soniox-mobile-to-mac/mac/src/AppDelegate.swift

import Cocoa
import AVFoundation
import ApplicationServices
import Network

public final class AppDelegate: NSObject, NSApplicationDelegate {
    private var statusBarController: StatusBarController!
    private var audioRecorder = AudioRecorder()
    private var activeSession: SonioxLiveSession?
    private var isBusyFinalizing = false
    private var currentAudioLevel: Float = 0.0
    private var latestPartialText: String? = nil
    
    // Remote Android Phone Local Receiver (Port 8999)
    private var remoteListener: NWListener?
    
    public func applicationDidFinishLaunching(_ notification: Notification) {
        // Completely silent operation by default
        UserDefaults.standard.set(false, forKey: "SonioxPlaySounds")
        
        statusBarController = StatusBarController()
        statusBarController.onToggleRecording = { [weak self] in
            self?.toggleRecording()
        }
        
        // Push-To-Talk / Toggle Hotkey Setup
        HotkeyManager.shared.onHotkeyPressed = { [weak self] in
            guard let self = self else { return }
            if HotkeyManager.shared.currentMode == .pushToTalk {
                if !self.audioRecorder.isRecording {
                    self.startRecording()
                }
            } else {
                self.toggleRecording()
            }
        }
        
        HotkeyManager.shared.onHotkeyReleased = { [weak self] in
            guard let self = self else { return }
            if HotkeyManager.shared.currentMode == .pushToTalk {
                if self.audioRecorder.isRecording {
                    self.stopRecordingAndTranscribe()
                }
            }
        }
        
        audioRecorder.onAudioLevelUpdate = { [weak self] level in
            guard let self = self else { return }
            self.currentAudioLevel = level
            if self.audioRecorder.isRecording {
                HUDOverlayController.shared.show(state: .recording(level: level, liveText: self.latestPartialText))
            }
        }
        
        audioRecorder.onAudioChunkAvailable = { [weak self] chunk in
            self?.activeSession?.sendAudioChunk(chunk)
        }
        
        SonioxSessionPool.shared.prewarmNextSession()
        startRemotePasteServer()
        
        // Connect to Linux Persistent Gateway (via local tunnel 18999 -> 8999)
        RelayClient.shared.onRemoteUpdateReceived = { text, cursor, isFullReplace in
            print("AppDelegate: 📥 Clean Remote Input Update: '\(text.prefix(30))...' (replace: \(isFullReplace))")
            
            // 1. FIRST apply the remote update directly into the active input while window is in pristine focus!
            FocusedInputSync.shared.applyRemoteUpdate(text: text, cursor: cursor, isFullReplace: isFullReplace)
            
            // 2. THEN show the floating success HUD notification
            DispatchQueue.main.asyncAfter(deadline: .now() + 0.05) {
                HUDOverlayController.shared.show(state: .success(text: text))
            }
        }
        RelayClient.shared.start()
        
        checkInitialPermissions()
    }
    
    private func startRemotePasteServer() {
        do {
            let port: NWEndpoint.Port = 8999
            let listener = try NWListener(using: .tcp, on: port)
            
            listener.newConnectionHandler = { [weak self] connection in
                guard let self = self else { return }
                connection.start(queue: .main)
                self.handleRemoteConnection(connection)
            }
            
            listener.start(queue: .main)
            self.remoteListener = listener
            print("RemotePasteServer: Listening on 0.0.0.0:8999")
        } catch {
            print("RemotePasteServer: Failed to bind port 8999:", error)
        }
    }
    
    private func handleRemoteConnection(_ connection: NWConnection) {
        connection.receive(minimumIncompleteLength: 1, maximumLength: 65536) { data, _, _, _ in
            guard let data = data, let reqStr = String(data: data, encoding: .utf8) else {
                connection.cancel()
                return
            }
            
            if reqStr.contains("GET /health") || reqStr.contains("GET /status") {
                let response = "HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nAccess-Control-Allow-Origin: *\r\nContent-Length: 15\r\n\r\n{\"status\":\"ok\"}"
                connection.send(content: response.data(using: .utf8), completion: .contentProcessed({ _ in
                    connection.cancel()
                }))
                return
            }
            
            if reqStr.contains("POST /paste") {
                if let bodyRange = reqStr.range(of: "\r\n\r\n") {
                    let bodyJsonStr = String(reqStr[bodyRange.upperBound...])
                    if let bodyData = bodyJsonStr.data(using: .utf8),
                       let json = try? JSONSerialization.jsonObject(with: bodyData) as? [String: Any],
                       let text = json["text"] as? String {
                        
                        DispatchQueue.main.async {
                            FocusedInputSync.shared.applyRemoteUpdate(text: text, cursor: nil, isFullReplace: true)
                        }
                    }
                }
                
                let response = "HTTP/1.1 200 OK\r\nContent-Type: application/json\r\nAccess-Control-Allow-Origin: *\r\nContent-Length: 19\r\n\r\n{\"status\":\"pasted\"}"
                connection.send(content: response.data(using: .utf8), completion: .contentProcessed({ _ in
                    connection.cancel()
                }))
                return
            }
            
            connection.cancel()
        }
    }
    
    private func checkInitialPermissions() {
        audioRecorder.requestMicrophonePermission { granted in
            if !granted {
                print("Warning: Microphone permission not granted.")
            }
        }
    }
    
    public func toggleRecording() {
        if audioRecorder.isRecording {
            stopRecordingAndTranscribe()
        } else {
            startRecording()
        }
    }
    
    public func startRecording() {
        guard !audioRecorder.isRecording, !isBusyFinalizing else { return }
        
        latestPartialText = nil
        
        let session = SonioxSessionPool.shared.acquireSession()
        self.activeSession = session
        
        session.onPartialText = { [weak self] liveText in
            guard let self = self, self.audioRecorder.isRecording else { return }
            self.latestPartialText = liveText
            HUDOverlayController.shared.show(state: .recording(level: self.currentAudioLevel, liveText: liveText))
        }
        
        session.onFinalResult = { [weak self] result in
            guard let self = self else { return }
            self.isBusyFinalizing = false
            self.activeSession = nil
            SonioxSessionPool.shared.prewarmNextSession()
            
            switch result {
            case .success(let text):
                FocusedInputSync.shared.applyRemoteUpdate(text: text, cursor: nil, isFullReplace: false)
            case .failure(let error):
                print("Soniox error:", error)
                HUDOverlayController.shared.show(state: .error(message: error.localizedDescription))
            }
        }
        
        do {
            try audioRecorder.startRecording()
            statusBarController.updateIcon(state: .recording)
            statusBarController.buildMenu(isRecording: true)
            HUDOverlayController.shared.show(state: .recording(level: 0.0, liveText: nil))
        } catch {
            HUDOverlayController.shared.show(state: .error(message: error.localizedDescription))
        }
    }
    
    public func stopRecordingAndTranscribe() {
        guard audioRecorder.isRecording, !isBusyFinalizing else { return }
        isBusyFinalizing = true
        
        _ = audioRecorder.stopRecording()
        statusBarController.updateIcon(state: .transcribing)
        statusBarController.buildMenu(isRecording: false)
        HUDOverlayController.shared.show(state: .transcribing)
        
        guard let session = self.activeSession else {
            self.isBusyFinalizing = false
            HUDOverlayController.shared.hide(animated: true)
            return
        }
        
        session.finalizeStream()
    }
}