mcp-baepsae 6.2.0 → 6.3.0

package/native/Sources/Commands/UICommands.swift

@@ -24,7 +24,16 @@ func handleDescribeUI(_ parsed: ParsedOptions) throws -> Int32 {
         // For macOS apps, start from window level, no iOSContentGroup filtering
         if !parsed.flags.contains("--all") {
             let windows = Children(appRoot)
-            if let firstWindow = windows.first {
+            if let windowSelector = parsed.options["--window"] {
+                if let idx = Int(windowSelector), idx < windows.count {
+                    targetRoot = windows[idx]
+                } else {
+                    targetRoot = windows.first(where: {
+                        let title = StringAttribute($0, kAXTitleAttribute as CFString) ?? ""
+                        return title.localizedCaseInsensitiveContains(windowSelector)
+                    }) ?? windows.first ?? appRoot
+                }
+            } else if let firstWindow = windows.first {
                 targetRoot = firstWindow
             }
         }
@@ -72,6 +81,31 @@ func handleDescribeUI(_ parsed: ParsedOptions) throws -> Int32 {
         }
     }
 
+    // Page-size based pagination for large trees
+    if let pageSizeStr = parsed.options["--page-size"], let pageSize = Int(pageSizeStr) {
+        let page = Int(parsed.options["--page"] ?? "0") ?? 0
+        var totalCount: CFIndex = 0
+        AXUIElementGetAttributeValueCount(targetRoot, kAXChildrenAttribute as CFString, &totalCount)
+        let offset = page * pageSize
+        if offset >= Int(totalCount) {
+            print("Page \(page) is beyond total children count (\(totalCount)).")
+            print("total=\(totalCount) pageSize=\(pageSize) pages=\((Int(totalCount) + pageSize - 1) / max(pageSize, 1))")
+            return 0
+        }
+        var pageChildren: CFArray?
+        let fetchCount = min(pageSize, Int(totalCount) - offset)
+        AXUIElementCopyAttributeValues(targetRoot, kAXChildrenAttribute as CFString, CFIndex(offset), CFIndex(fetchCount), &pageChildren)
+        let totalPages = (Int(totalCount) + pageSize - 1) / max(pageSize, 1)
+        print("page=\(page)/\(max(totalPages - 1, 0)) total=\(totalCount) pageSize=\(pageSize) showing=\(offset)..\(offset + fetchCount - 1)")
+        if let children = pageChildren as? [AXUIElement] {
+            for child in children {
+                let childLines = describeAccessibilityTree(from: child, options: descOpts)
+                print(childLines.joined(separator: "\n"))
+            }
+        }
+        return 0
+    }
+
     var lines = describeAccessibilityTree(from: targetRoot, options: descOpts)
     if lines.isEmpty {
         throw NativeError.commandFailed("No accessibility elements found.")
@@ -366,6 +400,34 @@ func handleType(_ parsed: ParsedOptions) throws -> Int32 {
     }
 
     let methodStr = parsed.options["--method"] ?? "auto"
+
+    if methodStr == "ax" {
+        let appRoot = try accessibilityRootElement(for: target)
+        var focusedRef: CFTypeRef?
+        let status = AXUIElementCopyAttributeValue(
+            appRoot,
+            "AXFocusedUIElement" as CFString,
+            &focusedRef
+        )
+        if status == .success, let focused = focusedRef {
+            let setResult = AXUIElementSetAttributeValue(
+                focused as! AXUIElement,
+                kAXValueAttribute as CFString,
+                text as CFTypeRef
+            )
+            if setResult == .success {
+                print("Set value via AX API.")
+                return 0
+            }
+            fputs("AX setValue failed (status \(setResult.rawValue)), falling back to paste\n", stderr)
+        } else {
+            fputs("No focused element found, falling back to paste\n", stderr)
+        }
+        // Fallback to paste
+        try pasteText(text, target: target)
+        return 0
+    }
+
     let usePaste: Bool
     switch methodStr {
     case "paste":
@@ -373,11 +435,7 @@ func handleType(_ parsed: ParsedOptions) throws -> Int32 {
     case "keyboard":
         usePaste = false
     default: // auto
-        if case .simulator = target {
-            usePaste = true
-        } else {
-            usePaste = false
-        }
+        usePaste = true  // paste is more reliable for all targets including CJK
     }
 
     if usePaste {
@@ -410,7 +468,7 @@ func pasteText(_ text: String, target: TargetApp) throws {
 
         // Cmd+V: Command keycode=55, V keycode=9
         sendKeyCombo(modifiers: [55], key: 9)
-        Thread.sleep(forTimeInterval: 0.15)
+        Thread.sleep(forTimeInterval: 0.3)
 
         // Restore original clipboard content
         pasteboard.clearContents()
@@ -462,6 +520,67 @@ func handleSwipe(_ parsed: ParsedOptions) throws -> Int32 {
     return 0
 }
 
+// MARK: - detect-dialog
+
+func handleDetectDialog(_ parsed: ParsedOptions) throws -> Int32 {
+    let target = try resolveTarget(from: parsed)
+    try ensureAccessibilityTrusted()
+    try activateTarget(target)
+    let appRoot = try accessibilityRootElement(for: target)
+    let windows = Children(appRoot)
+
+    var dialogs: [[String: Any]] = []
+    for window in windows {
+        let subrole = StringAttribute(window, kAXSubroleAttribute as CFString) ?? ""
+        let isDialog = ["AXDialog", "AXSheet", "AXSystemDialog", "AXSystemFloatingWindow"].contains(subrole)
+
+        var modalRef: CFTypeRef?
+        let modalStatus = AXUIElementCopyAttributeValue(window, kAXModalAttribute as CFString, &modalRef)
+        let isModal = (modalStatus == .success && (modalRef as? Bool) == true)
+
+        if isDialog || isModal {
+            let title = StringAttribute(window, kAXTitleAttribute as CFString) ?? ""
+            var buttons: [String] = []
+
+            func findButtons(in elements: [UIElement]) {
+                for element in elements {
+                    let role = StringAttribute(element, kAXRoleAttribute as CFString)
+                    if role == "AXButton" {
+                        if let btnTitle = StringAttribute(element, kAXTitleAttribute as CFString), !btnTitle.isEmpty {
+                            buttons.append(btnTitle)
+                        }
+                    }
+                    findButtons(in: Children(element))
+                }
+            }
+            findButtons(in: Children(window))
+
+            dialogs.append([
+                "type": subrole.isEmpty ? "modal" : subrole,
+                "title": title,
+                "isModal": isModal,
+                "buttons": buttons,
+            ])
+        }
+    }
+
+    if dialogs.isEmpty {
+        print("{\"hasDialog\":false,\"dialogs\":[]}")
+    } else {
+        var jsonParts: [String] = []
+        for d in dialogs {
+            let type = d["type"] as? String ?? ""
+            let title = (d["title"] as? String ?? "").replacingOccurrences(of: "\\", with: "\\\\").replacingOccurrences(of: "\"", with: "\\\"")
+            let isModal = d["isModal"] as? Bool ?? false
+            let btns = d["buttons"] as? [String] ?? []
+            let btnsJson = "[" + btns.map { "\"\($0.replacingOccurrences(of: "\\", with: "\\\\").replacingOccurrences(of: "\"", with: "\\\""))\"" }.joined(separator: ",") + "]"
+            jsonParts.append("{\"type\":\"\(type)\",\"title\":\"\(title)\",\"isModal\":\(isModal),\"buttons\":\(btnsJson)}")
+        }
+        print("{\"hasDialog\":true,\"dialogs\":[\(jsonParts.joined(separator: ","))]}")
+    }
+    return 0
+}
+
 func handleScroll(_ parsed: ParsedOptions) throws -> Int32 {
     let target = try resolveTarget(from: parsed)
     try ensureAccessibilityTrusted()
@@ -503,3 +622,305 @@ func handleScroll(_ parsed: ParsedOptions) throws -> Int32 {
     }
     return 0
 }
+
+// MARK: - handleSetUIValue (F010)
+
+func handleSetUIValue(_ parsed: ParsedOptions) throws -> Int32 {
+    let target = try resolveTarget(from: parsed)
+    try ensureAccessibilityTrusted()
+    try activateTarget(target)
+    let appRoot = try accessibilityRootElement(for: target)
+
+    let attributeStr = parsed.options["--attribute"] ?? "value"
+    let valueStr = try requiredOption("--value", from: parsed)
+
+    // Find element
+    let element: AXUIElement
+    if let accessibilityId = parsed.options["--id"] {
+        guard let found = findAccessibilityElement(in: appRoot, identifier: accessibilityId, label: nil) else {
+            throw NativeError.commandFailed("No element with id: \(accessibilityId)")
+        }
+        element = found
+    } else if let label = parsed.options["--label"] {
+        guard let found = findAccessibilityElement(in: appRoot, identifier: nil, label: label) else {
+            throw NativeError.commandFailed("No element with label: \(label)")
+        }
+        element = found
+    } else {
+        var focusedRef: CFTypeRef?
+        let status = AXUIElementCopyAttributeValue(appRoot, "AXFocusedUIElement" as CFString, &focusedRef)
+        guard status == .success, let ref = focusedRef else {
+            throw NativeError.commandFailed("No focused element and no --id or --label provided.")
+        }
+        element = ref as! AXUIElement
+    }
+
+    // 속성 이름 결정 (settable 체크와 set 호출 양쪽에 공통 사용)
+    let axAttributeName: String
+    switch attributeStr {
+    case "value":           axAttributeName = kAXValueAttribute as String
+    case "selectedTextRange": axAttributeName = "AXSelectedTextRange"
+    case "focused":         axAttributeName = kAXFocusedAttribute as String
+    default:
+        throw NativeError.invalidArguments("Unsupported attribute: \(attributeStr). Use: value, selectedTextRange, focused.")
+    }
+
+    // 쓰기 가능 여부 사전 확인
+    var settable: DarwinBoolean = false
+    AXUIElementIsAttributeSettable(element, axAttributeName as CFString, &settable)
+    if !settable.boolValue {
+        throw NativeError.commandFailed(
+            "Attribute '\(attributeStr)' is not writable on this element. "
+            + "Use enumerate_ui to discover which attributes are settable."
+        )
+    }
+
+    switch attributeStr {
+    case "value":
+        let result = AXUIElementSetAttributeValue(element, kAXValueAttribute as CFString, valueStr as CFTypeRef)
+        guard result == .success else {
+            throw NativeError.commandFailed("Failed to set value (AXError \(result.rawValue)). Element may not support writable value.")
+        }
+        print("Set value: \(valueStr)")
+
+    case "selectedTextRange":
+        // Parse "location,length" format
+        let parts = valueStr.split(separator: ",").compactMap { Int($0.trimmingCharacters(in: .whitespaces)) }
+        guard parts.count == 2 else {
+            throw NativeError.invalidArguments("selectedTextRange requires 'location,length' format (e.g. '10,5').")
+        }
+        var range = CFRange(location: parts[0], length: parts[1])
+        guard let axValue = AXValueCreate(.cfRange, &range) else {
+            throw NativeError.commandFailed("Failed to create AXValue for range.")
+        }
+        let result = AXUIElementSetAttributeValue(element, "AXSelectedTextRange" as CFString, axValue)
+        guard result == .success else {
+            throw NativeError.commandFailed("Failed to set selectedTextRange (AXError \(result.rawValue)).")
+        }
+        print("Set selectedTextRange: location=\(parts[0]), length=\(parts[1])")
+
+    case "focused":
+        let boolValue = (valueStr == "true" || valueStr == "1")
+        let result = AXUIElementSetAttributeValue(element, kAXFocusedAttribute as CFString, boolValue as CFTypeRef)
+        guard result == .success else {
+            throw NativeError.commandFailed("Failed to set focused (AXError \(result.rawValue)).")
+        }
+        print("Set focused: \(boolValue)")
+
+    default:
+        throw NativeError.invalidArguments("Unsupported attribute: \(attributeStr). Use: value, selectedTextRange, focused.")
+    }
+    return 0
+}
+
+// MARK: - handleHitTest (F003)
+
+func handleHitTest(_ parsed: ParsedOptions) throws -> Int32 {
+    try ensureAccessibilityTrusted()
+    guard let xStr = parsed.options["-x"], let yStr = parsed.options["-y"],
+          let x = Float(xStr), let y = Float(yStr) else {
+        throw NativeError.invalidArguments("hit-test requires -x and -y coordinates.")
+    }
+
+    let systemWide = AXUIElementCreateSystemWide()
+    var element: AXUIElement?
+    let status = AXUIElementCopyElementAtPosition(systemWide, x, y, &element)
+    guard status == .success, let found = element else {
+        throw NativeError.commandFailed("No accessibility element found at (\(x), \(y)). AXError: \(status.rawValue)")
+    }
+
+    let role = StringAttribute(found, kAXRoleAttribute as CFString) ?? ""
+    let subrole = StringAttribute(found, kAXSubroleAttribute as CFString) ?? ""
+    let title = StringAttribute(found, kAXTitleAttribute as CFString) ?? ""
+    let identifier = StringAttribute(found, "AXIdentifier" as CFString) ?? ""
+    let value = StringAttribute(found, kAXValueAttribute as CFString) ?? ""
+    let description = StringAttribute(found, kAXDescriptionAttribute as CFString) ?? ""
+
+    var frameStr = ""
+    if let frame = FrameAttribute(found) {
+        frameStr = "\(formatFloat(frame.origin.x)),\(formatFloat(frame.origin.y)),\(formatFloat(frame.width)),\(formatFloat(frame.height))"
+    }
+
+    var pid: pid_t = 0
+    AXUIElementGetPid(found, &pid)
+    let appName = NSWorkspace.shared.runningApplications.first(where: { $0.processIdentifier == pid })?.localizedName ?? ""
+
+    print("role=\(role) subrole=\(subrole) title=\(title) id=\(identifier) value=\(value) desc=\(description) frame=\(frameStr) pid=\(pid) app=\(appName)")
+    return 0
+}
+
+// MARK: - handleEnumerateUI (F005)
+
+func handleEnumerateUI(_ parsed: ParsedOptions) throws -> Int32 {
+    let target = try resolveTarget(from: parsed)
+    try ensureAccessibilityTrusted()
+    try activateTarget(target)
+    let appRoot = try accessibilityRootElement(for: target)
+
+    let element: AXUIElement
+    if let accessibilityId = parsed.options["--id"] {
+        guard let found = findAccessibilityElement(in: appRoot, identifier: accessibilityId, label: nil) else {
+            throw NativeError.commandFailed("No element with id: \(accessibilityId)")
+        }
+        element = found
+    } else if let label = parsed.options["--label"] {
+        guard let found = findAccessibilityElement(in: appRoot, identifier: nil, label: label) else {
+            throw NativeError.commandFailed("No element with label: \(label)")
+        }
+        element = found
+    } else {
+        var focusedRef: CFTypeRef?
+        let status = AXUIElementCopyAttributeValue(appRoot, "AXFocusedUIElement" as CFString, &focusedRef)
+        guard status == .success, let ref = focusedRef else {
+            throw NativeError.commandFailed("No focused element and no --id or --label provided.")
+        }
+        element = ref as! AXUIElement
+    }
+
+    // 속성 목록
+    var attrNames: CFArray?
+    AXUIElementCopyAttributeNames(element, &attrNames)
+    var attributes: [(name: String, settable: Bool)] = []
+    if let names = attrNames as? [String] {
+        for name in names {
+            var settable: DarwinBoolean = false
+            AXUIElementIsAttributeSettable(element, name as CFString, &settable)
+            attributes.append((name: name, settable: settable.boolValue))
+        }
+    }
+
+    // 액션 목록
+    var actionNames: CFArray?
+    AXUIElementCopyActionNames(element, &actionNames)
+    var actions: [(name: String, description: String)] = []
+    if let names = actionNames as? [String] {
+        for name in names {
+            var desc: CFString?
+            AXUIElementCopyActionDescription(element, name as CFString, &desc)
+            actions.append((name: name, description: (desc as String?) ?? ""))
+        }
+    }
+
+    // 파라미터화된 속성 목록
+    var paramNames: CFArray?
+    AXUIElementCopyParameterizedAttributeNames(element, &paramNames)
+    let parameterized = (paramNames as? [String]) ?? []
+
+    let attrJson = attributes.map { a in
+        let escapedName = a.name.replacingOccurrences(of: "\"", with: "\\\"")
+        return "{\"name\":\"\(escapedName)\",\"settable\":\(a.settable)}"
+    }.joined(separator: ",")
+
+    let actJson = actions.map { a in
+        let escapedName = a.name.replacingOccurrences(of: "\"", with: "\\\"")
+        let escapedDesc = a.description.replacingOccurrences(of: "\\", with: "\\\\").replacingOccurrences(of: "\"", with: "\\\"")
+        return "{\"name\":\"\(escapedName)\",\"description\":\"\(escapedDesc)\"}"
+    }.joined(separator: ",")
+
+    let paramJson = parameterized.map { "\"" + $0.replacingOccurrences(of: "\"", with: "\\\"") + "\"" }.joined(separator: ",")
+
+    print("{\"attributes\":[\(attrJson)],\"actions\":[\(actJson)],\"parameterizedAttributes\":[\(paramJson)]}")
+    return 0
+}
+
+// MARK: - handleReadUIParam (F012)
+
+func handleReadUIParam(_ parsed: ParsedOptions) throws -> Int32 {
+    let target = try resolveTarget(from: parsed)
+    try ensureAccessibilityTrusted()
+    try activateTarget(target)
+    let appRoot = try accessibilityRootElement(for: target)
+
+    let attributeStr = try requiredOption("--attribute", from: parsed)
+    let paramStr = try requiredOption("--param", from: parsed)
+
+    // Find element
+    let element: AXUIElement
+    if let accessibilityId = parsed.options["--id"] {
+        guard let found = findAccessibilityElement(in: appRoot, identifier: accessibilityId, label: nil) else {
+            throw NativeError.commandFailed("No element with id: \(accessibilityId)")
+        }
+        element = found
+    } else if let label = parsed.options["--label"] {
+        guard let found = findAccessibilityElement(in: appRoot, identifier: nil, label: label) else {
+            throw NativeError.commandFailed("No element with label: \(label)")
+        }
+        element = found
+    } else {
+        var focusedRef: CFTypeRef?
+        let status = AXUIElementCopyAttributeValue(appRoot, "AXFocusedUIElement" as CFString, &focusedRef)
+        guard status == .success, let ref = focusedRef else {
+            throw NativeError.commandFailed("No focused element and no --id or --label provided.")
+        }
+        element = ref as! AXUIElement
+    }
+
+    let axAttribute: String
+    let parameter: CFTypeRef
+
+    switch attributeStr {
+    case "stringForRange", "boundsForRange":
+        let parts = paramStr.split(separator: ",").compactMap { Int($0.trimmingCharacters(in: .whitespaces)) }
+        guard parts.count == 2 else {
+            throw NativeError.invalidArguments("\(attributeStr) requires 'location,length' format.")
+        }
+        var range = CFRange(location: parts[0], length: parts[1])
+        guard let axValue = AXValueCreate(.cfRange, &range) else {
+            throw NativeError.commandFailed("Failed to create AXValue for range.")
+        }
+        parameter = axValue
+        axAttribute = attributeStr == "stringForRange"
+            ? "AXStringForRange"
+            : "AXBoundsForRange"
+
+    case "lineForIndex":
+        guard let index = Int(paramStr) else {
+            throw NativeError.invalidArguments("lineForIndex requires a numeric index.")
+        }
+        parameter = index as CFTypeRef
+        axAttribute = "AXLineForIndex"
+
+    case "rangeForLine":
+        guard let line = Int(paramStr) else {
+            throw NativeError.invalidArguments("rangeForLine requires a numeric line number.")
+        }
+        parameter = line as CFTypeRef
+        axAttribute = "AXRangeForLine"
+
+    default:
+        throw NativeError.invalidArguments("Unsupported parameterized attribute: \(attributeStr). Use: stringForRange, boundsForRange, lineForIndex, rangeForLine.")
+    }
+
+    var resultRef: CFTypeRef?
+    let status = AXUIElementCopyParameterizedAttributeValue(element, axAttribute as CFString, parameter, &resultRef)
+    guard status == .success, let result = resultRef else {
+        throw NativeError.commandFailed("Failed to read \(attributeStr) (AXError \(status.rawValue)).")
+    }
+
+    if let str = result as? String {
+        print(str)
+    } else if let num = result as? NSNumber {
+        print(num.stringValue)
+    } else if CFGetTypeID(result) == AXValueGetTypeID() {
+        let axVal = result as! AXValue
+        let valType = AXValueGetType(axVal)
+        if valType == .cfRange {
+            var range = CFRange(location: 0, length: 0)
+            AXValueGetValue(axVal, .cfRange, &range)
+            print("\(range.location),\(range.length)")
+        } else if valType == .cgRect {
+            var rect = CGRect.zero
+            AXValueGetValue(axVal, .cgRect, &rect)
+            print("\(rect.origin.x),\(rect.origin.y),\(rect.width),\(rect.height)")
+        } else if valType == .cgPoint {
+            var point = CGPoint.zero
+            AXValueGetValue(axVal, .cgPoint, &point)
+            print("\(point.x),\(point.y)")
+        } else {
+            print(String(describing: result))
+        }
+    } else {
+        print(String(describing: result))
+    }
+    return 0
+}

package/native/Sources/Utils.swift

@@ -36,6 +36,17 @@ let supportedCommands: Set<String> = [
     "menu-action",
     "get-focused-app",
     "clipboard",
+    "list-input-sources",
+    "input-source",
+    "focus-window",
+    "read-ui-value",
+    "context-menu-action",
+    "detect-dialog",
+    "set-ui-value",
+    "read-ui-param",
+    "hit-test",
+    "enumerate-ui",
+    "watch-notification",
 ]
 
 // MARK: - Argument Parsing
@@ -291,7 +302,13 @@ func accessibilityRootElement(for target: TargetApp) throws -> UIElement {
     case .simulator:
         return try simulatorAccessibilityRootElement()
     case .macApp(let pid, _, _):
-        return AXUIElementCreateApplication(pid)
+        let appElement = AXUIElementCreateApplication(pid)
+        // 무거운 앱에서 기본 시스템 타임아웃(~6초)이 너무 짧은 경우 대비
+        // BAEPSAE_AX_TIMEOUT 환경변수(초)로 재정의 가능, 기본값 10초
+        let timeout = Float(ProcessInfo.processInfo.environment["BAEPSAE_AX_TIMEOUT"]
+            .flatMap(Double.init) ?? 10.0)
+        AXUIElementSetMessagingTimeout(appElement, timeout)
+        return appElement
     }
 }
 
@@ -1482,9 +1499,17 @@ func activateTarget(_ target: TargetApp) throws {
         try activateSimulator(udid: udid)
     case .macApp(let pid, _, _):
         let apps = NSWorkspace.shared.runningApplications.filter { $0.processIdentifier == pid }
-        if let app = apps.first {
-            app.activate(options: [.activateAllWindows])
+        guard let app = apps.first else {
+            throw NativeError.commandFailed("App with pid \(pid) not found.")
+        }
+        app.activate(options: [.activateAllWindows])
+        // Poll for activation (max 1 second)
+        let deadline = Date().addingTimeInterval(1.0)
+        while Date() < deadline {
+            if app.isActive { return }
+            Thread.sleep(forTimeInterval: 0.05)
         }
+        fputs("Warning: app activation may not have completed within timeout\n", stderr)
     }
 }
 
@@ -1520,6 +1545,7 @@ func postMouseEvent(type: CGEventType, point: CGPoint, button: CGMouseButton = .
 
 func sendClick(at point: CGPoint) {
     postMouseEvent(type: .leftMouseDown, point: point)
+    usleep(20_000) // 20ms prevents some apps from ignoring instant clicks
     postMouseEvent(type: .leftMouseUp, point: point)
 }
 
@@ -1607,6 +1633,21 @@ func sendDrag(from start: CGPoint, to end: CGPoint, holdDuration: Double, moveDu
     postMouseEvent(type: .leftMouseUp, point: end)
 }
 
+// MARK: - Keycode to CGEventFlags Mapping
+
+let keycodeToFlag: [Int: CGEventFlags] = [
+    0x37: .maskCommand,      // Left Command (55)
+    0x36: .maskCommand,      // Right Command (54)
+    0x38: .maskShift,        // Left Shift (56)
+    0x3C: .maskShift,        // Right Shift (60)
+    0x3A: .maskAlternate,    // Left Option (58)
+    0x3D: .maskAlternate,    // Right Option (61)
+    0x3B: .maskControl,      // Left Control (59)
+    0x3E: .maskControl,      // Right Control (62)
+    0x39: .maskAlphaShift,   // Caps Lock (57)
+    0x3F: .maskSecondaryFn,  // Fn (63)
+]
+
 // MARK: - Keyboard Events
 
 func sendKeyPress(keyCode: Int, duration: Double?) {
@@ -1622,34 +1663,54 @@ func sendKeyPress(keyCode: Int, duration: Double?) {
 
 func sendKeyCombo(modifiers: [Int], key: Int) {
     let source = CGEventSource(stateID: .hidSystemState)
+
+    // Convert keycodes to CGEventFlags
+    var flags: CGEventFlags = []
+    for modifier in modifiers {
+        if let flag = keycodeToFlag[modifier] {
+            flags.insert(flag)
+        }
+    }
+
+    // Send modifier key-down events with flags (backward compat with apps that watch key events)
     for modifier in modifiers {
         let event = CGEvent(keyboardEventSource: source, virtualKey: CGKeyCode(modifier), keyDown: true)
+        event?.flags = flags
         event?.post(tap: .cghidEventTap)
     }
+    usleep(30_000) // 30ms for modifiers to register
+
+    // Ensure modifier key-up always happens (prevent stuck modifiers)
+    defer {
+        usleep(30_000)
+        for modifier in modifiers.reversed() {
+            let event = CGEvent(keyboardEventSource: source, virtualKey: CGKeyCode(modifier), keyDown: false)
+            event?.post(tap: .cghidEventTap)
+        }
+    }
+
+    // Main key with flags set (critical: many apps only check flags, not key events)
     let keyDown = CGEvent(keyboardEventSource: source, virtualKey: CGKeyCode(key), keyDown: true)
-    let keyUp = CGEvent(keyboardEventSource: source, virtualKey: CGKeyCode(key), keyDown: false)
+    keyDown?.flags = flags
     keyDown?.post(tap: .cghidEventTap)
+
+    let keyUp = CGEvent(keyboardEventSource: source, virtualKey: CGKeyCode(key), keyDown: false)
+    keyUp?.flags = flags
     keyUp?.post(tap: .cghidEventTap)
-    for modifier in modifiers.reversed() {
-        let event = CGEvent(keyboardEventSource: source, virtualKey: CGKeyCode(modifier), keyDown: false)
-        event?.post(tap: .cghidEventTap)
-    }
 }
 
-func sendText(_ text: String) {
+func sendText(_ text: String, charDelay: Double = 0.01) {
     let source = CGEventSource(stateID: .hidSystemState)
-    for scalar in text.unicodeScalars {
-        let value = scalar.value
-        guard value <= UInt16.max else {
-            continue
-        }
-        var char = UniChar(value)
+    for char in text {
+        var utf16Chars = Array(char.utf16)
         let keyDown = CGEvent(keyboardEventSource: source, virtualKey: 0, keyDown: true)
-        keyDown?.keyboardSetUnicodeString(stringLength: 1, unicodeString: &char)
+        keyDown?.keyboardSetUnicodeString(stringLength: utf16Chars.count, unicodeString: &utf16Chars)
         keyDown?.post(tap: .cghidEventTap)
         let keyUp = CGEvent(keyboardEventSource: source, virtualKey: 0, keyDown: false)
-        keyUp?.keyboardSetUnicodeString(stringLength: 1, unicodeString: &char)
         keyUp?.post(tap: .cghidEventTap)
+        if charDelay > 0 {
+            Thread.sleep(forTimeInterval: charDelay)
+        }
     }
 }
 

package/native/Sources/Version.swift

@@ -1 +1 @@
-let BAEPSAE_VERSION = "6.2.0"
+let BAEPSAE_VERSION = "6.3.0"