refactor: 认证系统重构 + DevTools CLI 重写 + 文档全面更新
- auth: Login 简化为管理员始终通过 .env 验证,GetProfile 修正 admin DB 查询 - devtools: .sh/.bat 同步重写为完整 CLI (start/stop/status/logs/build/db:*) - docs: 新增 devtools.md,重写 Deploy.md (三种方式+Windows说明),更新 README/gateway-api - voice-service: DashScope 实时流式 STT 支持 - gateway: Phase 6 多模型配置 + 多端客户端管理 + WebSocket 增强 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
@@ -16,21 +16,27 @@ import (
|
||||
var SupportedLanguages = []string{"zh", "en", "ja", "ko", "auto"}
|
||||
|
||||
// STTService 语音转文字服务。
|
||||
// 优先使用 DashScope Gummy API,不可用时回退到本地 Whisper。
|
||||
// 优先使用 DashScope API,不可用时回退到本地 Whisper。
|
||||
type STTService struct {
|
||||
whisperBinary string
|
||||
whisperModel string
|
||||
language string
|
||||
dashscope *DashScopeSTT
|
||||
dashscope *DashScopeSTT // 实时 ASR (qwen3-asr-flash-realtime)
|
||||
}
|
||||
|
||||
// NewSTTService 创建 STT 服务。
|
||||
func NewSTTService(cfg *config.Config) *STTService {
|
||||
// 实时模型用于所有 WebSocket ASR 请求(支持 one-shot 和 streaming)
|
||||
// 离线模型 (qwen3-asr-flash-2026-02-10) 是 HTTP REST API,暂未实现
|
||||
model := cfg.DashScopeSTTRealtime
|
||||
if model == "" {
|
||||
model = cfg.DashScopeModel
|
||||
}
|
||||
return &STTService{
|
||||
whisperBinary: cfg.WhisperBinary,
|
||||
whisperModel: cfg.WhisperModel,
|
||||
language: cfg.WhisperLanguage,
|
||||
dashscope: NewDashScopeSTT(cfg.DashScopeAPIKey, cfg.DashScopeModel),
|
||||
dashscope: NewDashScopeSTT(cfg.DashScopeAPIKey, model),
|
||||
}
|
||||
}
|
||||
|
||||
@@ -58,15 +64,30 @@ func (s *STTService) Transcribe(audioData []byte, format string, language string
|
||||
ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
|
||||
defer cancel()
|
||||
text, err := s.dashscope.Transcribe(ctx, audioData, format, language)
|
||||
if err == nil && text != "" {
|
||||
if err == nil {
|
||||
return text, nil
|
||||
}
|
||||
// DashScope 失败,返回具体错误而不是回退到 Whisper
|
||||
return "", fmt.Errorf("语音识别失败: %w", err)
|
||||
}
|
||||
|
||||
// 回退到本地 Whisper
|
||||
return s.transcribeWhisper(audioData, format, language)
|
||||
}
|
||||
|
||||
// StartStreaming 创建持久的流式语音识别会话。
|
||||
func (s *STTService) StartStreaming(format, language string) (*StreamingSession, error) {
|
||||
if !s.dashscope.IsAvailable() {
|
||||
return nil, fmt.Errorf("流式识别需要 DashScope,请配置 DASHSCOPE_API_KEY")
|
||||
}
|
||||
if language == "" {
|
||||
language = s.language
|
||||
}
|
||||
ctx, cancel := context.WithTimeout(context.Background(), 15*time.Second)
|
||||
defer cancel()
|
||||
return s.dashscope.StartStreaming(ctx, format, language)
|
||||
}
|
||||
|
||||
// transcribeWhisper 使用本地 Whisper 引擎转录。
|
||||
func (s *STTService) transcribeWhisper(audioData []byte, format string, language string) (string, error) {
|
||||
if _, err := os.Stat(s.whisperBinary); err != nil {
|
||||
|
||||
Reference in New Issue
Block a user