audio_codebuddy/deploy-package/server/modules/tts/aliyun-realtime.provider.js

"use strict";
var __importDefault = (this && this.__importDefault) || function (mod) {
    return (mod && mod.__esModule) ? mod : { "default": mod };
};
Object.defineProperty(exports, "__esModule", { value: true });
exports.AliyunRealtimeTtsProvider = void 0;
const ws_1 = __importDefault(require("ws"));
const fs_1 = __importDefault(require("fs"));
const path_1 = __importDefault(require("path"));
const config_1 = require("../../config");
// 阿里云百炼 Qwen-TTS WebSocket 实时语音合成
// 文档: https://help.aliyun.com/zh/model-studio/qwen-tts-realtime
// 支持流式输入，适合长文本
class AliyunRealtimeTtsProvider {
    apiKey;
    model;
    voice;
    baseUrl = 'wss://dashscope.aliyuncs.com/api-ws/v1/realtime';
    constructor() {
        this.apiKey = config_1.config.dashscope.apiKey;
        this.model = config_1.config.dashscope.realtimeModel;
        this.voice = config_1.config.dashscope.voice;
    }
    // 语音合成（WebSocket 流式）
    async synthesize(text, voiceId, params, outputPath) {
        return new Promise((resolve, reject) => {
            const voice = voiceId || this.voice;
            const audioChunks = [];
            let sessionId;
            console.log('🔊 [Qwen Realtime TTS] 建立 WebSocket 连接...');
            const ws = new ws_1.default(this.baseUrl, {
                headers: {
                    'Authorization': `Bearer ${this.apiKey}`,
                    'X-DashScope-App': 'audio-tts',
                },
            });
            ws.on('open', () => {
                console.log('✅ [Qwen Realtime TTS] 连接已建立');
            });
            ws.on('message', (data) => {
                try {
                    const message = JSON.parse(data.toString());
                    switch (message.event) {
                        case 'session.configed':
                            // 配置成功，发送文本
                            sessionId = message.session_id;
                            console.log('📝 [Qwen Realtime TTS] Session ID:', sessionId);
                            // 构建指令
                            const instructions = [];
                            if (params.speed !== 1) {
                                const speedDesc = params.speed > 1 ? '语速较快' : '语速较慢';
                                instructions.push(speedDesc);
                            }
                            if (params.pitch !== 0) {
                                const pitchDesc = params.pitch > 0 ? '音调较高' : '音调较低';
                                instructions.push(pitchDesc);
                            }
                            // 使用 server_commit 模式：直接发送完整文本，服务端自动处理分段
                            const textMessage = {
                                event: 'input_text_buffer.append',
                                context: {
                                    text: text,
                                },
                            };
                            ws.send(JSON.stringify(textMessage));
                            console.log('📤 [Qwen Realtime TTS] 文本已发送，长度:', text.length);
                            // 发送 commit 触发合成
                            const commitMessage = {
                                event: 'input_text_buffer.commit',
                            };
                            ws.send(JSON.stringify(commitMessage));
                            console.log('📤 [Qwen Realtime TTS] 触发语音合成...');
                            break;
                        case 'audio.stream':
                            // 接收音频数据
                            if (message.data) {
                                const audioBuffer = Buffer.from(message.data, 'base64');
                                audioChunks.push(audioBuffer);
                            }
                            break;
                        case 'audio.finish':
                            // 合成完成
                            console.log('✅ [Qwen Realtime TTS] 合成完成');
                            // 保存文件
                            if (audioChunks.length > 0) {
                                const dir = path_1.default.dirname(outputPath);
                                if (!fs_1.default.existsSync(dir)) {
                                    fs_1.default.mkdirSync(dir, { recursive: true });
                                }
                                const finalPath = outputPath.endsWith('.mp3') ? outputPath : outputPath.replace(/\.[^.]+$/, '.mp3');
                                const finalBuffer = Buffer.concat(audioChunks);
                                fs_1.default.writeFileSync(finalPath, finalBuffer);
                                console.log(`✅ [Qwen Realtime TTS] 文件已保存: ${finalPath}, 大小: ${finalBuffer.length} bytes`);
                                ws.close();
                                resolve(finalPath);
                            }
                            else {
                                ws.close();
                                reject(new Error('Qwen Realtime TTS 未返回音频数据'));
                            }
                            break;
                        case 'task.finished':
                            // 任务完成
                            ws.close();
                            resolve(outputPath);
                            break;
                        case 'error':
                            console.error('❌ [Qwen Realtime TTS] 错误:', message.message);
                            ws.close();
                            reject(new Error(`Qwen Realtime TTS 错误: ${message.message}`));
                            break;
                        default:
                            // 忽略其他消息
                            break;
                    }
                }
                catch (err) {
                    console.error('❌ [Qwen Realtime TTS] 解析消息失败:', err);
                }
            });
            ws.on('error', (err) => {
                console.error('❌ [Qwen Realtime TTS] WebSocket 错误:', err.message);
                reject(err);
            });
            ws.on('close', () => {
                console.log('🔌 [Qwen Realtime TTS] 连接已关闭');
            });
            // 超时处理
            setTimeout(() => {
                if (ws.readyState === ws_1.default.OPEN) {
                    ws.close();
                    reject(new Error('Qwen Realtime TTS 请求超时'));
                }
            }, 120000); // 2分钟超时
        });
    }
}
exports.AliyunRealtimeTtsProvider = AliyunRealtimeTtsProvider;