Replace the external mask file upload workflow with an interactive
canvas-based brush tool. Users now paint directly on the image to
mark areas for erasure. Adds EraserCanvas component with adjustable
brush size, undo/clear, and mask export. Switch Python inpainting
from broken lama-cleaner to OpenCV cv2.inpaint (Telea algorithm).
Add before/after comparison slider after processing.
Remove approximate processing times from remove-bg quality options.
Update tool count from 37 to 33 across README, settings dialog, login
page, and i18n strings to reflect the accurate count.
Fix face detection failure caused by MediaPipe 0.10.33 removing the
mp.solutions API. Replace with OpenCV Haar cascade which works reliably
in headless Docker. Add autoOrient() call before detection to handle
EXIF-rotated phone photos. Remove technical jargon from UI.
Fix rotated output by calling autoOrient() before passing images to
the Python upscaler, correcting EXIF orientation metadata. Replace
2x/4x buttons with 2x-8x slider plus quick-select buttons (2x, 3x,
4x, 6x, 8x). Remove technical jargon about Real-ESRGAN/Lanczos from
the UI and progress bar.
Add CSS filter-based live preview for brightness, contrast, saturation,
color effects (grayscale/sepia/invert), and color channels (via SVG
feColorMatrix). The image in the right panel updates instantly as
sliders are adjusted. Adds cssFilter prop to ImageViewer, onPreviewFilter
callback to ColorSettings, and wires them through the tool page.
Add a miniature page preview in the settings panel that shows how the
image will look in the PDF. The preview updates in real-time as the
user changes page size, orientation, or margin. Uses the same scaling
logic as the backend to accurately represent image placement. Preview
tracks the currently selected image when multiple files are uploaded.
Replace the static angle display between Left/Right buttons with an
editable input field that accepts direct angle entry. Add - and +
buttons for 1-degree fine-tuned control. All controls update the
rotation state which triggers real-time CSS preview on the image.
- Make ensureDefaultAdmin idempotent with onConflictDoNothing (fixes
UNIQUE constraint error when parallel test files share a DB)
- Move duplicate-username check before user-limit check so 409 takes
priority over 403
- Bump MAX_USERS from 5 to 50 (tests create ~15 users across files)
Remove docs/superpowers/, .claude/ config, and PRD.md from version
control (kept locally via .gitignore). Update README, CHANGELOG,
VitePress docs, and .env.example to reflect recent features: Files
page, teams, admin settings, persistent storage, and various API
improvements.
checkAuth was defined as a plain function and used as a useEffect
dependency, causing it to fire on every render. Moved it inside the
effect with an empty dependency array so it runs once on mount.
- Add Teams section with CRUD (create, rename, delete) and member count display
- Add Tools section with per-tool enable/disable toggles grouped by category
- Add logo upload/delete, experimental tools toggle, file management settings to System Settings
- Replace free-text team input with dropdown populated from teams API in People section
- Filter disabled/experimental tools in tool panel, pipeline builder, and fullscreen grid
- Display custom logo in mobile header and sidebar when configured
Add feature flag support to skip disabled/experimental tools at startup
by reading disabledTools and enableExperimentalTools from the settings
table. Refactor cleanup.ts to read tempFileMaxAgeHours from DB settings
(with env var fallback) and respect the startupCleanup setting.
Replace static desktop-only layout with responsive version: mobile gets a
tab bar (Recent / Upload) and a bottom-sheet drawer for file details, while
desktop retains the existing three-column panel layout. FileDetails gains
updated mobile wrapper classes (removes border/width, uses flex gap).
Add serverFileId to FileEntry, pass fileId in tool processor formData when
a file originates from the Files page, and set serverFileId on the loaded
entry so subsequent tool runs create new file versions instead of duplicates.
Creates the full Files page UI (Tasks 8 & 9): FilesNav, FileListItem,
FileList, FileDetails, FileUploadArea components, FilesPage layout, and
wires up the /files route plus sidebar/mobile-nav entries.
Adds UserFile/UserFileDetail types plus apiListFiles, apiGetFileDetails,
apiUploadUserFiles, apiDeleteUserFiles, getFileThumbnailUrl, and
getFileDownloadUrl to support the Files page.
Adds optional fileId multipart field; when present, inserts a new versioned
userFiles record after successful processing and returns savedFileId in the
response. Non-fatal — tool processing succeeds even if the save fails.
Implements all 7 routes for the persistent file library:
list with pagination/search, upload (multi-file), details with recursive
version chain CTE, download stream, on-the-fly Sharp thumbnail, bulk delete
of entire version chains, and save-result for tool output versioning.
Registered in index.ts after fileRoutes.
Adds ensureStorageDir, saveFile, deleteStoredFile, and getStoredFilePath
utilities that manage the lifecycle of files on the local filesystem
under FILES_STORAGE_PATH.
Replace placeholder "default settings" message with actual form controls
when expanding pipeline steps. Each tool now shows its configurable fields
(dimensions, quality, colors, modes, etc.) directly in the pipeline builder.
Covers all 30+ tools with data-driven field definitions: number inputs,
select groups, checkboxes, color pickers, and conditional fields.
Extract EXIF auto-orientation logic into a shared auto-orient module
used by both single-tool and batch routes. This ensures camera photos
display correctly after processing regardless of entry point.
Also expands e2e and integration tests significantly.
- Remove @fastify/swagger and @fastify/swagger-ui (API docs live on GitHub Pages)
- Run typecheck, build, and docker CI jobs in parallel instead of sequentially
react-image-crop's onChange passes (pixelCrop, percentCrop) — we were
using the first arg (display pixels) but treating them as percentages,
causing values to exceed image dimensions.
Samsung Galaxy phones (and others) can have GPS EXIF tags with null
coordinate values when location is disabled. The server returned
_latitude: null, and the client guard used !== undefined which passed
for null, causing null.toFixed() to crash React with no ErrorBoundary.
- Server: validate GPS array values are actual numbers before computing
- Client: use != null guard (catches both null and undefined)
- App: add ErrorBoundary to prevent white screens from any future crash