Merge pull request #2344 from seefs001/feature/gemini-thinking-level

feat: gemini 3 thinking level gemini-3-pro-preview-high

relay/channel/gemini/adaptor.go

@@ -137,6 +137,8 @@ func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
 			info.UpstreamModelName = strings.TrimSuffix(info.UpstreamModelName, "-thinking")
 		} else if strings.HasSuffix(info.UpstreamModelName, "-nothinking") {
 			info.UpstreamModelName = strings.TrimSuffix(info.UpstreamModelName, "-nothinking")
+		} else if baseModel, level := parseThinkingLevelSuffix(info.UpstreamModelName); level != "" {
+			info.UpstreamModelName = baseModel
 		}
 	}
 

relay/channel/gemini/relay-gemini.go

@@ -19,8 +19,8 @@ import (
 	"github.com/QuantumNous/new-api/relay/helper"
 	"github.com/QuantumNous/new-api/service"
 	"github.com/QuantumNous/new-api/setting/model_setting"
+	"github.com/QuantumNous/new-api/setting/reasoning"
 	"github.com/QuantumNous/new-api/types"
-
 	"github.com/gin-gonic/gin"
 )
 
@@ -122,6 +122,14 @@ func clampThinkingBudgetByEffort(modelName string, effort string) int {
 	return clampThinkingBudget(modelName, maxBudget)
 }
 
+func parseThinkingLevelSuffix(modelName string) (string, string) {
+	base, level, ok := reasoning.TrimEffortSuffix(modelName)
+	if !ok {
+		return modelName, ""
+	}
+	return base, level
+}
+
 func ThinkingAdaptor(geminiRequest *dto.GeminiChatRequest, info *relaycommon.RelayInfo, oaiRequest ...dto.GeneralOpenAIRequest) {
 	if model_setting.GetGeminiSettings().ThinkingAdapterEnabled {
 		modelName := info.UpstreamModelName
@@ -178,6 +186,12 @@ func ThinkingAdaptor(geminiRequest *dto.GeminiChatRequest, info *relaycommon.Rel
 					ThinkingBudget: common.GetPointer(0),
 				}
 			}
+		} else if _, level := parseThinkingLevelSuffix(modelName); level != "" {
+			geminiRequest.GenerationConfig.ThinkingConfig = &dto.GeminiThinkingConfig{
+				IncludeThoughts: true,
+				ThinkingLevel:   level,
+			}
+			info.ReasoningEffort = level
 		}
 	}
 }