You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 

112 lines
3.7 KiB

package echomeet
import (
"context"
"yunyan/lego/sys/log"
"yunyan/pb"
"yunyan/sys/microsoft/speech"
"yunyan/sys/microsoft/translate"
"yunyan/sys/openai"
"fmt"
)
// newMicrosoftProviders 初始化微软系单元服务(转写=Azure Speech、翻译=Azure Translator、
// 总结=Azure OpenAI 兼容端点),返回组合后的转写/翻译/总结三个 provider。
func newMicrosoftProviders(opt ProviderOptions) (Transcriber, Translator, Summarizer) {
if err := speech.OnInit(opt.Transcribe); err != nil {
log.Errorf("echomeet: init microsoft speech failed: %v", err)
}
if err := translate.OnInit(opt.Translate); err != nil {
log.Errorf("echomeet: init microsoft translate failed: %v", err)
}
sys, err := openai.NewSys(
openai.SetBaseURL(opt.Summary.BaseURL),
openai.SetToken(opt.Summary.Token),
openai.SetModel(opt.Summary.Model),
openai.SetMaxfunccall(1),
)
if err != nil {
log.Errorf("echomeet: init microsoft(azure openai) summarizer failed: %v", err)
}
return &microsoftTranscriber{}, &microsoftTranslator{}, &microsoftSummarizer{sys: sys}
}
// Azure Speech Batch Transcription
type microsoftTranscriber struct{}
func (this *microsoftTranscriber) Name() pb.TranscribeService {
return pb.TranscribeService_TranscribeService_Microsoft
}
// Azure Speech 支持的语种较广,原生使用 BCP-47 locale,不做白名单限制;
// 不支持时由 Azure 直接返回错误,统一在 Submit 失败路径处理。
func (this *microsoftTranscriber) SupportsLanguage(_ string) bool {
return true
}
func (this *microsoftTranscriber) Submit(_ context.Context, req SubmitRequest) (*SubmitResult, error) {
taskID, err := speech.CreateTask(req.AudioURL, req.Language, req.EnableSpeaker, req.CallbackURL)
if err != nil {
return nil, fmt.Errorf("microsoft speech create task: %w", err)
}
return &SubmitResult{TaskID: taskID}, nil
}
func (this *microsoftTranscriber) Query(_ context.Context, taskID, _ string) (*QueryResult, error) {
status, contexts, err := speech.QueryTask(taskID)
switch status {
case speech.StatusSuccess:
out := make([]*pb.ContextStruct, 0, len(contexts))
for _, c := range contexts {
out = append(out, &pb.ContextStruct{
Content: c.Content,
Starttime: c.StartTime,
Endtime: c.EndTime,
Speaker: c.Speaker,
})
}
return &QueryResult{Status: StatusSuccess, Contexts: out}, nil
case speech.StatusRunning:
return &QueryResult{Status: StatusRunning}, nil
default:
return &QueryResult{Status: StatusFailed}, fmt.Errorf("microsoft speech query failed: %w", err)
}
}
// Azure Translator
type microsoftTranslator struct{}
func (this *microsoftTranslator) Name() pb.TranscribeService {
return pb.TranscribeService_TranscribeService_Microsoft
}
func (this *microsoftTranslator) Translate(ctx context.Context, from, to string, texts []string) ([]string, error) {
return translate.Translate(ctx, toMicrosoftTranslateLang(from), toMicrosoftTranslateLang(to), texts)
}
// Azure OpenAI 总结(OpenAI 兼容端点)
type microsoftSummarizer struct {
sys openai.ISys
}
func (this *microsoftSummarizer) Name() pb.TranscribeService {
return pb.TranscribeService_TranscribeService_Microsoft
}
func (this *microsoftSummarizer) Chat(ctx context.Context, messages []ChatMessage) (string, error) {
if this.sys == nil {
return "", fmt.Errorf("microsoft summarizer not initialized")
}
msgs := make([]openai.ChatReq, 0, len(messages))
for _, m := range messages {
msgs = append(msgs, openai.ChatReq{Role: m.Role, Content: m.Content, Images: m.Images})
}
resp, err := this.sys.Chat(ctx, msgs)
if err != nil {
return "", err
}
if resp == nil {
return "", fmt.Errorf("microsoft(azure openai) chat: empty response")
}
return resp.Content, nil
}