fix(admin): 修复退款金额精度问题

- 显示完整余额精度，避免四舍五入导致的退款失败
- 添加"全部"按钮，一键填入完整余额
- 移除最小金额限制，支持任意正数金额

README_CN.md

@@ -57,6 +57,13 @@ Sub2API 是一个 AI API 网关平台，用于分发和管理 AI 产品订阅（
 
 ---
 
+## OpenAI Responses 兼容注意事项
+
+- 当请求包含 `function_call_output` 时，需要携带 `previous_response_id`，或在 `input` 中包含带 `call_id` 的 `tool_call`/`function_call`，或带非空 `id` 且与 `function_call_output.call_id` 匹配的 `item_reference`。
+- 若依赖上游历史记录，网关会强制 `store=true` 并需要复用 `previous_response_id`，以避免出现 “No tool call found for function call output” 错误。
+
+---
+
 ## 部署方式
 
 ### 方式一：脚本安装（推荐）

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

backend/internal/handler/openai_gateway_handler.go

@@ -114,6 +114,26 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 
 	setOpsRequestContext(c, reqModel, reqStream, body)
 
+	// 提前校验 function_call_output 是否具备可关联上下文，避免上游 400。
+	// 要求 previous_response_id，或 input 内存在带 call_id 的 tool_call/function_call，
+	// 或带 id 且与 call_id 匹配的 item_reference。
+	if service.HasFunctionCallOutput(reqBody) {
+		previousResponseID, _ := reqBody["previous_response_id"].(string)
+		if strings.TrimSpace(previousResponseID) == "" && !service.HasToolCallContext(reqBody) {
+			if service.HasFunctionCallOutputMissingCallID(reqBody) {
+				log.Printf("[OpenAI Handler] function_call_output 缺少 call_id: model=%s", reqModel)
+				h.errorResponse(c, http.StatusBadRequest, "invalid_request_error", "function_call_output requires call_id or previous_response_id; if relying on history, ensure store=true and reuse previous_response_id")
+				return
+			}
+			callIDs := service.FunctionCallOutputCallIDs(reqBody)
+			if !service.HasItemReferenceForCallIDs(reqBody, callIDs) {
+				log.Printf("[OpenAI Handler] function_call_output 缺少匹配的 item_reference: model=%s", reqModel)
+				h.errorResponse(c, http.StatusBadRequest, "invalid_request_error", "function_call_output requires item_reference ids matching each call_id, or previous_response_id/tool_call context; if relying on history, ensure store=true and reuse previous_response_id")
+				return
+			}
+		}
+	}
+
 	// Track if we've started streaming (for error handling)
 	streamStarted := false
 

backend/internal/middleware/rate_limiter.go

@@ -2,7 +2,10 @@ package middleware
 
 import (
 	"context"
+	"fmt"
+	"log"
 	"net/http"
+	"strconv"
 	"time"
 
 	"github.com/gin-gonic/gin"
@@ -25,15 +28,34 @@ type RateLimitOptions struct {
 var rateLimitScript = redis.NewScript(`
 local current = redis.call('INCR', KEYS[1])
 local ttl = redis.call('PTTL', KEYS[1])
-if current == 1 or ttl == -1 then
+local repaired = 0
+if current == 1 then
   redis.call('PEXPIRE', KEYS[1], ARGV[1])
+elseif ttl == -1 then
+  redis.call('PEXPIRE', KEYS[1], ARGV[1])
+  repaired = 1
 end
-return current
+return {current, repaired}
 `)
 
 // rateLimitRun 允许测试覆写脚本执行逻辑
-var rateLimitRun = func(ctx context.Context, client *redis.Client, key string, windowMillis int64) (int64, error) {
+var rateLimitRun = func(ctx context.Context, client *redis.Client, key string, windowMillis int64) (int64, bool, error) {
-	return rateLimitScript.Run(ctx, client, []string{key}, windowMillis).Int64()
+	values, err := rateLimitScript.Run(ctx, client, []string{key}, windowMillis).Slice()
+	if err != nil {
+		return 0, false, err
+	}
+	if len(values) < 2 {
+		return 0, false, fmt.Errorf("rate limit script returned %d values", len(values))
+	}
+	count, err := parseInt64(values[0])
+	if err != nil {
+		return 0, false, err
+	}
+	repaired, err := parseInt64(values[1])
+	if err != nil {
+		return 0, false, err
+	}
+	return count, repaired == 1, nil
 }
 
 // RateLimiter Redis 速率限制器
@@ -74,8 +96,9 @@ func (r *RateLimiter) LimitWithOptions(key string, limit int, window time.Durati
 		windowMillis := windowTTLMillis(window)
 
 		// 使用 Lua 脚本原子操作增加计数并设置过期
-		count, err := rateLimitRun(ctx, r.redis, redisKey, windowMillis)
+		count, repaired, err := rateLimitRun(ctx, r.redis, redisKey, windowMillis)
 		if err != nil {
+			log.Printf("[RateLimit] redis error: key=%s mode=%s err=%v", redisKey, failureModeLabel(failureMode), err)
 			if failureMode == RateLimitFailClose {
 				abortRateLimit(c)
 				return
@@ -84,6 +107,9 @@ func (r *RateLimiter) LimitWithOptions(key string, limit int, window time.Durati
 			c.Next()
 			return
 		}
+		if repaired {
+			log.Printf("[RateLimit] ttl repaired: key=%s window_ms=%d", redisKey, windowMillis)
+		}
 
 		// 超过限制
 		if count > int64(limit) {
@@ -109,3 +135,27 @@ func abortRateLimit(c *gin.Context) {
 		"message": "Too many requests, please try again later",
 	})
 }
+
+func failureModeLabel(mode RateLimitFailureMode) string {
+	if mode == RateLimitFailClose {
+		return "fail-close"
+	}
+	return "fail-open"
+}
+
+func parseInt64(value any) (int64, error) {
+	switch v := value.(type) {
+	case int64:
+		return v, nil
+	case int:
+		return int64(v), nil
+	case string:
+		parsed, err := strconv.ParseInt(v, 10, 64)
+		if err != nil {
+			return 0, err
+		}
+		return parsed, nil
+	default:
+		return 0, fmt.Errorf("unexpected value type %T", value)
+	}
+}

backend/internal/middleware/rate_limiter_test.go

@@ -66,13 +66,13 @@ func TestRateLimiterSuccessAndLimit(t *testing.T) {
 	originalRun := rateLimitRun
 	counts := []int64{1, 2}
 	callIndex := 0
-	rateLimitRun = func(ctx context.Context, client *redis.Client, key string, windowMillis int64) (int64, error) {
+	rateLimitRun = func(ctx context.Context, client *redis.Client, key string, windowMillis int64) (int64, bool, error) {
 		if callIndex >= len(counts) {
-			return counts[len(counts)-1], nil
+			return counts[len(counts)-1], false, nil
 		}
 		value := counts[callIndex]
 		callIndex++
-		return value, nil
+		return value, false, nil
 	}
 	t.Cleanup(func() {
 		rateLimitRun = originalRun

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.

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()

backend/internal/service/openai_codex_transform.go

@@ -1,6 +1,7 @@
 package service
 
 import (
+	_ "embed"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -16,6 +17,9 @@ const (
 	codexCacheTTL          = 15 * time.Minute
 )
 
+//go:embed prompts/codex_cli_instructions.md
+var codexCLIInstructions string
+
 var codexModelMap = map[string]string{
 	"gpt-5.1-codex":             "gpt-5.1-codex",
 	"gpt-5.1-codex-low":         "gpt-5.1-codex",
@@ -70,6 +74,8 @@ type opencodeCacheMetadata struct {
 
 func applyCodexOAuthTransform(reqBody map[string]any) codexTransformResult {
 	result := codexTransformResult{}
+	// 工具续链需求会影响存储策略与 input 过滤逻辑。
+	needsToolContinuation := NeedsToolContinuation(reqBody)
 
 	model := ""
 	if v, ok := reqBody["model"].(string); ok {
@@ -84,6 +90,8 @@ func applyCodexOAuthTransform(reqBody map[string]any) codexTransformResult {
 		result.NormalizedModel = normalizedModel
 	}
 
+	// OAuth 走 ChatGPT internal API 时，store 必须为 false；显式 true 也会强制覆盖。
+	// 避免上游返回 "Store must be set to false"。
 	if v, ok := reqBody["store"].(bool); !ok || v {
 		reqBody["store"] = false
 		result.Modified = true
@@ -119,10 +127,18 @@ func applyCodexOAuthTransform(reqBody map[string]any) codexTransformResult {
 			reqBody["instructions"] = instructions
 			result.Modified = true
 		}
+	} else if existingInstructions == "" {
+		// 未获取到 opencode 指令时，回退使用 Codex CLI 指令。
+		codexInstructions := strings.TrimSpace(getCodexCLIInstructions())
+		if codexInstructions != "" {
+			reqBody["instructions"] = codexInstructions
+			result.Modified = true
+		}
 	}
 
+	// 续链场景保留 item_reference 与 id，避免 call_id 上下文丢失。
 	if input, ok := reqBody["input"].([]any); ok {
-		input = filterCodexInput(input)
+		input = filterCodexInput(input, needsToolContinuation)
 		reqBody["input"] = input
 		result.Modified = true
 	}
@@ -235,14 +251,75 @@ func getOpenCodeCachedPrompt(url, cacheFileName, metaFileName string) string {
 }
 
 func getOpenCodeCodexHeader() string {
-	return getOpenCodeCachedPrompt(opencodeCodexHeaderURL, "opencode-codex-header.txt", "opencode-codex-header-meta.json")
+	// 优先从 opencode 仓库缓存获取指令。
+	opencodeInstructions := getOpenCodeCachedPrompt(opencodeCodexHeaderURL, "opencode-codex-header.txt", "opencode-codex-header-meta.json")
+
+	// 若 opencode 指令可用，直接返回。
+	if opencodeInstructions != "" {
+		return opencodeInstructions
+	}
+
+	// 否则回退使用本地 Codex CLI 指令。
+	return getCodexCLIInstructions()
+}
+
+func getCodexCLIInstructions() string {
+	return codexCLIInstructions
 }
 
 func GetOpenCodeInstructions() string {
 	return getOpenCodeCodexHeader()
 }
 
-func filterCodexInput(input []any) []any {
+// GetCodexCLIInstructions 返回内置的 Codex CLI 指令内容。
+func GetCodexCLIInstructions() string {
+	return getCodexCLIInstructions()
+}
+
+// ReplaceWithCodexInstructions 将请求 instructions 替换为内置 Codex 指令（必要时）。
+func ReplaceWithCodexInstructions(reqBody map[string]any) bool {
+	codexInstructions := strings.TrimSpace(getCodexCLIInstructions())
+	if codexInstructions == "" {
+		return false
+	}
+
+	existingInstructions, _ := reqBody["instructions"].(string)
+	if strings.TrimSpace(existingInstructions) != codexInstructions {
+		reqBody["instructions"] = codexInstructions
+		return true
+	}
+
+	return false
+}
+
+// IsInstructionError 判断错误信息是否与指令格式/系统提示相关。
+func IsInstructionError(errorMessage string) bool {
+	if errorMessage == "" {
+		return false
+	}
+
+	lowerMsg := strings.ToLower(errorMessage)
+	instructionKeywords := []string{
+		"instruction",
+		"instructions",
+		"system prompt",
+		"system message",
+		"invalid prompt",
+		"prompt format",
+	}
+
+	for _, keyword := range instructionKeywords {
+		if strings.Contains(lowerMsg, keyword) {
+			return true
+		}
+	}
+
+	return false
+}
+
+// filterCodexInput 按需过滤 item_reference 与 id。
+// preserveReferences 为 true 时保持引用与 id，以满足续链请求对上下文的依赖。
+func filterCodexInput(input []any, preserveReferences bool) []any {
 	filtered := make([]any, 0, len(input))
 	for _, item := range input {
 		m, ok := item.(map[string]any)
@@ -250,15 +327,62 @@ func filterCodexInput(input []any) []any {
 			filtered = append(filtered, item)
 			continue
 		}
-		if typ, ok := m["type"].(string); ok && typ == "item_reference" {
+		typ, _ := m["type"].(string)
+		if typ == "item_reference" {
+			if !preserveReferences {
+				continue
+			}
+			newItem := make(map[string]any, len(m))
+			for key, value := range m {
+				newItem[key] = value
+			}
+			filtered = append(filtered, newItem)
 			continue
 		}
-		delete(m, "id")
+
-		filtered = append(filtered, m)
+		newItem := m
+		copied := false
+		// 仅在需要修改字段时创建副本，避免直接改写原始输入。
+		ensureCopy := func() {
+			if copied {
+				return
+			}
+			newItem = make(map[string]any, len(m))
+			for key, value := range m {
+				newItem[key] = value
+			}
+			copied = true
+		}
+
+		if isCodexToolCallItemType(typ) {
+			if callID, ok := m["call_id"].(string); !ok || strings.TrimSpace(callID) == "" {
+				if id, ok := m["id"].(string); ok && strings.TrimSpace(id) != "" {
+					ensureCopy()
+					newItem["call_id"] = id
+				}
+			}
+		}
+
+		if !preserveReferences {
+			ensureCopy()
+			delete(newItem, "id")
+			if !isCodexToolCallItemType(typ) {
+				delete(newItem, "call_id")
+			}
+		}
+
+		filtered = append(filtered, newItem)
 	}
 	return filtered
 }
 
+func isCodexToolCallItemType(typ string) bool {
+	if typ == "" {
+		return false
+	}
+	return strings.HasSuffix(typ, "_call") || strings.HasSuffix(typ, "_call_output")
+}
+
 func normalizeCodexTools(reqBody map[string]any) bool {
 	rawTools, ok := reqBody["tools"]
 	if !ok || rawTools == nil {

backend/internal/service/openai_codex_transform_test.go

@@ -0,0 +1,167 @@
+package service
+
+import (
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestApplyCodexOAuthTransform_ToolContinuationPreservesInput(t *testing.T) {
+	// 续链场景：保留 item_reference 与 id，但不再强制 store=true。
+	setupCodexCache(t)
+
+	reqBody := map[string]any{
+		"model": "gpt-5.2",
+		"input": []any{
+			map[string]any{"type": "item_reference", "id": "ref1", "text": "x"},
+			map[string]any{"type": "function_call_output", "call_id": "call_1", "output": "ok", "id": "o1"},
+		},
+		"tool_choice": "auto",
+	}
+
+	applyCodexOAuthTransform(reqBody)
+
+	// 未显式设置 store=true，默认为 false。
+	store, ok := reqBody["store"].(bool)
+	require.True(t, ok)
+	require.False(t, store)
+
+	input, ok := reqBody["input"].([]any)
+	require.True(t, ok)
+	require.Len(t, input, 2)
+
+	// 校验 input[0] 为 map，避免断言失败导致测试中断。
+	first, ok := input[0].(map[string]any)
+	require.True(t, ok)
+	require.Equal(t, "item_reference", first["type"])
+	require.Equal(t, "ref1", first["id"])
+
+	// 校验 input[1] 为 map，确保后续字段断言安全。
+	second, ok := input[1].(map[string]any)
+	require.True(t, ok)
+	require.Equal(t, "o1", second["id"])
+}
+
+func TestApplyCodexOAuthTransform_ExplicitStoreFalsePreserved(t *testing.T) {
+	// 续链场景：显式 store=false 不再强制为 true，保持 false。
+	setupCodexCache(t)
+
+	reqBody := map[string]any{
+		"model": "gpt-5.1",
+		"store": false,
+		"input": []any{
+			map[string]any{"type": "function_call_output", "call_id": "call_1"},
+		},
+		"tool_choice": "auto",
+	}
+
+	applyCodexOAuthTransform(reqBody)
+
+	store, ok := reqBody["store"].(bool)
+	require.True(t, ok)
+	require.False(t, store)
+}
+
+func TestApplyCodexOAuthTransform_ExplicitStoreTrueForcedFalse(t *testing.T) {
+	// 显式 store=true 也会强制为 false。
+	setupCodexCache(t)
+
+	reqBody := map[string]any{
+		"model": "gpt-5.1",
+		"store": true,
+		"input": []any{
+			map[string]any{"type": "function_call_output", "call_id": "call_1"},
+		},
+		"tool_choice": "auto",
+	}
+
+	applyCodexOAuthTransform(reqBody)
+
+	store, ok := reqBody["store"].(bool)
+	require.True(t, ok)
+	require.False(t, store)
+}
+
+func TestApplyCodexOAuthTransform_NonContinuationDefaultsStoreFalseAndStripsIDs(t *testing.T) {
+	// 非续链场景：未设置 store 时默认 false，并移除 input 中的 id。
+	setupCodexCache(t)
+
+	reqBody := map[string]any{
+		"model": "gpt-5.1",
+		"input": []any{
+			map[string]any{"type": "text", "id": "t1", "text": "hi"},
+		},
+	}
+
+	applyCodexOAuthTransform(reqBody)
+
+	store, ok := reqBody["store"].(bool)
+	require.True(t, ok)
+	require.False(t, store)
+
+	input, ok := reqBody["input"].([]any)
+	require.True(t, ok)
+	require.Len(t, input, 1)
+	// 校验 input[0] 为 map，避免类型不匹配触发 errcheck。
+	item, ok := input[0].(map[string]any)
+	require.True(t, ok)
+	_, hasID := item["id"]
+	require.False(t, hasID)
+}
+
+func TestFilterCodexInput_RemovesItemReferenceWhenNotPreserved(t *testing.T) {
+	input := []any{
+		map[string]any{"type": "item_reference", "id": "ref1"},
+		map[string]any{"type": "text", "id": "t1", "text": "hi"},
+	}
+
+	filtered := filterCodexInput(input, false)
+	require.Len(t, filtered, 1)
+	// 校验 filtered[0] 为 map，确保字段检查可靠。
+	item, ok := filtered[0].(map[string]any)
+	require.True(t, ok)
+	require.Equal(t, "text", item["type"])
+	_, hasID := item["id"]
+	require.False(t, hasID)
+}
+
+func TestApplyCodexOAuthTransform_EmptyInput(t *testing.T) {
+	// 空 input 应保持为空且不触发异常。
+	setupCodexCache(t)
+
+	reqBody := map[string]any{
+		"model": "gpt-5.1",
+		"input": []any{},
+	}
+
+	applyCodexOAuthTransform(reqBody)
+
+	input, ok := reqBody["input"].([]any)
+	require.True(t, ok)
+	require.Len(t, input, 0)
+}
+
+func setupCodexCache(t *testing.T) {
+	t.Helper()
+
+	// 使用临时 HOME 避免触发网络拉取 header。
+	tempDir := t.TempDir()
+	t.Setenv("HOME", tempDir)
+
+	cacheDir := filepath.Join(tempDir, ".opencode", "cache")
+	require.NoError(t, os.MkdirAll(cacheDir, 0o755))
+	require.NoError(t, os.WriteFile(filepath.Join(cacheDir, "opencode-codex-header.txt"), []byte("header"), 0o644))
+
+	meta := map[string]any{
+		"etag":        "",
+		"lastFetch":   time.Now().UTC().Format(time.RFC3339),
+		"lastChecked": time.Now().UnixMilli(),
+	}
+	data, err := json.Marshal(meta)
+	require.NoError(t, err)
+	require.NoError(t, os.WriteFile(filepath.Join(cacheDir, "opencode-codex-header-meta.json"), data, 0o644))
+}

backend/internal/service/openai_gateway_service.go

@@ -42,6 +42,7 @@ var openaiSSEDataRe = regexp.MustCompile(`^data:\s*`)
 var openaiAllowedHeaders = map[string]bool{
 	"accept-language": true,
 	"content-type":    true,
+	"conversation_id": true,
 	"user-agent":      true,
 	"originator":      true,
 	"session_id":      true,
@@ -545,16 +546,35 @@ 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)
+	// 对所有请求执行模型映射（包含 Codex CLI）。
-	mappedModel := reqModel
+	mappedModel := account.GetMappedModel(reqModel)
-	if !isCodexCLI {
+	if mappedModel != reqModel {
-		mappedModel = account.GetMappedModel(reqModel)
+		log.Printf("[OpenAI] Model mapping applied: %s -> %s (account: %s, isCodexCLI: %v)", reqModel, mappedModel, account.Name, isCodexCLI)
-		if mappedModel != reqModel {
+		reqBody["model"] = mappedModel
-			reqBody["model"] = mappedModel
+		bodyModified = true
+	}
+
+	// 针对所有 OpenAI 账号执行 Codex 模型名规范化，确保上游识别一致。
+	if model, ok := reqBody["model"].(string); ok {
+		normalizedModel := normalizeCodexModel(model)
+		if normalizedModel != "" && normalizedModel != model {
+			log.Printf("[OpenAI] Codex model normalization: %s -> %s (account: %s, type: %s, isCodexCLI: %v)",
+				model, normalizedModel, account.Name, account.Type, isCodexCLI)
+			reqBody["model"] = normalizedModel
+			mappedModel = normalizedModel
 			bodyModified = true
 		}
 	}
 
+	// 规范化 reasoning.effort 参数（minimal -> none），与上游允许值对齐。
+	if reasoning, ok := reqBody["reasoning"].(map[string]any); ok {
+		if effort, ok := reasoning["effort"].(string); ok && effort == "minimal" {
+			reasoning["effort"] = "none"
+			bodyModified = true
+			log.Printf("[OpenAI] Normalized reasoning.effort: minimal -> none (account: %s)", account.Name)
+		}
+	}
+
 	if account.Type == AccountTypeOAuth && !isCodexCLI {
 		codexResult := applyCodexOAuthTransform(reqBody)
 		if codexResult.Modified {
@@ -568,6 +588,44 @@ 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
@@ -747,9 +805,6 @@ func (s *OpenAIGatewayService) buildUpstreamRequest(ctx context.Context, c *gin.
 		if promptCacheKey != "" {
 			req.Header.Set("conversation_id", promptCacheKey)
 			req.Header.Set("session_id", promptCacheKey)
-		} else {
-			req.Header.Del("conversation_id")
-			req.Header.Del("session_id")
 		}
 	}
 

backend/internal/service/openai_tool_continuation.go

@@ -0,0 +1,213 @@
+package service
+
+import "strings"
+
+// NeedsToolContinuation 判定请求是否需要工具调用续链处理。
+// 满足以下任一信号即视为续链：previous_response_id、input 内包含 function_call_output/item_reference、
+// 或显式声明 tools/tool_choice。
+func NeedsToolContinuation(reqBody map[string]any) bool {
+	if reqBody == nil {
+		return false
+	}
+	if hasNonEmptyString(reqBody["previous_response_id"]) {
+		return true
+	}
+	if hasToolsSignal(reqBody) {
+		return true
+	}
+	if hasToolChoiceSignal(reqBody) {
+		return true
+	}
+	if inputHasType(reqBody, "function_call_output") {
+		return true
+	}
+	if inputHasType(reqBody, "item_reference") {
+		return true
+	}
+	return false
+}
+
+// HasFunctionCallOutput 判断 input 是否包含 function_call_output，用于触发续链校验。
+func HasFunctionCallOutput(reqBody map[string]any) bool {
+	if reqBody == nil {
+		return false
+	}
+	return inputHasType(reqBody, "function_call_output")
+}
+
+// HasToolCallContext 判断 input 是否包含带 call_id 的 tool_call/function_call，
+// 用于判断 function_call_output 是否具备可关联的上下文。
+func HasToolCallContext(reqBody map[string]any) bool {
+	if reqBody == nil {
+		return false
+	}
+	input, ok := reqBody["input"].([]any)
+	if !ok {
+		return false
+	}
+	for _, item := range input {
+		itemMap, ok := item.(map[string]any)
+		if !ok {
+			continue
+		}
+		itemType, _ := itemMap["type"].(string)
+		if itemType != "tool_call" && itemType != "function_call" {
+			continue
+		}
+		if callID, ok := itemMap["call_id"].(string); ok && strings.TrimSpace(callID) != "" {
+			return true
+		}
+	}
+	return false
+}
+
+// FunctionCallOutputCallIDs 提取 input 中 function_call_output 的 call_id 集合。
+// 仅返回非空 call_id，用于与 item_reference.id 做匹配校验。
+func FunctionCallOutputCallIDs(reqBody map[string]any) []string {
+	if reqBody == nil {
+		return nil
+	}
+	input, ok := reqBody["input"].([]any)
+	if !ok {
+		return nil
+	}
+	ids := make(map[string]struct{})
+	for _, item := range input {
+		itemMap, ok := item.(map[string]any)
+		if !ok {
+			continue
+		}
+		itemType, _ := itemMap["type"].(string)
+		if itemType != "function_call_output" {
+			continue
+		}
+		if callID, ok := itemMap["call_id"].(string); ok && strings.TrimSpace(callID) != "" {
+			ids[callID] = struct{}{}
+		}
+	}
+	if len(ids) == 0 {
+		return nil
+	}
+	result := make([]string, 0, len(ids))
+	for id := range ids {
+		result = append(result, id)
+	}
+	return result
+}
+
+// HasFunctionCallOutputMissingCallID 判断是否存在缺少 call_id 的 function_call_output。
+func HasFunctionCallOutputMissingCallID(reqBody map[string]any) bool {
+	if reqBody == nil {
+		return false
+	}
+	input, ok := reqBody["input"].([]any)
+	if !ok {
+		return false
+	}
+	for _, item := range input {
+		itemMap, ok := item.(map[string]any)
+		if !ok {
+			continue
+		}
+		itemType, _ := itemMap["type"].(string)
+		if itemType != "function_call_output" {
+			continue
+		}
+		callID, _ := itemMap["call_id"].(string)
+		if strings.TrimSpace(callID) == "" {
+			return true
+		}
+	}
+	return false
+}
+
+// HasItemReferenceForCallIDs 判断 item_reference.id 是否覆盖所有 call_id。
+// 用于仅依赖引用项完成续链场景的校验。
+func HasItemReferenceForCallIDs(reqBody map[string]any, callIDs []string) bool {
+	if reqBody == nil || len(callIDs) == 0 {
+		return false
+	}
+	input, ok := reqBody["input"].([]any)
+	if !ok {
+		return false
+	}
+	referenceIDs := make(map[string]struct{})
+	for _, item := range input {
+		itemMap, ok := item.(map[string]any)
+		if !ok {
+			continue
+		}
+		itemType, _ := itemMap["type"].(string)
+		if itemType != "item_reference" {
+			continue
+		}
+		idValue, _ := itemMap["id"].(string)
+		idValue = strings.TrimSpace(idValue)
+		if idValue == "" {
+			continue
+		}
+		referenceIDs[idValue] = struct{}{}
+	}
+	if len(referenceIDs) == 0 {
+		return false
+	}
+	for _, callID := range callIDs {
+		if _, ok := referenceIDs[callID]; !ok {
+			return false
+		}
+	}
+	return true
+}
+
+// inputHasType 判断 input 中是否存在指定类型的 item。
+func inputHasType(reqBody map[string]any, want string) bool {
+	input, ok := reqBody["input"].([]any)
+	if !ok {
+		return false
+	}
+	for _, item := range input {
+		itemMap, ok := item.(map[string]any)
+		if !ok {
+			continue
+		}
+		itemType, _ := itemMap["type"].(string)
+		if itemType == want {
+			return true
+		}
+	}
+	return false
+}
+
+// hasNonEmptyString 判断字段是否为非空字符串。
+func hasNonEmptyString(value any) bool {
+	stringValue, ok := value.(string)
+	return ok && strings.TrimSpace(stringValue) != ""
+}
+
+// hasToolsSignal 判断 tools 字段是否显式声明（存在且不为空）。
+func hasToolsSignal(reqBody map[string]any) bool {
+	raw, exists := reqBody["tools"]
+	if !exists || raw == nil {
+		return false
+	}
+	if tools, ok := raw.([]any); ok {
+		return len(tools) > 0
+	}
+	return false
+}
+
+// hasToolChoiceSignal 判断 tool_choice 是否显式声明（非空或非 nil）。
+func hasToolChoiceSignal(reqBody map[string]any) bool {
+	raw, exists := reqBody["tool_choice"]
+	if !exists || raw == nil {
+		return false
+	}
+	switch value := raw.(type) {
+	case string:
+		return strings.TrimSpace(value) != ""
+	case map[string]any:
+		return len(value) > 0
+	default:
+		return false
+	}
+}

backend/internal/service/openai_tool_continuation_test.go

@@ -0,0 +1,98 @@
+package service
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestNeedsToolContinuationSignals(t *testing.T) {
+	// 覆盖所有触发续链的信号来源，确保判定逻辑完整。
+	cases := []struct {
+		name string
+		body map[string]any
+		want bool
+	}{
+		{name: "nil", body: nil, want: false},
+		{name: "previous_response_id", body: map[string]any{"previous_response_id": "resp_1"}, want: true},
+		{name: "previous_response_id_blank", body: map[string]any{"previous_response_id": "  "}, want: false},
+		{name: "function_call_output", body: map[string]any{"input": []any{map[string]any{"type": "function_call_output"}}}, want: true},
+		{name: "item_reference", body: map[string]any{"input": []any{map[string]any{"type": "item_reference"}}}, want: true},
+		{name: "tools", body: map[string]any{"tools": []any{map[string]any{"type": "function"}}}, want: true},
+		{name: "tools_empty", body: map[string]any{"tools": []any{}}, want: false},
+		{name: "tools_invalid", body: map[string]any{"tools": "bad"}, want: false},
+		{name: "tool_choice", body: map[string]any{"tool_choice": "auto"}, want: true},
+		{name: "tool_choice_object", body: map[string]any{"tool_choice": map[string]any{"type": "function"}}, want: true},
+		{name: "tool_choice_empty_object", body: map[string]any{"tool_choice": map[string]any{}}, want: false},
+		{name: "none", body: map[string]any{"input": []any{map[string]any{"type": "text", "text": "hi"}}}, want: false},
+	}
+
+	for _, tt := range cases {
+		t.Run(tt.name, func(t *testing.T) {
+			require.Equal(t, tt.want, NeedsToolContinuation(tt.body))
+		})
+	}
+}
+
+func TestHasFunctionCallOutput(t *testing.T) {
+	// 仅当 input 中存在 function_call_output 才视为续链输出。
+	require.False(t, HasFunctionCallOutput(nil))
+	require.True(t, HasFunctionCallOutput(map[string]any{
+		"input": []any{map[string]any{"type": "function_call_output"}},
+	}))
+	require.False(t, HasFunctionCallOutput(map[string]any{
+		"input": "text",
+	}))
+}
+
+func TestHasToolCallContext(t *testing.T) {
+	// tool_call/function_call 必须包含 call_id，才能作为可关联上下文。
+	require.False(t, HasToolCallContext(nil))
+	require.True(t, HasToolCallContext(map[string]any{
+		"input": []any{map[string]any{"type": "tool_call", "call_id": "call_1"}},
+	}))
+	require.True(t, HasToolCallContext(map[string]any{
+		"input": []any{map[string]any{"type": "function_call", "call_id": "call_2"}},
+	}))
+	require.False(t, HasToolCallContext(map[string]any{
+		"input": []any{map[string]any{"type": "tool_call"}},
+	}))
+}
+
+func TestFunctionCallOutputCallIDs(t *testing.T) {
+	// 仅提取非空 call_id，去重后返回。
+	require.Empty(t, FunctionCallOutputCallIDs(nil))
+	callIDs := FunctionCallOutputCallIDs(map[string]any{
+		"input": []any{
+			map[string]any{"type": "function_call_output", "call_id": "call_1"},
+			map[string]any{"type": "function_call_output", "call_id": ""},
+			map[string]any{"type": "function_call_output", "call_id": "call_1"},
+		},
+	})
+	require.ElementsMatch(t, []string{"call_1"}, callIDs)
+}
+
+func TestHasFunctionCallOutputMissingCallID(t *testing.T) {
+	require.False(t, HasFunctionCallOutputMissingCallID(nil))
+	require.True(t, HasFunctionCallOutputMissingCallID(map[string]any{
+		"input": []any{map[string]any{"type": "function_call_output"}},
+	}))
+	require.False(t, HasFunctionCallOutputMissingCallID(map[string]any{
+		"input": []any{map[string]any{"type": "function_call_output", "call_id": "call_1"}},
+	}))
+}
+
+func TestHasItemReferenceForCallIDs(t *testing.T) {
+	// item_reference 需要覆盖所有 call_id 才视为可关联上下文。
+	require.False(t, HasItemReferenceForCallIDs(nil, []string{"call_1"}))
+	require.False(t, HasItemReferenceForCallIDs(map[string]any{}, []string{"call_1"}))
+	req := map[string]any{
+		"input": []any{
+			map[string]any{"type": "item_reference", "id": "call_1"},
+			map[string]any{"type": "item_reference", "id": "call_2"},
+		},
+	}
+	require.True(t, HasItemReferenceForCallIDs(req, []string{"call_1"}))
+	require.True(t, HasItemReferenceForCallIDs(req, []string{"call_1", "call_2"}))
+	require.False(t, HasItemReferenceForCallIDs(req, []string{"call_1", "call_3"}))
+}

backend/internal/service/prompts/codex_cli_instructions.md

@@ -0,0 +1,275 @@
+You are a coding agent running in the Codex CLI, a terminal-based coding assistant. Codex CLI is an open source project led by OpenAI. You are expected to be precise, safe, and helpful.
+
+Your capabilities:
+
+- Receive user prompts and other context provided by the harness, such as files in the workspace.
+- Communicate with the user by streaming thinking & responses, and by making & updating plans.
+- Emit function calls to run terminal commands and apply patches. Depending on how this specific run is configured, you can request that these function calls be escalated to the user for approval before running. More on this in the "Sandbox and approvals" section.
+
+Within this context, Codex refers to the open-source agentic coding interface (not the old Codex language model built by OpenAI).
+
+# How you work
+
+## Personality
+
+Your default personality and tone is concise, direct, and friendly. You communicate efficiently, always keeping the user clearly informed about ongoing actions without unnecessary detail. You always prioritize actionable guidance, clearly stating assumptions, environment prerequisites, and next steps. Unless explicitly asked, you avoid excessively verbose explanations about your work.
+
+# AGENTS.md spec
+- Repos often contain AGENTS.md files. These files can appear anywhere within the repository.
+- These files are a way for humans to give you (the agent) instructions or tips for working within the container.
+- Some examples might be: coding conventions, info about how code is organized, or instructions for how to run or test code.
+- Instructions in AGENTS.md files:
+    - The scope of an AGENTS.md file is the entire directory tree rooted at the folder that contains it.
+    - For every file you touch in the final patch, you must obey instructions in any AGENTS.md file whose scope includes that file.
+    - Instructions about code style, structure, naming, etc. apply only to code within the AGENTS.md file's scope, unless the file states otherwise.
+    - More-deeply-nested AGENTS.md files take precedence in the case of conflicting instructions.
+    - Direct system/developer/user instructions (as part of a prompt) take precedence over AGENTS.md instructions.
+- The contents of the AGENTS.md file at the root of the repo and any directories from the CWD up to the root are included with the developer message and don't need to be re-read. When working in a subdirectory of CWD, or a directory outside the CWD, check for any AGENTS.md files that may be applicable.
+
+## Responsiveness
+
+### Preamble messages
+
+Before making tool calls, send a brief preamble to the user explaining what you’re about to do. When sending preamble messages, follow these principles and examples:
+
+- **Logically group related actions**: if you’re about to run several related commands, describe them together in one preamble rather than sending a separate note for each.
+- **Keep it concise**: be no more than 1-2 sentences, focused on immediate, tangible next steps. (8–12 words for quick updates).
+- **Build on prior context**: if this is not your first tool call, use the preamble message to connect the dots with what’s been done so far and create a sense of momentum and clarity for the user to understand your next actions.
+- **Keep your tone light, friendly and curious**: add small touches of personality in preambles feel collaborative and engaging.
+- **Exception**: Avoid adding a preamble for every trivial read (e.g., `cat` a single file) unless it’s part of a larger grouped action.
+
+**Examples:**
+
+- “I’ve explored the repo; now checking the API route definitions.”
+- “Next, I’ll patch the config and update the related tests.”
+- “I’m about to scaffold the CLI commands and helper functions.”
+- “Ok cool, so I’ve wrapped my head around the repo. Now digging into the API routes.”
+- “Config’s looking tidy. Next up is patching helpers to keep things in sync.”
+- “Finished poking at the DB gateway. I will now chase down error handling.”
+- “Alright, build pipeline order is interesting. Checking how it reports failures.”
+- “Spotted a clever caching util; now hunting where it gets used.”
+
+## Planning
+
+You have access to an `update_plan` tool which tracks steps and progress and renders them to the user. Using the tool helps demonstrate that you've understood the task and convey how you're approaching it. Plans can help to make complex, ambiguous, or multi-phase work clearer and more collaborative for the user. A good plan should break the task into meaningful, logically ordered steps that are easy to verify as you go.
+
+Note that plans are not for padding out simple work with filler steps or stating the obvious. The content of your plan should not involve doing anything that you aren't capable of doing (i.e. don't try to test things that you can't test). Do not use plans for simple or single-step queries that you can just do or answer immediately.
+
+Do not repeat the full contents of the plan after an `update_plan` call — the harness already displays it. Instead, summarize the change made and highlight any important context or next step.
+
+Before running a command, consider whether or not you have completed the previous step, and make sure to mark it as completed before moving on to the next step. It may be the case that you complete all steps in your plan after a single pass of implementation. If this is the case, you can simply mark all the planned steps as completed. Sometimes, you may need to change plans in the middle of a task: call `update_plan` with the updated plan and make sure to provide an `explanation` of the rationale when doing so.
+
+Use a plan when:
+
+- The task is non-trivial and will require multiple actions over a long time horizon.
+- There are logical phases or dependencies where sequencing matters.
+- The work has ambiguity that benefits from outlining high-level goals.
+- You want intermediate checkpoints for feedback and validation.
+- When the user asked you to do more than one thing in a single prompt
+- The user has asked you to use the plan tool (aka "TODOs")
+- You generate additional steps while working, and plan to do them before yielding to the user
+
+### Examples
+
+**High-quality plans**
+
+Example 1:
+
+1. Add CLI entry with file args
+2. Parse Markdown via CommonMark library
+3. Apply semantic HTML template
+4. Handle code blocks, images, links
+5. Add error handling for invalid files
+
+Example 2:
+
+1. Define CSS variables for colors
+2. Add toggle with localStorage state
+3. Refactor components to use variables
+4. Verify all views for readability
+5. Add smooth theme-change transition
+
+Example 3:
+
+1. Set up Node.js + WebSocket server
+2. Add join/leave broadcast events
+3. Implement messaging with timestamps
+4. Add usernames + mention highlighting
+5. Persist messages in lightweight DB
+6. Add typing indicators + unread count
+
+**Low-quality plans**
+
+Example 1:
+
+1. Create CLI tool
+2. Add Markdown parser
+3. Convert to HTML
+
+Example 2:
+
+1. Add dark mode toggle
+2. Save preference
+3. Make styles look good
+
+Example 3:
+
+1. Create single-file HTML game
+2. Run quick sanity check
+3. Summarize usage instructions
+
+If you need to write a plan, only write high quality plans, not low quality ones.
+
+## Task execution
+
+You are a coding agent. Please keep going until the query is completely resolved, before ending your turn and yielding back to the user. Only terminate your turn when you are sure that the problem is solved. Autonomously resolve the query to the best of your ability, using the tools available to you, before coming back to the user. Do NOT guess or make up an answer.
+
+You MUST adhere to the following criteria when solving queries:
+
+- Working on the repo(s) in the current environment is allowed, even if they are proprietary.
+- Analyzing code for vulnerabilities is allowed.
+- Showing user code and tool call details is allowed.
+- Use the `apply_patch` tool to edit files (NEVER try `applypatch` or `apply-patch`, only `apply_patch`): {"command":["apply_patch","*** Begin Patch\\n*** Update File: path/to/file.py\\n@@ def example():\\n- pass\\n+ return 123\\n*** End Patch"]}
+
+If completing the user's task requires writing or modifying files, your code and final answer should follow these coding guidelines, though user instructions (i.e. AGENTS.md) may override these guidelines:
+
+- Fix the problem at the root cause rather than applying surface-level patches, when possible.
+- Avoid unneeded complexity in your solution.
+- Do not attempt to fix unrelated bugs or broken tests. It is not your responsibility to fix them. (You may mention them to the user in your final message though.)
+- Update documentation as necessary.
+- Keep changes consistent with the style of the existing codebase. Changes should be minimal and focused on the task.
+- Use `git log` and `git blame` to search the history of the codebase if additional context is required.
+- NEVER add copyright or license headers unless specifically requested.
+- Do not waste tokens by re-reading files after calling `apply_patch` on them. The tool call will fail if it didn't work. The same goes for making folders, deleting folders, etc.
+- Do not `git commit` your changes or create new git branches unless explicitly requested.
+- Do not add inline comments within code unless explicitly requested.
+- Do not use one-letter variable names unless explicitly requested.
+- NEVER output inline citations like "【F:README.md†L5-L14】" in your outputs. The CLI is not able to render these so they will just be broken in the UI. Instead, if you output valid filepaths, users will be able to click on them to open the files in their editor.
+
+## Validating your work
+
+If the codebase has tests or the ability to build or run, consider using them to verify that your work is complete. 
+
+When testing, your philosophy should be to start as specific as possible to the code you changed so that you can catch issues efficiently, then make your way to broader tests as you build confidence. If there's no test for the code you changed, and if the adjacent patterns in the codebases show that there's a logical place for you to add a test, you may do so. However, do not add tests to codebases with no tests.
+
+Similarly, once you're confident in correctness, you can suggest or use formatting commands to ensure that your code is well formatted. If there are issues you can iterate up to 3 times to get formatting right, but if you still can't manage it's better to save the user time and present them a correct solution where you call out the formatting in your final message. If the codebase does not have a formatter configured, do not add one.
+
+For all of testing, running, building, and formatting, do not attempt to fix unrelated bugs. It is not your responsibility to fix them. (You may mention them to the user in your final message though.)
+
+Be mindful of whether to run validation commands proactively. In the absence of behavioral guidance:
+
+- When running in non-interactive approval modes like **never** or **on-failure**, proactively run tests, lint and do whatever you need to ensure you've completed the task.
+- When working in interactive approval modes like **untrusted**, or **on-request**, hold off on running tests or lint commands until the user is ready for you to finalize your output, because these commands take time to run and slow down iteration. Instead suggest what you want to do next, and let the user confirm first.
+- When working on test-related tasks, such as adding tests, fixing tests, or reproducing a bug to verify behavior, you may proactively run tests regardless of approval mode. Use your judgement to decide whether this is a test-related task.
+
+## Ambition vs. precision
+
+For tasks that have no prior context (i.e. the user is starting something brand new), you should feel free to be ambitious and demonstrate creativity with your implementation.
+
+If you're operating in an existing codebase, you should make sure you do exactly what the user asks with surgical precision. Treat the surrounding codebase with respect, and don't overstep (i.e. changing filenames or variables unnecessarily). You should balance being sufficiently ambitious and proactive when completing tasks of this nature.
+
+You should use judicious initiative to decide on the right level of detail and complexity to deliver based on the user's needs. This means showing good judgment that you're capable of doing the right extras without gold-plating. This might be demonstrated by high-value, creative touches when scope of the task is vague; while being surgical and targeted when scope is tightly specified.
+
+## Sharing progress updates
+
+For especially longer tasks that you work on (i.e. requiring many tool calls, or a plan with multiple steps), you should provide progress updates back to the user at reasonable intervals. These updates should be structured as a concise sentence or two (no more than 8-10 words long) recapping progress so far in plain language: this update demonstrates your understanding of what needs to be done, progress so far (i.e. files explores, subtasks complete), and where you're going next.
+
+Before doing large chunks of work that may incur latency as experienced by the user (i.e. writing a new file), you should send a concise message to the user with an update indicating what you're about to do to ensure they know what you're spending time on. Don't start editing or writing large files before informing the user what you are doing and why.
+
+The messages you send before tool calls should describe what is immediately about to be done next in very concise language. If there was previous work done, this preamble message should also include a note about the work done so far to bring the user along.
+
+## Presenting your work and final message
+
+Your final message should read naturally, like an update from a concise teammate. For casual conversation, brainstorming tasks, or quick questions from the user, respond in a friendly, conversational tone. You should ask questions, suggest ideas, and adapt to the user’s style. If you've finished a large amount of work, when describing what you've done to the user, you should follow the final answer formatting guidelines to communicate substantive changes. You don't need to add structured formatting for one-word answers, greetings, or purely conversational exchanges.
+
+You can skip heavy formatting for single, simple actions or confirmations. In these cases, respond in plain sentences with any relevant next step or quick option. Reserve multi-section structured responses for results that need grouping or explanation.
+
+The user is working on the same computer as you, and has access to your work. As such there's no need to show the full contents of large files you have already written unless the user explicitly asks for them. Similarly, if you've created or modified files using `apply_patch`, there's no need to tell users to "save the file" or "copy the code into a file"—just reference the file path.
+
+If there's something that you think you could help with as a logical next step, concisely ask the user if they want you to do so. Good examples of this are running tests, committing changes, or building out the next logical component. If there’s something that you couldn't do (even with approval) but that the user might want to do (such as verifying changes by running the app), include those instructions succinctly.
+
+Brevity is very important as a default. You should be very concise (i.e. no more than 10 lines), but can relax this requirement for tasks where additional detail and comprehensiveness is important for the user's understanding.
+
+### Final answer structure and style guidelines
+
+You are producing plain text that will later be styled by the CLI. Follow these rules exactly. Formatting should make results easy to scan, but not feel mechanical. Use judgment to decide how much structure adds value.
+
+**Section Headers**
+
+- Use only when they improve clarity — they are not mandatory for every answer.
+- Choose descriptive names that fit the content
+- Keep headers short (1–3 words) and in `**Title Case**`. Always start headers with `**` and end with `**`
+- Leave no blank line before the first bullet under a header.
+- Section headers should only be used where they genuinely improve scanability; avoid fragmenting the answer.
+
+**Bullets**
+
+- Use `-` followed by a space for every bullet.
+- Merge related points when possible; avoid a bullet for every trivial detail.
+- Keep bullets to one line unless breaking for clarity is unavoidable.
+- Group into short lists (4–6 bullets) ordered by importance.
+- Use consistent keyword phrasing and formatting across sections.
+
+**Monospace**
+
+- Wrap all commands, file paths, env vars, and code identifiers in backticks (`` `...` ``).
+- Apply to inline examples and to bullet keywords if the keyword itself is a literal file/command.
+- Never mix monospace and bold markers; choose one based on whether it’s a keyword (`**`) or inline code/path (`` ` ``).
+
+**File References**
+When referencing files in your response, make sure to include the relevant start line and always follow the below rules:
+  * Use inline code to make file paths clickable.
+  * Each reference should have a stand alone path. Even if it's the same file.
+  * Accepted: absolute, workspace‑relative, a/ or b/ diff prefixes, or bare filename/suffix.
+  * Line/column (1‑based, optional): :line[:column] or #Lline[Ccolumn] (column defaults to 1).
+  * Do not use URIs like file://, vscode://, or https://.
+  * Do not provide range of lines
+  * Examples: src/app.ts, src/app.ts:42, b/server/index.js#L10, C:\repo\project\main.rs:12:5
+
+**Structure**
+
+- Place related bullets together; don’t mix unrelated concepts in the same section.
+- Order sections from general → specific → supporting info.
+- For subsections (e.g., “Binaries” under “Rust Workspace”), introduce with a bolded keyword bullet, then list items under it.
+- Match structure to complexity:
+  - Multi-part or detailed results → use clear headers and grouped bullets.
+  - Simple results → minimal headers, possibly just a short list or paragraph.
+
+**Tone**
+
+- Keep the voice collaborative and natural, like a coding partner handing off work.
+- Be concise and factual — no filler or conversational commentary and avoid unnecessary repetition
+- Use present tense and active voice (e.g., “Runs tests” not “This will run tests”).
+- Keep descriptions self-contained; don’t refer to “above” or “below”.
+- Use parallel structure in lists for consistency.
+
+**Don’t**
+
+- Don’t use literal words “bold” or “monospace” in the content.
+- Don’t nest bullets or create deep hierarchies.
+- Don’t output ANSI escape codes directly — the CLI renderer applies them.
+- Don’t cram unrelated keywords into a single bullet; split for clarity.
+- Don’t let keyword lists run long — wrap or reformat for scanability.
+
+Generally, ensure your final answers adapt their shape and depth to the request. For example, answers to code explanations should have a precise, structured explanation with code references that answer the question directly. For tasks with a simple implementation, lead with the outcome and supplement only with what’s needed for clarity. Larger changes can be presented as a logical walkthrough of your approach, grouping related steps, explaining rationale where it adds value, and highlighting next actions to accelerate the user. Your answers should provide the right level of detail while being easily scannable.
+
+For casual greetings, acknowledgements, or other one-off conversational messages that are not delivering substantive information or structured results, respond naturally without section headers or bullet formatting.
+
+# Tool Guidelines
+
+## Shell commands
+
+When using the shell, you must adhere to the following guidelines:
+
+- When searching for text or files, prefer using `rg` or `rg --files` respectively because `rg` is much faster than alternatives like `grep`. (If the `rg` command is not found, then use alternatives.)
+- Do not use python scripts to attempt to output larger chunks of a file.
+
+## `update_plan`
+
+A tool named `update_plan` is available to you. You can use it to keep an up‑to‑date, step‑by‑step plan for the task.
+
+To create a new plan, call `update_plan` with a short list of 1‑sentence steps (no more than 5-7 words each) with a `status` for each step (`pending`, `in_progress`, or `completed`).
+
+When steps have been completed, use `update_plan` to mark each finished step as `completed` and the next step you are working on as `in_progress`. There should always be exactly one `in_progress` step until everything is done. You can mark multiple items as complete in a single `update_plan` call.
+
+If all steps are complete, ensure you call `update_plan` to mark all steps as `completed`.

frontend/src/components/admin/user/UserBalanceModal.vue

@@ -3,14 +3,17 @@
     <form v-if="user" id="balance-form" @submit.prevent="handleBalanceSubmit" class="space-y-5">
       <div class="flex items-center gap-3 rounded-xl bg-gray-50 p-4 dark:bg-dark-700">
         <div class="flex h-10 w-10 items-center justify-center rounded-full bg-primary-100"><span class="text-lg font-medium text-primary-700">{{ user.email.charAt(0).toUpperCase() }}</span></div>
-        <div class="flex-1"><p class="font-medium text-gray-900">{{ user.email }}</p><p class="text-sm text-gray-500">{{ t('admin.users.currentBalance') }}: ${{ user.balance.toFixed(2) }}</p></div>
+        <div class="flex-1"><p class="font-medium text-gray-900">{{ user.email }}</p><p class="text-sm text-gray-500">{{ t('admin.users.currentBalance') }}: ${{ formatBalance(user.balance) }}</p></div>
       </div>
       <div>
         <label class="input-label">{{ operation === 'add' ? t('admin.users.depositAmount') : t('admin.users.withdrawAmount') }}</label>
-        <div class="relative"><div class="absolute left-3 top-1/2 -translate-y-1/2 font-medium text-gray-500">$</div><input v-model.number="form.amount" type="number" step="0.01" min="0.01" required class="input pl-8" /></div>
+        <div class="relative flex gap-2">
+          <div class="relative flex-1"><div class="absolute left-3 top-1/2 -translate-y-1/2 font-medium text-gray-500">$</div><input v-model.number="form.amount" type="number" step="any" min="0" required class="input pl-8" /></div>
+          <button v-if="operation === 'subtract'" type="button" @click="fillAllBalance" class="btn btn-secondary whitespace-nowrap">{{ t('admin.users.withdrawAll') }}</button>
+        </div>
       </div>
       <div><label class="input-label">{{ t('admin.users.notes') }}</label><textarea v-model="form.notes" rows="3" class="input"></textarea></div>
-      <div v-if="form.amount > 0" class="rounded-xl border border-blue-200 bg-blue-50 p-4"><div class="flex items-center justify-between text-sm"><span>{{ t('admin.users.newBalance') }}:</span><span class="font-bold">${{ calculateNewBalance().toFixed(2) }}</span></div></div>
+      <div v-if="form.amount > 0" class="rounded-xl border border-blue-200 bg-blue-50 p-4"><div class="flex items-center justify-between text-sm"><span>{{ t('admin.users.newBalance') }}:</span><span class="font-bold">${{ formatBalance(calculateNewBalance()) }}</span></div></div>
     </form>
     <template #footer>
       <div class="flex justify-end gap-3">
@@ -35,11 +38,30 @@ const emit = defineEmits(['close', 'success']); const { t } = useI18n(); const a
 const submitting = ref(false); const form = reactive({ amount: 0, notes: '' })
 watch(() => props.show, (v) => { if(v) { form.amount = 0; form.notes = '' } })
 
+// 格式化余额：显示完整精度，去除尾部多余的0
+const formatBalance = (value: number) => {
+  if (value === 0) return '0.00'
+  // 最多保留8位小数，去除尾部的0
+  const formatted = value.toFixed(8).replace(/\.?0+$/, '')
+  // 确保至少有2位小数
+  const parts = formatted.split('.')
+  if (parts.length === 1) return formatted + '.00'
+  if (parts[1].length === 1) return formatted + '0'
+  return formatted
+}
+
+// 填入全部余额
+const fillAllBalance = () => {
+  if (props.user) {
+    form.amount = props.user.balance
+  }
+}
+
 const calculateNewBalance = () => {
   if (!props.user) return 0
   const result = props.operation === 'add' ? props.user.balance + form.amount : props.user.balance - form.amount
   // 避免浮点数精度问题导致的 -0.00 显示
-  return result === 0 || Object.is(result, -0) ? 0 : result
+  return Math.abs(result) < 1e-10 ? 0 : result
 }
 const handleBalanceSubmit = async () => {
   if (!props.user) return
@@ -47,10 +69,8 @@ const handleBalanceSubmit = async () => {
     appStore.showError(t('admin.users.amountRequired'))
     return
   }
-  // 使用小数点后两位精度比较，避免浮点数精度问题
+  // 退款时验证金额不超过实际余额
-  const amount = Math.round(form.amount * 100) / 100
+  if (props.operation === 'subtract' && form.amount > props.user.balance) {
-  const balance = Math.round(props.user.balance * 100) / 100
-  if (props.operation === 'subtract' && amount > balance) {
     appStore.showError(t('admin.users.insufficientBalance'))
     return
   }

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,
+        apiKey
-        ...(platform === 'openai' ? { store: false } : {})
       }
     }
   }
   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,

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 = [

frontend/src/i18n/locales/en.ts

@@ -390,7 +390,7 @@ export default {
       opencode: {
         title: 'OpenCode Example',
         subtitle: 'opencode.json',
-        hint: 'Config path: ~/.config/opencode/opencode.json (create if not exists). This is an 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',
@@ -724,6 +724,7 @@ export default {
       withdraw: 'Withdraw',
       depositAmount: 'Deposit Amount',
       withdrawAmount: 'Withdraw Amount',
+      withdrawAll: 'All',
       currentBalance: 'Current Balance',
       depositNotesPlaceholder:
         'e.g., New user registration bonus, promotional credit, compensation, etc.',
@@ -1943,6 +1944,9 @@ 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',

frontend/src/i18n/locales/zh.ts

@@ -387,7 +387,7 @@ export default {
       opencode: {
         title: 'OpenCode 配置示例',
         subtitle: 'opencode.json',
-        hint: '配置文件路径：~/.config/opencode/opencode.json，不存在需手动创建。示例仅供参考，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',
@@ -780,6 +780,7 @@ export default {
       withdraw: '退款',
       depositAmount: '充值金额',
       withdrawAmount: '退款金额',
+      withdrawAll: '全部',
       depositNotesPlaceholder: '例如：新用户注册奖励、活动充值、补偿充值等',
       withdrawNotesPlaceholder: '例如：服务问题退款、错误充值退回、账户注销退款等',
       notesOptional: '备注为可选项，有助于未来查账',
@@ -2088,6 +2089,9 @@ 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',

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;
   }
 

frontend/src/views/admin/ops/OpsDashboard.vue

@@ -1,6 +1,6 @@
 <template>
-  <AppLayout>
+  <component :is="isFullscreen ? 'div' : AppLayout" :class="isFullscreen ? 'flex min-h-screen flex-col justify-center bg-gray-50 dark:bg-dark-950' : ''">
-    <div class="space-y-6 pb-12">
+    <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"
@@ -22,6 +22,7 @@
         :thresholds="metricThresholds"
         :auto-refresh-enabled="autoRefreshEnabled"
         :auto-refresh-countdown="autoRefreshCountdown"
+        :fullscreen="isFullscreen"
         @update:time-range="onTimeRangeChange"
         @update:platform="onPlatformChange"
         @update:group="onGroupChange"
@@ -31,6 +32,8 @@
         @open-error-details="openErrorDetails"
         @open-settings="showSettingsDialog = true"
         @open-alert-rules="showAlertRulesCard = true"
+        @enter-fullscreen="enterFullscreen"
+        @exit-fullscreen="exitFullscreen"
       />
 
       <!-- Row: Concurrency + Throughput -->
@@ -45,6 +48,7 @@
             :top-groups="throughputTrend?.top_groups ?? []"
             :loading="loadingTrend"
             :time-range="timeRange"
+            :fullscreen="isFullscreen"
             @select-platform="handleThroughputSelectPlatform"
             @select-group="handleThroughputSelectGroup"
             @open-details="handleOpenRequestDetails"
@@ -72,36 +76,37 @@
       <!-- Alert Events -->
       <OpsAlertEventsCard v-if="opsEnabled && !(loading && !hasLoadedOnce)" />
 
-      <!-- Settings Dialog -->
+      <!-- Settings Dialog (hidden in fullscreen mode) -->
-      <OpsSettingsDialog :show="showSettingsDialog" @close="showSettingsDialog = false" @saved="onSettingsSaved" />
+      <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">
-      <BaseDialog :show="showAlertRulesCard" :title="t('admin.ops.alertRules.title')" width="extra-wide" @close="showAlertRulesCard = false">
+          <OpsAlertRulesCard />
-        <OpsAlertRulesCard />
+        </BaseDialog>
-      </BaseDialog>
 
-      <OpsErrorDetailsModal
+        <OpsErrorDetailsModal
-        :show="showErrorDetails"
+          :show="showErrorDetails"
-        :time-range="timeRange"
+          :time-range="timeRange"
-        :platform="platform"
+          :platform="platform"
-        :group-id="groupId"
+          :group-id="groupId"
-        :error-type="errorDetailsType"
+          :error-type="errorDetailsType"
-        @update:show="showErrorDetails = $event"
+          @update:show="showErrorDetails = $event"
-        @openErrorDetail="openError"
+          @openErrorDetail="openError"
-      />
+        />
 
-      <OpsErrorDetailModal v-model:show="showErrorModal" :error-id="selectedErrorId" />
+        <OpsErrorDetailModal v-model:show="showErrorModal" :error-id="selectedErrorId" />
 
-      <OpsRequestDetailsModal
+        <OpsRequestDetailsModal
-        v-model="showRequestDetails"
+          v-model="showRequestDetails"
-        :time-range="timeRange"
+          :time-range="timeRange"
-        :preset="requestDetailsPreset"
+          :preset="requestDetailsPreset"
-        :platform="platform"
+          :platform="platform"
-        :group-id="groupId"
+          :group-id="groupId"
-        @openErrorDetail="openError"
+          @openErrorDetail="openError"
-      />
+        />
+      </template>
     </div>
-  </AppLayout>
+  </component>
 </template>
 
 <script setup lang="ts">
@@ -163,12 +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
 
@@ -603,6 +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')
@@ -637,6 +669,7 @@ async function loadThresholds() {
 }
 
 onUnmounted(() => {
+  window.removeEventListener('keydown', handleKeydown)
   abortDashboardFetch()
   pauseAutoRefresh()
   pauseCountdown()

frontend/src/views/admin/ops/components/OpsDashboardHeader.vue

@@ -25,6 +25,7 @@ interface Props {
   thresholds?: OpsMetricThresholds | null // 阈值配置
   autoRefreshEnabled?: boolean
   autoRefreshCountdown?: number
+  fullscreen?: boolean
 }
 
 interface Emits {
@@ -37,6 +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>()
@@ -391,15 +394,15 @@ const healthScoreClass = computed(() => {
   return 'text-red-500'
 })
 
-const circleSize = 100
+const circleSize = computed(() => props.fullscreen ? 140 : 100)
-const strokeWidth = 8
+const strokeWidth = computed(() => props.fullscreen ? 10 : 8)
-const radius = (circleSize - strokeWidth) / 2
+const radius = computed(() => (circleSize.value - strokeWidth.value) / 2)
-const circumference = 2 * Math.PI * radius
+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 {
@@ -814,7 +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>
@@ -830,7 +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>
@@ -863,28 +866,30 @@ function handleToolbarRefresh() {
       </div>
 
       <div class="flex flex-wrap items-center gap-3">
-        <Select
+        <template v-if="!props.fullscreen">
-          :model-value="platform"
+          <Select
-          :options="platformOptions"
+            :model-value="platform"
-          class="w-full sm:w-[140px]"
+            :options="platformOptions"
-          @update:model-value="handlePlatformChange"
+            class="w-full sm:w-[140px]"
-        />
+            @update:model-value="handlePlatformChange"
+          />
 
-        <Select
+          <Select
-          :model-value="groupId"
+            :model-value="groupId"
-          :options="groupOptions"
+            :options="groupOptions"
-          class="w-full sm:w-[160px]"
+            class="w-full sm:w-[160px]"
-          @update:model-value="handleGroupChange"
+            @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
+          <Select
-          :model-value="timeRange"
+            :model-value="timeRange"
-          :options="timeRangeOptions"
+            :options="timeRangeOptions"
-          class="relative w-full sm:w-[150px]"
+            class="relative w-full sm:w-[150px]"
-          @update:model-value="handleTimeRangeChange"
+            @update:model-value="handleTimeRangeChange"
-        />
+          />
+        </template>
 
         <Select
           v-if="false"
@@ -895,6 +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"
@@ -911,9 +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')"
@@ -925,7 +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')"
@@ -937,13 +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="['rounded-2xl bg-gray-50 dark:bg-dark-900 lg:col-span-5', props.fullscreen ? 'p-6' : 'p-4']">
-        <div class="grid grid-cols-1 gap-6 md:grid-cols-[200px_1fr] md:items-center">
+        <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"
@@ -1026,14 +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')" />
@@ -1051,7 +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">
@@ -1059,7 +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 -->
@@ -1079,18 +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="[props.fullscreen ? 'text-4xl' : 'text-xl sm:text-2xl', 'font-black text-gray-900 dark:text-white']">{{ displayRealTimeQps.toFixed(1) }}</span>
-                    <span class="text-xs font-bold text-gray-500">QPS</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="[props.fullscreen ? 'text-4xl' : 'text-xl sm:text-2xl', 'font-black text-gray-900 dark:text-white']">{{ displayRealTimeTps.toFixed(1) }}</span>
-                    <span class="text-xs font-bold text-gray-500">TPS</span>
+                    <span :class="[props.fullscreen ? 'text-sm' : 'text-xs', 'font-bold text-gray-500']">TPS</span>
                   </div>
                 </div>
               </div>
@@ -1099,8 +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="[props.fullscreen ? 'text-xs' : '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-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>
@@ -1114,8 +1137,8 @@ function handleToolbarRefresh() {
 
                 <!-- Average -->
                 <div>
-                  <div class="text-[10px] font-bold uppercase text-gray-400">{{ t('admin.ops.average') }}</div>
+                  <div :class="[props.fullscreen ? 'text-xs' : '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-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>
@@ -1156,15 +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') })"
@@ -1197,10 +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' })"
@@ -1227,9 +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) })"
@@ -1277,9 +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' })"
@@ -1327,9 +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>
@@ -1353,9 +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>
@@ -1383,12 +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 ? '-' : `${cpuPercentValue.toFixed(1)}%` }}
           </div>
-          <div class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
+          <div v-if="!props.fullscreen" class="mt-1 text-[10px] text-gray-500 dark:text-gray-400">
             {{ t('common.warning') }} 80% · {{ t('common.critical') }} 95%
           </div>
         </div>
@@ -1397,12 +1424,12 @@ function handleToolbarRefresh() {
         <div class="rounded-xl bg-gray-50 p-3 dark:bg-dark-900">
           <div class="flex items-center gap-1">
             <div class="text-[10px] font-bold uppercase tracking-wider text-gray-400">MEM</div>
-            <HelpTooltip :content="t('admin.ops.tooltips.memory')" />
+            <HelpTooltip v-if="!props.fullscreen" :content="t('admin.ops.tooltips.memory')" />
           </div>
           <div class="mt-1 text-lg font-black" :class="memPercentClass">
             {{ memPercentValue == null ? '-' : `${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
                 ? '-'
@@ -1415,12 +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 ?? '-' }}
@@ -1432,12 +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>
@@ -1448,12 +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>
@@ -1468,9 +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>
@@ -1479,7 +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>

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
+        <template v-if="!props.fullscreen">
-          type="button"
+          <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"
+            type="button"
-          :disabled="state !== 'ready'"
+            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"
-          :title="t('admin.ops.requestDetails.title')"
+            :disabled="state !== 'ready'"
-          @click="emit('openDetails')"
+            :title="t('admin.ops.requestDetails.title')"
-        >
+            @click="emit('openDetails')"
-          {{ t('admin.ops.requestDetails.details') }}
+          >
-        </button>
+            {{ t('admin.ops.requestDetails.details') }}
-        <button
+          </button>
-          type="button"
+          <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"
+            type="button"
-          :disabled="state !== 'ready'"
+            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"
-          :title="t('admin.ops.charts.resetZoomHint')"
+            :disabled="state !== 'ready'"
-          @click="resetZoom"
+            :title="t('admin.ops.charts.resetZoomHint')"
-        >
+            @click="resetZoom"
-          {{ t('admin.ops.charts.resetZoom') }}
+          >
-        </button>
+            {{ t('admin.ops.charts.resetZoom') }}
-        <button
+          </button>
-          type="button"
+          <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"
+            type="button"
-          :disabled="state !== 'ready'"
+            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"
-          :title="t('admin.ops.charts.downloadChartHint')"
+            :disabled="state !== 'ready'"
-          @click="downloadChart"
+            :title="t('admin.ops.charts.downloadChartHint')"
-        >
+            @click="downloadChart"
-          {{ t('admin.ops.charts.downloadChart') }}
+          >
-        </button>
+            {{ t('admin.ops.charts.downloadChart') }}
+          </button>
+        </template>
       </div>
     </div>