ten31-transcripts/Ten31Transcripts/Detection/CallDetector.swift

import AppKit
import CoreGraphics
import Combine

/// Detects when the user joins/leaves a call and reports it via callbacks.
///
/// Heuristic: the mic is live system-wide AND a known call app is present —
/// Zoom/Teams/Signal by bundle ID, or Google Meet by a browser window whose
/// title looks like a Meet call (read via `CGWindowList`, using the Screen
/// Recording permission). Debounced so a quick unrelated mic use doesn't trigger.
///
/// Main-actor: all evaluation runs on the main thread.
@MainActor
final class CallDetector: ObservableObject {

    enum DetectedApp: String, Equatable {
        case zoom, teams, signal, meet
        var label: String { rawValue }
        var display: String {
            switch self {
            case .zoom:   return "Zoom"
            case .teams:  return "Microsoft Teams"
            case .signal: return "Signal"
            case .meet:   return "Google Meet"
            }
        }
    }

    /// A detected call plus what to capture for visuals: the bundle ID of the owner
    /// (the native app for Zoom/Teams/Signal, or the *browser* hosting the Meet tab)
    /// and — for Meet — the exact `CGWindowID` of the matched call window, so the
    /// observer captures that window instead of guessing the browser's largest one.
    struct DetectedCall: Equatable {
        let app: DetectedApp
        let bundleID: String
        let windowID: CGWindowID?
    }

    enum Status: Equatable {
        case disabled
        case listening
        case inCall(DetectedApp)
    }

    @Published private(set) var status: Status = .disabled

    var onCallStart: ((DetectedCall) -> Void)?
    var onCallEnd: (() -> Void)?

    private let mic = MicActivityMonitor()
    private var pollTimer: Timer?
    private var openTimer: Timer?
    private var closeTimer: Timer?
    private var inCall = false
    private var currentCall: DetectedCall?
    private var enabled = false

    private let openDelay: TimeInterval = 2.0
    private let closeDelay: TimeInterval = 4.0
    private let pollInterval: TimeInterval = 3.0

    private static let nativeApps: [(id: String, app: DetectedApp)] = [
        ("us.zoom.xos", .zoom),
        ("com.microsoft.teams2", .teams),
        ("com.microsoft.teams", .teams),
        ("org.whispersystems.signal-desktop", .signal),
    ]
    private static let browserIDs: Set<String> = [
        "org.mozilla.firefox", "com.google.Chrome", "com.apple.Safari",
        "company.thebrowser.Browser", "com.brave.Browser", "com.microsoft.edgemac",
    ]

    func enable() {
        guard !enabled else { return }
        enabled = true
        mic.onChange = { [weak self] _ in self?.evaluate() }
        mic.start()
        status = .listening
        pollTimer = Timer.scheduledTimer(withTimeInterval: pollInterval, repeats: true) { [weak self] _ in
            Task { @MainActor in self?.evaluate() }
        }
        evaluate()
    }

    func disable() {
        guard enabled else { return }
        enabled = false
        mic.stop()
        pollTimer?.invalidate(); pollTimer = nil
        cancelOpen(); cancelClose()
        inCall = false
        currentCall = nil
        status = .disabled
    }

    // MARK: - Evaluation

    private func evaluate() {
        guard enabled else { return }
        let candidate = mic.isRunning ? detectApp() : nil

        if let candidate {
            cancelClose()
            if inCall {
                currentCall = candidate
                status = .inCall(candidate.app)
            } else if openTimer == nil {
                openTimer = Timer.scheduledTimer(withTimeInterval: openDelay, repeats: false) { [weak self] _ in
                    Task { @MainActor in self?.fireOpen() }
                }
            }
        } else {
            cancelOpen()
            if inCall && closeTimer == nil {
                closeTimer = Timer.scheduledTimer(withTimeInterval: closeDelay, repeats: false) { [weak self] _ in
                    Task { @MainActor in self?.fireClose() }
                }
            }
        }
    }

    private func fireOpen() {
        openTimer = nil
        // Re-resolve the app at fire time (the debounce window may have changed it).
        guard enabled, mic.isRunning, let call = detectApp(), !inCall else { return }
        inCall = true
        currentCall = call
        status = .inCall(call.app)
        onCallStart?(call)
    }

    private func fireClose() {
        closeTimer = nil
        guard enabled, inCall else { return }
        inCall = false
        currentCall = nil
        status = .listening
        onCallEnd?()
    }

    private func cancelOpen() { openTimer?.invalidate(); openTimer = nil }
    private func cancelClose() { closeTimer?.invalidate(); closeTimer = nil }

    // MARK: - App detection

    /// A call is active when a known call app is actually using the mic.
    /// On macOS 14+ we attribute mic usage per-process (robust start AND stop,
    /// works for Signal/Zoom/Teams/Meet, ignores our own recording). On macOS 13
    /// we fall back to the per-app call-window heuristic.
    private func detectApp() -> DetectedCall? {
        if #available(macOS 14.0, *) {
            return detectViaMicAttribution()
        }
        return detectViaWindowTitle()
    }

    @available(macOS 14.0, *)
    private func detectViaMicAttribution() -> DetectedCall? {
        let micPIDs = AudioInputProcesses.micUsingPIDs()
        guard !micPIDs.isEmpty else { return nil }
        let selfPID = NSRunningApplication.current.processIdentifier

        for app in NSWorkspace.shared.runningApplications {
            let pid = app.processIdentifier
            guard pid != selfPID, micPIDs.contains(pid), let id = app.bundleIdentifier else { continue }
            if let native = Self.nativeApps.first(where: { $0.id == id }) {
                return DetectedCall(app: native.app, bundleID: id, windowID: nil)   // native: capture largest owned window
            }
            // A browser using the mic + a Meet window = a Meet call. The mic state
            // gives reliable start/stop; the window check keeps non-Meet browser
            // mic use (other web apps) from being mislabeled as a Meet recording.
            // Capture that exact browser window (by ID), not just the browser.
            if Self.browserIDs.contains(id), let wid = meetWindowID(pid) {
                return DetectedCall(app: .meet, bundleID: id, windowID: wid)
            }
        }
        return nil
    }

    /// The `CGWindowID` of this PID's Google Meet call window (title "Meet - …"),
    /// or nil if none — also serves as the "is this a Meet call?" check.
    private func meetWindowID(_ pid: pid_t) -> CGWindowID? {
        guard let info = CGWindowListCopyWindowInfo([.excludeDesktopElements], kCGNullWindowID) as? [[String: Any]]
        else { return nil }
        for w in info {
            guard let wpid = w[kCGWindowOwnerPID as String] as? pid_t, wpid == pid,
                  let title = w[kCGWindowName as String] as? String, Self.looksLikeMeet(title) else { continue }
            return w[kCGWindowNumber as String] as? CGWindowID
        }
        return nil
    }

    /// macOS 13 fallback: detect by the presence of a call WINDOW per app.
    private func detectViaWindowTitle() -> DetectedCall? {
        var pidToApp: [pid_t: (app: DetectedApp, id: String)] = [:]
        var browserPIDs: [pid_t: String] = [:]
        for app in NSWorkspace.shared.runningApplications {
            guard let id = app.bundleIdentifier else { continue }
            if let native = Self.nativeApps.first(where: { $0.id == id }) {
                pidToApp[app.processIdentifier] = (native.app, id)
            } else if Self.browserIDs.contains(id) {
                browserPIDs[app.processIdentifier] = id
            }
        }
        guard !pidToApp.isEmpty || !browserPIDs.isEmpty else { return nil }
        guard let infoList = CGWindowListCopyWindowInfo([.excludeDesktopElements], kCGNullWindowID) as? [[String: Any]] else {
            return nil
        }
        for info in infoList {
            guard let pid = info[kCGWindowOwnerPID as String] as? pid_t,
                  let title = info[kCGWindowName as String] as? String,
                  !title.isEmpty else { continue }
            if let id = browserPIDs[pid], Self.looksLikeMeet(title) {
                return DetectedCall(app: .meet, bundleID: id,
                                    windowID: info[kCGWindowNumber as String] as? CGWindowID)
            }
            if let native = pidToApp[pid], Self.isCallWindow(native.app, title) {
                return DetectedCall(app: native.app, bundleID: native.id, windowID: nil)
            }
        }
        return nil
    }

    /// Per-app in-call window-title signatures (macOS 13 fallback only).
    private static func isCallWindow(_ app: DetectedApp, _ title: String) -> Bool {
        let t = title.lowercased()
        switch app {
        case .zoom:   return t.contains("zoom meeting") || t.contains("meeting")
        case .teams:  return t.contains("meeting")
        case .signal: return t.contains("signal call") || t.contains("group call")
        case .meet:   return false   // handled via the browser path above
        }
    }

    /// Match an ACTIVE Google Meet call. Verified against real Firefox titles:
    /// in a call the title is "Meet - <code>" (e.g. "Meet - tjh-pixe-ier"), while
    /// the home/lobby/"you left" pages are bare "Meet" or "Google Meet". Matching
    /// only the "Meet - …" form is what lets auto-STOP fire when you leave (and
    /// avoids false-starting on the home page). Also excludes "Zoom Meeting" etc.
    private static func looksLikeMeet(_ title: String) -> Bool {
        let t = title.lowercased()
        return t.hasPrefix("meet - ") || t.hasPrefix("meet – ") || t.hasPrefix("meet — ")
    }
}