Instead of transforming coordinates after Gemini returns crop_bounds, strip EXIF
orientation from image before sending to Gemini. This ensures:
- Gemini analyzes the same raw image as our backend
- crop_bounds are in raw image coordinate space
- No coordinate transformation needed
- Works for all images (with or without EXIF)
Added strip_exif_orientation() utility that removes orientation tag and
re-encodes image. Used in extract_label endpoint before sending to Gemini.
Gemini analyzes images with EXIF orientation applied (e.g., 90° CW rotation),
returning crop_bounds in that coordinate space. We need to transform them back
to raw image coordinates before cropping.
For EXIF orientation 6 (Rotate 90 CW):
- Raw: 4032×3024 (landscape)
- Displayed (after EXIF): 3024×4032 (portrait)
- Transform portrait coords back to landscape before cropping
This fixes the issue where crop was applied to wrong image region.
Add logging to compare file sizes:
- [EXTRACT] sent to Gemini
- [CREATE_ITEM] received when creating item
This will reveal if image is being processed/changed between extraction and local processing.
Gemini analyzes raw image and returns crop_bounds for that coordinate space.
Previous code applied EXIF rotation first, changing image dimensions, then
used crop_bounds on the rotated image (coordinate mismatch).
Now: crop on raw image (matches AI) → then apply EXIF + manual rotation.
This ensures cropped region contains the actual item, not background.
- Move manual rotation before cropping and text detection
- Detect text orientation on full rotated image (not just cropped region)
- This allows text angle detection to see full context and properly orient labels
- Crop happens after orientation correction for cleaner results
- Delete photo_path and photo_thumbnail_path files on item deletion
- Handle file not found gracefully with logging
- Preserves audit logs while removing actual image files
- Add rotation_degrees parameter to ImageProcessor.process_photo()
- Pass rotation through _auto_save_photo_from_extraction() to processor
- Allow no-crop fallback when crop_bounds is None
- Add buildPhotoUrl() helper to resolve backend URLs correctly
- Update frontend components to use backend URL for image sources
- Replace Use/Skip Photo buttons with checkbox in AI extraction UI
- Add images/ to .gitignore to prevent accidental commits
Addresses: rotation never applied, image 404s (relative to Next.js not backend), preview blank in edit form
- Extend ItemCreate schema with optional extracted_image_bytes (base64) and image_processing (dict)
- Update create_item endpoint to call _auto_save_photo_from_extraction after item creation
- Decode base64 image bytes and pass crop_bounds, rotation_degrees to helper
- Don't block item creation if photo save fails (log warning instead)
- Item returned with photo_path, photo_thumbnail_path populated if save succeeded
- Full backward compatibility: old clients without image fields work unchanged
- Add 5 integration tests covering all scenarios:
- Create item WITH image_processing → photo auto-saved
- Create item WITHOUT image_processing → no photo (backward compatible)
- Create item WITH invalid image_processing → item created, photo skipped
- Create item WITH crop_bounds=None → item created, photo skipped
- Create item WITH bytes but NO processing metadata → item created, photo skipped
- All 158 backend tests passing, zero regressions
- Added 11 comprehensive tests for image_processing parsing
- Tests validate crop_bounds structure: {x, y, width, height} all ints >= 0
- Tests validate rotation_degrees: int/float, -360 to +360
- Tests validate confidence: float, 0.0 to 1.0
- Tests graceful handling when image_processing field is missing
- Tests multiple items with image_processing data
- Tests partial data handling (optional fields)
- Tests with both Gemini and Claude providers
- Updated extract_label_info() to preserve and validate image_processing field
- All tests passing, no regressions
Temporarily using allow_origins=['*'] to debug whether CORS is blocking
LDAP login requests from VPN client. This is insecure for production.
TODO: Fix subnet pattern matching in ALLOWED_ORIGINS configuration.
- Simplify backend CORS middleware to use standard FastAPI implementation
- Keep subnet validation function for future use in route-level checks
- Add Tailscale subnet pattern to Next.js allowedDevOrigins config
- Both individual IPs and subnet configurations now work correctly
- Add ipaddress module for subnet parsing (10.0.0.0/24 format)
- Implement subnet validation in CORS middleware
- Separate individual IPs from subnet definitions in EXTRA_ALLOWED_ORIGINS
- Custom SubnetAwareCORSMiddleware for dynamic origin validation
- Support both exact IP matches and subnet ranges
- Backward compatible with existing ALLOWED_ORIGINS list
- Create ImageProcessor service with EXIF orientation detection
- Implement smart cropping via OpenCV contour detection (10% padding)
- Add text orientation detection using Hough line transform
- Resize and compress images to 1200px with 85% JPEG quality
- Generate 200px square thumbnails with center crop
- Fallback to Pillow if OpenCV fails
- Comprehensive test suite: 28 tests all passing
- File size validation (reject >10MB)
- Graceful error handling for corrupted/invalid images
- Update requirements.txt with opencv-python, piexif, python-magic
The N+1 optimization in save_image() pre-lowercases the existing_files list
before passing to get_unique_filename(). However, this broke the API contract:
the function should handle any-case input to remain robust.
Changed: get_unique_filename() now defensively lowercases the input list,
ensuring collision detection works regardless of input case.
Benefits:
- Fixes implicit API contract change (function expected any-case input)
- Maintains N+1 optimization (pre-lowercasing still works)
- Supports both optimization and edge cases (direct function calls)
- All 22 tests pass
- Add SIZE CONVERSION RULES section (critical) with explicit thresholds
- Convert ≥1600GB to TB format (e.g., 1600GB → 1.6TB)
- Convert memory ≥1024MB to GB format
- Update Item field examples with human-readable sizes
- Update Size field definition to emphasize HUMAN-READABLE format
- Update OCR key definition with size conversion examples
- Apply rules consistently across all size-related fields
This ensures AI-extracted items display sizes in user-friendly format
and improves OCR matching accuracy with normalized size representations.
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
New Component: ItemComparisonModal.tsx
- Shows existing vs new item side-by-side
- Highlights fields that are different (in yellow)
- Options to Update item or Skip (local-only save)
- Shows existing item ID and comparison details
Backend Changes:
- Updated error message to say 'Part Number' not 'barcode'
- 409 response includes existing item data for comparison
- Clear, user-friendly conflict messaging
Frontend Changes:
- New state for comparison modal (newItem, existingItem, existingId)
- handleOnboardingComplete() shows modal on 409 conflict
- handleComparisonUpdate() calls updateItem() API
- handleComparisonSkip() saves locally without syncing
- Better error handling distinguishes 409 from other failures
Workflow:
1. User imports item with Part Number that already exists
2. System shows comparison modal
3. User can:
- Update (merges new data into existing)
- Skip (saves locally, doesn't sync to cloud)
