♻️ refactor: 重构speech接口

common/client.go

@@ -103,6 +103,15 @@ func (c *Client) SendRequest(req *http.Request, response any) error {
 	return nil
 }
 
+func (c *Client) SendRequestRaw(req *http.Request) (body io.ReadCloser, err error) {
+	resp, err := HttpClient.Do(req)
+	if err != nil {
+		return
+	}
+
+	return resp.Body, nil
+}
+
 func IsFailureStatusCode(resp *http.Response) bool {
 	return resp.StatusCode < http.StatusOK || resp.StatusCode >= http.StatusBadRequest
 }

controller/relay-helper.go

@@ -13,7 +13,7 @@ import (
 	"github.com/gin-gonic/gin"
 )
 
-func relayTextHelper(c *gin.Context, relayMode int) *types.OpenAIErrorWithStatusCode {
+func relayHelper(c *gin.Context, relayMode int) *types.OpenAIErrorWithStatusCode {
 	// 获取请求参数
 	channelType := c.GetInt("channel")
 	channelId := c.GetInt("channel_id")
@@ -36,10 +36,9 @@ func relayTextHelper(c *gin.Context, relayMode int) *types.OpenAIErrorWithStatus
 		return types.ErrorWrapper(err, "unmarshal_model_mapping_failed", http.StatusInternalServerError)
 	}
 
-	var promptTokens int
 	quotaInfo := &QuotaInfo{
 		modelName:    "",
-		promptTokens: promptTokens,
+		promptTokens: 0,
 		userId:       userId,
 		channelId:    channelId,
 		tokenId:      tokenId,
@@ -57,6 +56,8 @@ func relayTextHelper(c *gin.Context, relayMode int) *types.OpenAIErrorWithStatus
 		usage, openAIErrorWithStatusCode = handleEmbeddings(c, provider, modelMap, quotaInfo, group)
 	case common.RelayModeModerations:
 		usage, openAIErrorWithStatusCode = handleModerations(c, provider, modelMap, quotaInfo, group)
+	case common.RelayModeAudioSpeech:
+		usage, openAIErrorWithStatusCode = handleSpeech(c, provider, modelMap, quotaInfo, group)
 	default:
 		return types.ErrorWrapper(errors.New("invalid relay mode"), "invalid_relay_mode", http.StatusBadRequest)
 	}
@@ -112,6 +113,7 @@ func handleChatCompletions(c *gin.Context, provider providers_base.ProviderInter
 	promptTokens := common.CountTokenMessages(chatRequest.Messages, chatRequest.Model)
 
 	quotaInfo.modelName = chatRequest.Model
+	quotaInfo.promptTokens = promptTokens
 	quotaInfo.initQuotaInfo(group)
 	quota_err := quotaInfo.preQuotaConsumption()
 	if quota_err != nil {
@@ -144,6 +146,7 @@ func handleCompletions(c *gin.Context, provider providers_base.ProviderInterface
 	promptTokens := common.CountTokenInput(completionRequest.Prompt, completionRequest.Model)
 
 	quotaInfo.modelName = completionRequest.Model
+	quotaInfo.promptTokens = promptTokens
 	quotaInfo.initQuotaInfo(group)
 	quota_err := quotaInfo.preQuotaConsumption()
 	if quota_err != nil {
@@ -176,6 +179,7 @@ func handleEmbeddings(c *gin.Context, provider providers_base.ProviderInterface,
 	promptTokens := common.CountTokenInput(embeddingsRequest.Input, embeddingsRequest.Model)
 
 	quotaInfo.modelName = embeddingsRequest.Model
+	quotaInfo.promptTokens = promptTokens
 	quotaInfo.initQuotaInfo(group)
 	quota_err := quotaInfo.preQuotaConsumption()
 	if quota_err != nil {
@@ -212,6 +216,7 @@ func handleModerations(c *gin.Context, provider providers_base.ProviderInterface
 	promptTokens := common.CountTokenInput(moderationRequest.Input, moderationRequest.Model)
 
 	quotaInfo.modelName = moderationRequest.Model
+	quotaInfo.promptTokens = promptTokens
 	quotaInfo.initQuotaInfo(group)
 	quota_err := quotaInfo.preQuotaConsumption()
 	if quota_err != nil {
@@ -219,3 +224,36 @@ func handleModerations(c *gin.Context, provider providers_base.ProviderInterface
 	}
 	return moderationProvider.ModerationAction(&moderationRequest, isModelMapped, promptTokens)
 }
+
+func handleSpeech(c *gin.Context, provider providers_base.ProviderInterface, modelMap map[string]string, quotaInfo *QuotaInfo, group string) (*types.Usage, *types.OpenAIErrorWithStatusCode) {
+	var speechRequest types.SpeechAudioRequest
+	isModelMapped := false
+	speechProvider, ok := provider.(providers_base.SpeechInterface)
+	if !ok {
+		return nil, types.ErrorWrapper(errors.New("channel not implemented"), "channel_not_implemented", http.StatusNotImplemented)
+	}
+
+	err := common.UnmarshalBodyReusable(c, &speechRequest)
+	if err != nil {
+		return nil, types.ErrorWrapper(err, "bind_request_body_failed", http.StatusBadRequest)
+	}
+
+	if speechRequest.Input == "" {
+		return nil, types.ErrorWrapper(errors.New("field input is required"), "required_field_missing", http.StatusBadRequest)
+	}
+
+	if modelMap != nil && modelMap[speechRequest.Model] != "" {
+		speechRequest.Model = modelMap[speechRequest.Model]
+		isModelMapped = true
+	}
+	promptTokens := len(speechRequest.Input)
+
+	quotaInfo.modelName = speechRequest.Model
+	quotaInfo.promptTokens = promptTokens
+	quotaInfo.initQuotaInfo(group)
+	quota_err := quotaInfo.preQuotaConsumption()
+	if quota_err != nil {
+		return nil, quota_err
+	}
+	return speechProvider.SpeechAction(&speechRequest, isModelMapped, promptTokens)
+}

controller/relay.go

@@ -235,13 +235,13 @@ func Relay(c *gin.Context) {
 		relayMode = common.RelayModeEmbeddings
 	} else if strings.HasPrefix(c.Request.URL.Path, "/v1/moderations") {
 		relayMode = common.RelayModeModerations
+	} else if strings.HasPrefix(c.Request.URL.Path, "/v1/audio/speech") {
+		relayMode = common.RelayModeAudioSpeech
 	}
 	// } else if strings.HasPrefix(c.Request.URL.Path, "/v1/images/generations") {
 	// 	relayMode = RelayModeImagesGenerations
 	// } else if strings.HasPrefix(c.Request.URL.Path, "/v1/edits") {
 	// 	relayMode = RelayModeEdits
-	// } else if strings.HasPrefix(c.Request.URL.Path, "/v1/audio/speech") {
-	// 	relayMode = RelayModeAudioSpeech
 	// } else if strings.HasPrefix(c.Request.URL.Path, "/v1/audio/transcriptions") {
 	// 	relayMode = RelayModeAudioTranscription
 	// } else if strings.HasPrefix(c.Request.URL.Path, "/v1/audio/translations") {
@@ -257,7 +257,7 @@ func Relay(c *gin.Context) {
 	// case RelayModeAudioTranscription:
 	// 	err = relayAudioHelper(c, relayMode)
 	default:
-		err = relayTextHelper(c, relayMode)
+		err = relayHelper(c, relayMode)
 	}
 	if err != nil {
 		requestId := c.GetString(common.RequestIdKey)

providers/base/common.go

@@ -95,6 +95,35 @@ func (p *BaseProvider) SendRequest(req *http.Request, response ProviderResponseH
 	return nil
 }
 
+func (p *BaseProvider) SendRequestRaw(req *http.Request) (openAIErrorWithStatusCode *types.OpenAIErrorWithStatusCode) {
+
+	// 发送请求
+	resp, err := common.HttpClient.Do(req)
+	if err != nil {
+		return types.ErrorWrapper(err, "http_request_failed", http.StatusInternalServerError)
+	}
+
+	defer resp.Body.Close()
+
+	// 处理响应
+	if common.IsFailureStatusCode(resp) {
+		return p.HandleErrorResp(resp)
+	}
+
+	for k, v := range resp.Header {
+		p.Context.Writer.Header().Set(k, v[0])
+	}
+
+	p.Context.Writer.WriteHeader(resp.StatusCode)
+
+	_, err = io.Copy(p.Context.Writer, resp.Body)
+	if err != nil {
+		return types.ErrorWrapper(err, "write_response_body_failed", http.StatusInternalServerError)
+	}
+
+	return nil
+}
+
 // 处理错误响应
 func (p *BaseProvider) HandleErrorResp(resp *http.Response) (openAIErrorWithStatusCode *types.OpenAIErrorWithStatusCode) {
 	openAIErrorWithStatusCode = &types.OpenAIErrorWithStatusCode{

providers/base/interface.go

@@ -38,6 +38,12 @@ type ModerationInterface interface {
 	ModerationAction(request *types.ModerationRequest, isModelMapped bool, promptTokens int) (usage *types.Usage, errWithCode *types.OpenAIErrorWithStatusCode)
 }
 
+// 文字转语音接口
+type SpeechInterface interface {
+	ProviderInterface
+	SpeechAction(request *types.SpeechAudioRequest, isModelMapped bool, promptTokens int) (usage *types.Usage, errWithCode *types.OpenAIErrorWithStatusCode)
+}
+
 // 余额接口
 type BalanceInterface interface {
 	BalanceAction(channel *model.Channel) (float64, error)

providers/openai/speech.go

@@ -0,0 +1,37 @@
+package openai
+
+import (
+	"net/http"
+	"one-api/common"
+	"one-api/types"
+)
+
+func (p *OpenAIProvider) SpeechAction(request *types.SpeechAudioRequest, isModelMapped bool, promptTokens int) (usage *types.Usage, errWithCode *types.OpenAIErrorWithStatusCode) {
+
+	requestBody, err := p.getRequestBody(&request, isModelMapped)
+	if err != nil {
+		return nil, types.ErrorWrapper(err, "json_marshal_failed", http.StatusInternalServerError)
+	}
+
+	fullRequestURL := p.GetFullRequestURL(p.AudioSpeech, request.Model)
+	headers := p.GetRequestHeaders()
+
+	client := common.NewClient()
+	req, err := client.NewRequest(p.Context.Request.Method, fullRequestURL, common.WithBody(requestBody), common.WithHeader(headers))
+	if err != nil {
+		return nil, types.ErrorWrapper(err, "new_request_failed", http.StatusInternalServerError)
+	}
+
+	errWithCode = p.SendRequestRaw(req)
+	if errWithCode != nil {
+		return
+	}
+
+	usage = &types.Usage{
+		PromptTokens:     promptTokens,
+		CompletionTokens: 0,
+		TotalTokens:      promptTokens,
+	}
+
+	return
+}