Phase 1: Bulkhead core editor, live on the subdomain

App

Bulkhead — browser deck-plan maker for sci-fi TTRPGs. Committed today (see FACTS.md). Store: itch.io downloadable build. Demo: bulkhead.bananafest-destiny.com.

Why this phase

Nothing sells until a stranger can draw a ship and get a file out. Phase 1 is the smallest editor that produces something a GM would actually print or drop into a VTT.

Scope (done means all of these)

1. app/ scaffold: single-page vanilla TS + Vite, no framework, no backend. Builds to dist/. Also builds a single-file bulkhead.html (all JS/CSS inlined) — that file is the itch.io download.
2. Editor, squares grid (default 5 ft / 1.5 m per square, configurable): - Tools: room (drag rectangle, auto-walls), corridor, erase, door / airlock / hatch, label, hull outline auto-derived from filled cells. - Pan/zoom, undo/redo, keyboard shortcuts. - Styles: "blueprint" (white on blue) and "print" (black on white).
3. Save/load as .bulkhead.json (local file, no server). Autosave to localStorage.
4. Export: - Universal VTT (.dd2vtt): PNG + wall segments from room edges, door portals from door/airlock cells, lights from lamp markers. Verified by importing into Owlbear Rodeo (free) with the uvtt-importer extension and confirming walls block fog and doors open. - PNG at chosen px-per-square (e.g. 70 px / 100 px). - PDF, A4/Letter, tiled at 1"-per-square for minis.
5. Deployed to bulkhead.bananafest-destiny.com via Workers static assets with the cider token. Opened in a browser and exercised end to end by me.
6. Demo build gate: the hosted demo watermarks exports; the downloadable build does not. (Same code, one build flag.)

Not in this phase

Furniture stamps, multi-deck ships (Phase 2 — this is the second differentiator vs NEWT), text tables/room keys, share links, itch.io listing copy, YouTube demo. Those are Phase 2 (stamps + keys + multi- deck) and Phase 3 (listing, video, launch).

Asks for the boss (batched)

1. GitHub remote URL for this repo — no remote is configured, so I cannot push. Blocks: publishing the log. Meanwhile: committing locally.
2. Open the itch.io account when convenient — not blocking Phase 1; needed before Phase 3. Nothing else.

Risk

NEWT Blueprint Builder (free, 5.0★) covers the basic editor. Bulkhead only earns money if the VTT export is flawless on first import; that is the acceptance test for this phase, not a nice-to-have.

PDF generation client-side: will use pdf-lib or jsPDF — decide by testing which handles large raster + vector lines cleanly. If both are poor, fall back to "print to PDF" via a print stylesheet.

Plan: plan/2026-09-14.md. Status: done, with one deviation noted below.

What shipped

• Live: https://bulkhead.bananafest-destiny.com (Workers static assets, custom domain on the zone). 404 page works.
• app/ — vanilla TS + Vite, no framework, no backend, ~22 KB of JS. Two builds: dist/ (demo, deployed) and dist-single/index.html (paid, one 33 KB self-contained file — the itch.io download).
• Editor: room (drag rectangle), corridor (paint), erase, door / airlock / hatch (click a wall; airlocks only on hull), label, light. Pan/zoom, undo/redo (60 steps), keyboard shortcuts, help dialog. Walls are derived, never drawn: any boundary between different regions is a wall, void boundaries are hull. Two styles: blueprint, print.
• Save/open .bulkhead.json (RLE cells); localStorage autosave survives reload.
• Exports: PNG at N px/square; Universal VTT .dd2vtt (image + walls + door portals + lights, Dungeondraft 0.3 layout); vector PDF via a 90-line hand-rolled writer — fit-to-page or tiled at 1″ / ½″ per square with page/col/row footers. Demo watermarks all three; paid does not.

How it was verified (by me, today)

• Playwright/Chromium against localhost, then against the live URL, then against the paid file opened from disk (file://): draw room, paint corridor, undo, redo, add label, switch style, export all four files, save, reload → autosave restored. Zero console errors in all three runs.
• .dd2vtt structure inspected: 39 walls, 6 portals, 2 lights, 2300×2600 PNG.
• Acceptance test: ran the Owlbear Rodeo uvtt-importer extension's own source (isUniversalVTTData, createWallItems, createDoorItems, createLightItems, with the OBR SDK bundled and the player id stubbed) against my file: validated true; produced 39 FOG-layer wall paths on exact grid pixels, 6 doors with rodeo.owlbear.dynamic-fog/doors metadata (closed), 2 lights at 400 px = 4 squares. Also read Foundry's Universal Battlemap Importer source: light colour is AARRGGBB ("#"+color.substring(2)), closed:true → door, range × grid distance → dim radius. My output matches both.
• PDFs rasterised with pdftoppm and eyeballed; page count correct after a float-rounding fix (fit-to-page briefly produced 2 pages).

Deviation from plan

• Plan said "verified by importing into Owlbear Rodeo". Creating an OBR scene needs a signed-in account, which I don't have and won't open. I verified against the importer's code instead (above). A live OBR import is still worth doing once before launch — it goes in the Phase 3 ask list, not now.

Fixed along the way

• Em-dash / middle-dot rendered as "?" in PDF (WinAnsi mapping added).
• Watermark was too large on big exports (4.5% → 2.2% of the short side).
• Deployed once without rebuilding — caught by re-running the live test.
• Starter ship had an airlock on an interior wall; the airlock tool itself already refused that, the seed data didn't.

Not done (as planned)

Furniture stamps, multi-deck, room keys, itch listing, YouTube demo.

Next

Phase 2: multi-deck ships (deck tabs, ladders/lifts linking decks, one file), a small stamp set (consoles, bunks, crates, cryopods, reactor, airlock markings), room key export. Then Phase 3: itch page + video + launch.

Asks

None new. Still open: itch.io account (before Phase 3).