Pup - the AI's own browser

@@ -1486,11 +1486,86 @@ function osSendWindowToBack(profileName) {   } catch (e) { console.error(`[bg] send-to-back failed for "${profileName}": ${e.message}`); return false; } } +// v1.8.24: capture the user's CURRENT foreground window BEFORE we launch a pup+// browser (which grabs foreground). We hand focus back to it right after, so+// opening a pup window NEVER steals focus. Returns a decimal HWND string or null.+function osGetForegroundWindow() {+  if (process.platform !== 'win32') return null;+  try {+    const scriptPath = path.join(require('os').tmpdir(), `_adom_fg_${process.pid}_${Date.now()}.ps1`);+    const ps = [+      'Add-Type @"',+      'using System; using System.Runtime.InteropServices;',+      'public class WinFgGet { [DllImport("user32.dll")] public static extern IntPtr GetForegroundWindow(); }',+      '"@',+      'Write-Output ([int64][WinFgGet]::GetForegroundWindow())',+    ].join('\r\n');+    fs.writeFileSync(scriptPath, ps);+    const out = execSync(`powershell.exe -NoProfile -ExecutionPolicy Bypass -File "${scriptPath}"`, { encoding: 'utf8', timeout: 4000 }).trim();+    return out && out !== '0' ? out : null;+  } catch (e) { return null; }+}++// v1.8.24: THE real "background by default" enforcement. osSendWindowToBack only+// lowers z-order — but a freshly-launched --start-maximized browser KEEPS foreground+// focus, so the window still sits on top of the user's work. This pushes the pup+// window (the given pid + its chrome/msedge children) to the BOTTOM without+// activating AND, if pup currently holds the foreground, hands focus straight back+// to `restoreHwnd` (the window the user was in). Called the instant the window+// appears (not after navigation), so there's no visible foreground pop.+// Takes a raw pid so it can run DURING launch, before the session is in `browsers`.+// cdpPort (optional) is the browser's CDP debug port: the process LISTENING on it is+// unambiguously the real browser with the window, which is more reliable than the+// spawned pid (Edge/Chrome relaunch to a different pid that isn't a child of it).+function osBackgroundWindowByPid(pid, restoreHwnd, cdpPort) {+  if (process.platform !== 'win32' || (!pid && !cdpPort)) return;+  try {+    const restore = (restoreHwnd && /^\d+$/.test(String(restoreHwnd))) ? String(restoreHwnd) : '0';+    const port = Number(cdpPort) || 0;+    const scriptPath = path.join(require('os').tmpdir(), `_adom_bgpid_${pid || port}.ps1`);+    const ps = [+      'Add-Type @"',+      'using System; using System.Runtime.InteropServices;',+      'public class WinBgP {',+      '  [DllImport("user32.dll")] public static extern bool SetWindowPos(IntPtr h, IntPtr a, int x, int y, int cx, int cy, uint f);',+      '  [DllImport("user32.dll")] public static extern bool SetForegroundWindow(IntPtr h);',+      '  [DllImport("user32.dll")] public static extern IntPtr GetForegroundWindow();',+      '}',+      '"@',+      `$targetPid = ${Number(pid) || 0}`,+      `$cdpPort = ${port}`,+      `$restore = [IntPtr]${restore}`,+      '$BOTTOM = [IntPtr]1',                 // HWND_BOTTOM+      '$F = 0x1 -bor 0x2 -bor 0x10',         // SWP_NOSIZE|SWP_NOMOVE|SWP_NOACTIVATE+      'function Back($h){ if($h -ne [IntPtr]::Zero){ [WinBgP]::SetWindowPos($h,$BOTTOM,0,0,0,0,$F) | Out-Null } }',+      '$handles = New-Object System.Collections.ArrayList',+      '$pids = New-Object System.Collections.ArrayList',+      'if ($targetPid -ne 0) { [void]$pids.Add($targetPid) }',+      // The process LISTENING on the CDP port is the real browser — most reliable anchor.+      'if ($cdpPort -ne 0) { try { $own = Get-NetTCPConnection -LocalPort $cdpPort -State Listen -ErrorAction SilentlyContinue | Select-Object -First 1 -ExpandProperty OwningProcess; if ($own) { [void]$pids.Add([int]$own) } } catch {} }',+      // For each anchor pid, take its own window + any chrome/msedge child windows.+      'foreach ($tp in $pids) {',+      '  $p = Get-Process -Id $tp -ErrorAction SilentlyContinue',+      '  if ($p -and $p.MainWindowHandle -ne 0) { [void]$handles.Add($p.MainWindowHandle) }',+      '  Get-CimInstance Win32_Process -Filter "ParentProcessId=$tp" -ErrorAction SilentlyContinue | Where-Object { $_.Name -in @("chrome.exe","msedge.exe") } | ForEach-Object { $cp = Get-Process -Id $_.ProcessId -ErrorAction SilentlyContinue; if ($cp -and $cp.MainWindowHandle -ne 0) { [void]$handles.Add($cp.MainWindowHandle) } }',+      '}',+      '$fg = [WinBgP]::GetForegroundWindow()',+      '$pupHadFg = $false',+      'foreach($h in $handles){ if($h -eq $fg){ $pupHadFg = $true }; Back $h }',+      // Hand focus back to the user's window ONLY if a pup window currently holds it.+      'if ($restore -ne [IntPtr]::Zero -and $pupHadFg) { [WinBgP]::SetForegroundWindow($restore) | Out-Null }',+      'Write-Output "ok"',+    ].join('\r\n');+    fs.writeFileSync(scriptPath, ps);+    execSync(`powershell.exe -NoProfile -ExecutionPolicy Bypass -File "${scriptPath}"`, { timeout: 6000, stdio: 'ignore' });+  } catch (e) { console.error(`[bg-restore] pid ${pid} port ${cdpPort}: ${e.message}`); }+}+ // v1.8.23: REAL taskbar flash (FlashWindowEx) for a session's window — the orange // nudge that tells the user "the agent touched this window" WITHOUT stealing focus.-// FLASHW_TRAY|FLASHW_TIMERNOFG flashes until the window is focused, then auto-clears-// (Windows does the clear-on-focus for us). Skips if the window is already foreground.-// Fire-and-forget (async exec) so it never adds latency to the driving verb.+// v1.8.24: a FINITE 4-blink pulse then stop (see Flash() below) — NOT a blink-forever+// nag. Skips if the window is already foreground. Fire-and-forget (async exec) so it+// never adds latency to the driving verb. function osFlashWindow(profileName) {   if (process.platform !== 'win32') return;   const be = browsers.get(profileName);@@ -1507,7 +1582,10 @@ function osFlashWindow(profileName) {       '  [DllImport("user32.dll")] public static extern bool FlashWindowEx(ref FLASHWINFO pwfi);',       '  [DllImport("user32.dll")] public static extern IntPtr GetForegroundWindow();',       '  public static void Flash(IntPtr h){ if(h==IntPtr.Zero) return; if(GetForegroundWindow()==h) return;',-      '    FLASHWINFO fi=new FLASHWINFO(); fi.cbSize=(uint)Marshal.SizeOf(fi); fi.hwnd=h; fi.dwFlags=0x2|0xC; fi.uCount=0; fi.dwTimeout=0; FlashWindowEx(ref fi); }',+      // v1.8.24: FINITE pulse — FLASHW_TRAY (taskbar button only), 4 blinks, then STOP.+      // NOT FLASHW_TIMERNOFG (which blinks forever until focused — an annoying nag). A short+      // orange pulse is a glance-and-forget "this window was updated" nudge, not a demand.+      '    FLASHWINFO fi=new FLASHWINFO(); fi.cbSize=(uint)Marshal.SizeOf(fi); fi.hwnd=h; fi.dwFlags=0x2; fi.uCount=4; fi.dwTimeout=0; FlashWindowEx(ref fi); }',       '}',       '"@',       `$targetPid = ${pid}`,@@ -1778,6 +1856,10 @@ async function getOrLaunchBrowser(profileName, { freshProfile = false, strictPer         `--user-data-dir=${userDataDir}`,       ];       console.log(`Chrome launch attempt ${attemptNum} for profile "${profileName}" on CDP port ${debugPort}`);+      // v1.8.24: remember the user's foreground window RIGHT BEFORE we launch, so we+      // can hand focus straight back once pup's window appears (background-by-default+      // must never steal focus, and --start-maximized grabs it on launch).+      const priorForeground = (!foreground) ? osGetForegroundWindow() : null;       const child = spawn(chromePath, chromeArgs, {         detached: true,         stdio: 'ignore',@@ -1835,6 +1917,17 @@ async function getOrLaunchBrowser(profileName, { freshProfile = false, strictPer       b._detachedPid = child.pid;       b._cdpPort = debugPort;       console.log(`Launched detached Chrome (PID ${child.pid}) on CDP port ${debugPort} for profile "${profileName}"`);+      // v1.8.24: ENFORCE background-by-default the instant the window exists — push it+      // behind the user's windows AND hand focus back to whatever they were in, so the+      // --start-maximized launch never steals focus. Retry a few times because Chrome/Edge+      // can re-raise itself during startup (first paint, restore prompts). Resolve the+      // window via the CDP-port owner (reliable across launcher→browser pid relaunches).+      if (!foreground && process.platform === 'win32') {+        osBackgroundWindowByPid(child.pid, priorForeground, debugPort);+        for (const delay of [250, 700, 1500]) {+          setTimeout(() => { try { osBackgroundWindowByPid(child.pid, priorForeground, debugPort); } catch (e) {} }, delay);+        }+      }       return b;     }; @@ -1940,7 +2033,7 @@ async function getOrLaunchBrowser(profileName, { freshProfile = false, strictPer // Launch a session as a tab in a Chrome browser // profile: which Chrome profile to use (defaults to sessionId) // If another session already uses that profile, this creates a NEW TAB in the same Chrome-async function launchSession(sessionId, url, { freshProfile = false, profile, strictPermissions = false, downloadPath = null, nativeSite = false, owner = null, takeover = false } = {}) {+async function launchSession(sessionId, url, { freshProfile = false, profile, strictPermissions = false, downloadPath = null, nativeSite = false, owner = null, takeover = false, foreground = false } = {}) {   const profileName = profile || sessionId;    // If session already exists and page is still alive, just navigate@@ -2834,6 +2927,10 @@ const server = http.createServer(async (req, res) => {               nativeSite: !!args.nativeSite,               owner: (typeof args.owner === 'string' && args.owner.trim()) ? args.owner.trim() : null,               takeover: !!args.takeover,+              // v1.8.24: background-by-default is enforced DURING launch (focus handed+              // back to the user's window the instant pup's window appears) unless the+              // caller explicitly wants it foreground.+              foreground: !!(args.foreground || args.raise || args.show),             });           } catch (e) {             if (e && e.code === 'session_owned_by_another_thread') {