- CropOverlay: rendered in its own Layer when crop tool is active,
cropState auto-initialized when switching to crop tool
- SelectionOverlay: marching ants rendered for active selections,
selection tool mouse events dispatched through canvas handler
- TransformToolTransformer: rendered when transform tool is active
- Selection tool handlers wired for marquee-rect, marquee-ellipse,
lasso-free, lasso-poly, and magic-wand tools
- Active selection preview rect shown while dragging
- Marquee: BoxSelect (dotted rectangle) instead of Square
- Lasso: Lasso icon instead of Pen
- Transform: Maximize2 (corner arrows) instead of Move
- Gradient: Blend instead of ArrowUpRight
- Blur Brush: Droplet (water drop) instead of ScanLine
- Sharpen Brush: Triangle instead of ScanLine
- Smudge: Fingerprint instead of ScanLine
- Burn: Flame instead of Sun (was same as Dodge)
- Sponge: Droplets instead of Sun (was same as Dodge)
- Shape: Hexagon instead of Square (was same as Marquee)
- Reorder groups to match standard Photoshop toolbar layout
- Merge Crop + Eyedropper into one group
- Separate Eraser into its own group
Brush and eraser mouseUp handlers were bumping _historyVersion a second
time (after addObject already bumped it in mouseDown). This created a
ghost history entry whose objects array held the same reference as the
current state. On undo, zundo restored that identical reference, so
zustand's Object.is selector equality saw no change and skipped the
canvas re-render -- the stroke stayed visible even though the history
panel moved backward.
Remove the redundant _historyVersion increment from both mouseUp
handlers. The addObject call in mouseDown already records the correct
pre-stroke snapshot. mouseUp now only patches lastAction for the
history label without creating a new temporal entry.
1. Export dialog (Ctrl+Shift+S): replaced react-hotkeys-hook handler with
a capture-phase keydown listener on window so the browser's native
"Save Page As" dialog is intercepted before it can fire.
2. Text tool: created useTextTool hook that spawns an inline textarea
overlay on canvas click, commits the text as a Konva Text object on
blur/Enter, and wired it into the useActiveToolHandlers dispatcher.
3. Undo (Ctrl+Z): removed the 500ms debounce from zundo's handleSet.
The debounce caused a race where calling undo before the timer fired
would discard the future-states stack, making undo appear to do
nothing. The equality function (keyed on _historyVersion) already
prevents intermediate states from being recorded, so the debounce
was redundant.
CRITICAL fixes:
- #14: Render source image on canvas via Konva Image + use-image hook
- #3: Wire move tool handlers (onClick, onDragEnd, onTransformEnd, draggable)
to all CanvasObjectRenderer shapes
- #4: Implement image object rendering for fill/gradient output
- #2: Show fallback text in histogram panel when no imageData provided
- #1: Forward all args in zundo handleSet debounce wrapper
HIGH fixes:
- #5: Track raw screen cursor position for brush overlay instead of
using canvas-space coordinates
- #6: Export dialog uses Konva stage.toDataURL via module-level ref
instead of DOM querySelector for correct export at any zoom/pan
- #7: Add _historyVersion increment to setAdjustment, resetAdjustments,
toggleFilter, and setFilterParam for undo tracking
- #8: Include lastAction in partialize so history labels display correctly
MEDIUM fixes:
- #10: Move useEditorShortcuts from EditorCanvas to EditorPage with
save/export callbacks
- #11: Remove _historyVersion increment from updateObject to prevent
brush strokes from flooding undo history
- #12: Apply Konva filters (Brighten, Contrast, HSL, Blur, Grayscale,
Sepia, Invert, Pixelate, Emboss, Posterize, Noise, Solarize,
Threshold, Kaleidoscope) to source image node based on store state
- Show desktop-recommended message on mobile viewports (<768px)
- Add LoadingOverlay component for canvas loading/progress states
- Add ARIA tablist/tab/aria-selected to right panel tabs
- Add "Open in Editor" link in review panel after tool processing
Implements Agent 2 deliverables for the image editor: move tool with Konva
Transformer and smart guide snapping, rectangular/elliptical marquee and
lasso selection with marching ants animation, magic wand flood fill, crop
tool with darkened overlay and rule-of-thirds grid, free transform with
numeric inputs, context menu with z-ordering and clipboard operations,
rulers with drag-to-create guides, smart alignment guides, and canvas/image
resize dialogs. Also creates the shared editor types and Zustand store with
full layer management, z-ordering, clipboard, guide, and document operations.
Implement Agent 7 features for the image editor:
- History panel with undo/redo list, action icons, and state jumping
- 51+ keyboard shortcuts via react-hotkeys-hook (tools, modifiers, clipboard)
- Export dialog with PNG/JPEG/WebP format, quality, resize, clipboard copy
- Navigator minimap with viewport rectangle, drag-to-pan, zoom slider
- Autosave/recovery system with localStorage and recovery banner
- Project save/load (.snapotter JSON format)
- Unsaved changes warning via beforeunload
- Paste from system clipboard support
Implements the layer management panel for the image editor (Features 16, 40):
- Blend mode dropdown (16 modes) and opacity slider for active layer
- Layer list with visibility/lock toggles, thumbnails, and inline rename
- Pointer-based drag reorder (not HTML5 drag API)
- Right-click context menu: Duplicate, Merge Down, Flatten All, Delete
- Layer effects section: Drop Shadow, Inner Shadow, Outer Glow, Stroke
- Each effect with enable/disable toggle and full parameter controls
- Wired into EditorRightPanel as the "Layers" tab content
Create the four-zone editor layout shell (toolbar, options bar, status
bar, right panel), welcome screen with drag-and-drop and new document
dialog, and the main editor page with clipboard paste and URL loading.
Add complete type system for the editor (ToolType, CanvasObject, EditorLayer,
EditorState, etc.) and the central Zustand store with zundo undo/redo middleware,
50-entry history limit, 500ms throttle, and O(1) _historyVersion equality checks.
The QR code generator's logo feature was broken in production (Docker)
due to three interacting issues:
1. The CSP connect-src directive did not include data:, so the
qr-code-styling library's internal XHR to convert logo data URLs to
blobs was silently blocked. The library has no onerror handler, so the
render promise hung forever after the container was already cleared.
2. crossOrigin: "anonymous" was unnecessarily set on imageOptions for
data URLs, which can cause canvas taint issues.
3. The logo options used a conditional spread that omitted the image key
when no logo was set. The library's update() deep-merges options, so
removing the logo preserved the stale data URL and the QR stayed
broken even after logo removal.
Closes#121
The desktop sidebar never rendered the custom logo because only mobile
views used the customLogo state. Added a logo section at the top of
the desktop sidebar that displays the custom logo (or the default
OtterLogo when none is set).
The upload handler used raw fetch() without checking response.ok, so
HTTP 4xx errors (e.g. file too large) were silently ignored and the UI
falsely reported success. Now checks response status and surfaces the
server error message.
Closes#125
The crop area flex container lacked min-h-0, preventing it from shrinking
below its content's intrinsic height. For tall images, react-image-crop's
CSS (max-height: inherit at specificity 0-1-1) overrides the Tailwind
max-h constraint, and without min-h-0 the flex item refuses to shrink,
causing viewport overflow.
Closes#122
The production CSP had connect-src/script-src/font-src set to 'self' only,
silently blocking all analytics and error reporting in production while
working fine in dev (where CSP is not applied).
CSP fixes:
- Add PostHog ingest + assets origins to connect-src and script-src
- Add Sentry ingest origin to connect-src
- Add Scalar fonts origin to font-src for API docs pages
- Extract CSP construction into testable buildCsp() function
Silent failure hardening:
- Settings/features stores now set loadError flag and allow retry on
subsequent fetch() calls instead of permanently caching failed state
- Analytics init no longer sets initialized=true before the try block,
allowing retry on failure
- Settings dialog Tools section disables save button when settings
failed to load, preventing accidental config wipe
- Branding logo storage moved from process.cwd() to FILES_STORAGE_PATH
so logos persist across Docker container recreation
Test coverage:
- 16 CSP directive tests covering all external service domains
- Store retry-on-error behavior tests for settings and features stores
- Analytics init retry-after-failure test
Sentry fixes:
- Only send 5xx errors to Sentry (was sending 4xx rate-limit, media type errors)
- Encode non-ASCII chars in X-Output-Filename header (encodeURIComponent)
- Handle FK constraint failures gracefully in file upload, pipeline save, API keys
- Harden getDirSize against ENOENT race on readdirSync
Test fixes:
- Wrap navbar test renders in act() to flush async useEffect state updates
- Add useEffect cleanup to navbar to prevent state updates on unmounted component
- Fixes timeout when running in full test suite
- README: Discord badge in header + community section link
- Docs: GitHub and Discord social links in VitePress config
- Landing: Discord in navbar, footer, open-source section, pricing copy
1. passport-photo 404 vs 501: add base route at /api/v1/tools/passport-photo
that returns 501 FEATURE_NOT_INSTALLED when the AI bundle is missing,
matching other AI tools. The /generate sub-route is Sharp-only (no
sidecar) so it correctly skips the isToolInstalled guard.
2. AuthGuard analytics consent race: don't evaluate shouldShowConsent()
until analyticsConfig has been fetched (guard on analyticsConfig !== null).
Prevents redirect to /analytics-consent before config is loaded.
3. Fragile sidebar Settings selector: add openSettings(page) helper to
E2E helpers that checks sidebar visibility with fallback to button role.
Replace all 134 occurrences of page.locator("aside").getByText("Settings")
across 18 test files.
Convert all 9 AI tool routes (colorize, restore-photo, remove-background,
enhance-faces, blur-faces, red-eye-removal, erase-object, noise-removal,
upscale) to async 202 processing so none are vulnerable to proxy
connection timeouts.
Also fixes:
- Replace basename() with sanitizeFilename() in all AI tool routes
(prevents double-extension attacks and adds length truncation)
- Add UUID format validation for clientJobId field
- Fix missing filename sanitization in noise-removal (was using raw
user-supplied filename with zero sanitization)
- Remove em dash from error message in use-tool-processor
SnapOtter