The border flood-fill on a full-res 2256x4000 (9M px) photo was too slow in
JS and the async mask pass timed out into the catch branch (overlay stayed
null, 'Mask settled' never rendered). Downscale to <=1200px for the fill,
then nearest-neighbor upsample the slab bitset back to full size so
minAreaRect/slabBounds stay in source-pixel coordinates.
The green/black-sheet luminance threshold could not separate the slab from
the 'black' sheet, which in real photos reads at lum ~50 and overlaps the
slab's bark/shadow tones (measured on the user's samples: 0% green, 60-80%
dark, no brightness gap) - it removed 100% of the image. Replace the mask
pass with connectivity-based background removal: sample the sheet tone from
the border, flood-fill from the border marking sheet-connected pixels as
background, keep the connected remainder as the slab. Robust to tonal
overlap. Adds bgRemoval.ts (removeBackground + sampleSheetTone) with unit
tests proving a sheet-toned slab pixel survives via enclosure.
Add global crop_fill percent (default 90, 1-100) as a settings key with
FastAPI 4xx validation. Wires the value through to the inventory PNG
encode (non-retro: stored PNGs untouched). Adds a throwaway
TapePrototype.svelte behind ?tape-prototype for the #99 design sandbox.
Gates: check_requirements PASS, check_coverage PASS, settings tests 8/8.
PhotoStage used new Image() while mask worker and calculate used bare
createImageBitmap; Orientation-8 matched-set JPEGs can disagree across
those paths so the approved pink axis is not the overlay (ux,uy).
One decodeOrientedBitmap helper always requests imageOrientation
from-image (with bare fallback), wired into PhotoStage, mask.worker,
and session calculate/siblings.
Add pure station/axis math and canvas burn-in for a second calibration
inventory PNG (results.overlayPng). Width stations start 6 in from the
EXIF-oriented photo bottom along the approved length axis (6, 18, … ≤
L−6); length graphic sits on axis-right outside the slab. Clean
inventoryPng and batchApply[cal]=null stay unchanged.
After calibration Calculate, reuse approved sheet/sensitivity on non-calibration
photos via the mask worker + inventory compose. Persist per-photo ok/error on
measure.batchApply and the device draft for #64; leave Mask UI unopened and keep
scale/widths/sqft/bdft on the calibration results only.
Document single-source M1 path after pickCalibration: Mask/Axis/Measure/
Calculate use the chosen Source only; batch siblings stay draft-untouched
until #62 apply. Tag encode-fallback inventory Blob as image/png (CAPTURE.md).
Replace single Add photo with Batch: accumulate 1–N photos, Continue
when ≥1 (M1 bridge: mask on first), persist batch blobs in the device
draft, Resume Continue restores. Keeps Mask/Axis/Measure intact.
Why
Calculate showed accurate numbers but the inventory PNG kept the sheet
background. Measure already uses approvedMask.bitset; encode did not.
Scope
Add composeInventoryFrame and inventory.test.ts alpha/crop proofs.
Wire encodeInventoryPng({ bitmap, mask, source }) at calculate().
This commit still paints the slabBounds AABB without the bitset gate so
the alpha contract fails until the follow-up fix.
Verification
node --import tsx --test src/png/inventory.test.ts
alpha contract fails (sheet green leaks); 3:4 / 80% / floor pass.
Length axis keeps the approved mask, replaces Flip 180 with a ± angle
slider, and sets scale from wood span along the approved direction rather
than bbox.longSidePx. Widths still sample every 6 in perpendicular to that
axis.
Why
Phone and PC need a single capture flow that ends at inventory PNG and
board-foot numbers with no backend, inference, or Woo create.
Scope
Add frontend/ Vite+Svelte capture UI with Resume, Add photo, Mask,
Length axis, and Measure. Mask runs in a worker with Auto/Green/Black
and a 2.0s sensitivity debounce. Measure computes scale, widths, sqft,
bdft, and a 3:4 alpha inventory PNG. Persist one device draft in
IndexedDB. Serve the static build from frontend/Dockerfile via nginx
(deploy/nginx-m1.conf). No FastAPI.
Verification
sudo docker build -f frontend/Dockerfile -t slabuploader-m1 .
sudo docker run -d --name slabuploader-m1 -p 0.0.0.0:18080:80 slabuploader-m1
curl http://127.0.0.1:18080/ returns 200
Playwright: synthetic green-sheet photo through Calculate shows sqft/bdft
Ty
grokbot_ty