feat: add tts design

Signed-off-by: Xinwei Xiong <[email protected]>

.env.example

@@ -40,6 +40,7 @@ VOICEFLOW_VOLCENGINE_STT_RESOURCE_ID='volc.bigasr.sauc.duration'
 VOICEFLOW_VOLCENGINE_TTS_WS_URL='wss://openspeech.bytedance.com/api/v1/tts/ws_binary'   # TTS WebSocket URL
 VOICEFLOW_VOLCENGINE_TTS_APP_ID=''                                                       # TTS 应用ID
 VOICEFLOW_VOLCENGINE_TTS_TOKEN=''                                                        # TTS 令牌
+VOICEFLOW_VOLCENGINE_TTS_CLUSTER='volcano_tts'                                          # TTS 集群名称
 VOICEFLOW_VOLCENGINE_TTS_VOICE_TYPE='zh_female_1'                                        # TTS 音色类型
 VOICEFLOW_VOLCENGINE_TTS_ENCODING='mp3'                                                  # TTS 音频编码
 VOICEFLOW_VOLCENGINE_TTS_SPEED_RATIO='1.0'                                              # TTS 语速比例

configs/config.yaml

@@ -4,15 +4,16 @@ server:
 
 minio:
   enabled: true
-  endpoint: ''
-  access_key: ""
-  secret_key: ""
-  bucket_name: 'telepace-pipeline'
+  endpoint: "localhost:9000"
+  access_key: "your_access_key"
+  secret_key: "your_secret_key"
+  bucket_name: "telepace-pipeline"
   secure: true
+  storage_path: "voiceflow/audio/"
 
 stt:
   # 可选值：azure、google、local、assemblyai、volcengine、 aws
-  provider: volcengine
+  provider: assemblyai
 
 tts:
   # 可选值：azure、google、local、volcengine
@@ -65,6 +66,7 @@ volcengine:
     ws_url: "wss://openspeech.bytedance.com/api/v1/tts/ws_binary"
     app_id: "your_app_id"
     token: "your_token"
+    cluster: "volcano_tts"
     voice_type: "zh_female_sajiaonvyou_moon_bigtts"
     encoding: "mp3"
     speed_ratio: 1.0

internal/storage/local.go

@@ -7,6 +7,8 @@ import (
 	"os"
 	"path/filepath"
 	"time"
+	"github.com/telepace/voiceflow/pkg/config"
+	"github.com/telepace/voiceflow/pkg/logger"
 )
 
 type LocalStorageService struct {
@@ -15,8 +17,12 @@ type LocalStorageService struct {
 
 // NewLocalStorageService 创建并返回一个新的本地存储服务
 func NewLocalStorageService() *LocalStorageService {
+	cfg, err := config.GetConfig()
+	if err != nil {
+		logger.Fatalf("配置初始化失败: %v", err)
+	}
 	return &LocalStorageService{
-		storagePath: "./audio_files", // 设置本地存储路径
+		storagePath: cfg.MinIO.StoragePath, // 使用相同的配置路径或单独配置
 	}
 }
 

internal/storage/minio.go

@@ -4,21 +4,23 @@ package storage
 import (
 	"context"
 	"fmt"
-	"github.com/google/uuid" // 用于生成唯一文件名
-	"github.com/minio/minio-go/v7"
-	"github.com/minio/minio-go/v7/pkg/credentials"
-	"github.com/telepace/voiceflow/pkg/config"
-	"github.com/telepace/voiceflow/pkg/logger"
 	"io/ioutil"
 	"log"
 	"net/url"
 	"os"
 	"time"
+
+	"github.com/google/uuid" // 用于生成唯一文件名
+	"github.com/minio/minio-go/v7"
+	"github.com/minio/minio-go/v7/pkg/credentials"
+	"github.com/telepace/voiceflow/pkg/config"
+	"github.com/telepace/voiceflow/pkg/logger"
 )
 
 type MinIOService struct {
-	client     *minio.Client
-	bucketName string
+	client      *minio.Client
+	bucketName  string
+	storagePath string
 }
 
 // NewMinIOService 创建并返回 MinIO 客户端
@@ -38,8 +40,9 @@ func NewMinIOService() *MinIOService {
 	}
 
 	return &MinIOService{
-		client:     minioClient,
-		bucketName: cfg.MinIO.BucketName,
+		client:      minioClient,
+		bucketName:  cfg.MinIO.BucketName,
+		storagePath: cfg.MinIO.StoragePath,
 	}
 }
 
@@ -61,8 +64,8 @@ func (m *MinIOService) StoreAudio(audioData []byte) (string, error) {
 		return "", fmt.Errorf("error writing audio to temp file: %v", err)
 	}
 
-	// 生成唯一文件名
-	objectName := uuid.New().String() + ".wav"
+	// 生成唯一文件名，并添加存储路径前缀
+	objectName := m.storagePath + uuid.New().String() + ".wav"
 
 	// 上传文件到 MinIO
 	info, err := m.client.FPutObject(ctx, m.bucketName, objectName, tempFile.Name(), minio.PutObjectOptions{

internal/tts/volcengine/volcengine.go

@@ -20,6 +20,7 @@ type VolcengineTTS struct {
 	wsURL      string
 	appID      string
 	token      string
+	cluster    string
 	voiceType  string
 	encoding   string
 	speedRatio float64
@@ -38,6 +39,7 @@ func NewVolcengineTTS() *VolcengineTTS {
 		wsURL:      ttsCfg.WsURL,
 		appID:      ttsCfg.AppID,
 		token:      ttsCfg.Token,
+		cluster:    ttsCfg.Cluster,
 		voiceType:  ttsCfg.VoiceType,
 		encoding:   ttsCfg.Encoding,
 		speedRatio: ttsCfg.SpeedRatio,
@@ -70,7 +72,7 @@ func (v *VolcengineTTS) Synthesize(text string) ([]byte, error) {
 		"app": {
 			"appid":   v.appID,
 			"token":   v.token,
-			"cluster": "default",
+			"cluster": v.cluster,
 		},
 		"user": {
 			"uid": fmt.Sprintf("user_%d", time.Now().UnixNano()),

pkg/config/config.go

@@ -79,6 +79,7 @@ type Config struct {
 			WsURL       string  `mapstructure:"ws_url"`
 			AppID       string  `mapstructure:"app_id"`
 			Token       string  `mapstructure:"token"`
+			Cluster     string  `mapstructure:"cluster"`
 			VoiceType   string  `mapstructure:"voice_type"`
 			Encoding    string  `mapstructure:"encoding"`
 			SpeedRatio  float64 `mapstructure:"speed_ratio"`
@@ -94,6 +95,7 @@ type Config struct {
 		SecretKey  string `mapstructure:"secret_key"`
 		UseSSL     bool   `mapstructure:"use_ssl"`
 		Secure     bool   `mapstructure:"secure"`
+		StoragePath string `mapstructure:"storage_path"`
 	}
 	Logging struct {
 		Level        string