with page-follow (auto-snapping to the current printed page), deep links, and clean UIs that drop into Squarespace (or any site builder) via a single code block.

the score PDF snaps to the correct printed page while audio plays, so evaluators aren’t lost—and you aren’t judged on secondary characteristics.

Presenting—or judging—score-based works on the web often collapses into: (1) the evaluator doesn’t know the score; (2) they get lost between time and page; (3) judgments drift to secondary aesthetics. Praetorius restores primary evidence: the work as score + sound, tightly synchronized.

• PDF page-follow: printed page ↔︎ media-time mapping.
• Works set authoring: a wizard writes your DB; a generator emits paste-ready assets.
• Squarespace-friendly: paste one snippet or host the assets—no theme collisions.
• Deep links: link straight to a work, with an optional start time.
• Light/Dark: strict tokens for predictable contrast.

Non-goals: hosting your media, analytics, auth. You own your assets; Praetorius orchestrates them.

# Global (recommended for authoring)
npm i -g praetorius

# Or one-off
npx praetorius@latest --help

mkdir my-portfolio && cd my-portfolio
prae init --out dist

This seeds .prae/works.json (if missing) and a minimal dist/ scaffold you can paste or host.

prae add
# Answer prompts for title, one-liner, audio, pdf, cues, and (optionally) page-follow mapping

# default skin = console
prae generate --out dist

# pick a specific skin
prae generate --skin console       --out dist
prae generate --skin vite-breeze   --out dist
prae generate --skin docs-reader   --out dist
prae generate --skin cards-tabs    --out dist
prae generate --skin kiosk         --out dist
prae generate --skin typefolio     --out dist
prae generate --skin typescatter   --out dist

# single pasteable snippet (for a Code block)
prae generate --embed > embed.html

• Squarespace (and similar): add a Code block and paste the embed or include the generated JS/CSS you’re hosting. This is the intended workflow; it avoids theme collisions.
• Static/JAMStack (any host): copy dist/* and reference in your template.
• PDF viewing: use the PDF.js viewer (?file=<url>) for reliable page navigation; Drive “preview” iframes don’t expose page control—see PDF Page-Follow below.

Minimal, human-readable schema:

{
  "version": 1,
  "works": [
    {
      "id": 1,
      "slug": "soundnoisemusic",
      "title": "String Quartet No. 2 — SOUNDNOISEMUSIC",
      "one": "A through-composed/indeterminate quartet…",
      "audio": "https://…/audio.mp3",
      "pdf": "https://…/score.pdf",
      "cues": [ { "label": "@10:30", "t": 630 } ],
      "score": {
        "pdfStartPage": 11,
        "mediaOffsetSec": 0,
        "pageMap": [
          { "at": "0:30", "page": 1 },
          { "at": "1:00", "page": 2 }
        ],
        "pdfDelta": 0
      }
    }
  ]
}

Notes

• id (int), slug, title, one (short one-liner).
• audio (URL or null), pdf (URL or null).
• cues[] may be { label, t } (seconds) or @mm:ss.
• score holds printed→PDF mapping: see next section.

Goal: While audio plays, the PDF viewer jumps to the correct printed page.

• pdfStartPage: the PDF page number that corresponds to printed page 1.
• mediaOffsetSec: if the recording’s bar 1 doesn’t start at 0:00 (pre-roll, announcements), offset can be positive or negative.
• pageMap: waypoints like { at: "7:49", page: 10 } defined in printed page numbers.
• Optional pdfDelta: adjust all computed PDF page numbers by a constant.

Praetorius renders the same data across different skins. Choose one at generate time with --skin. (You can regenerate in another skin at any time—your data stays the same.)

See the ui/ directory in the repo for each skin’s source and templates. (Skins currently listed in the repo include at least vite-breeze, docs-reader, cards-tabs, kiosk; additional skins here reflect the current codebase you’re generating with.)

All commands are available via praetorius or its short alias prae.

• prae init [--out dist] — create seed files (.prae/works.json, minimal output).
• prae add — add a new work (guided prompts).
• prae edit — edit an existing work (by id/slug; guided).
• prae rm — remove a work.
• prae order — reorder display order.
• prae list — print current DB for inspection.
• prae import <file.(json|csv)> — import works.
• prae export [--format json|csv] — export works.
• prae generate [--skin <name>] [--out dist] [--embed] — emit assets/snippet.
• prae doctor — validate schema, mapping, and links; print actionable guidance. (Doctor++ checks are implemented; use this when something feels off.)

Authoring is intended to be host-agnostic: generate static assets (dist/*) or a single --embed snippet you paste into a site builder.

• Squarespace/WordPress/Webflow/etc.: paste the embed or include your hosted script.js/styles.css. This avoids theme collisions and works anywhere a Code/HTML block is allowed.
• PDF hosting: prefer your own origin (or a CDN you control). When using Google Drive, convert links as noted under PDF Page-Follow.

You can link directly to a work section (e.g., #work-<slug>) and optionally include a start time query (e.g., ?t=630 for 10:30). Your UI can read these on load and activate the correct section + seek the audio.

• ARIA: preview/hud announce via aria-live; controls are labeled.
• Keyboard: arrow keys to change selection; Esc closes drawers.
• Performance: audio loads lazily; PDF viewer is iframed to isolate work.
• Color: strict light/dark tokens for predictable contrast.

404s for style.css/main.js (during dev or embed testing) If you see GET http://localhost:5173/style.css 404 or similar, you’re pointing at a dev server that isn’t running or a path that doesn’t exist in production. When embedding, either paste --embed output directly or host dist/* on your site and reference those hosted URLs.

PDF page-follow doesn’t work Use cdn-hosted links URL with a proper file= parameter (not a Drive “preview” URL). Verify your score mapping is monotonic (i.e. times increasing; pages ≥ 1).

Issues and PRs welcome—open an issue with a minimal repro (DB sample, links). Unit tests for time parsing and score validation are especially helpful.

MIT. © Seb Suarez.