Skip to content

phsl / new-api

Go to a project
Commit 0191a68d by Calcium-Ion Committed by GitHub
Options

Merge pull request #3400 from seefs001/fix/openai-usage 

refactor: optimize billing flow for OpenAI-to-Anthropic convert
parents 16221f82 9ecad906
Showing with 380 additions and 140 deletions
dto/openai_response.go
...@@ -220,10 +220,12 @@ type CompletionsStreamResponse struct { ...@@ -220,10 +220,12 @@ type CompletionsStreamResponse struct {
} }
type Usage struct { type Usage struct {
PromptTokens int `json:"prompt_tokens"` PromptTokens int `json:"prompt_tokens"`
CompletionTokens int `json:"completion_tokens"` CompletionTokens int `json:"completion_tokens"`
TotalTokens int `json:"total_tokens"` TotalTokens int `json:"total_tokens"`
PromptCacheHitTokens int `json:"prompt_cache_hit_tokens,omitempty"` PromptCacheHitTokens int `json:"prompt_cache_hit_tokens,omitempty"`
UsageSemantic string `json:"usage_semantic,omitempty"`
UsageSource string `json:"usage_source,omitempty"`
PromptTokensDetails InputTokenDetails `json:"prompt_tokens_details"` PromptTokensDetails InputTokenDetails `json:"prompt_tokens_details"`
CompletionTokenDetails OutputTokenDetails `json:"completion_tokens_details"` CompletionTokenDetails OutputTokenDetails `json:"completion_tokens_details"`
...@@ -251,7 +253,7 @@ type OpenAIVideoResponse struct { ...@@ -251,7 +253,7 @@ type OpenAIVideoResponse struct {
type InputTokenDetails struct { type InputTokenDetails struct {
CachedTokens int `json:"cached_tokens"` CachedTokens int `json:"cached_tokens"`
CachedCreationTokens int `json:"-"` CachedCreationTokens int `json:"cached_creation_tokens,omitempty"`
TextTokens int `json:"text_tokens"` TextTokens int `json:"text_tokens"`
AudioTokens int `json:"audio_tokens"` AudioTokens int `json:"audio_tokens"`
ImageTokens int `json:"image_tokens"` ImageTokens int `json:"image_tokens"`
......
relay/audio_handler.go
...@@ -70,7 +70,7 @@ func AudioHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *type ...@@ -70,7 +70,7 @@ func AudioHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *type
if usage.(*dto.Usage).CompletionTokenDetails.AudioTokens > 0 || usage.(*dto.Usage).PromptTokensDetails.AudioTokens > 0 { if usage.(*dto.Usage).CompletionTokenDetails.AudioTokens > 0 || usage.(*dto.Usage).PromptTokensDetails.AudioTokens > 0 {
service.PostAudioConsumeQuota(c, info, usage.(*dto.Usage), "") service.PostAudioConsumeQuota(c, info, usage.(*dto.Usage), "")
} else { } else {
postConsumeQuota(c, info, usage.(*dto.Usage)) service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
} }
return nil return nil
......
relay/channel/claude/relay-claude.go
...@@ -555,6 +555,35 @@ type ClaudeResponseInfo struct { ...@@ -555,6 +555,35 @@ type ClaudeResponseInfo struct {
Done bool Done bool
} }
func cacheCreationTokensForOpenAIUsage(usage *dto.Usage) int {
if usage == nil {
return 0
}
splitCacheCreationTokens := usage.ClaudeCacheCreation5mTokens + usage.ClaudeCacheCreation1hTokens
if splitCacheCreationTokens == 0 {
return usage.PromptTokensDetails.CachedCreationTokens
}
if usage.PromptTokensDetails.CachedCreationTokens > splitCacheCreationTokens {
return usage.PromptTokensDetails.CachedCreationTokens
}
return splitCacheCreationTokens
}
func buildOpenAIStyleUsageFromClaudeUsage(usage *dto.Usage) dto.Usage {
if usage == nil {
return dto.Usage{}
}
clone := *usage
cacheCreationTokens := cacheCreationTokensForOpenAIUsage(usage)
totalInputTokens := usage.PromptTokens + usage.PromptTokensDetails.CachedTokens + cacheCreationTokens
clone.PromptTokens = totalInputTokens
clone.InputTokens = totalInputTokens
clone.TotalTokens = totalInputTokens + usage.CompletionTokens
clone.UsageSemantic = "openai"
clone.UsageSource = "anthropic"
return clone
}
func buildMessageDeltaPatchUsage(claudeResponse *dto.ClaudeResponse, claudeInfo *ClaudeResponseInfo) *dto.ClaudeUsage { func buildMessageDeltaPatchUsage(claudeResponse *dto.ClaudeResponse, claudeInfo *ClaudeResponseInfo) *dto.ClaudeUsage {
usage := &dto.ClaudeUsage{} usage := &dto.ClaudeUsage{}
if claudeResponse != nil && claudeResponse.Usage != nil { if claudeResponse != nil && claudeResponse.Usage != nil {
...@@ -643,6 +672,7 @@ func FormatClaudeResponseInfo(claudeResponse *dto.ClaudeResponse, oaiResponse *d ...@@ -643,6 +672,7 @@ func FormatClaudeResponseInfo(claudeResponse *dto.ClaudeResponse, oaiResponse *d
// message_start, 获取usage // message_start, 获取usage
if claudeResponse.Message != nil && claudeResponse.Message.Usage != nil { if claudeResponse.Message != nil && claudeResponse.Message.Usage != nil {
claudeInfo.Usage.PromptTokens = claudeResponse.Message.Usage.InputTokens claudeInfo.Usage.PromptTokens = claudeResponse.Message.Usage.InputTokens
claudeInfo.Usage.UsageSemantic = "anthropic"
claudeInfo.Usage.PromptTokensDetails.CachedTokens = claudeResponse.Message.Usage.CacheReadInputTokens claudeInfo.Usage.PromptTokensDetails.CachedTokens = claudeResponse.Message.Usage.CacheReadInputTokens
claudeInfo.Usage.PromptTokensDetails.CachedCreationTokens = claudeResponse.Message.Usage.CacheCreationInputTokens claudeInfo.Usage.PromptTokensDetails.CachedCreationTokens = claudeResponse.Message.Usage.CacheCreationInputTokens
claudeInfo.Usage.ClaudeCacheCreation5mTokens = claudeResponse.Message.Usage.GetCacheCreation5mTokens() claudeInfo.Usage.ClaudeCacheCreation5mTokens = claudeResponse.Message.Usage.GetCacheCreation5mTokens()
...@@ -661,6 +691,7 @@ func FormatClaudeResponseInfo(claudeResponse *dto.ClaudeResponse, oaiResponse *d ...@@ -661,6 +691,7 @@ func FormatClaudeResponseInfo(claudeResponse *dto.ClaudeResponse, oaiResponse *d
} else if claudeResponse.Type == "message_delta" { } else if claudeResponse.Type == "message_delta" {
// 最终的usage获取 // 最终的usage获取
if claudeResponse.Usage != nil { if claudeResponse.Usage != nil {
claudeInfo.Usage.UsageSemantic = "anthropic"
if claudeResponse.Usage.InputTokens > 0 { if claudeResponse.Usage.InputTokens > 0 {
// 不叠加，只取最新的 // 不叠加，只取最新的
claudeInfo.Usage.PromptTokens = claudeResponse.Usage.InputTokens claudeInfo.Usage.PromptTokens = claudeResponse.Usage.InputTokens
...@@ -754,12 +785,16 @@ func HandleStreamFinalResponse(c *gin.Context, info *relaycommon.RelayInfo, clau ...@@ -754,12 +785,16 @@ func HandleStreamFinalResponse(c *gin.Context, info *relaycommon.RelayInfo, clau
} }
claudeInfo.Usage = service.ResponseText2Usage(c, claudeInfo.ResponseText.String(), info.UpstreamModelName, claudeInfo.Usage.PromptTokens) claudeInfo.Usage = service.ResponseText2Usage(c, claudeInfo.ResponseText.String(), info.UpstreamModelName, claudeInfo.Usage.PromptTokens)
} }
if claudeInfo.Usage != nil {
claudeInfo.Usage.UsageSemantic = "anthropic"
}
if info.RelayFormat == types.RelayFormatClaude { if info.RelayFormat == types.RelayFormatClaude {
// //
} else if info.RelayFormat == types.RelayFormatOpenAI { } else if info.RelayFormat == types.RelayFormatOpenAI {
if info.ShouldIncludeUsage { if info.ShouldIncludeUsage {
response := helper.GenerateFinalUsageResponse(claudeInfo.ResponseId, claudeInfo.Created, info.UpstreamModelName, *claudeInfo.Usage) openAIUsage := buildOpenAIStyleUsageFromClaudeUsage(claudeInfo.Usage)
response := helper.GenerateFinalUsageResponse(claudeInfo.ResponseId, claudeInfo.Created, info.UpstreamModelName, openAIUsage)
err := helper.ObjectData(c, response) err := helper.ObjectData(c, response)
if err != nil { if err != nil {
common.SysLog("send final response failed: " + err.Error()) common.SysLog("send final response failed: " + err.Error())
...@@ -810,6 +845,7 @@ func HandleClaudeResponseData(c *gin.Context, info *relaycommon.RelayInfo, claud ...@@ -810,6 +845,7 @@ func HandleClaudeResponseData(c *gin.Context, info *relaycommon.RelayInfo, claud
claudeInfo.Usage.PromptTokens = claudeResponse.Usage.InputTokens claudeInfo.Usage.PromptTokens = claudeResponse.Usage.InputTokens
claudeInfo.Usage.CompletionTokens = claudeResponse.Usage.OutputTokens claudeInfo.Usage.CompletionTokens = claudeResponse.Usage.OutputTokens
claudeInfo.Usage.TotalTokens = claudeResponse.Usage.InputTokens + claudeResponse.Usage.OutputTokens claudeInfo.Usage.TotalTokens = claudeResponse.Usage.InputTokens + claudeResponse.Usage.OutputTokens
claudeInfo.Usage.UsageSemantic = "anthropic"
claudeInfo.Usage.PromptTokensDetails.CachedTokens = claudeResponse.Usage.CacheReadInputTokens claudeInfo.Usage.PromptTokensDetails.CachedTokens = claudeResponse.Usage.CacheReadInputTokens
claudeInfo.Usage.PromptTokensDetails.CachedCreationTokens = claudeResponse.Usage.CacheCreationInputTokens claudeInfo.Usage.PromptTokensDetails.CachedCreationTokens = claudeResponse.Usage.CacheCreationInputTokens
claudeInfo.Usage.ClaudeCacheCreation5mTokens = claudeResponse.Usage.GetCacheCreation5mTokens() claudeInfo.Usage.ClaudeCacheCreation5mTokens = claudeResponse.Usage.GetCacheCreation5mTokens()
...@@ -819,7 +855,7 @@ func HandleClaudeResponseData(c *gin.Context, info *relaycommon.RelayInfo, claud ...@@ -819,7 +855,7 @@ func HandleClaudeResponseData(c *gin.Context, info *relaycommon.RelayInfo, claud
switch info.RelayFormat { switch info.RelayFormat {
case types.RelayFormatOpenAI: case types.RelayFormatOpenAI:
openaiResponse := ResponseClaude2OpenAI(&claudeResponse) openaiResponse := ResponseClaude2OpenAI(&claudeResponse)
openaiResponse.Usage = *claudeInfo.Usage openaiResponse.Usage = buildOpenAIStyleUsageFromClaudeUsage(claudeInfo.Usage)
responseData, err = json.Marshal(openaiResponse) responseData, err = json.Marshal(openaiResponse)
if err != nil { if err != nil {
return types.NewError(err, types.ErrorCodeBadResponseBody) return types.NewError(err, types.ErrorCodeBadResponseBody)
......
relay/channel/claude/relay_claude_test.go
...@@ -173,3 +173,85 @@ func TestFormatClaudeResponseInfo_ContentBlockDelta(t *testing.T) { ...@@ -173,3 +173,85 @@ func TestFormatClaudeResponseInfo_ContentBlockDelta(t *testing.T) {
t.Errorf("ResponseText = %q, want %q", claudeInfo.ResponseText.String(), "hello") t.Errorf("ResponseText = %q, want %q", claudeInfo.ResponseText.String(), "hello")
} }
} }
func TestBuildOpenAIStyleUsageFromClaudeUsage(t *testing.T) {
usage := &dto.Usage{
PromptTokens: 100,
CompletionTokens: 20,
PromptTokensDetails: dto.InputTokenDetails{
CachedTokens: 30,
CachedCreationTokens: 50,
},
ClaudeCacheCreation5mTokens: 10,
ClaudeCacheCreation1hTokens: 20,
UsageSemantic: "anthropic",
}
openAIUsage := buildOpenAIStyleUsageFromClaudeUsage(usage)
if openAIUsage.PromptTokens != 180 {
t.Fatalf("PromptTokens = %d, want 180", openAIUsage.PromptTokens)
}
if openAIUsage.InputTokens != 180 {
t.Fatalf("InputTokens = %d, want 180", openAIUsage.InputTokens)
}
if openAIUsage.TotalTokens != 200 {
t.Fatalf("TotalTokens = %d, want 200", openAIUsage.TotalTokens)
}
if openAIUsage.UsageSemantic != "openai" {
t.Fatalf("UsageSemantic = %s, want openai", openAIUsage.UsageSemantic)
}
if openAIUsage.UsageSource != "anthropic" {
t.Fatalf("UsageSource = %s, want anthropic", openAIUsage.UsageSource)
}
}
func TestBuildOpenAIStyleUsageFromClaudeUsagePreservesCacheCreationRemainder(t *testing.T) {
tests := []struct {
name string
cachedCreationTokens int
cacheCreationTokens5m int
cacheCreationTokens1h int
expectedTotalInputToken int
}{
{
name: "prefers aggregate when it includes remainder",
cachedCreationTokens: 50,
cacheCreationTokens5m: 10,
cacheCreationTokens1h: 20,
expectedTotalInputToken: 180,
},
{
name: "falls back to split tokens when aggregate missing",
cachedCreationTokens: 0,
cacheCreationTokens5m: 10,
cacheCreationTokens1h: 20,
expectedTotalInputToken: 160,
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
usage := &dto.Usage{
PromptTokens: 100,
CompletionTokens: 20,
PromptTokensDetails: dto.InputTokenDetails{
CachedTokens: 30,
CachedCreationTokens: tt.cachedCreationTokens,
},
ClaudeCacheCreation5mTokens: tt.cacheCreationTokens5m,
ClaudeCacheCreation1hTokens: tt.cacheCreationTokens1h,
UsageSemantic: "anthropic",
}
openAIUsage := buildOpenAIStyleUsageFromClaudeUsage(usage)
if openAIUsage.PromptTokens != tt.expectedTotalInputToken {
t.Fatalf("PromptTokens = %d, want %d", openAIUsage.PromptTokens, tt.expectedTotalInputToken)
}
if openAIUsage.InputTokens != tt.expectedTotalInputToken {
t.Fatalf("InputTokens = %d, want %d", openAIUsage.InputTokens, tt.expectedTotalInputToken)
}
})
}
}
relay/claude_handler.go
...@@ -122,7 +122,7 @@ func ClaudeHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ ...@@ -122,7 +122,7 @@ func ClaudeHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ
return newApiErr return newApiErr
} }
service.PostClaudeConsumeQuota(c, info, usage) service.PostTextConsumeQuota(c, info, usage, nil)
return nil return nil
} }
...@@ -190,6 +190,6 @@ func ClaudeHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ ...@@ -190,6 +190,6 @@ func ClaudeHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ
return newAPIError return newAPIError
} }
service.PostClaudeConsumeQuota(c, info, usage.(*dto.Usage)) service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
return nil return nil
} }
relay/embedding_handler.go
...@@ -82,6 +82,6 @@ func EmbeddingHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError * ...@@ -82,6 +82,6 @@ func EmbeddingHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
service.ResetStatusCode(newAPIError, statusCodeMappingStr) service.ResetStatusCode(newAPIError, statusCodeMappingStr)
return newAPIError return newAPIError
} }
postConsumeQuota(c, info, usage.(*dto.Usage)) service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
return nil return nil
} }
relay/gemini_handler.go
...@@ -194,7 +194,7 @@ func GeminiHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ ...@@ -194,7 +194,7 @@ func GeminiHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ
return openaiErr return openaiErr
} }
postConsumeQuota(c, info, usage.(*dto.Usage)) service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
return nil return nil
} }
...@@ -288,6 +288,6 @@ func GeminiEmbeddingHandler(c *gin.Context, info *relaycommon.RelayInfo) (newAPI ...@@ -288,6 +288,6 @@ func GeminiEmbeddingHandler(c *gin.Context, info *relaycommon.RelayInfo) (newAPI
return openaiErr return openaiErr
} }
postConsumeQuota(c, info, usage.(*dto.Usage)) service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
return nil return nil
} }
relay/image_handler.go
...@@ -141,6 +141,6 @@ func ImageHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *type ...@@ -141,6 +141,6 @@ func ImageHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *type
logContent = append(logContent, fmt.Sprintf("生成数量 %d", imageN)) logContent = append(logContent, fmt.Sprintf("生成数量 %d", imageN))
} }
postConsumeQuota(c, info, usage.(*dto.Usage), logContent...) service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), logContent)
return nil return nil
} }
relay/rerank_handler.go
...@@ -96,6 +96,6 @@ func RerankHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ ...@@ -96,6 +96,6 @@ func RerankHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ
service.ResetStatusCode(newAPIError, statusCodeMappingStr) service.ResetStatusCode(newAPIError, statusCodeMappingStr)
return newAPIError return newAPIError
} }
postConsumeQuota(c, info, usage.(*dto.Usage)) service.PostTextConsumeQuota(c, info, usage.(*dto.Usage), nil)
return nil return nil
} }
relay/responses_handler.go
...@@ -145,7 +145,7 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError * ...@@ -145,7 +145,7 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
info.PriceData = originPriceData info.PriceData = originPriceData
return types.NewError(err, types.ErrorCodeModelPriceError, types.ErrOptionWithSkipRetry()) return types.NewError(err, types.ErrorCodeModelPriceError, types.ErrOptionWithSkipRetry())
} }
postConsumeQuota(c, info, usageDto) service.PostTextConsumeQuota(c, info, usageDto, nil)
info.OriginModelName = originModelName info.OriginModelName = originModelName
info.PriceData = originPriceData info.PriceData = originPriceData
...@@ -155,7 +155,7 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError * ...@@ -155,7 +155,7 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
if strings.HasPrefix(info.OriginModelName, "gpt-4o-audio") { if strings.HasPrefix(info.OriginModelName, "gpt-4o-audio") {
service.PostAudioConsumeQuota(c, info, usageDto, "") service.PostAudioConsumeQuota(c, info, usageDto, "")
} else { } else {
postConsumeQuota(c, info, usageDto) service.PostTextConsumeQuota(c, info, usageDto, nil)
} }
return nil return nil
} }
service/convert.go
...@@ -223,6 +223,25 @@ func generateStopBlock(index int) *dto.ClaudeResponse { ...@@ -223,6 +223,25 @@ func generateStopBlock(index int) *dto.ClaudeResponse {
} }
} }
func buildClaudeUsageFromOpenAIUsage(oaiUsage *dto.Usage) *dto.ClaudeUsage {
if oaiUsage == nil {
return nil
}
usage := &dto.ClaudeUsage{
InputTokens: oaiUsage.PromptTokens,
OutputTokens: oaiUsage.CompletionTokens,
CacheCreationInputTokens: oaiUsage.PromptTokensDetails.CachedCreationTokens,
CacheReadInputTokens: oaiUsage.PromptTokensDetails.CachedTokens,
}
if oaiUsage.ClaudeCacheCreation5mTokens > 0 || oaiUsage.ClaudeCacheCreation1hTokens > 0 {
usage.CacheCreation = &dto.ClaudeCacheCreationUsage{
Ephemeral5mInputTokens: oaiUsage.ClaudeCacheCreation5mTokens,
Ephemeral1hInputTokens: oaiUsage.ClaudeCacheCreation1hTokens,
}
}
return usage
}
func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamResponse, info *relaycommon.RelayInfo) []*dto.ClaudeResponse { func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamResponse, info *relaycommon.RelayInfo) []*dto.ClaudeResponse {
if info.ClaudeConvertInfo.Done { if info.ClaudeConvertInfo.Done {
return nil return nil
...@@ -391,13 +410,8 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon ...@@ -391,13 +410,8 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
} }
if oaiUsage != nil { if oaiUsage != nil {
claudeResponses = append(claudeResponses, &dto.ClaudeResponse{ claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
Type: "message_delta", Type: "message_delta",
Usage: &dto.ClaudeUsage{ Usage: buildClaudeUsageFromOpenAIUsage(oaiUsage),
InputTokens: oaiUsage.PromptTokens,
OutputTokens: oaiUsage.CompletionTokens,
CacheCreationInputTokens: oaiUsage.PromptTokensDetails.CachedCreationTokens,
CacheReadInputTokens: oaiUsage.PromptTokensDetails.CachedTokens,
},
Delta: &dto.ClaudeMediaMessage{ Delta: &dto.ClaudeMediaMessage{
StopReason: common.GetPointer[string](stopReasonOpenAI2Claude(info.FinishReason)), StopReason: common.GetPointer[string](stopReasonOpenAI2Claude(info.FinishReason)),
}, },
...@@ -419,13 +433,8 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon ...@@ -419,13 +433,8 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
oaiUsage := info.ClaudeConvertInfo.Usage oaiUsage := info.ClaudeConvertInfo.Usage
if oaiUsage != nil { if oaiUsage != nil {
claudeResponses = append(claudeResponses, &dto.ClaudeResponse{ claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
Type: "message_delta", Type: "message_delta",
Usage: &dto.ClaudeUsage{ Usage: buildClaudeUsageFromOpenAIUsage(oaiUsage),
InputTokens: oaiUsage.PromptTokens,
OutputTokens: oaiUsage.CompletionTokens,
CacheCreationInputTokens: oaiUsage.PromptTokensDetails.CachedCreationTokens,
CacheReadInputTokens: oaiUsage.PromptTokensDetails.CachedTokens,
},
Delta: &dto.ClaudeMediaMessage{ Delta: &dto.ClaudeMediaMessage{
StopReason: common.GetPointer[string](stopReasonOpenAI2Claude(info.FinishReason)), StopReason: common.GetPointer[string](stopReasonOpenAI2Claude(info.FinishReason)),
}, },
...@@ -555,13 +564,8 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon ...@@ -555,13 +564,8 @@ func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamRespon
} }
if oaiUsage != nil { if oaiUsage != nil {
claudeResponses = append(claudeResponses, &dto.ClaudeResponse{ claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
Type: "message_delta", Type: "message_delta",
Usage: &dto.ClaudeUsage{ Usage: buildClaudeUsageFromOpenAIUsage(oaiUsage),
InputTokens: oaiUsage.PromptTokens,
OutputTokens: oaiUsage.CompletionTokens,
CacheCreationInputTokens: oaiUsage.PromptTokensDetails.CachedCreationTokens,
CacheReadInputTokens: oaiUsage.PromptTokensDetails.CachedTokens,
},
Delta: &dto.ClaudeMediaMessage{ Delta: &dto.ClaudeMediaMessage{
StopReason: common.GetPointer[string](stopReasonOpenAI2Claude(info.FinishReason)), StopReason: common.GetPointer[string](stopReasonOpenAI2Claude(info.FinishReason)),
}, },
......
service/log_info_generate.go
...@@ -73,6 +73,7 @@ func GenerateTextOtherInfo(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, m ...@@ -73,6 +73,7 @@ func GenerateTextOtherInfo(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, m
other["admin_info"] = adminInfo other["admin_info"] = adminInfo
appendRequestPath(ctx, relayInfo, other) appendRequestPath(ctx, relayInfo, other)
appendRequestConversionChain(relayInfo, other) appendRequestConversionChain(relayInfo, other)
appendFinalRequestFormat(relayInfo, other)
appendBillingInfo(relayInfo, other) appendBillingInfo(relayInfo, other)
appendParamOverrideInfo(relayInfo, other) appendParamOverrideInfo(relayInfo, other)
return other return other
...@@ -167,6 +168,17 @@ func appendRequestConversionChain(relayInfo *relaycommon.RelayInfo, other map[st ...@@ -167,6 +168,17 @@ func appendRequestConversionChain(relayInfo *relaycommon.RelayInfo, other map[st
other["request_conversion"] = chain other["request_conversion"] = chain
} }
func appendFinalRequestFormat(relayInfo *relaycommon.RelayInfo, other map[string]interface{}) {
if relayInfo == nil || other == nil {
return
}
if relayInfo.GetFinalRequestRelayFormat() == types.RelayFormatClaude {
// claude indicates the final upstream request format is Claude Messages.
// Frontend log rendering uses this to keep the original Claude input display.
other["claude"] = true
}
}
func GenerateWssOtherInfo(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage *dto.RealtimeUsage, modelRatio, groupRatio, completionRatio, audioRatio, audioCompletionRatio, modelPrice, userGroupRatio float64) map[string]interface{} { func GenerateWssOtherInfo(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage *dto.RealtimeUsage, modelRatio, groupRatio, completionRatio, audioRatio, audioCompletionRatio, modelPrice, userGroupRatio float64) map[string]interface{} {
info := GenerateTextOtherInfo(ctx, relayInfo, modelRatio, groupRatio, completionRatio, 0, 0.0, modelPrice, userGroupRatio) info := GenerateTextOtherInfo(ctx, relayInfo, modelRatio, groupRatio, completionRatio, 0, 0.0, modelPrice, userGroupRatio)
info["ws"] = true info["ws"] = true
......
service/quota.go
...@@ -235,108 +235,6 @@ func PostWssConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, mod ...@@ -235,108 +235,6 @@ func PostWssConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, mod
}) })
} }
func PostClaudeConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage *dto.Usage) {
if usage != nil {
ObserveChannelAffinityUsageCacheByRelayFormat(ctx, usage, relayInfo.GetFinalRequestRelayFormat())
}
useTimeSeconds := time.Now().Unix() - relayInfo.StartTime.Unix()
promptTokens := usage.PromptTokens
completionTokens := usage.CompletionTokens
modelName := relayInfo.OriginModelName
tokenName := ctx.GetString("token_name")
completionRatio := relayInfo.PriceData.CompletionRatio
modelRatio := relayInfo.PriceData.ModelRatio
groupRatio := relayInfo.PriceData.GroupRatioInfo.GroupRatio
modelPrice := relayInfo.PriceData.ModelPrice
cacheRatio := relayInfo.PriceData.CacheRatio
cacheTokens := usage.PromptTokensDetails.CachedTokens
cacheCreationRatio := relayInfo.PriceData.CacheCreationRatio
cacheCreationRatio5m := relayInfo.PriceData.CacheCreation5mRatio
cacheCreationRatio1h := relayInfo.PriceData.CacheCreation1hRatio
cacheCreationTokens := usage.PromptTokensDetails.CachedCreationTokens
cacheCreationTokens5m := usage.ClaudeCacheCreation5mTokens
cacheCreationTokens1h := usage.ClaudeCacheCreation1hTokens
if relayInfo.ChannelType == constant.ChannelTypeOpenRouter {
promptTokens -= cacheTokens
isUsingCustomSettings := relayInfo.PriceData.UsePrice || hasCustomModelRatio(modelName, relayInfo.PriceData.ModelRatio)
if cacheCreationTokens == 0 && relayInfo.PriceData.CacheCreationRatio != 1 && usage.Cost != 0 && !isUsingCustomSettings {
maybeCacheCreationTokens := CalcOpenRouterCacheCreateTokens(*usage, relayInfo.PriceData)
if maybeCacheCreationTokens >= 0 && promptTokens >= maybeCacheCreationTokens {
cacheCreationTokens = maybeCacheCreationTokens
}
}
promptTokens -= cacheCreationTokens
}
calculateQuota := 0.0
if !relayInfo.PriceData.UsePrice {
calculateQuota = float64(promptTokens)
calculateQuota += float64(cacheTokens) * cacheRatio
calculateQuota += float64(cacheCreationTokens5m) * cacheCreationRatio5m
calculateQuota += float64(cacheCreationTokens1h) * cacheCreationRatio1h
remainingCacheCreationTokens := cacheCreationTokens - cacheCreationTokens5m - cacheCreationTokens1h
if remainingCacheCreationTokens > 0 {
calculateQuota += float64(remainingCacheCreationTokens) * cacheCreationRatio
}
calculateQuota += float64(completionTokens) * completionRatio
calculateQuota = calculateQuota * groupRatio * modelRatio
} else {
calculateQuota = modelPrice * common.QuotaPerUnit * groupRatio
}
if modelRatio != 0 && calculateQuota <= 0 {
calculateQuota = 1
}
quota := int(calculateQuota)
totalTokens := promptTokens + completionTokens
var logContent string
// record all the consume log even if quota is 0
if totalTokens == 0 {
// in this case, must be some error happened
// we cannot just return, because we may have to return the pre-consumed quota
quota = 0
logContent += fmt.Sprintf("（可能是上游出错）")
logger.LogError(ctx, fmt.Sprintf("total tokens is 0, cannot consume quota, userId %d, channelId %d, "+
"tokenId %d, model %s， pre-consumed quota %d", relayInfo.UserId, relayInfo.ChannelId, relayInfo.TokenId, modelName, relayInfo.FinalPreConsumedQuota))
} else {
model.UpdateUserUsedQuotaAndRequestCount(relayInfo.UserId, quota)
model.UpdateChannelUsedQuota(relayInfo.ChannelId, quota)
}
if err := SettleBilling(ctx, relayInfo, quota); err != nil {
logger.LogError(ctx, "error settling billing: "+err.Error())
}
other := GenerateClaudeOtherInfo(ctx, relayInfo, modelRatio, groupRatio, completionRatio,
cacheTokens, cacheRatio,
cacheCreationTokens, cacheCreationRatio,
cacheCreationTokens5m, cacheCreationRatio5m,
cacheCreationTokens1h, cacheCreationRatio1h,
modelPrice, relayInfo.PriceData.GroupRatioInfo.GroupSpecialRatio)
model.RecordConsumeLog(ctx, relayInfo.UserId, model.RecordConsumeLogParams{
ChannelId: relayInfo.ChannelId,
PromptTokens: promptTokens,
CompletionTokens: completionTokens,
ModelName: modelName,
TokenName: tokenName,
Quota: quota,
Content: logContent,
TokenId: relayInfo.TokenId,
UseTimeSeconds: int(useTimeSeconds),
IsStream: relayInfo.IsStream,
Group: relayInfo.UsingGroup,
Other: other,
})
}
func CalcOpenRouterCacheCreateTokens(usage dto.Usage, priceData types.PriceData) int { func CalcOpenRouterCacheCreateTokens(usage dto.Usage, priceData types.PriceData) int {
if priceData.CacheCreationRatio == 1 { if priceData.CacheCreationRatio == 1 {
return 0 return 0
......
service/text_quota_test.go 0 → 100644
package service
import (
"net/http/httptest"
"testing"
"time"
"github.com/QuantumNous/new-api/dto"
relaycommon "github.com/QuantumNous/new-api/relay/common"
"github.com/QuantumNous/new-api/types"
"github.com/gin-gonic/gin"
"github.com/stretchr/testify/require"
)
func TestCalculateTextQuotaSummaryUnifiedForClaudeSemantic(t *testing.T) {
gin.SetMode(gin.TestMode)
w := httptest.NewRecorder()
ctx, _ := gin.CreateTestContext(w)
usage := &dto.Usage{
PromptTokens: 1000,
CompletionTokens: 200,
PromptTokensDetails: dto.InputTokenDetails{
CachedTokens: 100,
CachedCreationTokens: 50,
},
ClaudeCacheCreation5mTokens: 10,
ClaudeCacheCreation1hTokens: 20,
}
priceData := types.PriceData{
ModelRatio: 1,
CompletionRatio: 2,
CacheRatio: 0.1,
CacheCreationRatio: 1.25,
CacheCreation5mRatio: 1.25,
CacheCreation1hRatio: 2,
GroupRatioInfo: types.GroupRatioInfo{
GroupRatio: 1,
},
}
chatRelayInfo := &relaycommon.RelayInfo{
RelayFormat: types.RelayFormatOpenAI,
FinalRequestRelayFormat: types.RelayFormatClaude,
OriginModelName: "claude-3-7-sonnet",
PriceData: priceData,
StartTime: time.Now(),
}
messageRelayInfo := &relaycommon.RelayInfo{
RelayFormat: types.RelayFormatClaude,
FinalRequestRelayFormat: types.RelayFormatClaude,
OriginModelName: "claude-3-7-sonnet",
PriceData: priceData,
StartTime: time.Now(),
}
chatSummary := calculateTextQuotaSummary(ctx, chatRelayInfo, usage)
messageSummary := calculateTextQuotaSummary(ctx, messageRelayInfo, usage)
require.Equal(t, messageSummary.Quota, chatSummary.Quota)
require.Equal(t, messageSummary.CacheCreationTokens5m, chatSummary.CacheCreationTokens5m)
require.Equal(t, messageSummary.CacheCreationTokens1h, chatSummary.CacheCreationTokens1h)
require.True(t, chatSummary.IsClaudeUsageSemantic)
require.Equal(t, 1488, chatSummary.Quota)
}
func TestCalculateTextQuotaSummaryUsesSplitClaudeCacheCreationRatios(t *testing.T) {
gin.SetMode(gin.TestMode)
w := httptest.NewRecorder()
ctx, _ := gin.CreateTestContext(w)
relayInfo := &relaycommon.RelayInfo{
RelayFormat: types.RelayFormatOpenAI,
FinalRequestRelayFormat: types.RelayFormatClaude,
OriginModelName: "claude-3-7-sonnet",
PriceData: types.PriceData{
ModelRatio: 1,
CompletionRatio: 1,
CacheRatio: 0,
CacheCreationRatio: 1,
CacheCreation5mRatio: 2,
CacheCreation1hRatio: 3,
GroupRatioInfo: types.GroupRatioInfo{
GroupRatio: 1,
},
},
StartTime: time.Now(),
}
usage := &dto.Usage{
PromptTokens: 100,
CompletionTokens: 0,
PromptTokensDetails: dto.InputTokenDetails{
CachedCreationTokens: 10,
},
ClaudeCacheCreation5mTokens: 2,
ClaudeCacheCreation1hTokens: 3,
}
summary := calculateTextQuotaSummary(ctx, relayInfo, usage)
// 100 + remaining(5)*1 + 2*2 + 3*3 = 118
require.Equal(t, 118, summary.Quota)
}
func TestCalculateTextQuotaSummaryUsesAnthropicUsageSemanticFromUpstreamUsage(t *testing.T) {
gin.SetMode(gin.TestMode)
w := httptest.NewRecorder()
ctx, _ := gin.CreateTestContext(w)
relayInfo := &relaycommon.RelayInfo{
RelayFormat: types.RelayFormatOpenAI,
OriginModelName: "claude-3-7-sonnet",
PriceData: types.PriceData{
ModelRatio: 1,
CompletionRatio: 2,
CacheRatio: 0.1,
CacheCreationRatio: 1.25,
CacheCreation5mRatio: 1.25,
CacheCreation1hRatio: 2,
GroupRatioInfo: types.GroupRatioInfo{
GroupRatio: 1,
},
},
StartTime: time.Now(),
}
usage := &dto.Usage{
PromptTokens: 1000,
CompletionTokens: 200,
UsageSemantic: "anthropic",
PromptTokensDetails: dto.InputTokenDetails{
CachedTokens: 100,
CachedCreationTokens: 50,
},
ClaudeCacheCreation5mTokens: 10,
ClaudeCacheCreation1hTokens: 20,
}
summary := calculateTextQuotaSummary(ctx, relayInfo, usage)
require.True(t, summary.IsClaudeUsageSemantic)
require.Equal(t, "anthropic", summary.UsageSemantic)
require.Equal(t, 1488, summary.Quota)
}
func TestCacheWriteTokensTotal(t *testing.T) {
t.Run("split cache creation", func(t *testing.T) {
summary := textQuotaSummary{
CacheCreationTokens: 50,
CacheCreationTokens5m: 10,
CacheCreationTokens1h: 20,
}
require.Equal(t, 50, cacheWriteTokensTotal(summary))
})
t.Run("legacy cache creation", func(t *testing.T) {
summary := textQuotaSummary{CacheCreationTokens: 50}
require.Equal(t, 50, cacheWriteTokensTotal(summary))
})
t.Run("split cache creation without aggregate remainder", func(t *testing.T) {
summary := textQuotaSummary{
CacheCreationTokens5m: 10,
CacheCreationTokens1h: 20,
}
require.Equal(t, 30, cacheWriteTokensTotal(summary))
})
}
func TestCalculateTextQuotaSummaryHandlesLegacyClaudeDerivedOpenAIUsage(t *testing.T) {
gin.SetMode(gin.TestMode)
w := httptest.NewRecorder()
ctx, _ := gin.CreateTestContext(w)
relayInfo := &relaycommon.RelayInfo{
RelayFormat: types.RelayFormatOpenAI,
OriginModelName: "claude-3-7-sonnet",
PriceData: types.PriceData{
ModelRatio: 1,
CompletionRatio: 5,
CacheRatio: 0.1,
CacheCreationRatio: 1.25,
CacheCreation5mRatio: 1.25,
CacheCreation1hRatio: 2,
GroupRatioInfo: types.GroupRatioInfo{GroupRatio: 1},
},
StartTime: time.Now(),
}
usage := &dto.Usage{
PromptTokens: 62,
CompletionTokens: 95,
PromptTokensDetails: dto.InputTokenDetails{
CachedTokens: 3544,
},
ClaudeCacheCreation5mTokens: 586,
}
summary := calculateTextQuotaSummary(ctx, relayInfo, usage)
// 62 + 3544*0.1 + 586*1.25 + 95*5 = 1624.9 => 1624
require.Equal(t, 1624, summary.Quota)
}
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or sign in to comment