deepvoice_client/local_plugins/agent_service/ios/Classes/AgentServiceImpl.swift

import Foundation
import AVFoundation
import ble_service
import speech

/// 代理服务事件监听器接口
protocol AgentServiceListener {
    /// 当事件发生时调用
    /// - Parameters:
    ///   - eventName: 事件名称
    ///   - data: 事件数据
    func onEvent(eventName: String, data: [String: Any])
}

/// 代理服务实现类，负责语音识别、OpenAI对话和语音合成
class AgentServiceImpl: NSObject {
    // MARK: - 常量
    private let TAG = "AgentServiceImpl"
    
    // MARK: - 单例实现
    static let shared = AgentServiceImpl()
    
    // MARK: - 属性
    
    // 监听器集合
    private var listeners = Set<AgentServiceListener>()
    private let listenersLock = NSLock()
    
    // Azure 语音相关
    private var azureSpeechKey: String = ""
    private var azureSpeechRegion: String = ""
    private var isRecognizing: Bool = false
    private var isSpeaking: Bool = false
    
    // 语音识别和合成服务
    private var azureAsrHelper: AzureAsrHelper?
    private var azureTtsHelper: AzureTtsHelper?
    
    // OpenAI相关
    private var apiKey: String = ""
    private var baseUrl: String = "https://api.openai.com"
    private var model: String = "gpt-3.5-turbo"
    private var systemPrompt: String = ""
    private var chatHistory: [[String: Any]] = []
    
    // MCP服务相关
    private var mcpServer: String = ""
    
    // 状态标志
    private var isInitialized: Bool = false
    private var hasSpeechDetected: Bool = false
    private var isAiStreaming: Bool = false
    
    // 空闲检测相关
    private var idleCheckTimer: Timer?
    private let maxIdleSeconds: TimeInterval = 10 // 最大空闲秒数
    
    // 音频播放器
    private var audioPlayer: AudioPlayer?
    
    // MARK: - 初始化
    
    /// 私有初始化方法
    private override init() {
        super.init()
        
        // 设置音频会话
        setupAudioSession()
        
        // 初始化音频播放器
        audioPlayer = AudioPlayer()
        
        // 初始化语音助手服务
        azureAsrHelper = AzureAsrHelper()
        azureTtsHelper = AzureTtsHelper()
        
        // 设置BleService回调
        BleService.shared.setDelegate(self)
    }
    
    // MARK: - 监听器管理
    
    /// 添加事件监听器
    /// - Parameter listener: 要添加的监听器
    func addListener(_ listener: AgentServiceListener) {
        listenersLock.lock()
        listeners.insert(listener)
        listenersLock.unlock()
    }
    
    /// 移除事件监听器
    /// - Parameter listener: 要移除的监听器
    func removeListener(_ listener: AgentServiceListener) {
        listenersLock.lock()
        listeners.remove(listener)
        listenersLock.unlock()
    }
    
    /// 清除所有监听器
    func clearListeners() {
        listenersLock.lock()
        listeners.removeAll()
        listenersLock.unlock()
    }
    
    /// 向所有监听器发送事件
    /// - Parameters:
    ///   - eventName: 事件名称
    ///   - data: 事件数据
    private func sendEvent(name eventName: String, data: [String: Any]) {
        listenersLock.lock()
        let currentListeners = self.listeners
        listenersLock.unlock()
        
        for listener in currentListeners {
            listener.onEvent(eventName: eventName, data: data)
        }
    }
    
    /// 发送错误事件
    /// - Parameter message: 错误信息
    private func sendError(_ message: String, code: String = "ERROR") {
        sendEvent(name: "error", data: ["code": code, "message": message])
    }
    
    // MARK: - 初始化配置
    
    /// 初始化系统提示词
    private func initSystemPrompt() {
        systemPrompt = """
            你是小言，一个具备专业能力的智能助手，需根据用户场景灵活切换回答模式，确保服务精准高效。
            [系统提示词内容省略...]
            """
    }
    
    /// 初始化配置
    /// - Parameter config: 配置参数
    /// - Returns: 是否初始化成功
    func initialize(config: [String: Any]) -> Bool {
        if isInitialized { return true }
        
        // 设置OpenAI配置
        if let openaiApiKey = config["openaiApiKey"] as? String {
            self.apiKey = openaiApiKey
        } else {
            sendError("OpenAI API密钥缺失", code: "CONFIG_ERROR")
            return false
        }
        
        if let openaiBaseUrl = config["openaiBaseUrl"] as? String {
            self.baseUrl = openaiBaseUrl
        }
        
        if let openaiModel = config["openaiModel"] as? String {
            self.model = openaiModel
        }
        
        // 设置系统提示词
        if let customSystemPrompt = config["systemPrompt"] as? String, !customSystemPrompt.isEmpty {
            self.systemPrompt = customSystemPrompt
        } else {
            initSystemPrompt()
        }
        
        // 设置Azure语音配置
        if let azureSpeechKey = config["azureSpeechKey"] as? String {
            self.azureSpeechKey = azureSpeechKey
        } else {
            sendError("Azure语音密钥缺失", code: "CONFIG_ERROR")
            return false
        }
        
        if let azureSpeechRegion = config["azureSpeechRegion"] as? String {
            self.azureSpeechRegion = azureSpeechRegion
        } else {
            sendError("Azure语音区域缺失", code: "CONFIG_ERROR")
            return false
        }
        
        // 设置MCP服务
        if let mcpServer = config["mcpServer"] as? String {
            self.mcpServer = mcpServer
        }
        
        // 初始化Azure语音服务
        let asrInitSuccess = initializeAzureSpeech()
        
        isInitialized = asrInitSuccess
        return asrInitSuccess
    }
    
    // MARK: - 音频会话设置
    
    /// 设置音频会话
    private func setupAudioSession() {
        let audioSession = AVAudioSession.sharedInstance()
        do {
            try audioSession.setCategory(.playAndRecord, mode: .default, options: [.defaultToSpeaker, .allowBluetooth])
            try audioSession.setActive(true, options: .notifyOthersOnDeactivation)
        } catch {
            sendError("设置音频会话失败: \(error.localizedDescription)")
        }
    }
    
    // MARK: - Azure 语音服务
    
    /// 初始化Azure语音服务
    private func initializeAzureSpeech() -> Bool {
        // 添加事件监听器
        azureAsrHelper?.addListener(self)
        azureTtsHelper?.addListener(self)
        
        // 初始化ASR服务
        guard let asrSuccess = azureAsrHelper?.initialize(
            subscriptionKey: azureSpeechKey, 
            region: azureSpeechRegion, 
            supportedLanguages: ["zh-CN"]
        ), asrSuccess else {
            sendError("初始化语音识别服务失败", code: "ASR_INIT_ERROR")
            return false
        }
        
        // 初始化TTS服务
        guard let ttsSuccess = azureTtsHelper?.initialize(
            ttsAppId: "", // Azure TTS不需要appId
            ttsAppToken: azureSpeechKey,
            ttsResource: azureSpeechRegion,
            language: "zh-CN"
        ), ttsSuccess else {
            sendError("初始化语音合成服务失败", code: "TTS_INIT_ERROR")
            return false
        }
        
        // 设置TTS语音
        _ = azureTtsHelper?.setVoice("zh-CN-XiaoxiaoNeural")
        
        return true
    }
    
    // MARK: - 空闲检测
    
    /// 启动空闲检测
    private func startIdleCheck() {
        stopIdleCheck() // 先停止现有的检查
        
        if !isRecognizing { return }
        
        // 创建一个新的定时器
        idleCheckTimer = Timer.scheduledTimer(withTimeInterval: maxIdleSeconds, repeats: false) { [weak self] _ in
            guard let self = self else { return }
            
            if self.isRecognizing && !self.hasSpeechDetected && !self.isSpeaking && !self.isAiStreaming {
                self.stopRecognition()
                self.sendEvent(name: "auto_stop", data: [
                    "reason": "idle_timeout", 
                    "seconds": self.maxIdleSeconds
                ])
            }
        }
    }
    
    /// 停止空闲检测
    private func stopIdleCheck() {
        idleCheckTimer?.invalidate()
        idleCheckTimer = nil
    }
    
    /// 重启空闲检测
    private func restartIdleCheck() {
        if isRecognizing {
            startIdleCheck()
        }
    }
    
    // MARK: - 语音识别
    
    /// 开始语音识别
    /// - Returns: 是否成功启动识别
    func startRecognition() -> Bool {
        if !isInitialized {
            sendError("服务未初始化", code: "NOT_INITIALIZED")
            return false
        }
        
        // 防止重复启动
        if isRecognizing {
            stopRecognition()
        }
        
        // 启动语音识别
        guard let success = azureAsrHelper?.startContinuousRecognition(), success else {
            sendError("启动语音识别失败", code: "RECOGNITION_START_ERROR")
            return false
        }
        
        return true
    }
    
    /// 停止语音识别
    /// - Returns: 是否成功停止识别
    func stopRecognition() -> Bool {
        if !isRecognizing {
            return true
        }
        
        // 停止语音识别
        guard let success = azureAsrHelper?.stopContinuousRecognition(), success else {
            return false
        }
        
        return true
    }
    
    /// 推送音频数据
    /// - Parameter audioData: 音频数据
    /// - Returns: 是否成功推送
    func pushAudioData(_ audioData: Data) -> Bool {
        if !isRecognizing || !isInitialized {
            return false
        }
        
        // 将音频数据推送到语音识别引擎
        return azureAsrHelper?.pushAudioData(audioData: audioData) ?? false
    }
    
    // MARK: - 语音合成
    
    /// 语音合成播放文本
    /// - Parameter text: 要播放的文本
    /// - Returns: 是否成功开始播放
    func speakText(_ text: String) -> Bool {
        if !isInitialized {
            sendError("服务未初始化", code: "NOT_INITIALIZED")
            return false
        }
        
        if text.isEmpty {
            return false
        }
        
        // 使用TTS服务进行语音合成
        return azureTtsHelper?.speakOnce(text) ?? false
    }
    
    /// 停止语音合成
    /// - Returns: 是否成功停止合成
    func stopTts() -> Bool {
        if !isSpeaking {
            return true
        }
        
        // 停止TTS服务
        return azureTtsHelper?.stop() ?? false
    }
    
    /// 打断当前响应
    /// - Returns: 是否成功打断
    func interruptCurrentResponse() -> Bool {
        var interrupted = false
        
        // 停止TTS播放
        if isSpeaking {
            interrupted = stopTts() || interrupted
        }
        
        // 停止AI流输出
        if isAiStreaming {
            isAiStreaming = false
            interrupted = true
        }
        
        // 发送打断事件
        if interrupted {
            sendEvent(name: "response_interrupted", data: ["status": "interrupted"])
        }
        
        return interrupted
    }
    
    // MARK: - 文本处理
    
    /// 处理文本输入
    /// - Parameters:
    ///   - text: 文本内容
    ///   - speakResponse: 是否朗读响应
    /// - Returns: 是否成功处理
    func processTextInput(_ text: String, speakResponse: Bool) -> Bool {
        if !isInitialized {
            sendError("服务未初始化", code: "NOT_INITIALIZED")
            return false
        }
        
        if text.isEmpty {
            sendError("文本输入不能为空", code: "EMPTY_TEXT")
            return false
        }
        
        // TODO: 实现与OpenAI的通信
        // 这里只是模拟处理
        
        // 将用户消息添加到聊天历史
        let userMessage: [String: Any] = [
            "role": "user",
            "content": text
        ]
        chatHistory.append(userMessage)
        
        // 更新状态
        isAiStreaming = true
        
        // 模拟助手响应
        DispatchQueue.main.asyncAfter(deadline: .now() + 0.5) { [weak self] in
            guard let self = self else { return }
            
            // 模拟流式输出
            let response = "这是对"\(text)"的回复。我是AI助手，很高兴为您服务。"
            
            // 逐词输出模拟
            let words = response.split(separator: " ")
            for (index, word) in words.enumerated() {
                DispatchQueue.main.asyncAfter(deadline: .now() + Double(index) * 0.1) { [weak self] in
                    guard let self = self, self.isAiStreaming else { return }
                    self.sendEvent(name: "assistant_token", data: ["token": String(word) + " "])
                    
                    // 如果需要语音播报，则合成语音
                    if speakResponse {
                        self.azureTtsHelper?.speakStream(String(word) + " ")
                    }
                }
            }
            
            // 完整响应
            DispatchQueue.main.asyncAfter(deadline: .now() + Double(words.count) * 0.1 + 0.2) { [weak self] in
                guard let self = self else { return }
                
                // 结束AI流式输出
                self.isAiStreaming = false
                
                // 发送完整响应
                self.sendEvent(name: "assistant_response", data: ["text": response])
                
                // 将助手消息添加到聊天历史
                let assistantMessage: [String: Any] = [
                    "role": "assistant",
                    "content": response
                ]
                self.chatHistory.append(assistantMessage)
                
                // 如果需要语音播报，则完成语音流
                if speakResponse {
                    self.azureTtsHelper?.flushStream()
                }
            }
        }
        
        return true
    }
    
    // MARK: - 图片处理
    
    /// 处理图片输入
    /// - Parameters:
    ///   - imagePath: 图片路径
    ///   - text: 文本描述
    ///   - speakResponse: 是否朗读响应
    /// - Returns: 是否成功处理
    func processImageInput(imagePath: String, text: String, speakResponse: Bool) -> Bool {
        if !isInitialized {
            sendError("服务未初始化", code: "NOT_INITIALIZED")
            return false
        }
        
        if imagePath.isEmpty {
            sendError("图片路径不能为空", code: "EMPTY_IMAGE_PATH")
            return false
        }
        
        // 发送图片处理中事件
        sendEvent(name: "image_processing", data: [:])
        
        // TODO: 实现与OpenAI Vision API的通信
        // 这里只是模拟处理
        
        // 模拟图片处理
        DispatchQueue.main.asyncAfter(deadline: .now() + 1.0) { [weak self] in
            guard let self = self else { return }
            
            // 图片就绪事件
            self.sendEvent(name: "image_ready", data: [:])
            
            // 更新状态
            self.isAiStreaming = true
            
            // 模拟AI分析结果
            let imageDescription = "这是一张图片分析结果。"
            let additionalText = text.isEmpty ? "" : "关于您提问的"\(text)"，"
            let response = "\(imageDescription)\(additionalText)我可以看到这是一张图片。"
            
            // 发送AI助手token流
            let words = response.split(separator: " ")
            for (index, word) in words.enumerated() {
                DispatchQueue.main.asyncAfter(deadline: .now() + Double(index) * 0.1) { [weak self] in
                    guard let self = self, self.isAiStreaming else { return }
                    
                    let token = String(word) + " "
                    self.sendEvent(name: "assistant_token", data: ["token": token])
                    
                    // 如果需要语音播报，则合成语音
                    if speakResponse {
                        self.azureTtsHelper?.speakStream(token)
                    }
                }
            }
            
            // 发送完整响应
            DispatchQueue.main.asyncAfter(deadline: .now() + Double(words.count) * 0.1 + 0.2) { [weak self] in
                guard let self = self else { return }
                
                // 结束AI流式输出
                self.isAiStreaming = false
                
                self.sendEvent(name: "assistant_response", data: [
                    "text": response,
                    "userInput": text.isEmpty ? "[图片]" : text,
                    "hasImage": true
                ])
                
                // 如果需要语音播报，则完成语音
                if speakResponse {
                    self.azureTtsHelper?.flushStream()
                }
            }
        }
        
        return true
    }
    
    // MARK: - 聊天历史管理
    
    /// 清除聊天历史
    /// - Returns: 是否成功清除
    func clearChatHistory() -> Bool {
        chatHistory.removeAll()
        return true
    }
    
    // MARK: - 资源释放
    
    /// 释放资源
    /// - Returns: 是否成功释放
    func dispose() -> Bool {
        // 停止语音识别
        if isRecognizing {
            stopRecognition()
        }
        
        // 停止语音合成
        if isSpeaking {
            stopTts()
        }
        
        // 停止空闲检测
        stopIdleCheck()
        
        // 移除BleService代理
        BleService.shared.setDelegate(nil)
        
        // 释放Azure Speech资源
        azureAsrHelper?.removeListener(self)
        azureTtsHelper?.removeListener(self)
        azureAsrHelper?.dispose()
        azureTtsHelper?.dispose()
        
        // 释放音频播放器
        audioPlayer = nil
        
        // 清除监听器
        clearListeners()
        
        isInitialized = false
        
        return true
    }
}

// MARK: - AsrEventListener 实现
extension AgentServiceImpl: AsrEventListener {
    func onEvent(_ event: AsrEvent) {
        switch event.type {
        case .sessionStarted:
            sendEvent(name: "recognition_started", data: [:])
            isRecognizing = true
            startIdleCheck()
            audioPlayer?.playStartSound()
            
        case .sessionStopped:
            sendEvent(name: "recognition_stopped", data: [:])
            isRecognizing = false
            stopIdleCheck()
            audioPlayer?.playStopSound()
            
        case .recognizing:
            if let text = event.params["text"] as? String {
                // 检测到语音，更新状态
                let previousHasSpeech = hasSpeechDetected
                hasSpeechDetected = true
                
                if !previousHasSpeech {
                    restartIdleCheck()
                }
                
                var data: [String: Any] = ["text": text]
                if let language = event.params["language"] as? String, !language.isEmpty {
                    data["language"] = language
                }
                
                sendEvent(name: "recognizing", data: data)
                
                // 如果TTS正在播放或AI正在生成，则触发打断
                if isSpeaking || isAiStreaming {
                    interruptCurrentResponse()
                }
            }
            
        case .recognized:
            if let text = event.params["text"] as? String, !text.isEmpty {
                var data: [String: Any] = ["text": text]
                if let language = event.params["language"] as? String, !language.isEmpty {
                    data["language"] = language
                }
                
                sendEvent(name: "recognition_result", data: data)
                
                // 处理识别结果，发送到OpenAI
                processTextInput(text, speakResponse: true)
            }
            
            // 重置状态，继续识别
            let previousHasSpeech = hasSpeechDetected
            hasSpeechDetected = false
            
            if previousHasSpeech {
                restartIdleCheck()
            }
            
        case .canceled:
            var data: [String: Any] = [:]
            if let reason = event.params["reason"] as? String {
                data["reason"] = reason
            }
            if let errorDetails = event.params["errorDetails"] as? String {
                data["details"] = errorDetails
            }
            
            sendEvent(name: "recognition_canceled", data: data)
            isRecognizing = false
            stopIdleCheck()
            
        case .error:
            var data: [String: Any] = [:]
            if let errorCode = event.params["errorCode"] as? String {
                data["errorCode"] = errorCode
            }
            if let message = event.params["message"] as? String {
                data["message"] = message
            } else if let error = event.params["error"] as? String {
                data["message"] = error
            } else {
                data["message"] = "未知错误"
            }
            
            sendEvent(name: "recognition_error", data: data)
            isRecognizing = false
            stopIdleCheck()
            
        default:
            break
        }
    }
}

// MARK: - TtsEventListener 实现
extension AgentServiceImpl: TtsEventListener {
    func onEvent(_ event: TtsEvent) {
        switch event.type {
        case .synthesisStarted:
            isSpeaking = true
            restartIdleCheck()
            sendEvent(name: "tts_started", data: [:])
            
        case .synthesisCompleted:
            isSpeaking = false
            restartIdleCheck()
            sendEvent(name: "tts_completed", data: [:])
            
        case .synthesisCanceled:
            isSpeaking = false
            restartIdleCheck()
            
            var data: [String: Any] = [:]
            if let reason = event.params["reason"] as? String {
                data["reason"] = reason
            }
            if let errorDetails = event.params["errorDetails"] as? String {
                data["details"] = errorDetails
            }
            
            sendEvent(name: "tts_canceled", data: data)
            
        case .error:
            isSpeaking = false
            restartIdleCheck()
            
            var data: [String: Any] = [:]
            if let errorCode = event.params["errorCode"] as? String {
                data["errorCode"] = errorCode
            }
            if let errorMessage = event.params["errorMessage"] as? String {
                data["message"] = errorMessage
            } else {
                data["message"] = "TTS错误"
            }
            
            sendEvent(name: "error", data: data)
            
        default:
            break
        }
    }
}

// MARK: - BleService.Callback 实现
extension AgentServiceImpl: BleService.Callback {
    func onScanResult(devices: [[String: Any]]) {
        // 不处理扫描结果
    }
    
    func onConnectionStateChanged(state: Int) {
        // 可以在这里处理连接状态变化，例如通知UI更新
    }
    
    func onAudioDataReceived(data: Data) {
        // 直接将音频数据推送到Azure Speech进行识别
        pushAudioData(data)
    }
    
    func onWakeupSignalReceived() {
        NSLog("\(TAG): 收到唤醒信号")
        
        // 停止当前TTS，避免冲突
        stopTts()
        
        // 如果已经在识别，则重新启动
        stopRecognition()
        startRecognition()
    }
    
    func onDeviceInfoReceived(infoType: Int, infoData: [String: Any]) {
        // 可以处理设备信息，例如电量、设备名称等
    }
}

// MARK: - 音频播放器

/// 简单的音频播放器，用于播放提示音
class AudioPlayer {
    private var player: AVAudioPlayer?
    
    /// 播放开始识别提示音
    func playStartSound() {
        playSound(named: "start")
    }
    
    /// 播放停止识别提示音
    func playStopSound() {
        playSound(named: "stop")
    }
    
    /// 播放音频文件
    /// - Parameter named: 音频文件名
    private func playSound(named: String) {
        guard let path = Bundle.main.path(forResource: named, ofType: "wav") else {
            print("找不到音频文件: \(named).wav")
            return
        }
        
        let url = URL(fileURLWithPath: path)
        
        do {
            player = try AVAudioPlayer(contentsOf: url)
            player?.prepareToPlay()
            player?.play()
        } catch {
            print("播放音频失败: \(error.localizedDescription)")
        }
    }
}