fix: implement audio duration retrieval and add tests

relay/adaptor/openai/model.go

@@ -1,6 +1,10 @@
 package openai
 
-import "github.com/songquanpeng/one-api/relay/model"
+import (
+	"mime/multipart"
+
+	"github.com/songquanpeng/one-api/relay/model"
+)
 
 type TextContent struct {
 	Type string `json:"type,omitempty"`
@@ -71,6 +75,24 @@ type TextToSpeechRequest struct {
 	ResponseFormat string  `json:"response_format"`
 }
 
+type AudioTranscriptionRequest struct {
+	File                 *multipart.FileHeader `form:"file" binding:"required"`
+	Model                string                `form:"model" binding:"required"`
+	Language             string                `form:"language"`
+	Prompt               string                `form:"prompt"`
+	ReponseFormat        string                `form:"response_format" binding:"oneof=json text srt verbose_json vtt"`
+	Temperature          float64               `form:"temperature"`
+	TimestampGranularity []string              `form:"timestamp_granularity"`
+}
+
+type AudioTranslationRequest struct {
+	File           *multipart.FileHeader `form:"file" binding:"required"`
+	Model          string                `form:"model" binding:"required"`
+	Prompt         string                `form:"prompt"`
+	ResponseFormat string                `form:"response_format" binding:"oneof=json text srt verbose_json vtt"`
+	Temperature    float64               `form:"temperature"`
+}
+
 type UsageOrResponseText struct {
 	*model.Usage
 	ResponseText string

relay/controller/audio.go

@@ -8,6 +8,7 @@ import (
 	"fmt"
 	"io"
 	"math"
+	"mime/multipart"
 	"net/http"
 	"os"
 	"strings"
@@ -33,22 +34,40 @@ const (
 	TokensPerSecond = 1000 / 20 // $0.006 / minute -> $0.002 / 20 seconds -> $0.002 / 1K tokens
 )
 
+type commonAudioRequest struct {
+	File *multipart.FileHeader `form:"file" binding:"required"`
+}
+
 func countAudioTokens(c *gin.Context) (int, error) {
 	body, err := common.GetRequestBody(c)
 	if err != nil {
 		return 0, errors.WithStack(err)
 	}
 
+	reqBody := new(commonAudioRequest)
+	c.Request.Body = io.NopCloser(bytes.NewReader(body))
+	if err = c.ShouldBind(reqBody); err != nil {
+		return 0, errors.WithStack(err)
+	}
+
+	reqFp, err := reqBody.File.Open()
+	if err != nil {
+		return 0, errors.WithStack(err)
+	}
+
 	fp, err := os.CreateTemp("", "audio-*")
 	if err != nil {
 		return 0, errors.WithStack(err)
 	}
 	defer os.Remove(fp.Name())
 
-	_, err = io.Copy(fp, bytes.NewReader(body))
+	_, err = io.Copy(fp, reqFp)
 	if err != nil {
 		return 0, errors.WithStack(err)
 	}
+	if err = fp.Close(); err != nil {
+		return 0, errors.WithStack(err)
+	}
 
 	duration, err := helper.GetAudioDuration(c.Request.Context(), fp.Name())
 	if err != nil {