camera.follow({ position, viewFrom, minRange = 150, native? }) keeps a
moving target centred at an east-north-up offset the user may turn and
zoom; camera.following() and camera.followChanged report it; stop() frees
the camera and a camera command (setView, flyTo, flyToSphere, lookAt,
cancelFlight) ends it.
Probe ruling: Navara 0.2.1's cameraFollow holds the target but ignores the
user's drag and wheel, the aircraft sits ~13 px off centre with 7.6 px of
spread, and calling it without an offset each frame gives a NaN camera.
So Navara (and the fake engine) follow with src/globe/follow.js: each
frame, before Navara draws, the offset is re-read from where the user put
the camera relative to the previous target, kept outside the minimum
range and carried to the new target.
Cesium follows with viewer.trackedEntity: the layer's own Entity
(`native`) or a proxy Entity at position() with a 1 px transparent point,
since EntityView moves the camera only for an entity with a bounding
sphere. applyTrackedCameraFrame and the zoom limits moved to
src/globe/cesium/follow.js; src/data/trackedCamera.js keeps the pure parts
(the range clamp now lives in src/globe/follow.js). following() also
reports an entity the app tracked itself. Camera sensitivity claims made
on a viewer's Cesium camera now share the engine camera's ledger, and the
orientation controls and the world jump read the follow from the engine.
Contract (both engines): a target flying at 250 m/s stays within 3 px of
the centre for 2 s, zoomIn through it stops at 150 m, stop() and setView
end the follow. Cesium 71/71, Navara 77/77; shell 23/25/4.
Review of Task 0:
- build/navaraDependencies.js reads the dependency closure of
@navaramap/three and @navaramap/three-default-plugin from node_modules
(dependencies and peers) and drops what the app's other bundled
dependencies reach too (Cesium is served as a script, so draco3d, which
only Cesium shares, stays engine-only), plus three and postprocessing:
88 packages today (the @navaramap/* packages, @takram/*, @react-three/
fiber, @sparkjsdev/spark, astronomy-engine, d3*, draco3d, n8ao,
three-stdlib, …). The browser build's initialChunkGuard fails on a
module from any of them in an initial chunk. Tested on a fixture tree.
- check-main-chunk's budget measures the entry the page loads (the same
chunk the composition check reads), falling back to phase 4's largest
index-*.js only for a dist without a page.
Build clean: no Navara-only module in the initial chunks; main chunk
2,869,565 bytes (gzip 889,041).
Review of Task 0: the recorded live feeds left vessels (no AIS key here)
and local ADS-B (no receiver) empty, and flights held one adsb.lol
viewport answer. src/tooling/w1Feeds.js builds seeded answers relative to
one frozen instant (2026-10-16T12:00Z): a global /api/flights OpenSky
snapshot (8 000 aircraft, 2 000 within 60 km of JFK), /api/vessels with
12 000 AIS rows worldwide and the rows within the city area (the area
answer first, the global one last as the replay's fallback), and
/api/local-receivers/aircraft with 40 aircraft near JFK on 1090 MHz and
978 MHz UAT. Tests read them back through the app's own parsers.
qa-w1-baseline --synthesize=DIR writes them beside the recorded
military, satellites and launches; a replay now logs every feed request
and how it was answered (exact / as another key / none) into
summary.json, and --compare diffs those too.
Replayed on d50eaa28 and HEAD: flights 8 000, military 154, local ADS-B
40, vessels 12 000, satellites 824, launches 24; the runs agree (counts,
known objects, screen points, tracking poses, requests). The Cesium
reference is re-recorded this way before Tasks 10a, 12a and 13a.
Navara 0.2.1's transparent pass refreshes its copy of every scene light
each frame with copy(light, true). For the sun's cascaded shadow lights
(three-csm CascadedDirectionalLights) that copy clones every cascade, so
each frame brought new DirectionalLights with new ids and three's
WebGLLights uniforms cache (keyed by id) grew by about 1.2 KB a frame in
the looks whose sun casts shadows (minimal, photoreal).
src/globe/navara/cascadedLightCopy.js patches the copy to update the
cascades it has in place (pose, colour, intensity, target, shadow
settings and camera) when the counts match, and keeps the rebuild
otherwise; internals.js finds the class among the view's scene lights
(private view._scenes.light) and applies it when the globe is built.
Tests on three-csm's own class: the original copy changes the light ids,
the patched one keeps the lights and takes the values.
Post-GC JS heap (CDP HeapProfiler.collectGarbage every 10 s, the app on
?engine=navara rendering continuously at 60 fps): minimal 4.33 → 0.20
MB/min over 70 s and 0.06 MB/min over 180 s; photoreal 4.34 → 0.19
MB/min; cinematic (no shadow) 0.34. Added to the upstream asks.
input.on('click'|'doubleClick'|'move', fn, { pick, drill }) delivers
{ …event, picked, drilled }: PickResult { id, owner, instance, primitive
(Cesium) } names the mover set (and instance) or line hit. One pick (and
one drill, as deep as the deepest listener) per event, shared by its
listeners; moves with picks are coalesced to one delivery per frame with
the latest position (src/globe/pickDelivery.js). Option checks in the
contract (normalizeInputOptions).
- Cesium: scene.pick / drillPick, and the picker names the owner: movers
map billboards, points and models to their set and instance, lines
their primitives to the line.
- Navara: a click with pick takes Navara's GPU pick of the same gesture
(featureClick; its pick helper answers on the same pointerup, usually
before the engine's click, so a fresh answer waits for it; otherwise
the click waits up to 250 ms, then the CPU pick). Navara 0.2.1 answers
clicks in quick succession with the first click's pick: a GPU answer
whose instance is not drawn near the click is stale and the CPU pick
answers. Hover, drill and the synchronous picker.pick/drillPick are CPU
picks (src/globe/cpuPicks.js) of pickable mover sets and pickable lines
(within half the width + 3 px), behind the globe excluded, nearest
first; lines are picked on the CPU after an empty GPU answer.
picker.clickStats() (Navara only) counts the answers and latencies.
- The fake engine picks its own movers and lines on the CPU unless
scripted.
- The contract suite clicks with real pointer events (puppeteer) on both
engines: a sprite is picked with its owner and instance, the map picks
nothing, two sprites in one place drill both, a line is picked, a
hover picks the sprite under it. Navara, 20 clicks: payload p50 2.8 ms,
p95 4.1 ms after the click (puppeteer dispatch included), 20/20 GPU
answers. The Cesium harness keeps the first-run card closed.
engine.lines.createLine({ id, widthPx, color, depthFailAlpha, dash
({ lengthPx, gapPx } | 'tactical'), arc ('none'|'geodesic'), modelMatrix,
pickId }) → a Line: setPositions (Cartesian3[] or a flat Float64Array),
setColor, setModelMatrix, show, destroy; lines.list() for QA. Options,
flat positions, geodesic densification (great circles of the geodetic
normals, height interpolated, one step per degree) and the registry are
engine-neutral (src/globe/lines.js); the fake engine keeps every line.
- Cesium: a Primitive of a PolylineGeometry (synchronous, rebuilt when
the positions change) with a PolylineMaterialAppearance in an
engine-owned collection; depthFailAlpha is depthFailAppearance with the
alpha scaled; pixel dashes are PolylineDash (16-bit pattern); the
tactical dash is the launches' selected-orbit shader under its own type
(launches keep theirs until Task 15a moves them onto engine.lines).
- Navara: a public MeshDesc (transparent pass, unlit) holding a ribbon
of one quad per segment, widened to CSS pixels in the vertex shader;
a second copy drawn with GreaterDepth and the alpha scaled is the
depth-fail part; dashes are cut in the fragment shader (pixels along
the screen direction, or the tactical pattern over the fraction along
the line); positions relative to the first point, the model matrix on
the group. Geodesic arcs are densified on the CPU. Not pickable yet.
- The contract suite draws a three-point line on both engines: magenta
along both segments, nothing off it, 1° east once turned by its model
matrix, gone when hidden; a line 30 km under the ground shows only its
depth-fail copy, dimmer than the line.
Main chunk 2,867,499 bytes (+7,197, +0.25 %: the Cesium lines).
engine.movers.createSet({ id, kind: sprite|point|model, capacity, order,
depthTest, sprite|point|model, pickable }) → a MoverSet: set(index,
fields) stores an instance (position, screen rotation, orientation, rgba,
scale, atlas cell, show, pickId) in flat arrays, commit() uploads the
changed range once and requests one render, positionsView() reads the
positions, destroy() removes it; movers.list() for QA. The options, the
store and the registry are engine-neutral (src/globe/movers.js); the fake
engine keeps them for layer tests.
- Cesium: one primitive per set in an engine-owned PrimitiveCollection
added with the first set, in `order`: sprites a BillboardCollection
(one image per atlas cell, NearFarScalar, disableDepthTestDistance),
points a PointPrimitiveCollection, models a Model per instance.
- Navara: a public MeshDesc per set. Sprites and points are one
InstancedMesh of a quad, faced to the screen and sized in CSS pixels in
the vertex shader from uniforms (the pick pass draws the same
footprint), atlas uv, rotation, colour and scale per instance, drawn in
the transparent pass unlit; offsets from an origin that follows the
camera past max(1 km, 10 % of its height). Models: the glTF once, one
InstancedMesh per mesh, minimumPixelSize/maximumScale on the CPU per
drawn frame, colorBlend in the shader. Pickable sets register a
PickableInstancedMeshWrapper; pickIds.js maps batch ids back to
instances (used by Task 4).
- The contract suite draws three 24 px sprites on both engines: centred
within 2 px of worldToWindow, hidden by show:false, recoloured, gone on
destroy, and no frames added once the commit has drawn.
- qa-engine-bench --scenario=movers: 10 000 synthetic aircraft
(src/bench/syntheticFlights.js, moved from the Navara spike with its
test), every instance moved and committed every frame. p95 Navara
(look=minimal) / Cesium: global 5.2/7.0 ms, city 5.6/9.9, orbit 6.0/7.4
(0.74, 0.57, 0.81; gate 1.1). Cesium is measured without its 60 fps cap.
Main chunk 2,860,302 bytes (+9,706, +0.34 %: the Cesium movers).
Review of Task 0: the live baseline could not be compared between
commits. qa-w1-baseline now records every /api response of each layer's
run (--record=DIR, DIR/<layer>.json), replays them through puppeteer
request interception (--replay=DIR; an /api request outside the
recording answers 503) with the page clock frozen at the recording's time
(evaluateOnNewDocument: Date.now and new Date(), hence JulianDate.now and
the app clock), and diffs two runs (--compare=A,B: counts, known objects'
positions within 1e-7° and 5 cm, screen points within 0.5 px, tracking
poses within 1e-6 relative). The rules live in src/tooling/w1Baseline.js
with tests.
Recorded once (qa-artifacts/w1-feeds, git-ignored), replayed on
d50eaa28 and on HEAD: the runs agree within the tolerances for flights,
military, satellites (counts, five known objects each, tracking poses)
and launches (count). Local ADS-B has no receiver and vessels no AIS key
here: both stay empty. This is the reference for the extraction tasks.
Review of Task 0:
- build/initialChunkGuard.js, a Rollup plugin in the browser build, fails
the build when an initial chunk (an entry and what it imports
statically) holds a module from node_modules/three, @navaramap/* or
postprocessing. Checked by forcing it on satellite.js, which the main
chunk does hold: the build fails naming the chunk and module.
navaraAssets.test now pins both build plugins.
- scripts/check-main-chunk.mjs is one script for both main-chunk checks:
the composition (text markers in the initial chunks, sizes) and, when
scripts/main-chunk-baseline.json exists, the phase-4 branch's gzip-9
budget, with its exports (checkMainChunk, measureMainChunk) so the two
branches merge into this file. The composition test moved to
src/tooling/mainChunkComposition.test.mjs to leave phase 4's
mainChunk.test.mjs its name.
- CI runs npm run check:main-chunk after each npm run build (unit,
browser-gate and Windows jobs).
The phase-2 review fixes: look internals in internals.js, the groundFloor
resample throttle, tile height selection, the sharpen mirror, QA tile
cache limits, engine.groundFloor diagnostics. In navara/index.js both
sides are kept: the look gets internals.js's luminance uniforms and reads
the sun's time from the engine clock's frame time, so engine.clock stays
the one time source for the sun.
- The takram uniform paths move to internals.js (forEachLuminanceUniform,
loud in dev when a path is gone); look.js sets each factor to scale x
the recorded base value, so shared or copied vectors are never scaled
twice, and the stars keep the base value (own vector when shared).
Stars at intensity 120: a starfield shows from space.
- Both night-side shader variants compile on the first two frames.
- The clouds hide once the sun sets at the camera (dark puffs at dusk).
- No allocation on look frames that change nothing (frozen updates, a
scratch camera place from livePosition).
Gallery regenerated.
scripts/check-main-chunk.mjs (npm run check:main-chunk) finds the main
chunk of a build (the Vite manifest's entry, or the module script
dist/index.html loads), follows its static imports and fails when any of
them carries three or Navara markers; it prints the sizes. Base build:
main chunk 2,850,229 bytes (gzip 881,783), no static imports, clean.
scripts/qa-w1-baseline.mjs records Cesium mode for the W1 layers (flights,
military, local ADS-B, vessels, satellites, launches): counts, five known
objects with positions and screen points, JPEGs (path, size, sha256) at
the global / New York (JFK) / orbit poses, a tracked object's camera pose
after 2 s and what its share link restores. Live sources until the
fixture sources of Task 8.
qa-globe-contract checks the noon sky facing the horizon (daylight blue,
no star pixels) and stars at midnight; qa-engine-shell compares
sampleHeight at the Austin startup pose across engines and reports the
wheel floor's lowest frame. QA pages load OSM and keyless terrain tiles
through a disk cache with backoff on 429. Report updated.
The contract gains render.setSharpen({ enabled, amount }) and
render.sharpen(). Navara draws it as a custom EffectDesc: Cesium's unsharp
mask (3x3 box blur, gamma 2.2 round trip so one amount matches Cesium)
right after tone mapping and before SMAA/FXAA, skipping sky and cloud
pixels by scene depth. The Cesium engine keeps the state only; Cesium mode
still sharpens with its own post stage, unchanged. The fake engine records
it and the contract suite checks it.
Without a viewer the Sharpen slider and toggle drive the engine with the
same slider-to-amount mapping. On Navara the defaults, presets and share
links read Cesium's 49 % as 80 %.
Prettier on the 13 new test files that failed --check; format-scope.json
now lists every file phase 2 added (src/globe/**, engineChoice, navaraMode,
cameraAngles, the QA scripts and pages, build/navaraAssets and tests).
Contract: camera.position(result?) and camera.height() for per-frame
readers, render.request(reason?), maps.releasePhotorealTiles, sources
with a caller-owned ready provider, isGlobeEngine. engine.maps now hides
every tileset it has shown, lets the last imagery/photoreal call win, and
setPhotorealTiles(null) restores the last base imagery and its credit
instead of leaving the photoreal credit on a bare globe.
Cesium engine: reads the Viewer camera/canvas aliases, detaches bridged
events through the remover Cesium returns, raises frame timestamps
without allocating, re-aims the horizon occluder only when the camera
moved, and renders the scene directly when the governor governs another
viewer (or none yet). engineFromViewer returns the registered scene
engine of a viewer, else builds and caches one. The render governor stays
the Cesium render backend under engine.render and gains
isRenderGovernorInstalled.
The scene builds the page's Cesium engine right after the viewer
(src/app/globeEngine.js), names the startup Google tileset to it once
loaded, returns it beside the viewer and destroys it before the viewer.
The debug handle exposes it as window.__godsEyeView.engine. Nothing
consumes it yet; render holds taken before the governor is installed
apply at install.
qa-globe-contract now runs on the app's engine (self-built only when the
app has none); a throwaway engine checks creation and destroy leave the
scene untouched.
Map sources are applied through one src/maps/sceneApply.js helper shared by MapSourceController and engine.maps (live layer kept for an unchanged provider, providers cached per source). The contract gains flyTo endTransform and rectangle destinations, flyToSphere easing, drillPick width/height and pick content, with Cesium's own flyToSphere range. Input skips listeners removed during dispatch and stays inert after destroy.
createCesiumEngine wraps an existing viewer without changing it: camera, picker, projector, frames, render (through the render governor), surface, input, credits and maps, with geo values out and Cesium values in at the boundary. Map and surface code is shared with src/maps and services/groundSnap instead of copied. scripts/qa-globe-contract.mjs runs the contract suite against the engine in headless Chrome; the camera-events check now waits for the previous move to end, as Cesium's moveStart marks a move from rest.
src/globe/contract.js defines the engine-neutral GlobeEngine capabilities
(JSDoc), createEngineEvent with Cesium.Event semantics and
assertGlobeEngine. src/globe/testing adds a geo-based fake engine for shell
tests and an engine-agnostic contract suite that the fake passes in Node.
check-import-directions keeps the contract free of cesium, three and
@navaramap, and each engine directory free of the other engine.
Conflicts with upstream's Street Level provider composition and share-state
codec, resolved:
- Street View's switch moves from its own share option (`g`) to upstream's
provider-switch field (`r`, e.g. `0.r.google-1`); a link without it keeps
Street View off. Docs and the browser gate follow.
- createApplicationStreetLevel keeps upstream's injectable provider list;
its default, once the application supplies its Street Level source, is
Mapillary plus Google Street View. Without that source there are none, as
upstream intends.
- The coverage hint keeps upstream's empty-composition message and joins
every switched-on provider's hint otherwise.
- CHANGELOG, README and SECURITY keep both sides' entries.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Google's Maps Platform terms forbid Street View imagery beside non-Google
maps, so the provider declares requiresMapStack: 'photoreal': off Google
3D its chip is dashed with the reason, lookups and opens skip it, and
leaving Google 3D closes its panorama.
- Ground clicks: only a click on the map itself (scenePick.isSurfacePick:
nothing picked or Google 3D tile content) at street zoom opens Street
View, and the camera stays put; higher up the hint says to zoom in.
scenePick.pickGroundPosition replaces the copy in Directions.
- Viewer: poses go out at once and again when the date and address
arrive, so an open never hangs on that lookup; one lookup per panorama,
seeded by the nearest lookup; a close settles a pending open; failed
lookups are asked again; poses follow position changes, not pano_changed.
- Loader: the gm_authFailure handler is installed once, so a retried load
reports a refusal once.
- Viewer host: a prewarm that lands after another provider took the host
is released; FIT/FILL availability is set on every open.
- Provider: a refusal while switched off reports no error over another
photo; SINCE compares against the end of Google's month; messages name
the referrer and network causes; FLAT says Street View is 360° only.
- The layer row names the key of the providers that are on; the MCP panel
CSP admits Street View's tiles, control images and photos.
- Tests for each, against fakes that now follow Google's event order; the
browser gate restores its state, waits instead of sleeping, aims its
panel clicks and corner drags, and asserts the camera stays put.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Google Street View registers beside Mapillary as provider `google`. It runs
on the Maps JavaScript API with the browser GOOGLE_MAPS_API_KEY, loaded on
first use (mapsLoader.js), and uses Google's own StreetViewPanorama as its
viewer adapter, emitting provider-neutral poses (address and month in the
caption, a link to Google Maps, steps along Google's arrows, FOLLOW on
Google 3D). Its chip starts off because every panorama is billed to the
key; the share option is `g`.
Google offers no coverage layer for other maps, so the provider draws
none: with it on, a click on bare ground at street zoom opens its nearest
Google-collected outdoor panorama (groundClick.js). A key Google refuses
reads KEY REJECTED and fails a pending open at once; a lookup Google never
answers fails after 10 s.
Core: providers may declare `groundClick` and `defaultOn`; poses may carry
`capturedAtPrecision` and `title`; the coverage line keeps every
provider's hint; opening one provider's image releases another's
prewarmed viewer; FIT/FILL hides for viewers without render modes. CSP
script-src admits https://maps.googleapis.com. The browser gate gains two
Street View steps (the keyed one skips as 'no Google key' in CI).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Fixesbilawalsidhu/gods-eye-view#927. Each stdio MCP server made its own
random panel key, and panel_request accepts only the key in the panel
page. A client that starts two server processes for one connection and
reads the page from one while sending the page's requests to the other
(seen in Claude Desktop on Windows) therefore always got "Only the God's
Eye View panel may make this request".
The stdio entry now passes a key shared by the install, created once in
.gev-cache/mcp-panel-key. The file is staged 0600 and published with
link(), so processes starting together agree on one key and never read a
half-written one. If it cannot be stored, the process keeps a key of its
own, as before. createLocalMcpServer still makes a fresh key unless given
one, so in-process servers (the /mcp HTTP plugin, tests) are unchanged.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- The CI gate builds the app and serves it with `vite preview`, with a
dummy non-secret client token (a fork PR never gets secrets), and
answers the Graph API, image tiles and thumbnails from fixtures: 41
photos on a line under the parked camera and one generated 640x320 JPEG
(scripts/fixtures/street-level). All 7 photo steps now run in CI, plus a
real click on a coverage line and on a cone, the framing flight stopping
on close, the inert modal, the viewer returning into the panel, and a
reload keeping a collapsed panel collapsed (share ui token t).
- Fixture interception is limited to Mapillary hosts via CDP
Fetch.enable: page-wide request interception stalled Cesium's workers,
so the old fixture mode never drew anything on the globe.
- Failure evidence (screenshots, UI state, console, failed requests,
render errors) is written to qa-artifacts/ with keys and tokens redacted
and uploaded on failure; qa-artifacts/ is gitignored.
- Cesium render-loop errors fail the gate; Google 3D branches that cannot
run print as skips, and --strict rejects any other skip.
- Panel presses log their diagnostics, warn on a retry, and CI now fails
on any retry (QA_FAIL_ON_RETRY=1) since the rail settles; a press with
no pointerdown is a failure, not a retry.
- Assertion-gating sleeps 27 -> 1; the rejected-key "no requests" check
has a positive control.
Verified locally: strict fixture gate on the production build 28 ok + 2
"no Google 3D" skips (twice, fail-on-retry on); panel gate 5/5 plus no
render errors; live gate with a real token 25 ok + 3 fixture-only skips.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
From the IC8 review of #768, each with tests that fail before:
- P2: the expanded viewer is a real modal: while it is on screen every
other <body> child is inert (only the ones it set are restored, on
shrink, close and destroy, and when Clean View/recording/cockpit hide
it). The voice pill and toasts are not clickable while it is expanded.
- P2: the tile proxy follows redirects by hand, at most 3 hops, each one
HTTPS on exactly https://tiles.mapillary.com without credentials;
anything else is a 502 and the token never leaves that origin.
- P2: the real dev and preview server smoke tests request
/api/mapillary/status and a tile, and the browser gate first checks the
server's real status route before any interception, so CI fails if the
Mapillary plugin is not registered.
- P2: both gates press panel headers through one shared helper
(scripts/qa-panelDrag.mjs): re-resolve the header until it is under the
pointer, record what the press hit, wait on the real floating/docked
state, and retry only a press that provably missed the header.
Verified: unit 5910 pass; fixture gate 17/17; panel gate 5/5; live gate
with a real Mapillary token 22/22.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A duplicate scan of every module the PR adds (repeated 4-line blocks,
normalised) after gekh's dedup round left these:
- glyphs.js: the cone, ring and position glyphs share cachedGlyph(), wedge(),
ring() and dot(). Checked with a recording canvas: all 20 draw calls keep
the same path and fill/stroke state, so the glyphs are pixel-identical.
- panelPositionControls.js: one placeAt() for the three places that pin a
panel's top-left corner as a fixed window.
- tiles.js: one warn() for the three background-failure logs.
- Tests: src/testSupport/streetLevelFakes.mjs holds the provider snapshot
(uiState + presentation tests), the stand-in provider (core + controls
tests) and one real-ray pinhole camera (view + coverage tests, which had
each written their own). The coverage line-of-sight test still fails when
refresh() stops passing its focus.
- qa-street-level.mjs: one waitForCoverage() for the five copies of the
"on, drawn, settled, not key-gated" wait.
No behaviour change: unit 5862 pass, fixture gate 16/16, panel gate 5/5.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
From gekh's review of #768, each bug with a test that fails before:
- P2: a failed sequence load clears the previous sequence's cones, resets
the selection and can be retried; a second click while loading starts no
second lookup. sequences.js had no tests: sequences.test.mjs now covers
failure and retry, a superseded late answer, the cache and clearSelection.
- P2: closing a photo stops MapillaryJS sequence playback, which kept the
hidden viewer stepping and downloading images.
- openNearest without the panel reports it instead of loading forever, and
stops if the layer (or its provider) went off during the lookup; the
mesh sampler's misses are bounded and forgotten with its samples; the
sequence count is the union of ids across tiles; a provider withdraws
only its own error, so a sequence error cannot mask KEY REJECTED; hover
stays on (throttled) under an orbiting camera and re-picks when it
stops; "since N days" follows the clock in a long-open tab.
- A prewarm waits for one in flight and then builds if it gave up.
- New coverage for the layer core (credits, KEY REJECTED vs REQUIRED,
switching a provider off), the refresh wiring of the tilted-view fix and
the draped-to-cast swap completion.
- Cleanup: loading is derived from pending; freshStreet(); tile ranking
computes distances once and shares one path for the date line; shared
metresBetween / whenIdle / redrawTile; one NEAREST_RADIUS_M and one
source method list; unused contract fields, exports, getImage and the
decoder's image branch removed; PROVIDER_COLORS lists only Mapillary.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
From the IC8 review of #768 (P2s), each with a test that fails before:
- Switching the layer off and on during the first photo download no longer
leaves the viewer stuck on "not mounted": an outdated mount only unmounts
an adapter nobody else holds and subscribes poses only once current, and
the MapillaryJS adapter never builds a viewer after it was unmounted.
- Hover picks at most every 120 ms (the CCTV convention), not while the
camera moves or a button is held, and not after the layer went off.
- Esc shrinks the expanded viewer wherever focus is (document capture), so
the map keeps its selection, and Tab from outside returns into it.
Closing a photo moves focus to the panel instead of dropping it on body.
- Portable panels lift only after a mouse or pen moves past the drag
threshold: a click on a docked panel's edge or a touch swipe on its
header no longer turns it into a floating window, and pointercancel
saves nothing and puts the panel back.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
From the test-engineering review of #768: the rendering and panel code had
thin unit coverage (coverage.js 29% of lines, sequences/viewer/selection
~30%, cameraFollow 50%, streetLevelControls only source-text checks), and
two shipped bugs were caught only by the manual browser gate.
- coverage.test.mjs drives the real createCoverage against a fake scene and
a source whose tile requests the test resolves. It reproduces gekh's P0 #2:
a request superseded by retire() finishes after a newer one for the same
tile. Without the ownership checks it drops the newer request's key and a
refresh starts a duplicate whose lines nothing can remove (fails on the old
loadTile, passes now). It also covers the draped-to-cast swap keeping both
line sets on the globe until the cast one is built and clear() removing
both mid-swap, and a tile with nothing castable not being rebuilt.
- coverageFixture.mjs encodes Mapillary-shaped coverage tiles (sequence
lines, overview points) from lon/lat; the tests and the hermetic gate use it.
- cameraFollow.test.mjs pins the underground-camera fix: a -14,886 m sample
is ignored for bare earth, a plausible sample wins, and following stands at
eye height. Three of the four fail on the pre-fix module (the first reports
"centre at -14884 m").
- streetLevelControls.test.mjs is a contract test through the real
DataLayerManager and LayerStateCoordinator: a 360 deg click and a SINCE
release reach the durable layer state and the encoded share link under
Street Level's token. Against a stand-in layer it also covers the chip rule
(darkening the last lit chip turns the layer off, provider kept on; lighting
a dark chip is a user params request, then enable), the legend rebuilding
on a same-count swap (review #9) and the viewer resizing only on open
(review #8). With the old direct-to-layer routing, three of these fail,
including both share-link tests.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Norway's ~850 working road cameras load from one keyless GeoJSON request
to the agency's OGC API view of its DATEX 3.1 CCTV table, so no DATEX
account is needed. Faulty cameras are dropped, every camera sits under a
"Norway" category, and the ~135 that publish HLS play as live video
through the existing in-memory relay with the still as fallback. Frames
and manifests are pinned to each camera's own path on the agency hosts.
CCTV_VEGVESEN_ENABLED=0 disables the pack, CCTV_VEGVESEN_MAX_SOURCES
caps it (nearest the largest cities first) and CCTV_VEGVESEN_VIDEO=0
keeps stills only. Attributed under NLOD 2.0.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Brings in 123 commits from main, including the v0.2.0 and v0.2.1 releases
(MCP server and tool catalog, server security hardening, build hygiene).
Conflicts:
- package.json: kept main's script order and added this PR's
qa:panel-resize and qa:street-level; exports keep both sides (main's
tools/view entries and this PR's street-level and Mapillary entries).
- src/tooling/viteBuild.test.mjs: kept this PR's Mapillary-token define
assertion and main's new wildcard-host allowedHosts assertions.
- CHANGELOG.md: main now has release sections; this PR's entries go under
[Unreleased].
Checks after the merge: doctor, format:check, check:boundaries,
layer-token:check (28 published, 1 new: street-level 0), npm test (5726
pass), build, the Windows job's focused tests (70 pass), qa:panel-resize,
and the keyless street-level gate. The new CSP allows Mapillary (https:
connect/img, blob workers). The keyed gate stops at coverage locally
because the dev .env Mapillary token is invalid (upstream 403), not because
of the merge.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Binding HOST=0.0.0.0 no longer sets Vite's allowedHosts to true. The dev and
preview servers accept IP addresses, localhost and the LAN hostnames listed
in GEV_ALLOWED_HOSTS; suffix and wildcard entries are ignored. The
configuration factory takes the resolved list as an input, so build/vite.js
still reads no environment.
Ported from #97 to the current build/vite.js layout. Fixes#21.
Co-authored-by: Sameh Khamis <[email protected]>
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The OpenAI endpoints (/api/realtime/token, /api/openai/hud-summary) now
allow 30 requests per minute per client IP and the Google Places endpoints
120, the caps the Pinokio build already ships. GEV_RATELIMIT_OPENAI_PER_MIN
and GEV_RATELIMIT_GOOGLE_PER_MIN still override them and exactly 0 disables
them; an unreadable value falls back to the default rather than to
unlimited.
Salvaged from #683 onto current main.
Co-authored-by: Sameh Khamis <[email protected]>
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The cost-bearing endpoints (/api/realtime/token, /api/openai/hud-summary,
/api/google/nearby-places) and the voice debug log refuse cross-site browser
requests: a foreign or opaque Origin, a Sec-Fetch-Site other than
same-origin/none, or proxy forwarding headers. Loopback tools without an
Origin and the LAN opt-in keep working. The dev and preview servers send one
shared Content-Security-Policy, and embed-mode framing rewrites only the
frame-ancestors directive of it.
Salvaged from #242 onto current main.
Co-authored-by: Sameh Khamis <[email protected]>
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
show_on_globe replaces open_in_gods_eye_view and names an MCP Apps view, a
ui:// panel that loads the app in embed mode and posts each later view to it,
so the globe in the conversation changes without reloading. It also takes a
view another answer returned. Tools can declare a UI resource, and the MCP
adapter serves resources/list and resources/read and lists each tool's
resource in _meta.ui.resourceUri.
?embed=1 shows the globe alone: clean view with the HUD, panels, welcome and
setup prompts hidden, and provider attribution kept. The page that frames it
sends views with postMessage; the app applies each through its own actions
and reports every step. Annotations in any link are drawn once it restores.
Only embed-mode documents may be framed; every other page keeps
X-Frame-Options DENY and frame-ancestors 'none', and
GEV_EMBED_FRAME_ANCESTORS can restrict who frames them.
Tool answers with something to show now include data.view: the view with
the matching layers on, the area framed from above or a single aircraft or
satellite followed, and the link that opens it. Area framing moves to
src/tools/views.js so links and answers frame alike.
A view is a camera, data layers, visual style, map imagery and an aircraft,
military aircraft or satellite to follow, written in the share-link format
the app already restores. gods-eye-view/view builds, encodes and decodes one,
and share links take their style names from it. open_in_gods_eye_view now
builds its link from a full view: an area or camera, layers, style, map and
something to follow.
Brief sections received a named place as its bounding box, so they centered
on the box's middle: Tokyo's aircraft were sorted from a point in the
Pacific. Sections now receive the place name and resolve it like a direct
query, and place lookups are cached briefly so a brief resolves it once.
The raycast challenge on #768 showed bare-earth casting misplacing roads
that are not at grade: lines floated over freeways in trenches and steep
streets (rendered line pixels floating: Austin 8.5%, SF 27.1%, LA 43.8%),
because a ~111 m terrain grid cannot see a 30 m-wide cut or a slope between
its corners.
meshSampler.js samples the rendered Google 3D surface (scene.sampleHeight,
overlays excluded) once per ~11 m cell within 900 m of the camera, nearest
first, in idle time within a 6 ms budget; misses are retried, not latched.
groundCast.refineHeights blends it in, per point along a line:
- mesh up to 3 m above bare earth is the road: follow it, 1 m above;
- higher mesh (canopy, deck, roof): carry the road's offset from bare earth
across from the road samples on both sides (up to 8 points);
- otherwise bare earth + 2 m, as before.
Lines, cones and the marker use it; tiles touched by new samples are
redrawn one per idle slice with the no-blink swap.
Rendered line pixels on the surface / floating, before -> after:
Austin 90.0 -> 95.3% / 8.5 -> 0.7%; SF 69.0 -> 86.4% / 27.1 -> 9.1%;
LA 52.3 -> 91.0% / 43.8 -> 4.0%; Innsbruck 95.0 -> 95.7% / 0.2 -> 0.2%;
Sacramento 80.7 -> 79.3% / 3.8 -> 2.7%; buried lines stay under 1%.
Cones floating in SF 44 -> 5%, buried at LA 59 -> 28%. Main-thread long
tasks added during refinement: at most 134 ms (Austin, densest tile).
Roads on elevated decks are still drawn at ground level under the deck.
Unit tests cover the blend (trench, slope, canopy bridge, deck, bad probe),
the finer cast with meshAt, and the sampler (once per cell, nearest first,
overlays excluded, far cells skipped, misses not latched, idle when off).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
From the code review of #768:
- view.js measured the camera's height above ground with globe.getHeight,
but Google 3D hides the globe, so coverage picked its zoom from the
ellipsoidal height. Over a high city (Denver, ~1,600 m) a camera 500 m up
read 2,100 m: coverage stayed at z12/z13, whose tiles are too big to cast,
and lines stayed draped on the roofs. With the globe hidden, the bare-earth
height from the ground caster now answers; coverage re-picks its zoom when
the surface mode changes. Verified in Denver: terrain mode, z14, cast lines.
The core's surface check uses the same helper (and no longer subtracts the
2 m lift from the camera height).
- castTile rebuilt a tile even when no line could be cast (terrain proxy
down, tile too big); it now leaves such tiles as they are.
- The feature gate read Street Level's share token as a literal `0`; it now
takes it from LAYER_STATE_REGISTRY, since a rebase can reassign it.
New view.test.mjs covers the shown-globe, hidden-globe and no-sample cases.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The merge of main (c451478) appended main's new entries directly after
this branch's last entries in scripts/format-scope.json (after
streetLevel/index.test.mjs) and in two groups of
scripts/package-boundaries.json (after app/layers/streetLevel.js), without
the separating comma. Both files stopped parsing, so every CI job failed
at its first JSON read, the "Check adopted formatting" step.
With the commas back, the CI steps pass locally: doctor, format:check,
check:boundaries, layer-token:check against main, npm test (5475 pass),
build, and the Windows job's focused onboarding tests.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
src/tools/surfaces.js lists the tools a surface leaves out, and the MCP
server, the voice session's tool list and the browser catalog voice calls
all filter through it. Voice leaves out tools that answer with images, link
to the app, or repeat what its app actions already answer.
German Khokhlov