feat: add group monitoring module

controller/channel-test.go

@@ -862,6 +862,29 @@ func TestAllChannels(c *gin.Context) {
 	})
 }
 
+// MonitorTestResult holds the result of a channel test for the group monitor module.
+type MonitorTestResult struct {
+	Success   bool
+	LatencyMs int64
+	ErrorMsg  string
+}
+
+// TestChannelForMonitor runs a non-streaming test on the given channel and returns a simplified result.
+func TestChannelForMonitor(channel *model.Channel, testModel string) MonitorTestResult {
+	tik := time.Now()
+	result := testChannel(channel, testModel, "", false)
+	latencyMs := time.Since(tik).Milliseconds()
+	r := MonitorTestResult{LatencyMs: latencyMs}
+	if result.localErr != nil {
+		r.ErrorMsg = result.localErr.Error()
+	} else if result.newAPIError != nil {
+		r.ErrorMsg = result.newAPIError.Error()
+	} else {
+		r.Success = true
+	}
+	return r
+}
+
 var autoTestChannelsOnce sync.Once
 
 func AutomaticallyTestChannels() {

main.go

@@ -20,6 +20,7 @@ import (
 	"github.com/QuantumNous/new-api/model"
 	"github.com/QuantumNous/new-api/oauth"
 	"github.com/QuantumNous/new-api/router"
+	"github.com/QuantumNous/new-api/modules/group_monitor"
 	"github.com/QuantumNous/new-api/service"
 	_ "github.com/QuantumNous/new-api/setting/performance_setting"
 	"github.com/QuantumNous/new-api/setting/ratio_setting"
@@ -104,6 +105,7 @@ func main() {
 	}
 
 	go controller.AutomaticallyTestChannels()
+	go group_monitor.AutomaticallyGroupMonitor()
 
 	// Codex credential auto-refresh check every 10 minutes, refresh when expires within 1 day
 	service.StartCodexCredentialAutoRefreshTask()
@@ -299,5 +301,8 @@ func InitResources() error {
 		// Don't return error, custom OAuth is not critical
 	}
 
+	// Migrate group monitor tables
+	group_monitor.Migrate()
+
 	return nil
 }

modules/group_monitor/controller.go

@@ -0,0 +1,181 @@
+package group_monitor
+
+import (
+	"net/http"
+	"strconv"
+
+	"github.com/QuantumNous/new-api/common"
+
+	"github.com/gin-gonic/gin"
+)
+
+// GetGroupMonitorLogs 分页查询监控日志（管理员）
+func GetGroupMonitorLogsHandler(c *gin.Context) {
+	pageInfo := common.GetPageQuery(c)
+	groupName := c.Query("group")
+	startTs, _ := strconv.ParseInt(c.Query("start_timestamp"), 10, 64)
+	endTs, _ := strconv.ParseInt(c.Query("end_timestamp"), 10, 64)
+
+	logs, total, err := GetGroupMonitorLogs(groupName, startTs, endTs, pageInfo.GetStartIdx(), pageInfo.GetPageSize())
+	if err != nil {
+		common.ApiError(c, err)
+		return
+	}
+	pageInfo.SetTotal(int(total))
+	pageInfo.SetItems(logs)
+	common.ApiSuccess(c, pageInfo)
+}
+
+// GetGroupMonitorLatestHandler 获取所有分组最新状态（管理员）
+func GetGroupMonitorLatestHandler(c *gin.Context) {
+	logs, err := GetGroupMonitorLatest()
+	if err != nil {
+		common.ApiError(c, err)
+		return
+	}
+	common.ApiSuccess(c, logs)
+}
+
+// GetGroupMonitorStatsHandler 获取聚合统计（管理员）
+func GetGroupMonitorStatsHandler(c *gin.Context) {
+	startTs, _ := strconv.ParseInt(c.Query("start_timestamp"), 10, 64)
+	endTs, _ := strconv.ParseInt(c.Query("end_timestamp"), 10, 64)
+
+	// 默认查询最近 1 小时
+	if startTs == 0 {
+		startTs = common.GetTimestamp() - 3600
+	}
+
+	stats, err := GetGroupMonitorStats(startTs, endTs)
+	if err != nil {
+		common.ApiError(c, err)
+		return
+	}
+	common.ApiSuccess(c, stats)
+}
+
+// GetGroupMonitorTimeSeriesHandler 获取时间序列数据（趋势图）
+func GetGroupMonitorTimeSeriesHandler(c *gin.Context) {
+	groupName := c.Query("group")
+	startTs, _ := strconv.ParseInt(c.Query("start_timestamp"), 10, 64)
+	endTs, _ := strconv.ParseInt(c.Query("end_timestamp"), 10, 64)
+
+	// 默认最近 1 小时
+	if startTs == 0 {
+		startTs = common.GetTimestamp() - 3600
+	}
+
+	logs, err := GetGroupMonitorTimeSeries(groupName, startTs, endTs)
+	if err != nil {
+		common.ApiError(c, err)
+		return
+	}
+	common.ApiSuccess(c, logs)
+}
+
+// GetGroupMonitorConfigsHandler 获取所有分组监控配置（管理员）
+func GetGroupMonitorConfigsHandler(c *gin.Context) {
+	configs, err := GetAllGroupMonitorConfigs()
+	if err != nil {
+		common.ApiError(c, err)
+		return
+	}
+	common.ApiSuccess(c, configs)
+}
+
+// SaveGroupMonitorConfigHandler 保存分组监控配置（管理员）
+func SaveGroupMonitorConfigHandler(c *gin.Context) {
+	var cfg GroupMonitorConfig
+	if err := c.ShouldBindJSON(&cfg); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{
+			"success": false,
+			"message": "invalid request body",
+		})
+		return
+	}
+
+	if cfg.GroupName == "" {
+		c.JSON(http.StatusBadRequest, gin.H{
+			"success": false,
+			"message": "group_name is required",
+		})
+		return
+	}
+
+	if err := SaveGroupMonitorConfig(&cfg); err != nil {
+		common.ApiError(c, err)
+		return
+	}
+	common.ApiSuccess(c, nil)
+}
+
+// DeleteGroupMonitorConfigHandler 删除分组监控配置（管理员）
+func DeleteGroupMonitorConfigHandler(c *gin.Context) {
+	groupName := c.Param("group")
+	if groupName == "" {
+		c.JSON(http.StatusBadRequest, gin.H{
+			"success": false,
+			"message": "group name is required",
+		})
+		return
+	}
+
+	if err := DeleteGroupMonitorConfig(groupName); err != nil {
+		common.ApiError(c, err)
+		return
+	}
+	common.ApiSuccess(c, nil)
+}
+
+// GetGroupMonitorStatusHandler 用户可见的简化状态
+func GetGroupMonitorStatusHandler(c *gin.Context) {
+	// 获取最近 1 小时的聚合统计
+	startTs := common.GetTimestamp() - 3600
+	stats, err := GetGroupMonitorStats(startTs, 0)
+	if err != nil {
+		common.ApiError(c, err)
+		return
+	}
+
+	// 获取每个分组的最新记录
+	latest, err := GetGroupMonitorLatest()
+	if err != nil {
+		common.ApiError(c, err)
+		return
+	}
+
+	type GroupStatus struct {
+		GroupName      string  `json:"group_name"`
+		LatestLatency  int64   `json:"latest_latency"`
+		LatestSuccess  bool    `json:"latest_success"`
+		LatestTime     int64   `json:"latest_time"`
+		AvgLatency     float64 `json:"avg_latency"`
+		Availability   float64 `json:"availability"` // 百分比
+		TotalChecks    int64   `json:"total_checks"`
+	}
+
+	// 构建 stats map
+	statsMap := make(map[string]*GroupMonitorStat)
+	for i := range stats {
+		statsMap[stats[i].GroupName] = &stats[i]
+	}
+
+	var result []GroupStatus
+	for _, log := range latest {
+		status := GroupStatus{
+			GroupName:     log.GroupName,
+			LatestLatency: log.LatencyMs,
+			LatestSuccess: log.Success,
+			LatestTime:    log.CreatedAt,
+		}
+		if stat, ok := statsMap[log.GroupName]; ok {
+			status.AvgLatency = stat.AvgLatency
+			status.TotalChecks = stat.TotalCount
+			if stat.TotalCount > 0 {
+				status.Availability = float64(stat.SuccessCount) / float64(stat.TotalCount) * 100
+			}
+		}
+		result = append(result, status)
+	}
+	common.ApiSuccess(c, result)
+}

modules/group_monitor/migrate.go

@@ -0,0 +1,10 @@
+package group_monitor
+
+import (
+	"github.com/QuantumNous/new-api/model"
+)
+
+// Migrate 执行分组监控模块的数据库迁移
+func Migrate() error {
+	return model.DB.AutoMigrate(&GroupMonitorLog{}, &GroupMonitorConfig{})
+}