Files
meeting-assistant/MeetingAssistant/Native/MacOsDesktopControls/main.swift
T
dh 728e66dd72
PR and Push Build/Test / windows-build-and-test (push) Failing after 38m25s
PR and Push Build/Test / portable-build-and-test (push) Successful in 8m9s
PR and Push Build/Test / macos-native-full (push) Skipped
ci: consolidate platform pipelines and test prerequisites on macOS support
2026-10-06 09:02:19 +02:00

517 lines
21 KiB
Swift

import AppKit
import Carbon
import Foundation
import WebKit
import UserNotifications
struct Manifest: Decodable {
let baseUrl: String
let statusUrl: String
let openAgentUrl: String
let profiles: [Profile]
let hotkeys: [Hotkey]
}
struct Profile: Decodable {
let name: String
let toggleShortcut: String
}
struct Hotkey: Decodable {
let shortcut: String
let endpoint: String
}
enum RecordingState: Int, Decodable {
case idle = 0
case recording = 1
case summarizing = 2
}
struct RecordingStatus: Decodable {
let isRecording: Bool
let state: RecordingState
let launchProfile: String?
}
@main
@MainActor
final class DesktopControls: NSObject, NSApplicationDelegate, NSMenuDelegate, UNUserNotificationCenterDelegate {
private let manifest: Manifest
private var status = RecordingStatus(isRecording: false, state: .idle, launchProfile: nil)
private var statusItem: NSStatusItem!
private var registeredHotkeys: [UInt32: Hotkey] = [:]
private var timer: Timer?
private var agentWindow: NSWindow?
private var independentAgentWindows: [String: NSWindow] = [:]
private var agentWindowState = DesktopAgentWindowState()
private var requests = DesktopUserRequestState()
private var fetchingRequests = false
private var fetchingWindows = false
private var authorizationInFlight = false
private var authorizationChecked = false
private var notificationsAuthorized = false
private lazy var notificationCenter = UNUserNotificationCenter.current()
override init() {
guard CommandLine.arguments.count == 2,
let data = try? Data(contentsOf: URL(fileURLWithPath: CommandLine.arguments[1])),
let decoded = try? JSONDecoder().decode(Manifest.self, from: data) else {
fputs("Usage: macos-desktop-controls <manifest.json>\n", stderr)
exit(2)
}
manifest = decoded
super.init()
}
static func main() {
if CommandLine.arguments.count == 2 && CommandLine.arguments[1] == "--self-test" {
runDesktopControlSelfTests()
return
}
// Notification activation after shutdown has no trusted service manifest.
if CommandLine.arguments.count == 1 {
fputs("Ignoring desktop notification activation without an active service manifest.\n", stderr)
return
}
guard Bundle.main.bundleIdentifier == "cloud.schweigert.meeting-assistant.desktop-controls" else {
fputs("Desktop controls must run from the signed MeetingAssistantDesktopControls.app bundle.\n", stderr)
exit(2)
}
let application = NSApplication.shared
let delegate = DesktopControls()
application.delegate = delegate
application.setActivationPolicy(.accessory)
application.run()
}
func applicationWillFinishLaunching(_ notification: Notification) {
notificationCenter.delegate = self
notificationCenter.removeAllPendingNotificationRequests()
notificationCenter.removeAllDeliveredNotifications()
}
func applicationDidFinishLaunching(_ notification: Notification) {
statusItem = NSStatusBar.system.statusItem(withLength: NSStatusItem.squareLength)
statusItem.button?.image = NSImage(
systemSymbolName: "waveform.circle",
accessibilityDescription: "Meeting Assistant")
let menu = NSMenu()
menu.delegate = self
statusItem.menu = menu
registerHotkeys()
refresh()
timer = Timer.scheduledTimer(withTimeInterval: 2, repeats: true) { [weak self] _ in
DispatchQueue.main.async { self?.refresh() }
}
}
func applicationWillTerminate(_ notification: Notification) {
timer?.invalidate()
notificationCenter.removeAllPendingNotificationRequests()
notificationCenter.removeAllDeliveredNotifications()
}
func menuWillOpen(_ menu: NSMenu) {
removeNotifications(requests.reconcile(requests.pending, now: Date()))
rebuildMenu(menu)
refresh()
}
private func rebuildMenu(_ menu: NSMenu) {
menu.removeAllItems()
let stateItem = NSMenuItem(title: tooltip, action: nil, keyEquivalent: "")
stateItem.isEnabled = false
menu.addItem(stateItem)
menu.addItem(.separator())
menu.addItem(item("Open agent", #selector(openAgent)))
if status.isRecording {
menu.addItem(item("Stop meeting recording and transcribe", #selector(stopRecording)))
menu.addItem(item("Cancel meeting recording and discard", #selector(abortRecording)))
} else {
for (index, profile) in manifest.profiles.enumerated() {
let entry = item(
"Start meeting recording (\(profile.name))",
#selector(startProfile(_:)))
entry.tag = index
entry.keyEquivalent = menuKey(from: profile.toggleShortcut)
entry.keyEquivalentModifierMask = menuModifiers(from: profile.toggleShortcut)
menu.addItem(entry)
}
}
menu.addItem(.separator())
let pending = NSMenuItem(title: "Pending requests (\(requests.pending.count))", action: nil, keyEquivalent: "")
let pendingMenu = NSMenu()
if requests.pending.isEmpty {
let empty = NSMenuItem(title: "No pending requests", action: nil, keyEquivalent: "")
empty.isEnabled = false
pendingMenu.addItem(empty)
}
for request in requests.pending {
let entry = NSMenuItem(title: request.title, action: nil, keyEquivalent: "")
entry.toolTip = request.message
let decisions = NSMenu()
let message = NSMenuItem(title: request.message, action: nil, keyEquivalent: "")
message.isEnabled = false
decisions.addItem(message)
decisions.addItem(.separator())
for action in request.actions {
let decision = item(action.label, #selector(answerRequest(_:)))
decision.representedObject = DesktopRequestDecision(request.id, action.id)
decision.isEnabled = !requests.isResponding(request.id)
decisions.addItem(decision)
}
entry.submenu = decisions
pendingMenu.addItem(entry)
}
pending.submenu = pendingMenu
menu.addItem(pending)
menu.addItem(.separator())
menu.addItem(item("Exit", #selector(exitApplication)))
}
private var tooltip: String {
let suffix = requests.pending.isEmpty ? "" : " — \(requests.pending.count) pending request(s)"
if status.isRecording {
return (status.launchProfile.map { "Meeting Assistant — recording (\($0))" }
?? "Meeting Assistant — recording") + suffix
}
return (status.state == .summarizing
? "Meeting Assistant — summarizing"
: "Meeting Assistant — idle") + suffix
}
private func item(_ title: String, _ action: Selector) -> NSMenuItem {
let result = NSMenuItem(title: title, action: action, keyEquivalent: "")
result.target = self
return result
}
@objc private func openAgent() {
if agentWindow == nil, let url = URL(string: manifest.openAgentUrl) {
agentWindow = makeAgentWindow(url)
}
NSApplication.shared.activate(ignoringOtherApps: true)
agentWindow?.makeKeyAndOrderFront(nil)
}
private func makeAgentWindow(_ url: URL) -> NSWindow {
let webView = WKWebView(frame: .zero)
let window = NSWindow(
contentRect: NSRect(x: 0, y: 0, width: 840, height: 760),
styleMask: [.titled, .closable, .miniaturizable, .resizable],
backing: .buffered,
defer: false)
window.title = "Meeting Summary Agent"
window.contentView = webView
window.center()
window.isReleasedWhenClosed = false
webView.load(URLRequest(url: url))
return window
}
@objc private func answerRequest(_ sender: NSMenuItem) {
guard let decision = sender.representedObject as? DesktopRequestDecision else { return }
submitResponse(requestId: decision.requestId, action: decision.action)
}
private func submitResponse(requestId: String, action: String) {
guard let response = requests.beginResponse(baseUrl: manifest.baseUrl, requestId: requestId, action: action, now: Date()) else { return }
URLSession.shared.dataTask(with: response) { [weak self] _, response, error in
let accepted = error == nil && (response as? HTTPURLResponse).map { (200..<300).contains($0.statusCode) } == true
DispatchQueue.main.async {
guard let self else { return }
self.requests.completeResponse(requestId, accepted: accepted)
if accepted { self.removeNotifications([requestId]) }
if let error { desktopLog("User response failed: \(error.localizedDescription)") }
self.refresh()
}
}.resume()
}
nonisolated func userNotificationCenter(_ center: UNUserNotificationCenter, willPresent notification: UNNotification, withCompletionHandler completionHandler: @escaping (UNNotificationPresentationOptions) -> Void) {
completionHandler([.banner, .list, .sound])
}
nonisolated func userNotificationCenter(_ center: UNUserNotificationCenter, didReceive response: UNNotificationResponse, withCompletionHandler completionHandler: @escaping () -> Void) {
// Closing the announcement is independent of deciding its menu request.
guard response.actionIdentifier != UNNotificationDismissActionIdentifier,
response.actionIdentifier != UNNotificationDefaultActionIdentifier,
let id = response.notification.request.content.userInfo["requestId"] as? String else {
completionHandler()
return
}
let action = response.actionIdentifier
DispatchQueue.main.async { [weak self] in self?.submitResponse(requestId: id, action: action) }
completionHandler()
}
@objc private func startProfile(_ sender: NSMenuItem) {
let profile = manifest.profiles[sender.tag]
post(path: "/profiles/\(escaped(profile.name))/recording/start")
}
@objc private func stopRecording() {
post(path: "/recording/stop")
}
@objc private func abortRecording() {
post(path: "/recording/abort")
}
@objc private func exitApplication() {
post(path: "/application/exit")
DispatchQueue.main.asyncAfter(deadline: .now() + 0.3) {
NSApplication.shared.terminate(nil)
}
}
private func refresh() {
removeNotifications(requests.reconcile(requests.pending, now: Date()))
updateStatusItem()
refreshStatus()
refreshRequests()
refreshAgentWindows()
}
private func updateStatusItem() {
statusItem.button?.image = NSImage(
systemSymbolName: status.isRecording ? "record.circle.fill" : "waveform.circle",
accessibilityDescription: tooltip)
statusItem.button?.toolTip = tooltip
}
private func refreshRequests() {
guard !fetchingRequests, let url = URL(string: manifest.baseUrl + "/desktop/requests") else { return }
fetchingRequests = true
URLSession.shared.dataTask(with: url) { [weak self] data, response, error in
var snapshot: [DesktopUserRequest]?
if let data, error == nil, (response as? HTTPURLResponse)?.statusCode == 200 {
do { snapshot = try decodeDesktopRequests(data) }
catch { desktopLog("Cannot decode pending desktop requests: \(error)") }
}
DispatchQueue.main.async {
guard let self else { return }
self.fetchingRequests = false
guard let snapshot else { return }
self.removeNotifications(self.requests.reconcile(snapshot, now: Date()))
self.updateStatusItem()
self.announcePendingRequests()
}
}.resume()
}
private func removeNotifications(_ ids: [String]) {
guard !ids.isEmpty else { return }
let identifiers = ids.map { "desktop-request." + $0 }
notificationCenter.removePendingNotificationRequests(withIdentifiers: identifiers)
notificationCenter.removeDeliveredNotifications(withIdentifiers: identifiers)
}
private func announcePendingRequests() {
guard !requests.notificationsToAnnounce().isEmpty else { return }
if !authorizationChecked {
guard !authorizationInFlight else { return }
authorizationInFlight = true
notificationCenter.getNotificationSettings { [weak self] settings in
DispatchQueue.main.async {
guard let self else { return }
if settings.authorizationStatus == .notDetermined {
self.notificationCenter.requestAuthorization(options: [.alert, .sound]) { [weak self] granted, error in
DispatchQueue.main.async {
if let error { desktopLog("Notification authorization failed: \(error.localizedDescription)") }
self?.finishNotificationAuthorization(granted)
}
}
} else {
self.finishNotificationAuthorization(settings.authorizationStatus == .authorized || settings.authorizationStatus == .provisional)
}
}
}
return
}
guard notificationsAuthorized else { return }
notificationCenter.setNotificationCategories(Set(requests.pending.map { $0.notificationCategory() }))
for request in requests.notificationsToAnnounce() {
requests.markAnnounced(request.id)
notificationCenter.add(request.notificationRequest()) { error in
if let error { desktopLog("Cannot schedule desktop notification \(request.id): \(error.localizedDescription)") }
else { desktopLog("Scheduled desktop notification \(request.id)") }
}
}
}
private func finishNotificationAuthorization(_ granted: Bool) {
authorizationChecked = true
authorizationInFlight = false
notificationsAuthorized = granted
desktopLog(granted ? "Desktop notifications authorized." : "Desktop notifications unavailable; pending requests remain in the menu.")
removeNotifications(requests.reconcile(requests.pending, now: Date()))
announcePendingRequests()
}
private func refreshAgentWindows() {
guard !fetchingWindows, let url = URL(string: manifest.baseUrl + "/desktop/agent-windows") else { return }
fetchingWindows = true
URLSession.shared.dataTask(with: url) { [weak self] data, response, error in
let windows = data.flatMap { data in
error == nil && (response as? HTTPURLResponse)?.statusCode == 200
? try? JSONDecoder().decode([DesktopAgentWindow].self, from: data) : nil
}
DispatchQueue.main.async {
guard let self else { return }
self.fetchingWindows = false
guard let windows else { return }
for requested in self.agentWindowState.takeUnopened(windows) {
guard let url = URL(string: requested.url) else { continue }
let window = self.makeAgentWindow(url)
self.independentAgentWindows[requested.id] = window
NSApplication.shared.activate(ignoringOtherApps: true)
window.makeKeyAndOrderFront(nil)
desktopLog("Opened independent workflow agent window \(requested.id)")
}
}
}.resume()
}
private func refreshStatus() {
guard let url = URL(string: manifest.statusUrl) else { return }
URLSession.shared.dataTask(with: url) { [weak self] data, _, _ in
guard let data,
let next = try? JSONDecoder().decode(RecordingStatus.self, from: data) else { return }
DispatchQueue.main.async {
self?.status = next
self?.updateStatusItem()
}
}.resume()
}
private func post(path: String) {
post(url: manifest.baseUrl + path)
}
private func post(url value: String) {
guard let url = URL(string: value) else { return }
var request = URLRequest(url: url)
request.httpMethod = "POST"
URLSession.shared.dataTask(with: request) { [weak self] _, _, _ in
DispatchQueue.main.asyncAfter(deadline: .now() + 0.2) {
self?.refresh()
}
}.resume()
}
private func registerHotkeys() {
var eventType = EventTypeSpec(
eventClass: OSType(kEventClassKeyboard),
eventKind: OSType(kEventHotKeyPressed))
InstallEventHandler(
GetApplicationEventTarget(),
{ _, event, context in
guard let event, let context else { return noErr }
var identifier = EventHotKeyID()
GetEventParameter(
event,
EventParamName(kEventParamDirectObject),
EventParamType(typeEventHotKeyID),
nil,
MemoryLayout<EventHotKeyID>.size,
nil,
&identifier)
let owner = Unmanaged<DesktopControls>.fromOpaque(context).takeUnretainedValue()
if let hotkey = owner.registeredHotkeys[identifier.id] {
owner.post(path: hotkey.endpoint)
}
return noErr
},
1,
&eventType,
Unmanaged.passUnretained(self).toOpaque(),
nil)
for (index, hotkey) in manifest.hotkeys.enumerated() {
guard let keyCode = carbonKeyCode(from: hotkey.shortcut) else { continue }
let identifier = EventHotKeyID(signature: fourCC("MTGA"), id: UInt32(index + 1))
var reference: EventHotKeyRef?
let result = RegisterEventHotKey(
keyCode,
carbonModifiers(from: hotkey.shortcut),
identifier,
GetApplicationEventTarget(),
0,
&reference)
if result == noErr {
registeredHotkeys[identifier.id] = hotkey
}
}
}
private func escaped(_ value: String) -> String {
value.addingPercentEncoding(withAllowedCharacters: .urlPathAllowed) ?? value
}
}
private final class DesktopRequestDecision: NSObject {
let requestId: String
let action: String
init(_ requestId: String, _ action: String) {
self.requestId = requestId
self.action = action
}
}
private func desktopLog(_ message: String) {
fputs("\(message)\n", stderr)
}
private func tokens(_ shortcut: String) -> [String] {
shortcut.uppercased().split(separator: "+").map(String.init)
}
private func carbonModifiers(from shortcut: String) -> UInt32 {
let parts = tokens(shortcut)
var result: UInt32 = 0
if parts.contains("CTRL") || parts.contains("CONTROL") { result |= UInt32(controlKey) }
if parts.contains("ALT") || parts.contains("OPTION") { result |= UInt32(optionKey) }
if parts.contains("SHIFT") { result |= UInt32(shiftKey) }
if parts.contains("WIN") || parts.contains("WINDOWS") || parts.contains("CMD") || parts.contains("COMMAND") {
result |= UInt32(cmdKey)
}
return result
}
private func menuModifiers(from shortcut: String) -> NSEvent.ModifierFlags {
let parts = tokens(shortcut)
var result: NSEvent.ModifierFlags = []
if parts.contains("CTRL") || parts.contains("CONTROL") { result.insert(.control) }
if parts.contains("ALT") || parts.contains("OPTION") { result.insert(.option) }
if parts.contains("SHIFT") { result.insert(.shift) }
if parts.contains("WIN") || parts.contains("WINDOWS") || parts.contains("CMD") || parts.contains("COMMAND") {
result.insert(.command)
}
return result
}
private func menuKey(from shortcut: String) -> String {
tokens(shortcut).last?.lowercased() ?? ""
}
private func carbonKeyCode(from shortcut: String) -> UInt32? {
guard let key = tokens(shortcut).last, key.count == 1, let scalar = key.unicodeScalars.first else {
return nil
}
let codes: [Character: UInt32] = [
"A": 0, "S": 1, "D": 2, "F": 3, "H": 4, "G": 5, "Z": 6, "X": 7,
"C": 8, "V": 9, "B": 11, "Q": 12, "W": 13, "E": 14, "R": 15,
"Y": 16, "T": 17, "M": 46, "L": 37, "N": 45, "P": 35, "O": 31,
"I": 34, "U": 32, "J": 38, "K": 40
]
return codes[Character(String(scalar))]
}
private func fourCC(_ value: String) -> OSType {
value.utf8.reduce(0) { ($0 << 8) + OSType($1) }
}