@remcp/runtime 0.2.211 → 0.2.213

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@remcp/runtime",
-  "version": "0.2.211",
+  "version": "0.2.213",
   "description": "First-party ReMCP local device runtime: filesystem, terminal, computer-use, browser, diagnostics and document tools over MCP.",
   "type": "module",
   "license": "MIT",

package/src/extended/browser-actions.mjs

@@ -51,8 +51,10 @@ export const PAGE_SCROLL_ACTION = `function(selector, needle, deltaX, deltaY) {
 
   const beforeX = Number(scroller.scrollLeft || 0);
   const beforeY = Number(scroller.scrollTop || 0);
-  scroller.scrollLeft = beforeX + Number(deltaX || 0);
-  scroller.scrollTop = beforeY + Number(deltaY || 0);
+  // CSS scroll-behavior:smooth can defer these changes indefinitely in a
+  // background target. Apply the requested move before reading its result.
+  scroller.scrollTo({left:beforeX + Number(deltaX || 0),
+    top:beforeY + Number(deltaY || 0), behavior:'instant'});
   return {
     target: target ? target.tagName.toLowerCase() : 'page',
     scroller: scroller === document.scrollingElement || scroller === document.documentElement ? 'page' : scroller.tagName.toLowerCase(),

package/src/extended/browser-endpoint.mjs

@@ -35,6 +35,7 @@ export function browserExecutable() {
           path.join(process.env['PROGRAMFILES(X86)'] || 'C:\\Program Files (x86)', 'Google', 'Chrome', 'Application', 'chrome.exe'),
           path.join(process.env.LOCALAPPDATA || '', 'Google', 'Chrome', 'Application', 'chrome.exe'),
           path.join(process.env.PROGRAMFILES || 'C:\\Program Files', 'Microsoft', 'Edge', 'Application', 'msedge.exe'),
+          path.join(process.env['PROGRAMFILES(X86)'] || 'C:\\Program Files (x86)', 'Microsoft', 'Edge', 'Application', 'msedge.exe'),
         ]
       : ['/opt/google/chrome/chrome','/usr/bin/google-chrome','/usr/bin/google-chrome-stable','chromium','chromium-browser','microsoft-edge','microsoft-edge-stable'];
   return candidates.find(candidate => path.isAbsolute(candidate) ? existsSync(candidate) : commandExists(candidate)) || null;

package/src/extended/common.mjs

@@ -197,11 +197,15 @@ export async function runFileHeadLines(file, args = [], lineLimit = 200, options
   });
 }
 
+// Windows PowerShell writes redirected stdout in the OEM code page by default,
+// which corrupts Unicode names and clipboard text. Emit UTF-8 without a BOM.
+const WINDOWS_UTF8_OUTPUT = '[Console]::OutputEncoding=New-Object System.Text.UTF8Encoding $false;';
+
 export async function runPowerShell(script, options = {}) {
   const executable = process.env.SystemRoot
     ? path.join(process.env.SystemRoot, 'System32', 'WindowsPowerShell', 'v1.0', 'powershell.exe')
     : 'powershell.exe';
-  return runFile(executable, ['-NoLogo', '-NoProfile', '-NonInteractive', '-ExecutionPolicy', 'Bypass', '-Command', script], {
+  return runFile(executable, ['-NoLogo', '-NoProfile', '-NonInteractive', '-ExecutionPolicy', 'Bypass', '-Command', WINDOWS_UTF8_OUTPUT + script], {
     ...options,
     label: options.label || 'PowerShell',
   });

package/src/extended/desktop-computer-snapshot.mjs

@@ -16,15 +16,31 @@ import { findOcrText, ocrImage } from './ocr.mjs';
 import { boundedComputerSnapshot, clipboard, cursorPosition, displayInventory, listWindows, objectRecord, resultValue, screenshotRegion, snapshotError, suggestedUiAction, toolResultMessage, uiSnapshot } from './desktop-core.mjs';
 import { activeWindowCaptureBounds, browserTitleHint, computerScreenshotTask, hasExplicitComputerScreenshotTarget, ocrScreenshotResult } from './desktop-vision.mjs';
 
+// Several simultaneous Windows PowerShell starts can stall until their shared
+// timeout on low-core hosts, which made the whole snapshot report every probe
+// as unavailable. Queue the probes on Windows; other platforms stay parallel.
+export async function settleComputerSnapshotProbes(tasks, { serial = process.platform === 'win32' } = {}) {
+  if (!serial) return Promise.allSettled(tasks.map((task) => task()));
+  const results = [];
+  for (const task of tasks) {
+    try {
+      results.push({ status: 'fulfilled', value: await task() });
+    } catch (reason) {
+      results.push({ status: 'rejected', reason });
+    }
+  }
+  return results;
+}
+
 export async function computerSnapshot(args = {}) {
   const desktopAvailable = desktopSessionAvailable();
   const screenshotRequested = args.include_screenshot !== false || args.include_ocr === true;
   const captureScreenshot = desktopAvailable && screenshotRequested;
   const includeUi = args.include_ui !== false;
-  const [windowsResult, displaysResult, uiResult, clipboardResult, cursorResult, shotResult, readinessResult] = await Promise.allSettled([
-    desktopAvailable ? listWindows() : Promise.resolve(null),
-    desktopAvailable ? displayInventory() : Promise.resolve(null),
-    includeUi && desktopAvailable
+  const [windowsResult, displaysResult, uiResult, clipboardResult, cursorResult, shotResult, readinessResult] = await settleComputerSnapshotProbes([
+    () => desktopAvailable ? listWindows() : Promise.resolve(null),
+    () => desktopAvailable ? displayInventory() : Promise.resolve(null),
+    () => includeUi && desktopAvailable
       ? uiSnapshot({
           max_nodes:args.max_ui_nodes || 800,
           max_depth:args.max_ui_depth || 12,
@@ -33,10 +49,10 @@ export async function computerSnapshot(args = {}) {
           ...(args.ui_browser_dom === true ? { browser_dom:true } : {}),
         })
       : Promise.resolve(null),
-    desktopAvailable ? clipboard({ action:'read' }) : Promise.resolve(null),
-    desktopAvailable ? cursorPosition() : Promise.resolve(null),
-    captureScreenshot ? computerScreenshotTask(args) : Promise.resolve(null),
-    args.browser_readiness === true ? probeBrowserReadiness() : Promise.resolve(null),
+    () => desktopAvailable ? clipboard({ action:'read' }) : Promise.resolve(null),
+    () => desktopAvailable ? cursorPosition() : Promise.resolve(null),
+    () => captureScreenshot ? computerScreenshotTask(args) : Promise.resolve(null),
+    () => args.browser_readiness === true ? probeBrowserReadiness() : Promise.resolve(null),
   ]);
   const parse = (result, fallback = null) => {
     if (result.status !== 'fulfilled' || !result.value || result.value.isError === true) return fallback;

package/src/extended/desktop-windows.mjs

@@ -197,7 +197,7 @@ while($q.Count -gt 0 -and $seen -lt ${maxNodes}){
   return jsonResult(parsed);
 }
 
-export async function uiAction(args = {}) {
+export async function uiAction(args = {}, { runPowerShellFn = runPowerShell } = {}) {
   const action = requireEnum(args.action, 'action', [
     'click','invoke','focus','set_value','select','toggle','expand','collapse',
     'scroll_into_view','set_range_value','add_to_selection','remove_from_selection',
@@ -207,7 +207,8 @@ export async function uiAction(args = {}) {
   const role = escapePowerShellSingle(optionalString(args.role) || '');
   const automationId = escapePowerShellSingle(optionalString(args.automation_id || args.automationId) || '');
   const value = escapePowerShellSingle(String(args.value ?? ''));
-  const numericValue = Number(args.value);
+  // The entire switch must parse even when the selected action has no numeric value.
+  const numericValue = action === 'set_range_value' ? Number(args.value) : 0;
   if (action === 'set_range_value' && !Number.isFinite(numericValue)) throw new Error('set_range_value requires a finite numeric value');
   const script = String.raw`
 Add-Type -AssemblyName UIAutomationClient;Add-Type -AssemblyName UIAutomationTypes
@@ -278,16 +279,40 @@ switch('${action}'){
  }
 }
 [pscustomobject]@{id=('win:'+($found.GetRuntimeId()-join '.'));name=$found.Current.Name;role=$found.Current.ControlType.ProgrammaticName.Replace('ControlType.','');action='${action}';backend=$used}|ConvertTo-Json -Compress`;
-  const raw = (await runPowerShell(script, { label:'UI action', timeout:30_000 })).stdout.trim();
+  const raw = (await runPowerShellFn(script, { label:'UI action', timeout:30_000 })).stdout.trim();
   return jsonResult(raw ? JSON.parse(raw) : { action });
 }
 
+export function buildWindowsClipboardPowerShell(action, value = '') {
+  // Set-Clipboard rejects an empty value in Windows PowerShell 5.1.
+  // Clear the clipboard using the OS clipboard API instead of treating it as a write.
+  if (action === 'clear' || value.length === 0) {
+    return 'Add-Type -AssemblyName System.Windows.Forms;[System.Windows.Forms.Clipboard]::Clear()';
+  }
+  const encoded = Buffer.from(value, 'utf16le').toString('base64');
+  return `$b=[Convert]::FromBase64String('${encoded}');Set-Clipboard -Value ([Text.Encoding]::Unicode.GetString($b))`;
+}
+
+export function buildWindowsClipboardReadPowerShell() {
+  // Windows PowerShell writes redirected stdout in the OEM code page, which
+  // corrupts non-ASCII clipboard text. Return UTF-8 bytes as base64 instead.
+  return "$v=Get-Clipboard -Raw;if($null -eq $v){$v=''};[Convert]::ToBase64String([Text.Encoding]::UTF8.GetBytes($v))";
+}
+
+export function decodeWindowsClipboardRead(stdout = '') {
+  const encoded = String(stdout).trim();
+  if (!encoded) return '';
+  return Buffer.from(encoded, 'base64').toString('utf8').replace(/\r?\n$/, '');
+}
+
 export async function clipboard(args = {}) {
   const action=requireEnum(args.action,'action',['read','write','clear']);
-  if(action==='read')return text((await runPowerShell("$v=Get-Clipboard -Raw;if($null -eq $v){$v=''};$v",{label:'clipboard read'})).stdout.replace(/\r?\n$/,''));
+  if(action==='read'){
+    const {stdout}=await runPowerShell(buildWindowsClipboardReadPowerShell(),{label:'clipboard read'});
+    return text(decodeWindowsClipboardRead(stdout));
+  }
   const value=action==='clear'?'':String(args.text??args.value??'');
-  const encoded=Buffer.from(value,'utf16le').toString('base64');
-  await runPowerShell(`$b=[Convert]::FromBase64String('${encoded}');Set-Clipboard -Value ([Text.Encoding]::Unicode.GetString($b))`,{label:'clipboard write'});
+  await runPowerShell(buildWindowsClipboardPowerShell(action, value),{label: action === 'clear' ? 'clipboard clear' : 'clipboard write'});
   return jsonResult({action,length:value.length});
 }
 
@@ -310,8 +335,10 @@ export async function typeTextKeys(value) {
 }
 
 function mousePrelude() {
+  // dwData is a signed wheel delta. Declaring it as uint made the generated
+  // negative literal cast throw before the wheel event was sent.
   return String.raw`Add-Type @'
-using System;using System.Runtime.InteropServices;public static class RMMouse{[DllImport("user32.dll")]public static extern bool SetCursorPos(int x,int y);[DllImport("user32.dll")]public static extern void mouse_event(uint f,uint x,uint y,uint d,UIntPtr i);}
+using System;using System.Runtime.InteropServices;public static class RMMouse{[DllImport("user32.dll")]public static extern bool SetCursorPos(int x,int y);[DllImport("user32.dll")]public static extern void mouse_event(uint f,uint x,uint y,int d,UIntPtr i);}
 '@`;
 }
 
@@ -337,10 +364,15 @@ export async function dragDrop(args = {}) {
   return jsonResult({from:[args.from_x,args.from_y],to:[args.to_x,args.to_y]});
 }
 
+export function buildWindowsScrollPowerShell(dx, dy) {
+  const wheel = Math.trunc(Number(dy) || 0);
+  const horizontal = Math.trunc(Number(dx) || 0);
+  return `${mousePrelude()}\n[RMMouse]::mouse_event(0x0800,0,0,${-wheel},[UIntPtr]::Zero);[RMMouse]::mouse_event(0x01000,0,0,${horizontal},[UIntPtr]::Zero)`;
+}
+
 export async function scroll(args = {}) {
   const dx=Number(args.delta_x||0),dy=Number(args.delta_y??args.delta??0);
-  const script=`${mousePrelude()}\n[RMMouse]::mouse_event(0x0800,0,0,[uint32]([int]${Math.trunc(-dy)}),[UIntPtr]::Zero);[RMMouse]::mouse_event(0x01000,0,0,[uint32]([int]${Math.trunc(dx)}),[UIntPtr]::Zero)`;
-  await runPowerShell(script,{label:'scroll'});
+  await runPowerShell(buildWindowsScrollPowerShell(dx,dy),{label:'scroll'});
   return jsonResult({delta_x:dx,delta_y:dy});
 }
 

package/src/screenshot-portal.mjs

@@ -210,9 +210,19 @@ async function copyPortalFile(source, destination) {
   try {
     sourceParent = await openDirectoryNoFollow(path.dirname(source));
     destinationParent = await openDirectoryNoFollow(path.dirname(destination));
-    input = await open(path.join(sourceParent.anchor, path.basename(source)), constants.O_RDONLY | constants.O_NOFOLLOW);
-    const info = await input.stat();
-    if (!info.isFile()) throw new Error('desktop portal returned a non-regular screenshot file');
+    const sourcePath = path.join(sourceParent.anchor, path.basename(source));
+    // On Windows O_NOFOLLOW is not available: inspect the leaf explicitly.
+    // BigInt identifiers avoid rounding NTFS file IDs above 2^53.
+    const before = await lstat(sourcePath, { bigint: true });
+    if (!before.isFile() || before.isSymbolicLink())
+      throw new Error('desktop portal screenshot source cannot be a symbolic link');
+    input = await open(sourcePath, constants.O_RDONLY | (constants.O_NOFOLLOW || 0));
+    const info = await input.stat({ bigint: true });
+    const current = await lstat(sourcePath, { bigint: true });
+    if (!info.isFile() || current.isSymbolicLink() || !current.isFile() ||
+        info.dev !== before.dev || info.ino !== before.ino ||
+        current.dev !== before.dev || current.ino !== before.ino)
+      throw new Error('desktop portal screenshot source changed or is not regular');
     const destinationPath = path.join(destinationParent.anchor, path.basename(destination));
     const existing = await lstat(destinationPath).catch(() => null);
     if (existing?.isSymbolicLink()) throw new Error('desktop portal screenshot destination cannot be a symbolic link');

package/src/tools/files-bulk-operations.mjs

@@ -339,7 +339,10 @@ export async function setPermissionsTool(args) {
       }
     }
   }
-  const summary = `Set mode ${raw}${uid !== null || gid !== null ? ` (uid ${uid ?? '-'} gid ${gid ?? '-'})` : ''} on ${changed} path(s) starting at ${displayPath(absolute)}.`;
+  const windowsNote = process.platform === 'win32'
+    ? '\nWindows applies only the read-only attribute; POSIX group/other bits and ownership are not changed.'
+    : '';
+  const summary = `Set mode ${raw}${uid !== null || gid !== null ? ` (uid ${uid ?? '-'} gid ${gid ?? '-'})` : ''} on ${changed} path(s) starting at ${displayPath(absolute)}.${windowsNote}`;
   if (!failures.length) return text(summary);
   return text(`${summary}\n${failures.length} path(s) could not be changed:\n${failures.slice(0, 20).join('\n')}`, true);
 }

package/src/tools/search.mjs

@@ -184,7 +184,9 @@ function runRipgrep(session, { target, pattern, searchType, filePattern, ignoreC
   const searchTarget = target.anchor;
   if (ignoreCase) args.push('--ignore-case');
   if (includeHidden) args.push('--hidden');
-  if (!includeIgnored) for (const glob of SKIP_GLOBS) args.push('-g', glob);
+  // includeIgnored must override both our skip globs and ripgrep's own ignore files.
+  if (includeIgnored) args.push('--no-ignore');
+  else for (const glob of SKIP_GLOBS) args.push('-g', glob);
   for (const glob of splitGlobs(filePattern)) args.push('-g', glob);
   if (searchType === 'files') {
     args.push('--files');

package/src/tools/terminal.mjs

@@ -24,8 +24,16 @@ function shellCommand() {
   return process.env.SHELL || '/bin/bash';
 }
 
-function shellArgs(command) {
-  if (process.platform === 'win32') return ['/d', '/s', '/c', command];
+function usesWindowsCmd(shell) {
+  return process.platform === 'win32' && /(?:^|[\\/])cmd(?:\.exe)?$/i.test(shell);
+}
+
+function shellArgs(command, shell) {
+  // CMD parses its own quoting: wrap the full command and disable CRT escaping below.
+  if (usesWindowsCmd(shell)) return ['/d', '/s', '/c', '"' + command + '"'];
+  if (process.platform === 'win32' && /(?:^|[\\/])(?:powershell|pwsh)(?:\.exe)?$/i.test(shell)) {
+    return ['-NoLogo', '-NoProfile', '-NonInteractive', '-Command', command];
+  }
   return ['-c', command];
 }
 
@@ -90,11 +98,12 @@ export async function startProcessTool(args, extra = {}) {
   const shell = shellCommand();
   // `detached` gives the child its own process group so a session can be stopped as a
   // tree; a pipeline such as `sleep 20 | cat` otherwise survives force_terminate.
-  const child = spawn(shell, shellArgs(command), {
+  const child = spawn(shell, shellArgs(command, shell), {
     cwd: process.cwd(),
     env: process.env,
     stdio: ['pipe', 'pipe', 'pipe'],
     windowsHide: true,
+    windowsVerbatimArguments: usesWindowsCmd(shell),
     detached: process.platform !== 'win32',
   });
   // The error listener has to exist before anything can throw: a bogus REMCP_RUNTIME_SHELL