feat: Agent CLI 与终端审计回放全面优化

- Agent CLI: 多轮对话、SSH 长连接、流式推理、Evolver 按用户隔离、历史侧栏
- 终端审计: 回放播放器、命令书签、筛选分页导出、命令输出统计、录制落盘修复
- 修复 session_commands 表迁移、Kill 竞态导致录制丢失、无录制会话回放提示
- 审计历史页只读（移除删除），移除回放弹框录制搜索

backend/controller/llm_chat.go

@@ -14,18 +14,6 @@ import (
 	"github.com/gin-gonic/gin"
 )
 
-// buildLLMClient 从数据库配置构造 LLM 客户端（按 provider 协议自动路由 OpenAI/Anthropic） // Logan.Li
-func (h *LLMHandler) buildLLMClient(cfg model.LLMConfig) llm.Client {
-	return llm.NewClient(llm.Config{
-		Protocol:    resolveProtocol(cfg.Provider),
-		Endpoint:    resolveEndpoint(cfg.Provider, cfg.Endpoint),
-		APIKey:      cfg.APIKey,
-		Model:       cfg.Model,
-		Temperature: cfg.Temperature,
-		MaxTokens:   cfg.MaxTokens,
-	})
-}
-
 // toLLMMessages model.ChatMessage[] → llm.Message[]（智能助手纯对话，仅填 Role/Content） // Logan.Li
 func toLLMMessages(msgs []model.ChatMessage) []llm.Message {
 	out := make([]llm.Message, len(msgs))
@@ -57,7 +45,7 @@ func (h *LLMHandler) Chat(c *gin.Context) {
 	}
 	messages := append([]model.ChatMessage{{Role: "system", Content: systemPrompt}}, req.Messages...)
 
-	resp, err := h.buildLLMClient(cfg).Complete(c.Request.Context(), &llm.CompletionRequest{
+	resp, err := buildLLMClientFromConfig(cfg).Complete(c.Request.Context(), &llm.CompletionRequest{
 		Messages: toLLMMessages(messages),
 	})
 	if err != nil {
@@ -127,7 +115,7 @@ func (h *LLMHandler) ChatStream(c *gin.Context) {
 
 	// 流式补全：逐 token 回调转发 SSE；ctx 取消（客户端断开）由 llm.Client 内部终止上游读取
 	var contentBuilder strings.Builder
-	_, streamErr := h.buildLLMClient(cfg).Stream(c.Request.Context(), &llm.CompletionRequest{
+	_, streamErr := buildLLMClientFromConfig(cfg).Stream(c.Request.Context(), &llm.CompletionRequest{
 		Messages: toLLMMessages(messages),
 	}, func(delta string) {
 		contentBuilder.WriteString(delta)
@@ -188,7 +176,7 @@ func writeSSEError(c *gin.Context, code int, message string) {
 	c.Writer.Flush()
 }
 
-// TestLLMConnection 测试 LLM 连接（发送一条 ping 消息） // Logan.Li
+// TestLLMConnection 测试 LLM 连接（询问模型身份并返回回答） // Logan.Li
 // POST /api/v1/aiops/llm-config/test
 func (h *LLMHandler) TestLLMConnection(c *gin.Context) {
 	var cfg model.LLMConfig
@@ -196,12 +184,22 @@ func (h *LLMHandler) TestLLMConnection(c *gin.Context) {
 		c.JSON(http.StatusOK, model.Response{Code: 400, Message: "请先保存配置"})
 		return
 	}
-	_, err := h.buildLLMClient(cfg).Complete(c.Request.Context(), &llm.CompletionRequest{
-		Messages: []llm.Message{{Role: "user", Content: "ping"}},
+	if !cfg.Enabled {
+		c.JSON(http.StatusOK, model.Response{Code: 400, Message: "请先启用 LLM 服务"})
+		return
+	}
+	resp, err := buildLLMClientFromConfig(cfg).Complete(c.Request.Context(), &llm.CompletionRequest{
+		Messages: []llm.Message{{Role: "user", Content: "你是什么模型？请简短回答。"}},
 	})
 	if err != nil {
 		c.JSON(http.StatusOK, model.Response{Code: 502, Message: "连接失败: " + err.Error()})
 		return
 	}
-	c.JSON(http.StatusOK, model.Response{Code: 200, Message: "连接成功"})
+	c.JSON(http.StatusOK, model.Response{Code: 200, Message: "连接成功", Data: gin.H{
+		"answer":        resp.Message.Content,
+		"model":         resp.Model,
+		"tokens_used":   resp.Usage.TotalTokens,
+		"provider":      cfg.Provider,
+		"reasoning_enabled": cfg.ReasoningEnabled,
+	}})
 }