github/new-api
1
0
Fork 0
mirror of https://github.com/QuantumNous/new-api.git synced 2026-04-19 14:38:38 +00:00
Code Issues 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 21cfc1ca38
commit 0e9198e9b5
52 changed files with 523 additions and 232 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
relay/channel/claude/relay-claude.go
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")