Move the world overlay into scene.services.overlay. Each attachment owns its
sources, collision domains, paint lanes, canvases, observers, timers and
accessible action subscribers. Keep both canvases inside the engine container,
with detection beside the card root so its screen blend still reaches the
scene. Release shared font measurements only after the last host detaches.
Resolve retained layer publications against their attached engine, clear the
outgoing source before releasing its renderer, and reconnect the shell's action
subscription on attachment. Keep destroyed hosts inert and give parked engines
fresh services. Bind annotation rectangles and the existing tracked bridge to
their engine's overlay. The tracked bridge itself remains page-owned.
Preserve Cesium's projection, card placement, arbitration, fades, hit tests,
blend order and render demand. Navara keeps the same shared behaviour. No
user-visible change is intended. Exercise simultaneous hosts, either teardown,
stale handles, reattachment and a retained selected card on the fake and
headless Cesium engines. Update QA reads and package ownership for the service.
No coupling metric falls or grows: renderer lines remain 0 Cesium and 189
Navara; external Cesium import files 3; viewer lines 891; Navara kind checks 1;
engine sniff calls 0; extension reads 3; ground passes 1; SDK loads 3;
ScreenSpaceEventHandler files 4; engine-name lines 32.
Each host now takes the page's overlay ids from the host it replaces and gives
them back to a remaining host when destroyed, and the chrome's static overlay
placeholders are gone. The new hosts had left the ids on those empty
placeholders, so the voice pointer policy, the id selectors and the QA probes
that read #world-overlay-canvas missed the drawn canvas.
Checked in the browser, keyless: the overlay baselines on both engines, CCTV
v2 (58 checks), military names, and the engine, voice, ground-layer and Ghost
Maps switches pass. The transit script fails the same 25 checks as before.
Preserve the camera, map source and layer selection across engine sessions. Restore the previous engine if startup fails and retain local history navigation.
The CCTV layer kept every loaded camera live: each view change touched every
icon, every notification rebuilt every camera's state, and the picker listed
the whole catalog. With the default 4,000 cameras, a continental drag ran at
152-179 fps with 18-24 frames over 30 ms (worst 112 ms). It now runs at
235 fps with none. The layer now:
- shows icons only on or near the screen, and when zoomed out keeps one camera
per world-anchored cell (healthy feeds first), so panning keeps the same
cameras;
- grounds and draws coverage only for the cameras in view, the active camera
and its coverage neighbours, retrying a camera at most twice per view regime;
- reruns the view pass while the camera moves, paced by its own cost, and
projects with one matrix per pass instead of a Cesium call per camera;
- keeps the billboard buffer-usage hint fixed, so showing or hiding a few
icons no longer rewrites the vertex data of every billboard;
- lists the cameras in view (active camera first) in the picker, closing with
the count of cameras outside the view, and builds the full camera list in
the UI state only when it is read.
Keep shared CCTV logic independent of Cesium, release the original renderer's resources, and guard retired initialization/media work. Preserve native marker, plane, volume, calibration and input behavior.
Record keyless dev/preview validation and retain unresolved shell idle/terrain and photoreal acceptance gaps. Routine QA builds use empty Google and Cesium credentials and block paid maps before navigation.
Wind (NOAA GFS and ECMWF IFS: animated particles, colour field, a reading at map centre), observed weather (NOAA MRMS radar, GOES and NESDIS infrared clouds, lightning density) on one shared history timeline, NHC and CPHC cyclone advisories with tracks and cones, and a WEATHER panel in the right rail. On 3D Tiles the observed products render as raised shells with a sharper window around the view.
Co-authored-by: Gustavo Beneduzi <[email protected]>
Co-authored-by: Daniel Slay <[email protected]>
Co-authored-by: Rehaan Delmotra <[email protected]>
Co-authored-by: Cursor Agent <[email protected]>