ios: guided wall scan from the meter to a battery spot by SamGu-NRX · Pull Request #10 · SamGu-NRX/BaseScanning

SamGu-NRX · 2026-09-26T07:45:58Z

What changed

The iOS app runs the homeowner's whole scan, from finding the electric meter to seeing where a Base Power battery could go. It sends the server the measurements it needs to decide, and says what it didn't see.

Meter. The homeowner points at the meter and taps. The scan anchors on the wall surface there: a plane ARKit detected, or ARKit's estimated vertical plane with a wider meter error in the export. The close-up photo takes itself once the picture has been sharp, lit and centred for 0.6 s. Vision's accurate mode reads the meter number, with the focus and size checks from the close-up eval (experiments/meter-closeup on t3/meter-closeup). Then the app asks "Which number is on your meter?" with up to three candidates, barcode-confirmed first, and "None of these". It never fills the number in. A close-up that passes the focus check also counts as one camera position toward coverage.
Wall walk. The homeowner walks left, then right, along the wall. The phone keeps a photo every 0.5 m or 15° of turn, or when the frame shows wall it hasn't seen. It keeps photos only while tracking is normal and the picture is sharp.
- A haze covers wall and ground the phone hasn't seen, and a strip at the bottom shows the wall unrolled with the meter at zero.
- A dotted path and an aim ring show where to stand and where to point. Coaching replaces the instruction while something is wrong.
- The first request is the ground in front of the meter, the stretch every check needs. "Wall ends here" is offered whenever the homeowner is walking a side, and "Can't get there" ends the wall where the phone is, never beyond the path already walked. Before either, the strip shows a dashed line where the end would land and dims what it would leave out.
- After ending, the homeowner says whether the wall turns a corner or something blocks it. At a corner they mark the next wall and keep walking, and it becomes the next entry in walls[].
- Before leaving, they tilt up over the meter and say whether anything is overhead.
Marking. They can pin a gas meter, AC unit, door, window, driveway edge or fence as they pass. Marks past an end still export, because the server places them on the wall's continued line and decides which it can use.
The ground question (new). The feature review opens with "What's on the ground along this wall?": grass, mulch, gravel, concrete, driveway, deck, or "Not sure". The camera can't tell mulch from soil, and without a ground type every real scan ended with the server's "Ground under the battery" check unsure. An answer is exported as a ground patch of that type covering only the ground the coverage saw. "Not sure" sends no patch. The question doesn't block "Looks complete".
What the server needs. coverage.observed follows "What settles each check" in server/README.md (t3/server at 9e42970):
- wall entries with out_ft, the height up the wall the phone saw from two positions;
- ground entries with how far out the ground was seen;
- the clear space in front of the wall that the walked path proves, less the wall's position error for its source;
- the overhead band from the tilt-up;
- ground past a limit end.
  Each wall says how its line was found in walls[].source. A wall through a plane ARKit detected is plane, and a tap on an estimated plane or a replayed wall is tap. Nothing yet builds a wall line from the LiDAR mesh. Ground entries are split at corners, and every span and reach rounds inward, so no entry claims ground the camera didn't sample (see Why it works this way).
Depth on LiDAR phones. The session runs scene depth and mesh reconstruction. A cell counts only where depth at its projection matches its distance. A nearer reading marks it hidden (a bush or bin in front of the wall), and a later view without depth can't clear that. The walk then asks the homeowner to look past it.
One more view. After the walk, the phone asks for the most useful stretch it hasn't seen from two positions, in one sentence, and moves on once the camera has covered it. It does the same for evidence the server lists in missing_evidence, including a wall seen too low: those requests carry the height the view must exceed. "I can't get there" leaves the gap for an installer.
Result. The phone posts scene.json to POST /v1/placements on the hosted server (Config/Shared.xcconfig) and shows the answer two ways. The first is a 3D model of the wall with the battery, the cable and the clearances. The second is the battery on the live camera, placed on the meter's anchor, where people and objects in front of it hide it. The cable now bends at every corner of the wall chain it passes, where before it was projected onto the battery's own wall. Unknown missing_evidence kinds from a newer server go to installer review; the whole answer no longer fails on them.
Getting a scan off the phone. "Share scan" shares a capture packet 1.1 (S4's spec, packet/README.md on t3/packet at d5439cf). New since 1.0:
- top-level planes with boundary_m;
- confidence on all ARKit depth;
- session.device.mesh_classification_enabled;
- depth_frames between photos at 2 Hz. They are written to disk as they arrive, stop at 300 frames, and are recorded only during the close-up, the walk and "one more view".
- a writer entry point for S6's on-device estimated depth with sigma.
Photos leave the phone only through this share sheet, and no location or heading is recorded.

The capture logic is a Swift package, ios/HouseScanKit. It imports only Foundation, simd, CryptoKit (packet hashes) and ImageIO, so its 366 tests run on macOS. HouseScan/Runtime feeds it ARKit or a replay and runs the flow; HouseScan/UI draws the screens. ScanContract.swift is the whole interface between the two.

For the Simulator (contract C4):

-replay <folder> plays a Measure Lab v2 session instead of the camera.
-autopilot drives every step, including a gap request, and answers the ground question with mulch.
-serverURL <url> picks the server; -sampleResult answers with a bundled sample that every screen labels as an example.
-uiDemo… arguments open single states. -uiDemoGroundQuestion is new.

Each screen change is logged as STATE=<name>. ios/README.md lists every argument and the steps to run on an iPhone.

Aiden's PRs into this branch are part of it:

ios: three small fixes from field test run 1 (#25, #26, #31) #34 (feb35a0): "AC unit" keeps its capitals, "Slow down" stays hidden while the mark tray is open, and the upload moves on to "Check clearances" once the body is sent.
ios: result rule lines say at least / at most (B-14, #36) #37: result rule lines say "at least" or "at most".
ios: past-end requests keep their end; no requests past a wall end (#35, #38) #47: past-end requests keep their end, and nothing is requested past a wall end.
ios: end the side the phone is on, and no zero-length walls (#24) #48: "Wall ends here" ends the side the phone is on, and a walk can't finish with zero-length walls.
ios: AR result is hidden by people and objects in front of it (#41) #51: the AR result is drawn in the camera's RealityKit scene, so people and, with LiDAR, the mesh hide it.

#37, #47, #48 and #51 were merged before their reviews cleared. Their open findings are listed under Not done. #49, #52, #53, #55, #56 and #57 are open against this branch.

Why it works this way

Coverage counts a cell only when two positions saw it. Each 6-inch cell is covered once each of its rows has been seen from two camera positions at least 0.25 m apart with normal tracking. The coverage map keeps the cameras and rebuilds itself when the wall moves (a corner, a better meter anchor).
The phone reports the height it saw; the walk asks for less. The server counts a wall stretch as seen for a check only when out_ft exceeds that check's height: 6.5 ft for gas and openings, 3.29 ft for the battery's backing, 1 ft for the cable.
- What gets reported: per stretch, the highest row seen from two positions with no gap below it.
- While the ground is still a chest-height guess: the guess's 0.3 m error comes off the reported height, and the rows down to that error must be seen too.
- What the walk asks for: the wall up to 4.5 ft (wallWalkHeight). That is the most a view from the 2 m stand-off can show while it also shows the ground in front of the wall; the long-side half angle of a phone's image is about 30° to 31°.
- Above 4.5 ft: where a check needs more, the server's answer asks for that stretch at the height it needs, and the gap loop raises it as the next view. Rows are sampled up to 7.5 ft (wallCaptureHeight), so a tilt-up can show up to 7.5 ft.
- Neither number is measured: no calibration run chose 4.5 ft or 7.5 ft.
The ground patch stops where coverage stops. The server treats unrecorded ground within its error margin of the footprint as unsure, so a patch drawn wider than what was seen would pass ground nobody looked at. The patch is built from the ground entries as written. It leaves out the wedge outside a convex corner, and it is cut at the neighbouring wall at an inside corner. Its wall-side edge starts 0.001 ft behind the wall line. That is inside the house, where the server models no ground: on the hosted server a rotated wall whose patch started exactly on the line left ground_surface unsure.
Ground entries are split at corners. When one ground entry crosses a corner, the server's band_polygon includes the sector outside it, and no camera sample lies there. The export splits every ground entry at the chain's corners after merging.
Wall source says how the line was found, and the phone subtracts that source's error. The server gives a detected-plane wall 0.75 ft of default error and a tap 0.3 ft. The walked clearance the phone reports subtracts the larger error of the two wall pieces at each cell. The phone mirrors those defaults from rules.yaml in ServerErrorDefaults.swift; they go stale if the rules change.
Capture overlays are projected from the camera pose. The haze, path and ring are world points projected through each frame's pose and intrinsics. The same code runs on a live frame and a replayed keyframe, so the Simulator tests the real overlay geometry. Nothing occludes them. The AR result is the exception since ios: AR result is hidden by people and objects in front of it (#41) #51: on a live camera it is RealityKit content on the meter's anchor, and over a replay it is still projected.
An end goes where the homeowner is. In the first phone run, "Can't get there" put each end at the unbroken covered stretch from the meter. That stretch was 0 because nothing had asked for the ground by the meter, so every photo past the ends was dropped. An end now goes at the phone's position along the wall, kept within the path it walked on that side (WalkedEnd). Unseen stretches in between stay unseen, and the server asks for them. A photo taken past an unexplored end that sees nothing between the ends isn't kept, and the card says so.
Losing tracking late doesn't cost the scan. If ARKit can't relocalize within 20 s, the find-meter, close-up and walk steps start again from the meter. On the feature review the scan now stays: "Add something" waits for tracking, and "Looks complete" uploads, because the upload needs no live frame. During "one more view" the timeout counts as "I can't get there".

UI changes

New since the last description, all from the scripted demo engine at dd44fc3. The ground question is new. The number picker shipped earlier but had no capture here. These three images still need uploading to GitHub from a signed-in browser. Until then they are local files on the build Mac:

Ground question	Answered	Meter number picker
`/tmp/hs-pr-shots/19-ground-question.png`	`/tmp/hs-pr-shots/20-ground-answered.png`	`/tmp/hs-pr-shots/21-number-picker.png`
Six ground types in two columns, "Not sure" across the width, one line on why	Folds to one row with "Change"; the review below moves up with it	Up to three candidates, the barcode match first; the numbers are made up by the demo engine

The rest are from earlier commits and still show the current screens. The replay rows are the real engine on the synthetic replay at beede15 with the bundled sample result. Row 9 is a real server answer from 0876e03. "Demo" rows come from the scripted demo engine. No image shows a real home or dataset photos.



1. Onboarding	2. Find the meter	3. Close-up, "Can't get a clear shot" (demo)

4. Wall walk	5. End of the wall (demo)	6. Feature review, before the ground question

7. One more view, satisfied	8. No server: "Making a sample result"	9. A server's answer (`0876e03`)

10. The spot on the wall	11. Coaching (demo)	12. Lost place (demo)

13. A refused mark (demo)	14. Offline (demo)	15. Camera denied (demo)

16. A pass (demo)	17. Walk at AX5 (demo)	18. Result at AX5 (demo)

The autopilot on the synthetic replay at cbe0440, from onboarding to the AR view, at 3× replay speed:

house-scan-flow.mp4

Verification

swift test --package-path ios/HouseScanKit -Xswiftc -warnings-as-errors: 366 tests in 58 suites pass at d02cc11. They include:
- every exported scene validated against the vendored scene.schema.json;
- drift tests comparing the vendored schemas byte for byte with the heads of t3/server (9e42970) and t3/packet (d5439cf). CI now fetches both branches, so these compare instead of skipping; a server schema change fails every iOS run until the copies are refreshed.
- a port of the server's settlement rules, checked against server/scene.py;
- ground patches round convex, concave and sharp corners and at chain ends;
- NoUnseenClaimsTests, one case per overclaim path the review found.
CI (Xcode 26.6) runs those tests, the XcodeGen drift check, the device build and every UI test with each screen's accessibility audit enforced. Green at d02cc11 (run 36279804539).
Local, Xcode 27, iPhone 17 Simulator (iOS 26.5), at e9e405d, before the field fix:
- testFullFlowFromReplay and testFullFlowFromLidarReplay pass.
- S4's validator (uv run python -m packet validate from packet/ at d5439cf) accepts both exported packets as 1.1: 32 photos each, and 38 depth frames from the LiDAR replay.
- A package-built packet with ARKit and estimated depth, a mesh and two planes (one with an L-shaped boundary) also passes. The same packet with its boundary left in anchor coordinates fails, as it should.
- The exported scene.json validates against the server schema. It carries source: tap, a mulch patch and out_ft on every wall entry.
The hosted server (redeployed at 930e8e5) answers that scene manual_review. Its three requests are correct for the fixture:
- a wall view with out_ft 6.500001 over −3.5 to 6.5 ft, for the gas, opening and wall-equipment checks, because the synthetic camera sees the wall only 6.5 ft up;
- ground out to 12.6 ft;
- the unexplored left end.
With the mulch patch, ground_surface is unsure only because the footprint sits within measurement error of the patch edge (margin); before, no surface was recorded at all. wall_backing, facing_gap, headroom and both route checks pass.
Two fresh reviews at 35ac3f2: an Opus review of the whole diff and an Astra review of one question, whether any path exports coverage, a ground patch or a wall the camera didn't see. Fixed from them:
- a relocalization timeout on the review screen wiped the scan;
- a spatial reset left marks, skipped gaps, the result and the old close-up behind;
- live frames could reach the engine out of order;
- depth frames and sensor streams were recorded in the wrong phases and after a failed upload;
- ground entries crossing a corner claimed the unseen outside sector;
- a view without depth could cover a cell depth had marked hidden;
- a blurry close-up, or one being retaken, could count toward coverage;
- facing clearance used the tap error on plane walls;
- a late photo save could join the walked path across a tracking break;
- spans and patch edges rounded outward;
- two drift tests passed without comparing anything.
The Astra reviewer then re-checked at e9e405d. Its fixes held, and it found four more paths, fixed in b7f51ba:
- the server works out each corner from the rounded baseline, so a split entry could still cross the corner by 3e-5 ft;
- a view without depth kept its credit after a depth view proved the row hidden;
- a guessed ground that sits too high left the real wall foot unseen;
- a pending overhead photo joined the walked path across a tracking break.
The caretaker's Astra review of the field fix found two more, fixed in 5c8c739: a nearly parallel aim could put an end hundreds of metres away, and an object past an end before a corner named a wall the scene leaves out.

Not done, and what could go wrong

One real-iPhone run so far, and it lost the walk. Aiden ran TestFlight build 657ab28 at a mock wall (experiments: device field test runbook and run 1 of the TestFlight build #23, experiments/device-field-test/README.md on t3/device-field-test). "Can't get there" set each end at the unbroken covered stretch from the meter, which was 0 because the walk never asked for the ground in front of the meter. The scene therefore reported no wall and no ground. The fix (3c7b83a, merged at 75dea41) is described under Why it works this way; issues "Can't get there" during the walk can put the wall's end at the meter, and the rest of the walk is dropped #24, Tilt-down prompts never ask for the ground in front of the meter #28 and The "Is this the end of the wall?" prompt never appeared on a 19 ft walk #29 carry the details for Aiden's retest. His marks against a tape, from the meter's centre line: window right edge −0.5 in, sill +1.9 in, left edge +5.4 in, gas stand-in 4.8 in nearer the meter. All four are inside the server's allowance for AR taps (0.3 ft plus 0.16 ft per ft from the meter). These are the first device measurements.
Most of the live path hasn't run on a device. Four things have only built and run in the Simulator: the live ARKit path (meter raycast, anchor refresh, ground refinement, relocalization), the plane pose and boundary_m math, depth-frame recording and the 7.5 ft capture height. Apple's documentation matches the plane convention the writer uses: since iOS 16 the anchor transform isn't rotated, and the extent's rotationOnYAxis is applied to the geometry. No device has confirmed its direction. A wrong sign would drop boundaries (logged), not break the packet. Every capture threshold is a guess, stated where it is defined.
The 4.5 ft walking height rests on an assumed field of view. If a phone's view is narrower than the 30° to 31° half angle assumed, the walk will still ask for some tilt-ups. The camera intrinsics recorded in runs 2 and 3 can settle this.
The fixture never settles the 6.5 ft checks. The synthetic camera sees the wall 6.5 ft up at most, and its three tilt-up frames share one position. So every replay answer asks for one more wall view, and the autopilot says "I can't get there". A second tilt-up position in Tools/make-synthetic-replay.swift would fix that.
Coverage still stands in for visibility in three places:
- A cell is judged by points on its row edges, not its whole area, so a small obstruction between sampled rows can go unnoticed.
- The walked path counts as clear space between the wall and the phone even if a hedge stands in it.
- The overhead answer covers the upper wall, not the volume above the battery; the homeowner's "nothing overhead" carries that.
S6's Map3D (ios: live 3D map as the coverage model where depth exists #21) replaces this coverage model. This PR stops adding occlusion to the old one.
One ground answer for the whole wall. A mixed wall can't be described. Answering "Driveway" or "Deck" fails the ground check at every spot, which the rules intend. Bare soil has no answer of its own, so a homeowner with a dirt bed will likely pick "Not sure".
Wall source and error. A wall on a detected plane gets 0.75 ft of default error on the server, while a tap on ARKit's estimated plane gets 0.3 ft. The estimated tap is likely the less accurate of the two. The phone reports the source truthfully and leaves the error values to the server's rules.
The UI tests never see a real server's request. They run on the sample result. The real-answer path is covered by package tests on a recorded server answer.
Timing flakes. ScreenStatesUITests/testAddSomethingFromTheReviewMarksOnTheCamera failed once locally, tapping a chip while the review fades back in, and then passed 3 of 3. The full-flow audit now waits 6 s before its second pass, after a system notification banner failed a contrast check on CI (run 36264789032).
Main-thread copy at upload. Copying the mesh and planes for the packet runs on the main actor. Nobody has measured how long that stalls on a large mesh.
The ADVIO walk can't close its gap loop. Under the two-position rule it never covers the ground beside any assumed meter, so the autopilot skips the gap. That replay looks along a path, not at a wall.
Open findings on merged PRs, posted to Aiden as follow-ups:
- ios: end the side the phone is on, and no zero-length walls (#24) #48: the short-wall refusal copies the server's battery width, and it clears an answered limit end; WalkedEnd.side can fall back to the other side; the card loses the task's detail while wallTooShort is set.
- ios: past-end requests keep their end; no requests past a wall end (#35, #38) #47: an end the app infers is written into the packet as a homeowner wall_end mark.
- ios: AR result is hidden by people and objects in front of it (#41) #51: support is checked for person segmentation alone but it runs with scene depth (a crash risk on some phones); the lawn mesh can hide the clearance zones; no automated test reaches the RealityKit path.
- ios: result rule lines say at least / at most (B-14, #36) #37: the demo's overlap example draws a gap beside an overlap warning.
The 3D view frames the whole marked wall. On a long wall the battery looks small.

Checklist

The scope is focused; related changes belong together. This is one feature, the capture app.
I explained what changed and why
I included before/after images for visible changes: the three new captures above still need uploading
I included a video for motion or interaction changes

🤖 Generated with Claude Code

vercel · 2026-09-26T07:46:03Z

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
house-scanning-web	Ignored	Preview	Sep 26, 2026 11:12am UTC

coderabbitai · 2026-09-26T07:46:06Z

Important

Review skipped

Too many files!

This PR contains 330 files, which is 230 over the limit of 100.

To get a review, reduce the PR to 100 files or fewer by splitting it into smaller PRs or changing its base branch.

Upgrade to a paid plan to raise the limit.

Usage-priced reviews support at most 300 files.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: 47bb261d-0380-4e02-8cc5-70ad458154d2

📥 Commits

Reviewing files that changed from the base of the PR and between f852207 and c820cf1.

⛔ Files ignored due to path filters (123)

ios/HouseScan/Assets.xcassets/AppIcon.appiconset/AppIcon.png is excluded by !**/*.png
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00001.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00002.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00003.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00004.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00005.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00006.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00007.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00008.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00009.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00010.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00011.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00012.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00013.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00014.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00015.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00016.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00017.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00018.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00019.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00020.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00021.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00022.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00023.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00024.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00025.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00026.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00027.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00028.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00029.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00030.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00031.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00032.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00033.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00034.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00035.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00036.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00037.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00038.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00039.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00040.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00041.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00042.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00043.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00044.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00045.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00046.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00047.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00048.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00049.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00050.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00051.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00052.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00053.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00054.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00055.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00056.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00057.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00058.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00059.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00060.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00061.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00001.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00002.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00003.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00004.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00005.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00006.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00007.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00008.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00009.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00010.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00011.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00012.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00013.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00014.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00015.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00016.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00017.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00018.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00019.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00020.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00021.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00022.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00023.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00024.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00025.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00026.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00027.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00028.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00029.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00030.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00031.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00032.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00033.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00034.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00035.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00036.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00037.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00038.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00039.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00040.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00041.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00042.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00043.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00044.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00045.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00046.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00047.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00048.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00049.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00050.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00051.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00052.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00053.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00054.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00055.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00056.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00057.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00058.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00059.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00060.jpg is excluded by !**/*.jpg
ios/HouseScanUITests/Fixtures/synthetic-wall/keyframes/k00061.jpg is excluded by !**/*.jpg

📒 Files selected for processing (330)

.github/labels.json
.github/workflows/ios.yml
CONTRIBUTING.md
Makefile
ios/.gitignore
ios/Config/Info.plist
ios/Config/Shared.xcconfig
ios/HouseScan.xcodeproj/project.pbxproj
ios/HouseScan.xcodeproj/xcshareddata/xcschemes/HouseScan.xcscheme
ios/HouseScan/ARCaptureView.swift
ios/HouseScan/Assets.xcassets/AccentColor.colorset/Contents.json
ios/HouseScan/Assets.xcassets/AppIcon.appiconset/Contents.json
ios/HouseScan/Assets.xcassets/Canvas.colorset/Contents.json
ios/HouseScan/Assets.xcassets/Caution.colorset/Contents.json
ios/HouseScan/Assets.xcassets/Chalk.colorset/Contents.json
ios/HouseScan/Assets.xcassets/Contents.json
ios/HouseScan/Assets.xcassets/CoverageCovered.colorset/Contents.json
ios/HouseScan/Assets.xcassets/CoverageHidden.colorset/Contents.json
ios/HouseScan/Assets.xcassets/CoverageSeen.colorset/Contents.json
ios/HouseScan/Assets.xcassets/CoverageSkipped.colorset/Contents.json
ios/HouseScan/Assets.xcassets/CoverageUnseen.colorset/Contents.json
ios/HouseScan/Assets.xcassets/Danger.colorset/Contents.json
ios/HouseScan/Assets.xcassets/Ink.colorset/Contents.json
ios/HouseScan/Assets.xcassets/Muted.colorset/Contents.json
ios/HouseScan/Assets.xcassets/Signal.colorset/Contents.json
ios/HouseScan/Assets.xcassets/SignalText.colorset/Contents.json
ios/HouseScan/Assets.xcassets/Surface.colorset/Contents.json
ios/HouseScan/CaptureScreen.swift
ios/HouseScan/CaptureSessionModel.swift
ios/HouseScan/Contract/ResultPresentation+Reading.swift
ios/HouseScan/Contract/ScanContract.swift
ios/HouseScan/HouseScanApp.swift
ios/HouseScan/Runtime/AppRuntime.swift
ios/HouseScan/Runtime/Autopilot.swift
ios/HouseScan/Runtime/CaptureRecorder.swift
ios/HouseScan/Runtime/GuidanceLog.swift
ios/HouseScan/Runtime/KeyframeStore.swift
ios/HouseScan/Runtime/LaunchOptions.swift
ios/HouseScan/Runtime/LiveCapture.swift
ios/HouseScan/Runtime/MeterNumberReading.swift
ios/HouseScan/Runtime/MeterReader/VisionMeterNumberReader.swift
ios/HouseScan/Runtime/ReplayPlayer.swift
ios/HouseScan/Runtime/ResultClient.swift
ios/HouseScan/Runtime/SampleResult.json
ios/HouseScan/Runtime/ScanEngine+Actions.swift
ios/HouseScan/Runtime/ScanEngine+Confirm.swift
ios/HouseScan/Runtime/ScanEngine+Ends.swift
ios/HouseScan/Runtime/ScanEngine+Export.swift
ios/HouseScan/Runtime/ScanEngine+GuidanceLog.swift
ios/HouseScan/Runtime/ScanEngine+Packet.swift
ios/HouseScan/Runtime/ScanEngine+Snapshot.swift
ios/HouseScan/Runtime/ScanEngine+Stamp.swift
ios/HouseScan/Runtime/ScanEngine.swift
ios/HouseScan/Runtime/SourceFrame.swift
ios/HouseScan/SessionDelegate.swift
ios/HouseScan/UI/Camera/CameraBackdrop.swift
ios/HouseScan/UI/Camera/CameraOverlays.swift
ios/HouseScan/UI/Camera/FogOverlay.swift
ios/HouseScan/UI/Camera/WallMarksOverlay.swift
ios/HouseScan/UI/Camera/WallProjection.swift
ios/HouseScan/UI/Camera/WallTape.swift
ios/HouseScan/UI/Camera/WayfindingOverlay.swift
ios/HouseScan/UI/Components/AnswerButton.swift
ios/HouseScan/UI/Components/Buttons.swift
ios/HouseScan/UI/Components/CameraChrome.swift
ios/HouseScan/UI/Components/CenteredScroll.swift
ios/HouseScan/UI/Components/GroundAnswers.swift
ios/HouseScan/UI/Components/InstructionCard.swift
ios/HouseScan/UI/Components/OverheadAnswers.swift
ios/HouseScan/UI/Components/PhotoCounter.swift
ios/HouseScan/UI/Components/SavedMeterPhoto.swift
ios/HouseScan/UI/Components/Scrim.swift
ios/HouseScan/UI/Components/ShareScanButton.swift
ios/HouseScan/UI/Components/TapRipple.swift
ios/HouseScan/UI/Copy/Distance.swift
ios/HouseScan/UI/Copy/ScanCopy.swift
ios/HouseScan/UI/Copy/SpotCheckCopy.swift
ios/HouseScan/UI/Preview/DemoEngine.swift
ios/HouseScan/UI/Preview/DemoScene.swift
ios/HouseScan/UI/Preview/UIDemo.swift
ios/HouseScan/UI/Result/ResultScene3D.swift
ios/HouseScan/UI/ScanRootView.swift
ios/HouseScan/UI/Screens/FindMeterScreen.swift
ios/HouseScan/UI/Screens/GapRequestScreen.swift
ios/HouseScan/UI/Screens/MarkFeaturesScreen.swift
ios/HouseScan/UI/Screens/MeterCloseUpScreen.swift
ios/HouseScan/UI/Screens/OnboardingScreen.swift
ios/HouseScan/UI/Screens/ResultARScreen.swift
ios/HouseScan/UI/Screens/ResultScreen.swift
ios/HouseScan/UI/Screens/SpotConfirmScreen.swift
ios/HouseScan/UI/Screens/UnsupportedScreen.swift
ios/HouseScan/UI/Screens/UploadingScreen.swift
ios/HouseScan/UI/Screens/WallWalkScreen.swift
ios/HouseScan/UI/Theme/Theme.swift
ios/HouseScanKit/Package.swift
ios/HouseScanKit/Sources/HouseScanKit/Capture/AutoCapture.swift
ios/HouseScanKit/Sources/HouseScanKit/Capture/CaptureSourceState.swift
ios/HouseScanKit/Sources/HouseScanKit/Capture/CloseUpCredit.swift
ios/HouseScanKit/Sources/HouseScanKit/Capture/CloseUpGate.swift
ios/HouseScanKit/Sources/HouseScanKit/Capture/FrameQuality.swift
ios/HouseScanKit/Sources/HouseScanKit/Confirm/SpotConfirmation.swift
ios/HouseScanKit/Sources/HouseScanKit/Coverage/CoverageMap.swift
ios/HouseScanKit/Sources/HouseScanKit/Depth/DepthImage.swift
ios/HouseScanKit/Sources/HouseScanKit/Geometry/Camera.swift
ios/HouseScanKit/Sources/HouseScanKit/Geometry/GroundPlaneChoice.swift
ios/HouseScanKit/Sources/HouseScanKit/Geometry/MeterAnchorTracking.swift
ios/HouseScanKit/Sources/HouseScanKit/Geometry/SpatialUpdate.swift
ios/HouseScanKit/Sources/HouseScanKit/Geometry/WallFrame.swift
ios/HouseScanKit/Sources/HouseScanKit/Guidance/GapPlanner.swift
ios/HouseScanKit/Sources/HouseScanKit/Guidance/GuidancePlanner.swift
ios/HouseScanKit/Sources/HouseScanKit/Guidance/WalkedEnd.swift
ios/HouseScanKit/Sources/HouseScanKit/HouseScanKit.swift
ios/HouseScanKit/Sources/HouseScanKit/Mesh/MeshProbe.swift
ios/HouseScanKit/Sources/HouseScanKit/Mesh/TriangleMesh.swift
ios/HouseScanKit/Sources/HouseScanKit/Meter/MeterBrand.swift
ios/HouseScanKit/Sources/HouseScanKit/Meter/MeterNumberRanking.swift
ios/HouseScanKit/Sources/HouseScanKit/Meter/MeterPhotoChecks.swift
ios/HouseScanKit/Sources/HouseScanKit/Packet/MeterFrame.swift
ios/HouseScanKit/Sources/HouseScanKit/Packet/PacketDepthFrames.swift
ios/HouseScanKit/Sources/HouseScanKit/Packet/PacketFiles.swift
ios/HouseScanKit/Sources/HouseScanKit/Packet/PacketManifest.swift
ios/HouseScanKit/Sources/HouseScanKit/Packet/PacketMarks.swift
ios/HouseScanKit/Sources/HouseScanKit/Packet/PacketPlane.swift
ios/HouseScanKit/Sources/HouseScanKit/Packet/PacketSharpness.swift
ios/HouseScanKit/Sources/HouseScanKit/Packet/PacketStreams.swift
ios/HouseScanKit/Sources/HouseScanKit/Packet/PacketWriter.swift
ios/HouseScanKit/Sources/HouseScanKit/Replay/ReplayPlanning.swift
ios/HouseScanKit/Sources/HouseScanKit/Replay/ReplaySession.swift
ios/HouseScanKit/Sources/HouseScanKit/Scene/AnswerFreshness.swift
ios/HouseScanKit/Sources/HouseScanKit/Scene/GroundPatches.swift
ios/HouseScanKit/Sources/HouseScanKit/Scene/PlacementResult.swift
ios/HouseScanKit/Sources/HouseScanKit/Scene/ResultReading.swift
ios/HouseScanKit/Sources/HouseScanKit/Scene/ScanFolderCleanup.swift
ios/HouseScanKit/Sources/HouseScanKit/Scene/ScanStamp.swift
ios/HouseScanKit/Sources/HouseScanKit/Scene/SceneExport.swift
ios/HouseScanKit/Sources/HouseScanKit/Scene/ServerErrorDefaults.swift
ios/HouseScanKit/Sources/HouseScanKit/Scene/ZipWriter.swift
ios/HouseScanKit/Sources/HouseScanKit/Upload/UploadFailureKind.swift
ios/HouseScanKit/Tests/HouseScanKitTests/AnchorRepairTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/AutoCaptureTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/CloseUpCoverageTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/CloseUpGateTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/CornerTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/CornerWalkedPathTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/CoverageDepthTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/CoverageMapTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/DepthTestSupport.swift
ios/HouseScanKit/Tests/HouseScanKitTests/FacingTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/FenceAcrossCornerTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/FrameQualityTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/GapPlannerTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/GeometryTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/GroundDepthTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/GroundPatchTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/GroundSettlementTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/GuidancePlannerTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/JSONSchemaValidator.swift
ios/HouseScanKit/Tests/HouseScanKitTests/JSONSchemaValidatorTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/KeptScansTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/MeshProbeTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/MeterBrandTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/MeterNumberRankingTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/MeterPhotoChecksTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/NoUnseenClaimsTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/OverheadTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/PacketFileTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/PacketFrameTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/PacketWriterTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/PlacementResultTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/RealReplayTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/ReplayPlanningTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/ReplaySessionTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/ResultReadingTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/SceneExportTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/Schemas/example-scene.json
ios/HouseScanKit/Tests/HouseScanKitTests/Schemas/manifest.schema.json
ios/HouseScanKit/Tests/HouseScanKitTests/Schemas/result.schema.json
ios/HouseScanKit/Tests/HouseScanKitTests/Schemas/scene.schema.json
ios/HouseScanKit/Tests/HouseScanKitTests/Schemas/server-answer-synthetic-wall.json
ios/HouseScanKit/Tests/HouseScanKitTests/ServerErrorDefaultsTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/ServerGapTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/ServerRequestSettlementTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/SpatialSourceTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/SpotConfirmationTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/TestSupport.swift
ios/HouseScanKit/Tests/HouseScanKitTests/TwoCornerObjectTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/UploadFailureKindTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/UploadSnapshotTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/UpperWallFixtureTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/VersionTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/WalkedEndTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/WalkingHeightTests.swift
ios/HouseScanKit/Tests/HouseScanKitTests/ZipWriterTests.swift
ios/HouseScanUITests/AccessibilityAudit.swift
ios/HouseScanUITests/Fixtures/results/no-clean-spot.json
ios/HouseScanUITests/Fixtures/results/pass.json
ios/HouseScanUITests/Fixtures/results/reject-nearest.json
ios/HouseScanUITests/Fixtures/results/unsure-view.json
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00001.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00001.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00002.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00002.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00003.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00003.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00004.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00004.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00005.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00005.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00006.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00006.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00007.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00007.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00008.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00008.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00009.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00009.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00010.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00010.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00011.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00011.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00012.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00012.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00013.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00013.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00014.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00014.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00015.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00015.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00016.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00016.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00017.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00017.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00018.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00018.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00019.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00019.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00020.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00020.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00021.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00021.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00022.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00022.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00023.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00023.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00024.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00024.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00025.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00025.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00026.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00026.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00027.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00027.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00028.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00028.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00029.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00029.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00030.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00030.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00031.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00031.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00032.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00032.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00033.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00033.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00034.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00034.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00035.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00035.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00036.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00036.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00037.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00037.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00038.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00038.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00039.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00039.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00040.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00040.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00041.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00041.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00042.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00042.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00043.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00043.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00044.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00044.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00045.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00045.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00046.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00046.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00047.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00047.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00048.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00048.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00049.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00049.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00050.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00050.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00051.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00051.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00052.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00052.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00053.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00053.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00054.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00054.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00055.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00055.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00056.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00056.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00057.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00057.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00058.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00058.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00059.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00059.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00060.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00060.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00061.confidence.u8
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/keyframes/k00061.depth.f32
ios/HouseScanUITests/Fixtures/synthetic-wall-lidar/session.json
ios/HouseScanUITests/Fixtures/synthetic-wall/session.json
ios/HouseScanUITests/FullFlowUITests.swift
ios/HouseScanUITests/ScreenStatesUITests.swift
ios/HouseScanUITests/SourceRestartUITests.swift
ios/README.md
ios/Tools/check-app-scene.sh
ios/Tools/make-app-icon.swift
ios/Tools/make-synthetic-replay.swift
ios/project.yml

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

The ground under the wall was looked up once, when the meter was tapped, and fell back to camera height minus 1.4 m for the rest of the scan. The lookup now re-runs whenever ARKit's horizontal planes change, replacing the guess (or an earlier plane) and republishing the wall. While the ground is still a guess, scene.json carries plus_minus_ft of 0.3 m on the meter and every object: the guess's error enters every height, and the schema has no other place to say the ground was estimated. A replay's recorded wall counts as measured; an assumed one does not. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

After 20 s of relocalizing the engine threw away the scan in every phase, including an upload in flight (its answer was then dropped) and the result. The spatial reset now runs only from findMeter through gapRequest. Uploading and the result keep the scan and the answer; the AR result already hides its overlay while tracking is not normal. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

"Capture it now" on a past_end item cleared that end and started a gap step, but markWallEnd only worked in the walk, so the end could never be marked again. During a past_end request, markWallEnd now accepts that side; answering the end question settles the request and goes on to the upload. If the request closes by coverage instead, the end stays cleared and exports as unexplored. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

- Sweep zones: start_ft is a range of battery left edges, so a zone runs to the last start plus the spot's width. - Keyframe writes in flight are counted per scan generation; resetAll no longer zeroes a counter that a finishing write then drove negative. - Entering the close-up clears a pending shot from an earlier visit. - A refined meter anchor or measured ground moves the coverage cells and marked ends with the meter (whole cells; the sub-cell remainder carries over) and re-projects every marked feature from its tapped world points. - An UNSURE check without unsure_cause needs a person. - Missing evidence is capturable only when a gap request can be built from it. - The unseen side ignores an unexplored end beyond cable reach. - A keyframe write that finishes after its world frame was discarded is dropped. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…calization, past-end, fence depth, guidance arrow Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

… one Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…time The planner's dwell ran on frame timestamps. On a replay at 3x speed that let the card change every second or so (walk left, tilt down, walk left), which the accessibility audit caught mid-transition and the verify lane's I6 check reads on the wall clock. Live, both clocks agree. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…ks inside and scalable CI's Simulator showed a system banner over the photo count during an audit, and once the accessibility tree changed while an audit read it. The screen-state audit now fails on issues found again six seconds later, and a failed snapshot is retried once in both UI tests. On the ADVIO walk the meter sits at the strip's end, and its mark hung half outside the strip, which the audit reported as clipped text. Marks now stay inside the strip, and their glyphs follow the text size up to 16 pt instead of fixed 9 and 10 pt. The walk's distance hint counts down in 5 ft steps above 5 ft, so the card no longer changes every stride. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

codemetrics-ai · 2026-09-26T11:57:59Z

Note

Review skipped

Codemetrics is installed on this repository, but it hasn't been added to your team yet.

To enable AI code reviews for this repo, go to your dashboard and add house-scanning to your team's repositories.

… picker Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…es to replace Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

… reader Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

updateWall only shifted cells along s, so after the ground height was corrected the export claimed wall and ground rows sampled in the air. The map now keeps the camera of every normally tracked keyframe and replays them against the new frame, keeping skipped cells and ends. refineGround republishes the strip after the rebuild. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…vives tap geometry POST {serverURL}/v1/placements now carries scene.json as application/json; the solver reads no photos, so keyframes stay on the phone and scan.zip is still written for replay and debugging but never uploaded or awaited. Network failures and 5xx become failed (Try again); 4xx, an unreadable answer and a scene that can't be exported become rejected with homeowner wording, the technical detail going to the log. backToReview returns a rejected upload to the feature review with the scan kept. The export clamps opening heights, moves ground taps behind the wall line to just in front of it, and drops a zero-length driveway edge instead of throwing. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…s as strings The app never reads them, so a new server value there should not fail every decode. Decision, outcome and unsure_cause stay strict. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

HouseScanServerURL = $(HOUSESCAN_SERVER_URL), set in Shared.xcconfig to the deployed server. -serverURL overrides it and -sampleResult still forces the bundled sample, which the full-flow UI test passes so it stays offline. An empty or non-http(s) value means no server and falls back to the sample. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

scene.schema.json now closes attrs to operable and well; the export only writes operable, and a test checks another key is rejected. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

A denied camera, a failed AR session and an unreadable replay set state.failure but left the flow where it was; only an unsupported phone reached .unsupported. All four now go through one fail(_:) that sets the failure and the phase. Checked with -replay /nonexistent: STATE=onboarding, then STATE=unsupported. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Every new KeyframeStore (at launch, and on Start over) deletes the other folders under Caches/Scans off the main actor. Before, each Start over left its previous Caches/Scans/<UUID> behind and nothing ever removed one. Checked on the Simulator: 30 folders before launch, 1 after. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Coverage marks what the camera pointed at. The server does not read images, so nothing downstream corrects a bush or bin counted as seen. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

A failed try needed one problem lasting 4 s with no good frame between, so frames alternating sharp and blurry never produced a photo or a failed try, and "Can't get a clear shot" never appeared. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…otos stay on the phone The second page said only "Your phone takes the photos". Uploads now carry the wall's measurements, so the page says what leaves the phone and that the photos are deleted on Start over, before the camera is asked for. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…e recording Both failures showed the engine's message, which is system error text (ARKit's localizedDescription, or String(describing:) of the replay loader's error). They now say what happened in the homeowner's words, and Start over, the only way on, is the screen's primary button. Open Settings uses the primary style too, so it gets the pressed state a Link lacked. The demo can open on either failure (-uiDemoFailure sessionFailed|replayUnreadable). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

# Conflicts: # ios/HouseScanUITests/ScreenStatesUITests.swift

chatgpt-codex-connector · 2026-09-27T11:05:37Z

Codex Review: Didn't find any major issues. You're on a roll.

Reviewed commit: 29ebe846a3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

The photo was chosen with its raw camera against the corrected wall, and the outline drawn through the raw pose: after a 0.30 m anchor correction the outline sat tens of pixels off the spot while the whole-area check passed. Each candidate is now judged with its camera corrected by the anchor corrections applied since it was taken (SpotPhotoCandidate.camera(correctedBy:)), and the choice carries that camera for the outline. The photo and camera are fixed together after the decode, and judged and drawn again whenever the wall moves while the question is up (refreshSpotPhoto, from publishWall); a photo that no longer shows the whole area turns the question into leaving the area out. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

The planner joined ground spans to the 125-entry band budget, but the export joins ground to 124 on a chain with one corner, to leave room for cutting it there. At the boundary the export merged two neighbours (2.5 ft and 3 ft) that the planner kept, and a 3 ft request settled on the phone while the scene sent 2.5 ft (caretaker review, GapPlanner P2). SceneExport.groundEntries is the one normalization; the export and the planner's settlement both read ground through it. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…ngine

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

SamGu-NRX · 2026-09-27T11:14:18Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 77d820bb69

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

SamGu-NRX · 2026-09-27T11:35:07Z

@greptileai

The LiDAR replay at 9b4c57d failed contrast on action.cannotAccess. Its recording shows "Can't get there" and "Can't see past it" drawn over each other half-transparent as the walk moved to the see-behind step: the reply stayed one button, so its new title cross-faded inside the card's instruction animation. It now takes a new identity per title and swaps at once, as the message above it does. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

A scan's first bundle write creates scan.zip.partial; if the app died before scan.zip existed, the next launch counted that folder as completed and newest, so it took one of the two slots and evicted a real completed scan (Greptile 4115191959). Now only scan.zip marks and dates a completed scan; a folder with both counts by its scan.zip. A partial-only folder is kept apart, uncounted: the newest one, while it is newer than every completed scan, so its photos last until the next scan completes. Older ones go. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

scene.json, the mesh measurement and the packet read the live wall, ground, marks and ARKit's current mesh at different moments, so a correction or ground change between reads made them describe different walls. - UploadSnapshot (ScanEngine+Snapshot.swift) is taken on the main actor in ingest, right after a frame's anchor correction and ground update. It holds the coverage map, ground, marks as world points, raw keyframes with the pose corrections, stills, end kinds, spot ground patches, and that frame's mesh and planes. - LiveCapture reads the mesh and planes from the delivered ARFrame's own anchors when asked, stamped with its timestamp. The snapshot keeps them only when that time is the frame's, and goes without them if no such frame arrives within 2 s. There is no more currentFrame read and no retry loop. - sceneJSON, packetInputs and the mesh measurement take the snapshot; nothing after it reads live state. - Pending taps are kept as world points and moved by corrections, so a mark's heights come from the snapshot's ground. - spatialRevision counts corrections, ground changes, new walls and followed corners. A stale answer is dropped and the scan sent once more; a second stale answer is not shown and the upload fails with "still adjusting, try again" (AnswerFreshness). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

680e4b7: only a real scan.zip makes a completed scan; a partial-only folder takes no slot. 4ef41b3: one upload snapshot from a frame-stamped input; the mesh only from the matching frame; stale answers resent once; mark heights from world points against the snapshot's ground. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Start over sits under the result card's collapsed Details (ResultScreen), as ScreenStatesUITests already reaches it; 1c496fe stopped at the result but still tapped a hidden button. The test also lets the spot question go on. Found by the #10 caretaker from the b33e214 run. Not rerun. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…ply title Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

SamGu-NRX · 2026-09-27T12:08:39Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4f1fb8ebec

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-09-27T12:20:43Z

+        // A fence's feet on two pieces of the wall would be sent as one depth that misses how
+        // close its line comes to the wall by the corner: refused, and asked for per side.
+        if marking.kind == .fence, let first = pendingTaps.first, !wall.onSamePiece(first, wall.world(hit)) {
+            marking.refusal = .fenceAcrossCorner


Reject openings whose taps cross a wall corner

When the homeowner places a door or window's two taps on different pieces of a followed corner, this same-piece validation runs only for fences, so the opening is accepted with one continuous span. SceneExport then assigns that span a single wall_id, and the result renderer likewise draws the entire opening on the midpoint's wall, misplacing or losing part of the marked obstacle for clearance checks. Apply the same corner validation to two-tap wall openings, or split them into per-wall features.

Useful? React with 👍 / 👎.

…ey measure A gas meter or AC unit was one tap, exported as a 0.3 m square standing 0.3 m off the wall: a guessed width and depth. - Gas meter: two taps where it meets the wall, bottom-left then top-right, as a window or #52's box. Exported as a wall object with its span, bottom_ft and top_ft and no footprint, since a wall-plane tap says nothing about how far it stands out. The server measures to its stretch of wall line. - AC unit: two taps on the ground where its front meets it (a ray aimed at a unit meets the ground in front of the wall, #163). Exported with its span and a footprint from the wall line to that tapped front edge, so its depth is measured and the space behind it counts as taken. Height is not measured and is left out. - Refused and marked again: taps closer than 1 cm (SceneExport.minObjectSize), a single tap, and AC corners on two pieces of the wall. - The packet manifest (1.1) gives these kinds one point, so each mark carries the middle of its two corners; the packet's scene.json has the span and footprint. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

A gas meter takes two taps on the wall (span and heights, no footprint); an AC unit two taps on the ground at its front (footprint to the wall line, no heights). The old 0.3 m square is gone, and a one-tap or cross-corner mark is refused, never exported as a sized object. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…lt and AR The server's answer was checked against the spatial revision only when it arrived. A ground plane revoked or refined while the spot check is open raised the revision, but finishing on a path that adds no patch ("It's clear" then "Not sure") still went straight to the result with the old answer. Every step that would present the accepted answer as current (the spot check, the result after it, opening and closing the AR view) now asks AnswerFreshness.step: an unchanged revision shows it; a changed one drops the answer and sends the scan again from a fresh snapshot, once, then shows the stale-answer failure, as at the POST. A step that changes the scan uploads anyway. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

An accepted answer is shown only while the spatial revision it was built on still holds; otherwise the scan is sent again once, then fails as stale. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

SamGu-NRX · 2026-09-27T13:47:11Z

@codex review

chatgpt-codex-connector · 2026-09-27T13:47:22Z

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.

…ft are rounding - make sim-app at c820cf1 on A4EE0B20, synthetic replay, hosted public server: 17 states, a server gap request answered "can't get there", three spot checks, the result and AR. S4-2 is met there and S4-1 partial (the real ADVIO replay is blocked by its data). - The scan's less-coverage resend flagged wall_equipment_above for a 0.0037 ft sliver past the seen wall. The server's missing() drops 1D gaps under 0.01 ft (COVERAGE_TOLERANCE_FT); the wall, route, facing and overhead checks now do too. Tested at 0.0037 and 0.011 ft. e2e at 3baa338 with that scan: 49 of 50, the ETH3D server issue. #11 moved to 6c7bea1 during the run, so S2-6 and S2-8 read stale until the turn-2 re-pin. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

SamGu-NRX · 2026-10-01T19:38:45Z

@greptileai review

Please review the current head c820cf150a73cb97527428d580c330f631253f9a against this PR's actual base. Catch-up review requested by Sam after enabling BaseScanning in Greptile.

greptile-apps · 2026-10-01T19:45:54Z

+                    heading
+                    featureList
+                    addSomething


Wall-wide ground question is missing. The review shows the feature list and “Looks complete” without asking what is on the ground along the wall. Ground type is asked only later for a server-selected spot, and its answer covers only that footprint. If the server returns no spot, or no kept photo shows the whole spot-check area, the homeowner cannot provide a ground type, leaving the ground-surface check unsure even when they know the answer.

Prompt To Fix With AI

This is a comment left during a code review. Path: ios/HouseScan/UI/Screens/MarkFeaturesScreen.swift Line: 54-56 Comment: **Wall-wide ground question is missing.** The review shows the feature list and “Looks complete” without asking what is on the ground along the wall. Ground type is asked only later for a server-selected spot, and its answer covers only that footprint. If the server returns no spot, or no kept photo shows the whole spot-check area, the homeowner cannot provide a ground type, leaving the ground-surface check unsure even when they know the answer. --- For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Caretaker check at c820cf1: not a defect in the code. The wall-wide ground question was removed on purpose after Codex 4114010727 showed one answer could label lawn as concrete (Sam, 4114020413 and 4114023377). With no spot, ScanEngine+Confirm.swift:94-100 shows the result without a ground question, and SpotConfirmation exports only answered footprints. The scene contract makes ground optional, and an unrecorded surface stays UNSURE, which is the intended outcome for unseen ground.

What is stale: the PR body and the header comment at MarkFeaturesScreen.swift:3 still describe a wall-wide question. That is description drift, recorded for #10's owner; #10 is frozen, so no change from here.

You're right — this is not a defect in MarkFeaturesScreen or the export flow. The wall-wide ground question was intentionally removed because a single answer could overgeneralize across the wall, and leaving unseen ground as UNSURE is the intended contract behavior. The stale parts are the PR description and the screen's header comment, so I’m recording this as description drift for the frozen PR rather than requesting a code change.

greptile-apps · 2026-10-01T19:45:54Z

+    types: [opened, synchronize, reopened, ready_for_review, labeled]
    paths:
      - "ios/**"
+      - "!ios/**/*.md"
      - ".github/workflows/ios.yml"


Server schema changes skip drift check. The new iOS contract comparison runs only when an iOS non-Markdown file or this workflow changes. A pull request that changes only a server schema will not trigger it, so an incompatible vendored iOS schema can go undetected by this check. Include the server contract paths in the workflow triggers.

Prompt To Fix With AI

This is a comment left during a code review. Path: .github/workflows/ios.yml Line: 13-17 Comment: **Server schema changes skip drift check.** The new iOS contract comparison runs only when an iOS non-Markdown file or this workflow changes. A pull request that changes only a server schema will not trigger it, so an incompatible vendored iOS schema can go undetected by this check. Include the server contract paths in the workflow triggers. --- For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Caretaker check at c820cf1: confirmed, and already true on main. A ready PR that changes only server/schemas/*.schema.json matches none of ios.yml:14-17, and server.yml runs only lint and pytest, so vendoredCopiesMatchServer never runs. Adding server/schemas/** and server/tests/fixtures/example-scene.json to both path lists closes the gap, but it also puts a 45-minute macOS job on every server-schema PR. That cost trade-off belongs to #10's owner, and #10 is frozen, so it is recorded for them rather than changed here. A cheaper alternative is a separate Linux job that only diffs the vendored copies.

That’s a fair resolution. The finding remains valid, but since #10 is frozen, I won’t recommend changing this workflow here. Recording the two concrete options for the owner is sufficient: add the server contract paths and accept the macOS cost, or add a cheaper standalone Linux vendored-schema drift job.

* Sync c3-vlm breadth work from the Project store (result file C3-vlm-baseline.md) * Sync c3-vlm breadth work from the Project store (result file C3-vlm-baseline.md)

github-actions Bot added size:XL 500-999 effective changed lines. size:XXL 1,000+ effective changed lines. and removed size:XL 500-999 effective changed lines. labels Sep 26, 2026

SamGu-NRX and others added 8 commits September 26, 2026 05:22

Merge fixes from the fresh review: coverage rows, ground height, relo…

fe0a016

…calization, past-end, fence depth, guidance arrow Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Runtime: drop the stand-in answerWallEnd now that every conformer has…

178add4

… one Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

SamGu-NRX changed the title ~~ios: guided wall scan, from meter to battery spot~~ ios: guided wall scan from the meter to a battery spot Sep 26, 2026

SamGu-NRX marked this pull request as ready for review September 26, 2026 11:57

SamGu-NRX and others added 14 commits September 26, 2026 07:07

contract: final upload failures, back to review, and the meter-number…

1cc297f

… picker Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Compile against contract 1cc297f with marked placeholders for the lan…

51dceb9

…es to replace Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Runtime: the interface between the close-up flow and the meter-number…

a39d0a5

… reader Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Schemas: re-vendor from origin/t3/server at 739fb6f

dc6d077

scene.schema.json now closes attrs to operable and well; the export only writes operable, and a test checks another key is rejected. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Coverage: say plainly that occlusion is not modelled

166d249

Coverage marks what the camera pointed at. The server does not read images, so nothing downstream corrects a bush or bin counted as seen. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Merge remote-tracking branch 'origin/t3/ios-mvf' into t3/ios-mvf-r9-ui

c2304e2

# Conflicts: # ios/HouseScanUITests/ScreenStatesUITests.swift

SamGu-NRX and others added 4 commits September 27, 2026 06:10

Merge remote-tracking branch 'origin/t3/ios-mvf' into t3/ios-mvf-r7-e…

842cdd0

…ngine

Merge the spot photo's corrected camera

77d820b

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

chatgpt-codex-connector Bot reviewed Sep 27, 2026

View reviewed changes

Comment thread ios/HouseScan/UI/Result/ResultScene3D.swift

greptile-apps Bot reviewed Sep 27, 2026

View reviewed changes

Comment thread ios/HouseScanKit/Sources/HouseScanKit/Scene/ScanFolderCleanup.swift Outdated

SamGu-NRX and others added 3 commits September 27, 2026 06:50

Merge the UI lane's CI fixes: the map legend, the picker wait, the re…

6be087d

…ply title Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

chatgpt-codex-connector Bot reviewed Sep 27, 2026

View reviewed changes

SamGu-NRX and others added 2 commits September 27, 2026 07:26

SamGu-NRX mentioned this pull request Sep 27, 2026

verification: independent end-to-end checks (S4) #13

Merged

2 tasks done

SamGu-NRX and others added 2 commits September 27, 2026 08:14

Merge answer freshness through the spot check, the result and AR

c820cf1

An accepted answer is shown only while the spatial revision it was built on still holds; otherwise the scan is sent again once, then fails as stale. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

greptile-apps Bot reviewed Oct 1, 2026

View reviewed changes

obvious-autobuild Bot mentioned this pull request Oct 9, 2026

[obvious-fleet] docs: bring README, AGENTS.md and CONTRIBUTING.md up to what main holds #219

Draft

SamGu-NRX pushed a commit that referenced this pull request Oct 15, 2026

breadth: C3-vlm-baseline results (#10)

a579ea8

* Sync c3-vlm breadth work from the Project store (result file C3-vlm-baseline.md) * Sync c3-vlm breadth work from the Project store (result file C3-vlm-baseline.md)

Conversation

SamGu-NRX commented Sep 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What changed

Why it works this way

UI changes

Verification

Not done, and what could go wrong

Checklist

Uh oh!

vercel Bot commented Sep 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

coderabbitai Bot commented Sep 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Uh oh!

codemetrics-ai Bot commented Sep 26, 2026

Uh oh!

chatgpt-codex-connector Bot commented Sep 27, 2026

Uh oh!

SamGu-NRX commented Sep 27, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

SamGu-NRX commented Sep 27, 2026

Uh oh!

Uh oh!

SamGu-NRX commented Sep 27, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot Sep 27, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

SamGu-NRX commented Sep 27, 2026

Uh oh!

chatgpt-codex-connector Bot commented Sep 27, 2026

Uh oh!

SamGu-NRX commented Oct 1, 2026

Uh oh!

greptile-apps Bot Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

SamGu-NRX Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

SamGu-NRX Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

SamGu-NRX commented Sep 26, 2026 •

edited

Loading

vercel Bot commented Sep 26, 2026 •

edited

Loading

coderabbitai Bot commented Sep 26, 2026 •

edited

Loading