Merge pull request #551 from SilentFlower/opus4.6-think

[UPDATE] 增强 Claude Thinking 模式支持与 Opus 4.6 动态预算适配

backend/internal/pkg/antigravity/claude_types.go

@@ -27,7 +27,7 @@ type ClaudeMessage struct {
 
 // ThinkingConfig Thinking 配置
 type ThinkingConfig struct {
-	Type         string `json:"type"`                    // "enabled" or "disabled"
+	Type         string `json:"type"`                    // "enabled" / "adaptive" / "disabled"
 	BudgetTokens int    `json:"budget_tokens,omitempty"` // thinking budget
 }
 

backend/internal/pkg/antigravity/request_transformer.go

@@ -64,6 +64,10 @@ const MaxTokensBudgetPadding = 1000
 // Gemini 2.5 Flash thinking budget 上限
 const Gemini25FlashThinkingBudgetLimit = 24576
 
+// 对于 Antigravity 的 Claude（budget-only）模型，该语义最终等价为 thinkingBudget=24576。
+// 这里复用相同数值以保持行为一致。
+const ClaudeAdaptiveHighThinkingBudgetTokens = Gemini25FlashThinkingBudgetLimit
+
 // ensureMaxTokensGreaterThanBudget 确保 max_tokens > budget_tokens
 // Claude API 要求启用 thinking 时，max_tokens 必须大于 thinking.budget_tokens
 // 返回调整后的 maxTokens 和是否进行了调整
@@ -96,7 +100,7 @@ func TransformClaudeToGeminiWithOptions(claudeReq *ClaudeRequest, projectID, map
 	}
 
 	// 检测是否启用 thinking
-	isThinkingEnabled := claudeReq.Thinking != nil && claudeReq.Thinking.Type == "enabled"
+	isThinkingEnabled := claudeReq.Thinking != nil && (claudeReq.Thinking.Type == "enabled" || claudeReq.Thinking.Type == "adaptive")
 
 	// 只有 Gemini 模型支持 dummy thought workaround
 	// Claude 模型通过 Vertex/Google API 需要有效的 thought signatures
@@ -198,8 +202,7 @@ type modelInfo struct {
 
 // modelInfoMap 模型前缀 → 模型信息映射
 // 只有在此映射表中的模型才会注入身份提示词
-// 注意：当前 claude-opus-4-6 会被映射到 claude-opus-4-5-thinking，
-// 但保留此条目以便后续 Antigravity 上游支持 4.6 时快速切换
+// 注意：模型映射逻辑在网关层完成；这里仅用于按模型前缀判断是否注入身份提示词。
 var modelInfoMap = map[string]modelInfo{
 	"claude-opus-4-5":   {DisplayName: "Claude Opus 4.5", CanonicalID: "claude-opus-4-5-20250929"},
 	"claude-opus-4-6":   {DisplayName: "Claude Opus 4.6", CanonicalID: "claude-opus-4-6"},
@@ -593,6 +596,10 @@ func maxOutputTokensLimit(model string) int {
 	return maxOutputTokensUpperBound
 }
 
+func isAntigravityOpus46Model(model string) bool {
+	return strings.HasPrefix(strings.ToLower(model), "claude-opus-4-6")
+}
+
 func buildGenerationConfig(req *ClaudeRequest) *GeminiGenerationConfig {
 	maxLimit := maxOutputTokensLimit(req.Model)
 	config := &GeminiGenerationConfig{
@@ -606,25 +613,36 @@ func buildGenerationConfig(req *ClaudeRequest) *GeminiGenerationConfig {
 	}
 
 	// Thinking 配置
-	if req.Thinking != nil && req.Thinking.Type == "enabled" {
+	if req.Thinking != nil && (req.Thinking.Type == "enabled" || req.Thinking.Type == "adaptive") {
 		config.ThinkingConfig = &GeminiThinkingConfig{
 			IncludeThoughts: true,
 		}
+
+		// - thinking.type=enabled：budget_tokens>0 用显式预算
+		// - thinking.type=adaptive：仅在 Antigravity 的 Opus 4.6 上覆写为 （24576）
+		budget := -1
 		if req.Thinking.BudgetTokens > 0 {
-			budget := req.Thinking.BudgetTokens
+			budget = req.Thinking.BudgetTokens
+		}
+		if req.Thinking.Type == "adaptive" && isAntigravityOpus46Model(req.Model) {
+			budget = ClaudeAdaptiveHighThinkingBudgetTokens
+		}
+
+		// 正预算需要做上限与 max_tokens 约束；动态预算（-1）直接透传给上游。
+		if budget > 0 {
 			// gemini-2.5-flash 上限
 			if strings.Contains(req.Model, "gemini-2.5-flash") && budget > Gemini25FlashThinkingBudgetLimit {
 				budget = Gemini25FlashThinkingBudgetLimit
 			}
-			config.ThinkingConfig.ThinkingBudget = budget
 
-			// 自动修正：max_tokens 必须大于 budget_tokens
+			// 自动修正：max_tokens 必须大于 budget_tokens（Claude 上游要求）
 			if adjusted, ok := ensureMaxTokensGreaterThanBudget(config.MaxOutputTokens, budget); ok {
 				log.Printf("[Antigravity] Auto-adjusted max_tokens from %d to %d (must be > budget_tokens=%d)",
 					config.MaxOutputTokens, adjusted, budget)
 				config.MaxOutputTokens = adjusted
 			}
 		}
+		config.ThinkingConfig.ThinkingBudget = budget
 	}
 
 	if config.MaxOutputTokens > maxLimit {

backend/internal/pkg/antigravity/request_transformer_test.go

@@ -259,3 +259,93 @@ func TestBuildTools_CustomTypeTools(t *testing.T) {
 		})
 	}
 }
+
+func TestBuildGenerationConfig_ThinkingDynamicBudget(t *testing.T) {
+	tests := []struct {
+		name        string
+		model       string
+		thinking    *ThinkingConfig
+		wantBudget  int
+		wantPresent bool
+	}{
+		{
+			name:        "enabled without budget defaults to dynamic (-1)",
+			model:       "claude-opus-4-6-thinking",
+			thinking:    &ThinkingConfig{Type: "enabled"},
+			wantBudget:  -1,
+			wantPresent: true,
+		},
+		{
+			name:        "enabled with budget uses the provided value",
+			model:       "claude-opus-4-6-thinking",
+			thinking:    &ThinkingConfig{Type: "enabled", BudgetTokens: 1024},
+			wantBudget:  1024,
+			wantPresent: true,
+		},
+		{
+			name:        "enabled with -1 budget uses dynamic (-1)",
+			model:       "claude-opus-4-6-thinking",
+			thinking:    &ThinkingConfig{Type: "enabled", BudgetTokens: -1},
+			wantBudget:  -1,
+			wantPresent: true,
+		},
+		{
+			name:        "adaptive on opus4.6 maps to high budget (24576)",
+			model:       "claude-opus-4-6-thinking",
+			thinking:    &ThinkingConfig{Type: "adaptive", BudgetTokens: 20000},
+			wantBudget:  ClaudeAdaptiveHighThinkingBudgetTokens,
+			wantPresent: true,
+		},
+		{
+			name:        "adaptive on non-opus model keeps default dynamic (-1)",
+			model:       "claude-sonnet-4-5-thinking",
+			thinking:    &ThinkingConfig{Type: "adaptive"},
+			wantBudget:  -1,
+			wantPresent: true,
+		},
+		{
+			name:        "disabled does not emit thinkingConfig",
+			model:       "claude-opus-4-6-thinking",
+			thinking:    &ThinkingConfig{Type: "disabled", BudgetTokens: 1024},
+			wantBudget:  0,
+			wantPresent: false,
+		},
+		{
+			name:        "nil thinking does not emit thinkingConfig",
+			model:       "claude-opus-4-6-thinking",
+			thinking:    nil,
+			wantBudget:  0,
+			wantPresent: false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			req := &ClaudeRequest{
+				Model:    tt.model,
+				Thinking: tt.thinking,
+			}
+			cfg := buildGenerationConfig(req)
+			if cfg == nil {
+				t.Fatalf("expected non-nil generationConfig")
+			}
+
+			if tt.wantPresent {
+				if cfg.ThinkingConfig == nil {
+					t.Fatalf("expected thinkingConfig to be present")
+				}
+				if !cfg.ThinkingConfig.IncludeThoughts {
+					t.Fatalf("expected includeThoughts=true")
+				}
+				if cfg.ThinkingConfig.ThinkingBudget != tt.wantBudget {
+					t.Fatalf("expected thinkingBudget=%d, got %d", tt.wantBudget, cfg.ThinkingConfig.ThinkingBudget)
+				}
+				return
+			}
+
+			if cfg.ThinkingConfig != nil {
+				t.Fatalf("expected thinkingConfig to be nil, got %+v", cfg.ThinkingConfig)
+			}
+		})
+	}
+}

backend/internal/service/antigravity_gateway_service.go

@@ -1309,7 +1309,7 @@ func (s *AntigravityGatewayService) Forward(ctx context.Context, c *gin.Context,
 		return nil, s.writeClaudeError(c, http.StatusForbidden, "permission_error", fmt.Sprintf("model %s not in whitelist", claudeReq.Model))
 	}
 	// 应用 thinking 模式自动后缀：如果 thinking 开启且目标是 claude-sonnet-4-5，自动改为 thinking 版本
-	thinkingEnabled := claudeReq.Thinking != nil && claudeReq.Thinking.Type == "enabled"
+	thinkingEnabled := claudeReq.Thinking != nil && (claudeReq.Thinking.Type == "enabled" || claudeReq.Thinking.Type == "adaptive")
 	mappedModel = applyThinkingModelSuffix(mappedModel, thinkingEnabled)
 
 	// 获取 access_token

backend/internal/service/gateway_request.go

@@ -101,9 +101,9 @@ func ParseGatewayRequest(body []byte, protocol string) (*ParsedRequest, error) {
 		}
 	}
 
-	// thinking: {type: "enabled"}
+	// thinking: {type: "enabled" | "adaptive"}
 	if rawThinking, ok := req["thinking"].(map[string]any); ok {
-		if t, ok := rawThinking["type"].(string); ok && t == "enabled" {
+		if t, ok := rawThinking["type"].(string); ok && (t == "enabled" || t == "adaptive") {
 			parsed.ThinkingEnabled = true
 		}
 	}
@@ -161,9 +161,9 @@ func parseIntegralNumber(raw any) (int, bool) {
 // Returns filtered body or original body if filtering fails (fail-safe)
 // This prevents 400 errors from invalid thinking block signatures
 //
-// Strategy:
-//   - When thinking.type != "enabled": Remove all thinking blocks
-//   - When thinking.type == "enabled": Only remove thinking blocks without valid signatures
+// 策略：
+//   - 当 thinking.type 不是 "enabled"/"adaptive"：移除所有 thinking 相关块
+//   - 当 thinking.type 是 "enabled"/"adaptive"：仅移除缺失/无效 signature 的 thinking 块（避免 400）
 //     (blocks with missing/empty/dummy signatures that would cause 400 errors)
 func FilterThinkingBlocks(body []byte) []byte {
 	return filterThinkingBlocksInternal(body, false)
@@ -489,9 +489,9 @@ func FilterSignatureSensitiveBlocksForRetry(body []byte) []byte {
 }
 
 // filterThinkingBlocksInternal removes invalid thinking blocks from request
-// Strategy:
-//   - When thinking.type != "enabled": Remove all thinking blocks
-//   - When thinking.type == "enabled": Only remove thinking blocks without valid signatures
+// 策略：
+//   - 当 thinking.type 不是 "enabled"/"adaptive"：移除所有 thinking 相关块
+//   - 当 thinking.type 是 "enabled"/"adaptive"：仅移除缺失/无效 signature 的 thinking 块
 func filterThinkingBlocksInternal(body []byte, _ bool) []byte {
 	// Fast path: if body doesn't contain "thinking", skip parsing
 	if !bytes.Contains(body, []byte(`"type":"thinking"`)) &&
@@ -511,7 +511,7 @@ func filterThinkingBlocksInternal(body []byte, _ bool) []byte {
 	// Check if thinking is enabled
 	thinkingEnabled := false
 	if thinking, ok := req["thinking"].(map[string]any); ok {
-		if thinkType, ok := thinking["type"].(string); ok && thinkType == "enabled" {
+		if thinkType, ok := thinking["type"].(string); ok && (thinkType == "enabled" || thinkType == "adaptive") {
 			thinkingEnabled = true
 		}
 	}

backend/internal/service/gateway_request_test.go

@@ -29,6 +29,14 @@ func TestParseGatewayRequest_ThinkingEnabled(t *testing.T) {
 	require.True(t, parsed.ThinkingEnabled)
 }
 
+func TestParseGatewayRequest_ThinkingAdaptiveEnabled(t *testing.T) {
+	body := []byte(`{"model":"claude-sonnet-4-5","thinking":{"type":"adaptive"},"messages":[{"content":"hi"}]}`)
+	parsed, err := ParseGatewayRequest(body, "")
+	require.NoError(t, err)
+	require.Equal(t, "claude-sonnet-4-5", parsed.Model)
+	require.True(t, parsed.ThinkingEnabled)
+}
+
 func TestParseGatewayRequest_MaxTokens(t *testing.T) {
 	body := []byte(`{"model":"claude-haiku-4-5","max_tokens":1}`)
 	parsed, err := ParseGatewayRequest(body, "")
@@ -209,6 +217,16 @@ func TestFilterThinkingBlocks(t *testing.T) {
 			input:        `{"model":"claude-3-5-sonnet-20241022","messages":[{"role":"user","content":[{"type":"text","text":"Hello"},{"type":"thinking","thinking":"internal","signature":"invalid"},{"type":"text","text":"World"}]}]}`,
 			shouldFilter: true,
 		},
+		{
+			name:         "does not filter signed thinking blocks when thinking adaptive",
+			input:        `{"thinking":{"type":"adaptive"},"messages":[{"role":"assistant","content":[{"type":"thinking","thinking":"ok","signature":"sig_real_123"},{"type":"text","text":"B"}]}]}`,
+			shouldFilter: false,
+		},
+		{
+			name:         "filters unsigned thinking blocks when thinking adaptive",
+			input:        `{"thinking":{"type":"adaptive"},"messages":[{"role":"assistant","content":[{"type":"thinking","thinking":"internal","signature":""},{"type":"text","text":"B"}]}]}`,
+			shouldFilter: true,
+		},
 		{
 			name:         "handles no thinking blocks",
 			input:        `{"model":"claude-3-5-sonnet-20241022","messages":[{"role":"user","content":[{"type":"text","text":"Hello"}]}]}`,

backend/internal/service/gateway_service.go

@@ -3631,7 +3631,8 @@ func requestNeedsBetaFeatures(body []byte) bool {
 	if tools.Exists() && tools.IsArray() && len(tools.Array()) > 0 {
 		return true
 	}
-	if strings.EqualFold(gjson.GetBytes(body, "thinking.type").String(), "enabled") {
+	thinkingType := gjson.GetBytes(body, "thinking.type").String()
+	if strings.EqualFold(thinkingType, "enabled") || strings.EqualFold(thinkingType, "adaptive") {
 		return true
 	}
 	return false