mirror of
https://github.com/songquanpeng/one-api.git
synced 2026-04-07 02:44:25 +08:00
Compare commits
4 Commits
v0.6.10-al
...
b7e5d3b761
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
b7e5d3b761 | ||
|
|
7c8628bd95 | ||
|
|
6ab87f8a08 | ||
|
|
e9d4226e7d |
27
middleware/gzip.go
Normal file
27
middleware/gzip.go
Normal file
@@ -0,0 +1,27 @@
|
||||
package middleware
|
||||
|
||||
import (
|
||||
"compress/gzip"
|
||||
"github.com/gin-gonic/gin"
|
||||
"io"
|
||||
"net/http"
|
||||
)
|
||||
|
||||
func GzipDecodeMiddleware() gin.HandlerFunc {
|
||||
return func(c *gin.Context) {
|
||||
if c.GetHeader("Content-Encoding") == "gzip" {
|
||||
gzipReader, err := gzip.NewReader(c.Request.Body)
|
||||
if err != nil {
|
||||
c.AbortWithStatus(http.StatusBadRequest)
|
||||
return
|
||||
}
|
||||
defer gzipReader.Close()
|
||||
|
||||
// Replace the request body with the decompressed data
|
||||
c.Request.Body = io.NopCloser(gzipReader)
|
||||
}
|
||||
|
||||
// Continue processing the request
|
||||
c.Next()
|
||||
}
|
||||
}
|
||||
@@ -71,9 +71,9 @@ func RelayAudioHelper(c *gin.Context, relayMode int) *relaymodel.ErrorWithStatus
|
||||
return openai.ErrorWrapper(err, "get_user_quota_failed", http.StatusInternalServerError)
|
||||
}
|
||||
|
||||
// Check if user quota is enough
|
||||
// Check if user quota is enough Message prompts the user to use Chinese
|
||||
if userQuota-preConsumedQuota < 0 {
|
||||
return openai.ErrorWrapper(errors.New("user quota is not enough"), "insufficient_user_quota", http.StatusForbidden)
|
||||
return openai.ErrorWrapper(errors.New("请移步充值页面进行充值,可在日志中查阅使用明细"), "insufficient_user_quota", http.StatusForbidden)
|
||||
}
|
||||
err = model.CacheDecreaseUserQuota(userId, preConsumedQuota)
|
||||
if err != nil {
|
||||
|
||||
@@ -69,8 +69,10 @@ func preConsumeQuota(ctx context.Context, textRequest *relaymodel.GeneralOpenAIR
|
||||
if err != nil {
|
||||
return preConsumedQuota, openai.ErrorWrapper(err, "get_user_quota_failed", http.StatusInternalServerError)
|
||||
}
|
||||
// Check if user quota is enough Message prompts the user to use Chinese
|
||||
if userQuota-preConsumedQuota < 0 {
|
||||
return preConsumedQuota, openai.ErrorWrapper(errors.New("user quota is not enough"), "insufficient_user_quota", http.StatusForbidden)
|
||||
return preConsumedQuota, openai.ErrorWrapper(errors.New("请移步充值页面进行充值,可在日志中查阅使用明细"), "insufficient_user_quota", http.StatusForbidden)
|
||||
|
||||
}
|
||||
err = model.CacheDecreaseUserQuota(meta.UserId, preConsumedQuota)
|
||||
if err != nil {
|
||||
@@ -91,7 +93,7 @@ func preConsumeQuota(ctx context.Context, textRequest *relaymodel.GeneralOpenAIR
|
||||
return preConsumedQuota, nil
|
||||
}
|
||||
|
||||
func postConsumeQuota(ctx context.Context, usage *relaymodel.Usage, meta *meta.Meta, textRequest *relaymodel.GeneralOpenAIRequest, ratio float64, preConsumedQuota int64, modelRatio float64, groupRatio float64) {
|
||||
func postConsumeQuota(ctx context.Context, usage *relaymodel.Usage, meta *meta.Meta, textRequest *relaymodel.GeneralOpenAIRequest, ratio float64, preConsumedQuota int64, modelRatio float64, groupRatio float64, systemPromptReset bool) {
|
||||
if usage == nil {
|
||||
logger.Error(ctx, "usage is nil, which is unexpected")
|
||||
return
|
||||
@@ -119,7 +121,11 @@ func postConsumeQuota(ctx context.Context, usage *relaymodel.Usage, meta *meta.M
|
||||
if err != nil {
|
||||
logger.Error(ctx, "error update user quota cache: "+err.Error())
|
||||
}
|
||||
logContent := fmt.Sprintf("模型倍率 %.2f,分组倍率 %.2f,补全倍率 %.2f", modelRatio, groupRatio, completionRatio)
|
||||
var extraLog string
|
||||
if systemPromptReset {
|
||||
extraLog = " (注意系统提示词已被重置)"
|
||||
}
|
||||
logContent := fmt.Sprintf("模型倍率 %.2f,分组倍率 %.2f,补全倍率 %.2f%s", modelRatio, groupRatio, completionRatio, extraLog)
|
||||
model.RecordConsumeLog(ctx, meta.UserId, meta.ChannelId, promptTokens, completionTokens, textRequest.Model, meta.TokenName, quota, logContent)
|
||||
model.UpdateUserUsedQuotaAndRequestCount(meta.UserId, quota)
|
||||
model.UpdateChannelUsedQuota(meta.ChannelId, quota)
|
||||
@@ -156,21 +162,22 @@ func isErrorHappened(meta *meta.Meta, resp *http.Response) bool {
|
||||
return false
|
||||
}
|
||||
|
||||
func setSystemPrompt(ctx context.Context, request *relaymodel.GeneralOpenAIRequest, prompt string) {
|
||||
func setSystemPrompt(ctx context.Context, request *relaymodel.GeneralOpenAIRequest, prompt string) (reset bool) {
|
||||
if prompt == "" {
|
||||
return
|
||||
return false
|
||||
}
|
||||
if len(request.Messages) == 0 {
|
||||
return
|
||||
return false
|
||||
}
|
||||
if request.Messages[0].Role == role.System {
|
||||
request.Messages[0].Content = prompt
|
||||
logger.Infof(ctx, "rewrite system prompt")
|
||||
return
|
||||
return true
|
||||
}
|
||||
request.Messages = append([]relaymodel.Message{{
|
||||
Role: role.System,
|
||||
Content: prompt,
|
||||
}}, request.Messages...)
|
||||
logger.Infof(ctx, "add system prompt")
|
||||
return true
|
||||
}
|
||||
|
||||
@@ -173,9 +173,9 @@ func RelayImageHelper(c *gin.Context, relayMode int) *relaymodel.ErrorWithStatus
|
||||
userQuota, err := model.CacheGetUserQuota(ctx, meta.UserId)
|
||||
|
||||
quota := int64(ratio*imageCostRatio*1000) * int64(imageRequest.N)
|
||||
|
||||
// Check if user quota is enough Message prompts the user to use Chinese
|
||||
if userQuota-quota < 0 {
|
||||
return openai.ErrorWrapper(errors.New("user quota is not enough"), "insufficient_user_quota", http.StatusForbidden)
|
||||
return openai.ErrorWrapper(errors.New("请移步充值页面进行充值,可在日志中查阅使用明细"), "insufficient_user_quota", http.StatusForbidden)
|
||||
}
|
||||
|
||||
// do request
|
||||
|
||||
@@ -37,7 +37,7 @@ func RelayTextHelper(c *gin.Context) *model.ErrorWithStatusCode {
|
||||
textRequest.Model, _ = getMappedModelName(textRequest.Model, meta.ModelMapping)
|
||||
meta.ActualModelName = textRequest.Model
|
||||
// set system prompt if not empty
|
||||
setSystemPrompt(ctx, textRequest, meta.SystemPrompt)
|
||||
systemPromptReset := setSystemPrompt(ctx, textRequest, meta.SystemPrompt)
|
||||
// get model ratio & group ratio
|
||||
modelRatio := billingratio.GetModelRatio(textRequest.Model, meta.ChannelType)
|
||||
groupRatio := billingratio.GetGroupRatio(meta.Group)
|
||||
@@ -82,7 +82,7 @@ func RelayTextHelper(c *gin.Context) *model.ErrorWithStatusCode {
|
||||
return respErr
|
||||
}
|
||||
// post-consume quota
|
||||
go postConsumeQuota(ctx, usage, meta, textRequest, ratio, preConsumedQuota, modelRatio, groupRatio)
|
||||
go postConsumeQuota(ctx, usage, meta, textRequest, ratio, preConsumedQuota, modelRatio, groupRatio, systemPromptReset)
|
||||
return nil
|
||||
}
|
||||
|
||||
|
||||
@@ -9,6 +9,7 @@ import (
|
||||
|
||||
func SetRelayRouter(router *gin.Engine) {
|
||||
router.Use(middleware.CORS())
|
||||
router.Use(middleware.GzipDecodeMiddleware())
|
||||
// https://platform.openai.com/docs/api-reference/introduction
|
||||
modelsRouter := router.Group("/v1/models")
|
||||
modelsRouter.Use(middleware.TokenAuth())
|
||||
|
||||
Reference in New Issue
Block a user