kitchen/
├── scripts/
│   ├── generate-manifest.ts   # static parser → feature-manifest.json
│   └── validate-manifest.ts   # consistency checks
├── src/
│   ├── bun/index.ts           # Electrobun bun-side entrypoint
│   ├── generated/
│   │   ├── manifest-types.ts  # TypeScript types for the manifest
│   │   └── feature-manifest.json  # generated feature index
│   ├── test-framework/        # test executor & result types
│   ├── test-runner/           # test runner UI (views://test-runner/)
│   ├── test-harness/          # webview used by automated tests
│   ├── tests/
│   │   ├── index.ts           # aggregates all suites
│   │   ├── *.test.ts          # automated test suites
│   │   └── interactive/       # interactive test suites
│   └── playgrounds/           # standalone feature demo windows
└── electrobun.config.ts

Running the Kitchen Sink

cd kitchen && bun install

# Dev mode — opens test runner window
electrobun dev

# Watch mode — auto-rebuilds on file change
electrobun dev --watch

# Auto-run all automated tests and exit (CI / headless)
AUTO_RUN=1 electrobun dev

# Run a single test by name
AUTO_RUN_TEST_NAME="Window creation with URL" electrobun dev

Exit codes: 0 = all pass, 1 = any failure.

Feature Manifest

Generated at kitchen/src/generated/feature-manifest.json. Each entry:

{
  "id": "window-creation-with-url",         // stable slug
  "title": "Window creation with URL",      // human-readable
  "category": "BrowserWindow",              // API grouping
  "description": "Test creating a window with a URL",
  "interactive": false,                     // true = requires human input
  "testFile": "src/tests/window.test.ts",   // source location
  "playgroundRoute": null,                  // e.g. "playgrounds/clipboard/index.html"
  "apiSurface": ["BrowserWindow"],          // Electrobun APIs exercised
  "uiSelectors": [],                        // CSS selectors used in playground
  "platformCaveats": []                     // e.g. ["requires CEF renderer"]
}

Regenerate / Validate

cd kitchen
npx tsx scripts/generate-manifest.ts   # parse defineTest() calls, write manifest
npx tsx scripts/validate-manifest.ts   # check all tests have manifest entries, no dups

Generator does static analysis only — no Electrobun runtime required.

Test Suites

Automated (run unattended)

Suite	Key APIs
RPC	BrowserView.defineRPC, Electroview.defineRPC
BrowserWindow	new BrowserWindow, all lifecycle methods
Navigation	setNavigationRules, will-navigate, did-navigate
Utils	clipboard, paths, openFileDialog, showNotification
Screen	getPrimaryDisplay, getAllDisplays, getCursorScreenPoint
Session	Session.fromPartition, cookies
Events	Electrobun.events.on (all global events)
Preload	preload script injection
Updater	checkForUpdate, downloadUpdate, applyUpdate
Sandbox	sandbox: true isolation
Tray	new Tray, setMenu, tray-clicked
WGPU FFI	WGPUBridge low-level FFI
WGPU Adapter	GpuWindow, WGPUView adapter layer
Babylon Adapter	BabylonJS WebGPU integration
WGPU Adapter Extended	extended WGPU surface coverage

Interactive (require human)

Suite	Feature area
Dialogs	openFileDialog, showMessageBox
Tray	Tray visibility, icon, menu
Shortcuts	GlobalShortcut register/unregister
Webview Tag	`<electrobun-webview>` masks, passthrough, navigation
Clipboard	clipboardRead/WriteText/Image
Menus	ApplicationMenu, ContextMenu
Window Events	move, resize, blur, focus events
Chromeless	titleBarStyle: hidden, draggable regions
Multiwindow CEF	CEF renderer multi-window
Quit test	before-quit lifecycle
Webview settings	renderer options, sandbox
Webview cleanup	view lifecycle and removal
WGPU View	WGPUView interactive
WGPU Tag	`<wgpu-view>` HTML tag
Fullsize frame repro	window frame edge cases

Test Runner UI — CSS Selectors

Main window: views://test-runner/index.html

Controls

Selector	Action
`#btn-run-all`	`runAllAutomated()` — runs all non-interactive tests
`#btn-run-interactive`	`runInteractiveTests()`
`.run-btn[data-test-id]`	`runTest({ testId })` — run one test
`#test-search`	Fuzzy filter by name/category/description
`#update-btn`	`applyUpdate()`
`#update-history-toggle`	Toggle update status history panel
`#update-history-clear`	`clearUpdateStatusHistory()`

Status elements

Selector	Shows
`#total-count`	Total test count
`#passed-count`	Passed
`#failed-count`	Failed
`#pending-count`	Pending/not-run
`#test-list`	Category groups + test cards
`#search-meta`	Total or filtered count

Interactive modal

Selector	Purpose
`#interactive-modal`	Modal container
`#modal-title`	Test name
`#modal-instructions`	Instructions text
`#btn-start`	`submitReady({ testId })`
`#btn-pass`	`submitVerification({ testId, action: 'pass' })`
`#btn-fail`	`submitVerification({ testId, action: 'fail' })`
`#btn-retest`	`submitVerification({ testId, action: 'retest' })`
`#notes-input`	Optional notes

Bun↔UI RPC Contract

Requests (bun exposes to UI)

getTests() → test list
runTest({ testId }) → run one test
runAllAutomated() → run all automated
runInteractiveTests() → start interactive sequence
submitInteractiveResult({ testId, passed, notes }) → submit result
submitReady({ testId }) → signal ready for interactive step
submitVerification({ testId, action, notes }) → pass/fail/retest
applyUpdate() → apply downloaded update
getUpdateStatusHistory() / clearUpdateStatusHistory()
getTestRunnerPreferences() / setTestRunnerPreferences({ searchQuery })

Messages (bun sends to UI)

testStarted — test execution began
testCompleted — result with pass/fail/duration
testLog — log line
allCompleted — all tests done
interactiveWaiting — waiting for user to click Start
interactiveReady — user ready, perform the action
interactiveVerify — show pass/fail/retest controls
buildConfig — app metadata
updateStatus / updateStatusEntry — updater state

Interactive Test Flow (Two-Step)

1. User clicks Open on an interactive test
2. Runner shows modal: instructions + #btn-start
3. User reads instructions, clicks Start → submitReady()
4. A secondary playground window opens (bun-side)
5. User performs the described action in that window
6. Modal shows: #btn-pass, #btn-fail, #btn-retest
7. User marks result → submitVerification()
8. Modal closes, result recorded

Consuming the Manifest (agent usage)

# Read the manifest
cat kitchen/src/generated/feature-manifest.json

# Filter: automated tests only
jq '[.[] | select(.interactive == false)]' feature-manifest.json

# Filter: tests for a specific API
jq '[.[] | select(.apiSurface | contains(["BrowserWindow"]))]' feature-manifest.json

# Map test to source file
jq '.[] | select(.id == "window-creation-with-url") | .testFile' feature-manifest.json

Playground Windows

Each interactive suite opens a dedicated playground window. Key ones:

Playground	URL	Key Controls
File dialog	`views://playgrounds/file-dialog/`	`#openDialogBtn`, `#result`, `#history`
App menu	`views://playgrounds/application-menu/`	Sets app menu via RPC
Context menu	`views://playgrounds/context-menu/`	Shows context menu via RPC
Webview tag	`views://playgrounds/webviewtag/`	masks, passthrough, navigation
Session/partition	`views://playgrounds/session/`	Compares cross-partition localStorage
Window events	`views://playgrounds/window-events-*/`	move/resize/blur/focus detection
Clipboard	`views://playgrounds/clipboard/`	read/write text and image
Tray	`views://playgrounds/tray/`	tray icon and menu

Platform Caveats in Tests

Zoom: exact zoom level only verified on macOS; defaults to 1.0 on other platforms
Minimize/unminimize: marked unreliable on some Linux window managers
Session cookies: some tests commented out due to ARM Windows VM crashes
Application menus: intentionally skipped on Linux in interactive tests
Context menus: marked macOS-only in interactive tests
CEF devtools: only available in dev builds, at http://localhost:9222

electrobun-kitchen-sink

Electrobun Kitchen Sink

Project Layout