feat: web sources frame-captured via composition (CoreWebView2CompositionController → Windows.Graphics.Capture); PNG poll + CaptureScheduler deleted

The ~30Hz CapturePreviewAsync PNG poll capped real cadence at ~20Hz
(35–165ms full-HD encode+decode), so a 60fps widget still juddered at ~1/6
speed. Replaced polling with frame-driven capture of the composition
controller's root visual — the mechanism WebView2CompositionControl and
Flutter's webview_windows use (graphics_context.cc captures the root
surface_ visual via CreateGraphicsCaptureItemFromVisual; reference:
github.com/microsoft/Windows.UI.Composition.WinUI / flutter-internal
webview_windows). Frames now arrive at the renderer's own pace; capture
memory is epoch'd ring reuse + one crop-sized shared WriteableBitmap.

New Services/WebCaptureFrameSource.cs owns GraphicsCaptureItem + free-
threaded Direct3D11CaptureFramePool + session (Straight alpha readback,
per-frame FindContentBounds → CropBounds). WebView2Manager reworked around
per-session composition controllers + one UI-thread Compositor created via
the CoreMessaging CreateDispatcherQueueController P/Invoke (the 19041
projection lacks CreateOnCurrentThread); internal seam ctor
(Dispatcher, Func<string,IScreenCaptureSource>?) for hermetic tests.
CaptureScheduler.cs deleted; the three SetCaptureInterval cadence hooks
removed; InitWebView2() moved from MainWindow ctor to Loaded (a parent
HWND must exist for the composition controller); the hidden WebViewHostPanel
overlay deleted. TransparentBackgroundScript unchanged.

Tests: WebView2ManagerTests reworked — 4 control-size + scheduler tests
dropped, FindContentBounds tests moved to WebCaptureFrameSource, ONE
integration test (Frames_PublishCroppedPreview_And_CoalesceToLatest_CarryingCropBounds)
drives the seam with a FakeWebSource + background-STA DispatcherPump.
Suite 293/293, 0 warnings.

NOTE: composition path NOT yet verified on a device — the take is the
next step. Web work committed locally only (no push per standing rule).

Docs same-commit: ai.md Slice 14 + supersede marker on Slice 11, HANDOFF,
MyMistakes (CoreMessaging DQ + namespace-landmine recipe), TASK 17,
Controls/ViewModels/Services indexes.

HANDOFF.md

@@ -1,14 +1,10 @@
-# HANDOFF — 2026-09-14 (signed audio sync SHIPPED + PUSHED: −500..+500, negative advances by eating the stream head; slider locked while live/recording)
+# HANDOFF — 2026-09-14 (web capture REDESIGNED: composition-capture, PNG polling + CaptureScheduler deleted — NOT yet verified on device, NOT pushed)
 
 ## Branch / Commit State
 
-`main` HEAD = **`b22d08e`** (signed audio-sync, committed AND **pushed** to origin/main with the
-user's greenlight — 29 commits pushed in one go). Working tree **clean**.
-
-DB now: **`Audio.SyncOffsetMs = 0`** (confirmed via sqlite3 — the creator slid the sync to 0
-and the write finally stuck; sound is correct at 0). The 0.54s residual in the 1128 take ≈ 300ms
-injected offset (leftover DB value) + ~240ms natural (partly webcam clap-quantization ±40–85ms,
-partly measurement).
+`main` HEAD = **`b22d08e`** (signed audio-sync, already pushed). Working tree **DIRTY** with the
+web-frame-capture redesign (below). Commits for it land **LOCALLY only** — the web work was never
+pushed; the user greenlights pushes at checkpoints.
 
 ## ⚠️ Branding (2026-09-14, creator-corrected): product = **llamacasty**, internals = ytLive
 
@@ -17,72 +13,84 @@ The product is **llamacasty**; the repo path, csproj `AssemblyName`/`RootNamespa
 language must say "llamacasty"; code/assembly/repo names stay ytLive. Full detail in
 `ai.md` → Brand → "Product name vs repo/assembly branding".
 
-## ✅ SHIPPED (this dirty tree) — signed audio sync, −500..+500
+## 💥 The web-capture take history (why this redesign exists)
 
-**What:** the audio-sync control is now a SIGNED offset. Positive = delay the mix (audio runs
-AHEAD of video — existing `AudioSyncDelay` behavior, unchanged and live-reactive). Negative =
-**advance** the audio (audio runs BEHIND video): OBS's "eat the head of the buffer" fix —
-the mixer drops the first |N| ms of the written stream at the pipe, re-anchoring the audio
-stream so every event lands |N| ms EARLIER relative to video.
+The recording was 60fps but web widgets ran at ~1/6 speed: capture was a `CapturePreviewAsync` PNG
+poll; every full-HD encode+decode cost **35–165ms** so the session budget capped real cadence at
+~20Hz — a 60fps widget still juddered. Slice 11's `CaptureScheduler` (30Hz) was the wrong lever:
+the fix is frame-driven, not faster polling.
 
-**Mechanism (AudioMixer):** `StartLive` arms `_advanceSamplesRemaining = |N| ms → samples` at
-go-live (a negative offset can only eat the HEAD of the stream; it is armed once, not
-live-reactive). `LiveLoopAsync` skips `min(budget, mixBuffer.Length)` samples off each write
-head while the budget lasts — the pipe writer accepts a partial chunk via `AsMemory(writeFrom)`.
-Positive path untouched (delay line still re-reads the Func every tick).
+## 🔬 IN FLIGHT (this dirty tree, NOT committed, NOT pushed) — composition capture
 
-**UI/plumbing:**
-- `MainViewModel.Audio.cs` — clamp `Math.Clamp(value, -500, 500)`, doc updated.
-- `LayoutStore.Settings.cs` — `LoadAudioSyncOffsetMs`/`SaveAudioSyncOffsetMs` clamp −500..500.
-- `PreviewPane.xaml` — label **SYNC → "AUDIO SYNC"**, `Minimum="-500"`, tooltip explains both
-  directions (calibrate with a clap: clap late → negative; early → positive), and
-  **`IsEnabled="{Binding IsEditMode}"`** — the slider locks during live AND recording (gun
-  safety, same property `IsRecording`/`StreamStatus` already raise PropertyChanged for).
-- `AudioSyncDelay` unchanged (still clamps negative→0 internally; header doc updated to point
-  at the mixer for the advance side).
+**What:** web sources now render through `CoreWebView2CompositionController` into
+`Windows.Graphics.Capture` (the Flutter `webview_windows` / `WebView2CompositionControl`
+mechanism) — frame-driven at the renderer's pace instead of polling PNGs.
 
-**Regression test (the ONE integration test for this change):**
-`StartLive_NegativeOffset_AdvancesAudio_ByDroppingTheStreamHead` in `AudioPipelineTests.cs` —
-−40 ms advance (8-tick budget at 5ms interval), emits 6×0.9 fresh right after StartLive (≤
-budget, so the head MUST be eaten) then a long 0.2 bed; asserts the wire max ≈ 0.2 (< 0.3).
-Fails loudly if −N no longer drops the head (0.9 leaks).
+**Pipeline (per session):** `CreateCoreWebView2CompositionControllerAsync(appHwnd)` →
+`RootVisualTarget` = a `RelativeSizeAdjustment=1,1` child under a 1920×1080 root `ContainerVisual`
+on one `Compositor` (CoreMessaging DQ P/Invoke recipe — see ai.md Slice 14 / MyMistakes) →
+`GraphicsCaptureItem.CreateFromVisual(root)` → free-threaded `Direct3D11CaptureFramePool` (2
+buffers) → `SoftwareBitmap.CreateCopyFromSurfaceAsync` (`BitmapAlphaMode.Straight`) →
+per-frame `FindContentBounds` bbox → `CropBounds` → epoch'd ring → dispatcher-coalesced crop copy
+into ONE shared `WriteableBitmap`.
 
-**User directives (this session):**
-- Signed −500..+500 with positive=delay / negative=advance, relabel "AUDIO SYNC", default 0.
-- **Lock the sync control when live or recording** (done — `IsEditMode`).
-- **NOTE ONLY, no fix:** the live recording is completely different from the "recording
-  results" shown in Chat view (recorded in `bugs.md` — do not rediscover as a surprise).
-- **Before 1.0:** write a detailed USER-DOC tutorial on the audio-sync feature (see TASKS.md
-  note; add it to the gold-pass/1.0 checklist). `docs/` currently holds only the README image.
+**Files:** NEW `Services/WebCaptureFrameSource.cs`; reworked `Services/WebView2Manager.cs` (internal
+seam ctor `(Dispatcher, Func<string, IScreenCaptureSource>?)` for hermetic tests); DELETED
+`Services/CaptureScheduler.cs`; `MainViewModel.Web.cs` (`InitWebView2()` no-arg),
+`MainViewModel.Streaming.Operations.cs` (three `SetCaptureInterval` hooks gone — verified zero
+remaining), `MainWindow.xaml(.cs)` (`InitWebView2()` moved ctor→**Loaded**; `WebViewHostPanel`
+overlay deleted). `TransparentBackgroundScript` unchanged.
 
-## Take verification so far (ring-backlog fix)
+**Tests:** reworked `ytLive.Tests/WebView2ManagerTests.cs` — dropped the 4 control-size tests +
+`CaptureScheduler_Drops…`; `FindContentBounds` tests moved to `WebCaptureFrameSource`; the ONE
+integration test (`Frames_PublishCroppedPreview_And_CoalesceToLatest_CarryingCropBounds`) drives the
+seam with `FakeWebSource` + background-STA `DispatcherPump`: crop-sized preview published once →
+`CropBounds` + `IsOpaque=false` → back-to-back frames coalesce to latest. `RoundClipInteractionTests`
+comment updated (WebViewHostPanel gone). Suite **293/293 green**, app build **0 warnings**. Docs
+slant (ai.md Slice 14, HANDOFF, MyMistakes, TASK 17, Controls/ViewModels/Services indexes) updated in
+the same working set.
 
-Take `ty-20260914-1128-0000-2.mp4` (1920×1080@60fps, 356 frames / 5.93s): audio clap RMS peak
-3.260s; video diff-frame 163 → 2.717s; **offset ≈ +0.54s** — collapsed from +2.11/2.22s; the
-remaining ~300ms was the baked-in `SyncOffsetMs=300` (now 0). Both streams `start_time=0` — not
-an `avoid_negative_ts` artifact.
+## ⚠️ Open items on this change (before it is PUSHABLE)
+
+- **Device verification:** the composition path has never run against a real widget. Verify ~60fps
+  web animation in a take (also: the OBS de-throttle flags stay; hidden-page JS throttling is fixed
+  by them, capture pacing is now renderer-driven).
+- **No push yet** — web work is commit-local until the user says push.
+
+## ✅ SHIPPED + PUSHED — signed audio sync, −500..+500 (`b22d08e`)
+
+Positive = delay the mix (audio AHEAD — the `AudioSyncDelay` line, live-reactive). Negative =
+**advance** (audio BEHIND): OBS-style "eat the stream head" — `AudioMixer.StartLive` arms
+`_advanceSamplesRemaining = |N| ms`, `LiveLoopAsync` skips `min(budget, mixBuffer.Length)` off each
+write head while it lasts. Slider relabeled **AUDIO SYNC**, `−500..500`, locked while
+live/recording (`IsEditMode`). One regression test:
+`StartLive_NegativeOffset_AdvancesAudio_ByDroppingTheStreamHead`.
+
+Take `ty-20260914-1128-0000-2.mp4`: **offset ≈ +0.54s** (collapsed from +2.11/2.22s; residual was
+the baked-in `SyncOffsetMs=300`, now 0 — confirmed via sqlite3). Both streams `start_time=0`.
 
 ## Open threads
 
-- **Verify signed sync on device:** re-run the clap take with a NEGATIVE offset to confirm the
-  advance direction end-to-end (the regression test proves the mixer; a take proves the file).
-- **Audio-silence verification** — fixed code (`724af14`) confirmed; creator heard real audio.
+- **Verify signed sync negative direction on device** (clap take with a negative offset).
+- **Audio-silence verification** — fixed (`724af14`); creator heard real audio.
 - **Webcam MJPG missing / ~10–14Hz**, layer SortOrder, truncation-with-dynamic-scenes — queued.
-- **Sync control tutorial in user docs — REQUIRED before 1.0** (creator directive).
+- **Sync control user-doc tutorial — REQUIRED before 1.0** (creator directive; TASK 22 file).
+- **Device verify the composition web capture (this change).**
 
 ## Landmines
 
 - testhost shares startup.log — filter by time.
 - `cmd.exe /c "taskkill /F /IM ytLive.exe"` (WSL double-slashes mangle) before rebuilds.
-- Build/tests: **Windows dotnet host** (`/mnt/c/Program Files/dotnet/dotnet.exe`). 0 warnings.
+- Build/tests: **Windows dotnet host** (`/mnt/c/Program Files/dotnet/dotnet.exe`). 0 warnings —
+  only `./scripts/verify.sh "<files>"`'s clean build counts.
 - ffmpeg/ffprobe: `/mnt/c/Program Files/Krita (x64)/bin/` with Windows paths.
-- `MyMistakes.md` has the **audio/video sync measurement recipe** (claps + cross-correlation)
-  — grep it before re-deriving.
+- `MyMistakes.md` has the **audio/video sync measurement recipe** AND the **composition-capture
+  CoreMessaging DQ recipe (2026-09-14)** — grep before re-deriving.
 - sqlite3 lives at `/home/gramps/android-sdk/platform-tools/sqlite3` (WSL) for the DB at
   `/mnt/c/Users/gramp/AppData/Roaming/ytLlive/ytLlive.db`.
 
 ## Next step
 
-Open queue question: user-doc tutorial for the audio-sync feature (creator directive, REQUIRED
-before 1.0 — recorded in `TASKS/task-22-audio-sync-offset.md`). Also asked 2026-09-14 whether
-other root md files need the TASKS-style split — answered: no (see below for the criterion).
+From this dirty state: run `./scripts/verify.sh "<all changed files>"`, review the diff, commit
+locally (web work — NO push, per standing directive). Then a device take to verify ~60fps web
+animation, then present for push review.