Repository navigation
Conversation
The gate compares each frame with the median of its last 15, so steady blur stops reading as blurry within a second, and with blur shown only after 2 s its coaching never came up. The debouncer now also counts a frame as blurry when it is under the gate's ratio of the median sharpness from the 1.5 s before the spell began, held fixed while the spell lasts. A short drop, such as a tilt, still says nothing, and steady blur shows from 2 s in until the photos are sharp again. A package test runs the real gate into the debouncer. The dark walk card is also checked at AX5. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
|
Important Draft PR not reviewedDraft PRs are not automatically reviewed by default.
To automatically review draft PRs, update your CodeRabbit configuration: reviews:
auto_review:
drafts: true
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. Comment |
|
@codex review |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Codex Review: Something went wrong. Try again later by commenting “@codex review”. ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
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". |
|
@codex review |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7094fbf426
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".
…indow Codex on #317: frames with no sharpness measurement append nothing and never pruned the buffer, so after a gap the median came from sharpness older than blurReferenceWindow. A scene merely softer than the one before the gap then read as blur and was coached. Prune by time before taking the median. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: df44d5e089
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".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Fixed in 8f32ed5. Supplemental sharpness blur evidence now runs only when the gate reached sharpness, and rejected-frame sharpness values stay out of the reference buffer. Regression: tooDarkFramesDoNotCreateBlur. Tests: heavy swift test --filter CoachingDebouncerTests (17 passed); heavy swift test (973 passed, 1 known issue). |
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8f32ed5139
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".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
@codex review |
|
Codex Review: Didn't find any major issues. Keep it up! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
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". |
Ports a097c66 from #104 to alpha
Finding: CoachingDebouncer's rolling 15-frame median catches up with steady blur within about one second. Since blur coaching appears after two seconds, steady blur never produced coaching. This port holds a pre-spell 1.5-second sharpness median as a fixed reference and counts frames below the gate ratio against it. A package test covers the real AutoCapture gate, sustained blur, and recovery.
Conflict resolutions: cherry-pick applied cleanly on alpha, so no conflicts required resolution. The resulting ScanEngine.swift changes pass frame sharpness into the debouncer while preserving alpha's surrounding code. ScreenStatesUITests.swift retains alpha's existing state list and adds the reference card assertion.
Failing-first evidence: on unchanged alpha,
~/.long-run/bin/heavy swift test --filter Coachingfailed to compile because the new test passedsharpnessto the old debouncer API (extra argument 'sharpness').Counts using
~/.long-run/bin/heavy swift test --filter Coaching: alpha baseline 16 tests in 2 suites passed; changed branch 18 tests in 2 suites passed.Counts using
~/.long-run/bin/heavy swift test: alpha baseline 969 tests in 130 suites passed with 1 known issue; changed branch 971 tests in 130 suites passed with 1 known issue.No Xcode app build or simulator run was performed. The app-target change was not independently compiled by the package tests.
🤖 Generated with Claude Code
Repeated blur recovery and app build verification
adcd0918 excludes frames judged blurry from the rolling reference buffer. Sharp recovery frames remain eligible during the coaching's clearing delay, so a brief recovery can establish the next spell's reference without inheriting the previous spell's low sharpness.
eachBlurSpellUsesSharpnessBeforeThatSpellfails if a second sustained blur after a short sharp gap receives no coaching. It uses the existing Frame/run helpers with 2 seconds sharp, 3 seconds blurry, a 0.6-second sharp gap, then 3 seconds blurry. On8f32ed51, the first blur was coached but the final assertion failed because the second blur returned nil. The same test passes with the fix.Checks ran with
NODE_ENVunset and the heavy wrapper. Capacity checks before each test run and the app build showed at least 6 GB free and memory pressure 1.ios/HouseScanKit,~/.long-run/bin/heavy swift test --jobs 2 --filter CoachingDebouncerTests: before the fix, 18 tests with the expected regression failure; after the fix, all 18 passed.ios/HouseScanKit,~/.long-run/bin/heavy swift test --jobs 2: 974 tests in 130 suites passed with 1 known issue.ios,~/.long-run/bin/heavy xcodebuild -listfound the app schemeHouseScan.ios,~/.long-run/bin/heavy xcodebuild -scheme HouseScan -destination 'generic/platform=iOS Simulator' -derivedDataPath /tmp/bs-blur-dd build CODE_SIGNING_ALLOWED=NO: BUILD SUCCEEDED. The app-target changes compile; no compile fixes were needed. This build does not compile or run the UI-test target, and no simulator was booted.Derived data and
HouseScanKit/.buildwere deleted after the build. The Codex repeated-blur review thread was answered and resolved.