@momasiku/dsh-pilot 0.5.17 → 0.6.0

package/lib/scripts/_dsh-capture.ps1

@@ -1,4 +1,4 @@
-# pilot / _dsh-capture.ps1 - one capture, executed in whatever process called it.
+# pilot / _dsh-capture.ps1 - one capture, executed in whatever process called it.
 #
 # Extracted from desktop-probe.ps1 so the same code serves two callers:
 #   * desktop-probe.ps1  - one frame per process (the fallback path);

package/lib/scripts/_dsh-overlay.ps1

@@ -0,0 +1,301 @@
+
+# pilot / _dsh-overlay.ps1 - the visible part of a takeover.
+#
+# Ring around every monitor, a caption naming the operator, and a halo that follows
+# the pointer so it is obvious the movement is not yours. Esc stamps the cancel token
+# the host published.
+#
+# The halo is DRAWING, not the pointer itself: this process never calls SetCursorPos
+# and never installs an input hook, because either would add exactly the input lag
+# that makes a person's own mouse feel broken while an agent drives. Windows draws
+# the real cursor above every window, so the ring and glow sit under it and the
+# pointer reads as larger and lit without being replaced.
+param(
+    [Parameter(Mandatory = $true)][string] $StateDir
+)
+$ErrorActionPreference = 'Stop'
+
+Add-Type -AssemblyName PresentationFramework, PresentationCore, WindowsBase
+Add-Type -TypeDefinition @"
+using System;
+using System.Runtime.InteropServices;
+public class DshOv {
+    [StructLayout(LayoutKind.Sequential)] public struct POINT { public int X; public int Y; }
+    [DllImport("user32.dll")] public static extern bool GetCursorPos(out POINT p);
+    [DllImport("user32.dll")] public static extern int GetWindowLong(IntPtr h, int i);
+    [DllImport("user32.dll")] public static extern int SetWindowLong(IntPtr h, int i, int v);
+    [DllImport("user32.dll")] public static extern bool RegisterHotKey(IntPtr h, int id, uint mod, uint vk);
+    [DllImport("user32.dll")] public static extern bool UnregisterHotKey(IntPtr h, int id);
+    public const int GWL_EXSTYLE = -20;
+    public static void ClickThrough(IntPtr h) {
+        int ex = GetWindowLong(h, GWL_EXSTYLE);
+        SetWindowLong(h, GWL_EXSTYLE, ex | 0x20 | 0x80 | 0x08000000);
+    }
+}
+"@
+
+$showFile = Join-Path $StateDir 'show.txt'
+$quitFile = Join-Path $StateDir 'quit.txt'
+$hitFile  = Join-Path $StateDir 'esc.txt'
+$logFile  = Join-Path $StateDir 'status.txt'
+if (-not (Test-Path $StateDir)) { New-Item -ItemType Directory -Force -Path $StateDir | Out-Null }
+function Log([string] $m) {
+    try { [System.IO.File]::AppendAllText($logFile, ((Get-Date).ToString('HH:mm:ss.fff') + ' ' + $m + [Environment]::NewLine)) } catch { }
+}
+
+$global:DshOvVsL = [System.Windows.SystemParameters]::VirtualScreenLeft
+$global:DshOvVsT = [System.Windows.SystemParameters]::VirtualScreenTop
+$vw = [System.Windows.SystemParameters]::VirtualScreenWidth
+$vh = [System.Windows.SystemParameters]::VirtualScreenHeight
+$unit = [Math]::Min($vw / 1920.0, $vh / 1080.0)
+if ($unit -lt 0.75) { $unit = 0.75 }
+
+$coreThickness = [Math]::Max(6, [Math]::Round(8 * $unit))
+$glowThickness = [Math]::Max(18, [Math]::Round(36 * $unit))
+$blurRadius    = [Math]::Max(18, [Math]::Round(34 * $unit))
+$radius        = [Math]::Round(22 * $unit)
+$fontSize      = [Math]::Max(30, [Math]::Round(40 * $unit))
+$spotD         = [Math]::Round(120 * $unit)
+$global:DshOvSpotD = $spotD
+$haloD         = [Math]::Round(58 * $unit)
+
+function SB([int] $a, [int] $r, [int] $g, [int] $b) {
+    New-Object System.Windows.Media.SolidColorBrush ([System.Windows.Media.Color]::FromArgb($a, $r, $g, $b))
+}
+
+$window = New-Object System.Windows.Window
+$window.WindowStyle = 'None'
+$window.AllowsTransparency = $true
+$window.Background = [System.Windows.Media.Brushes]::Transparent
+$window.Topmost = $true
+$window.ShowInTaskbar = $false
+$window.ShowActivated = $false
+$window.ResizeMode = 'NoResize'
+$window.Left = $global:DshOvVsL
+$window.Top = $global:DshOvVsT
+$window.Width = $vw
+$window.Height = $vh
+
+$root = New-Object System.Windows.Controls.Grid
+
+# ---- halo that follows the pointer ---------------------------------------
+# Its own window, not a child of the ring window: a child Grid with runtime
+# Margin never painted on this layered surface, while a window that is simply
+# moved does. Moving a window is also cheaper than re-laying out a full-screen
+# visual tree 30 times a second.
+$haloWin = New-Object System.Windows.Window
+$haloWin.WindowStyle = 'None'
+$haloWin.AllowsTransparency = $true
+$haloWin.Background = [System.Windows.Media.Brushes]::Transparent
+$haloWin.Topmost = $true
+$haloWin.ShowInTaskbar = $false
+$haloWin.ShowActivated = $false
+$haloWin.ResizeMode = 'NoResize'
+$haloWin.Width = $spotD
+$haloWin.Height = $spotD
+$haloWin.Left = -10000
+$haloWin.Top = -10000
+
+$halo = New-Object System.Windows.Controls.Grid
+$spot = New-Object System.Windows.Shapes.Ellipse
+$spot.Width = $spotD
+$spot.Height = $spotD
+$spotFill = New-Object System.Windows.Media.RadialGradientBrush
+$stop1 = New-Object System.Windows.Media.GradientStop ([System.Windows.Media.Color]::FromArgb(155, 70, 150, 255)), 0.0
+$stop2 = New-Object System.Windows.Media.GradientStop ([System.Windows.Media.Color]::FromArgb(70, 70, 150, 255)), 0.55
+$stop3 = New-Object System.Windows.Media.GradientStop ([System.Windows.Media.Color]::FromArgb(0, 70, 150, 255)), 1.0
+$spotFill.GradientStops.Add($stop1) | Out-Null
+$spotFill.GradientStops.Add($stop2) | Out-Null
+$spotFill.GradientStops.Add($stop3) | Out-Null
+$spot.Fill = $spotFill
+$halo.Children.Add($spot) | Out-Null
+
+$haloRing = New-Object System.Windows.Shapes.Ellipse
+$haloRing.Width = $haloD
+$haloRing.Height = $haloD
+$haloRing.Stroke = SB 235 236 246 255
+$haloRing.StrokeThickness = [Math]::Max(2, [Math]::Round(3 * $unit))
+$haloRing.HorizontalAlignment = 'Center'
+$haloRing.VerticalAlignment = 'Center'
+$halo.Children.Add($haloRing) | Out-Null
+
+$haloWin.Content = $halo
+$haloWin.Add_SourceInitialized({
+    # This window sits exactly where the pointer is. Without WS_EX_TRANSPARENT it
+    # eats the very click it is drawn around.
+    $hh = (New-Object System.Windows.Interop.WindowInteropHelper($haloWin)).Handle
+    [DshOv]::ClickThrough($hh) | Out-Null
+}.GetNewClosure())
+$global:DshOvHaloWin = $haloWin
+
+
+# ---- caption -------------------------------------------------------------
+$sans = New-Object System.Windows.Media.FontFamily('Microsoft YaHei UI, Microsoft YaHei, Segoe UI Variable Display, Segoe UI')
+$label = New-Object System.Windows.Controls.TextBlock
+# Built from UTF-8 bytes on purpose: this file lost its BOM once and PowerShell
+# 5.1 then read the literal as GBK, turning the caption into mojibake. Bytes cannot
+# be mis-decoded.
+$captionBytes = [byte[]]@(0x44,0x53,0x48,0x20,0xE6,0xAD,0xA3,0xE5,0x9C,0xA8,0xE6,0x93,0x8D,0xE4,0xBD,0x9C,0xE4,0xBD,0xA0,0xE7,0x9A,0x84,0xE7,0x94,0xB5,0xE8,0x84,0x91,0x20,0x20,0x20,0xE6,0x8C,0x89,0x20,0x45,0x73,0x63,0x20,0xE5,0x8F,0x96,0xE6,0xB6,0x88)
+$label.Text = [System.Text.Encoding]::UTF8.GetString($captionBytes)
+$label.FontSize = $fontSize
+$label.FontFamily = $sans
+$label.FontWeight = [System.Windows.FontWeights]::Bold
+$label.Foreground = [System.Windows.Media.Brushes]::White
+$label.Opacity = 0.9
+$label.HorizontalAlignment = 'Center'
+$label.VerticalAlignment = 'Top'
+$label.Margin = [System.Windows.Thickness]::new(0, [Math]::Round(34 * $unit), 0, 0)
+$capShadow = New-Object System.Windows.Media.Effects.DropShadowEffect
+$capShadow.Color = [System.Windows.Media.Colors]::Black
+$capShadow.BlurRadius = [Math]::Round(26 * $unit)
+$capShadow.ShadowDepth = 3
+$capShadow.Opacity = 0.9
+$label.Effect = $capShadow
+$root.Children.Add($label) | Out-Null
+
+$window.Content = $root
+
+$global:DshOvHwnd = [IntPtr]::Zero
+$global:DshOvCancelTarget = ''
+$global:DshOvStoppedFor = ''
+$global:DshOvHotkeyOn = $false
+
+function Enable-Hotkey {
+    if ($global:DshOvHotkeyOn -or $global:DshOvHwnd -eq [IntPtr]::Zero) { return }
+    $global:DshOvHotkeyOn = [DshOv]::RegisterHotKey($global:DshOvHwnd, 1, 0, 0x1B)
+}
+function Disable-Hotkey {
+    if (-not $global:DshOvHotkeyOn -or $global:DshOvHwnd -eq [IntPtr]::Zero) { return }
+    [DshOv]::UnregisterHotKey($global:DshOvHwnd, 1) | Out-Null
+    $global:DshOvHotkeyOn = $false
+}
+
+function Stop-Takeover {
+    $global:DshOvStoppedFor = $global:DshOvCancelTarget
+    if ($global:DshOvCancelTarget -ne '') {
+        try { [System.IO.File]::WriteAllText($global:DshOvCancelTarget, 'stop', [System.Text.Encoding]::UTF8) } catch { }
+    }
+    try { [System.IO.File]::WriteAllText($hitFile, (Get-Date).ToString('o'), [System.Text.Encoding]::UTF8) } catch { }
+    if ($window.IsVisible) { $window.Hide(); Disable-Hotkey }
+    $global:DshOvCancelTarget = ''
+}
+
+$script:haloErrors = 0
+function Update-Halo {
+    # No latching. A failed GetCursorPos (the desktop can be mid-switch between a
+    # locked and an active session) used to disable the halo for the rest of the
+    # process - which is exactly "it worked at first and then stopped".
+    try {
+        # The nested struct has to be named, not written as DshOv+POINT: unquoted,
+        # PowerShell reads the plus as arithmetic and the call throws on every tick.
+        $p = [DshOv+POINT]::new()
+        # A frozen position makes the halo verifiable without racing a human hand.
+        $freeze = Join-Path $StateDir 'freeze.txt'
+        $fx = $null
+        if (Test-Path -LiteralPath $freeze) {
+            try {
+                $parts = ([System.IO.File]::ReadAllText($freeze)).Trim() -split ','
+                if ($parts.Count -ge 2) { $fx = @([int] $parts[0], [int] $parts[1]) }
+            } catch { }
+        }
+        $ok = $false
+        if ($null -ne $fx) { $p.X = $fx[0]; $p.Y = $fx[1]; $ok = $true } else { $ok = [DshOv]::GetCursorPos([ref] $p) }
+        if ($ok) {
+            $h = $global:DshOvHaloWin
+            if ($null -eq $h) { if (-not $script:haloLogged) { $script:haloLogged = $true; Log 'halo missing' }; return }
+            $h.Left = $p.X - ($global:DshOvSpotD / 2)
+            $h.Top = $p.Y - ($global:DshOvSpotD / 2)
+            $script:haloMoves = [int] $script:haloMoves + 1
+            if ($script:haloMoves -le 3 -or ($script:haloMoves % 40) -eq 0) {
+                Log ("halo move " + $script:haloMoves + " -> " + $p.X + "," + $p.Y + " margin=" + $h.Margin.Left + "," + $h.Margin.Top)
+            }
+        }
+    } catch {
+        $script:haloErrors++
+        if ($script:haloErrors -le 3 -or ($script:haloErrors % 300) -eq 0) {
+            Log ("halo error " + $script:haloErrors + ": " + $_.Exception.Message)
+        }
+    }
+}
+
+function Update-State {
+    $show = Test-Path -LiteralPath $showFile
+    if ($show) {
+        $raw = ''
+        try { $raw = [System.IO.File]::ReadAllText($showFile, [System.Text.Encoding]::UTF8).Trim() } catch { }
+        if ($raw -eq $global:DshOvStoppedFor) { $show = $false } else { $global:DshOvCancelTarget = $raw }
+    }
+    if ($show) {
+        if (-not $window.IsVisible) {
+            # The desktop can change size between shows (a reconnect, a display mode
+            # switch); re-fit rather than draw a 1080p ring on a 1600p desktop.
+            $w = [System.Windows.SystemParameters]::VirtualScreenWidth
+            $h = [System.Windows.SystemParameters]::VirtualScreenHeight
+            if ($window.Width -ne $w -or $window.Height -ne $h) {
+                $window.Left = [System.Windows.SystemParameters]::VirtualScreenLeft
+                $window.Top = [System.Windows.SystemParameters]::VirtualScreenTop
+                $window.Width = $w
+                $window.Height = $h
+            }
+            $window.Show(); Enable-Hotkey
+            if (-not $global:DshOvHaloWin.IsVisible) { $global:DshOvHaloWin.Show() }
+            Log ("shown " + $w + "x" + $h)
+        }
+        Update-Halo
+    } elseif ($window.IsVisible) {
+        $window.Hide()
+        if ($global:DshOvHaloWin.IsVisible) { $global:DshOvHaloWin.Hide() }
+        Disable-Hotkey
+        $global:DshOvCancelTarget = ''
+    }
+    if (Test-Path -LiteralPath $quitFile) { $window.Close() }
+}
+
+$window.Add_SourceInitialized({
+    $global:DshOvHwnd = (New-Object System.Windows.Interop.WindowInteropHelper($window)).Handle
+    [DshOv]::ClickThrough($global:DshOvHwnd) | Out-Null
+    $source = [System.Windows.Interop.HwndSource]::FromHwnd($global:DshOvHwnd)
+    $source.AddHook([System.Windows.Interop.HwndSourceHook]{
+        param($msgHwnd, $msg, $wParam, $lParam, [ref] $handled)
+        if ($msg -eq 0x0312 -and [int] $wParam -eq 1) { Stop-Takeover; $handled.Value = $true }
+        return [IntPtr]::Zero
+    })
+}.GetNewClosure())
+
+$haloWin.Add_Closed({ })
+$window.Add_Closed({
+    if ($global:DshOvHaloWin.IsVisible) { $global:DshOvHaloWin.Close() }
+    if ($global:DshOvHwnd -ne [IntPtr]::Zero) { [DshOv]::UnregisterHotKey($global:DshOvHwnd, 1) | Out-Null }
+})
+
+# No parent watch: a host that dies leaves no trace to watch, and a false positive
+# kills the ring the moment the launcher is a short-lived shell. An idle timeout
+# bounds the leftover process without ever firing while it is in use.
+$global:DshOvTicks = 0
+$global:DshOvIdleTicks = 0
+
+$timer = New-Object System.Windows.Threading.DispatcherTimer
+$timer.Interval = [TimeSpan]::FromMilliseconds(33)
+$timer.Add_Tick({
+    Update-State
+    $global:DshOvTicks++
+    if ($window.IsVisible) { $global:DshOvIdleTicks = 0 } else { $global:DshOvIdleTicks++ }
+    # 33 ms * 18000 = ~10 minutes hidden before the process gives the memory back.
+    if ($global:DshOvIdleTicks -gt 18000) { $window.Close() }
+})
+$timer.Start()
+
+# Create the handle and strip activation BEFORE the first Show. Applying the
+# extended style from SourceInitialized is too late: Windows has already offered
+# the window the foreground by then, and a topmost window that takes focus breaks
+# the very actions it is drawn for.
+$pre = New-Object System.Windows.Interop.WindowInteropHelper($window)
+$preHwnd = $pre.EnsureHandle()
+[DshOv]::ClickThrough($preHwnd) | Out-Null
+$preHalo = New-Object System.Windows.Interop.WindowInteropHelper($haloWin)
+$preHaloHwnd = $preHalo.EnsureHandle()
+[DshOv]::ClickThrough($preHaloHwnd) | Out-Null
+
+$window.Show()
+$window.Hide()
+[System.Windows.Threading.Dispatcher]::Run()

package/lib/scripts/_dsh-uia.ps1

@@ -1,4 +1,4 @@
-# pilot / _dsh-uia.ps1 - read a window's UI Automation tree, and resolve a node by index.
+# pilot / _dsh-uia.ps1 - read a window's UI Automation tree, and resolve a node by index.
 #
 # Why this exists: screen_view returns pixels. Finding a control in pixels costs the
 # model a guess at coordinates plus another frame to confirm the guess. UI Automation

package/lib/scripts/_dsh-win32.ps1

@@ -1,4 +1,4 @@
-# pilot / _dsh-win32.ps1 - the Win32 bridge and per-monitor DPI awareness, shared
+# pilot / _dsh-win32.ps1 - the Win32 bridge and per-monitor DPI awareness, shared
 # by the one-shot scripts and by the persistent worker.
 #
 # Every block is idempotent: a process that already has the type does not compile

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@momasiku/dsh-pilot",
-  "version": "0.5.17",
+  "version": "0.6.0",
   "description": "Desktop vision and control for DeepSeek Harness - the whole Windows machine, any application, not browser automation: screen_view captures the live screen as an image the model can see, desktop_control drives the mouse, keyboard, and windows against the exact same pixel space; desktop_sequence runs a whole batch of actions in one call and looks once at the end.",
   "license": "MIT",
   "author": "momasiku",