chaos/chaos-api
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: preserve explicit zero values in native relay requests

Browse Source
This commit is contained in:
Seefs
2026-03-01 15:47:03 +08:00
parent e22f59e449
commit 2cf3c1836c
52 changed files with 523 additions and 232 deletions
Download Patch File Download Diff File Expand all files Collapse all files
relay/channel/ali/image.go
+2 -1
View File
@@ -18,6 +18,7 @@ import (
"github.com/QuantumNous/new-api/types"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
func oaiImage2AliImageRequest(info *relaycommon.RelayInfo, request dto.ImageRequest, isSync bool) (*AliImageRequest, error) {
@@ -34,7 +35,7 @@ func oaiImage2AliImageRequest(info *relaycommon.RelayInfo, request dto.ImageRequ
// 兼容没有parameters字段的情况，从openai标准字段中提取参数
imageRequest.Parameters = AliImageParameters{
Size: strings.Replace(request.Size, "x", "*", -1),
N: int(request.N),
N: int(lo.FromPtrOr(request.N, uint(1))),
Watermark: request.Watermark,
}
}
relay/channel/ali/image_wan.go
+2 -1
View File
@@ -9,6 +9,7 @@ import (
relaycommon "github.com/QuantumNous/new-api/relay/common"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
func oaiFormEdit2WanxImageEdit(c *gin.Context, info *relaycommon.RelayInfo, request dto.ImageRequest) (*AliImageRequest, error) {
@@ -31,7 +32,7 @@ func oaiFormEdit2WanxImageEdit(c *gin.Context, info *relaycommon.RelayInfo, requ
//}
imageRequest.Input = wanInput
imageRequest.Parameters = AliImageParameters{
N: int(request.N),
N: int(lo.FromPtrOr(request.N, uint(1))),
}
info.PriceData.AddOtherRatio("n", float64(imageRequest.Parameters.N))
relay/channel/ali/rerank.go
+1 -1
View File
@@ -26,7 +26,7 @@ func ConvertRerankRequest(request dto.RerankRequest) *AliRerankRequest {
Documents: request.Documents,
},
Parameters: AliRerankParameters{
TopN: &request.TopN,
TopN: request.TopN,
ReturnDocuments: returnDocuments,
},
}
relay/channel/ali/text.go
+6 -4
View File
@@ -2,6 +2,7 @@ package ali
import (
"github.com/QuantumNous/new-api/dto"
"github.com/samber/lo"
)
// https://help.aliyun.com/document_detail/613695.html?spm=a2c4g.2399480.0.0.1adb778fAdzP9w#341800c0f8w0r
@@ -9,10 +10,11 @@ import (
const EnableSearchModelSuffix = "-internet"
func requestOpenAI2Ali(request dto.GeneralOpenAIRequest) *dto.GeneralOpenAIRequest {
if request.TopP >= 1 {
request.TopP = 0.999
} else if request.TopP <= 0 {
request.TopP = 0.001
topP := lo.FromPtrOr(request.TopP, 0)
if topP >= 1 {
request.TopP = lo.ToPtr(0.999)
} else if topP <= 0 {
request.TopP = lo.ToPtr(0.001)
}
return &request
}
relay/channel/aws/dto.go
+7 -7
View File
@@ -94,19 +94,19 @@ func convertToNovaRequest(req *dto.GeneralOpenAIRequest) *NovaRequest {
}
// 设置推理配置
if req.MaxTokens != 0 || (req.Temperature != nil && *req.Temperature != 0) || req.TopP != 0 || req.TopK != 0 || req.Stop != nil {
if (req.MaxTokens != nil && *req.MaxTokens != 0) || (req.Temperature != nil && *req.Temperature != 0) || (req.TopP != nil && *req.TopP != 0) || (req.TopK != nil && *req.TopK != 0) || req.Stop != nil {
novaReq.InferenceConfig = &NovaInferenceConfig{}
if req.MaxTokens != 0 {
novaReq.InferenceConfig.MaxTokens = int(req.MaxTokens)
if req.MaxTokens != nil && *req.MaxTokens != 0 {
novaReq.InferenceConfig.MaxTokens = int(*req.MaxTokens)
}
if req.Temperature != nil && *req.Temperature != 0 {
novaReq.InferenceConfig.Temperature = *req.Temperature
}
if req.TopP != 0 {
novaReq.InferenceConfig.TopP = req.TopP
if req.TopP != nil && *req.TopP != 0 {
novaReq.InferenceConfig.TopP = *req.TopP
}
if req.TopK != 0 {
novaReq.InferenceConfig.TopK = req.TopK
if req.TopK != nil && *req.TopK != 0 {
novaReq.InferenceConfig.TopK = *req.TopK
}
if req.Stop != nil {
if stopSequences := parseStopSequences(req.Stop); len(stopSequences) > 0 {
relay/channel/baidu/relay-baidu.go
+4 -3
View File
@@ -17,6 +17,7 @@ import (
"github.com/QuantumNous/new-api/relay/helper"
"github.com/QuantumNous/new-api/service"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/gin-gonic/gin"
)
@@ -28,9 +29,9 @@ var baiduTokenStore sync.Map
func requestOpenAI2Baidu(request dto.GeneralOpenAIRequest) *BaiduChatRequest {
baiduRequest := BaiduChatRequest{
Temperature: request.Temperature,
TopP: request.TopP,
PenaltyScore: request.FrequencyPenalty,
Stream: request.Stream,
TopP: lo.FromPtrOr(request.TopP, 0),
PenaltyScore: lo.FromPtrOr(request.FrequencyPenalty, 0),
Stream: lo.FromPtrOr(request.Stream, false),
DisableSearch: false,
EnableCitation: false,
UserId: request.User,
relay/channel/claude/relay-claude.go
+20 -11
View File
@@ -123,14 +123,22 @@ func RequestOpenAI2ClaudeMessage(c *gin.Context, textRequest dto.GeneralOpenAIRe
claudeRequest := dto.ClaudeRequest{
Model: textRequest.Model,
MaxTokens: textRequest.GetMaxTokens(),
StopSequences: nil,
Temperature: textRequest.Temperature,
TopP: textRequest.TopP,
TopK: textRequest.TopK,
Stream: textRequest.Stream,
Tools: claudeTools,
}
if maxTokens := textRequest.GetMaxTokens(); maxTokens > 0 {
claudeRequest.MaxTokens = common.GetPointer(maxTokens)
}
if textRequest.TopP != nil {
claudeRequest.TopP = common.GetPointer(*textRequest.TopP)
}
if textRequest.TopK != nil {
claudeRequest.TopK = common.GetPointer(*textRequest.TopK)
}
if textRequest.IsStream(nil) {
claudeRequest.Stream = common.GetPointer(true)
}
// 处理 tool_choice 和 parallel_tool_calls
if textRequest.ToolChoice != nil || textRequest.ParallelTooCalls != nil {
@@ -140,8 +148,9 @@ func RequestOpenAI2ClaudeMessage(c *gin.Context, textRequest dto.GeneralOpenAIRe
}
}
if claudeRequest.MaxTokens == 0 {
claudeRequest.MaxTokens = uint(model_setting.GetClaudeSettings().GetDefaultMaxTokens(textRequest.Model))
if claudeRequest.MaxTokens == nil || *claudeRequest.MaxTokens == 0 {
defaultMaxTokens := uint(model_setting.GetClaudeSettings().GetDefaultMaxTokens(textRequest.Model))
claudeRequest.MaxTokens = &defaultMaxTokens
}
if baseModel, effortLevel, ok := reasoning.TrimEffortSuffix(textRequest.Model); ok && effortLevel != "" &&
@@ -151,24 +160,24 @@ func RequestOpenAI2ClaudeMessage(c *gin.Context, textRequest dto.GeneralOpenAIRe
Type: "adaptive",
}
claudeRequest.OutputConfig = json.RawMessage(fmt.Sprintf(`{"effort":"%s"}`, effortLevel))
claudeRequest.TopP = 0
claudeRequest.TopP = common.GetPointer[float64](0)
claudeRequest.Temperature = common.GetPointer[float64](1.0)
} else if model_setting.GetClaudeSettings().ThinkingAdapterEnabled &&
strings.HasSuffix(textRequest.Model, "-thinking") {
// 因为BudgetTokens 必须大于1024
if claudeRequest.MaxTokens < 1280 {
claudeRequest.MaxTokens = 1280
if claudeRequest.MaxTokens == nil || *claudeRequest.MaxTokens < 1280 {
claudeRequest.MaxTokens = common.GetPointer[uint](1280)
}
// BudgetTokens 为 max_tokens 的 80%
claudeRequest.Thinking = &dto.Thinking{
Type: "enabled",
BudgetTokens: common.GetPointer[int](int(float64(claudeRequest.MaxTokens) * model_setting.GetClaudeSettings().ThinkingAdapterBudgetTokensPercentage)),
BudgetTokens: common.GetPointer[int](int(float64(*claudeRequest.MaxTokens) * model_setting.GetClaudeSettings().ThinkingAdapterBudgetTokensPercentage)),
}
// TODO: 临时处理
// https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking#important-considerations-when-using-extended-thinking
claudeRequest.TopP = 0
claudeRequest.TopP = common.GetPointer[float64](0)
claudeRequest.Temperature = common.GetPointer[float64](1.0)
if !model_setting.ShouldPreserveThinkingSuffix(textRequest.Model) {
claudeRequest.Model = strings.TrimSuffix(textRequest.Model, "-thinking")
relay/channel/cloudflare/relay_cloudflare.go
+2 -1
View File
@@ -14,6 +14,7 @@ import (
"github.com/QuantumNous/new-api/relay/helper"
"github.com/QuantumNous/new-api/service"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/gin-gonic/gin"
)
@@ -23,7 +24,7 @@ func convertCf2CompletionsRequest(textRequest dto.GeneralOpenAIRequest) *CfReque
return &CfRequest{
Prompt: p,
MaxTokens: textRequest.GetMaxTokens(),
Stream: textRequest.Stream,
Stream: lo.FromPtrOr(textRequest.Stream, false),
Temperature: textRequest.Temperature,
}
}
relay/channel/codex/adaptor.go
+1 -1
View File
@@ -102,7 +102,7 @@ func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommo
// codex: store must be false
request.Store = json.RawMessage("false")
// rm max_output_tokens
request.MaxOutputTokens = 0
request.MaxOutputTokens = nil
request.Temperature = nil
return request, nil
}
relay/channel/cohere/relay-cohere.go
+6 -4
View File
@@ -16,6 +16,7 @@ import (
"github.com/QuantumNous/new-api/types"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
func requestOpenAI2Cohere(textRequest dto.GeneralOpenAIRequest) *CohereRequest {
@@ -23,7 +24,7 @@ func requestOpenAI2Cohere(textRequest dto.GeneralOpenAIRequest) *CohereRequest {
Model: textRequest.Model,
ChatHistory: []ChatHistory{},
Message: "",
Stream: textRequest.Stream,
Stream: lo.FromPtrOr(textRequest.Stream, false),
MaxTokens: textRequest.GetMaxTokens(),
}
if common.CohereSafetySetting != "NONE" {
@@ -55,14 +56,15 @@ func requestOpenAI2Cohere(textRequest dto.GeneralOpenAIRequest) *CohereRequest {
}
func requestConvertRerank2Cohere(rerankRequest dto.RerankRequest) *CohereRerankRequest {
if rerankRequest.TopN == 0 {
rerankRequest.TopN = 1
topN := lo.FromPtrOr(rerankRequest.TopN, 1)
if topN <= 0 {
topN = 1
}
cohereReq := CohereRerankRequest{
Query: rerankRequest.Query,
Documents: rerankRequest.Documents,
Model: rerankRequest.Model,
TopN: rerankRequest.TopN,
TopN: topN,
ReturnDocuments: true,
}
return &cohereReq
relay/channel/coze/relay-coze.go
+2 -1
View File
@@ -15,6 +15,7 @@ import (
"github.com/QuantumNous/new-api/relay/helper"
"github.com/QuantumNous/new-api/service"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/gin-gonic/gin"
)
@@ -40,7 +41,7 @@ func convertCozeChatRequest(c *gin.Context, request dto.GeneralOpenAIRequest) *C
BotId: c.GetString("bot_id"),
UserId: user,
AdditionalMessages: messages,
Stream: request.Stream,
Stream: lo.FromPtrOr(request.Stream, false),
}
return cozeRequest
}
relay/channel/dify/relay-dify.go
+2 -1
View File
@@ -18,6 +18,7 @@ import (
"github.com/QuantumNous/new-api/relay/helper"
"github.com/QuantumNous/new-api/service"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/gin-gonic/gin"
)
@@ -168,7 +169,7 @@ func requestOpenAI2Dify(c *gin.Context, info *relaycommon.RelayInfo, request dto
difyReq.Query = content.String()
difyReq.Files = files
mode := "blocking"
if request.Stream {
if lo.FromPtrOr(request.Stream, false) {
mode = "streaming"
}
difyReq.ResponseMode = mode
relay/channel/gemini/adaptor.go
+5 -3
View File
@@ -17,6 +17,7 @@ import (
"github.com/QuantumNous/new-api/types"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
type Adaptor struct {
@@ -91,7 +92,7 @@ func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInf
},
},
Parameters: dto.GeminiImageParameters{
SampleCount: int(request.N),
SampleCount: int(lo.FromPtrOr(request.N, uint(1))),
AspectRatio: aspectRatio,
PersonGeneration: "allow_adult", // default allow adult
},
@@ -223,8 +224,9 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
switch info.UpstreamModelName {
case "text-embedding-004", "gemini-embedding-exp-03-07", "gemini-embedding-001":
// Only newer models introduced after 2024 support OutputDimensionality
if request.Dimensions > 0 {
geminiRequest["outputDimensionality"] = request.Dimensions
dimensions := lo.FromPtrOr(request.Dimensions, 0)
if dimensions > 0 {
geminiRequest["outputDimensionality"] = dimensions
}
}
geminiRequests = append(geminiRequests, geminiRequest)
relay/channel/gemini/relay-gemini.go
+17 -6
View File
@@ -24,6 +24,7 @@ import (
"github.com/QuantumNous/new-api/setting/reasoning"
"github.com/QuantumNous/new-api/types"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
// https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/inference?hl=zh-cn#blob
@@ -167,8 +168,8 @@ func ThinkingAdaptor(geminiRequest *dto.GeminiChatRequest, info *relaycommon.Rel
geminiRequest.GenerationConfig.ThinkingConfig = &dto.GeminiThinkingConfig{
IncludeThoughts: true,
}
if geminiRequest.GenerationConfig.MaxOutputTokens > 0 {
budgetTokens := model_setting.GetGeminiSettings().ThinkingAdapterBudgetTokensPercentage * float64(geminiRequest.GenerationConfig.MaxOutputTokens)
if geminiRequest.GenerationConfig.MaxOutputTokens != nil && *geminiRequest.GenerationConfig.MaxOutputTokens > 0 {
budgetTokens := model_setting.GetGeminiSettings().ThinkingAdapterBudgetTokensPercentage * float64(*geminiRequest.GenerationConfig.MaxOutputTokens)
clampedBudget := clampThinkingBudget(modelName, int(budgetTokens))
geminiRequest.GenerationConfig.ThinkingConfig.ThinkingBudget = common.GetPointer(clampedBudget)
} else {
@@ -200,13 +201,23 @@ func CovertOpenAI2Gemini(c *gin.Context, textRequest dto.GeneralOpenAIRequest, i
geminiRequest := dto.GeminiChatRequest{
Contents: make([]dto.GeminiChatContent, 0, len(textRequest.Messages)),
GenerationConfig: dto.GeminiChatGenerationConfig{
Temperature: textRequest.Temperature,
TopP: textRequest.TopP,
MaxOutputTokens: textRequest.GetMaxTokens(),
Seed: int64(textRequest.Seed),
Temperature: textRequest.Temperature,
},
}
if textRequest.TopP != nil && *textRequest.TopP > 0 {
geminiRequest.GenerationConfig.TopP = common.GetPointer(*textRequest.TopP)
}
if maxTokens := textRequest.GetMaxTokens(); maxTokens > 0 {
geminiRequest.GenerationConfig.MaxOutputTokens = common.GetPointer(maxTokens)
}
if textRequest.Seed != nil && *textRequest.Seed != 0 {
geminiSeed := int64(lo.FromPtr(textRequest.Seed))
geminiRequest.GenerationConfig.Seed = common.GetPointer(geminiSeed)
}
attachThoughtSignature := (info.ChannelType == constant.ChannelTypeGemini ||
info.ChannelType == constant.ChannelTypeVertexAi) &&
model_setting.GetGeminiSettings().FunctionCallThoughtSignatureEnabled
relay/channel/minimax/adaptor.go
+2 -1
View File
@@ -17,6 +17,7 @@ import (
"github.com/QuantumNous/new-api/types"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
type Adaptor struct {
@@ -37,7 +38,7 @@ func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInf
}
voiceID := request.Voice
speed := request.Speed
speed := lo.FromPtrOr(request.Speed, 0.0)
outputFormat := request.ResponseFormat
minimaxRequest := MiniMaxTTSRequest{
relay/channel/mistral/text.go
+6 -2
View File
@@ -66,14 +66,18 @@ func requestOpenAI2Mistral(request *dto.GeneralOpenAIRequest) *dto.GeneralOpenAI
ToolCallId: message.ToolCallId,
})
}
return &dto.GeneralOpenAIRequest{
out := &dto.GeneralOpenAIRequest{
Model: request.Model,
Stream: request.Stream,
Messages: messages,
Temperature: request.Temperature,
TopP: request.TopP,
MaxTokens: request.GetMaxTokens(),
Tools: request.Tools,
ToolChoice: request.ToolChoice,
}
if request.MaxTokens != nil || request.MaxCompletionTokens != nil {
maxTokens := request.GetMaxTokens()
out.MaxTokens = &maxTokens
}
return out
}
relay/channel/ollama/relay-ollama.go
+36 -34
View File
@@ -16,12 +16,13 @@ import (
"github.com/QuantumNous/new-api/types"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
func openAIChatToOllamaChat(c *gin.Context, r *dto.GeneralOpenAIRequest) (*OllamaChatRequest, error) {
chatReq := &OllamaChatRequest{
Model: r.Model,
Stream: r.Stream,
Stream: lo.FromPtrOr(r.Stream, false),
Options: map[string]any{},
Think: r.Think,
}
@@ -41,20 +42,20 @@ func openAIChatToOllamaChat(c *gin.Context, r *dto.GeneralOpenAIRequest) (*Ollam
if r.Temperature != nil {
chatReq.Options["temperature"] = r.Temperature
}
if r.TopP != 0 {
chatReq.Options["top_p"] = r.TopP
if r.TopP != nil {
chatReq.Options["top_p"] = lo.FromPtr(r.TopP)
}
if r.TopK != 0 {
chatReq.Options["top_k"] = r.TopK
if r.TopK != nil {
chatReq.Options["top_k"] = lo.FromPtr(r.TopK)
}
if r.FrequencyPenalty != 0 {
chatReq.Options["frequency_penalty"] = r.FrequencyPenalty
if r.FrequencyPenalty != nil {
chatReq.Options["frequency_penalty"] = lo.FromPtr(r.FrequencyPenalty)
}
if r.PresencePenalty != 0 {
chatReq.Options["presence_penalty"] = r.PresencePenalty
if r.PresencePenalty != nil {
chatReq.Options["presence_penalty"] = lo.FromPtr(r.PresencePenalty)
}
if r.Seed != 0 {
chatReq.Options["seed"] = int(r.Seed)
if r.Seed != nil {
chatReq.Options["seed"] = int(lo.FromPtr(r.Seed))
}
if mt := r.GetMaxTokens(); mt != 0 {
chatReq.Options["num_predict"] = int(mt)
@@ -155,7 +156,7 @@ func openAIChatToOllamaChat(c *gin.Context, r *dto.GeneralOpenAIRequest) (*Ollam
func openAIToGenerate(c *gin.Context, r *dto.GeneralOpenAIRequest) (*OllamaGenerateRequest, error) {
gen := &OllamaGenerateRequest{
Model: r.Model,
Stream: r.Stream,
Stream: lo.FromPtrOr(r.Stream, false),
Options: map[string]any{},
Think: r.Think,
}
@@ -193,20 +194,20 @@ func openAIToGenerate(c *gin.Context, r *dto.GeneralOpenAIRequest) (*OllamaGener
if r.Temperature != nil {
gen.Options["temperature"] = r.Temperature
}
if r.TopP != 0 {
gen.Options["top_p"] = r.TopP
if r.TopP != nil {
gen.Options["top_p"] = lo.FromPtr(r.TopP)
}
if r.TopK != 0 {
gen.Options["top_k"] = r.TopK
if r.TopK != nil {
gen.Options["top_k"] = lo.FromPtr(r.TopK)
}
if r.FrequencyPenalty != 0 {
gen.Options["frequency_penalty"] = r.FrequencyPenalty
if r.FrequencyPenalty != nil {
gen.Options["frequency_penalty"] = lo.FromPtr(r.FrequencyPenalty)
}
if r.PresencePenalty != 0 {
gen.Options["presence_penalty"] = r.PresencePenalty
if r.PresencePenalty != nil {
gen.Options["presence_penalty"] = lo.FromPtr(r.PresencePenalty)
}
if r.Seed != 0 {
gen.Options["seed"] = int(r.Seed)
if r.Seed != nil {
gen.Options["seed"] = int(lo.FromPtr(r.Seed))
}
if mt := r.GetMaxTokens(); mt != 0 {
gen.Options["num_predict"] = int(mt)
@@ -237,26 +238,27 @@ func requestOpenAI2Embeddings(r dto.EmbeddingRequest) *OllamaEmbeddingRequest {
if r.Temperature != nil {
opts["temperature"] = r.Temperature
}
if r.TopP != 0 {
opts["top_p"] = r.TopP
if r.TopP != nil {
opts["top_p"] = lo.FromPtr(r.TopP)
}
if r.FrequencyPenalty != 0 {
opts["frequency_penalty"] = r.FrequencyPenalty
if r.FrequencyPenalty != nil {
opts["frequency_penalty"] = lo.FromPtr(r.FrequencyPenalty)
}
if r.PresencePenalty != 0 {
opts["presence_penalty"] = r.PresencePenalty
if r.PresencePenalty != nil {
opts["presence_penalty"] = lo.FromPtr(r.PresencePenalty)
}
if r.Seed != 0 {
opts["seed"] = int(r.Seed)
if r.Seed != nil {
opts["seed"] = int(lo.FromPtr(r.Seed))
}
if r.Dimensions != 0 {
opts["dimensions"] = r.Dimensions
dimensions := lo.FromPtrOr(r.Dimensions, 0)
if r.Dimensions != nil {
opts["dimensions"] = dimensions
}
input := r.ParseInput()
if len(input) == 1 {
return &OllamaEmbeddingRequest{Model: r.Model, Input: input[0], Options: opts, Dimensions: r.Dimensions}
return &OllamaEmbeddingRequest{Model: r.Model, Input: input[0], Options: opts, Dimensions: dimensions}
}
return &OllamaEmbeddingRequest{Model: r.Model, Input: input, Options: opts, Dimensions: r.Dimensions}
return &OllamaEmbeddingRequest{Model: r.Model, Input: input, Options: opts, Dimensions: dimensions}
}
func ollamaEmbeddingHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {
relay/channel/openai/adaptor.go
+5 -4
View File
@@ -29,6 +29,7 @@ import (
"github.com/QuantumNous/new-api/service"
"github.com/QuantumNous/new-api/setting/model_setting"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/gin-gonic/gin"
)
@@ -314,9 +315,9 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
}
if strings.HasPrefix(info.UpstreamModelName, "o") || strings.HasPrefix(info.UpstreamModelName, "gpt-5") {
if request.MaxCompletionTokens == 0 && request.MaxTokens != 0 {
if lo.FromPtrOr(request.MaxCompletionTokens, uint(0)) == 0 && lo.FromPtrOr(request.MaxTokens, uint(0)) != 0 {
request.MaxCompletionTokens = request.MaxTokens
request.MaxTokens = 0
request.MaxTokens = nil
}
if strings.HasPrefix(info.UpstreamModelName, "o") {
@@ -326,8 +327,8 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
// gpt-5系列模型适配 归零不再支持的参数
if strings.HasPrefix(info.UpstreamModelName, "gpt-5") {
request.Temperature = nil
request.TopP = 0 // oai 的 top_p 默认值是 1.0，但是为了 omitempty 属性直接不传，这里显式设置为 0
request.LogProbs = false
request.TopP = nil
request.LogProbs = nil
}
// 转换模型推理力度后缀
relay/channel/perplexity/adaptor.go
+3 -2
View File
@@ -12,6 +12,7 @@ import (
relaycommon "github.com/QuantumNous/new-api/relay/common"
relayconstant "github.com/QuantumNous/new-api/relay/constant"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/gin-gonic/gin"
)
@@ -59,8 +60,8 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
if request == nil {
return nil, errors.New("request is nil")
}
if request.TopP >= 1 {
request.TopP = 0.99
if lo.FromPtrOr(request.TopP, 0) >= 1 {
request.TopP = lo.ToPtr(0.99)
}
return requestOpenAI2Perplexity(*request), nil
}
relay/channel/perplexity/relay-perplexity.go
+6 -2
View File
@@ -10,13 +10,12 @@ func requestOpenAI2Perplexity(request dto.GeneralOpenAIRequest) *dto.GeneralOpen
Content: message.Content,
})
}
return &dto.GeneralOpenAIRequest{
req := &dto.GeneralOpenAIRequest{
Model: request.Model,
Stream: request.Stream,
Messages: messages,
Temperature: request.Temperature,
TopP: request.TopP,
MaxTokens: request.GetMaxTokens(),
FrequencyPenalty: request.FrequencyPenalty,
PresencePenalty: request.PresencePenalty,
SearchDomainFilter: request.SearchDomainFilter,
@@ -25,4 +24,9 @@ func requestOpenAI2Perplexity(request dto.GeneralOpenAIRequest) *dto.GeneralOpen
ReturnRelatedQuestions: request.ReturnRelatedQuestions,
SearchMode: request.SearchMode,
}
if request.MaxTokens != nil || request.MaxCompletionTokens != nil {
maxTokens := request.GetMaxTokens()
req.MaxTokens = &maxTokens
}
return req
}
relay/channel/replicate/adaptor.go
+3 -2
View File
@@ -22,6 +22,7 @@ import (
"github.com/QuantumNous/new-api/types"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
type Adaptor struct {
@@ -115,8 +116,8 @@ func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInf
}
}
if request.N > 0 {
inputPayload["num_outputs"] = int(request.N)
if imageN := lo.FromPtrOr(request.N, uint(0)); imageN > 0 {
inputPayload["num_outputs"] = int(imageN)
}
if strings.EqualFold(request.Quality, "hd") || strings.EqualFold(request.Quality, "high") {
relay/channel/siliconflow/adaptor.go
+4 -1
View File
@@ -15,6 +15,7 @@ import (
"github.com/QuantumNous/new-api/types"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
type Adaptor struct {
@@ -53,7 +54,9 @@ func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInf
sfRequest.ImageSize = request.Size
}
if sfRequest.BatchSize == 0 {
sfRequest.BatchSize = request.N
if request.N != nil {
sfRequest.BatchSize = lo.FromPtr(request.N)
}
}
return sfRequest, nil
relay/channel/tencent/relay-tencent.go
+3 -3
View File
@@ -37,12 +37,12 @@ func requestOpenAI2Tencent(a *Adaptor, request dto.GeneralOpenAIRequest) *Tencen
})
}
var req = TencentChatRequest{
Stream: &request.Stream,
Stream: request.Stream,
Messages: messages,
Model: &request.Model,
}
if request.TopP != 0 {
req.TopP = &request.TopP
if request.TopP != nil {
req.TopP = request.TopP
}
req.Temperature = request.Temperature
return &req
relay/channel/vertex/adaptor.go
+5 -4
View File
@@ -21,6 +21,7 @@ import (
"github.com/QuantumNous/new-api/types"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
const (
@@ -292,11 +293,11 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
imgReq := dto.ImageRequest{
Model: request.Model,
Prompt: prompt,
N: 1,
N: lo.ToPtr(uint(1)),
Size: "1024x1024",
}
if request.N > 0 {
imgReq.N = uint(request.N)
if request.N != nil && *request.N > 0 {
imgReq.N = lo.ToPtr(uint(*request.N))
}
if request.Size != "" {
imgReq.Size = request.Size
@@ -305,7 +306,7 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
var extra map[string]any
if err := json.Unmarshal(request.ExtraBody, &extra); err == nil {
if n, ok := extra["n"].(float64); ok && n > 0 {
imgReq.N = uint(n)
imgReq.N = lo.ToPtr(uint(n))
}
if size, ok := extra["size"].(string); ok {
imgReq.Size = size
relay/channel/vertex/dto.go
+4 -4
View File
@@ -10,12 +10,12 @@ type VertexAIClaudeRequest struct {
AnthropicVersion string `json:"anthropic_version"`
Messages []dto.ClaudeMessage `json:"messages"`
System any `json:"system,omitempty"`
MaxTokens uint `json:"max_tokens,omitempty"`
MaxTokens *uint `json:"max_tokens,omitempty"`
StopSequences []string `json:"stop_sequences,omitempty"`
Stream bool `json:"stream,omitempty"`
Stream *bool `json:"stream,omitempty"`
Temperature *float64 `json:"temperature,omitempty"`
TopP float64 `json:"top_p,omitempty"`
TopK int `json:"top_k,omitempty"`
TopP *float64 `json:"top_p,omitempty"`
TopK *int `json:"top_k,omitempty"`
Tools any `json:"tools,omitempty"`
ToolChoice any `json:"tool_choice,omitempty"`
Thinking *dto.Thinking `json:"thinking,omitempty"`
relay/channel/volcengine/adaptor.go
+2 -1
View File
@@ -21,6 +21,7 @@ import (
"github.com/QuantumNous/new-api/types"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
const (
@@ -56,7 +57,7 @@ func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInf
}
voiceType := mapVoiceType(request.Voice)
speedRatio := request.Speed
speedRatio := lo.FromPtrOr(request.Speed, 0.0)
encoding := mapEncoding(request.ResponseFormat)
c.Set(contextKeyResponseFormat, encoding)
relay/channel/xai/adaptor.go
+4 -3
View File
@@ -15,6 +15,7 @@ import (
"github.com/QuantumNous/new-api/relay/constant"
"github.com/gin-gonic/gin"
"github.com/samber/lo"
)
type Adaptor struct {
@@ -40,7 +41,7 @@ func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInf
xaiRequest := ImageRequest{
Model: request.Model,
Prompt: request.Prompt,
N: int(request.N),
N: int(lo.FromPtrOr(request.N, uint(1))),
ResponseFormat: request.ResponseFormat,
}
return xaiRequest, nil
@@ -73,9 +74,9 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
return toMap, nil
}
if strings.HasPrefix(request.Model, "grok-3-mini") {
if request.MaxCompletionTokens == 0 && request.MaxTokens != 0 {
if lo.FromPtrOr(request.MaxCompletionTokens, uint(0)) == 0 && lo.FromPtrOr(request.MaxTokens, uint(0)) != 0 {
request.MaxCompletionTokens = request.MaxTokens
request.MaxTokens = 0
request.MaxTokens = lo.ToPtr(uint(0))
}
if strings.HasSuffix(request.Model, "-high") {
request.ReasoningEffort = "high"
relay/channel/xunfei/relay-xunfei.go
+2 -1
View File
@@ -16,6 +16,7 @@ import (
"github.com/QuantumNous/new-api/dto"
"github.com/QuantumNous/new-api/relay/helper"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/gin-gonic/gin"
"github.com/gorilla/websocket"
@@ -48,7 +49,7 @@ func requestOpenAI2Xunfei(request dto.GeneralOpenAIRequest, xunfeiAppId string,
xunfeiRequest.Header.AppId = xunfeiAppId
xunfeiRequest.Parameter.Chat.Domain = domain
xunfeiRequest.Parameter.Chat.Temperature = request.Temperature
xunfeiRequest.Parameter.Chat.TopK = request.N
xunfeiRequest.Parameter.Chat.TopK = lo.FromPtrOr(request.N, 0)
xunfeiRequest.Parameter.Chat.MaxTokens = request.GetMaxTokens()
xunfeiRequest.Payload.Message.Text = messages
return &xunfeiRequest
relay/channel/zhipu/adaptor.go
+3 -2
View File
@@ -10,6 +10,7 @@ import (
"github.com/QuantumNous/new-api/relay/channel"
relaycommon "github.com/QuantumNous/new-api/relay/common"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/gin-gonic/gin"
)
@@ -60,8 +61,8 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
if request == nil {
return nil, errors.New("request is nil")
}
if request.TopP >= 1 {
request.TopP = 0.99
if lo.FromPtrOr(request.TopP, 0) >= 1 {
request.TopP = lo.ToPtr(0.99)
}
return requestOpenAI2Zhipu(*request), nil
}
relay/channel/zhipu/relay-zhipu.go
+2 -1
View File
@@ -16,6 +16,7 @@ import (
"github.com/QuantumNous/new-api/relay/helper"
"github.com/QuantumNous/new-api/service"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/gin-gonic/gin"
"github.com/golang-jwt/jwt/v5"
@@ -98,7 +99,7 @@ func requestOpenAI2Zhipu(request dto.GeneralOpenAIRequest) *ZhipuRequest {
return &ZhipuRequest{
Prompt: messages,
Temperature: request.Temperature,
TopP: request.TopP,
TopP: lo.FromPtrOr(request.TopP, 0),
Incremental: false,
}
}
relay/channel/zhipu_4v/adaptor.go
+3 -2
View File
@@ -14,6 +14,7 @@ import (
relaycommon "github.com/QuantumNous/new-api/relay/common"
relayconstant "github.com/QuantumNous/new-api/relay/constant"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/gin-gonic/gin"
)
@@ -83,8 +84,8 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
if request == nil {
return nil, errors.New("request is nil")
}
if request.TopP >= 1 {
request.TopP = 0.99
if lo.FromPtrOr(request.TopP, 0) >= 1 {
request.TopP = lo.ToPtr(0.99)
}
return requestOpenAI2Zhipu(*request), nil
}
relay/channel/zhipu_4v/relay-zhipu_v4.go
+6 -2
View File
@@ -41,16 +41,20 @@ func requestOpenAI2Zhipu(request dto.GeneralOpenAIRequest) *dto.GeneralOpenAIReq
} else {
Stop, _ = request.Stop.([]string)
}
return &dto.GeneralOpenAIRequest{
out := &dto.GeneralOpenAIRequest{
Model: request.Model,
Stream: request.Stream,
Messages: messages,
Temperature: request.Temperature,
TopP: request.TopP,
MaxTokens: request.GetMaxTokens(),
Stop: Stop,
Tools: request.Tools,
ToolChoice: request.ToolChoice,
THINKING: request.THINKING,
}
if request.MaxTokens != nil || request.MaxCompletionTokens != nil {
maxTokens := request.GetMaxTokens()
out.MaxTokens = &maxTokens
}
return out
}
relay/claude_handler.go
+8 -7
View File
@@ -47,8 +47,9 @@ func ClaudeHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ
}
adaptor.Init(info)
if request.MaxTokens == 0 {
request.MaxTokens = uint(model_setting.GetClaudeSettings().GetDefaultMaxTokens(request.Model))
if request.MaxTokens == nil || *request.MaxTokens == 0 {
defaultMaxTokens := uint(model_setting.GetClaudeSettings().GetDefaultMaxTokens(request.Model))
request.MaxTokens = &defaultMaxTokens
}
if baseModel, effortLevel, ok := reasoning.TrimEffortSuffix(request.Model); ok && effortLevel != "" &&
@@ -58,25 +59,25 @@ func ClaudeHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ
Type: "adaptive",
}
request.OutputConfig = json.RawMessage(fmt.Sprintf(`{"effort":"%s"}`, effortLevel))
request.TopP = 0
request.TopP = common.GetPointer[float64](0)
request.Temperature = common.GetPointer[float64](1.0)
info.UpstreamModelName = request.Model
} else if model_setting.GetClaudeSettings().ThinkingAdapterEnabled &&
strings.HasSuffix(request.Model, "-thinking") {
if request.Thinking == nil {
// 因为BudgetTokens 必须大于1024
if request.MaxTokens < 1280 {
request.MaxTokens = 1280
if request.MaxTokens == nil || *request.MaxTokens < 1280 {
request.MaxTokens = common.GetPointer[uint](1280)
}
// BudgetTokens 为 max_tokens 的 80%
request.Thinking = &dto.Thinking{
Type: "enabled",
BudgetTokens: common.GetPointer[int](int(float64(request.MaxTokens) * model_setting.GetClaudeSettings().ThinkingAdapterBudgetTokensPercentage)),
BudgetTokens: common.GetPointer[int](int(float64(*request.MaxTokens) * model_setting.GetClaudeSettings().ThinkingAdapterBudgetTokensPercentage)),
}
// TODO: 临时处理
// https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking#important-considerations-when-using-extended-thinking
request.TopP = 0
request.TopP = common.GetPointer[float64](0)
request.Temperature = common.GetPointer[float64](1.0)
}
if !model_setting.ShouldPreserveThinkingSuffix(info.OriginModelName) {
relay/compatible_handler.go
+2 -1
View File
@@ -21,6 +21,7 @@ import (
"github.com/QuantumNous/new-api/setting/operation_setting"
"github.com/QuantumNous/new-api/setting/ratio_setting"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/shopspring/decimal"
@@ -56,7 +57,7 @@ func TextHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *types
}
// 如果不支持StreamOptions，将StreamOptions设置为nil
if !info.SupportStreamOptions || !request.Stream {
if !info.SupportStreamOptions || !lo.FromPtrOr(request.Stream, false) {
request.StreamOptions = nil
} else {
// 如果支持StreamOptions，且请求中没有设置StreamOptions，根据配置文件设置StreamOptions
relay/helper/valid_request.go
+7 -6
View File
@@ -12,6 +12,7 @@ import (
"github.com/QuantumNous/new-api/logger"
relayconstant "github.com/QuantumNous/new-api/relay/constant"
"github.com/QuantumNous/new-api/types"
"github.com/samber/lo"
"github.com/gin-gonic/gin"
)
@@ -151,7 +152,7 @@ func GetAndValidOpenAIImageRequest(c *gin.Context, relayMode int) (*dto.ImageReq
formData := c.Request.PostForm
imageRequest.Prompt = formData.Get("prompt")
imageRequest.Model = formData.Get("model")
imageRequest.N = uint(common.String2Int(formData.Get("n")))
imageRequest.N = common.GetPointer(uint(common.String2Int(formData.Get("n"))))
imageRequest.Quality = formData.Get("quality")
imageRequest.Size = formData.Get("size")
if imageValue := formData.Get("image"); imageValue != "" {
@@ -163,8 +164,8 @@ func GetAndValidOpenAIImageRequest(c *gin.Context, relayMode int) (*dto.ImageReq
imageRequest.Quality = "standard"
}
}
if imageRequest.N == 0 {
imageRequest.N = 1
if imageRequest.N == nil || *imageRequest.N == 0 {
imageRequest.N = common.GetPointer(uint(1))
}
hasWatermark := formData.Has("watermark")
@@ -218,8 +219,8 @@ func GetAndValidOpenAIImageRequest(c *gin.Context, relayMode int) (*dto.ImageReq
// return nil, errors.New("prompt is required")
//}
if imageRequest.N == 0 {
imageRequest.N = 1
if imageRequest.N == nil || *imageRequest.N == 0 {
imageRequest.N = common.GetPointer(uint(1))
}
}
@@ -260,7 +261,7 @@ func GetAndValidateTextRequest(c *gin.Context, relayMode int) (*dto.GeneralOpenA
textRequest.Model = c.Param("model")
}
if textRequest.MaxTokens > math.MaxInt32/2 {
if lo.FromPtrOr(textRequest.MaxTokens, uint(0)) > math.MaxInt32/2 {
return nil, errors.New("max_tokens is invalid")
}
if textRequest.Model == "" {
relay/image_handler.go
+8 -4
View File
@@ -113,11 +113,15 @@ func ImageHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *type
return newAPIError
}
imageN := uint(1)
if request.N != nil {
imageN = *request.N
}
if usage.(*dto.Usage).TotalTokens == 0 {
usage.(*dto.Usage).TotalTokens = int(request.N)
usage.(*dto.Usage).TotalTokens = int(imageN)
}
if usage.(*dto.Usage).PromptTokens == 0 {
usage.(*dto.Usage).PromptTokens = int(request.N)
usage.(*dto.Usage).PromptTokens = int(imageN)
}
quality := "standard"
@@ -133,8 +137,8 @@ func ImageHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *type
if len(quality) > 0 {
logContent = append(logContent, fmt.Sprintf("品质 %s", quality))
}
if request.N > 0 {
logContent = append(logContent, fmt.Sprintf("生成数量 %d", request.N))
if imageN > 0 {
logContent = append(logContent, fmt.Sprintf("生成数量 %d", imageN))
}
postConsumeQuota(c, info, usage.(*dto.Usage), logContent...)