feat: add SSML processing capabilities and configuration support

configs/config.yaml

@@ -28,3 +28,30 @@ tts:
     shimmer: "zh-CN-XiaomoNeural"     # 温柔女声
 openai:
   api_key: ''
+
+ssml:
+  preserve_tags:
+    - name: break
+      pattern: <break\s+[^>]*/>
+    - name: speak
+      pattern: <speak>|</speak>
+    - name: prosody
+      pattern: <prosody\s+[^>]*>|</prosody>
+    - name: emphasis
+      pattern: <emphasis\s+[^>]*>|</emphasis>
+    - name: voice
+      pattern: <voice\s+[^>]*>|</voice>
+    - name: say-as
+      pattern: <say-as\s+[^>]*>|</say-as>
+    - name: phoneme
+      pattern: <phoneme\s+[^>]*>|</phoneme>
+    - name: audio
+      pattern: <audio\s+[^>]*>|</audio>
+    - name: p
+      pattern: <p>|</p>
+    - name: s
+      pattern: <s>|</s>
+    - name: sub
+      pattern: <sub\s+[^>]*>|</sub>
+    - name: mstts
+      pattern: <mstts:[^>]*>|</mstts:[^>]*>

internal/config/config.go

@@ -2,6 +2,8 @@ package config
 
 import (
 	"fmt"
+	"html"
+	"regexp"
 	"strings"
 	"sync"
 
@@ -13,6 +15,7 @@ type Config struct {
 	Server ServerConfig `mapstructure:"server"`
 	TTS    TTSConfig    `mapstructure:"tts"`
 	OpenAI OpenAIConfig `mapstructure:"openai"`
+	SSML   SSMLConfig   `mapstructure:"ssml"`
 }
 
 // OpenAIConfig 包含OpenAI API配置
@@ -89,3 +92,69 @@ func Load(configPath string) (*Config, error) {
 func Get() *Config {
 	return &config
 }
+
+// TagPattern 定义标签模式及其名称
+type TagPattern struct {
+	Name    string `mapstructure:"name"`    // 标签名称，用于日志和调试
+	Pattern string `mapstructure:"pattern"` // 标签的正则表达式模式
+}
+
+// SSMLConfig 存储SSML标签配置
+type SSMLConfig struct {
+	// PreserveTags 包含所有需要保留的标签的正则表达式模式
+	PreserveTags []TagPattern `mapstructure:"preserve_tags"`
+}
+
+// SSMLProcessor 处理SSML内容
+type SSMLProcessor struct {
+	config       *SSMLConfig
+	patternCache map[string]*regexp.Regexp
+}
+
+// NewSSMLProcessor 从配置对象创建SSMLProcessor
+func NewSSMLProcessor(config *SSMLConfig) (*SSMLProcessor, error) {
+	processor := &SSMLProcessor{
+		config:       config,
+		patternCache: make(map[string]*regexp.Regexp),
+	}
+
+	// 预编译正则表达式
+	for _, tagPattern := range config.PreserveTags {
+		regex, err := regexp.Compile(tagPattern.Pattern)
+		if err != nil {
+			return nil, fmt.Errorf("编译正则表达式'%s'失败: %w", tagPattern.Name, err)
+		}
+		processor.patternCache[tagPattern.Name] = regex
+	}
+
+	return processor, nil
+}
+
+// EscapeSSML 转义SSML内容，但保留配置的标签
+func (p *SSMLProcessor) EscapeSSML(ssml string) string {
+	// 使用占位符替换标签
+	placeholders := make(map[string]string)
+	processedSSML := ssml
+
+	counter := 0
+
+	// 处理所有配置的标签
+	for name, pattern := range p.patternCache {
+		processedSSML = pattern.ReplaceAllStringFunc(processedSSML, func(match string) string {
+			placeholder := fmt.Sprintf("__SSML_PLACEHOLDER_%s_%d__", name, counter)
+			placeholders[placeholder] = match
+			counter++
+			return placeholder
+		})
+	}
+
+	// 对处理后的文本进行HTML转义
+	escapedContent := html.EscapeString(processedSSML)
+
+	// 恢复所有标签占位符
+	for placeholder, tag := range placeholders {
+		escapedContent = strings.Replace(escapedContent, placeholder, tag, 1)
+	}
+
+	return escapedContent
+}

internal/tts/microsoft/client.go

@@ -6,7 +6,6 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
-	"html"
 	"io"
 	"log"
 	"net/http"
@@ -50,10 +49,16 @@ type Client struct {
 	endpoint       map[string]interface{}
 	endpointMu     sync.RWMutex
 	endpointExpiry time.Time
+	ssmProcessor   *config.SSMLProcessor
 }
 
 // NewClient 创建一个新的Microsoft TTS客户端
 func NewClient(cfg *config.Config) *Client {
+	// 从Viper配置中创建SSML处理器
+	ssmProcessor, err := config.NewSSMLProcessor(&cfg.SSML)
+	if err != nil {
+		log.Fatalf("创建SSML处理器失败: %v", err)
+	}
 	client := &Client{
 		defaultVoice:  cfg.TTS.DefaultVoice,
 		defaultRate:   cfg.TTS.DefaultRate,
@@ -65,6 +70,7 @@ func NewClient(cfg *config.Config) *Client {
 		},
 		voicesCacheExpiry: time.Time{}, // 初始时缓存为空
 		endpointExpiry:    time.Time{}, // 初始时端点为空
+		ssmProcessor:      ssmProcessor,
 	}
 
 	return client
@@ -245,8 +251,7 @@ func (c *Client) createTTSRequest(ctx context.Context, req models.TTSRequest) (*
 	}
 
 	// 对文本进行HTML转义，防止XML解析错误
-
+	escapedText := c.ssmProcessor.EscapeSSML(req.Text)
-	escapedText := html.EscapeString(req.Text)
 
 	// 准备SSML内容
 	ssml := fmt.Sprintf(ssmlTemplate, locale, voice, style, rate, pitch, escapedText)