Merge main into codex/fix-cctv-session-admission

Main's Street View budget tests in the media feed-type file now mount the
provider through the shared CCTV test harness, which also resets the Google
rate-limit setting.
This commit is contained in:
German Khokhlov committed 2026-11-20 17:39:46 +04:00
commit 23efd3b343
196 files changed
+21064 -562

No files matched your search

+138
View File
@@ -0,0 +1,138 @@
# GEV cameras in Blender
Export the cameras of one saved Director scene as a camera-only USD stage
(`.usda`, plain text) and open it with Blender's built-in importer. The export
needs no API key, network, browser or Blender add-on and contains no imagery or
geometry, so use independently licensed assets for your set.
## Export a saved scene
Save a Director project as JSON, then run from the repository root:
```sh
mkdir -p output
node tools/export-blender-camera.mjs --input scene.json --output output/cameras.usda --scene my-scene-id --origin 39.7,-105,1600 --fps 30000/1001 --width 1920 --height 1080 --horizontal-fov-deg 70
```
Paths are relative to the current directory; `output/` is ignored by git.
Omit `--scene` only when the document contains exactly one scene. `--output`
is required and must be a new file ending in `.usda`; the tool refuses to
overwrite it. No environment files are read.
`--origin` is latitude, longitude and **WGS84 ellipsoid height in meters**.
Positions and sampled paths must remain within 100 km of that origin. Heights
are never implicitly converted from mean sea level or above ground level.
Positions must be complete; legacy documents without saved IDs or positions
need to be saved again in GEV before export.
Projection and fps are required because Director documents do not record a
viewport lens, aspect ratio or video frame rate. `--horizontal-fov-deg` is the
horizontal perspective FOV in degrees, even for portrait output. If deriving it
from a Cesium `PerspectiveFrustum`, use
`2 * atan(tan(fovy / 2) * aspectRatio) * 180 / π` (`fovy` is in radians).
Do not assume `frustum.fov` is horizontal: it is the wider-dimension FOV.
The optional `--sensor-width-mm` defaults to 36; focal length is
`sensorWidthMm / (2 * tan(horizontalFovRadians / 2))`.
The exporter supports integer fps or an integer ratio such as `30000/1001`,
with an effective rate from 1 to 240. Image dimensions are integers from 4 to
16384, pixels are square, horizontal FOV is 1 to 179 degrees and sensor width
is 1 to 100 mm. A lens shorter than Blender's 1 mm minimum is rejected.
Input must be a regular file within the existing 5 MiB scene-document limit;
output is limited to 100,000 frames and 64 MiB. Every shot needs at least one
sampled frame.
## Timing and camera semantics
Timing is the Director's **runtime** seek timing. The exporter loads the
document as Director does, so ordinary shots fly for at least 0.2 s, and takes
each shot's length and hold from `shotRuntimeTiming` in
`src/scenes/shotPresentation.js` with the default scene packs. Registered hold
overrides, minimum holds and evidence-sequence durations therefore move later
shots as they do when seeking in the app. Live playback can run longer when it
waits for a layer's media. Saved shots are never rewritten; each camera
records both the authored and the runtime hold.
Each frame's camera is the Director's `sceneSeekState` pose at that time.
Named anchors and explicit `move.from` endpoints resolve through the existing
camera owner, and explicit linear/cubic moves use its sampler. Ordinary shots
interpolate cubically, by the shortest angle, from the previous shot's camera;
a shot that opens the scene starts at its own camera. A cut can be authored as
an explicit move with identical start/end poses.
The bridge does not reproduce Cesium's implicit live `flyTo` trajectory,
provider readiness, layer-owned camera motion, lens changes, visual effects,
media or geometry. Export authored camera moves when repeatable playback
framing is required. This feature does not depend on GEV video export.
Frame 1 samples time zero. Frame `n` samples `(n - 1) * denominator / numerator`
seconds. The final sample precedes the exclusive scene end; frame count is the
ceiling of duration times fps, with rounding noise at integer boundaries removed.
Shot ranges are half open: `frameStart <= frame < frameEndExclusive`. A shot
boundary switches at the first sampled time at or after the boundary.
## What the USD stage contains
- Stage metadata `upAxis = "Z"` and `metersPerUnit = 1`. Scene X/Y/Z are
east/north/up in meters around `--origin`, in the geodetic ENU frame (axes
from the WGS84 ellipsoid normal at the origin's latitude and longitude), so
Blender applies no axis or unit conversion.
- USD time codes equal frame numbers: `startTimeCode = 1`, `endTimeCode` is the
last frame, and `timeCodesPerSecond`/`framesPerSecond` hold the effective
rate. The exact fps ratio is kept as text.
- A root `GEV` transform at identity, with one perspective `Camera` per shot
named `Shot_001_<id>`, `Shot_002_<id>` and so on. Each camera has an
`xformOp:transform` matrix time sample on every frame of its own shot range,
and holds its first or last pose outside that range. Camera local X points
right, local Y up and local -Z forward, which is both the USD and the Blender
camera convention.
- Lens attributes in USD's units: `focalLength`, `horizontalAperture` and
`verticalAperture` are in **tenths of a scene unit**, so with
`metersPerUnit = 1` one unit of these attributes is 100 mm. A 25 mm lens on
a 36 mm sensor is written as `focalLength = 0.25` and
`horizontalAperture = 0.36`. The vertical aperture is the sensor width times
`height / width`, so it matches the output aspect.
- Clipping range 0.01 m to 200 km; adjust it for your set if needed.
- Source scene and shot IDs, titles, pack identifiers, authored and runtime
timing, frame ranges, origin, fps ratio, image size and data-pack
attribution text, as `userProperties:` attributes. No referenced assets are
fetched.
## Import into Blender
1. In Blender, choose **File > Import > Universal Scene Description**, select
the `.usda` file and keep **Set Frame Range** enabled.
2. Set **Output Properties > Format** to the exported `--width`, `--height` and
frame rate, with **Aspect X:Y** at 1:1. USD carries the time codes but not
render resolution, and Blender does not take its scene frame rate from the
file.
3. Pick a shot camera as the active camera, or bind cameras to timeline markers
at each shot's `gevFrameStart` (shown under **Object Data Properties >
Custom Properties**).
How Blender maps the lens: its USD importer multiplies `focalLength` and the
apertures by `100 * metersPerUnit` to get millimeters. With this exporter's
`metersPerUnit = 1` that is ×100, so the imported **Focal Length** and
**Sensor Width** equal the `lensMm` and `--sensor-width-mm` values. Blender
chooses horizontal sensor fit when the horizontal aperture is at least as large
as the vertical one and vertical fit otherwise, so portrait exports import with
vertical fit and the same field of view. This mapping is in Blender's
`usd_reader_camera.cc`.
Blender imports animated USD transforms as a **Transform Cache** constraint
that keeps the absolute path of the imported `.usda`. Put the file in its
permanent place before importing, or update the path in the constraint after
moving it. For a self-contained file, select the cameras and use **Object >
Animation > Bake Action…** with **Visual Keying** and **Clear Constraints**.
Blender stores object transforms in single precision: up to about 4 mm of
rounding at 100 km, less closer to the origin.
## Checks
```sh
node --test src/tooling/blender*.test.mjs
```
The tests compare the export with an independent Cesium camera and the
Director's runtime timing, and parse the stage with OpenUSD's `usdcat` when it
is installed.
+276 -17
View File
@@ -1577,6 +1577,10 @@ seek and keeps the last settled position. Undeclared operator layers remain
untouched, and the shot is not rebuilt.
See [camera directions](DIRECTOR-CAMERA.md).
An opt-in offline [Blender camera bridge](BLENDER-CAMERAS.md) exports a saved
Director scene as a camera-only USD stage (`.usda`) with Director seek
interpolation and runtime shot timing; format, units and limits are in that guide.
Realtime voice composes separate connection, response/tool, Radio handoff, input
and audio-meter, cost, viewport-context and diagnostic owners. The existing
controller exports and session/backend contracts remain available. Tool protocol,
@@ -2025,7 +2029,13 @@ and follows redirects only on the camera's origin. Catalog image URLs, and
redirects on those hosts, are pinned by parsed URL to the agency's image
directory. Every `/api/cctv` route refuses cross-site browser requests but
accepts forwarding headers, so cameras work behind a reverse proxy; a forwarded
request skips the key-bearing Street View fallback and gets the placeholder.
request skips the key-bearing Street View fallback and gets the placeholder. The
fallback answers only for cameras the server lists, at their catalog position,
so position calibration never reaches it. Heading, field of view (clamped to
20–120°) and pitch (clamped to −40–20°) come from the request, else from the
catalog pose. A bundled seed camera, shown when the server catalog is empty,
gets the placeholder. The fallback counts in its own per-IP Google window (see
the cost-bearing proxies below).
The CCTV layer draws only the cameras on or near the screen. Zoomed out, a
ground cell with two or more cameras shows one marker at their mean position. A
@@ -2055,6 +2065,19 @@ in view finishes loading. The camera picker lists the active camera first, then
one camera per visible cell, preferring healthy feeds. It ends with the count of
catalog cameras missing from the list.
Turning CCTV on opens on the camera nearest the view. Until someone picks a
camera, each settled move, such as a shared view's flight landing, moves it to
the camera then nearest the view. Automatic selection ends at the first pick:
NEAREST, PREV or NEXT, FOCUS, a camera chosen in the picker, a click on a
camera, a click on empty ground that deselects, an auto-hop step, or a call to
`selectCamera`, `focusNearest`, `focusNearestToPoint` or
`setParams({ selectedCameraId })`, which voice and scenes use. Picking the
camera that is already active counts. The picker shows the automatic camera
as its own option, and a native list sends no change for the option it
already shows, so the list cannot pick that camera: FOCUS does, and so does
NEAREST when that camera is the nearest. Closing the list without a choice
keeps automatic selection.
Traffic owns its rendered-surface observer only while enabled, removes it on disable/destroy, and rebuilds no cached heights on frames where the surface identity is unchanged. Its dots and heat-lines are two renderer draw sets, it animates on the frame port's preRender, and it follows the camera through the camera port at its own 5% threshold rather than the shared `percentageChanged` ledger. The surface identity is the globe port's `surfaceVersion()` and tile readiness its `isSettled('surface')`, so the layer no longer reaches Cesium. Heat-lines are always asked for draped: where the Cesium renderer cannot drape (no depth textures) its draw port draws them (and every other draped line) over the surface, in their colour where the surface hides them and hidden past the horizon, recolouring a line in place when only its colour changes, rather than the layer dropping them.
Traffic and bikeshare expose factories through `./layers/traffic` and
@@ -2092,6 +2115,213 @@ tracked satellite's dot a set of its own drawn where the follow camera framed
it, and the orbit rings a line set dimmed behind the globe and turned with the
Earth each second. A caller handing it a Cesium viewer, with no
renderer bound, loads the `./render/cesium/unbound` export once first.
The ISS (NORAD 25544, by number, not name) is drawn as NASA's 3D model within
1,500 km of the camera and as its dot beyond; the two split the distance range,
so exactly one draws (the tracked dot too). The model is at its real 109 m span
with a 72 px floor, flies the station's usual attitude (modules along the
velocity, the same side to the Earth) from its own per-frame SGP4 sample, sits
on the follow camera's sample while tracked, loads only once the ISS is shown,
and hides with the dot's own rules. A click on it tracks the ISS.
The station has several catalog entries that are one body: the ISS itself
(listed as ISS (ZARYA)), ISS (NAUKA) and whatever vehicle is docked, each with
a dot of its own, so a click on the station's dot can land on any of them.
`issTrackedOnStation` (`policy.js`) treats a tracked entry as the station when
it is the ISS or the ISS shares its position (`_dockedCompanions`). For such
an entry the model sits on the tracked entry's own per-frame sample, as it
does for the ISS (drawn from a later sample it stood a frame of orbital
motion, over 100 m, off the camera's centre); its tracked dot and orbit ring
leave the near range to the model, its card keeps clear of the model, and the
command panel and the live video show as they do for the ISS. Such an entry
draws no orbit ring of its own (its elements would put a second line beside
the ISS's through the same station): when tracking judges it docked (both
orbits at one instant within 5 km) the ISS's ring shows for it, even with the
Stations class hidden.
In the Cockpit, a station entry is ridden from the station itself instead of
the straight-down picture: the camera sits at a point on the model
(`issMounts.js`), with the model drawn, and looks out, away from the station.
There are six views, the same whichever entry was selected: from Nauka (the
one it opens on; its lower end, out ahead to port), from Zarya (its port
side, out across the track), from Poisk (its top, out behind to starboard),
from the Cupola (just below it, out aft past the Russian segment), from Kibo
(outside the two windows on its port end, out over the Exposed Facility;
the port solar arrays turn through it as they follow the Sun, as they do past
the real windows)
and, while a vehicle is docked, from that vehicle (behind Zvezda's aft port,
out behind the station; the catalog does not say which port a vehicle is
on). The Cupola and Kibo have no catalog entries: their views ride the ISS's
and show cards of their own (`ISS_MOUNT_CARDS`: the Cupola as ESA's
observation module with 7 windows, launched 8 Feb 2010 on STS-130, on
Tranquility's Earth-facing port; Kibo as JAXA's laboratory module with 2
windows, launched 31 May 2008 on STS-124, on Harmony's port side) in place of
the entry's NORAD number, owner and TLE; altitude, velocity and the
sub-satellite point stay the station's.
Each looks 20° to 30° below level, which keeps the Earth's limb in the frame.
The points were chosen by casting rays across the view against the model at
rest: over a meter clear of every surface, at most a tenth of the frame
meeting the station and none of it nearer than 12 m. The satellites layer's
`getCockpitMount(key)` gives the eye point, the view frame, the view shown,
the catalog entry it belongs to (ISS (NAUKA), the ISS itself for Zarya,
POISK, the docked vehicle) and the views on offer, worked out from the same
sample the model is then placed on, so the camera stays on the model. The
Cockpit rides the view's own entry (`switchOrbitalSatellite`, as riding
another satellite from Live Signals), so the readout, the TLE and the signals
describe the module the view is from; a view whose entry is missing or cannot
be tracked keeps the entry ridden, and is not retried until the view changes.
Leaving the Cockpit from a station view selects Nauka (when the catalog has
it) and puts the follow camera on it, whichever view it ended on; leaving
any other satellite's view refocuses that satellite as before. `cockpitOrbitalView.js` puts the
camera there, turns the arrow-key look about the view's own frame, keeps the
chosen view for the session (also while the model is away) and publishes its
name (`mount`) and the number on offer (`mountViews`). The HUD shows STATION
VIEW with the view's name between two arrows that step back and forth
through them, wrapping round,
and leaves out the nadir reticle, north pointer, swath and sensor strip. A
ground focus (a click on a mapped installation, or its Live Signals row)
works from a station view as from the nadir picture: it aims from the
view's camera point and turns about the view's own frame, with the same
±45° reach, so the sites listed and focusable are those within reach of
where the view looks; their rows still give the angle from straight down.
Switching views ends a focus.
Until the model has loaded, or with the model not shown, the entry keeps the
nadir picture, and the TRACK ACQUIRED line follows the change. Zoom, the
place line and the identity readouts are unchanged.
While the station is selected on the map, Context gains a LIVE section
(`src/ui/liveControls.js`, `LivePanel.vue`, the `gev-live` store): the live
streams attached to the selected object. The list is fixed
(`src/data/liveStreams.js`: NASA's exterior cameras, and its live video with
space-to-ground audio, both on YouTube) and each stream names the subject
keys it belongs to; a layer names its selected subject's keys with
`getLiveSubjectKeys` (the satellites layer answers `station:iss` for the ISS,
Nauka and a docked vehicle, and `station:tiangong` for Tiangong's entries,
which no stream is listed for yet), and the keys are read again every 2 s while
something is selected, since a docked entry is judged so only once both
orbits have been compared. The section shows the subject, the stream's name
with two arrows when there are several, a screen and a status line. It
starts collapsed, and nothing loads on selection: the screen is a poster
with Play, under the line "Video plays from YouTube, which sees your IP
address." Play is the one thing that loads the player:
`youtubeLivePlayer.js` builds YouTube's player (the IFrame API from
`www.youtube.com`, the video on `www.youtube-nocookie.com`), muted, and reads
its state back. Collapsing the section, collapsing or closing Context, or the
section leaving the page removes the player, so nothing plays unseen. ON AIR
with a blinking red dot while it plays; LOADING with an amber dot while it
starts; an offline card ("Stream is offline", with Try again) when
the player reports the video ended, errored or did not start within 25 s,
which is how an off-air broadcast shows, or when YouTube's player API does not
load (Try again loads it afresh); "This stream cannot play here" when the
owner refuses embedding. Stepping to another
stream hands a playing one over; a selection that moves to an object
without that stream, clearing it or turning the layer off removes the
player, while another entry of the same station (a module, a docked vehicle,
the Cockpit stepping between views) keeps it. In the Cockpit the rail is hidden, so the
briefing card gains a fourth page, LIVE (SIG / NEWS / LOCAL / LIVE,
`CockpitLivePage.vue`), with the same subject line, arrows, screen, states
and credit, and the same poster with Play under the IP line; it is offered
only while the ridden subject has a stream, and with none the briefing keeps
its three pages. There is one player: entering or leaving the Cockpit
removes it (an embedded player cannot change parents without reloading, and
the view it was in is hidden), and so does leaving the LIVE page; the view
now shown offers Play again. Previous, Next and the tabs reach the LIVE page;
the automatic 9-second cycle never turns to it and waits while it is shown.
The credit links the owner and the stream's own page.
The model is NASA/JSC IGOAL's current-configuration station, reduced to 3 MB
with its joints kept as nodes (`public/models/README.md`), fetched only once
the camera nears its range, and drawn as an item of a `models` set of the
renderer's draw port, its joints posed through the item's `nodes`.
`issArticulation.js` turns them every frame, by default from computation: the two alpha joints and eight beta joints face
the arrays to the Sun (feathered along the flight direction in the Earth's
shadow, taken as a cylinder), the two radiator joints keep the radiators
edge-on to it, the two Ku-band dishes aim at the highest TDRS nominal slots
(41°W, 171°W, 85°E) above the station's horizon plane, and Canadarm2, Dextre's
arms and the Japanese arm sway a few degrees about NASA's filed pose, which is
illustrative only (no public source gives their real pose; the European arm has
no joints in the file). Joints turn at most 12°/s of wall time. Docked objects
are found by comparing both orbits at one instant, since dots are moved at
different moments. Selecting the station on the map also gives Context a
STATION section (`stationPanel.js` presents it through the `gev-station`
store; `StationPanel.vue` renders it; `issStation.js` describes the ISS, and
another station plugs in with a description of its own): the station's badge
and the selected entry's name, **Automatic** / **Manual**, a status line, and
**Zoom to the ISS** (offered beyond 1 km while the viewer's follow camera is
on the station's entity; it shortens the camera's offset from the station,
a step per frame, to 260 m, and never moves along the view direction, which
ran away when the camera was not looking at the station).
Automatic (the default) leaves the joints to their rules and says where the
angles come from: "Telemetry · updated 4 s ago" with a green dot while the
relay's answer is fresh, "No signal since 12:04 · computed values" when the
feed lost the station, "Telemetry unavailable · last update 3 min ago ·
computed values" when the relay cannot be reached, "Telemetry not enabled on
this server · computed values" when it is switched off, and a lit/shadow
line. Manual pauses telemetry ("Manual · telemetry paused") and offers the
controls: replace the Sun with a chosen position or a 40 s orbit, fold the
arrays and radiators as a labelled demonstration, aim the dishes and pose
Canadarm2 by joint. Automatic, a new selection, the Cockpit (the rail is
hidden there, and the section with it) or turning the layer off hand every
joint back. `issCommand.js` keeps the rules these share (the zoom's range and
step, the fast orbit, the status wording).
With `ISS_TELEMETRY_ENABLED=1` (off by default: the feed publishes no terms)
`server/providers/space/issTelemetry.js` relays NASA's public ISSLive telemetry
at `/api/iss-telemetry`. It speaks Lightstreamer's text protocol over `ws`,
holds one connection for the whole server, opens it on the first request,
closes it after two minutes without one, retries on a 5 s to 5 min ladder and
recycles a socket that is silent for 30 s. The body is
`{ configured, status, generatedAt, items }`: `status` is `live` only while the
station clock ticked with signal in the last 20 s, and each item is
`[value, epoch ms it was true at]`. On a socket loss the relay discards that
connection's joint samples and station-clock credit, so fresh signal after
reconnection cannot present old angles as live telemetry. Unset, the route answers
`{ configured: false }` without contacting the feed, and the browser asks once.
`issTelemetry.js` in the satellites layer polls it every 5 s while the model is
drawn and, in Automatic, replaces the computed targets with the feed's: each
alpha joint its own angle (port `+`, starboard `-`), each beta joint
`±angle + 90°`, each radiator joint its angle, and the first dish the direction
its elevation and cross-elevation give (the second holds still: the feed
reports one dish and not which). The mapping was fitted on a recording of the
feed: the dish direction stayed within 0.3° of a TDRS satellite across a
hand-over, and the arrays sat 5° to 9° off the Sun; the radiator joints did not
move in the recording, so their sign is not fitted. With no relay, no signal,
an answer older than 20 s, Manual, or a Sun chosen by hand, the joints keep the
computed rules, and the STATION section's status line and note say which
holds. The arms still sway for illustration.
Tiangong is the second station, drawn the same way: `stationModel.js` is the
station-independent model part (load on approach, placement from its own SGP4
sample or the tracked entry's, dot and companion hand-over, ring hidden up
close, card anchor and clearance), driven by a spec; `issModel.js` gives the
ISS's and `tiangongModel.js` Tiangong's (`CSS (TIANHE)`, NORAD 48274, with
`CSS (WENTIAN)`, `CSS (MENGTIAN)` and the Shenzhou and Tianzhou ships as its
docked entries). `policy.js` lists both stations (`STATIONS`) and
`stationModels.js` says which station a tracked entry belongs to, so the card,
the detection marks, the LIVE keys, the Cockpit's mounts and the STATION
section work per station rather than for the ISS alone (tracking records
`_trackedDockedToStation`; the ISS's ring keeps its own rules). The model
(`public/models/tiangong.glb`, 382 KB, 26 k triangles) is the project's own,
built from primitives with procedural textures, in meters and in the ISS
file's frame, so it takes the ISS's flight-axes mapping: the station is
assumed to fly with Tianhe along the velocity, the node forward, the labs
across the orbit and nadir to the Earth. `tiangongJoints.js` carries the
file's 37 joint and dock nodes (pivot, axis, kind) and
`tiangongArticulation.js` drives them, all computed, since China publishes no
telemetry: Tianhe's and the ships' one-axis wings turn about their booms to
face the Sun, the lab wings turn about the lab's axis then their own (the
SARJ/BGA pair), in the Earth's shadow every wing turns flat to the Earth
(edge into the flight direction), the high-gain antenna aims at the highest
Tianlian II nominal slot (79.9°E, 171.04°E, 10.5°E) above the horizon plane,
and the two arms sway a few degrees on one side of their folded pose (so they
lift off the hull), at the ISS's 12°/s cap. The STATION section for Tiangong
(`tiangongStation.js`) says "No public telemetry · computed values", offers
the Sun positions and the fast orbit, every wing on a slider of its own
(Tianhe's two; the labs' four, alpha and beta each: a wing set by hand holds
its angle, reached at the joint rate, while the rest keep following the Sun,
and "Follow the Sun" frees them), and the Tianhe and Wentian arms by joint;
**Zoom to Tiangong** uses the ISS's zoom range. In the Cockpit, Tiangong is
ridden from views on its model (`tiangongMounts.js`), as the ISS is: from
Tianhe (the one it opens on; its port side, out across the track), from the
outer end of Wentian and of Mengtian (out past their wings, which turn through
the view as they follow the Sun) and from the docked vehicle (standing in on
the aft port, looking behind the station); each rides its own catalog entry,
so no cards are needed, and leaving selects Tianhe. The visiting ships are
always drawn; they are not yet toggled by what the catalog shows docked.
The mission factory separates paths, replay, camera, panel, overlays and placement.
Its markers, lines, live rings and pad zone are renderer draw sets, and it
follows a mission through the camera port, so a caller handing it a Cesium
@@ -2815,9 +3045,16 @@ frame the arrows tilt, so the site stays at the centre as the satellite
moves. The turn onto it (and between sites) eases what is left of the swing
with the arrows' easing on a slower clock (250 ms time constant, about
0.75 s to 95%), then holds the site exactly, and cuts under
`prefers-reduced-motion`. A single click on a site's marker focuses it too:
the pick is resolved through the layer's own pick ids (`findByPick`: a named
marker's `{ installationId }` or a site entity). A click anywhere else
`prefers-reduced-motion`. A single click on a site's marker, or on the label
box the overlay paints for it, focuses it too: the click is resolved through
the layer's own lookup (`findAt`: `scene.pick` for a named marker's
`{ installationId }` or a site entity, else the box the overlay painted for
the label; from orbit the dot is 7 px, so the label is what a click lands
on). The map's own click selection reads the label the same way, through a
click claim (`registerClickClaim`) rather than as an interactive card: the
label takes a click only where nothing another layer draws is under it, so
an aircraft or vessel drawn under a base's label stays selectable, and
another layer's card painted over the label keeps its click. A click anywhere else
(a site beyond the reach included), the
focused site's row, NADIR, Home or `0` ends the focus and the view eases
back to nadir on the same clock; an arrow ends it (or the ease back) and
@@ -3579,6 +3816,9 @@ unaffected. `GEV_RATELIMIT_OPENAI_PER_MIN` and `GEV_RATELIMIT_GOOGLE_PER_MIN` ov
those; exactly `0` disables the limiter, while a value that cannot be read as a
number falls back to the default rather than to unlimited. Over-limit requests
receive a sanitized `429` with `Retry-After: 5` and never reach the provider.
The CCTV Street View fallback counts in a separate window with the Google cap.
An IPv6 /64 counts as one client. Over the cap, a frame gets the placeholder
image instead of a `429`.
Typed chat (`/api/openai/chat`, `/api/gemini/chat`) counts in its own window,
set by `GEV_RATELIMIT_OPENAI_CHAT_PER_MIN` and `GEV_RATELIMIT_GEMINI_CHAT_PER_MIN`.
Left unset, each takes its provider's resolved cap — `GEV_RATELIMIT_OPENAI_PER_MIN`
@@ -5162,9 +5402,20 @@ down | auth-failed` (plus the unchanged `missing-key`/`unsupported`) with
> stacking. Narrow mobile Cockpit
> viewports suppress the legacy layer stack, peripheral HUD, and secondary
> Context/briefing panes so the flight instruments and primary controls remain
> unobstructed. Outside Cockpit, Radio starts collapsed while off and lives
> only in its section inside Context; the detailed panel's normal accordion
> control owns its disclosure and its Enable/Disable action owns power.
> unobstructed. In the phone layout (720 px wide and below) the flight and
> bridge views put the collapsed Display and Radio launchers in one row
> centred on the Reset/Exit axis, above the reticle and clear of the
> destination cue: below the roll arc in portrait, over the arc's lower tick
> ends in landscape. On short landscape screens the row stays
> on screen and tappable below the vision controls but may cover the top of
> the speed and altitude readouts. Windows under 240 px tall, and those under
> 390 px wide and 450 px tall, keep the row above Reset/Exit at the right.
> Satellite and provider views keep the row
> above Reset/Exit at the right, where their own layouts expect it. The
> first-visit Stories announcement waits, unread, until Cockpit closes.
> Outside Cockpit, Radio starts collapsed while off and lives only in its
> section inside Context; the detailed panel's normal accordion control
> owns its disclosure and its Enable/Disable action owns power.
> Radio volume in the full and Cockpit surfaces, plus Space Mission
> replay speed, use Display / Sharpen's muted 3px rail, circular cyan thumb,
> glow, and mono value treatment. Their larger transparent hit areas, keyboard
@@ -7593,7 +7844,7 @@ silently demoting every later lookup for the session.
requires a developer key, so the pack needs `ONTARIO_511_API_KEY` and is skipped without one
(#1228). All sources are RAW PRIOR
poses, and the layer is stills-first.
sources. Fintraffic Finland road weather cameras (2026-09-13; cap 300, `CCTV_FINTRAFFIC_MAX_SOURCES`,
Fintraffic Finland road weather cameras (2026-09-13; cap 300, `CCTV_FINTRAFFIC_MAX_SOURCES`,
kill switch `CCTV_FINTRAFFIC_ENABLED=0`) are the fourth pack: one keyless GeoJSON station list
covering the whole country, where one _preset_ (a station's fixed view) is one camera — 806
GATHERING stations carry 2,256 in-collection presets, prioritized to 300 against seven anchors
@@ -7608,11 +7859,17 @@ silently demoting every later lookup for the session.
empty keeps the old catalog and is retried after another 15 minutes, not on the next
request. Estonia's Tarktee DATEX feeds, which answer in 2 to 16.6 s, get 30 s
(`TARKTEE_FETCH_TIMEOUT_MS`) where other packs get 15 s.
Both Tarktee XML catalogs refuse redirects and have a 4 MiB per-document
limit, enforced against declared length and streamed bytes. A rejected fetch,
unsuccessful response, or body-read failure releases both available response
bodies; read failures also cancel an active sibling reader. Cameras sharing
an image number are deduplicated before the Tarktee pack cap is applied.
Ontario catalog requests refuse redirects and log only the error class and
system code for transport/JSON failures, never the request URL or key.
Both Tarktee XML catalogs share that 30 s deadline, refuse redirects and
have a 4 MiB per-document limit, enforced against declared length and
streamed bytes. A failed request, unsuccessful response or body read aborts
the other feed at once and releases every response body, including one that
arrives after the failure. The Lithuania pack reads its two catalogs
through the same helper (`readCatalogsTogether`,
`server/providers/cctv/catalogPair.js`) with its 15 s deadline. An Ontario
error response has its body released. Cameras sharing an image number are
deduplicated before the Tarktee pack cap is applied.
- **CCTV v3 UX — viewshed + calibration gizmo** (built 2026-07-05 and field
validated 2026-07-21): the COVERAGE toggle is a
tri-state cycle `OFF → ON → VIEWSHED`; viewshed mode renders each visible camera's frustum
@@ -8400,7 +8657,8 @@ Providers normalize coordinates, canonical name, label, place types and optional
bounds. Camera framing, nearby landmark recovery and footprint selection remain
in their consumers, including the Capitol identity/containment safeguards.
Radio keeps localized country names in labels rather than station filters.
Nearby/text-search endpoints retain their existing behavior.
Nearby/text-search cache and error rules are listed under Proxy/Security
Baseline below.
**Google geocoding is a server route, not a browser call** (#363). Forward and
reverse lookups go to `/api/google/geocode` and `/api/google/reverse-geocode`,
@@ -8608,7 +8866,7 @@ pins and share links already supply complete boxes.
- The layer sits in the Data Layers panel's Cameras group beside Cameras (CCTV); Mapped ALPR Cameras moved to the Infrastructure group to make room.
- DATA ▸ **Recent Imagery** (`recent-imagery`, `src/layers/recentImagery/`) lists the last 30 UTC days of imagery for a selected box and drapes the days you pick: HLS S30 (Sentinel-2) and L30 (Landsat 8/9) at 30 m, and the VIIRS NOAA-21 daily overview at 250 m. Keyless and browser-direct: NASA CMR for which days have HLS granules in the box (`catalog.js`, both collections concurrently, paged with `CMR-Search-After` up to 2,000 records each, `truncated` against CMR `hits`), GIBS WMTS for tiles, Worldview Snapshots for thumbnails and exports. Worldview's `Data-Present` header is the only availability signal: GIBS answers an empty tile with HTTP 200, and the JPEG overview would drape black, so a day confirmed empty never drapes.
- **Box.** SELECT BOX (`src/ui/imageryBoxTool.js`) claims the pointer like Draw, borrows the stock click actions and freezes the camera only mid-drag; USE VIEW takes the camera rectangle; `boxFromPinAt` grows a 10 km box. Boxes must be ≤ 1,000 km a side, on one side of the dateline and inside the Web-Mercator limit (`validateBox`). A refusal (`boxError`, including the tool's own via `reportBoxRefusal`) stays at the top of the panel until a box succeeds; a view over the cap is refused with its size, never shrunk. An oversized box or view (`Box is … km wide · limit 1,000 km`, `View is … wide · …`) sets the hint to `Zoom in or draw a smaller box` and enables ZOOM IN: `zoomToFit()` flies top-down (1.2 s) to the refused box's centre, or the ground under the canvas centre for a view, at the height whose view is about 400 km wide for the canvas aspect and the camera's vertical FOV (`fitViewHeightM`, clamped 5–400 km), and returns that height.
- **Modes, pins and the preview.** The mode is explicit: `image` (one pinned day, slot `a`), `basemap` (that day on the left of a swipe, the basemap on the right: slot `a` splits LEFT and there is no second layer) and `ab` (day `a` left, day `b` right). `setAssignment(slot, key | null)` / `toggleAssignment(slot, key)` pin and unpin; a day is in one slot at most (pinning it in the other slot moves it) and an empty day cannot be pinned. `setMode()` keeps both pins; `b` only drapes in `ab`. While the mode's first empty slot exists the focused day previews in it: a click at once, arrow keys after 250 ms of quiet (`FOCUS_DEBOUNCE_MS`, one timer, last day wins); an unprobed day is followed and previews when its probe says present, an empty or unknown day never drapes and the previous preview stays. A pin change drops the preview until focus moves again, and unpinning takes that layer off at once. `rankLatest` picks START HERE coverage first (a present HLS day whose footprints cover the box, or have none, beats a newer sliver), then newest with all granules ≤ 20% known cloud, then newest; slivers only as `partial`; else the newest non-empty VIIRS day; it is focused and previewed after each search when nothing is pinned. Escape clears the preview (`clearPreview()`), then cancels the box tool, then blurs — with SELECT BOX armed too, through the tool's `onEscape` interceptor wired in `createApplicationTools`. Pins, mode, box and split changes made in the panel are published with `adoptLayerParams`, never while a share link is being applied.
- **Modes, pins and the preview.** The mode is explicit: `image` (one pinned day, slot `a`), `basemap` (that day on the left of a swipe, the basemap on the right: slot `a` splits LEFT and there is no second layer) and `ab` (day `a` left, day `b` right). `setAssignment(slot, key | null)` / `toggleAssignment(slot, key)` pin and unpin; a day is in one slot at most (pinning it in the other slot moves it) and an empty day cannot be pinned. `setMode()` keeps both pins; `b` only drapes in `ab`. While the mode's first empty slot exists the focused day previews in it: a click at once, arrow keys after 250 ms of quiet (`FOCUS_DEBOUNCE_MS`, one timer, last day wins); an unprobed day is followed and previews when its probe says present, an empty or unknown day never drapes and the previous preview stays. A pin change drops the preview until focus moves again, and unpinning takes that layer off at once. `rankLatest` picks START HERE coverage first (a present HLS day whose footprints cover the box, or have none, beats a newer sliver), then newest with all granules ≤ 20% known cloud, then newest; slivers only as `partial`; else the newest non-empty VIIRS day; it is focused and previewed after each search when nothing is pinned. When that day probes empty the next ranked day takes over, and focus still on the empty day moves with it, so the successor drapes once its probe says present; a day the user picked meanwhile, still being probed or debounced, keeps the slot instead (and the Esri lease while it waits), and if that pick probes empty, fails, loses its source or gives way to an empty card the hand-over goes ahead. Escape clears the preview (`clearPreview()`), then cancels the box tool, then blurs — with SELECT BOX armed too, through the tool's `onEscape` interceptor wired in `createApplicationTools`. Pins, mode, box and split changes made in the panel are published with `adoptLayerParams`, never while a share link is being applied.
- **Rendering.** `rendering.js` owns at most two GIBS layers on the host `src/maps/imageryHost.js` resolves: the globe port's imagery while the globe shows, the shown photoreal tileset while it is hidden, or none (the panel says so). Layers are rebuilt on a new host, never re-parented. `MapSourceController.subscribe()` fires after every settled stack switch (silent switches and fallbacks included) and the layer rebinds on it. Tile requests pass a six-slot semaphore of our own (Cesium's global limits are untouched). Under `requestRenderMode` a deferred tile needs a later frame, but a deferral must never request one itself or a stalled tile keeps the scene rendering every frame: a request that finds the semaphore full asks for nothing (the settling request asks for one frame), and a tile Cesium itself deferred goes through one coalesced 250 ms retry. Thumbnails: three fetches at once, sixteen decoded images, least-recently-used where a `get` read is never a use; eviction keeps the probe result. The strip's visible window is probed around the focused card; the pinned and previewed days are probed wherever they sit, at the focused card's priority, so a pin restored or assigned off-screen still drapes. A thumbnail request pass over a non-empty strip ends with `retain()`, which cancels queued and in-flight work for days it no longer wants (focus, visible range, pins, sources, the empty-day filter, a share link); a cancelled day already probed present keeps its status. When an automatic day probes empty, its successor is requested again, since scrolling may have cancelled that probe.
- **Esri surface and the swipe.** While any image is on the map (a preview included) the layer holds the map lease (`acquireImageryComparison`, owner `recent-imagery`, `esri-if-photoreal`): on Google 3D the map switches to Esri, because 30 m pixels over 3D buildings do not read, and Cesium's split only works on the globe. The lease is handed back, and Google 3D restored unless the operator changed maps meanwhile, when the last image leaves the map or the layer is disabled; the notice line says so while it holds. While a day is shown the imagery keeps Esri: a manual switch that lands on Google 3D is answered at once with a new lease under the same policy (`keepEsri`), the notice reads `Imagery stays on Esri · CLEAR to use Google 3D`, and the divider never suspends because of a map switch. It re-leases once per switch generation and only for a manual one (`MapSourceController#getSwitchOrigin()` is `manual` for an outside `setStack`, `automatic` for the controller's own fallback or recovery), never for the lease's own switch and never while releasing, so CLEAR and disable still return to Google 3D; a lease another owner holds stays refused. When Esri fails and the controller falls back to Google 3D on its own, the layer lets the lease go without restoring anything, both days drape on the tileset without a swipe and the guidance reads `Esri map unavailable · no swipe`; only the next manual switch to Google 3D takes Esri again, so a failing Esri map is never retried in a loop. Switching between globe stacks (Esri, OSM, Bing) only rebinds. On a tileset host both images drape without a swipe. The split control (`createImagerySplit`, labels A / B, or IMAGE / BASEMAP) exists only while a swipe is live, recentres on every new comparison (not while scrubbing the second day), except once for a split restored from a link. There is no keyboard shortcut (Space is the app's push-to-talk): SWAP (`#ri-swap`, in the controls row, its slot reserved and the button disabled while no swipe is live) trades the two sides, A to the right and B to the left, or the image to the right of the basemap, and a second click restores; the hint reads `Drag the divider · SWAP trades sides`. The swap is not in the share link, and every new comparison starts unswapped. One opacity applies to both drapes. Another owner holding the lease is reported and the images still drape.
- **Panel.** `#recent-imagery-panel` is rail markup in the Context view (`src/ui/views/ContextView.vue`) beside `#weather-panel`, registered in `PanelLayoutController`, `PanelChrome` and the share panel state (`ui` token `i`), hidden while the layer is off and opened on first appearance unless a stored or shared collapse choice exists (`dataset.collapsedPreference`). `LayerPanel.attachRecentImagery(factory)` mounts the readout (`createRecentImageryPanel` in `src/ui/recentImagery.js`) into its `data-rail-scroller` body, as the weather readout is mounted. The body is a fixed stack in which no block changes height and no control moves in any state: `#ri-actions` (SELECT BOX · USE VIEW · CLEAR, rail-card actions), `#ri-notice` (one line: `#ri-notice-text` with the refusal, error, CLEAR notice or the Esri note, and `#ri-zoom-in` in a fixed slot at its right end, reserved unless an oversized box or view can be fitted), `#ri-strip` (day cards newest first: thumbnail, date, sensor, cloud, START HERE, PREVIEW, and a SHOW chip or A and B chips; ← → Home End move, S / A / B pin, Enter previews), `#ri-hint` (one line: the next step), `#ri-selection` (`#ri-mode` IMAGE · VS BASEMAP · A / B, then `#ri-slot-a` and `#ri-slot-b` with `#ri-unpin-a` / `#ri-unpin-b` kept in place when empty), `#ri-controls` (`#ri-opacity`, EXPORT / EXPORT A and EXPORT B) and `#ri-details`, a collapsed rail card holding the focused day's acquisition, coverage and granules, catalog notes, the empty-days toggle and the imagery credit. Absent controls are reserved (invisible), never removed. Every render preserves the body's `scrollTop`; opening DETAILS then scrolls its card into view once.
@@ -8720,6 +8978,7 @@ apply to `public/` assets or built-preview serving.
- `AISSTREAM_API_KEY` is server-side only; the browser reads the same-origin `/api/vessels` cache.
- `/api/google/nearby-places` keeps the Google key out of Places requests issued for voice scene context.
- `/api/google/text-search` keeps the Google key server-side for view-biased Places recovery used by annotation resolution.
- The Places and geocode routes cache only an answer (`private, max-age=300`). For geocode, that means an HTTP success with `OK` or `ZERO_RESULTS`, because Google sends `REQUEST_DENIED` and `OVER_QUERY_LIMIT` as HTTP 200. Every other reply, including an unreadable Places 200 (now a 502), is `no-store`.
- `/api/overpass` is bounded by body/response caps, per-client/global rate limits, concurrency limits, operator-configured upstreams, in-flight dedupe, cache bounds, and static validation that every selector is spatially bounded.
- `/api/military-installations` uses an independent limiter with the same 90-per-client/300-global one-minute bounds, so viewport installation refreshes never consume `/api/overpass` annotation capacity.
- `/api/route` proxies bounded OSRM route requests for annotation routes, with profile allowlisting, distance caps, response caps, caching, and sanitized "no route found" errors; routing outages return 502.
@@ -8820,7 +9079,7 @@ apply to `public/` assets or built-preview serving.
The ship bridge and the orbital view have their own own-layer gates
(`bridgeEntryAllowed`, `orbitEntryAllowed`; see
[Ship bridge view](#ship-bridge-view) and [Orbital view](#orbital-view)).
- **Aircraft cockpit view:** selecting a commercial or military aircraft reveals a `COCKPIT` action once that aircraft's own layer is on; Contacts is not required. The action shows in Contacts (`#cockpit-entry`) and under the Intel HUD's bottom-right readout (`#cockpit-entry-readout`, with the HUD on, in its Tactical layout), so entry needs no Contacts panel. Turning off the ridden aircraft's layer ends the view (the layer clears its tracking). When an air feed the surrounding picture uses is off, the Cockpit says so instead of blocking: Live Signals leads with `MILITARY FEED OFF · PARTIAL PICTURE` or `FLIGHTS FEED OFF · PARTIAL PICTURE`, and the Contact card shows that line in place of its inputs status, so a partial picture is never presented as complete. `control_cockpit` refuses entry with `entryBlockedReason` `contacts-starting` (a Context change in flight), `subject-layer-off` (the aircraft's layer, or with none tracked any aircraft layer, is off) or `no-tracked-aircraft`. Cockpit mode temporarily releases Cesium's orbit-follow transform and drives a first-person camera from the tracked aircraft's existing smoothed display position and course. Its concave helmet-visor HUD shows callsign, UTC time, coordinates, curved roll/pitch guides, and a seven-division heading tape. Ambient commercial and military AIR contacts use a Cockpit near/far band selected by the shared Display 3D mode: Proximity admits at 150 km and retains to 185 km; All admits at 400 km and retains to 450 km. The shared 3D toggle now applies in Cockpit: Off keeps in-range contacts as rotating 2D aircraft silhouettes, while On lets a ready admitted glTF take over without a drawing gap. The Cockpit model cap remains 60 and can only lower the map budget; in-range contacts that are capped or still loading remain 2D silhouettes instead of degrading to out-of-range dots. Contacts outside the selected band use small rotation-free cyan-white pips for civilian aircraft and amber pips for military. The pilot's own airframe is not drawn in first person, and exiting clears the Cockpit band and restores normal map silhouettes/models. The normal left accordion remains available for Layers and Authoring and keeps the same 26vh HUD-aligned anchor used in map mode, independent of whether the bottom-left Contact card is expanded or collapsed; the right-rail CCTV and Context chooser are hidden to avoid duplicating or overlapping the cockpit presentation. When the intelligence HUD is enabled, its classification, scene summary, collection/orbit metadata, coordinates, and imaging-status text remain visible as reduced peripheral cockpit telemetry; the optical center and flight instruments stay clear, and turning the HUD off still hides it. Ground speed, exact heading, and rendered altitude form one compact lower-center instrument cluster, keeping the horizon and peripheral view open. A second live altitude tape hugs the inside-right visor rim: its rail and nine moving ticks are derived from the same responsive keyhole radius, remain 20 px inside the circular edge even on wide displays, fade deeply at both vertical ends, and move continuously behind a fixed current-altitude pointer; its interval tightens automatically near the surface and it is suppressed on narrow screens. Cockpit mode also has a weather-backed transparent volumetric-cloud pass derived from the supplied FBM/domain-warped R&D shader. It renders at no more than 520×320, uses 24 ray steps/three FBM octaves at 12 FPS, is clipped to the visor, fails clear when Open-Meteo is unavailable, and stops its animation completely on cockpit exit. It does not restore the prior CPU weather canvases, precipitation, scene fog, or any map-mode effect. When Contacts is active for the tracked aircraft, the compact `CONTACT` rail adds the 250 km subject window, four cohort counts, nearest observed/mapped example with relative bearing and distance, freshness, explicit uncertainty, and Previous and Next controls plus its own collapse control. The mirrored right rail is a three-page **cockpit briefing carousel**: source-backed live signals, location-matched regional headlines, and physical-region/current-weather context from bundled Natural Earth polygons and Open-Meteo. It is manual-first: Previous, Next, and direct page controls are always available, and the visible `CYCLE OFF` / `CYCLE ON` control starts or stops the nine-second page cycle. The cycle pauses on hover or keyboard focus and stops while collapsed, hidden, or outside cockpit mode; live signal data continues refreshing either way. Regional News holds its list to the last 24 hours (`newsWindowHours` in the brief; the client drops anything older or undated even if a server sends it) and shows four distinct states: fresh results with no banner and a `{SOURCE} · LAST 24 H` footer; earlier results kept after a failed refresh under `CACHED · REFRESH FAILED` with the cached age in the footer; `NO NEWS IN THE LAST 24 H` when a source answered with nothing Line truncated
- **Aircraft cockpit view:** selecting a commercial or military aircraft reveals a `COCKPIT` action once that aircraft's own layer is on; Contacts is not required. The action shows in Contacts (`#cockpit-entry`) and under the Intel HUD's bottom-right readout (`#cockpit-entry-readout`, with the HUD on, in its Tactical layout), so entry needs no Contacts panel. Turning off the ridden aircraft's layer ends the view (the layer clears its tracking). When an air feed the surrounding picture uses is off, the Cockpit says so instead of blocking: Live Signals leads with `MILITARY FEED OFF · PARTIAL PICTURE` or `FLIGHTS FEED OFF · PARTIAL PICTURE`, and the Contact card shows that line in place of its inputs status, so a partial picture is never presented as complete. `control_cockpit` refuses entry with `entryBlockedReason` `contacts-starting` (a Context change in flight), `subject-layer-off` (the aircraft's layer, or with none tracked any aircraft layer, is off) or `no-tracked-aircraft`. Cockpit mode temporarily releases Cesium's orbit-follow transform and drives a first-person camera from the tracked aircraft's existing smoothed display position and course. Its concave helmet-visor HUD shows callsign, UTC time, coordinates, curved roll/pitch guides, and a seven-division heading tape. Ambient commercial and military AIR contacts use a Cockpit near/far band selected by the shared Display 3D mode: Proximity admits at 150 km and retains to 185 km; All admits at 400 km and retains to 450 km. The shared 3D toggle now applies in Cockpit: Off keeps in-range contacts as rotating 2D aircraft silhouettes, while On lets a ready admitted glTF take over without a drawing gap. The Cockpit model cap remains 60 and can only lower the map budget; in-range contacts that are capped or still loading remain 2D silhouettes instead of degrading to out-of-range dots. Contacts outside the selected band use small rotation-free cyan-white pips for civilian aircraft and amber pips for military. The pilot's own airframe is not drawn in first person, and exiting clears the Cockpit band and restores normal map silhouettes/models. The normal left accordion remains available for Layers and Authoring and keeps the same 26vh HUD-aligned anchor used in map mode, independent of whether the bottom-left Contact card is expanded or collapsed; the right-rail CCTV and Context chooser are hidden to avoid duplicating or overlapping the cockpit presentation. When the intelligence HUD is enabled, its classification, scene summary, collection/orbit metadata, coordinates, and imaging-status text remain visible as reduced peripheral cockpit telemetry; the optical center and flight instruments stay clear, and turning the HUD off still hides it. Ground speed, exact heading, and rendered altitude form one compact lower-center instrument cluster, keeping the horizon and peripheral view open. A second live altitude tape hugs the inside-right visor rim: its rail and nine moving ticks are derived from the same responsive keyhole radius, remain 20 px inside the circular edge even on wide displays, fade deeply at both vertical ends, and move continuously behind a fixed current-altitude pointer; its interval tightens automatically near the surface and it is suppressed on narrow screens. Cockpit mode also has a weather-backed transparent volumetric-cloud pass derived from the supplied FBM/domain-warped R&D shader. It renders at no more than 520×320, uses 24 ray steps/three FBM octaves at 12 FPS, is clipped to the visor, fails clear when Open-Meteo is unavailable, and stops its animation completely on cockpit exit. It does not restore the prior CPU weather canvases, precipitation, scene fog, or any map-mode effect. When Contacts is active for the tracked aircraft, the compact `CONTACT` rail adds the 250 km subject window, four cohort counts, nearest observed/mapped example with relative bearing and distance, freshness, explicit uncertainty, and Previous and Next controls plus its own collapse control. The mirrored right rail is a **cockpit briefing carousel** of three pages: source-backed live signals, location-matched regional headlines, and physical-region/current-weather context from bundled Natural Earth polygons and Open-Meteo; a fourth page, LIVE, is offered while the ridden subject has a live stream (the ISS; see the LIVE section) and shows the same muted player. It is manual-first: Previous, Next, and direct page controls are always available, and the visible `CYCLE OFF` / `CYCLE ON` control starts or stops the nine-second page cycle. The cycle pauses on hover or keyboard focus, never turns to the LIVE page and waits while it is shown, and stops while collapsed, hidden, or outside cockpit mode; live signal data continues refreshing either way. Regional News holds its list to the last 24 hours (`newsWindowHours` in the brief; the client drops anything older or undated even if a server sends it) and shows four distinct states: fresh results with no banner and a `{SOLine truncated
- **Cockpit left-panel clearance:** the Cockpit Contact card and peripheral HUD participate in the adaptive left accordion's live obstacle measurements, including live viewport-height changes. Expanding Layers or Authoring keeps the active panel in the available upper-left corridor with internal scrolling; it does not cover the Contact card, lower Cockpit controls, or Cesium credit line. Outside Cockpit the hidden card does not alter the normal corridor.
- **Cockpit Context scope:** the 250 km radius applies to the air/sea proximity cohorts. Installation counts come only from the currently loaded viewport and are labeled `CURRENT VIEWPORT ONLY` in the cockpit as well as the normal Context panel; neither surface presents them as a complete 250 km installation survey.
- **Cockpit camera anchor:** first-person mode does not write feed-boundary corrections directly into the camera. A cockpit-only inertial anchor advances from the selected aircraft's displayed course and speed, then converges toward the authoritative delayed track with correction capped below forward motion. The displayed kinematics are derived from the same consecutive fix segment as the rendered position, with raw feed speed/course used only as fallback; a transient zero/missing feed speed therefore cannot freeze a visibly moving aircraft after layer enable or a map/cockpit handoff. Rendered altitude continues to come from that interpolated track position. Late ADS-B fixes and short render stalls can remove drift without accelerating or reversing the view. Camera placement runs before scene update/culling at a bounded 20 Hz so a moving cockpit does not force Photoreal 3D Tiles to retraverse on every display frame; textual instruments update at 10 Hz and context/layout work at 4 Hz. Every far Cockpit contact pip shares one stable Cesium texture-atlas entry and skips unused screen-projected course calculations, while only in-range 2D aircraft silhouettes pay the screen-projected rotation cost; ambient glTF collections are hidden/retained rather than synchronously destroyed at cockpit entry, and context rails lay out only on explicit content/state changes and viewport resize. The deliberate 15/30-second layer interpolation delays and per-Cesium-frame position caches remain unchanged.
@@ -8831,7 +9090,7 @@ apply to `public/` assets or built-preview serving.
- **Cockpit weather status:** the earlier multi-canvas atmospheric compositor remains fail-closed and is not attached to the live viewer. Cockpit clouds are a separate transparent WebGL pass with a capped 520×320 framebuffer, 24 ray steps, three FBM octaves, and a 12 FPS ceiling. It defaults off and starts only when local storage explicitly contains the persisted `WX ON` opt-in (`'1'`). When opted in, observations refresh after five minutes or 25 km of aircraft movement, fail transparent when unavailable or clear, and stop on exit or disable. `WX OFF` governs atmospheric rendering only: the briefing still fetches source-backed Natural Earth region, headline, and Open-Meteo local-information data, aborting and replacing any in-flight request when the selected aircraft changes. No weather effect runs in map mode and no synthetic fallback is shown.
- **Cockpit trail visibility:** entering cockpit hides the selected aircraft's trail body and head so they cannot cross the first-person view; exit restores them. This cockpit-only presentation change does not alter the normal map-mode invariant that aircraft trails render through terrain using their depth-fail material.
- **Aircraft course slew:** civilian and military 3D models retain the 60°/s course limiter, but each rendered frame can consume at most 250 ms of accumulated slew time. A long tile/render stall therefore catches up over multiple visible frames instead of turning one delayed frame into a heading snap.
- **Manual-first cockpit briefing:** the right-side Live Signals / Regional News / Local Info carousel does not advance automatically on page load. Previous, Next, and direct page controls remain available; the visible `CYCLE OFF` / `CYCLE ON` toggle explicitly starts or stops the nine-second page cycle, which still pauses on hover/focus and while collapsed, hidden, or outside cockpit mode. Live signal data continues refreshing in either state.
- **Manual-first cockpit briefing:** the right-side Live Signals / Regional News / Local Info carousel (plus LIVE while the ridden subject has a stream) does not advance automatically on page load. Previous, Next, and direct page controls remain available; the visible `CYCLE OFF` / `CYCLE ON` toggle explicitly starts or stops the nine-second page cycle, which still pauses on hover/focus and while collapsed, hidden, or outside cockpit mode, and which never turns to the LIVE page. Live signal data continues refreshing in either state.
- **Photoreal horizon blend:** Cesium's sky atmosphere remains enabled behind the hidden base globe, but its light intensity, saturation, and brightness are reduced from the library defaults so the distant Google Photorealistic 3D Tiles boundary blends into the sky instead of producing a bright cyan horizon seam.
- **Cockpit direction and speed tapes:** plausible destination metadata now drives one translucent, isometric visor chevron labeled directly below with the estimated geographic bearing; the prior full geodesic dashed path is not rendered. A mirrored live ground-speed tape follows the inside-left keyhole rim using the same responsive curve, end fades, fixed pointer, and fractional tick motion as the altitude tape on the right. Speed values scroll upward as they increase while altitude values scroll downward. Its tick endpoints and current-speed pointer share the rail's inset-circle origin, so the markings stay attached to the visible curve rather than drifting inward with the text-label gutter.
- Voice control UI (`#gev-voice-control`) shows status states OFF / CONNECTING / LISTENING / EXECUTING / ERROR.
+2 -2
View File
@@ -83,6 +83,6 @@ Consumers use the package exports, which do not import standalone application
globals.
Dataset files and source/license notices remain under `src/data/local_data/`.
Their input counts are 4,352 datacenter features and 704 dam features. A multipart
Their input counts are 4,352 datacenter features and 706 dam features. A multipart
geometry draws as several parts, each with its own stem; part and stem counts are
different measurements from feature counts. This change does not refresh or relicense the datasets.
different measurements from feature counts.
+11
View File
@@ -343,6 +343,17 @@ correction turns them to, paints the mask in a custom shader per colour
clipping plane through the model's origin. The checks `draw.models.bytes`
and `draw.models.fleet` cover them.
An articulated model's joints are its item's `nodes`: transforms of the
model's named glTF nodes, each a column-major 4×4 matrix in the node's
parent frame that replaces the node's own (the ISS's solar arrays,
radiators, dishes and arms). They change live; a node whose name leaves the
map goes back to its authored pose, and names the model lacks are ignored.
A renderer that poses nodes lists `modelNodes`; one without it draws the
model as authored. The Cesium port lists it, and writes each node's matrix
once the model is ready, and again only when its numbers change. The check
`draw.models.nodes` moves the harness model's node off the screen's centre
and back.
The Cesium port draws timed imagery (`imageryFrames`): a layer given a timed
source hides its own imagery and draws each frame its clock shows as a Cesium
imagery layer just above it (in `blend` mode the later frame over the earlier