Skip to content

fix(catalog): refine industrial-plant material finishes - #1507

Draft
w1ne wants to merge 64 commits into
developfrom
feat/catalog-2000
Draft

w1ne wants to merge 64 commits into
developfrom
feat/catalog-2000

Conversation

@w1ne

@w1ne w1ne commented Nov 17, 2026 •

Copy link
Copy Markdown
Owner

Summary

The 17 Nov five-preview audit found that the palletizer robot cell's full floor slab inherited safetyPaint. It now has a separate concrete control and preset colors; the fence posts and robot pedestal retain safety paint. Geometry and triangle budget are unchanged.

This branch also contains the preceding industrial-plant material corrections for the roller conveyor, steam manifold, scrubber, inline flow meter, reactor vessel, and process blower.

Verification

  • Focused exported-GLB material regression: 1/1 passed.
  • ESLint on the industrial-plant kit and new test: clean.
  • catalog/index.json regenerated; the diff contains only the palletizer row and its source SHA matches the source. Preview path exists.
  • Industrial-plant quality gate: 60/60 (100%). Global quality scan: 2,398/2,402 (99.83%); four unchanged standard-components files fail the material-count rule and match origin/develop.
  • Five reviewed previews, before/after source and PNGs, MP4 demo, settings, and verification evidence: handoff bundle.

The branch is not deployed. The demo shows local source editing, software-rendered previews, and a GLB regression only; it is not simulation or hardware acceptance.

Latest round — scissor lift table arm finishes (2026-11-17)

The five-preview audit selected industrial-plant/scissor-lift-table: its matched scissor-arm-b parts used safetyPaint, while the corresponding scissor-arm-a parts used steel. Both B arms now use the existing steel control; geometry, parameters, presets, and the 196-triangle mesh are unchanged.

  • The GLB material regression fails on the original source and passes after the correction (1/1).
  • Regenerated the scissor-lift preview and catalog/index.json; the generated diff changes only this asset row and its source hash.
  • ESLint is clean. Industrial-plant quality: 60/60 (100%). Global structural quality: 2,398/2,402 (99.83%); the four unchanged exceptions are in standard-components.
  • Full catalog integration passes: 141 files / 211 tests, including the index freshness check.
  • Five-preview ledger, before/after preview, source snapshots, verification, and inspected 14-second MP4.

The current public detail page and old preview returned HTTP 200 before this update, with the live preview matching the baseline image. This branch is not deployed; the new preview has not been verified on the deployed page. The MP4 shows a local source edit, software-rendered preview, and GLB regression only; it is not simulation or hardware acceptance.

Latest round — pipe-rack crossbeam finish (2026-11-18)

The five-preview industrial-plant review selected pipe-rack-bay: its two crossbeams and two mounting baseplates were safety-painted while the matching columns and braces were structural steel. Those four frame parts now use the existing steel finish; the yellow center utility pipe and flanges remain yellow. Geometry, parameters, and presets are unchanged; the mesh is 360 triangles under its 1,400-triangle budget.

  • Added exported-GLB material regression: fails on the original crossbeam assignment and passes after the correction.
  • Regenerated preview and catalog/index.json; the index source SHA matches the edited source.
  • ESLint is clean. Full catalog integration passed with 152 files / 222 tests, including index freshness. The default parallel run had one 300-second shape-handle timeout under load; that check passed in isolation and in the full two-worker rerun.
  • Five-preview ledger, before/after source and previews, verification, and inspected 15-second MP4 demo.

The page has not been deployed, so no post-deployment page verification is claimed. The demo shows local source, software-rendered previews, and an exported-GLB regression only; it is not simulation or hardware acceptance.

Latest round — assay sample divider bucket material (2026-11-19)

The five-preview mining audit selected assay-sample-divider: its eight open rotary collection buckets used the pale plastic finish while the adjacent turntable and feed surfaces used stainless steel. The buckets now use the existing stainlessSteel material control; sample ore remains separate. Geometry, envelope, triangle budget, presets and divider/bucket choices are unchanged.

  • The exported-GLB regression failed on the original material assignment and passes after the fix (1/1). The full catalog integration suite passed 161 files / 231 tests, including source/export/index freshness.
  • Mining quality: 50/50 (100%). Global quality: 2,398/2,402 (99.83%); four unchanged below-gate assets remain in standard-components.
  • Re-rendered the target preview and regenerated catalog/index.json; the target source hash and preview path are current, and its triangle count and bounds are unchanged. Three post-render previews were opened and read clearly in color.
  • Five-preview ledger, before/after source and images, verification evidence, and inspected local MP4.

The existing public routes matched the pre-change baseline before editing. The branch has not been deployed, so no post-deployment check is claimed. The MP4 documents local software previews and the GLB regression only; it does not claim simulation, equipment performance, or hardware acceptance.

Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 17, 2026

Copy link
Copy Markdown
Owner Author

Semantic preview review — 17 November 2026

Reviewed five industrial-plant previews and recorded the full ledger, before/after images, source snapshots, local checks and the 14-second MP4 in the handoff bundle.

One material mismatch was fixed in industrial-plant/valve-manifold-station: its three outlet flange rings inherited yellow safetyPaint while the adjacent branch outlets and valve bodies use steel. The flanges now use the existing steel control. The skid and handwheels remain safety-painted; geometry stays at 740 triangles under the 1,400 budget. The index was regenerated and changes only this asset row.

Verification: the exported-GLB regression failed on the baseline flange color and passes after the fix; ESLint is clean; industrial-plant quality gate is 60/60 (100%); npx vitest run tests/integration/catalog/ passed 134 files / 204 tests, including the full index-freshness rebuild. The live detail page still serves the baseline because this branch is not deployed. The MP4 shows local software previews and a software regression, not deployment, simulation, or hardware acceptance.

The required render inspect attempt used the existing Chrome CDP session, timed out during its handshake, and did not launch a second browser. The catalog preview and demo video were inspected directly. PR checks for this push are starting; no review has been submitted yet.

@w1ne

w1ne commented Nov 17, 2026

Copy link
Copy Markdown
Owner Author

Final CI update for 29f2fecda8eafe84dbdec41c2f0b63be08b6cb3b:

  • build-and-checks, changed-tests, lint, test (draft-light), and changes passed.
  • catalog-index, external-tools, geometry-proofs, reference-parts, solidcraft-ledger, and test-shard were skipped by path filters.
  • The full local catalog integration suite passed: 134 files / 204 tests.

The audit ledger, before/after previews, regression evidence, and inspected demo are in the handoff bundle. Public deployment was not changed; the live catalog page remains on its previous preview. No review has been submitted, and GitHub currently reports this PR as blocked from merge.

Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 17, 2026

Copy link
Copy Markdown
Owner Author

Semantic quality review round (commit f5167a953)

  • Checked PR catalog: rebuild 30 box-stack assets as real machines (music, amusement, sports, solar, construction) #1258 first: merged with all 17 reported CI checks successful.
  • Reviewed five rendered industrial-plant assets; exact-ID open issue/PR searches were clear. The ledger and exclusion evidence are in the review bundle.
  • Corrected the motor-control-center material mismatch: the center cabinet bay now uses the same steel finish as the two neighboring sections. The plinth and operator panels retain safety paint. No geometry or asset count changed; the model remains at 396 triangles.
  • Re-rendered the preview and regenerated catalog/index.json; only the motor-control-center row changed and its source hash matches.
  • Focused GLB regression passed 1/1 after failing on the baseline. ESLint passed. npx vitest run tests/integration/catalog/ passed 136 files / 206 tests, including the catalog index freshness check.
  • The public page still serves the baseline preview because this branch has not been deployed.

Review bundle: README · 14-second MP4

Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 17, 2026

Copy link
Copy Markdown
Owner Author

Semantic preview audit follow-up — 17 November 2026

Reviewed five rendered industrial/manufacturing assets: forklift truck, centrifugal pump, distillation column, industrial boiler, and bridge crane. The handoff bundle has the five-preview ledger, source snapshots, before/after render, regression evidence, and an inspected 14-second MP4:

Bridge crane hoist-motor material review

One material mismatch was fixed: bridge-crane/hoist-motor used the girder finish. It now uses the existing hoist control. Geometry and envelope are unchanged (4,104 triangles / 6,000 budget); bridge girders retain their finish.

Local verification:

  • Regression failed on the baseline source and passed after the change (1/1).
  • Industrial-plant ESLint: clean.
  • tests/integration/catalog/: 140 files / 210 tests passed, including catalog source/export/index freshness.
  • Regenerated catalog/index.json; only industrial-plant/bridge-crane changed and the source SHA matches.
  • Mechanical quality gate: 2,398/2,402 pass (99.83%); four unchanged exceptions are in standard-components.

The live detail page and baseline preview returned HTTP 200 before this push; the changed preview is not deployed. The MP4 documents local software rendering and the exported-GLB regression only, not simulation or hardware acceptance.

@w1ne

w1ne commented Nov 17, 2026

Copy link
Copy Markdown
Owner Author

CI follow-up for ca24dba41: run #44067490196 passed build-and-checks, changed-tests, lint, and the aggregate test check. catalog-index and other specialized jobs were skipped by the changed-file selection; the local catalog suite passed 140 files / 210 tests, including index freshness.

Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 19, 2026

Copy link
Copy Markdown
Owner Author

Semantic preview audit — 19 November 2026 (shipyard fixtures and support stands)

Commit bd99908d5 reviews five shipyard previews and corrects one material mismatch in panel-line-strongback: both longitudinal rails now use the yardPaint control. Before, longitudinal-rail-1 used structuralSteel, so two identical rails rendered in two finishes. Geometry is unchanged (476 triangles, 3 × 0.54 × 1.52 m).

Evidence (local):

  • Focused exported-GLB regression: fails on the baseline at longitudinal-rail-1, passes after the fix (1/1).
  • ESLint clean. Shipyard quality gate 70/70. Global structural gate 2,398/2,402 (99.83%); the four exceptions are unchanged standard-components assets.
  • Full catalog integration suite: 165 files / 235 tests pass with --maxWorkers=2, including index freshness.
  • Index row check: source SHA matches, only the target row changed.

Reviewed, not changed: stern-section-support-frame, bow-section-support-stand, ship-plate-staging-pallet, low-profile-material-skid. Noted limits: three descriptions name parts the models do not contain (strongback end stops, bow-stand saddle pad, pallet corner restraint blocks); the plate stack renders as one block; the skid lifting eyes are solid posts.

Ledger, before/after previews and MP4: artifacts/catalog-demos/shipyard-panel-line-strongback-rail-material-review-20261119/. The branch is not deployed; no simulation, load or hardware claim is made.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 19, 2026

Copy link
Copy Markdown
Owner Author

Semantic preview audit — 19 November 2026 (data-center overhead services and telecom mast frames)

Commit 5ace9580d reviews five infrastructure previews and corrects one material mismatch in overhead-cable-ladder: both hanger caps now use the polymer control. Before, hanger-cap-right used steel, so two identical caps rendered amber and blue-grey. Geometry, the 292-triangle mesh and the bounds are unchanged.

Verified locally: the new exported-GLB regression fails on the baseline and passes after the fix (1/1); ESLint is clean; the kit quality gate passes 50/50; the catalog integration suite passes 166 files / 236 tests; only the target index row changed.

Observed and not corrected in this round: liquid-cooling-manifold, overhead-coolant-header, lattice-mast-section and mast-outrigger-frame each show one identical part pair with mixed finishes. The ledger is in artifacts/catalog-demos/data-center-overhead-cable-ladder-hanger-cap-material-review-20261119/preview-review.md.

Not deployed. No simulation, load or hardware claim.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 19, 2026

Copy link
Copy Markdown
Owner Author

Semantic preview audit — 19 November 2026 (emergency-stop stations and instrument posts)

Commit 49fcb5ad9 reviews five control-station and instrument previews and corrects one defect in data-center-infrastructure/emergency-power-off-station: the stop actuator is now a red mushroom head that faces the operator on a yellow collar. Before, push-button used the yellow body polymer control and stood upright on a red steel cup that passed through the hood.

Change:

  • New actuatorPolymer colour control (#d42a20, red in both presets) for push-button; actuator-base now uses polymer.
  • Both cylinders now run along -Y, out of the wall. This is a geometry change: depth 0.2895 m -> 0.2675 m. Width, height and the 136-triangle mesh are unchanged.

Verification (local):

  • Focused exported-GLB regression: fails on the pre-fix source, passes after the fix (1/1).
  • ESLint clean. Kit quality gate 50/50 (100%). Global structural gate 2,398/2,402 (99.83%); the four exceptions are unchanged standard-components assets.
  • Full catalog integration suite: 167 files / 237 tests pass with --maxWorkers=2, including index freshness.
  • Target preview re-rendered and opened; catalog/index.json regenerated; only the target row changed.

Noted, not corrected: ev-emergency-stop-station and sorter-emergency-stop-pedestal have the same upright, amber actuator. The power-off station guard hood is a flat plate and its label panel cannot be seen.

Evidence: bundle README, ledger, before/after and inspected local MP4. The branch is not deployed. The colour convention is a visual rule for the model; no safety or compliance claim is made.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 19, 2026

Copy link
Copy Markdown
Owner Author

Semantic preview audit — 19 November 2026 (recycling-facility site equipment)

Commit cbfe86f5f reviews five recycling-facility previews (vehicle-barrier-gate, tyre-dropoff-rack, truck-wheel-wash, roller-conveyor-section, walking-floor-trailer). It found no material mismatch in the five. It corrects one physical-feature defect in vehicle-barrier-gate: both photo-eye posts stood inside the 3.64 m drive lane, 0.87 m from the cabinet plinth, with the beam along the direction of travel.

Change:

  • eye-post-0 now stands 160 mm clear of the cabinet plinth and eye-post-1 stands in line with the rest post. Both are 0.3 m ahead of the boom on the road side. The lenses face each other across the lane.
  • Part names, the four colour controls and the three presets are unchanged. The mesh keeps 1,050 triangles (budget 2,400). Depth changes from 1.31 m to 0.595 m; width and height are unchanged.

Verified locally:

  • The new exported-GLB regression fails on the baseline (depth 1.31 m, expected 0.595 m) and passes after the correction (1/1).
  • ESLint is clean. Recycling-facility quality gate: 153/153 (100%). Global gate: 2,398/2,402 (99.83%); the four exceptions are unchanged standard-components assets.
  • Full catalog integration: 168 files / 238 tests pass with --maxWorkers=2, including index freshness.
  • The preview was re-rendered and catalog/index.json regenerated; only the target row changed and its source SHA matches.
  • Five-preview ledger, before/after source and previews, verification and inspected 20-second MP4.

Noted, not corrected: the far photo-eye post stays when the optional rest post is off; truck-wheel-wash pump-motor uses the tank colour control; the tyre-dropoff-rack pallet is a solid slab.

The branch is not deployed, so no post-deployment page check is claimed. The MP4 shows local source, software-rendered previews and a GLB regression only; it is not simulation, a safety assessment or hardware acceptance. The photo-eye position is a modelling choice, not a vendor installation dimension.

…post

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 19, 2026

Copy link
Copy Markdown
Owner Author

Round: RFID access reader status lamp (ae1567d)

Result: one material mismatch corrected in ev-charging-infrastructure/rfid-access-reader-pedestal.

Defect: status-lamp used the housing powderCoat control, stood upright and floated 34 mm in front of the post. The preview showed a dark teal stub, not a lamp.

Correction: new indicatorPolymer control (default #48b866, set in both presets). The lamp axis now runs out of the post face; the lamp back face lies on the post, 34 mm below the reader housing. Triangles (104) and envelope (0.36 × 0.98 × 0.36 m) are unchanged.

Five previews reviewed: rfid-access-reader-pedestal, parking-wheel-stop, charging-payment-terminal, charger-cable-holster-bollard, ev-parking-bay-sign. Ledger: artifacts/catalog-demos/ev-charging-rfid-access-reader-status-lamp-material-review-20261119/preview-review.md.

Verification (local):

  • Focused GLB regression: fails on the baseline, passes after the correction (1/1).
  • ESLint: clean.
  • Quality gate: kit 50/50; global 2,398/2,402 (unchanged; four standard-components exceptions).
  • Catalog integration suite: 169 files / 239 tests pass with --maxWorkers=2.
  • Index: only the target row changed; source SHA matches.

Noted, not corrected: charging-payment-terminal contactless reader floats 17 mm off the column and its bezel uses the glass control; ev-parking-bay-sign cap floats 60 mm above the post; parking-wheel-stop description names bevels and recesses that the model lacks. The kit README row for emergency-power-off-station was stale from 49fcb5a and is updated here.

Not deployed. No simulation, compliance or hardware claim.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 19, 2026

Copy link
Copy Markdown
Owner Author

Round: cleanroom door push plate (lab-infrastructure)

Result: one material mismatch corrected in cleanroom-single-door-module. door-push-plate used the leaf enamel control, was a 40 mm bar and floated 3 mm off the leaf. It now has a stainless control and is a 100 mm × 300 mm × 4 mm sheet on the leaf face.

Audited (5): cleanroom-single-door-module, cold-room-door-module, laboratory-wall-eyewash-station, laboratory-water-service-console, cleanroom-pass-through-chamber.

Verification (local):

  • Focused GLB regression: fails on the baseline, passes after the fix (1/1).
  • ESLint clean. Kit gate 60/60. Global gate 2,398/2,402 (99.83%; four unchanged standard-components exceptions).
  • Catalog integration suite: 170 files / 240 tests pass (--maxWorkers=2).
  • Index: only the target row changed; source SHA and preview path match.

Noted, not corrected: cold-room-door-module has the same push-plate defect and its description names a pull and a kick plate the model lacks. cleanroom-pass-through-chamber doors rise 286 mm above the chamber. Water-console spouts run parallel to the splash panel. The new plate has low contrast against the default enamel.

Bundle: artifacts/catalog-demos/lab-cleanroom-single-door-push-plate-material-review-20261119/. The branch is not deployed.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 19, 2026

Copy link
Copy Markdown
Owner Author

Round: wind tower section shell (4f4b17a)

Result: one material mismatch corrected in wind-energy/tower-section.

Defect: shaft, lower-collar and upper-band used a stainlessSteel control; a tower shell is painted carbon steel. The "Tapered tower section" shaft was a straight cylinder.

Correction: the shell has a towerPaint control (#e2e5e3) and the flanges have a flangeSteel control (#7f8b8e). The stainlessSteel control is removed, so the control set changes from two colours to three. The shell tapers from a 150 mm base radius to a 110 mm top radius. Triangles (144) and bounds (0.36 × 0.36 × 1.75 m) are unchanged.

Evidence (local):

  • Focused GLB regression: fails on the baseline, passes after the correction (1/1).
  • ESLint clean. Quality gate: wind-energy 50/50; catalog 2,398/2,402 (99.83%, four unchanged standard-components exceptions).
  • Catalog integration suite: 171 files / 241 tests pass (--maxWorkers=2).
  • Index: only the tower-section row changed; source SHA matches.
  • Bundle: artifacts/catalog-demos/wind-energy-tower-section-shell-material-review-20261119/ (ledger, before/after previews, MP4).

Audited, no change: wind-farm-meter-pedestal (status light floats below the head), aviation-light, turbine-door-frame (100 mm gap between leaf and header), foundation-anchor-cage (stainless label on anchor bolts; central drum stands above the bolts). The ledger records these limits.

Limits: in the preview the visible side of the pale shell is in shade and reads as mid-grey; the taper is the more visible change. The branch is not deployed.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 19, 2026

Copy link
Copy Markdown
Owner Author

Round: ferric dosing tank materials (eec7fc3)

Result: one material mismatch corrected in water-treatment/ferric-dosing-tank.

Defect: the tank, lid, bund and level column used a coatedSteel control; the vent and outlet used a stainlessSteel control. Ferric chloride attacks both metals. The level column stood 40 mm off the shell, and the outlet was an upright stub inside the tank wall.

Correction: four plastic controls replace the two steel controls: tankPolymer (#e6dfc4), bundPolymer (#5f7f95), pipePvc (#8b9499) and sightTube (#9a5a2a). The sight tube now touches the shell. The outlet is a horizontal stub from the shell to the bund edge. Triangles (360) and bounds (0.65 × 0.52 × 0.845 m) are unchanged.

Evidence (local):

  • Focused GLB regression: fails on the baseline, passes after the correction (1/1).
  • ESLint clean. Quality gate: water-treatment 50/50; catalog 2,398/2,402 (99.83%, four unchanged standard-components exceptions).
  • Catalog integration suite: 172 files / 242 tests pass (--maxWorkers=2).
  • Index: only the ferric-dosing-tank row changed; source SHA matches.
  • Bundle: artifacts/catalog-demos/water-treatment-ferric-dosing-tank-polymer-material-review-20261119/ (ledger, before/after previews, MP4).

Audited, no change: process-pipe-rack (the four process lines are upright rods), pump-station-control-panel and sample-tap-station (cabinet floats above the plinth; the hood is a band at mid-height), fine-bubble-diffuser-grid (diffusers on one lateral only). The ledger records these limits.

Limits: the control set changes, so a saved customisation of the two steel controls no longer applies. The bund is a flat slab with no walls. The branch is not deployed.

…ipe stubs

The rising stem used the painted coatedSteel control; it now uses the
existing stainlessSteel control. The two pipe stubs stood upright beside
the body and now run along the pipe axis. The body stands on its mount
foot. Adds a GLB regression test and the five-preview audit bundle.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 19, 2026

Copy link
Copy Markdown
Owner Author

Round: knife gate valve stem (e4b81bb)

Result: one material mismatch corrected in water-treatment/knife-gate-valve-assembly.

Defect: the rising stem used the painted coatedSteel control. A rising stem slides through the packing gland, so it is bare stainless steel. The two pipe stubs stood upright beside the body, and the body floated 20 mm above the mount foot.

Correction: stem uses the existing stainlessSteel control. pipe-left and pipe-right run along the pipe axis from each body face. gate-body stands on the foot. Colour controls, presets, part names, triangle count (356) and bounds (0.38 m × 0.64 m × 0.685 m) are unchanged.

Evidence:

  • Focused GLB regression knifeGateValveStemMaterial.test.ts: failed on the baseline, passes after the fix (1/1).
  • ESLint: clean. Quality gate: water-treatment 50/50; catalog 2,398/2,402 (99.83%; four unchanged standard-components exceptions).
  • Catalog integration suite: 173 files / 243 tests pass (--maxWorkers=2).
  • Index: only the target row changed; source SHA matches.
  • Bundle: artifacts/catalog-demos/water-treatment-knife-gate-valve-stem-material-review-20261119/ (ledger, before/after previews, MP4).

Reviewed, not changed: check-valve-assembly (cover plate on the pipe flange face; painted body with stainless flanges), lime-slurry-tank, coagulant-mixing-tank, flocculation-tank ("open" tanks with solid tops; paddles above or sunk in the top; flocculation description names three paddle arms, model has two).

Limits: the valve has no gate blade or yoke and its body is a 300 mm cube. Not deployed; no simulation or hardware claim.

…izontal body

The adjusting stem used the painted coatedSteel control; it now uses the
existing stainlessSteel control. The body stood upright on one pipe flange
with the bonnet on the face of the other; the body and both flanges now lie
on the pipe run axis and the bonnet rises from the top of the body. Adds a
GLB regression test and the five-preview audit bundle.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 19, 2026

Copy link
Copy Markdown
Owner Author

Round: pressure reducing valve stem (90ce135)

Result: one material mismatch corrected in water-treatment/pressure-reducing-valve-assembly.

Defect: the adjusting stem used the painted coatedSteel control. The stem turns through the actuator cover, so it is bare stainless steel. The body also stood upright on flange-left, with the bonnet on the face of flange-right, so the preview read as a pedestal.

Correction: stem uses the existing stainlessSteel control. The body and both flanges lie on the pipe run axis (X); the bonnet rises from the top of the body. Controls, presets, part names and the 532-triangle count are unchanged. Bounds change from 0.38 × 0.38 × 0.735 m to 0.34 × 0.38 × 0.735 m.

Evidence (local, verified):

  • Focused GLB regression: failed on the baseline source, passes after the correction (1/1).
  • ESLint clean. Kit quality gate 50/50 (100%). Global gate 2,398/2,402 (99.83%); the four exceptions are unchanged standard-components assets.
  • Catalog integration suite: 174 files / 244 tests pass (--maxWorkers=2), which includes the index freshness check.
  • Only the target index row changed; its source SHA matches.
  • Bundle: artifacts/catalog-demos/water-treatment-pressure-reducing-valve-stem-material-review-20261119/ (ledger, before/after previews, MP4, inspection).

Reviewed, no change (limits in the ledger): air-release-valve-assembly (upright "side" vent stub), static-mixer-spool (upright spool, box housing), flow-meter-spool (sensor boss and transmitter hidden inside the spool; the preview is a plain pipe), intake-gate-hoist (drum axis vertical, chain ends 50 mm above the gate).

Scope: local source, software-rendered preview and GLB regression only. Not deployed. No standards, simulation or hardware claim.

The eight anchor parts used a stainlessSteel control and sat hidden inside
the flange, so the control had no visible effect. A new anchorSteel control
(galvanized anchor steel) replaces it; each anchor is now a stud that stands
above the flange with a nut on the flange top, around a wider shell stub.
Adds a GLB regression test and the five-preview audit bundle.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 20, 2026

Copy link
Copy Markdown
Owner Author

Round: tower base flange anchors (1021a9f)

Result: one material mismatch corrected in wind-energy/tower-base-flange.

Defect: the eight anchor-* parts used a stainlessSteel control. Wind tower anchor bolts are galvanized or coated high-strength carbon steel. Each anchor was a 70 mm boss inside the 90 mm flange, so no anchor showed in the preview and the control had no visible effect. The sleeve was a 240 mm post on a 620 mm flange.

Correction: anchorSteel ("Galvanized anchor steel", #a9b6b7) replaces stainlessSteel. Each anchor is a stud that stands 60 mm above the flange, with a new anchor-nut-* part on the flange top. The shell stub is 380 mm wide. Triangles 440 → 792; budget 900 → 1,100; bounds unchanged.

Compatibility: the stainlessSteel control is removed from this asset; a saved scene must set anchorSteel.

Audited, not changed: turbine-blade-root-sleeve, blade-root-protector, tower-transport-saddle, turbine-number-placard. Limits are in the ledger (hidden protector tabs, flat saddle supports with upright "ties", placard with no number).

Verification (local):

  • Focused GLB regression: failed on the baseline, passes after (1/1).
  • ESLint: clean.
  • Quality gate: wind-energy 50/50; catalog 2,398/2,402 (four unchanged standard-components exceptions).
  • Catalog integration suite: 175 files / 245 tests pass (--maxWorkers=2).
  • Index: only the tower-base-flange row changed; source SHA matches.

Evidence: artifacts/catalog-demos/wind-energy-tower-base-flange-anchor-material-review-20261119/ (ledger, before/after previews, MP4, verification logs). Not deployed.

The two reflector bands used a stainlessSteel control and rendered as grey
steel collars. A new reflectiveBand control (reflective sheeting, default
red) replaces it; each band is now an 80 mm wrap that stands 3 mm proud of
the post. Triangle count and bounds are unchanged.
Adds a GLB regression test and the five-preview audit bundle.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 20, 2026

Copy link
Copy Markdown
Owner Author

Round: site bollard reflector bands (a7b66a9)

Result: one material mismatch corrected in wind-energy/wind-site-bollard.

Defect: the two reflector bands (band-low, band-high) used a stainlessSteel control. A reflector band is retroreflective sheeting. The preview showed two grey steel collars, 8 mm proud of the post, and no reflector.

Correction: reflectiveBand ("Reflective sheeting", default red #d5322c) replaces stainlessSteel. Each band is an 80 mm wrap, 3 mm proud of the post, at the same centre height. Triangles (188), budget (400) and bounds are unchanged. A white default was tried first; the renderer shades it to mid grey, so it still read as steel.

Compatibility: the stainlessSteel control is removed from this asset; a saved scene must set reflectiveBand.

Audited, not changed: crane-outrigger-pad, cable-trench-cover, foundation-inspection-hatch, turbine-junction-box. Limits are in the ledger (steel top plate hidden inside the pad, trench-cover "slots" drawn as painted patches, hatch bolts hidden inside the lid, junction-box lid drawn as a mid-body band with floating feet).

Verification (local):

  • Focused GLB regression: failed on the baseline, passes after (1/1).
  • ESLint: clean.
  • Quality gate: wind-energy 50/50; catalog 2,398/2,402 (four unchanged standard-components exceptions).
  • Catalog integration suite: 176 files / 246 tests pass (--maxWorkers=2).
  • Index: only the wind-site-bollard row changed; source SHA matches.

Evidence: artifacts/catalog-demos/wind-energy-site-bollard-reflector-band-material-review-20261120/ (ledger, before/after previews, MP4, verification logs). Not deployed.

…aterials

The drive-over cable protector was a stainless steel slab with powder-coated
metal blocks on it. rampElastomer, coverPolymer and anchorSteel controls
replace the two metal controls. The body is now a 70 mm ramp with sloped
flanks and sloped end caps, the two lids lie on the deck, and four anchor
heads stand on the flanks where they show. Triangles go from 160 to 240;
bounds go from 1.33 x 0.36 x 0.15 m to 1.24 x 0.50 x 0.078 m.
Adds a GLB regression test and the five-preview audit bundle.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 20, 2026

Copy link
Copy Markdown
Owner Author

Round: parking cable protection ramp materials (2edc71a)

Result: one material mismatch corrected in ev-charging-infrastructure/parking-cable-protection-ramp.

Defect: the drive-over cable protector used a stainlessSteel control for its base and anchors and a powderCoat control for its covers, divider and end slopes. A cable protector is a moulded rubber body with polymer lids. The preview showed a grey steel slab with vertical sides and teal metal blocks; the "end slopes" were square blocks and both anchors were hidden inside the covers.

Correction: rampElastomer ("Moulded rubber", #474b50), coverPolymer ("High-visibility polymer lid", #f2b824) and anchorSteel ("Galvanized anchor steel", #b5c0c2) replace the two metal controls. The body is a 70 mm ramp with flanks that slope to the ground (500 mm at the ground, 220 mm at the deck), two lids 6 mm proud of the deck, sloped end caps, and four anchor heads on the flanks. Triangles: 160 → 240 (budget 2,400).

Compatibility: powderCoat and stainlessSteel are removed from this asset; a saved scene must set the new controls. The eight earlier part names are kept; anchor-left-rear and anchor-right-rear are added. Bounds change from 1.33 × 0.36 × 0.15 m to 1.24 × 0.50 × 0.078 m.

Audited, not changed: charging-island-plinth, ev-charger-protection-bollard, charging-station-direction-sign, wall-mounted-connector-holster. Limits are in the ledger. The plinth has a second material mismatch (site island and curb labelled powder-coated metal, not concrete); it is the best next correction in this kit. The bollard flange is round where the description says square. The direction-sign top cap floats 60 mm above the post and the arrow has no head.

Verification (local):

  • Focused GLB regression: failed on the baseline, passes after (1/1).
  • ESLint: clean.
  • Quality gate: ev-charging-infrastructure 50/50; catalog 2,398/2,402 (four unchanged standard-components exceptions).
  • Catalog integration suite: 177 files / 247 tests pass (--maxWorkers=2).
  • Index: only the parking-cable-protection-ramp row changed; source SHA matches.

Not updated: docs/benchmarks/token-efficiency/sample-manifest.json holds a frozen hash of the earlier source of this asset. No test reads it.

Evidence: artifacts/catalog-demos/ev-charging-parking-cable-protection-ramp-elastomer-material-review-20261120/ (ledger, before/after previews, MP4, verification logs). Not deployed.

…onnectors

The two antennas used the powder-coated metal control of the enclosure and
floated 18 mm above the top cap. An antennaPolymer control now colours two
slim whips, each on a new stainless connector that stands on the cap. The
mounting plate moves from the middle of the enclosure to its rear wall and
the status button sits on the door with its axis toward the operator.
Triangles go from 240 to 328; bounds go from 0.56 x 0.352 x 1.10 m to
0.56 x 0.3585 x 1.082 m.
Adds a GLB regression test and the five-preview audit bundle.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 20, 2026

Copy link
Copy Markdown
Owner Author

Round: site energy controller antennas (3915f3c)

Result: one material mismatch corrected in ev-charging-infrastructure/site-energy-management-controller.

Defect: antenna-left and antenna-right used the powderCoat control ("Powder-coated metal") of the enclosure. An antenna whip has a polymer radome on a metal connector. The preview showed two thick teal rods that floated 18 mm above the top cap. The mounting plate cut through the middle of the enclosure, and the status button floated 9 mm in front of the door.

Correction: a new antennaPolymer control ("Antenna radome polymer", #2b3236) colours two 22 mm whips. Two new antenna-mount-* stainless connectors stand on the top cap. mounting-plate is now on the enclosure rear wall. status-button sits on the door with its axis toward the operator. Triangles: 240 → 328 (budget 2,400).

Compatibility: the four earlier colour controls keep their names, labels and defaults. powderCoat no longer colours the antennas. The twelve earlier part names are kept; two connector parts are added. Bounds change from 0.56 × 0.352 × 1.10 m to 0.56 × 0.3585 × 1.082 m.

Audited, not changed: charger-canopy-column-base, curbside-lamp-charger, streetlight-charger-arm, fleet-charging-canopy. Limits are in the ledger. The column base has a second material mismatch (teal powder-coated gussets on a stainless structural column). The two lamp chargers differ only in mast height, and each arm sticks out 265 mm behind the mast. The canopy cable drops are horizontal rods and its "rafters" are roof slabs.

Verification (local):

  • Focused GLB regression: failed on the baseline, passes after (1/1).
  • ESLint: clean.
  • Quality gate: ev-charging-infrastructure 50/50; catalog 2,398/2,402 (four unchanged standard-components exceptions).
  • Catalog integration suite: 178 files / 248 tests pass (--maxWorkers=2).
  • Index: only the site-energy-management-controller row changed; source SHA matches.

Not updated: docs/benchmarks/token-efficiency/sample-manifest.json holds a frozen hash of the earlier source of this asset. No test reads it.

Evidence: artifacts/catalog-demos/ev-charging-site-energy-controller-antenna-material-review-20261120/ (ledger, before/after previews, MP4, verification logs). Not deployed.

The two wheels used the powder-coated metal control of the enclosure. They
were upright drums in front of the door, and the enclosure stood on the
ground through the chassis. A tyreRubber control now colours four tyres on
horizontal axles beside a lowered chassis, each with a painted hub. The
enclosure stands on the chassis, two brackets carry the push handle, the
cable reel touches the rear wall and the status button sits on the door.
Triangles go from 384 to 672; bounds go from 0.82 x 1.178 x 1.182 m to
0.956 x 1.05 x 1.182 m.
Adds a GLB regression test and the five-preview audit bundle.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>
@w1ne

w1ne commented Nov 20, 2026

Copy link
Copy Markdown
Owner Author

Round: mobile fast-charge cart tyres (47044ac)

Result: one material mismatch corrected in ev-charging-infrastructure/mobile-fast-charge-cart.

Defect: left-wheel and right-wheel used the powderCoat control ("Powder-coated metal") of the enclosure. A cart wheel has a rubber tyre on a metal hub. The preview showed two upright teal drums in front of the door, 30 mm above the ground. The enclosure stood on the ground through the chassis, so the cart could not roll. The push handle, the cable reel and the status button floated.

Correction: a new tyreRubber control ("Wheel tyre rubber", #23282b) colours four 200 mm tyres on horizontal axles beside the chassis. Four new hub parts use powderCoat. The chassis is at 70 mm to 160 mm and the enclosure stands on it. Two brackets carry the push handle, the cable reel touches the rear wall, and the status button sits on the door. Triangles: 384 → 672 (budget 2,400).

Compatibility: the four earlier colour controls keep their names, labels and defaults. powderCoat no longer colours the tyres. The sixteen earlier part names are kept; eight parts are added. Bounds change from 0.82 × 1.178 × 1.182 m to 0.956 × 1.05 × 1.182 m (width × depth × height). The description no longer says "casters", because the wheels do not swivel.

Audited, not changed: coiled-charging-cable-reel, charger-cable-retractor, ev-metering-pedestal, home-ac-wallbox. Limits are in the ledger. The reel has no cable on its drum and its spool core is at a right angle to the axle. The retractor is a floor mast under an "overhead" description, with a vertical-axis pulley. The meter dial lies flat. The wallbox cable hook floats 41 mm beside the enclosure.

Verification (local):

  • Focused GLB regression: failed on the baseline, passes after (1/1).
  • ESLint: clean.
  • Quality gate: ev-charging-infrastructure 50/50; catalog 2,398/2,402 (four unchanged standard-components exceptions).
  • Catalog integration suite: 179 files / 249 tests pass (--maxWorkers=2).
  • Index: only the mobile-fast-charge-cart row changed; source SHA matches.

Not updated: docs/benchmarks/token-efficiency/sample-manifest.json holds a frozen hash of the earlier source of this asset. No test reads it.

Evidence: artifacts/catalog-demos/ev-charging-mobile-fast-charge-cart-tyre-material-review-20261120/ (ledger, before/after previews, MP4, verification logs). Not deployed.

The display bezel used the display glass control of the screen, so the
frame and the glass rendered as one pale block. The screen also sat on the
lower cabinet door at 354 mm, although the asset is a wayfinding-height
display post. A bezelPolymer control now colours a frame 20 mm wider than
the glass on each side. The bezel and glass move to a new powder-coated
display housing on the column head, with the glass bottom edge at 1.45 m
and the glass 3 mm proud of the bezel. Triangles go from 252 to 264; depth
goes from 0.32 m to 0.321 m.
Adds a GLB regression test and the five-preview audit bundle.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: w1ne <41898282+w1ne@users.noreply.github.com>

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant