fix: Gemini OpenCode 教程 baseURL 改为 v1beta

Merge pull request #226 from xilu0/main
feat(gateway): 优化 Antigravity/Gemini 思考块处理此提交解决了思考块 (thinking blocks) 在转发过程中的兼容性问题
2026-04-06 16:30:22 +08:00 · 2026-01-13 09:52:37 +08:00 · 2026-01-13 09:39:43 +08:00 · 2026-01-13 09:31:00 +08:00 · 2026-01-13 09:30:26 +08:00 · 2026-01-13 09:29:25 +08:00
26 changed files with 705 additions and 157 deletions
--- a/backend/internal/handler/auth_handler.go
+++ b/backend/internal/handler/auth_handler.go
@@ -3,6 +3,7 @@ package handler
 import (
 	"github.com/Wei-Shaw/sub2api/internal/config"
 	"github.com/Wei-Shaw/sub2api/internal/handler/dto"
+	"github.com/Wei-Shaw/sub2api/internal/pkg/ip"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/response"
 	middleware2 "github.com/Wei-Shaw/sub2api/internal/server/middleware"
 	"github.com/Wei-Shaw/sub2api/internal/service"
@@ -76,7 +77,7 @@ func (h *AuthHandler) Register(c *gin.Context) {

 	// Turnstile 验证（当提供了邮箱验证码时跳过，因为发送验证码时已验证过）
 	if req.VerifyCode == "" {
-		if err := h.authService.VerifyTurnstile(c.Request.Context(), req.TurnstileToken, c.ClientIP()); err != nil {
+		if err := h.authService.VerifyTurnstile(c.Request.Context(), req.TurnstileToken, ip.GetClientIP(c)); err != nil {
 			response.ErrorFrom(c, err)
 			return
 		}
@@ -105,7 +106,7 @@ func (h *AuthHandler) SendVerifyCode(c *gin.Context) {
 	}

 	// Turnstile 验证
-	if err := h.authService.VerifyTurnstile(c.Request.Context(), req.TurnstileToken, c.ClientIP()); err != nil {
+	if err := h.authService.VerifyTurnstile(c.Request.Context(), req.TurnstileToken, ip.GetClientIP(c)); err != nil {
 		response.ErrorFrom(c, err)
 		return
 	}
@@ -132,7 +133,7 @@ func (h *AuthHandler) Login(c *gin.Context) {
 	}

 	// Turnstile 验证
-	if err := h.authService.VerifyTurnstile(c.Request.Context(), req.TurnstileToken, c.ClientIP()); err != nil {
+	if err := h.authService.VerifyTurnstile(c.Request.Context(), req.TurnstileToken, ip.GetClientIP(c)); err != nil {
 		response.ErrorFrom(c, err)
 		return
 	}
--- a/backend/internal/handler/gateway_handler.go
+++ b/backend/internal/handler/gateway_handler.go
@@ -275,12 +275,11 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 				var failoverErr *service.UpstreamFailoverError
 				if errors.As(err, &failoverErr) {
 					failedAccountIDs[account.ID] = struct{}{}
+					lastFailoverStatus = failoverErr.StatusCode
 					if switchCount >= maxAccountSwitches {
-						lastFailoverStatus = failoverErr.StatusCode
 						h.handleFailoverExhausted(c, lastFailoverStatus, streamStarted)
 						return
 					}
-					lastFailoverStatus = failoverErr.StatusCode
 					switchCount++
 					log.Printf("Account %d: upstream error %d, switching account %d/%d", account.ID, failoverErr.StatusCode, switchCount, maxAccountSwitches)
 					continue
@@ -409,12 +408,11 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 			var failoverErr *service.UpstreamFailoverError
 			if errors.As(err, &failoverErr) {
 				failedAccountIDs[account.ID] = struct{}{}
+				lastFailoverStatus = failoverErr.StatusCode
 				if switchCount >= maxAccountSwitches {
-					lastFailoverStatus = failoverErr.StatusCode
 					h.handleFailoverExhausted(c, lastFailoverStatus, streamStarted)
 					return
 				}
-				lastFailoverStatus = failoverErr.StatusCode
 				switchCount++
 				log.Printf("Account %d: upstream error %d, switching account %d/%d", account.ID, failoverErr.StatusCode, switchCount, maxAccountSwitches)
 				continue
--- a/backend/internal/handler/gemini_v1beta_handler.go
+++ b/backend/internal/handler/gemini_v1beta_handler.go
@@ -12,6 +12,7 @@ import (
 	"github.com/Wei-Shaw/sub2api/internal/pkg/antigravity"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/gemini"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/googleapi"
+	"github.com/Wei-Shaw/sub2api/internal/pkg/ip"
 	"github.com/Wei-Shaw/sub2api/internal/server/middleware"
 	"github.com/Wei-Shaw/sub2api/internal/service"

@@ -316,7 +317,7 @@ func (h *GatewayHandler) GeminiV1BetaModels(c *gin.Context) {

 		// 捕获请求信息（用于异步记录，避免在 goroutine 中访问 gin.Context）
 		userAgent := c.GetHeader("User-Agent")
-		clientIP := c.ClientIP()
+		clientIP := ip.GetClientIP(c)

 		// 6) record usage async
 		go func(result *service.ForwardResult, usedAccount *service.Account, ua, ip string) {
--- a/backend/internal/handler/openai_gateway_handler.go
+++ b/backend/internal/handler/openai_gateway_handler.go
@@ -12,6 +12,7 @@ import (
 	"time"

 	"github.com/Wei-Shaw/sub2api/internal/config"
+	"github.com/Wei-Shaw/sub2api/internal/pkg/ip"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/openai"
 	middleware2 "github.com/Wei-Shaw/sub2api/internal/server/middleware"
 	"github.com/Wei-Shaw/sub2api/internal/service"
@@ -265,7 +266,7 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {

 		// 捕获请求信息（用于异步记录，避免在 goroutine 中访问 gin.Context）
 		userAgent := c.GetHeader("User-Agent")
-		clientIP := c.ClientIP()
+		clientIP := ip.GetClientIP(c)

 		// Async record usage
 		go func(result *service.OpenAIForwardResult, usedAccount *service.Account, ua, ip string) {
--- a/backend/internal/handler/ops_error_logger.go
+++ b/backend/internal/handler/ops_error_logger.go
@@ -15,6 +15,7 @@ import (
 	"unicode/utf8"

 	"github.com/Wei-Shaw/sub2api/internal/pkg/ctxkey"
+	"github.com/Wei-Shaw/sub2api/internal/pkg/ip"
 	middleware2 "github.com/Wei-Shaw/sub2api/internal/server/middleware"
 	"github.com/Wei-Shaw/sub2api/internal/service"
 	"github.com/gin-gonic/gin"
@@ -489,6 +490,7 @@ func OpsErrorLoggerMiddleware(ops *service.OpsService) gin.HandlerFunc {
 				Severity:          classifyOpsSeverity("upstream_error", effectiveUpstreamStatus),
 				StatusCode:        status,
 				IsBusinessLimited: false,
+				IsCountTokens:     isCountTokensRequest(c),

 				ErrorMessage: recoveredMsg,
 				ErrorBody:    "",
@@ -521,7 +523,7 @@ func OpsErrorLoggerMiddleware(ops *service.OpsService) gin.HandlerFunc {
 			}

 			var clientIP string
-			if ip := strings.TrimSpace(c.ClientIP()); ip != "" {
+			if ip := strings.TrimSpace(ip.GetClientIP(c)); ip != "" {
 				clientIP = ip
 				entry.ClientIP = &clientIP
 			}
@@ -598,6 +600,7 @@ func OpsErrorLoggerMiddleware(ops *service.OpsService) gin.HandlerFunc {
 			Severity:          classifyOpsSeverity(parsed.ErrorType, status),
 			StatusCode:        status,
 			IsBusinessLimited: isBusinessLimited,
+			IsCountTokens:     isCountTokensRequest(c),

 			ErrorMessage: parsed.Message,
 			// Keep the full captured error body (capture is already capped at 64KB) so the
@@ -680,7 +683,7 @@ func OpsErrorLoggerMiddleware(ops *service.OpsService) gin.HandlerFunc {
 		}

 		var clientIP string
-		if ip := strings.TrimSpace(c.ClientIP()); ip != "" {
+		if ip := strings.TrimSpace(ip.GetClientIP(c)); ip != "" {
 			clientIP = ip
 			entry.ClientIP = &clientIP
 		}
@@ -704,6 +707,14 @@ var opsRetryRequestHeaderAllowlist = []string{
 	"anthropic-version",
 }

+// isCountTokensRequest checks if the request is a count_tokens request
+func isCountTokensRequest(c *gin.Context) bool {
+	if c == nil || c.Request == nil || c.Request.URL == nil {
+		return false
+	}
+	return strings.Contains(c.Request.URL.Path, "/count_tokens")
+}
+
 func extractOpsRetryRequestHeaders(c *gin.Context) *string {
 	if c == nil || c.Request == nil {
 		return nil
--- a/backend/internal/repository/ops_repo.go
+++ b/backend/internal/repository/ops_repo.go
@@ -46,6 +46,7 @@ INSERT INTO ops_error_logs (
  severity,
  status_code,
  is_business_limited,
+  is_count_tokens,
  error_message,
  error_body,
  error_source,
@@ -64,7 +65,7 @@ INSERT INTO ops_error_logs (
  retry_count,
  created_at
 ) VALUES (
-  $1,$2,$3,$4,$5,$6,$7,$8,$9,$10,$11,$12,$13,$14,$15,$16,$17,$18,$19,$20,$21,$22,$23,$24,$25,$26,$27,$28,$29,$30,$31,$32,$33,$34
+  $1,$2,$3,$4,$5,$6,$7,$8,$9,$10,$11,$12,$13,$14,$15,$16,$17,$18,$19,$20,$21,$22,$23,$24,$25,$26,$27,$28,$29,$30,$31,$32,$33,$34,$35
 ) RETURNING id`

 	var id int64
@@ -88,6 +89,7 @@ INSERT INTO ops_error_logs (
 		opsNullString(input.Severity),
 		opsNullInt(input.StatusCode),
 		input.IsBusinessLimited,
+		input.IsCountTokens,
 		opsNullString(input.ErrorMessage),
 		opsNullString(input.ErrorBody),
 		opsNullString(input.ErrorSource),
--- a/backend/internal/repository/ops_repo_dashboard.go
+++ b/backend/internal/repository/ops_repo_dashboard.go
@@ -964,8 +964,8 @@ func buildErrorWhere(filter *service.OpsDashboardFilter, start, end time.Time, s
 	}

 	idx := startIndex
-	clauses := make([]string, 0, 4)
-	args = make([]any, 0, 4)
+	clauses := make([]string, 0, 5)
+	args = make([]any, 0, 5)

 	args = append(args, start)
 	clauses = append(clauses, fmt.Sprintf("created_at >= $%d", idx))
@@ -974,6 +974,8 @@ func buildErrorWhere(filter *service.OpsDashboardFilter, start, end time.Time, s
 	clauses = append(clauses, fmt.Sprintf("created_at < $%d", idx))
 	idx++

+	clauses = append(clauses, "is_count_tokens = FALSE")
+
 	if groupID != nil && *groupID > 0 {
 		args = append(args, *groupID)
 		clauses = append(clauses, fmt.Sprintf("group_id = $%d", idx))
--- a/backend/internal/repository/ops_repo_preagg.go
+++ b/backend/internal/repository/ops_repo_preagg.go
@@ -78,7 +78,9 @@ error_base AS (
    status_code AS client_status_code,
    COALESCE(upstream_status_code, status_code, 0) AS effective_status_code
  FROM ops_error_logs
+  -- Exclude count_tokens requests from error metrics as they are informational probes
  WHERE created_at >= $1 AND created_at < $2
+    AND is_count_tokens = FALSE
 ),
 error_agg AS (
  SELECT
--- a/backend/internal/repository/ops_repo_trends.go
+++ b/backend/internal/repository/ops_repo_trends.go
@@ -170,6 +170,7 @@ error_totals AS (
  FROM ops_error_logs
  WHERE created_at >= $1 AND created_at < $2
    AND COALESCE(status_code, 0) >= 400
+    AND is_count_tokens = FALSE  -- 排除 count_tokens 请求的错误
  GROUP BY 1
 ),
 combined AS (
@@ -243,6 +244,7 @@ error_totals AS (
    AND platform = $3
    AND group_id IS NOT NULL
    AND COALESCE(status_code, 0) >= 400
+    AND is_count_tokens = FALSE  -- 排除 count_tokens 请求的错误
  GROUP BY 1
 ),
 combined AS (
--- a/backend/internal/service/antigravity_gateway_service.go
+++ b/backend/internal/service/antigravity_gateway_service.go
@@ -523,6 +523,9 @@ func (s *AntigravityGatewayService) Forward(ctx context.Context, c *gin.Context,
 		proxyURL = account.Proxy.URL()
 	}

+	// Sanitize thinking blocks (clean cache_control and flatten history thinking)
+	sanitizeThinkingBlocks(&claudeReq)
+
 	// 获取转换选项
 	// Antigravity 上游要求必须包含身份提示词，否则会返回 429
 	transformOpts := s.getClaudeTransformOptions(ctx)
@@ -534,6 +537,9 @@ func (s *AntigravityGatewayService) Forward(ctx context.Context, c *gin.Context,
 		return nil, fmt.Errorf("transform request: %w", err)
 	}

+	// Safety net: ensure no cache_control leaked into Gemini request
+	geminiBody = cleanCacheControlFromGeminiJSON(geminiBody)
+
 	// Antigravity 上游只支持流式请求，统一使用 streamGenerateContent
 	// 如果客户端请求非流式，在响应处理阶段会收集完整流式响应后转换返回
 	action := "streamGenerateContent"
@@ -903,6 +909,143 @@ func extractAntigravityErrorMessage(body []byte) string {
 	return ""
 }

+// cleanCacheControlFromGeminiJSON removes cache_control from Gemini JSON (emergency fix)
+// This should not be needed if transformation is correct, but serves as a safety net
+func cleanCacheControlFromGeminiJSON(body []byte) []byte {
+	// Try a more robust approach: parse and clean
+	var data map[string]any
+	if err := json.Unmarshal(body, &data); err != nil {
+		log.Printf("[Antigravity] Failed to parse Gemini JSON for cache_control cleaning: %v", err)
+		return body
+	}
+
+	cleaned := removeCacheControlFromAny(data)
+	if !cleaned {
+		return body
+	}
+
+	if result, err := json.Marshal(data); err == nil {
+		log.Printf("[Antigravity] Successfully cleaned cache_control from Gemini JSON")
+		return result
+	}
+
+	return body
+}
+
+// removeCacheControlFromAny recursively removes cache_control fields
+func removeCacheControlFromAny(v any) bool {
+	cleaned := false
+
+	switch val := v.(type) {
+	case map[string]any:
+		for k, child := range val {
+			if k == "cache_control" {
+				delete(val, k)
+				cleaned = true
+			} else if removeCacheControlFromAny(child) {
+				cleaned = true
+			}
+		}
+	case []any:
+		for _, item := range val {
+			if removeCacheControlFromAny(item) {
+				cleaned = true
+			}
+		}
+	}
+
+	return cleaned
+}
+
+// sanitizeThinkingBlocks cleans cache_control and flattens history thinking blocks
+// Thinking blocks do NOT support cache_control field (Anthropic API/Vertex AI requirement)
+// Additionally, history thinking blocks are flattened to text to avoid upstream validation errors
+func sanitizeThinkingBlocks(req *antigravity.ClaudeRequest) {
+	if req == nil {
+		return
+	}
+
+	log.Printf("[Antigravity] sanitizeThinkingBlocks: processing request with %d messages", len(req.Messages))
+
+	// Clean system blocks
+	if len(req.System) > 0 {
+		var systemBlocks []map[string]any
+		if err := json.Unmarshal(req.System, &systemBlocks); err == nil {
+			for i := range systemBlocks {
+				if blockType, _ := systemBlocks[i]["type"].(string); blockType == "thinking" || systemBlocks[i]["thinking"] != nil {
+					if removeCacheControlFromAny(systemBlocks[i]) {
+						log.Printf("[Antigravity] Deep cleaned cache_control from thinking block in system[%d]", i)
+					}
+				}
+			}
+			// Marshal back
+			if cleaned, err := json.Marshal(systemBlocks); err == nil {
+				req.System = cleaned
+			}
+		}
+	}
+
+	// Clean message content blocks and flatten history
+	lastMsgIdx := len(req.Messages) - 1
+	for msgIdx := range req.Messages {
+		raw := req.Messages[msgIdx].Content
+		if len(raw) == 0 {
+			continue
+		}
+
+		// Try to parse as blocks array
+		var blocks []map[string]any
+		if err := json.Unmarshal(raw, &blocks); err != nil {
+			continue
+		}
+
+		cleaned := false
+		for blockIdx := range blocks {
+			blockType, _ := blocks[blockIdx]["type"].(string)
+
+			// Check for thinking blocks (typed or untyped)
+			if blockType == "thinking" || blocks[blockIdx]["thinking"] != nil {
+				// 1. Clean cache_control
+				if removeCacheControlFromAny(blocks[blockIdx]) {
+					log.Printf("[Antigravity] Deep cleaned cache_control from thinking block in messages[%d].content[%d]", msgIdx, blockIdx)
+					cleaned = true
+				}
+
+				// 2. Flatten to text if it's a history message (not the last one)
+				if msgIdx < lastMsgIdx {
+					log.Printf("[Antigravity] Flattening history thinking block to text at messages[%d].content[%d]", msgIdx, blockIdx)
+
+					// Extract thinking content
+					var textContent string
+					if t, ok := blocks[blockIdx]["thinking"].(string); ok {
+						textContent = t
+					} else {
+						// Fallback for non-string content (marshal it)
+						if b, err := json.Marshal(blocks[blockIdx]["thinking"]); err == nil {
+							textContent = string(b)
+						}
+					}
+
+					// Convert to text block
+					blocks[blockIdx]["type"] = "text"
+					blocks[blockIdx]["text"] = textContent
+					delete(blocks[blockIdx], "thinking")
+					delete(blocks[blockIdx], "signature")
+					delete(blocks[blockIdx], "cache_control") // Ensure it's gone
+					cleaned = true
+				}
+			}
+		}
+
+		// Marshal back if modified
+		if cleaned {
+			if marshaled, err := json.Marshal(blocks); err == nil {
+				req.Messages[msgIdx].Content = marshaled
+			}
+		}
+	}
+}
+
 // stripThinkingFromClaudeRequest converts thinking blocks to text blocks in a Claude Messages request.
 // This preserves the thinking content while avoiding signature validation errors.
 // Note: redacted_thinking blocks are removed because they cannot be converted to text.
--- a/backend/internal/service/gateway_service.go
+++ b/backend/internal/service/gateway_service.go
@@ -1227,6 +1227,9 @@ func enforceCacheControlLimit(body []byte) []byte {
 		return body
 	}

+	// 清理 thinking 块中的非法 cache_control（thinking 块不支持该字段）
+	removeCacheControlFromThinkingBlocks(data)
+
 	// 计算当前 cache_control 块数量
 	count := countCacheControlBlocks(data)
 	if count <= maxCacheControlBlocks {
@@ -1254,6 +1257,7 @@ func enforceCacheControlLimit(body []byte) []byte {
 }

 // countCacheControlBlocks 统计 system 和 messages 中的 cache_control 块数量
+// 注意：thinking 块不支持 cache_control，统计时跳过
 func countCacheControlBlocks(data map[string]any) int {
 	count := 0

@@ -1261,6 +1265,10 @@ func countCacheControlBlocks(data map[string]any) int {
 	if system, ok := data["system"].([]any); ok {
 		for _, item := range system {
 			if m, ok := item.(map[string]any); ok {
+				// thinking 块不支持 cache_control，跳过
+				if blockType, _ := m["type"].(string); blockType == "thinking" {
+					continue
+				}
 				if _, has := m["cache_control"]; has {
 					count++
 				}
@@ -1275,6 +1283,10 @@ func countCacheControlBlocks(data map[string]any) int {
 				if content, ok := msgMap["content"].([]any); ok {
 					for _, item := range content {
 						if m, ok := item.(map[string]any); ok {
+							// thinking 块不支持 cache_control，跳过
+							if blockType, _ := m["type"].(string); blockType == "thinking" {
+								continue
+							}
 							if _, has := m["cache_control"]; has {
 								count++
 							}
@@ -1290,6 +1302,7 @@ func countCacheControlBlocks(data map[string]any) int {

 // removeCacheControlFromMessages 从 messages 中移除一个 cache_control（从头开始）
 // 返回 true 表示成功移除，false 表示没有可移除的
+// 注意：跳过 thinking 块（它不支持 cache_control）
 func removeCacheControlFromMessages(data map[string]any) bool {
 	messages, ok := data["messages"].([]any)
 	if !ok {
@@ -1307,6 +1320,10 @@ func removeCacheControlFromMessages(data map[string]any) bool {
 		}
 		for _, item := range content {
 			if m, ok := item.(map[string]any); ok {
+				// thinking 块不支持 cache_control，跳过
+				if blockType, _ := m["type"].(string); blockType == "thinking" {
+					continue
+				}
 				if _, has := m["cache_control"]; has {
 					delete(m, "cache_control")
 					return true
@@ -1319,6 +1336,7 @@ func removeCacheControlFromMessages(data map[string]any) bool {

 // removeCacheControlFromSystem 从 system 中移除一个 cache_control（从尾部开始，保护注入的 prompt）
 // 返回 true 表示成功移除，false 表示没有可移除的
+// 注意：跳过 thinking 块（它不支持 cache_control）
 func removeCacheControlFromSystem(data map[string]any) bool {
 	system, ok := data["system"].([]any)
 	if !ok {
@@ -1328,6 +1346,10 @@ func removeCacheControlFromSystem(data map[string]any) bool {
 	// 从尾部开始移除，保护开头注入的 Claude Code prompt
 	for i := len(system) - 1; i >= 0; i-- {
 		if m, ok := system[i].(map[string]any); ok {
+			// thinking 块不支持 cache_control，跳过
+			if blockType, _ := m["type"].(string); blockType == "thinking" {
+				continue
+			}
 			if _, has := m["cache_control"]; has {
 				delete(m, "cache_control")
 				return true
@@ -1337,6 +1359,44 @@ func removeCacheControlFromSystem(data map[string]any) bool {
 	return false
 }

+// removeCacheControlFromThinkingBlocks 强制清理所有 thinking 块中的非法 cache_control
+// thinking 块不支持 cache_control 字段，这个函数确保所有 thinking 块都不含该字段
+func removeCacheControlFromThinkingBlocks(data map[string]any) {
+	// 清理 system 中的 thinking 块
+	if system, ok := data["system"].([]any); ok {
+		for _, item := range system {
+			if m, ok := item.(map[string]any); ok {
+				if blockType, _ := m["type"].(string); blockType == "thinking" {
+					if _, has := m["cache_control"]; has {
+						delete(m, "cache_control")
+						log.Printf("[Warning] Removed illegal cache_control from thinking block in system")
+					}
+				}
+			}
+		}
+	}
+
+	// 清理 messages 中的 thinking 块
+	if messages, ok := data["messages"].([]any); ok {
+		for msgIdx, msg := range messages {
+			if msgMap, ok := msg.(map[string]any); ok {
+				if content, ok := msgMap["content"].([]any); ok {
+					for contentIdx, item := range content {
+						if m, ok := item.(map[string]any); ok {
+							if blockType, _ := m["type"].(string); blockType == "thinking" {
+								if _, has := m["cache_control"]; has {
+									delete(m, "cache_control")
+									log.Printf("[Warning] Removed illegal cache_control from thinking block in messages[%d].content[%d]", msgIdx, contentIdx)
+								}
+							}
+						}
+					}
+				}
+			}
+		}
+	}
+}
+
 // Forward 转发请求到Claude API
 func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *Account, parsed *ParsedRequest) (*ForwardResult, error) {
 	startTime := time.Now()
--- a/backend/internal/service/openai_gateway_service.go
+++ b/backend/internal/service/openai_gateway_service.go
@@ -545,14 +545,12 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco

 	isCodexCLI := openai.IsCodexCLIRequest(c.GetHeader("User-Agent"))

-	// Apply model mapping (skip for Codex CLI for transparent forwarding)
-	mappedModel := reqModel
-	if !isCodexCLI {
-		mappedModel = account.GetMappedModel(reqModel)
-		if mappedModel != reqModel {
-			reqBody["model"] = mappedModel
-			bodyModified = true
-		}
+	// Apply model mapping for all requests (including Codex CLI)
+	mappedModel := account.GetMappedModel(reqModel)
+	if mappedModel != reqModel {
+		log.Printf("[OpenAI] Model mapping applied: %s -> %s (account: %s, isCodexCLI: %v)", reqModel, mappedModel, account.Name, isCodexCLI)
+		reqBody["model"] = mappedModel
+		bodyModified = true
 	}

 	if account.Type == AccountTypeOAuth && !isCodexCLI {
@@ -568,6 +566,44 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco
 		}
 	}

+	// Handle max_output_tokens based on platform and account type
+	if !isCodexCLI {
+		if maxOutputTokens, hasMaxOutputTokens := reqBody["max_output_tokens"]; hasMaxOutputTokens {
+			switch account.Platform {
+			case PlatformOpenAI:
+				// For OpenAI API Key, remove max_output_tokens (not supported)
+				// For OpenAI OAuth (Responses API), keep it (supported)
+				if account.Type == AccountTypeAPIKey {
+					delete(reqBody, "max_output_tokens")
+					bodyModified = true
+				}
+			case PlatformAnthropic:
+				// For Anthropic (Claude), convert to max_tokens
+				delete(reqBody, "max_output_tokens")
+				if _, hasMaxTokens := reqBody["max_tokens"]; !hasMaxTokens {
+					reqBody["max_tokens"] = maxOutputTokens
+				}
+				bodyModified = true
+			case PlatformGemini:
+				// For Gemini, remove (will be handled by Gemini-specific transform)
+				delete(reqBody, "max_output_tokens")
+				bodyModified = true
+			default:
+				// For unknown platforms, remove to be safe
+				delete(reqBody, "max_output_tokens")
+				bodyModified = true
+			}
+		}
+
+		// Also handle max_completion_tokens (similar logic)
+		if _, hasMaxCompletionTokens := reqBody["max_completion_tokens"]; hasMaxCompletionTokens {
+			if account.Type == AccountTypeAPIKey || account.Platform != PlatformOpenAI {
+				delete(reqBody, "max_completion_tokens")
+				bodyModified = true
+			}
+		}
+	}
+
 	// Re-serialize body only if modified
 	if bodyModified {
 		var err error
--- a/backend/internal/service/ops_port.go
+++ b/backend/internal/service/ops_port.go
@@ -73,6 +73,7 @@ type OpsInsertErrorLogInput struct {
 	Severity          string
 	StatusCode        int
 	IsBusinessLimited bool
+	IsCountTokens     bool // 是否为 count_tokens 请求

 	ErrorMessage string
 	ErrorBody    string
--- a/backend/internal/service/ops_settings.go
+++ b/backend/internal/service/ops_settings.go
@@ -368,6 +368,9 @@ func defaultOpsAdvancedSettings() *OpsAdvancedSettings {
 		Aggregation: OpsAggregationSettings{
 			AggregationEnabled: false,
 		},
+		IgnoreCountTokensErrors: false,
+		AutoRefreshEnabled:      false,
+		AutoRefreshIntervalSec:  30,
 	}
 }

@@ -388,6 +391,10 @@ func normalizeOpsAdvancedSettings(cfg *OpsAdvancedSettings) {
 	if cfg.DataRetention.HourlyMetricsRetentionDays <= 0 {
 		cfg.DataRetention.HourlyMetricsRetentionDays = 30
 	}
+	// Normalize auto refresh interval (default 30 seconds)
+	if cfg.AutoRefreshIntervalSec <= 0 {
+		cfg.AutoRefreshIntervalSec = 30
+	}
 }

 func validateOpsAdvancedSettings(cfg *OpsAdvancedSettings) error {
@@ -403,6 +410,9 @@ func validateOpsAdvancedSettings(cfg *OpsAdvancedSettings) error {
 	if cfg.DataRetention.HourlyMetricsRetentionDays < 1 || cfg.DataRetention.HourlyMetricsRetentionDays > 365 {
 		return errors.New("hourly_metrics_retention_days must be between 1 and 365")
 	}
+	if cfg.AutoRefreshIntervalSec < 15 || cfg.AutoRefreshIntervalSec > 300 {
+		return errors.New("auto_refresh_interval_seconds must be between 15 and 300")
+	}
 	return nil
 }

--- a/backend/internal/service/ops_settings_models.go
+++ b/backend/internal/service/ops_settings_models.go
@@ -79,8 +79,11 @@ type OpsAlertRuntimeSettings struct {

 // OpsAdvancedSettings stores advanced ops configuration (data retention, aggregation).
 type OpsAdvancedSettings struct {
-	DataRetention OpsDataRetentionSettings `json:"data_retention"`
-	Aggregation   OpsAggregationSettings   `json:"aggregation"`
+	DataRetention           OpsDataRetentionSettings `json:"data_retention"`
+	Aggregation             OpsAggregationSettings   `json:"aggregation"`
+	IgnoreCountTokensErrors bool                     `json:"ignore_count_tokens_errors"`
+	AutoRefreshEnabled      bool                     `json:"auto_refresh_enabled"`
+	AutoRefreshIntervalSec  int                      `json:"auto_refresh_interval_seconds"`
 }

 type OpsDataRetentionSettings struct {
--- a/backend/migrations/036_ops_error_logs_add_is_count_tokens.sql
+++ b/backend/migrations/036_ops_error_logs_add_is_count_tokens.sql
@@ -0,0 +1,16 @@
+-- Migration: 添加 is_count_tokens 字段到 ops_error_logs 表
+-- Purpose: 标记 count_tokens 请求的错误，以便在统计和告警中根据配置动态过滤
+-- Author: System
+-- Date: 2026-01-12
+
+-- Add is_count_tokens column to ops_error_logs table
+ALTER TABLE ops_error_logs
+ADD COLUMN is_count_tokens BOOLEAN NOT NULL DEFAULT FALSE;
+
+-- Add comment
+COMMENT ON COLUMN ops_error_logs.is_count_tokens IS '是否为 count_tokens 请求的错误（用于统计过滤）';
+
+-- Create index for filtering (optional, improves query performance)
+CREATE INDEX IF NOT EXISTS idx_ops_error_logs_is_count_tokens
+ON ops_error_logs(is_count_tokens)
+WHERE is_count_tokens = TRUE;
--- a/frontend/src/api/admin/ops.ts
+++ b/frontend/src/api/admin/ops.ts
@@ -734,6 +734,9 @@ export interface OpsAlertRuntimeSettings {
 export interface OpsAdvancedSettings {
  data_retention: OpsDataRetentionSettings
  aggregation: OpsAggregationSettings
+  ignore_count_tokens_errors: boolean
+  auto_refresh_enabled: boolean
+  auto_refresh_interval_seconds: number
 }

 export interface OpsDataRetentionSettings {
--- a/frontend/src/components/keys/UseKeyModal.vue
+++ b/frontend/src/components/keys/UseKeyModal.vue
@@ -376,6 +376,10 @@ const currentFiles = computed((): FileConfig[] => {
    const trimmed = `${baseRoot}/antigravity`.replace(/\/+$/, '')
    return trimmed.endsWith('/v1beta') ? trimmed : `${trimmed}/v1beta`
  })()
+  const geminiBase = (() => {
+    const trimmed = baseRoot.replace(/\/+$/, '')
+    return trimmed.endsWith('/v1beta') ? trimmed : `${trimmed}/v1beta`
+  })()

  if (activeClientTab.value === 'opencode') {
    switch (props.platform) {
@@ -384,7 +388,7 @@ const currentFiles = computed((): FileConfig[] => {
      case 'openai':
        return [generateOpenCodeConfig('openai', apiBase, apiKey)]
      case 'gemini':
-        return [generateOpenCodeConfig('gemini', apiBase, apiKey)]
+        return [generateOpenCodeConfig('gemini', geminiBase, apiKey)]
      case 'antigravity':
        return [
          generateOpenCodeConfig('antigravity-claude', antigravityBase, apiKey, 'opencode.json (Claude)'),
@@ -525,14 +529,16 @@ function generateOpenCodeConfig(platform: string, baseUrl: string, apiKey: strin
    [platform]: {
      options: {
        baseURL: baseUrl,
-        apiKey,
-        ...(platform === 'openai' ? { store: false } : {})
+        apiKey
      }
    }
  }
  const openaiModels = {
    'gpt-5.2-codex': {
      name: 'GPT-5.2 Codex',
+      options: {
+        store: false
+      },
      variants: {
        low: {},
        medium: {},
@@ -574,9 +580,26 @@ function generateOpenCodeConfig(platform: string, baseUrl: string, apiKey: strin
    provider[platform].models = openaiModels
  }

+  const agent =
+    platform === 'openai'
+      ? {
+          build: {
+            options: {
+              store: false
+            }
+          },
+          plan: {
+            options: {
+              store: false
+            }
+          }
+        }
+      : undefined
+
  const content = JSON.stringify(
    {
      provider,
+      ...(agent ? { agent } : {}),
      $schema: 'https://opencode.ai/config.json'
    },
    null,
--- a/frontend/src/composables/useModelWhitelist.ts
+++ b/frontend/src/composables/useModelWhitelist.ts
@@ -13,7 +13,17 @@ const openaiModels = [
  'o1', 'o1-preview', 'o1-mini', 'o1-pro',
  'o3', 'o3-mini', 'o3-pro',
  'o4-mini',
-  'gpt-5', 'gpt-5-mini', 'gpt-5-nano',
+  // GPT-5 系列（同步后端定价文件）
+  'gpt-5', 'gpt-5-2025-08-07', 'gpt-5-chat', 'gpt-5-chat-latest',
+  'gpt-5-codex', 'gpt-5-pro', 'gpt-5-pro-2025-10-06',
+  'gpt-5-mini', 'gpt-5-mini-2025-08-07',
+  'gpt-5-nano', 'gpt-5-nano-2025-08-07',
+  // GPT-5.1 系列
+  'gpt-5.1', 'gpt-5.1-2025-11-13', 'gpt-5.1-chat-latest',
+  'gpt-5.1-codex', 'gpt-5.1-codex-max', 'gpt-5.1-codex-mini',
+  // GPT-5.2 系列
+  'gpt-5.2', 'gpt-5.2-2025-12-11', 'gpt-5.2-chat-latest',
+  'gpt-5.2-codex', 'gpt-5.2-pro', 'gpt-5.2-pro-2025-12-11',
  'chatgpt-4o-latest',
  'gpt-4o-audio-preview', 'gpt-4o-realtime-preview'
 ]
@@ -211,7 +221,10 @@ const openaiPresetMappings = [
  { label: 'GPT-4.1', from: 'gpt-4.1', to: 'gpt-4.1', color: 'bg-indigo-100 text-indigo-700 hover:bg-indigo-200 dark:bg-indigo-900/30 dark:text-indigo-400' },
  { label: 'o1', from: 'o1', to: 'o1', color: 'bg-purple-100 text-purple-700 hover:bg-purple-200 dark:bg-purple-900/30 dark:text-purple-400' },
  { label: 'o3', from: 'o3', to: 'o3', color: 'bg-emerald-100 text-emerald-700 hover:bg-emerald-200 dark:bg-emerald-900/30 dark:text-emerald-400' },
-  { label: 'GPT-5', from: 'gpt-5', to: 'gpt-5', color: 'bg-amber-100 text-amber-700 hover:bg-amber-200 dark:bg-amber-900/30 dark:text-amber-400' }
+  { label: 'GPT-5', from: 'gpt-5', to: 'gpt-5', color: 'bg-amber-100 text-amber-700 hover:bg-amber-200 dark:bg-amber-900/30 dark:text-amber-400' },
+  { label: 'GPT-5.1', from: 'gpt-5.1', to: 'gpt-5.1', color: 'bg-orange-100 text-orange-700 hover:bg-orange-200 dark:bg-orange-900/30 dark:text-orange-400' },
+  { label: 'GPT-5.2', from: 'gpt-5.2', to: 'gpt-5.2', color: 'bg-red-100 text-red-700 hover:bg-red-200 dark:bg-red-900/30 dark:text-red-400' },
+  { label: 'GPT-5.1 Codex', from: 'gpt-5.1-codex', to: 'gpt-5.1-codex', color: 'bg-cyan-100 text-cyan-700 hover:bg-cyan-200 dark:bg-cyan-900/30 dark:text-cyan-400' }
 ]

 const geminiPresetMappings = [
--- a/frontend/src/i18n/locales/en.ts
+++ b/frontend/src/i18n/locales/en.ts
@@ -390,7 +390,7 @@ export default {
      opencode: {
        title: 'OpenCode Example',
        subtitle: 'opencode.json',
-        hint: 'This is a group configuration example. Adjust model and options as needed.',
+        hint: 'Config path: ~/.config/opencode/opencode.json (or opencode.jsonc), create if not exists. Use default providers (openai/anthropic/google) or custom provider_id. API Key can be configured directly or via /connect command. This is an example, adjust models and options as needed.',
      },
    },
    customKeyLabel: 'Custom Key',
@@ -1943,6 +1943,9 @@ export default {
        '6h': 'Last 6 hours',
        '24h': 'Last 24 hours'
      },
+      fullscreen: {
+        enter: 'Enter Fullscreen'
+      },
      diagnosis: {
        title: 'Smart Diagnosis',
        footer: 'Automated diagnostic suggestions based on current metrics',
--- a/frontend/src/i18n/locales/zh.ts
+++ b/frontend/src/i18n/locales/zh.ts
@@ -387,7 +387,7 @@ export default {
      opencode: {
        title: 'OpenCode 配置示例',
        subtitle: 'opencode.json',
-        hint: '示例仅用于演示分组配置，模型与选项可按需调整。',
+        hint: '配置文件路径：~/.config/opencode/opencode.json（或 opencode.jsonc），不存在需手动创建。可使用默认 provider（openai/anthropic/google）或自定义 provider_id。API Key 支持直接配置或通过客户端 /connect 命令配置。示例仅供参考，模型与选项可按需调整。',
      },
    },
    customKeyLabel: '自定义密钥',
@@ -2088,6 +2088,9 @@ export default {
        '6h': '近6小时',
        '24h': '近24小时'
      },
+      fullscreen: {
+        enter: '进入全屏'
+      },
      diagnosis: {
        title: '智能诊断',
        footer: '基于当前指标的自动诊断建议',
--- a/frontend/src/style.css
+++ b/frontend/src/style.css
@@ -19,7 +19,22 @@
    @apply min-h-screen;
  }

-  /* 自定义滚动条 */
+  /* 自定义滚动条 - 默认隐藏，悬停或滚动时显示 */
+  * {
+    scrollbar-width: thin;
+    scrollbar-color: transparent transparent;
+  }
+
+  *:hover,
+  *:focus-within {
+    scrollbar-color: rgba(156, 163, 175, 0.5) transparent;
+  }
+
+  .dark *:hover,
+  .dark *:focus-within {
+    scrollbar-color: rgba(75, 85, 99, 0.5) transparent;
+  }
+
  ::-webkit-scrollbar {
    @apply h-2 w-2;
  }
@@ -29,10 +44,15 @@
  }

  ::-webkit-scrollbar-thumb {
-    @apply rounded-full bg-gray-300 dark:bg-dark-600;
+    @apply rounded-full bg-transparent;
+    transition: background-color 0.2s ease;
  }

-  ::-webkit-scrollbar-thumb:hover {
+  *:hover::-webkit-scrollbar-thumb {
+    @apply bg-gray-300/50 dark:bg-dark-600/50;
+  }
+
+  *:hover::-webkit-scrollbar-thumb:hover {
    @apply bg-gray-400 dark:bg-dark-500;
  }

--- a/frontend/src/views/admin/ops/OpsDashboard.vue
+++ b/frontend/src/views/admin/ops/OpsDashboard.vue
@@ -1,6 +1,6 @@
 <template>
-  <AppLayout>
-    <div class="space-y-6 pb-12">
+  <component :is="isFullscreen ? 'div' : AppLayout" :class="isFullscreen ? 'flex min-h-screen flex-col justify-center bg-gray-50 dark:bg-dark-950' : ''">
+    <div :class="[isFullscreen ? 'p-4 md:p-6' : '', 'space-y-6 pb-12']">
      <div
        v-if="errorMessage"
        class="rounded-2xl bg-red-50 p-4 text-sm text-red-600 dark:bg-red-900/20 dark:text-red-400"
@@ -20,6 +20,9 @@
        :loading="loading"
        :last-updated="lastUpdated"
        :thresholds="metricThresholds"
+        :auto-refresh-enabled="autoRefreshEnabled"
+        :auto-refresh-countdown="autoRefreshCountdown"
+        :fullscreen="isFullscreen"
        @update:time-range="onTimeRangeChange"
        @update:platform="onPlatformChange"
        @update:group="onGroupChange"
@@ -29,6 +32,8 @@
        @open-error-details="openErrorDetails"
        @open-settings="showSettingsDialog = true"
        @open-alert-rules="showAlertRulesCard = true"
+        @enter-fullscreen="enterFullscreen"
+        @exit-fullscreen="exitFullscreen"
      />

      <!-- Row: Concurrency + Throughput -->
@@ -43,6 +48,7 @@
            :top-groups="throughputTrend?.top_groups ?? []"
            :loading="loadingTrend"
            :time-range="timeRange"
+            :fullscreen="isFullscreen"
            @select-platform="handleThroughputSelectPlatform"
            @select-group="handleThroughputSelectGroup"
            @open-details="handleOpenRequestDetails"
@@ -70,41 +76,42 @@
      <!-- Alert Events -->
      <OpsAlertEventsCard v-if="opsEnabled && !(loading && !hasLoadedOnce)" />

-      <!-- Settings Dialog -->
-      <OpsSettingsDialog :show="showSettingsDialog" @close="showSettingsDialog = false" @saved="onSettingsSaved" />
+      <!-- Settings Dialog (hidden in fullscreen mode) -->
+      <template v-if="!isFullscreen">
+        <OpsSettingsDialog :show="showSettingsDialog" @close="showSettingsDialog = false" @saved="onSettingsSaved" />

-      <!-- Alert Rules Dialog -->
-      <BaseDialog :show="showAlertRulesCard" :title="t('admin.ops.alertRules.title')" width="extra-wide" @close="showAlertRulesCard = false">
-        <OpsAlertRulesCard />
-      </BaseDialog>
+        <BaseDialog :show="showAlertRulesCard" :title="t('admin.ops.alertRules.title')" width="extra-wide" @close="showAlertRulesCard = false">
+          <OpsAlertRulesCard />
+        </BaseDialog>

-      <OpsErrorDetailsModal
-        :show="showErrorDetails"
-        :time-range="timeRange"
-        :platform="platform"
-        :group-id="groupId"
-        :error-type="errorDetailsType"
-        @update:show="showErrorDetails = $event"
-        @openErrorDetail="openError"
-      />
+        <OpsErrorDetailsModal
+          :show="showErrorDetails"
+          :time-range="timeRange"
+          :platform="platform"
+          :group-id="groupId"
+          :error-type="errorDetailsType"
+          @update:show="showErrorDetails = $event"
+          @openErrorDetail="openError"
+        />

-      <OpsErrorDetailModal v-model:show="showErrorModal" :error-id="selectedErrorId" />
+        <OpsErrorDetailModal v-model:show="showErrorModal" :error-id="selectedErrorId" />

-      <OpsRequestDetailsModal
-        v-model="showRequestDetails"
-        :time-range="timeRange"
-        :preset="requestDetailsPreset"
-        :platform="platform"
-        :group-id="groupId"
-        @openErrorDetail="openError"
-      />
+        <OpsRequestDetailsModal
+          v-model="showRequestDetails"
+          :time-range="timeRange"
+          :preset="requestDetailsPreset"
+          :platform="platform"
+          :group-id="groupId"
+          @openErrorDetail="openError"
+        />
+      </template>
    </div>
-  </AppLayout>
+  </component>
 </template>

 <script setup lang="ts">
 import { computed, onMounted, onUnmounted, ref, watch } from 'vue'
-import { useDebounceFn } from '@vueuse/core'
+import { useDebounceFn, useIntervalFn } from '@vueuse/core'
 import { useI18n } from 'vue-i18n'
 import { useRoute, useRouter } from 'vue-router'
 import AppLayout from '@/components/layout/AppLayout.vue'
@@ -161,12 +168,36 @@ const QUERY_KEYS = {
  timeRange: 'tr',
  platform: 'platform',
  groupId: 'group_id',
-  queryMode: 'mode'
+  queryMode: 'mode',
+  fullscreen: 'fullscreen'
 } as const

 const isApplyingRouteQuery = ref(false)
 const isSyncingRouteQuery = ref(false)

+// Fullscreen mode
+const isFullscreen = computed(() => {
+  const val = route.query[QUERY_KEYS.fullscreen]
+  return val === '1' || val === 'true'
+})
+
+function exitFullscreen() {
+  const nextQuery = { ...route.query }
+  delete nextQuery[QUERY_KEYS.fullscreen]
+  router.replace({ query: nextQuery })
+}
+
+function enterFullscreen() {
+  const nextQuery = { ...route.query, [QUERY_KEYS.fullscreen]: '1' }
+  router.replace({ query: nextQuery })
+}
+
+function handleKeydown(e: KeyboardEvent) {
+  if (e.key === 'Escape' && isFullscreen.value) {
+    exitFullscreen()
+  }
+}
+
 let dashboardFetchController: AbortController | null = null
 let dashboardFetchSeq = 0

@@ -287,6 +318,45 @@ const requestDetailsPreset = ref<OpsRequestDetailsPreset>({
 const showSettingsDialog = ref(false)
 const showAlertRulesCard = ref(false)

+// Auto refresh settings
+const autoRefreshEnabled = ref(false)
+const autoRefreshIntervalMs = ref(30000) // default 30 seconds
+const autoRefreshCountdown = ref(0)
+
+// Auto refresh timer
+const { pause: pauseAutoRefresh, resume: resumeAutoRefresh } = useIntervalFn(
+  () => {
+    if (autoRefreshEnabled.value && opsEnabled.value && !loading.value) {
+      fetchData()
+    }
+  },
+  autoRefreshIntervalMs,
+  { immediate: false }
+)
+
+// Countdown timer (updates every second)
+const { pause: pauseCountdown, resume: resumeCountdown } = useIntervalFn(
+  () => {
+    if (autoRefreshEnabled.value && autoRefreshCountdown.value > 0) {
+      autoRefreshCountdown.value--
+    }
+  },
+  1000,
+  { immediate: false }
+)
+
+// Load auto refresh settings from backend
+async function loadAutoRefreshSettings() {
+  try {
+    const settings = await opsAPI.getAdvancedSettings()
+    autoRefreshEnabled.value = settings.auto_refresh_enabled
+    autoRefreshIntervalMs.value = settings.auto_refresh_interval_seconds * 1000
+    autoRefreshCountdown.value = settings.auto_refresh_interval_seconds
+  } catch (err) {
+    console.error('[OpsDashboard] Failed to load auto refresh settings', err)
+  }
+}
+
 function handleThroughputSelectPlatform(nextPlatform: string) {
  platform.value = nextPlatform || ''
  groupId.value = null
@@ -510,6 +580,10 @@ async function fetchData() {
    ])
    if (fetchSeq !== dashboardFetchSeq) return
    lastUpdated.value = new Date()
+    // Reset auto refresh countdown after successful fetch
+    if (autoRefreshEnabled.value) {
+      autoRefreshCountdown.value = Math.floor(autoRefreshIntervalMs.value / 1000)
+    }
  } catch (err) {
    if (!isOpsDisabledError(err)) {
      console.error('[ops] failed to fetch dashboard data', err)
@@ -558,6 +632,9 @@ watch(
 )

 onMounted(async () => {
+  // Fullscreen mode: listen for ESC key
+  window.addEventListener('keydown', handleKeydown)
+
  await adminSettingsStore.fetch()
  if (!adminSettingsStore.opsMonitoringEnabled) {
    await router.replace('/admin/settings')
@@ -567,9 +644,18 @@ onMounted(async () => {
  // Load thresholds configuration
  loadThresholds()

+  // Load auto refresh settings
+  await loadAutoRefreshSettings()
+
  if (opsEnabled.value) {
    await fetchData()
  }
+
+  // Start auto refresh if enabled
+  if (autoRefreshEnabled.value) {
+    resumeAutoRefresh()
+    resumeCountdown()
+  }
 })

 async function loadThresholds() {
@@ -583,6 +669,29 @@ async function loadThresholds() {
 }

 onUnmounted(() => {
+  window.removeEventListener('keydown', handleKeydown)
  abortDashboardFetch()
+  pauseAutoRefresh()
+  pauseCountdown()
+})
+
+// Watch auto refresh settings changes
+watch(autoRefreshEnabled, (enabled) => {
+  if (enabled) {
+    autoRefreshCountdown.value = Math.floor(autoRefreshIntervalMs.value / 1000)
+    resumeAutoRefresh()
+    resumeCountdown()
+  } else {
+    pauseAutoRefresh()
+    pauseCountdown()
+    autoRefreshCountdown.value = 0
+  }
+})
+
+// Reload auto refresh settings after settings dialog is closed
+watch(showSettingsDialog, async (show) => {
+  if (!show) {
+    await loadAutoRefreshSettings()
+  }
 })
 </script>
--- a/frontend/src/views/admin/ops/components/OpsDashboardHeader.vue
+++ b/frontend/src/views/admin/ops/components/OpsDashboardHeader.vue
@@ -23,6 +23,9 @@ interface Props {
  loading: boolean
  lastUpdated: Date | null
  thresholds?: OpsMetricThresholds | null // 阈值配置
+  autoRefreshEnabled?: boolean
+  autoRefreshCountdown?: number
+  fullscreen?: boolean
 }

 interface Emits {
@@ -35,6 +38,8 @@ interface Emits {
  (e: 'openErrorDetails', kind: 'request' | 'upstream'): void
  (e: 'openSettings'): void
  (e: 'openAlertRules'): void
+  (e: 'enterFullscreen'): void
+  (e: 'exitFullscreen'): void
 }

 const props = defineProps<Props>()
@@ -389,15 +394,15 @@ const healthScoreClass = computed(() => {
  return 'text-red-500'
 })

-const circleSize = 100
-const strokeWidth = 8
-const radius = (circleSize - strokeWidth) / 2
-const circumference = 2 * Math.PI * radius
+const circleSize = computed(() => props.fullscreen ? 140 : 100)
+const strokeWidth = computed(() => props.fullscreen ? 10 : 8)
+const radius = computed(() => (circleSize.value - strokeWidth.value) / 2)
+const circumference = computed(() => 2 * Math.PI * radius.value)
 const dashOffset = computed(() => {
  if (isSystemIdle.value) return 0
  if (healthScoreValue.value == null) return 0
  const score = Math.max(0, Math.min(100, healthScoreValue.value))
-  return circumference - (score / 100) * circumference
+  return circumference.value - (score / 100) * circumference.value
 })

 interface DiagnosisItem {
@@ -812,7 +817,7 @@ function handleToolbarRefresh() {
 </script>

 <template>
-  <div class="flex flex-col gap-4 rounded-3xl bg-white p-6 shadow-sm ring-1 ring-gray-900/5 dark:bg-dark-800 dark:ring-dark-700">
+  <div :class="['flex flex-col gap-4 rounded-3xl bg-white shadow-sm ring-1 ring-gray-900/5 dark:bg-dark-800 dark:ring-dark-700', props.fullscreen ? 'p-8' : 'p-6']">
    <!-- Top Toolbar -->
    <div class="flex flex-wrap items-center justify-between gap-4 border-b border-gray-100 pb-4 dark:border-dark-700">
      <div>
@@ -828,7 +833,7 @@ function handleToolbarRefresh() {
          {{ t('admin.ops.title') }}
        </h1>

-        <div class="mt-1 flex items-center gap-3 text-xs text-gray-500 dark:text-gray-400">
+        <div v-if="!props.fullscreen" class="mt-1 flex items-center gap-3 text-xs text-gray-500 dark:text-gray-400">
          <span class="flex items-center gap-1.5" :title="props.loading ? t('admin.ops.loadingText') : t('admin.ops.ready')">
            <span class="relative flex h-2 w-2">
              <span class="relative inline-flex h-2 w-2 rounded-full" :class="props.loading ? 'bg-gray-400' : 'bg-green-500'"></span>
@@ -839,6 +844,17 @@ function handleToolbarRefresh() {
          <span>·</span>
          <span>{{ t('common.refresh') }}: {{ updatedAtLabel }}</span>

+          <template v-if="props.autoRefreshEnabled && props.autoRefreshCountdown !== undefined">
+            <span>·</span>
+            <span class="flex items-center gap-1">
+              <svg class="h-3 w-3 animate-spin text-blue-500" fill="none" viewBox="0 0 24 24">
+                <circle class="opacity-25" cx="12" cy="12" r="10" stroke="currentColor" stroke-width="4"></circle>
+                <path class="opacity-75" fill="currentColor" d="M4 12a8 8 0 018-8V0C5.373 0 0 5.373 0 12h4zm2 5.291A7.962 7.962 0 014 12H0c0 3.042 1.135 5.824 3 7.938l3-2.647z"></path>
+              </svg>
+              <span>自动刷新: {{ props.autoRefreshCountdown }}s</span>
+            </span>
+          </template>
+
          <template v-if="systemMetrics">
            <span>·</span>
            <span>
@@ -850,28 +866,30 @@ function handleToolbarRefresh() {
      </div>

      <div class="flex flex-wrap items-center gap-3">
-        <Select
-          :model-value="platform"
-          :options="platformOptions"
-          class="w-full sm:w-[140px]"
-          @update:model-value="handlePlatformChange"
-        />
+        <template v-if="!props.fullscreen">
+          <Select
+            :model-value="platform"
+            :options="platformOptions"
+            class="w-full sm:w-[140px]"
+            @update:model-value="handlePlatformChange"
+          />

-        <Select
-          :model-value="groupId"
-          :options="groupOptions"
-          class="w-full sm:w-[160px]"
-          @update:model-value="handleGroupChange"
-        />
+          <Select
+            :model-value="groupId"
+            :options="groupOptions"
+            class="w-full sm:w-[160px]"
+            @update:model-value="handleGroupChange"
+          />

-        <div class="mx-1 hidden h-4 w-[1px] bg-gray-200 dark:bg-dark-700 sm:block"></div>
+          <div class="mx-1 hidden h-4 w-[1px] bg-gray-200 dark:bg-dark-700 sm:block"></div>

-        <Select
-          :model-value="timeRange"
-          :options="timeRangeOptions"
-          class="relative w-full sm:w-[150px]"
-          @update:model-value="handleTimeRangeChange"
-        />
+          <Select
+            :model-value="timeRange"
+            :options="timeRangeOptions"
+            class="relative w-full sm:w-[150px]"
+            @update:model-value="handleTimeRangeChange"
+          />
+        </template>

        <Select
          v-if="false"
@@ -882,6 +900,7 @@ function handleToolbarRefresh() {
        />

        <button
+          v-if="!props.fullscreen"
          type="button"
          class="flex h-8 w-8 items-center justify-center rounded-lg bg-gray-100 text-gray-500 transition-colors hover:bg-gray-200 dark:bg-dark-700 dark:text-gray-400 dark:hover:bg-dark-600"
          :disabled="loading"
@@ -898,9 +917,11 @@ function handleToolbarRefresh() {
          </svg>
        </button>

-        <div class="mx-1 hidden h-4 w-[1px] bg-gray-200 dark:bg-dark-700 sm:block"></div>
+        <div v-if="!props.fullscreen" class="mx-1 hidden h-4 w-[1px] bg-gray-200 dark:bg-dark-700 sm:block"></div>

+        <!-- Alert Rules Button (hidden in fullscreen) -->
        <button
+          v-if="!props.fullscreen"
          type="button"
          class="flex h-8 items-center gap-1.5 rounded-lg bg-blue-100 px-3 text-xs font-bold text-blue-700 transition-colors hover:bg-blue-200 dark:bg-blue-900/30 dark:text-blue-400 dark:hover:bg-blue-900/50"
          :title="t('admin.ops.alertRules.title')"
@@ -912,7 +933,9 @@ function handleToolbarRefresh() {
          <span class="hidden sm:inline">{{ t('admin.ops.alertRules.manage') }}</span>
        </button>

+        <!-- Settings Button (hidden in fullscreen) -->
        <button
+          v-if="!props.fullscreen"
          type="button"
          class="flex h-8 items-center gap-1.5 rounded-lg bg-gray-100 px-3 text-xs font-bold text-gray-700 transition-colors hover:bg-gray-200 dark:bg-dark-700 dark:text-gray-300 dark:hover:bg-dark-600"
          :title="t('admin.ops.settings.title')"
@@ -924,13 +947,26 @@ function handleToolbarRefresh() {
          </svg>
          <span class="hidden sm:inline">{{ t('common.settings') }}</span>
        </button>
+
+        <!-- Enter Fullscreen Button (hidden in fullscreen mode) -->
+        <button
+          v-if="!props.fullscreen"
+          type="button"
+          class="flex h-8 w-8 items-center justify-center rounded-lg bg-gray-100 text-gray-700 transition-colors hover:bg-gray-200 dark:bg-dark-700 dark:text-gray-300 dark:hover:bg-dark-600"
+          :title="t('admin.ops.fullscreen.enter')"
+          @click="emit('enterFullscreen')"
+        >
+          <svg class="h-4 w-4" fill="none" viewBox="0 0 24 24" stroke="currentColor">
+            <path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M4 8V4m0 0h4M4 4l5 5m11-1V4m0 0h-4m4 0l-5 5M4 16v4m0 0h4m-4 0l5-5m11 5l-5-5m5 5v-4m0 4h-4" />
+          </svg>
+        </button>
      </div>
    </div>

    <div v-if="overview" class="grid grid-cols-1 gap-6 lg:grid-cols-12">
      <!-- Left: Health + Realtime -->
-      <div class="rounded-2xl bg-gray-50 p-4 dark:bg-dark-900 lg:col-span-5">
-        <div class="grid grid-cols-1 gap-6 md:grid-cols-[200px_1fr] md:items-center">
+      <div :class="['rounded-2xl bg-gray-50 dark:bg-dark-900 lg:col-span-5', props.fullscreen ? 'p-6' : 'p-4']">
+        <div class="grid h-full grid-cols-1 gap-6 md:grid-cols-[200px_1fr] md:items-center">
          <!-- 1) Health Score -->
          <div
            class="group relative flex cursor-pointer flex-col items-center justify-center rounded-xl py-2 transition-all hover:bg-white/60 dark:hover:bg-dark-800/60 md:border-r md:border-gray-200 md:pr-6 dark:md:border-dark-700"
@@ -1013,14 +1049,14 @@ function handleToolbarRefresh() {
              </svg>

              <div class="absolute flex flex-col items-center">
-                <span class="text-3xl font-black" :class="healthScoreClass">
+                <span :class="[props.fullscreen ? 'text-5xl' : 'text-3xl', 'font-black', healthScoreClass]">
                  {{ isSystemIdle ? t('admin.ops.idleStatus') : (overview.health_score ?? '--') }}
                </span>
-                <span class="text-[10px] font-bold uppercase tracking-wider text-gray-400">{{ t('admin.ops.health') }}</span>
+                <span :class="[props.fullscreen ? 'text-xs' : 'text-[10px]', 'font-bold uppercase tracking-wider text-gray-400']">{{ t('admin.ops.health') }}</span>
              </div>
            </div>

-            <div class="mt-4 text-center">
+            <div class="mt-4 text-center" v-if="!props.fullscreen">
              <div class="flex items-center justify-center gap-1 text-xs font-medium text-gray-500">
                {{ t('admin.ops.healthCondition') }}
                <HelpTooltip :content="t('admin.ops.healthHelp')" />
@@ -1038,7 +1074,7 @@ function handleToolbarRefresh() {
          </div>

          <!-- 2) Realtime Traffic -->
-          <div class="flex flex-col justify-center py-2">
+          <div class="flex h-full flex-col justify-center py-2">
            <div class="mb-3 flex flex-wrap items-center justify-between gap-2">
              <div class="flex items-center gap-2">
                <div class="relative flex h-3 w-3 shrink-0">
@@ -1046,7 +1082,7 @@ function handleToolbarRefresh() {
                  <span class="relative inline-flex h-3 w-3 rounded-full bg-blue-500"></span>
                </div>
                <h3 class="text-xs font-bold uppercase tracking-wider text-gray-400">{{ t('admin.ops.realtime.title') }}</h3>
-                <HelpTooltip :content="t('admin.ops.tooltips.qps')" />
+                <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.qps')" />
              </div>

              <!-- Time Window Selector -->
@@ -1066,18 +1102,18 @@ function handleToolbarRefresh() {
              </div>
            </div>

-            <div class="space-y-3">
+            <div :class="props.fullscreen ? 'space-y-4' : 'space-y-3'">
              <!-- Row 1: Current -->
              <div>
-                <div class="text-[10px] font-bold uppercase text-gray-400">{{ t('admin.ops.current') }}</div>
+                <div :class="[props.fullscreen ? 'text-xs' : 'text-[10px]', 'font-bold uppercase text-gray-400']">{{ t('admin.ops.current') }}</div>
                <div class="mt-1 flex flex-wrap items-baseline gap-x-4 gap-y-2">
                  <div class="flex items-baseline gap-1.5">
-                    <span class="text-xl font-black text-gray-900 dark:text-white sm:text-2xl">{{ displayRealTimeQps.toFixed(1) }}</span>
-                    <span class="text-xs font-bold text-gray-500">QPS</span>
+                    <span :class="[props.fullscreen ? 'text-4xl' : 'text-xl sm:text-2xl', 'font-black text-gray-900 dark:text-white']">{{ displayRealTimeQps.toFixed(1) }}</span>
+                    <span :class="[props.fullscreen ? 'text-sm' : 'text-xs', 'font-bold text-gray-500']">QPS</span>
                  </div>
                  <div class="flex items-baseline gap-1.5">
-                    <span class="text-xl font-black text-gray-900 dark:text-white sm:text-2xl">{{ displayRealTimeTps.toFixed(1) }}</span>
-                    <span class="text-xs font-bold text-gray-500">TPS</span>
+                    <span :class="[props.fullscreen ? 'text-4xl' : 'text-xl sm:text-2xl', 'font-black text-gray-900 dark:text-white']">{{ displayRealTimeTps.toFixed(1) }}</span>
+                    <span :class="[props.fullscreen ? 'text-sm' : 'text-xs', 'font-bold text-gray-500']">TPS</span>
                  </div>
                </div>
              </div>
@@ -1086,8 +1122,8 @@ function handleToolbarRefresh() {
              <div class="grid grid-cols-2 gap-3">
                <!-- Peak -->
                <div>
-                  <div class="text-[10px] font-bold uppercase text-gray-400">{{ t('admin.ops.peak') }}</div>
-                  <div class="mt-1 space-y-0.5 text-sm font-medium text-gray-600 dark:text-gray-400">
+                  <div :class="[props.fullscreen ? 'text-xs' : 'text-[10px]', 'font-bold uppercase text-gray-400']">{{ t('admin.ops.peak') }}</div>
+                  <div :class="[props.fullscreen ? 'text-base' : 'text-sm', 'mt-1 space-y-0.5 font-medium text-gray-600 dark:text-gray-400']">
                    <div class="flex items-baseline gap-1.5">
                      <span class="font-black text-gray-900 dark:text-white">{{ realtimeQpsPeakLabel }}</span>
                      <span class="text-xs">QPS</span>
@@ -1101,8 +1137,8 @@ function handleToolbarRefresh() {

                <!-- Average -->
                <div>
-                  <div class="text-[10px] font-bold uppercase text-gray-400">{{ t('admin.ops.average') }}</div>
-                  <div class="mt-1 space-y-0.5 text-sm font-medium text-gray-600 dark:text-gray-400">
+                  <div :class="[props.fullscreen ? 'text-xs' : 'text-[10px]', 'font-bold uppercase text-gray-400']">{{ t('admin.ops.average') }}</div>
+                  <div :class="[props.fullscreen ? 'text-base' : 'text-sm', 'mt-1 space-y-0.5 font-medium text-gray-600 dark:text-gray-400']">
                    <div class="flex items-baseline gap-1.5">
                      <span class="font-black text-gray-900 dark:text-white">{{ realtimeQpsAvgLabel }}</span>
                      <span class="text-xs">QPS</span>
@@ -1143,15 +1179,16 @@ function handleToolbarRefresh() {
      </div>

      <!-- Right: 6 cards (3 cols x 2 rows) -->
-      <div class="grid grid-cols-1 gap-4 sm:grid-cols-2 lg:col-span-7 lg:grid-cols-3">
+      <div class="grid h-full grid-cols-1 content-center gap-4 sm:grid-cols-2 lg:col-span-7 lg:grid-cols-3">
        <!-- Card 1: Requests -->
        <div class="rounded-2xl bg-gray-50 p-4 dark:bg-dark-900">
          <div class="flex items-center justify-between">
            <div class="flex items-center gap-1">
              <span class="text-[10px] font-bold uppercase text-gray-400">{{ t('admin.ops.requestsTitle') }}</span>
-              <HelpTooltip :content="t('admin.ops.tooltips.totalRequests')" />
+              <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.totalRequests')" />
            </div>
            <button
+              v-if="!props.fullscreen"
              class="text-[10px] font-bold text-blue-500 hover:underline"
              type="button"
              @click="openDetails({ title: t('admin.ops.requestDetails.title') })"
@@ -1184,10 +1221,11 @@ function handleToolbarRefresh() {
          <div class="flex items-center justify-between">
            <div class="flex items-center gap-2">
              <span class="text-[10px] font-bold uppercase text-gray-400">SLA</span>
-              <HelpTooltip :content="t('admin.ops.tooltips.sla')" />
+              <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.sla')" />
              <span class="h-1.5 w-1.5 rounded-full" :class="isSLABelowThreshold(slaPercent) ? 'bg-red-500' : (slaPercent ?? 0) >= 99.5 ? 'bg-green-500' : 'bg-yellow-500'"></span>
            </div>
            <button
+              v-if="!props.fullscreen"
              class="text-[10px] font-bold text-blue-500 hover:underline"
              type="button"
              @click="openDetails({ title: t('admin.ops.requestDetails.title'), kind: 'error' })"
@@ -1214,9 +1252,10 @@ function handleToolbarRefresh() {
          <div class="flex items-center justify-between">
            <div class="flex items-center gap-1">
              <span class="text-[10px] font-bold uppercase text-gray-400">{{ t('admin.ops.latencyDuration') }}</span>
-              <HelpTooltip :content="t('admin.ops.tooltips.latency')" />
+              <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.latency')" />
            </div>
            <button
+              v-if="!props.fullscreen"
              class="text-[10px] font-bold text-blue-500 hover:underline"
              type="button"
              @click="openDetails({ title: t('admin.ops.latencyDuration'), sort: 'duration_desc', min_duration_ms: Math.max(Number(durationP99Ms ?? 0), 0) })"
@@ -1264,9 +1303,10 @@ function handleToolbarRefresh() {
          <div class="flex items-center justify-between">
            <div class="flex items-center gap-1">
              <span class="text-[10px] font-bold uppercase text-gray-400">TTFT</span>
-              <HelpTooltip :content="t('admin.ops.tooltips.ttft')" />
+              <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.ttft')" />
            </div>
            <button
+              v-if="!props.fullscreen"
              class="text-[10px] font-bold text-blue-500 hover:underline"
              type="button"
              @click="openDetails({ title: 'TTFT', sort: 'duration_desc' })"
@@ -1314,9 +1354,9 @@ function handleToolbarRefresh() {
          <div class="flex items-center justify-between">
            <div class="flex items-center gap-1">
              <span class="text-[10px] font-bold uppercase text-gray-400">{{ t('admin.ops.requestErrors') }}</span>
-              <HelpTooltip :content="t('admin.ops.tooltips.errors')" />
+              <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.errors')" />
            </div>
-            <button class="text-[10px] font-bold text-blue-500 hover:underline" type="button" @click="openErrorDetails('request')">
+            <button v-if="!props.fullscreen" class="text-[10px] font-bold text-blue-500 hover:underline" type="button" @click="openErrorDetails('request')">
              {{ t('admin.ops.requestDetails.details') }}
            </button>
          </div>
@@ -1340,9 +1380,9 @@ function handleToolbarRefresh() {
          <div class="flex items-center justify-between">
            <div class="flex items-center gap-1">
              <span class="text-[10px] font-bold uppercase text-gray-400">{{ t('admin.ops.upstreamErrors') }}</span>
-              <HelpTooltip :content="t('admin.ops.tooltips.upstreamErrors')" />
+              <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.upstreamErrors')" />
            </div>
-            <button class="text-[10px] font-bold text-blue-500 hover:underline" type="button" @click="openErrorDetails('upstream')">
+            <button v-if="!props.fullscreen" class="text-[10px] font-bold text-blue-500 hover:underline" type="button" @click="openErrorDetails('upstream')">
              {{ t('admin.ops.requestDetails.details') }}
            </button>
          </div>
@@ -1370,12 +1410,12 @@ function handleToolbarRefresh() {
        <div class="rounded-xl bg-gray-50 p-3 dark:bg-dark-900">
          <div class="flex items-center gap-1">
            <div class="text-[10px] font-bold uppercase tracking-wider text-gray-400">CPU</div>
-            <HelpTooltip :content="t('admin.ops.tooltips.cpu')" />
+            <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.cpu')" />
          </div>
          <div class="mt-1 text-lg font-black" :class="cpuPercentClass">
            {{ cpuPercentValue == null ? '-' : `${cpuPercentValue.toFixed(1)}%` }}
          </div>
-          <div class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
+          <div v-if="!props.fullscreen" class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
            {{ t('common.warning') }} 80% · {{ t('common.critical') }} 95%
          </div>
        </div>
@@ -1384,12 +1424,12 @@ function handleToolbarRefresh() {
        <div class="rounded-xl bg-gray-50 p-3 dark:bg-dark-900">
          <div class="flex items-center gap-1">
            <div class="text-[10px] font-bold uppercase tracking-wider text-gray-400">MEM</div>
-            <HelpTooltip :content="t('admin.ops.tooltips.memory')" />
+            <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.memory')" />
          </div>
          <div class="mt-1 text-lg font-black" :class="memPercentClass">
            {{ memPercentValue == null ? '-' : `${memPercentValue.toFixed(1)}%` }}
          </div>
-          <div class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
+          <div v-if="!props.fullscreen" class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
            {{
              systemMetrics?.memory_used_mb == null || systemMetrics?.memory_total_mb == null
                ? '-'
@@ -1402,12 +1442,12 @@ function handleToolbarRefresh() {
        <div class="rounded-xl bg-gray-50 p-3 dark:bg-dark-900">
          <div class="flex items-center gap-1">
            <div class="text-[10px] font-bold uppercase tracking-wider text-gray-400">DB</div>
-            <HelpTooltip :content="t('admin.ops.tooltips.db')" />
+            <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.db')" />
          </div>
          <div class="mt-1 text-lg font-black" :class="dbMiddleClass">
            {{ dbMiddleLabel }}
          </div>
-          <div class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
+          <div v-if="!props.fullscreen" class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
            {{ t('admin.ops.conns') }} {{ dbConnOpenValue ?? '-' }} / {{ dbMaxOpenConnsValue ?? '-' }}
            · {{ t('admin.ops.active') }} {{ dbConnActiveValue ?? '-' }}
            · {{ t('admin.ops.idle') }} {{ dbConnIdleValue ?? '-' }}
@@ -1419,12 +1459,12 @@ function handleToolbarRefresh() {
        <div class="rounded-xl bg-gray-50 p-3 dark:bg-dark-900">
          <div class="flex items-center gap-1">
            <div class="text-[10px] font-bold uppercase tracking-wider text-gray-400">Redis</div>
-            <HelpTooltip :content="t('admin.ops.tooltips.redis')" />
+            <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.redis')" />
          </div>
          <div class="mt-1 text-lg font-black" :class="redisMiddleClass">
            {{ redisMiddleLabel }}
          </div>
-          <div class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
+          <div v-if="!props.fullscreen" class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
            {{ t('admin.ops.conns') }} {{ redisConnTotalValue ?? '-' }} / {{ redisPoolSizeValue ?? '-' }}
            <span v-if="redisConnActiveValue != null"> · {{ t('admin.ops.active') }} {{ redisConnActiveValue }} </span>
            <span v-if="redisConnIdleValue != null"> · {{ t('admin.ops.idle') }} {{ redisConnIdleValue }} </span>
@@ -1435,12 +1475,12 @@ function handleToolbarRefresh() {
        <div class="rounded-xl bg-gray-50 p-3 dark:bg-dark-900">
          <div class="flex items-center gap-1">
            <div class="text-[10px] font-bold uppercase tracking-wider text-gray-400">{{ t('admin.ops.goroutines') }}</div>
-            <HelpTooltip :content="t('admin.ops.tooltips.goroutines')" />
+            <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.goroutines')" />
          </div>
          <div class="mt-1 text-lg font-black" :class="goroutineStatusClass">
            {{ goroutineStatusLabel }}
          </div>
-          <div class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
+          <div v-if="!props.fullscreen" class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
            {{ t('admin.ops.current') }} <span class="font-mono">{{ goroutineCountValue ?? '-' }}</span>
            · {{ t('common.warning') }} <span class="font-mono">{{ goroutinesWarnThreshold }}</span>
            · {{ t('common.critical') }} <span class="font-mono">{{ goroutinesCriticalThreshold }}</span>
@@ -1455,9 +1495,9 @@ function handleToolbarRefresh() {
          <div class="flex items-center justify-between gap-2">
            <div class="flex items-center gap-1">
              <div class="text-[10px] font-bold uppercase tracking-wider text-gray-400">{{ t('admin.ops.jobs') }}</div>
-              <HelpTooltip :content="t('admin.ops.tooltips.jobs')" />
+              <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.jobs')" />
            </div>
-            <button class="text-[10px] font-bold text-blue-500 hover:underline" type="button" @click="openJobsDetails">
+            <button v-if="!props.fullscreen" class="text-[10px] font-bold text-blue-500 hover:underline" type="button" @click="openJobsDetails">
              {{ t('admin.ops.requestDetails.details') }}
            </button>
          </div>
@@ -1466,7 +1506,7 @@ function handleToolbarRefresh() {
            {{ jobsStatusLabel }}
          </div>

-          <div class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
+          <div v-if="!props.fullscreen" class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
            {{ t('common.total') }} <span class="font-mono">{{ jobHeartbeats.length }}</span>
            · {{ t('common.warning') }} <span class="font-mono">{{ jobsWarnCount }}</span>
          </div>
--- a/frontend/src/views/admin/ops/components/OpsSettingsDialog.vue
+++ b/frontend/src/views/admin/ops/components/OpsSettingsDialog.vue
@@ -487,6 +487,48 @@ async function saveAllSettings() {
              <Toggle v-model="advancedSettings.aggregation.aggregation_enabled" />
            </div>
          </div>
+
+          <!-- 错误过滤 -->
+          <div class="space-y-3">
+            <h5 class="text-xs font-semibold text-gray-700 dark:text-gray-300">错误过滤</h5>
+
+            <div class="flex items-center justify-between">
+              <div>
+                <label class="text-sm font-medium text-gray-700 dark:text-gray-300">忽略 count_tokens 错误</label>
+                <p class="mt-1 text-xs text-gray-500">
+                  启用后，count_tokens 请求的错误将不计入运维监控的统计和告警中（但仍会存储在数据库中）
+                </p>
+              </div>
+              <Toggle v-model="advancedSettings.ignore_count_tokens_errors" />
+            </div>
+          </div>
+
+          <!-- 自动刷新 -->
+          <div class="space-y-3">
+            <h5 class="text-xs font-semibold text-gray-700 dark:text-gray-300">自动刷新</h5>
+
+            <div class="flex items-center justify-between">
+              <div>
+                <label class="text-sm font-medium text-gray-700 dark:text-gray-300">启用自动刷新</label>
+                <p class="mt-1 text-xs text-gray-500">
+                  自动刷新仪表板数据，启用后会定期拉取最新数据
+                </p>
+              </div>
+              <Toggle v-model="advancedSettings.auto_refresh_enabled" />
+            </div>
+
+            <div v-if="advancedSettings.auto_refresh_enabled">
+              <label class="input-label">刷新间隔</label>
+              <Select
+                v-model="advancedSettings.auto_refresh_interval_seconds"
+                :options="[
+                  { value: 15, label: '15 秒' },
+                  { value: 30, label: '30 秒' },
+                  { value: 60, label: '60 秒' }
+                ]"
+              />
+            </div>
+          </div>
        </div>
      </details>
    </div>
--- a/frontend/src/views/admin/ops/components/OpsThroughputTrendChart.vue
+++ b/frontend/src/views/admin/ops/components/OpsThroughputTrendChart.vue
@@ -19,6 +19,7 @@ interface Props {
  timeRange: string
  byPlatform?: OpsThroughputPlatformBreakdownItem[]
  topGroups?: OpsThroughputGroupBreakdownItem[]
+  fullscreen?: boolean
 }

 const props = defineProps<Props>()
@@ -179,38 +180,40 @@ function downloadChart() {
          <path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M13 7h8m0 0v8m0-8l-8 8-4-4-6 6" />
        </svg>
        {{ t('admin.ops.throughputTrend') }}
-        <HelpTooltip :content="t('admin.ops.tooltips.throughputTrend')" />
+        <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.throughputTrend')" />
      </h3>
      <div class="flex items-center gap-2 text-xs text-gray-500 dark:text-gray-400">
        <span class="flex items-center gap-1"><span class="h-2 w-2 rounded-full bg-blue-500"></span>{{ t('admin.ops.qps') }}</span>
        <span class="flex items-center gap-1"><span class="h-2 w-2 rounded-full bg-green-500"></span>{{ t('admin.ops.tpsK') }}</span>
-        <button
-          type="button"
-          class="ml-2 inline-flex items-center rounded-lg border border-gray-200 bg-white px-2 py-1 text-[11px] font-semibold text-gray-600 hover:bg-gray-50 disabled:opacity-50 dark:border-dark-700 dark:bg-dark-900 dark:text-gray-300 dark:hover:bg-dark-800"
-          :disabled="state !== 'ready'"
-          :title="t('admin.ops.requestDetails.title')"
-          @click="emit('openDetails')"
-        >
-          {{ t('admin.ops.requestDetails.details') }}
-        </button>
-        <button
-          type="button"
-          class="ml-2 inline-flex items-center rounded-lg border border-gray-200 bg-white px-2 py-1 text-[11px] font-semibold text-gray-600 hover:bg-gray-50 disabled:opacity-50 dark:border-dark-700 dark:bg-dark-900 dark:text-gray-300 dark:hover:bg-dark-800"
-          :disabled="state !== 'ready'"
-          :title="t('admin.ops.charts.resetZoomHint')"
-          @click="resetZoom"
-        >
-          {{ t('admin.ops.charts.resetZoom') }}
-        </button>
-        <button
-          type="button"
-          class="inline-flex items-center rounded-lg border border-gray-200 bg-white px-2 py-1 text-[11px] font-semibold text-gray-600 hover:bg-gray-50 disabled:opacity-50 dark:border-dark-700 dark:bg-dark-900 dark:text-gray-300 dark:hover:bg-dark-800"
-          :disabled="state !== 'ready'"
-          :title="t('admin.ops.charts.downloadChartHint')"
-          @click="downloadChart"
-        >
-          {{ t('admin.ops.charts.downloadChart') }}
-        </button>
+        <template v-if="!props.fullscreen">
+          <button
+            type="button"
+            class="ml-2 inline-flex items-center rounded-lg border border-gray-200 bg-white px-2 py-1 text-[11px] font-semibold text-gray-600 hover:bg-gray-50 disabled:opacity-50 dark:border-dark-700 dark:bg-dark-900 dark:text-gray-300 dark:hover:bg-dark-800"
+            :disabled="state !== 'ready'"
+            :title="t('admin.ops.requestDetails.title')"
+            @click="emit('openDetails')"
+          >
+            {{ t('admin.ops.requestDetails.details') }}
+          </button>
+          <button
+            type="button"
+            class="ml-2 inline-flex items-center rounded-lg border border-gray-200 bg-white px-2 py-1 text-[11px] font-semibold text-gray-600 hover:bg-gray-50 disabled:opacity-50 dark:border-dark-700 dark:bg-dark-900 dark:text-gray-300 dark:hover:bg-dark-800"
+            :disabled="state !== 'ready'"
+            :title="t('admin.ops.charts.resetZoomHint')"
+            @click="resetZoom"
+          >
+            {{ t('admin.ops.charts.resetZoom') }}
+          </button>
+          <button
+            type="button"
+            class="inline-flex items-center rounded-lg border border-gray-200 bg-white px-2 py-1 text-[11px] font-semibold text-gray-600 hover:bg-gray-50 disabled:opacity-50 dark:border-dark-700 dark:bg-dark-900 dark:text-gray-300 dark:hover:bg-dark-800"
+            :disabled="state !== 'ready'"
+            :title="t('admin.ops.charts.downloadChartHint')"
+            @click="downloadChart"
+          >
+            {{ t('admin.ops.charts.downloadChart') }}
+          </button>
+        </template>
      </div>
    </div>
Author	SHA1	Message	Date
shaw	93db889a10	fix: Gemini OpenCode 教程 baseURL 改为 v1beta	2026-01-13 09:52:37 +08:00
Wesley Liddick	0df7385c4e	Merge pull request #226 from xilu0/main feat(gateway): 优化 Antigravity/Gemini 思考块处理此提交解决了思考块 (thinking blocks) 在转发过程中的兼容性问题	2026-01-13 09:39:43 +08:00
Wesley Liddick	1a3fa6411c	Merge pull request #260 from IanShaw027/fix/sync-openai-gpt5-models fix: 同步 OpenAI GPT-5 模型列表并完善参数处理	2026-01-13 09:31:00 +08:00
Wesley Liddick	64614756d1	Merge pull request #259 from cyhhao/main fix: adjust OpenCode OpenAI example store placement	2026-01-13 09:30:26 +08:00
Wesley Liddick	bb1fd54d4d	Merge pull request #257 from Edric-Li/feat/ops-fullscreen-scrollbar feat(ops): 添加运维监控全屏模式 & 优化滚动条	2026-01-13 09:29:25 +08:00
ianshaw	d85288a6c0	Revert "fix(gateway): 修复 base_url 包含 /chat/completions 时路径拼接错误" This reverts commit `7fdc25df3c`.	2026-01-12 13:29:04 -08:00
ianshaw	3402acb606	feat(gateway): 对所有请求（包括 Codex CLI）应用模型映射 - 移除 Codex CLI 的模型映射跳过逻辑 - 添加详细的模型映射日志，包含账号名称和请求类型 - 确保所有 OpenAI 请求都能正确应用账号配置的模型映射	2026-01-12 13:23:05 -08:00
ianshaw	7fdc25df3c	fix(gateway): 修复 base_url 包含 /chat/completions 时路径拼接错误问题： - 当账号的 base_url 配置为 https://example.com/v1/chat/completions 时 - 代码直接追加 /responses，导致路径变成 /v1/chat/completions/responses - 上游返回 404 错误修复： - 在追加 /responses 前，先移除 base_url 中的 /chat/completions 后缀 - 确保最终路径为 https://example.com/v1/responses 影响范围： - OpenAI API Key 账号的测试接口 - OpenAI API Key 账号的实际网关请求 Related-to: #231	2026-01-12 11:39:45 -08:00
ianshaw	ea699cbdc2	docs(frontend): 完善 OpenCode 配置说明更新 API 密钥页面 OpenCode 配置提示信息： - 补充支持 opencode.jsonc 后缀名 - 说明可使用默认 provider（openai/anthropic/google）或自定义 provider_id - 说明 API Key 支持直接配置或通过 /connect 命令配置 - 保留"示例仅供参考，模型与选项可按需调整"的提示配置文件路径：~/.config/opencode/opencode.json（或 opencode.jsonc）	2026-01-12 11:17:47 -08:00
ianshaw	fe6a3f4267	fix(gateway): 完善 max_output_tokens 参数处理逻辑根据不同平台和账号类型处理 max_output_tokens 参数： - OpenAI OAuth (Responses API): 保留 max_output_tokens（支持） - OpenAI API Key: 删除 max_output_tokens（不支持） - Anthropic (Claude): 转换 max_output_tokens 为 max_tokens - Gemini: 删除 max_output_tokens（由 Gemini 专用转换处理） - 其他平台: 删除（安全起见）同时处理 max_completion_tokens 参数，仅在 OpenAI OAuth 时保留。修复客户端（如 OpenCode）发送不支持参数导致上游返回 400 错误的问题。 Related-to: #231	2026-01-12 11:08:28 -08:00
ianshaw	fe8198c8cd	fix(frontend): 同步 OpenAI GPT-5 系列模型列表修复编辑账号页面 GPT-5 模型只显示 3 个的问题： - 原来只有: gpt-5, gpt-5-mini, gpt-5-nano - 现在添加完整的 22 个模型，包括： * GPT-5 系列: gpt-5, gpt-5-codex, gpt-5-chat, gpt-5-pro, gpt-5-mini, gpt-5-nano 及各时间戳版本 * GPT-5.1 系列: gpt-5.1, gpt-5.1-codex, gpt-5.1-codex-max, gpt-5.1-codex-mini 及各版本 * GPT-5.2 系列: gpt-5.2, gpt-5.2-codex, gpt-5.2-pro 及各版本 - 更新快捷预设按钮，新增 GPT-5.1, GPT-5.2, GPT-5.1 Codex 选项与后端定价文件 (model_prices_and_context_window.json) 保持一致。 Fixes issue introduced in `fb86002` (feat: 添加模型白名单选择器组件) Related-to: `fb86002ef9`	2026-01-12 10:14:50 -08:00
cyhhao	675e61385f	Merge branch 'main' of github.com:Wei-Shaw/sub2api	2026-01-12 22:36:14 +08:00
cyhhao	67acac1082	fix: adjust OpenCode OpenAI example store placement	2026-01-12 22:31:43 +08:00
Edric Li	d02e1db018	style: 优化滚动条自动隐藏效果 - 默认隐藏滚动条，悬停时显示 - 支持 Webkit (Chrome/Safari/Edge) 和 Firefox - 滚动条样式与暗色主题适配	2026-01-12 22:10:59 +08:00
Edric Li	0da515071b	feat(ops): 添加运维监控全屏模式 - 支持通过 URL 参数 ?fullscreen=1 进入全屏模式 - 全屏模式下隐藏非必要 UI 元素（选择器、按钮、提示等） - 增大健康评分圆环和字体以提升可读性 - 支持 ESC 键退出全屏 - 添加全屏按钮的 i18n 翻译	2026-01-12 22:10:59 +08:00
xiluo	524d80ae1c	feat(gateway): 优化 Antigravity/Gemini 思考块处理此提交解决了思考块 (thinking blocks) 在转发过程中的兼容性问题。主要变更： 1. 思考块优化 (Thinking Blocks)： - 在 AntigravityGatewayService 中增加了 sanitizeThinkingBlocks 处理，强制移除思考块中不支持的 cache_control 字段（避免 Anthropic/Vertex AI 报错） - 实现历史思考块展平 (Flattening)：将非最后一条消息中的思考块转换为普通文本块，以绕过上游对历史思考块签名的严格校验 - 增加 cleanCacheControlFromGeminiJSON 作为最后一道防线，确保转换后的 Gemini 请求中不残留非法的 cache_control 2. GatewayService 缓存控制优化： - 更新缓存控制逻辑，跳过 thinking 块（thinking 块不支持 cache_control 字段） - 增加 removeCacheControlFromThinkingBlocks 函数强制清理关联 Issue: #225	2026-01-12 13:36:59 +00:00
shaw	3b71bc3df1	feat: OpenCode 配置提示添加配置文件路径说明	2026-01-12 20:49:54 +08:00
shaw	22ef9534e0	fix: 修复反向代理下客户端 IP 获取错误	2026-01-12 20:44:38 +08:00
Wesley Liddick	c206d12d5c	Merge pull request #254 from IanShaw027/feat/ops-count-tokens-filter-and-auto-refresh feat(ops): count_tokens 错误过滤和自动刷新功能	2026-01-12 17:31:54 +08:00
IanShaw027	6ad29a470c	style(ops): 移除未使用的 isAutoRefreshActive 变量	2026-01-12 17:28:25 +08:00
IanShaw027	2d45e61a9b	style(ops): 修复代码格式问题以通过 golangci-lint	2026-01-12 17:18:49 +08:00
IanShaw027	b98fb013ae	feat(ops): 添加自动刷新配置功能功能特性： - 支持配置启用/禁用自动刷新 - 可配置刷新间隔（15秒/30秒/60秒） - 实时倒计时显示，用户可见下次刷新时间 - 手动刷新自动重置倒计时 - 页面卸载时自动清理定时器用户体验： - 默认禁用，用户可根据需求开启 - 与现有 OpsConcurrencyCard 5秒刷新保持一致 - 倒计时带旋转动画，视觉反馈清晰 - 配置修改后立即生效，无需刷新页面技术实现： - ops.ts: 添加 auto_refresh_enabled 和 auto_refresh_interval_seconds 配置 - OpsSettingsDialog.vue: 添加自动刷新配置界面 - OpsDashboard.vue: 实现主刷新逻辑和双定时器设计 - OpsDashboardHeader.vue: 倒计时显示组件配置说明： - auto_refresh_enabled: 是否启用（默认 false） - auto_refresh_interval_seconds: 刷新间隔（默认 30 秒，范围 15-300 秒）	2026-01-12 17:07:07 +08:00
IanShaw027	345a965fa3	feat(ops): 添加 count_tokens 错误过滤功能功能特性： - 自动识别并标记 count_tokens 请求的错误 - 支持配置是否在统计中忽略 count_tokens 错误 - 错误数据完整保留，仅在统计时动态过滤技术实现： - ops_error_logger.go: 自动标记 count_tokens 请求 - ops_repo.go: INSERT 语句添加 is_count_tokens 字段 - ops_repo_dashboard.go: buildErrorWhere 核心过滤函数 - ops_repo_preagg.go: 预聚合统计中添加过滤 - ops_repo_trends.go: 趋势统计查询添加过滤（2 处） - ops_settings_models.go: 添加 ignore_count_tokens_errors 配置 - ops_settings.go: 配置验证和默认值设置 - ops_port.go: 错误日志模型添加 IsCountTokens 字段业务价值： - count_tokens 是探测性请求，其错误不影响真实业务 SLA - 用户可根据需求灵活控制是否计入统计 - 提升错误率、告警等运维指标的准确性影响范围： - Dashboard 概览统计 - 错误趋势图表 - 告警规则评估 - 预聚合指标（hourly/daily） - 健康分数计算	2026-01-12 17:06:12 +08:00
IanShaw027	c02c120579	feat(ops): 添加 count_tokens 错误标记数据库迁移 - 新增 is_count_tokens 布尔字段到 ops_error_logs 表 - 默认值为 false - 支持后续动态过滤统计	2026-01-12 17:06:12 +08:00