Merge branch 'router-for-me:main' into main

internal/runtime/executor/aistudio_executor.go

@@ -309,11 +309,9 @@ func (e *AIStudioExecutor) translateRequest(req cliproxyexecutor.Request, opts c
 	to := sdktranslator.FromString("gemini")
 	payload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream)
 	payload = applyThinkingMetadata(payload, req.Metadata, req.Model)
+	payload = util.ApplyDefaultThinkingIfNeeded(req.Model, payload)
 	payload = util.ConvertThinkingLevelToBudget(payload)
-	if budget := gjson.GetBytes(payload, "generationConfig.thinkingConfig.thinkingBudget"); budget.Exists() {
-		normalized := util.NormalizeThinkingBudget(req.Model, int(budget.Int()))
-		payload, _ = sjson.SetBytes(payload, "generationConfig.thinkingConfig.thinkingBudget", normalized)
-	}
+	payload = util.NormalizeGeminiThinkingBudget(req.Model, payload)
 	payload = util.StripThinkingConfigIfUnsupported(req.Model, payload)
 	payload = fixGeminiImageAspectRatio(req.Model, payload)
 	payload = applyPayloadConfig(e.cfg, req.Model, payload)

internal/runtime/executor/antigravity_executor.go

@@ -81,6 +81,7 @@ func (e *AntigravityExecutor) Execute(ctx context.Context, auth *cliproxyauth.Au
 	translated := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 
 	translated = applyThinkingMetadataCLI(translated, req.Metadata, req.Model)
+	translated = util.ApplyDefaultThinkingIfNeededCLI(req.Model, translated)
 	translated = normalizeAntigravityThinking(req.Model, translated)
 
 	baseURLs := antigravityBaseURLFallbackOrder(auth)
@@ -175,6 +176,7 @@ func (e *AntigravityExecutor) ExecuteStream(ctx context.Context, auth *cliproxya
 	translated := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 
 	translated = applyThinkingMetadataCLI(translated, req.Metadata, req.Model)
+	translated = util.ApplyDefaultThinkingIfNeededCLI(req.Model, translated)
 	translated = normalizeAntigravityThinking(req.Model, translated)
 
 	baseURLs := antigravityBaseURLFallbackOrder(auth)
@@ -377,9 +379,14 @@ func FetchAntigravityModels(ctx context.Context, auth *cliproxyauth.Auth, cfg *c
 		for originalName := range result.Map() {
 			aliasName := modelName2Alias(originalName)
 			if aliasName != "" {
+				cfg := modelConfig[aliasName]
+				modelName := aliasName
+				if cfg != nil && cfg.Name != "" {
+					modelName = cfg.Name
+				}
 				modelInfo := &registry.ModelInfo{
 					ID:          aliasName,
-					Name:        aliasName,
+					Name:        modelName,
 					Description: aliasName,
 					DisplayName: aliasName,
 					Version:     aliasName,
@@ -389,7 +396,7 @@ func FetchAntigravityModels(ctx context.Context, auth *cliproxyauth.Auth, cfg *c
 					Type:        antigravityAuthType,
 				}
 				// Look up Thinking support from static config using alias name
-				if cfg, ok := modelConfig[aliasName]; ok {
+				if cfg != nil {
 					if cfg.Thinking != nil {
 						modelInfo.Thinking = cfg.Thinking
 					}
@@ -842,9 +849,12 @@ func normalizeAntigravityThinking(model string, payload []byte) []byte {
 		effectiveMax, setDefaultMax := antigravityEffectiveMaxTokens(model, payload)
 		if effectiveMax > 0 && normalized >= effectiveMax {
 			normalized = effectiveMax - 1
-			if normalized < 1 {
-				normalized = 1
-			}
+		}
+		minBudget := antigravityMinThinkingBudget(model)
+		if minBudget > 0 && normalized >= 0 && normalized < minBudget {
+			// Budget is below minimum, remove thinking config entirely
+			payload, _ = sjson.DeleteBytes(payload, "request.generationConfig.thinkingConfig")
+			return payload
 		}
 		if setDefaultMax {
 			if res, errSet := sjson.SetBytes(payload, "request.generationConfig.maxOutputTokens", effectiveMax); errSet == nil {
@@ -872,3 +882,12 @@ func antigravityEffectiveMaxTokens(model string, payload []byte) (max int, fromM
 	}
 	return 0, false
 }
+
+// antigravityMinThinkingBudget returns the minimum thinking budget for a model.
+// Falls back to -1 if no model info is found.
+func antigravityMinThinkingBudget(model string) int {
+	if modelInfo := registry.GetGlobalRegistry().GetModelInfo(model); modelInfo != nil && modelInfo.Thinking != nil {
+		return modelInfo.Thinking.Min
+	}
+	return -1
+}

internal/runtime/executor/gemini_cli_executor.go

@@ -64,6 +64,8 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 	to := sdktranslator.FromString("gemini-cli")
 	basePayload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 	basePayload = applyThinkingMetadataCLI(basePayload, req.Metadata, req.Model)
+	basePayload = util.ApplyDefaultThinkingIfNeededCLI(req.Model, basePayload)
+	basePayload = util.NormalizeGeminiCLIThinkingBudget(req.Model, basePayload)
 	basePayload = util.StripThinkingConfigIfUnsupported(req.Model, basePayload)
 	basePayload = fixGeminiCLIImageAspectRatio(req.Model, basePayload)
 	basePayload = applyPayloadConfigWithRoot(e.cfg, req.Model, "gemini", "request", basePayload)
@@ -199,6 +201,8 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 	to := sdktranslator.FromString("gemini-cli")
 	basePayload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 	basePayload = applyThinkingMetadataCLI(basePayload, req.Metadata, req.Model)
+	basePayload = util.ApplyDefaultThinkingIfNeededCLI(req.Model, basePayload)
+	basePayload = util.NormalizeGeminiCLIThinkingBudget(req.Model, basePayload)
 	basePayload = util.StripThinkingConfigIfUnsupported(req.Model, basePayload)
 	basePayload = fixGeminiCLIImageAspectRatio(req.Model, basePayload)
 	basePayload = applyPayloadConfigWithRoot(e.cfg, req.Model, "gemini", "request", basePayload)

internal/runtime/executor/gemini_executor.go

@@ -80,6 +80,8 @@ func (e *GeminiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 	body = applyThinkingMetadata(body, req.Metadata, req.Model)
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
@@ -169,6 +171,8 @@ func (e *GeminiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 	body = applyThinkingMetadata(body, req.Metadata, req.Model)
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)

internal/runtime/executor/gemini_vertex_executor.go

@@ -296,6 +296,8 @@ func (e *GeminiVertexExecutor) executeWithServiceAccount(ctx context.Context, au
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
@@ -391,6 +393,8 @@ func (e *GeminiVertexExecutor) executeWithAPIKey(ctx context.Context, auth *clip
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
@@ -487,6 +491,8 @@ func (e *GeminiVertexExecutor) executeStreamWithServiceAccount(ctx context.Conte
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
@@ -599,6 +605,8 @@ func (e *GeminiVertexExecutor) executeStreamWithAPIKey(ctx context.Context, auth
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)