Merge branch 'main' into feat/cctv-vegvesen-kartverket-heights

This commit is contained in:
Said Turksever authored and GitHub committed 2026-10-19 23:15:41 +01:00
commit 3744a97f00
83 files changed
+17445 -385

No files matched your search

+5
View File
@@ -0,0 +1,5 @@
- Color CCTV cameras with a video feed violet. Cameras with an HLS, MP4 or
WebM feed were the same cyan as stills, so a DelDOT or Statens vegvesen
video camera could only be found by opening it. Stills stay cyan and the
active camera stays amber, and focus de-emphasis fades the violet the same
way (Danielslay86, #1200).
+4
View File
@@ -0,0 +1,4 @@
- Draw CCTV stills at their source aspect on the monitor plane instead of
stretching them to the 16:9 canvas. Most DOT cameras publish 4:3 frames, so
every still was widened; they are now centered and pillarboxed on the dark
the placeholder already ends on (Danielslay86, #1201).
+3
View File
@@ -0,0 +1,3 @@
- Avoid starting or queuing a reply when the local channel rejects a typed OpenAI
command, while preserving an earlier accepted deferred input
(Manjunath, #1220).
@@ -0,0 +1,5 @@
- Serve the camera catalog while it refreshes. An expired catalog made every
camera request wait for the slowest pack to reload; it is now answered at
once while one background refresh runs, so only the first load waits. Estonia's
slow Tarktee feeds get 30 s instead of 15 s, so the pack no longer drops out
on about a third of loads (quanticsoul4772, #1234).
@@ -0,0 +1,7 @@
- Close the compact Radio disclosure when an adaptive tactical right-rail
restore directly expands Context, before Space Missions can be covered.
A repeated Context-open request also clears stale disclosure state.
Real PanelChrome constructor wiring and the rail reflow path are covered
by focused regressions. Credits: kvnloo (#816), raysung-2026's original
#804 report, asturksever's review, and quanticsoul4772's independent
Windows diagnosis.
+4
View File
@@ -0,0 +1,4 @@
- Add optional Gemini Live voice to the standalone app, with a provider chooser,
server-held API key, microphone streaming, shared map actions and captions.
Switching providers ends the old conversation and preserves the input mode
(Manjunath, #872).
+6
View File
@@ -0,0 +1,6 @@
- Extract a reusable model-free World Packet from the existing voice
`get_current_view_state` owner without restoring disabled layers to the
model context. Keep the enabled-only payload, disabled-layer count, source
provenance, action result and view-state contract; read the layer roster once
per packet; voice retains the speech/card envelope at its own boundary.
Credits: original GEV voice/layer contributors and #880 implementation.