diff --git a/README_CN.md b/README_CN.md
index bec7fe86..9dd69226 100644
--- a/README_CN.md
+++ b/README_CN.md
@@ -404,6 +404,14 @@ gateway:
 - `server.trusted_proxies` 启用可信代理解析 X-Forwarded-For
 - `turnstile.required` 在 release 模式强制启用 Turnstile
 
+**网关防御纵深建议（重点）**
+
+- `gateway.upstream_response_read_max_bytes`：限制非流式上游响应读取大小（默认 `8MB`），用于防止异常响应导致内存放大。
+- `gateway.proxy_probe_response_read_max_bytes`：限制代理探测响应读取大小（默认 `1MB`）。
+- `gateway.gemini_debug_response_headers`：默认 `false`，仅在排障时短时开启，避免高频请求日志开销。
+- `/auth/register`、`/auth/login`、`/auth/login/2fa`、`/auth/send-verify-code` 已提供服务端兜底限流（Redis 故障时 fail-close）。
+- 推荐将 WAF/CDN 作为第一层防护，服务端限流与响应读取上限作为第二层兜底；两层同时保留，避免旁路流量与误配置风险。
+
 **⚠️ 安全警告：HTTP URL 配置**
 
 当 `security.url_allowlist.enabled=false` 时，系统默认执行最小 URL 校验，**拒绝 HTTP URL**，仅允许 HTTPS。要允许 HTTP URL（例如用于开发或内网测试），必须显式设置：
diff --git a/backend/internal/config/config.go b/backend/internal/config/config.go
index 861351de..b9f31ba9 100644
--- a/backend/internal/config/config.go
+++ b/backend/internal/config/config.go
@@ -308,6 +308,12 @@ type GatewayConfig struct {
 	ResponseHeaderTimeout int `mapstructure:"response_header_timeout"`
 	// 请求体最大字节数，用于网关请求体大小限制
 	MaxBodySize int64 `mapstructure:"max_body_size"`
+	// 非流式上游响应体读取上限（字节），用于防止无界读取导致内存放大
+	UpstreamResponseReadMaxBytes int64 `mapstructure:"upstream_response_read_max_bytes"`
+	// 代理探测响应体读取上限（字节）
+	ProxyProbeResponseReadMaxBytes int64 `mapstructure:"proxy_probe_response_read_max_bytes"`
+	// Gemini 上游响应头调试日志开关（默认关闭，避免高频日志开销）
+	GeminiDebugResponseHeaders bool `mapstructure:"gemini_debug_response_headers"`
 	// ConnectionPoolIsolation: 上游连接池隔离策略（proxy/account/account_proxy）
 	ConnectionPoolIsolation string `mapstructure:"connection_pool_isolation"`
 	// ForceCodexCLI: 强制将 OpenAI `/v1/responses` 请求按 Codex CLI 处理。
@@ -1059,6 +1065,9 @@ func setDefaults() {
 	viper.SetDefault("gateway.openai_passthrough_allow_timeout_headers", false)
 	viper.SetDefault("gateway.antigravity_fallback_cooldown_minutes", 1)
 	viper.SetDefault("gateway.max_body_size", int64(100*1024*1024))
+	viper.SetDefault("gateway.upstream_response_read_max_bytes", int64(8*1024*1024))
+	viper.SetDefault("gateway.proxy_probe_response_read_max_bytes", int64(1024*1024))
+	viper.SetDefault("gateway.gemini_debug_response_headers", false)
 	viper.SetDefault("gateway.sora_max_body_size", int64(256*1024*1024))
 	viper.SetDefault("gateway.sora_stream_timeout_seconds", 900)
 	viper.SetDefault("gateway.sora_request_timeout_seconds", 180)
@@ -1465,6 +1474,12 @@ func (c *Config) Validate() error {
 	if c.Gateway.MaxBodySize <= 0 {
 		return fmt.Errorf("gateway.max_body_size must be positive")
 	}
+	if c.Gateway.UpstreamResponseReadMaxBytes <= 0 {
+		return fmt.Errorf("gateway.upstream_response_read_max_bytes must be positive")
+	}
+	if c.Gateway.ProxyProbeResponseReadMaxBytes <= 0 {
+		return fmt.Errorf("gateway.proxy_probe_response_read_max_bytes must be positive")
+	}
 	if c.Gateway.SoraMaxBodySize < 0 {
 		return fmt.Errorf("gateway.sora_max_body_size must be non-negative")
 	}
diff --git a/backend/internal/handler/gateway_handler.go b/backend/internal/handler/gateway_handler.go
index 598eb4b3..ca5ee9d7 100644
--- a/backend/internal/handler/gateway_handler.go
+++ b/backend/internal/handler/gateway_handler.go
@@ -418,8 +418,12 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 					}
 					continue
 				}
-				// 错误响应已在Forward中处理，这里只记录日志
-				reqLog.Error("gateway.forward_failed", zap.Int64("account_id", account.ID), zap.Error(err))
+				wroteFallback := h.ensureForwardErrorResponse(c, streamStarted)
+				reqLog.Error("gateway.forward_failed",
+					zap.Int64("account_id", account.ID),
+					zap.Bool("fallback_error_response_written", wroteFallback),
+					zap.Error(err),
+				)
 				return
 			}
 
@@ -683,8 +687,12 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 					}
 					continue
 				}
-				// 错误响应已在Forward中处理，这里只记录日志
-				reqLog.Error("gateway.forward_failed", zap.Int64("account_id", account.ID), zap.Error(err))
+				wroteFallback := h.ensureForwardErrorResponse(c, streamStarted)
+				reqLog.Error("gateway.forward_failed",
+					zap.Int64("account_id", account.ID),
+					zap.Bool("fallback_error_response_written", wroteFallback),
+					zap.Error(err),
+				)
 				return
 			}
 
@@ -1117,6 +1125,15 @@ func (h *GatewayHandler) handleStreamingAwareError(c *gin.Context, status int, e
 	h.errorResponse(c, status, errType, message)
 }
 
+// ensureForwardErrorResponse 在 Forward 返回错误但尚未写响应时补写统一错误响应。
+func (h *GatewayHandler) ensureForwardErrorResponse(c *gin.Context, streamStarted bool) bool {
+	if c == nil || c.Writer == nil || c.Writer.Written() {
+		return false
+	}
+	h.handleStreamingAwareError(c, http.StatusBadGateway, "upstream_error", "Upstream request failed", streamStarted)
+	return true
+}
+
 // errorResponse 返回Claude API格式的错误响应
 func (h *GatewayHandler) errorResponse(c *gin.Context, status int, errType, message string) {
 	c.JSON(status, gin.H{
diff --git a/backend/internal/handler/gateway_handler_error_fallback_test.go b/backend/internal/handler/gateway_handler_error_fallback_test.go
new file mode 100644
index 00000000..4fce5ec1
--- /dev/null
+++ b/backend/internal/handler/gateway_handler_error_fallback_test.go
@@ -0,0 +1,49 @@
+package handler
+
+import (
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"github.com/gin-gonic/gin"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestGatewayEnsureForwardErrorResponse_WritesFallbackWhenNotWritten(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	c.Request = httptest.NewRequest(http.MethodGet, "/", nil)
+
+	h := &GatewayHandler{}
+	wrote := h.ensureForwardErrorResponse(c, false)
+
+	require.True(t, wrote)
+	require.Equal(t, http.StatusBadGateway, w.Code)
+
+	var parsed map[string]any
+	err := json.Unmarshal(w.Body.Bytes(), &parsed)
+	require.NoError(t, err)
+	assert.Equal(t, "error", parsed["type"])
+	errorObj, ok := parsed["error"].(map[string]any)
+	require.True(t, ok)
+	assert.Equal(t, "upstream_error", errorObj["type"])
+	assert.Equal(t, "Upstream request failed", errorObj["message"])
+}
+
+func TestGatewayEnsureForwardErrorResponse_DoesNotOverrideWrittenResponse(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	c.Request = httptest.NewRequest(http.MethodGet, "/", nil)
+	c.String(http.StatusTeapot, "already written")
+
+	h := &GatewayHandler{}
+	wrote := h.ensureForwardErrorResponse(c, false)
+
+	require.False(t, wrote)
+	require.Equal(t, http.StatusTeapot, w.Code)
+	assert.Equal(t, "already written", w.Body.String())
+}
diff --git a/backend/internal/handler/openai_gateway_handler.go b/backend/internal/handler/openai_gateway_handler.go
index 470eab45..f5db385b 100644
--- a/backend/internal/handler/openai_gateway_handler.go
+++ b/backend/internal/handler/openai_gateway_handler.go
@@ -365,8 +365,12 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 				)
 				continue
 			}
-			// Error response already handled in Forward, just log
-			reqLog.Error("openai.forward_failed", zap.Int64("account_id", account.ID), zap.Error(err))
+			wroteFallback := h.ensureForwardErrorResponse(c, streamStarted)
+			reqLog.Error("openai.forward_failed",
+				zap.Int64("account_id", account.ID),
+				zap.Bool("fallback_error_response_written", wroteFallback),
+				zap.Error(err),
+			)
 			return
 		}
 
@@ -521,6 +525,15 @@ func (h *OpenAIGatewayHandler) handleStreamingAwareError(c *gin.Context, status
 	h.errorResponse(c, status, errType, message)
 }
 
+// ensureForwardErrorResponse 在 Forward 返回错误但尚未写响应时补写统一错误响应。
+func (h *OpenAIGatewayHandler) ensureForwardErrorResponse(c *gin.Context, streamStarted bool) bool {
+	if c == nil || c.Writer == nil || c.Writer.Written() {
+		return false
+	}
+	h.handleStreamingAwareError(c, http.StatusBadGateway, "upstream_error", "Upstream request failed", streamStarted)
+	return true
+}
+
 // errorResponse returns OpenAI API format error response
 func (h *OpenAIGatewayHandler) errorResponse(c *gin.Context, status int, errType, message string) {
 	c.JSON(status, gin.H{
diff --git a/backend/internal/handler/openai_gateway_handler_test.go b/backend/internal/handler/openai_gateway_handler_test.go
index 65296da4..1ca52c2d 100644
--- a/backend/internal/handler/openai_gateway_handler_test.go
+++ b/backend/internal/handler/openai_gateway_handler_test.go
@@ -105,6 +105,42 @@ func TestOpenAIHandleStreamingAwareError_NonStreaming(t *testing.T) {
 	assert.Equal(t, "test error", errorObj["message"])
 }
 
+func TestOpenAIEnsureForwardErrorResponse_WritesFallbackWhenNotWritten(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	c.Request = httptest.NewRequest(http.MethodGet, "/", nil)
+
+	h := &OpenAIGatewayHandler{}
+	wrote := h.ensureForwardErrorResponse(c, false)
+
+	require.True(t, wrote)
+	require.Equal(t, http.StatusBadGateway, w.Code)
+
+	var parsed map[string]any
+	err := json.Unmarshal(w.Body.Bytes(), &parsed)
+	require.NoError(t, err)
+	errorObj, ok := parsed["error"].(map[string]any)
+	require.True(t, ok)
+	assert.Equal(t, "upstream_error", errorObj["type"])
+	assert.Equal(t, "Upstream request failed", errorObj["message"])
+}
+
+func TestOpenAIEnsureForwardErrorResponse_DoesNotOverrideWrittenResponse(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	c.Request = httptest.NewRequest(http.MethodGet, "/", nil)
+	c.String(http.StatusTeapot, "already written")
+
+	h := &OpenAIGatewayHandler{}
+	wrote := h.ensureForwardErrorResponse(c, false)
+
+	require.False(t, wrote)
+	require.Equal(t, http.StatusTeapot, w.Code)
+	assert.Equal(t, "already written", w.Body.String())
+}
+
 // TestOpenAIHandler_GjsonExtraction 验证 gjson 从请求体中提取 model/stream 的正确性
 func TestOpenAIHandler_GjsonExtraction(t *testing.T) {
 	tests := []struct {
diff --git a/backend/internal/pkg/ip/ip.go b/backend/internal/pkg/ip/ip.go
index 6ab2ff72..3f05ac41 100644
--- a/backend/internal/pkg/ip/ip.go
+++ b/backend/internal/pkg/ip/ip.go
@@ -44,6 +44,16 @@ func GetClientIP(c *gin.Context) string {
 	return normalizeIP(c.ClientIP())
 }
 
+// GetTrustedClientIP 从 Gin 的可信代理解析链提取客户端 IP。
+// 该方法依赖 gin.Engine.SetTrustedProxies 配置，不会优先直接信任原始转发头值。
+// 适用于 ACL / 风控等安全敏感场景。
+func GetTrustedClientIP(c *gin.Context) string {
+	if c == nil {
+		return ""
+	}
+	return normalizeIP(c.ClientIP())
+}
+
 // normalizeIP 规范化 IP 地址，去除端口号和空格。
 func normalizeIP(ip string) string {
 	ip = strings.TrimSpace(ip)
diff --git a/backend/internal/pkg/ip/ip_test.go b/backend/internal/pkg/ip/ip_test.go
index c3c90c74..3839403c 100644
--- a/backend/internal/pkg/ip/ip_test.go
+++ b/backend/internal/pkg/ip/ip_test.go
@@ -3,8 +3,10 @@
 package ip
 
 import (
+	"net/http/httptest"
 	"testing"
 
+	"github.com/gin-gonic/gin"
 	"github.com/stretchr/testify/require"
 )
 
@@ -49,3 +51,25 @@ func TestIsPrivateIP(t *testing.T) {
 		})
 	}
 }
+
+func TestGetTrustedClientIPUsesGinClientIP(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	r := gin.New()
+	require.NoError(t, r.SetTrustedProxies(nil))
+
+	r.GET("/t", func(c *gin.Context) {
+		c.String(200, GetTrustedClientIP(c))
+	})
+
+	w := httptest.NewRecorder()
+	req := httptest.NewRequest("GET", "/t", nil)
+	req.RemoteAddr = "9.9.9.9:12345"
+	req.Header.Set("X-Forwarded-For", "1.2.3.4")
+	req.Header.Set("X-Real-IP", "1.2.3.4")
+	req.Header.Set("CF-Connecting-IP", "1.2.3.4")
+	r.ServeHTTP(w, req)
+
+	require.Equal(t, 200, w.Code)
+	require.Equal(t, "9.9.9.9", w.Body.String())
+}
diff --git a/backend/internal/repository/proxy_probe_service.go b/backend/internal/repository/proxy_probe_service.go
index 513e929c..54de2897 100644
--- a/backend/internal/repository/proxy_probe_service.go
+++ b/backend/internal/repository/proxy_probe_service.go
@@ -19,10 +19,14 @@ func NewProxyExitInfoProber(cfg *config.Config) service.ProxyExitInfoProber {
 	insecure := false
 	allowPrivate := false
 	validateResolvedIP := true
+	maxResponseBytes := defaultProxyProbeResponseMaxBytes
 	if cfg != nil {
 		insecure = cfg.Security.ProxyProbe.InsecureSkipVerify
 		allowPrivate = cfg.Security.URLAllowlist.AllowPrivateHosts
 		validateResolvedIP = cfg.Security.URLAllowlist.Enabled
+		if cfg.Gateway.ProxyProbeResponseReadMaxBytes > 0 {
+			maxResponseBytes = cfg.Gateway.ProxyProbeResponseReadMaxBytes
+		}
 	}
 	if insecure {
 		log.Printf("[ProxyProbe] Warning: insecure_skip_verify is not allowed and will cause probe failure.")
@@ -31,11 +35,13 @@ func NewProxyExitInfoProber(cfg *config.Config) service.ProxyExitInfoProber {
 		insecureSkipVerify: insecure,
 		allowPrivateHosts:  allowPrivate,
 		validateResolvedIP: validateResolvedIP,
+		maxResponseBytes:   maxResponseBytes,
 	}
 }
 
 const (
-	defaultProxyProbeTimeout = 30 * time.Second
+	defaultProxyProbeTimeout          = 30 * time.Second
+	defaultProxyProbeResponseMaxBytes = int64(1024 * 1024)
 )
 
 // probeURLs 按优先级排列的探测 URL 列表
@@ -52,6 +58,7 @@ type proxyProbeService struct {
 	insecureSkipVerify bool
 	allowPrivateHosts  bool
 	validateResolvedIP bool
+	maxResponseBytes   int64
 }
 
 func (s *proxyProbeService) ProbeProxy(ctx context.Context, proxyURL string) (*service.ProxyExitInfo, int64, error) {
@@ -98,10 +105,17 @@ func (s *proxyProbeService) probeWithURL(ctx context.Context, client *http.Clien
 		return nil, latencyMs, fmt.Errorf("request failed with status: %d", resp.StatusCode)
 	}
 
-	body, err := io.ReadAll(resp.Body)
+	maxResponseBytes := s.maxResponseBytes
+	if maxResponseBytes <= 0 {
+		maxResponseBytes = defaultProxyProbeResponseMaxBytes
+	}
+	body, err := io.ReadAll(io.LimitReader(resp.Body, maxResponseBytes+1))
 	if err != nil {
 		return nil, latencyMs, fmt.Errorf("failed to read response: %w", err)
 	}
+	if int64(len(body)) > maxResponseBytes {
+		return nil, latencyMs, fmt.Errorf("proxy probe response exceeds limit: %d", maxResponseBytes)
+	}
 
 	switch parser {
 	case "ip-api":
diff --git a/backend/internal/server/http.go b/backend/internal/server/http.go
index d2d8ed40..a8034e98 100644
--- a/backend/internal/server/http.go
+++ b/backend/internal/server/http.go
@@ -51,6 +51,9 @@ func ProvideRouter(
 		if err := r.SetTrustedProxies(nil); err != nil {
 			log.Printf("Failed to disable trusted proxies: %v", err)
 		}
+		if cfg.Server.Mode == "release" {
+			log.Printf("Warning: server.trusted_proxies is empty in release mode; client IP trust chain is disabled")
+		}
 	}
 
 	return SetupRouter(r, handlers, jwtAuth, adminAuth, apiKeyAuth, apiKeyService, subscriptionService, opsService, settingService, cfg, redisClient)
diff --git a/backend/internal/server/middleware/api_key_auth.go b/backend/internal/server/middleware/api_key_auth.go
index 8e03f785..7aad1699 100644
--- a/backend/internal/server/middleware/api_key_auth.go
+++ b/backend/internal/server/middleware/api_key_auth.go
@@ -96,7 +96,7 @@ func apiKeyAuthWithSubscription(apiKeyService *service.APIKeyService, subscripti
 		// 检查 IP 限制（白名单/黑名单）
 		// 注意：错误信息故意模糊，避免暴露具体的 IP 限制机制
 		if len(apiKey.IPWhitelist) > 0 || len(apiKey.IPBlacklist) > 0 {
-			clientIP := ip.GetClientIP(c)
+			clientIP := ip.GetTrustedClientIP(c)
 			allowed, _ := ip.CheckIPRestriction(clientIP, apiKey.IPWhitelist, apiKey.IPBlacklist)
 			if !allowed {
 				AbortWithError(c, 403, "ACCESS_DENIED", "Access denied")
diff --git a/backend/internal/server/middleware/api_key_auth_test.go b/backend/internal/server/middleware/api_key_auth_test.go
index 3e33c7e3..f3a6f076 100644
--- a/backend/internal/server/middleware/api_key_auth_test.go
+++ b/backend/internal/server/middleware/api_key_auth_test.go
@@ -300,6 +300,57 @@ func TestAPIKeyAuthOverwritesInvalidContextGroup(t *testing.T) {
 	require.Equal(t, http.StatusOK, w.Code)
 }
 
+func TestAPIKeyAuthIPRestrictionDoesNotTrustSpoofedForwardHeaders(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	user := &service.User{
+		ID:          7,
+		Role:        service.RoleUser,
+		Status:      service.StatusActive,
+		Balance:     10,
+		Concurrency: 3,
+	}
+	apiKey := &service.APIKey{
+		ID:          100,
+		UserID:      user.ID,
+		Key:         "test-key",
+		Status:      service.StatusActive,
+		User:        user,
+		IPWhitelist: []string{"1.2.3.4"},
+	}
+
+	apiKeyRepo := &stubApiKeyRepo{
+		getByKey: func(ctx context.Context, key string) (*service.APIKey, error) {
+			if key != apiKey.Key {
+				return nil, service.ErrAPIKeyNotFound
+			}
+			clone := *apiKey
+			return &clone, nil
+		},
+	}
+
+	cfg := &config.Config{RunMode: config.RunModeSimple}
+	apiKeyService := service.NewAPIKeyService(apiKeyRepo, nil, nil, nil, nil, nil, cfg)
+	router := gin.New()
+	require.NoError(t, router.SetTrustedProxies(nil))
+	router.Use(gin.HandlerFunc(NewAPIKeyAuthMiddleware(apiKeyService, nil, cfg)))
+	router.GET("/t", func(c *gin.Context) {
+		c.JSON(http.StatusOK, gin.H{"ok": true})
+	})
+
+	w := httptest.NewRecorder()
+	req := httptest.NewRequest(http.MethodGet, "/t", nil)
+	req.RemoteAddr = "9.9.9.9:12345"
+	req.Header.Set("x-api-key", apiKey.Key)
+	req.Header.Set("X-Forwarded-For", "1.2.3.4")
+	req.Header.Set("X-Real-IP", "1.2.3.4")
+	req.Header.Set("CF-Connecting-IP", "1.2.3.4")
+	router.ServeHTTP(w, req)
+
+	require.Equal(t, http.StatusForbidden, w.Code)
+	require.Contains(t, w.Body.String(), "ACCESS_DENIED")
+}
+
 func newAuthTestRouter(apiKeyService *service.APIKeyService, subscriptionService *service.SubscriptionService, cfg *config.Config) *gin.Engine {
 	router := gin.New()
 	router.Use(gin.HandlerFunc(NewAPIKeyAuthMiddleware(apiKeyService, subscriptionService, cfg)))
diff --git a/backend/internal/server/routes/auth.go b/backend/internal/server/routes/auth.go
index 26d79605..c168820c 100644
--- a/backend/internal/server/routes/auth.go
+++ b/backend/internal/server/routes/auth.go
@@ -24,10 +24,19 @@ func RegisterAuthRoutes(
 	// 公开接口
 	auth := v1.Group("/auth")
 	{
-		auth.POST("/register", h.Auth.Register)
-		auth.POST("/login", h.Auth.Login)
-		auth.POST("/login/2fa", h.Auth.Login2FA)
-		auth.POST("/send-verify-code", h.Auth.SendVerifyCode)
+		// 注册/登录/2FA/验证码发送均属于高风险入口，增加服务端兜底限流（Redis 故障时 fail-close）
+		auth.POST("/register", rateLimiter.LimitWithOptions("auth-register", 5, time.Minute, middleware.RateLimitOptions{
+			FailureMode: middleware.RateLimitFailClose,
+		}), h.Auth.Register)
+		auth.POST("/login", rateLimiter.LimitWithOptions("auth-login", 20, time.Minute, middleware.RateLimitOptions{
+			FailureMode: middleware.RateLimitFailClose,
+		}), h.Auth.Login)
+		auth.POST("/login/2fa", rateLimiter.LimitWithOptions("auth-login-2fa", 20, time.Minute, middleware.RateLimitOptions{
+			FailureMode: middleware.RateLimitFailClose,
+		}), h.Auth.Login2FA)
+		auth.POST("/send-verify-code", rateLimiter.LimitWithOptions("auth-send-verify-code", 5, time.Minute, middleware.RateLimitOptions{
+			FailureMode: middleware.RateLimitFailClose,
+		}), h.Auth.SendVerifyCode)
 		// Token刷新接口添加速率限制：每分钟最多 30 次（Redis 故障时 fail-close）
 		auth.POST("/refresh", rateLimiter.LimitWithOptions("refresh-token", 30, time.Minute, middleware.RateLimitOptions{
 			FailureMode: middleware.RateLimitFailClose,
diff --git a/backend/internal/server/routes/auth_rate_limit_integration_test.go b/backend/internal/server/routes/auth_rate_limit_integration_test.go
new file mode 100644
index 00000000..8a0ef860
--- /dev/null
+++ b/backend/internal/server/routes/auth_rate_limit_integration_test.go
@@ -0,0 +1,111 @@
+//go:build integration
+
+package routes
+
+import (
+	"context"
+	"fmt"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+	"strconv"
+	"strings"
+	"testing"
+
+	"github.com/redis/go-redis/v9"
+	"github.com/stretchr/testify/require"
+	tcredis "github.com/testcontainers/testcontainers-go/modules/redis"
+)
+
+const authRouteRedisImageTag = "redis:8.4-alpine"
+
+func TestAuthRegisterRateLimitThresholdHitReturns429(t *testing.T) {
+	ctx := context.Background()
+	rdb := startAuthRouteRedis(t, ctx)
+
+	router := newAuthRoutesTestRouter(rdb)
+	const path = "/api/v1/auth/register"
+
+	for i := 1; i <= 6; i++ {
+		req := httptest.NewRequest(http.MethodPost, path, strings.NewReader(`{}`))
+		req.Header.Set("Content-Type", "application/json")
+		req.RemoteAddr = "198.51.100.10:23456"
+
+		w := httptest.NewRecorder()
+		router.ServeHTTP(w, req)
+
+		if i <= 5 {
+			require.Equal(t, http.StatusBadRequest, w.Code, "第 %d 次请求应先进入业务校验", i)
+			continue
+		}
+		require.Equal(t, http.StatusTooManyRequests, w.Code, "第 6 次请求应命中限流")
+		require.Contains(t, w.Body.String(), "rate limit exceeded")
+	}
+}
+
+func startAuthRouteRedis(t *testing.T, ctx context.Context) *redis.Client {
+	t.Helper()
+	ensureAuthRouteDockerAvailable(t)
+
+	redisContainer, err := tcredis.Run(ctx, authRouteRedisImageTag)
+	require.NoError(t, err)
+	t.Cleanup(func() {
+		_ = redisContainer.Terminate(ctx)
+	})
+
+	redisHost, err := redisContainer.Host(ctx)
+	require.NoError(t, err)
+	redisPort, err := redisContainer.MappedPort(ctx, "6379/tcp")
+	require.NoError(t, err)
+
+	rdb := redis.NewClient(&redis.Options{
+		Addr: fmt.Sprintf("%s:%d", redisHost, redisPort.Int()),
+		DB:   0,
+	})
+	require.NoError(t, rdb.Ping(ctx).Err())
+	t.Cleanup(func() {
+		_ = rdb.Close()
+	})
+	return rdb
+}
+
+func ensureAuthRouteDockerAvailable(t *testing.T) {
+	t.Helper()
+	if authRouteDockerAvailable() {
+		return
+	}
+	t.Skip("Docker 未启用，跳过认证限流集成测试")
+}
+
+func authRouteDockerAvailable() bool {
+	if os.Getenv("DOCKER_HOST") != "" {
+		return true
+	}
+
+	socketCandidates := []string{
+		"/var/run/docker.sock",
+		filepath.Join(os.Getenv("XDG_RUNTIME_DIR"), "docker.sock"),
+		filepath.Join(authRouteUserHomeDir(), ".docker", "run", "docker.sock"),
+		filepath.Join(authRouteUserHomeDir(), ".docker", "desktop", "docker.sock"),
+		filepath.Join("/run/user", strconv.Itoa(os.Getuid()), "docker.sock"),
+	}
+
+	for _, socket := range socketCandidates {
+		if socket == "" {
+			continue
+		}
+		if _, err := os.Stat(socket); err == nil {
+			return true
+		}
+	}
+	return false
+}
+
+func authRouteUserHomeDir() string {
+	home, err := os.UserHomeDir()
+	if err != nil {
+		return ""
+	}
+	return home
+}
diff --git a/backend/internal/server/routes/auth_rate_limit_test.go b/backend/internal/server/routes/auth_rate_limit_test.go
new file mode 100644
index 00000000..5ce8497c
--- /dev/null
+++ b/backend/internal/server/routes/auth_rate_limit_test.go
@@ -0,0 +1,67 @@
+package routes
+
+import (
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/Wei-Shaw/sub2api/internal/handler"
+	servermiddleware "github.com/Wei-Shaw/sub2api/internal/server/middleware"
+	"github.com/gin-gonic/gin"
+	"github.com/redis/go-redis/v9"
+	"github.com/stretchr/testify/require"
+)
+
+func newAuthRoutesTestRouter(redisClient *redis.Client) *gin.Engine {
+	gin.SetMode(gin.TestMode)
+	router := gin.New()
+	v1 := router.Group("/api/v1")
+
+	RegisterAuthRoutes(
+		v1,
+		&handler.Handlers{
+			Auth:    &handler.AuthHandler{},
+			Setting: &handler.SettingHandler{},
+		},
+		servermiddleware.JWTAuthMiddleware(func(c *gin.Context) {
+			c.Next()
+		}),
+		redisClient,
+	)
+
+	return router
+}
+
+func TestAuthRoutesRateLimitFailCloseWhenRedisUnavailable(t *testing.T) {
+	rdb := redis.NewClient(&redis.Options{
+		Addr:         "127.0.0.1:1",
+		DialTimeout:  50 * time.Millisecond,
+		ReadTimeout:  50 * time.Millisecond,
+		WriteTimeout: 50 * time.Millisecond,
+	})
+	t.Cleanup(func() {
+		_ = rdb.Close()
+	})
+
+	router := newAuthRoutesTestRouter(rdb)
+	paths := []string{
+		"/api/v1/auth/register",
+		"/api/v1/auth/login",
+		"/api/v1/auth/login/2fa",
+		"/api/v1/auth/send-verify-code",
+	}
+
+	for _, path := range paths {
+		req := httptest.NewRequest(http.MethodPost, path, strings.NewReader(`{}`))
+		req.Header.Set("Content-Type", "application/json")
+		req.RemoteAddr = "203.0.113.10:12345"
+
+		w := httptest.NewRecorder()
+		router.ServeHTTP(w, req)
+
+		require.Equal(t, http.StatusTooManyRequests, w.Code, "path=%s", path)
+		require.Contains(t, w.Body.String(), "rate limit exceeded", "path=%s", path)
+	}
+}
diff --git a/backend/internal/service/gateway_service.go b/backend/internal/service/gateway_service.go
index 83cde19e..0502d352 100644
--- a/backend/internal/service/gateway_service.go
+++ b/backend/internal/service/gateway_service.go
@@ -3332,7 +3332,7 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 
 		// 不需要重试（成功或不可重试的错误），跳出循环
 		// DEBUG: 输出响应 headers（用于检测 rate limit 信息）
-		if account.Platform == PlatformGemini && resp.StatusCode < 400 {
+		if account.Platform == PlatformGemini && resp.StatusCode < 400 && s.cfg != nil && s.cfg.Gateway.GeminiDebugResponseHeaders {
 			logger.LegacyPrintf("service.gateway", "[DEBUG] Gemini API Response Headers for account %d:", account.ID)
 			for k, v := range resp.Header {
 				logger.LegacyPrintf("service.gateway", "[DEBUG]   %s: %v", k, v)
@@ -4467,8 +4467,19 @@ func (s *GatewayService) handleNonStreamingResponse(ctx context.Context, resp *h
 	// 更新5h窗口状态
 	s.rateLimitService.UpdateSessionWindow(ctx, account, resp.Header)
 
-	body, err := io.ReadAll(resp.Body)
+	maxBytes := resolveUpstreamResponseReadLimit(s.cfg)
+	body, err := readUpstreamResponseBodyLimited(resp.Body, maxBytes)
 	if err != nil {
+		if errors.Is(err, ErrUpstreamResponseBodyTooLarge) {
+			setOpsUpstreamError(c, http.StatusBadGateway, "upstream response too large", "")
+			c.JSON(http.StatusBadGateway, gin.H{
+				"type": "error",
+				"error": gin.H{
+					"type":    "upstream_error",
+					"message": "Upstream response too large",
+				},
+			})
+		}
 		return nil, err
 	}
 
@@ -4990,9 +5001,15 @@ func (s *GatewayService) ForwardCountTokens(ctx context.Context, c *gin.Context,
 	}
 
 	// 读取响应体
-	respBody, err := io.ReadAll(resp.Body)
+	maxReadBytes := resolveUpstreamResponseReadLimit(s.cfg)
+	respBody, err := readUpstreamResponseBodyLimited(resp.Body, maxReadBytes)
 	_ = resp.Body.Close()
 	if err != nil {
+		if errors.Is(err, ErrUpstreamResponseBodyTooLarge) {
+			setOpsUpstreamError(c, http.StatusBadGateway, "upstream response too large", "")
+			s.countTokensError(c, http.StatusBadGateway, "upstream_error", "Upstream response too large")
+			return err
+		}
 		s.countTokensError(c, http.StatusBadGateway, "upstream_error", "Failed to read response")
 		return err
 	}
@@ -5007,9 +5024,14 @@ func (s *GatewayService) ForwardCountTokens(ctx context.Context, c *gin.Context,
 			retryResp, retryErr := s.httpUpstream.DoWithTLS(retryReq, proxyURL, account.ID, account.Concurrency, account.IsTLSFingerprintEnabled())
 			if retryErr == nil {
 				resp = retryResp
-				respBody, err = io.ReadAll(resp.Body)
+				respBody, err = readUpstreamResponseBodyLimited(resp.Body, maxReadBytes)
 				_ = resp.Body.Close()
 				if err != nil {
+					if errors.Is(err, ErrUpstreamResponseBodyTooLarge) {
+						setOpsUpstreamError(c, http.StatusBadGateway, "upstream response too large", "")
+						s.countTokensError(c, http.StatusBadGateway, "upstream_error", "Upstream response too large")
+						return err
+					}
 					s.countTokensError(c, http.StatusBadGateway, "upstream_error", "Failed to read response")
 					return err
 				}
diff --git a/backend/internal/service/gemini_messages_compat_service.go b/backend/internal/service/gemini_messages_compat_service.go
index 2fe55137..8670f99a 100644
--- a/backend/internal/service/gemini_messages_compat_service.go
+++ b/backend/internal/service/gemini_messages_compat_service.go
@@ -2358,29 +2358,36 @@ type UpstreamHTTPResult struct {
 }
 
 func (s *GeminiMessagesCompatService) handleNativeNonStreamingResponse(c *gin.Context, resp *http.Response, isOAuth bool) (*ClaudeUsage, error) {
-	// Log response headers for debugging
-	logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ========== Response Headers ==========")
-	for key, values := range resp.Header {
-		if strings.HasPrefix(strings.ToLower(key), "x-ratelimit") {
-			logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] %s: %v", key, values)
+	if s.cfg != nil && s.cfg.Gateway.GeminiDebugResponseHeaders {
+		logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ========== Response Headers ==========")
+		for key, values := range resp.Header {
+			if strings.HasPrefix(strings.ToLower(key), "x-ratelimit") {
+				logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] %s: %v", key, values)
+			}
 		}
+		logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ========================================")
 	}
-	logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ========================================")
 
-	respBody, err := io.ReadAll(resp.Body)
+	maxBytes := resolveUpstreamResponseReadLimit(s.cfg)
+	respBody, err := readUpstreamResponseBodyLimited(resp.Body, maxBytes)
 	if err != nil {
+		if errors.Is(err, ErrUpstreamResponseBodyTooLarge) {
+			setOpsUpstreamError(c, http.StatusBadGateway, "upstream response too large", "")
+			c.JSON(http.StatusBadGateway, gin.H{
+				"error": gin.H{
+					"type":    "upstream_error",
+					"message": "Upstream response too large",
+				},
+			})
+		}
 		return nil, err
 	}
 
-	var parsed map[string]any
 	if isOAuth {
 		unwrappedBody, uwErr := unwrapGeminiResponse(respBody)
 		if uwErr == nil {
 			respBody = unwrappedBody
 		}
-		_ = json.Unmarshal(respBody, &parsed)
-	} else {
-		_ = json.Unmarshal(respBody, &parsed)
 	}
 
 	responseheaders.WriteFilteredHeaders(c.Writer.Header(), resp.Header, s.cfg.Security.ResponseHeaders)
@@ -2398,14 +2405,15 @@ func (s *GeminiMessagesCompatService) handleNativeNonStreamingResponse(c *gin.Co
 }
 
 func (s *GeminiMessagesCompatService) handleNativeStreamingResponse(c *gin.Context, resp *http.Response, startTime time.Time, isOAuth bool) (*geminiNativeStreamResult, error) {
-	// Log response headers for debugging
-	logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ========== Streaming Response Headers ==========")
-	for key, values := range resp.Header {
-		if strings.HasPrefix(strings.ToLower(key), "x-ratelimit") {
-			logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] %s: %v", key, values)
+	if s.cfg != nil && s.cfg.Gateway.GeminiDebugResponseHeaders {
+		logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ========== Streaming Response Headers ==========")
+		for key, values := range resp.Header {
+			if strings.HasPrefix(strings.ToLower(key), "x-ratelimit") {
+				logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] %s: %v", key, values)
+			}
 		}
+		logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ====================================================")
 	}
-	logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ====================================================")
 
 	if s.cfg != nil {
 		responseheaders.WriteFilteredHeaders(c.Writer.Header(), resp.Header, s.cfg.Security.ResponseHeaders)
diff --git a/backend/internal/service/gemini_messages_compat_service_test.go b/backend/internal/service/gemini_messages_compat_service_test.go
index c5888d88..7560f480 100644
--- a/backend/internal/service/gemini_messages_compat_service_test.go
+++ b/backend/internal/service/gemini_messages_compat_service_test.go
@@ -3,10 +3,15 @@ package service
 import (
 	"encoding/json"
 	"fmt"
+	"io"
+	"net/http"
+	"net/http/httptest"
 	"strings"
 	"testing"
 	"time"
 
+	"github.com/Wei-Shaw/sub2api/internal/config"
+	"github.com/gin-gonic/gin"
 	"github.com/stretchr/testify/require"
 )
 
@@ -133,6 +138,38 @@ func TestConvertClaudeToolsToGeminiTools_CustomType(t *testing.T) {
 	}
 }
 
+func TestGeminiHandleNativeNonStreamingResponse_DebugDisabledDoesNotEmitHeaderLogs(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	logSink, restore := captureStructuredLog(t)
+	defer restore()
+
+	svc := &GeminiMessagesCompatService{
+		cfg: &config.Config{
+			Gateway: config.GatewayConfig{
+				GeminiDebugResponseHeaders: false,
+			},
+		},
+	}
+
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	c.Request = httptest.NewRequest(http.MethodPost, "/v1/messages", nil)
+
+	resp := &http.Response{
+		StatusCode: http.StatusOK,
+		Header: http.Header{
+			"Content-Type":      []string{"application/json"},
+			"X-RateLimit-Limit": []string{"60"},
+		},
+		Body: io.NopCloser(strings.NewReader(`{"usageMetadata":{"promptTokenCount":10,"candidatesTokenCount":2}}`)),
+	}
+
+	usage, err := svc.handleNativeNonStreamingResponse(c, resp, false)
+	require.NoError(t, err)
+	require.NotNil(t, usage)
+	require.False(t, logSink.ContainsMessage("[GeminiAPI]"), "debug 关闭时不应输出 Gemini 响应头日志")
+}
+
 func TestConvertClaudeMessagesToGeminiGenerateContent_AddsThoughtSignatureForToolUse(t *testing.T) {
 	claudeReq := map[string]any{
 		"model":      "claude-haiku-4-5-20251001",
diff --git a/backend/internal/service/openai_gateway_service.go b/backend/internal/service/openai_gateway_service.go
index 157506a6..ac93aa0c 100644
--- a/backend/internal/service/openai_gateway_service.go
+++ b/backend/internal/service/openai_gateway_service.go
@@ -1741,8 +1741,18 @@ func (s *OpenAIGatewayService) handleNonStreamingResponsePassthrough(
 	resp *http.Response,
 	c *gin.Context,
 ) (*OpenAIUsage, error) {
-	body, err := io.ReadAll(resp.Body)
+	maxBytes := resolveUpstreamResponseReadLimit(s.cfg)
+	body, err := readUpstreamResponseBodyLimited(resp.Body, maxBytes)
 	if err != nil {
+		if errors.Is(err, ErrUpstreamResponseBodyTooLarge) {
+			setOpsUpstreamError(c, http.StatusBadGateway, "upstream response too large", "")
+			c.JSON(http.StatusBadGateway, gin.H{
+				"error": gin.H{
+					"type":    "upstream_error",
+					"message": "Upstream response too large",
+				},
+			})
+		}
 		return nil, err
 	}
 
@@ -2371,8 +2381,18 @@ func (s *OpenAIGatewayService) parseSSEUsage(data string, usage *OpenAIUsage) {
 }
 
 func (s *OpenAIGatewayService) handleNonStreamingResponse(ctx context.Context, resp *http.Response, c *gin.Context, account *Account, originalModel, mappedModel string) (*OpenAIUsage, error) {
-	body, err := io.ReadAll(resp.Body)
+	maxBytes := resolveUpstreamResponseReadLimit(s.cfg)
+	body, err := readUpstreamResponseBodyLimited(resp.Body, maxBytes)
 	if err != nil {
+		if errors.Is(err, ErrUpstreamResponseBodyTooLarge) {
+			setOpsUpstreamError(c, http.StatusBadGateway, "upstream response too large", "")
+			c.JSON(http.StatusBadGateway, gin.H{
+				"error": gin.H{
+					"type":    "upstream_error",
+					"message": "Upstream response too large",
+				},
+			})
+		}
 		return nil, err
 	}
 
@@ -2930,6 +2950,25 @@ func normalizeOpenAIPassthroughOAuthBody(body []byte) ([]byte, bool, error) {
 	return normalized, changed, nil
 }
 
+func detectOpenAIPassthroughInstructionsRejectReason(reqModel string, body []byte) string {
+	model := strings.ToLower(strings.TrimSpace(reqModel))
+	if !strings.Contains(model, "codex") {
+		return ""
+	}
+
+	instructions := gjson.GetBytes(body, "instructions")
+	if !instructions.Exists() {
+		return "instructions_missing"
+	}
+	if instructions.Type != gjson.String {
+		return "instructions_not_string"
+	}
+	if strings.TrimSpace(instructions.String()) == "" {
+		return "instructions_empty"
+	}
+	return ""
+}
+
 func extractOpenAIReasoningEffortFromBody(body []byte, requestedModel string) *string {
 	reasoningEffort := strings.TrimSpace(gjson.GetBytes(body, "reasoning.effort").String())
 	if reasoningEffort == "" {
@@ -3002,22 +3041,3 @@ func normalizeOpenAIReasoningEffort(raw string) string {
 		return ""
 	}
 }
-func detectOpenAIPassthroughInstructionsRejectReason(reqModel string, body []byte) string {
-	model := strings.ToLower(strings.TrimSpace(reqModel))
-	if !strings.Contains(model, "codex") {
-		return ""
-	}
-
-	instructions := gjson.GetBytes(body, "instructions")
-	if !instructions.Exists() {
-		return "instructions_missing"
-	}
-	if instructions.Type != gjson.String {
-		return "instructions_not_string"
-	}
-	if strings.TrimSpace(instructions.String()) == "" {
-		return "instructions_empty"
-	}
-	return ""
-}
-
diff --git a/backend/internal/service/upstream_response_limit.go b/backend/internal/service/upstream_response_limit.go
new file mode 100644
index 00000000..aecf69a3
--- /dev/null
+++ b/backend/internal/service/upstream_response_limit.go
@@ -0,0 +1,38 @@
+package service
+
+import (
+	"errors"
+	"fmt"
+	"io"
+
+	"github.com/Wei-Shaw/sub2api/internal/config"
+)
+
+var ErrUpstreamResponseBodyTooLarge = errors.New("upstream response body too large")
+
+const defaultUpstreamResponseReadMaxBytes int64 = 8 * 1024 * 1024
+
+func resolveUpstreamResponseReadLimit(cfg *config.Config) int64 {
+	if cfg != nil && cfg.Gateway.UpstreamResponseReadMaxBytes > 0 {
+		return cfg.Gateway.UpstreamResponseReadMaxBytes
+	}
+	return defaultUpstreamResponseReadMaxBytes
+}
+
+func readUpstreamResponseBodyLimited(reader io.Reader, maxBytes int64) ([]byte, error) {
+	if reader == nil {
+		return nil, errors.New("response body is nil")
+	}
+	if maxBytes <= 0 {
+		maxBytes = defaultUpstreamResponseReadMaxBytes
+	}
+
+	body, err := io.ReadAll(io.LimitReader(reader, maxBytes+1))
+	if err != nil {
+		return nil, err
+	}
+	if int64(len(body)) > maxBytes {
+		return nil, fmt.Errorf("%w: limit=%d", ErrUpstreamResponseBodyTooLarge, maxBytes)
+	}
+	return body, nil
+}
diff --git a/backend/internal/service/upstream_response_limit_test.go b/backend/internal/service/upstream_response_limit_test.go
new file mode 100644
index 00000000..b9e5cc6d
--- /dev/null
+++ b/backend/internal/service/upstream_response_limit_test.go
@@ -0,0 +1,37 @@
+package service
+
+import (
+	"bytes"
+	"errors"
+	"testing"
+
+	"github.com/Wei-Shaw/sub2api/internal/config"
+	"github.com/stretchr/testify/require"
+)
+
+func TestResolveUpstreamResponseReadLimit(t *testing.T) {
+	t.Run("use default when config missing", func(t *testing.T) {
+		require.Equal(t, defaultUpstreamResponseReadMaxBytes, resolveUpstreamResponseReadLimit(nil))
+	})
+
+	t.Run("use configured value", func(t *testing.T) {
+		cfg := &config.Config{}
+		cfg.Gateway.UpstreamResponseReadMaxBytes = 1234
+		require.Equal(t, int64(1234), resolveUpstreamResponseReadLimit(cfg))
+	})
+}
+
+func TestReadUpstreamResponseBodyLimited(t *testing.T) {
+	t.Run("within limit", func(t *testing.T) {
+		body, err := readUpstreamResponseBodyLimited(bytes.NewReader([]byte("ok")), 2)
+		require.NoError(t, err)
+		require.Equal(t, []byte("ok"), body)
+	})
+
+	t.Run("exceeds limit", func(t *testing.T) {
+		body, err := readUpstreamResponseBodyLimited(bytes.NewReader([]byte("toolong")), 3)
+		require.Nil(t, body)
+		require.Error(t, err)
+		require.True(t, errors.Is(err, ErrUpstreamResponseBodyTooLarge))
+	})
+}
diff --git a/deploy/config.example.yaml b/deploy/config.example.yaml
index f016fd49..9fd2d391 100644
--- a/deploy/config.example.yaml
+++ b/deploy/config.example.yaml
@@ -146,6 +146,15 @@ gateway:
   # Max request body size in bytes (default: 100MB)
   # 请求体最大字节数（默认 100MB）
   max_body_size: 104857600
+  # Max bytes to read for non-stream upstream responses (default: 8MB)
+  # 非流式上游响应体读取上限（默认 8MB）
+  upstream_response_read_max_bytes: 8388608
+  # Max bytes to read for proxy probe responses (default: 1MB)
+  # 代理探测响应体读取上限（默认 1MB）
+  proxy_probe_response_read_max_bytes: 1048576
+  # Enable Gemini upstream response header debug logs (default: false)
+  # 是否开启 Gemini 上游响应头调试日志（默认 false）
+  gemini_debug_response_headers: false
   # Sora max request body size in bytes (0=use max_body_size)
   # Sora 请求体最大字节数（0=使用 max_body_size）
   sora_max_body_size: 268435456