@livedesk/client 0.1.91 → 0.1.93

package/README.md

@@ -39,8 +39,11 @@ the matching callback URL to Supabase Auth redirect URLs.
 The client registers the device, sends status heartbeats, can return
 Hub-requested thumbnails, can stream a focused view-only live screen, and
 can receive safe task-only instructions. The C# RemoteFast engine also supports
-keyboard/mouse control, file transfer, and Windows system-audio loopback
-playback. When file transfer is enabled by the Hub, received files are saved
+keyboard/mouse control on Windows, macOS, and Linux/X11, file transfer, and
+Windows system-audio loopback playback. macOS control requires Accessibility
+permission for the terminal app that started LiveDesk; screen streaming requires
+Screen Recording permission. LiveDesk opens the matching System Settings page
+only when that permission is first needed. When file transfer is enabled by the Hub, received files are saved
 to `Desktop/LiveDeskFiles` unless the Hub sets another destination folder.
 
 By default, the launcher uses the packaged C# RemoteFast engine when supported.
@@ -70,7 +73,7 @@ Frame pipeline roadmap:
 
 - Mode 1: `mode1-jpeg` - current test path using screen capture, resize, and JPEG binary frames.
 - Mode 2: `mode2-lz4` - Windows Fast path using raw BGRA frames compressed as an LZ4 block.
-- Mode 3: `mode3-h264-hw` - OS-specific ffmpeg hardware H.264 path. Windows tries Media Foundation/NVENC/QSV/AMF, macOS tries VideoToolbox, and Linux tries NVENC/VAAPI/QSV. The launcher uses the bundled ffmpeg first, unless `LIVEDESK_FFMPEG` points to a custom binary. On macOS, set `LIVEDESK_FFMPEG_AVFOUNDATION_INPUT` if your screen input is not `1:none`.
+- Mode 3: `mode3-h264-hw` - OS-specific hardware H.264 path. Windows tries Media Foundation/NVENC/QSV/AMF, macOS prefers ScreenCaptureKit plus VideoToolbox, and Linux tries NVENC/VAAPI/QSV. macOS emits one startup key frame and then an approximately one-second GOP. The launcher uses bundled ffmpeg for fallback paths unless `LIVEDESK_FFMPEG` points to a custom binary. On macOS, set `LIVEDESK_FFMPEG_AVFOUNDATION_INPUT` only when the ScreenCaptureKit helper is unavailable and the AVFoundation fallback input is not `1:none`.
 
 Legacy mode names such as `remote-fast` and `remote-quality` are treated as
 Mode 1 aliases.

package/fast/osx-arm64/MindExec.RemoteAgent.Fast/MacScreenCaptureKitH264.swift

@@ -37,7 +37,16 @@ final class LiveDeskScreenCaptureKitEncoder: NSObject, SCStreamOutput, SCStreamD
             throw LiveDeskSCKError.noDisplays
         }
 
-        let display = shareableContent.displays[min(displayIndex, shareableContent.displays.count - 1)]
+        let mainDisplayID = CGMainDisplayID()
+        let orderedDisplays = shareableContent.displays.sorted { left, right in
+            let leftIsMain = left.displayID == mainDisplayID
+            let rightIsMain = right.displayID == mainDisplayID
+            if leftIsMain != rightIsMain {
+                return leftIsMain
+            }
+            return left.displayID < right.displayID
+        }
+        let display = orderedDisplays[min(displayIndex, orderedDisplays.count - 1)]
         let filter = SCContentFilter(display: display, excludingWindows: [])
         let configuration = SCStreamConfiguration()
         configuration.width = width
@@ -79,8 +88,8 @@ final class LiveDeskScreenCaptureKitEncoder: NSObject, SCStreamOutput, SCStreamD
         VTSessionSetProperty(session, key: kVTCompressionPropertyKey_RealTime, value: kCFBooleanTrue)
         VTSessionSetProperty(session, key: kVTCompressionPropertyKey_AllowFrameReordering, value: kCFBooleanFalse)
         VTSessionSetProperty(session, key: kVTCompressionPropertyKey_ProfileLevel, value: kVTProfileLevel_H264_Baseline_AutoLevel)
-        VTSessionSetProperty(session, key: kVTCompressionPropertyKey_MaxKeyFrameInterval, value: NSNumber(value: 1))
-        VTSessionSetProperty(session, key: kVTCompressionPropertyKey_MaxKeyFrameIntervalDuration, value: NSNumber(value: max(0.001, 1.0 / Double(fps))))
+        VTSessionSetProperty(session, key: kVTCompressionPropertyKey_MaxKeyFrameInterval, value: NSNumber(value: fps))
+        VTSessionSetProperty(session, key: kVTCompressionPropertyKey_MaxKeyFrameIntervalDuration, value: NSNumber(value: 1.0))
         VTSessionSetProperty(session, key: kVTCompressionPropertyKey_ExpectedFrameRate, value: NSNumber(value: fps))
         VTSessionSetProperty(session, key: kVTCompressionPropertyKey_AverageBitRate, value: NSNumber(value: bitrateKbps * 1000))
         VTCompressionSessionPrepareToEncodeFrames(session)
@@ -111,8 +120,12 @@ final class LiveDeskScreenCaptureKitEncoder: NSObject, SCStreamOutput, SCStreamD
         frameCount += 1
         let pts = CMSampleBufferGetPresentationTimeStamp(sampleBuffer)
         let duration = CMTime(value: 1, timescale: CMTimeScale(fps))
-        let frameProperties = NSMutableDictionary()
-        frameProperties[kVTEncodeFrameOptionKey_ForceKeyFrame] = kCFBooleanTrue
+        var frameProperties: CFDictionary?
+        if frameCount == 1 {
+            let properties = NSMutableDictionary()
+            properties[kVTEncodeFrameOptionKey_ForceKeyFrame] = kCFBooleanTrue
+            frameProperties = properties
+        }
 
         let status = VTCompressionSessionEncodeFrame(
             session,

package/fast/osx-x64/MindExec.RemoteAgent.Fast/MacScreenCaptureKitH264.swift

@@ -37,7 +37,16 @@ final class LiveDeskScreenCaptureKitEncoder: NSObject, SCStreamOutput, SCStreamD
             throw LiveDeskSCKError.noDisplays
         }
 
-        let display = shareableContent.displays[min(displayIndex, shareableContent.displays.count - 1)]
+        let mainDisplayID = CGMainDisplayID()
+        let orderedDisplays = shareableContent.displays.sorted { left, right in
+            let leftIsMain = left.displayID == mainDisplayID
+            let rightIsMain = right.displayID == mainDisplayID
+            if leftIsMain != rightIsMain {
+                return leftIsMain
+            }
+            return left.displayID < right.displayID
+        }
+        let display = orderedDisplays[min(displayIndex, orderedDisplays.count - 1)]
         let filter = SCContentFilter(display: display, excludingWindows: [])
         let configuration = SCStreamConfiguration()
         configuration.width = width
@@ -79,8 +88,8 @@ final class LiveDeskScreenCaptureKitEncoder: NSObject, SCStreamOutput, SCStreamD
         VTSessionSetProperty(session, key: kVTCompressionPropertyKey_RealTime, value: kCFBooleanTrue)
         VTSessionSetProperty(session, key: kVTCompressionPropertyKey_AllowFrameReordering, value: kCFBooleanFalse)
         VTSessionSetProperty(session, key: kVTCompressionPropertyKey_ProfileLevel, value: kVTProfileLevel_H264_Baseline_AutoLevel)
-        VTSessionSetProperty(session, key: kVTCompressionPropertyKey_MaxKeyFrameInterval, value: NSNumber(value: 1))
-        VTSessionSetProperty(session, key: kVTCompressionPropertyKey_MaxKeyFrameIntervalDuration, value: NSNumber(value: max(0.001, 1.0 / Double(fps))))
+        VTSessionSetProperty(session, key: kVTCompressionPropertyKey_MaxKeyFrameInterval, value: NSNumber(value: fps))
+        VTSessionSetProperty(session, key: kVTCompressionPropertyKey_MaxKeyFrameIntervalDuration, value: NSNumber(value: 1.0))
         VTSessionSetProperty(session, key: kVTCompressionPropertyKey_ExpectedFrameRate, value: NSNumber(value: fps))
         VTSessionSetProperty(session, key: kVTCompressionPropertyKey_AverageBitRate, value: NSNumber(value: bitrateKbps * 1000))
         VTCompressionSessionPrepareToEncodeFrames(session)
@@ -111,8 +120,12 @@ final class LiveDeskScreenCaptureKitEncoder: NSObject, SCStreamOutput, SCStreamD
         frameCount += 1
         let pts = CMSampleBufferGetPresentationTimeStamp(sampleBuffer)
         let duration = CMTime(value: 1, timescale: CMTimeScale(fps))
-        let frameProperties = NSMutableDictionary()
-        frameProperties[kVTEncodeFrameOptionKey_ForceKeyFrame] = kCFBooleanTrue
+        var frameProperties: CFDictionary?
+        if frameCount == 1 {
+            let properties = NSMutableDictionary()
+            properties[kVTEncodeFrameOptionKey_ForceKeyFrame] = kCFBooleanTrue
+            frameProperties = properties
+        }
 
         let status = VTCompressionSessionEncodeFrame(
             session,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@livedesk/client",
-  "version": "0.1.91",
+  "version": "0.1.93",
   "description": "LiveDesk local remote client",
   "type": "module",
   "bin": {