github/new-api
1
0
Fork 0
mirror of https://github.com/QuantumNous/new-api.git synced 2026-05-03 22:14:38 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat: Add ContextKeyLocalCountTokens and update ResponseText2Usage to use context in multiple channels

Browse Source
This commit is contained in:
CaIon
2025-11-21 18:16:40 +08:00
parent ef0647285c
commit 84745d5ca4
17 changed files with 96 additions and 127 deletions
Download Patch File Download Diff File Expand all files Collapse all files

72
relay/channel/gemini/relay-gemini-native.go
View File

@@ -3,7 +3,6 @@ package gemini
import (
"io"
"net/http"
"strings"
"github.com/QuantumNous/new-api/common"
"github.com/QuantumNous/new-api/dto"
@@ -13,8 +12,6 @@ import (
"github.com/QuantumNous/new-api/service"
"github.com/QuantumNous/new-api/types"
"github.com/pkg/errors"
"github.com/gin-gonic/gin"
)
@@ -97,80 +94,15 @@ func NativeGeminiEmbeddingHandler(c *gin.Context, resp *http.Response, info *rel
}
func GeminiTextGenerationStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {
var usage = &dto.Usage{}
var imageCount int
helper.SetEventStreamHeaders(c)
responseText := strings.Builder{}
helper.StreamScannerHandler(c, resp, info, func(data string) bool {
var geminiResponse dto.GeminiChatResponse
err := common.UnmarshalJsonStr(data, &geminiResponse)
if err != nil {
logger.LogError(c, "error unmarshalling stream response: "+err.Error())
return false
}
// 统计图片数量
for _, candidate := range geminiResponse.Candidates {
for _, part := range candidate.Content.Parts {
if part.InlineData != nil && part.InlineData.MimeType != "" {
imageCount++
}
if part.Text != "" {
responseText.WriteString(part.Text)
}
}
}
// 更新使用量统计
if geminiResponse.UsageMetadata.TotalTokenCount != 0 {
usage.PromptTokens = geminiResponse.UsageMetadata.PromptTokenCount
usage.CompletionTokens = geminiResponse.UsageMetadata.CandidatesTokenCount + geminiResponse.UsageMetadata.ThoughtsTokenCount
usage.TotalTokens = geminiResponse.UsageMetadata.TotalTokenCount
usage.CompletionTokenDetails.ReasoningTokens = geminiResponse.UsageMetadata.ThoughtsTokenCount
for _, detail := range geminiResponse.UsageMetadata.PromptTokensDetails {
if detail.Modality == "AUDIO" {
usage.PromptTokensDetails.AudioTokens = detail.TokenCount
} else if detail.Modality == "TEXT" {
usage.PromptTokensDetails.TextTokens = detail.TokenCount
}
}
}
return geminiStreamHandler(c, info, resp, func(data string, geminiResponse *dto.GeminiChatResponse) bool {
// 直接发送 GeminiChatResponse 响应
err = helper.StringData(c, data)
err := helper.StringData(c, data)
if err != nil {
logger.LogError(c, err.Error())
}
info.SendResponseCount++
return true
})
if info.SendResponseCount == 0 {
return nil, types.NewOpenAIError(errors.New("no response received from Gemini API"), types.ErrorCodeEmptyResponse, http.StatusInternalServerError)
}
if imageCount != 0 {
if usage.CompletionTokens == 0 {
usage.CompletionTokens = imageCount * 258
}
}
// 如果usage.CompletionTokens为0，则使用本地统计的completion tokens
if usage.CompletionTokens == 0 {
str := responseText.String()
if len(str) > 0 {
usage = service.ResponseText2Usage(responseText.String(), info.UpstreamModelName, info.PromptTokens)
} else {
// 空补全，不需要使用量
usage = &dto.Usage{}
}
}
// 移除流式响应结尾的[Done]，因为Gemini API没有发送Done的行为
//helper.Done(c)
return usage, nil
}

97
relay/channel/gemini/relay-gemini.go
View File

@@ -954,14 +954,10 @@ func handleFinalStream(c *gin.Context, info *relaycommon.RelayInfo, resp *dto.Ch
return nil
}
func GeminiChatStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {
// responseText := ""
id := helper.GetResponseID(c)
createAt := common.GetTimestamp()
responseText := strings.Builder{}
func geminiStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response, callback func(data string, geminiResponse *dto.GeminiChatResponse) bool) (*dto.Usage, *types.NewAPIError) {
var usage = &dto.Usage{}
var imageCount int
finishReason := constant.FinishReasonStop
responseText := strings.Builder{}
helper.StreamScannerHandler(c, resp, info, func(data string) bool {
var geminiResponse dto.GeminiChatResponse
@@ -971,6 +967,7 @@ func GeminiChatStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *
return false
}
// 统计图片数量
for _, candidate := range geminiResponse.Candidates {
for _, part := range candidate.Content.Parts {
if part.InlineData != nil && part.InlineData.MimeType != "" {
@@ -982,14 +979,10 @@ func GeminiChatStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *
}
}
response, isStop := streamResponseGeminiChat2OpenAI(&geminiResponse)
response.Id = id
response.Created = createAt
response.Model = info.UpstreamModelName
// 更新使用量统计
if geminiResponse.UsageMetadata.TotalTokenCount != 0 {
usage.PromptTokens = geminiResponse.UsageMetadata.PromptTokenCount
usage.CompletionTokens = geminiResponse.UsageMetadata.CandidatesTokenCount
usage.CompletionTokens = geminiResponse.UsageMetadata.CandidatesTokenCount + geminiResponse.UsageMetadata.ThoughtsTokenCount
usage.CompletionTokenDetails.ReasoningTokens = geminiResponse.UsageMetadata.ThoughtsTokenCount
usage.TotalTokens = geminiResponse.UsageMetadata.TotalTokenCount
for _, detail := range geminiResponse.UsageMetadata.PromptTokensDetails {
@@ -1000,6 +993,45 @@ func GeminiChatStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *
}
}
}
return callback(data, &geminiResponse)
})
if imageCount != 0 {
if usage.CompletionTokens == 0 {
usage.CompletionTokens = imageCount * 1400
}
}
usage.PromptTokensDetails.TextTokens = usage.PromptTokens
if usage.TotalTokens > 0 {
usage.CompletionTokens = usage.TotalTokens - usage.PromptTokens
}
if usage.CompletionTokens <= 0 {
str := responseText.String()
if len(str) > 0 {
usage = service.ResponseText2Usage(c, responseText.String(), info.UpstreamModelName, info.PromptTokens)
} else {
usage = &dto.Usage{}
}
}
return usage, nil
}
func GeminiChatStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {
id := helper.GetResponseID(c)
createAt := common.GetTimestamp()
finishReason := constant.FinishReasonStop
usage, err := geminiStreamHandler(c, info, resp, func(data string, geminiResponse *dto.GeminiChatResponse) bool {
response, isStop := streamResponseGeminiChat2OpenAI(geminiResponse)
response.Id = id
response.Created = createAt
response.Model = info.UpstreamModelName
logger.LogDebug(c, fmt.Sprintf("info.SendResponseCount = %d", info.SendResponseCount))
if info.SendResponseCount == 0 {
// send first response
@@ -1015,7 +1047,7 @@ func GeminiChatStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *
emptyResponse.Choices[0].Delta.ToolCalls = copiedToolCalls
}
finishReason = constant.FinishReasonToolCalls
err = handleStream(c, info, emptyResponse)
err := handleStream(c, info, emptyResponse)
if err != nil {
logger.LogError(c, err.Error())
}
@@ -1025,14 +1057,14 @@ func GeminiChatStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *
response.Choices[0].FinishReason = nil
}
} else {
err = handleStream(c, info, emptyResponse)
err := handleStream(c, info, emptyResponse)
if err != nil {
logger.LogError(c, err.Error())
}
}
}
err = handleStream(c, info, response)
err := handleStream(c, info, response)
if err != nil {
logger.LogError(c, err.Error())
}
@@ -1042,40 +1074,15 @@ func GeminiChatStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *
return true
})
if info.SendResponseCount == 0 {
// 空补全，报错不计费
// empty response, throw an error
return nil, types.NewOpenAIError(errors.New("no response received from Gemini API"), types.ErrorCodeEmptyResponse, http.StatusInternalServerError)
}
if imageCount != 0 {
if usage.CompletionTokens == 0 {
usage.CompletionTokens = imageCount * 258
}
}
usage.PromptTokensDetails.TextTokens = usage.PromptTokens
usage.CompletionTokens = usage.TotalTokens - usage.PromptTokens
if usage.CompletionTokens == 0 {
str := responseText.String()
if len(str) > 0 {
usage = service.ResponseText2Usage(responseText.String(), info.UpstreamModelName, info.PromptTokens)
} else {
// 空补全，不需要使用量
usage = &dto.Usage{}
}
if err != nil {
return usage, err
}
response := helper.GenerateFinalUsageResponse(id, createAt, info.UpstreamModelName, *usage)
err := handleFinalStream(c, info, response)
if err != nil {
common.SysLog("send final response failed: " + err.Error())
handleErr := handleFinalStream(c, info, response)
if handleErr != nil {
common.SysLog("send final response failed: " + handleErr.Error())
}
//if info.RelayFormat == relaycommon.RelayFormatOpenAI {
// helper.Done(c)
//}
//resp.Body.Close()
return usage, nil
}