fix(api): handle zero-baseline edge cases in trend calculation

Introduce `CalculateTrendFloatWithBaseline` to correctly handle scenarios where previous period metrics (Error Rate, Latency) are zero or missing. This prevents arithmetic errors and distinguishes between "new" data and actual increases ("up") when starting from zero.

Also updates the admin panel dashboard documentation to reflect current project status.

internal/api/dashboard_handler.go

@@ -89,6 +89,31 @@ func CalculateTrendFloat(current, previous float64) TrendInfo {
 	return TrendInfo{Delta: &delta, Direction: direction}
 }
 
+// CalculateTrendFloatWithBaseline calculates trend when baseline existence is explicit.
+func CalculateTrendFloatWithBaseline(current, previous float64, hasBaseline bool) TrendInfo {
+	if !hasBaseline {
+		if current == 0 {
+			return TrendInfo{Direction: "stable"}
+		}
+		return TrendInfo{Direction: "new"}
+	}
+	if previous == 0 {
+		if current == 0 {
+			return TrendInfo{Direction: "stable"}
+		}
+		return TrendInfo{Direction: "up"}
+	}
+	delta := (current - previous) / previous * 100
+	delta = math.Round(delta*10) / 10
+	direction := "stable"
+	if delta > 0.5 {
+		direction = "up"
+	} else if delta < -0.5 {
+		direction = "down"
+	}
+	return TrendInfo{Delta: &delta, Direction: direction}
+}
+
 // RequestStats contains request-related statistics
 type RequestStats struct {
 	Total     int64   `json:"total"`
@@ -302,20 +327,21 @@ func (h *DashboardHandler) GetSummary(c *gin.Context) {
 		if !prevStart.IsZero() && !prevEnd.IsZero() {
 			prevStats, err := h.aggregateFromLogRecords(prevStart, prevEnd)
 			if err == nil {
+				hasBaseline := prevStats.Requests > 0
 				prevErrorRate := 0.0
-				if prevStats.Requests > 0 {
+				if hasBaseline {
 					prevErrorRate = float64(prevStats.Failed) / float64(prevStats.Requests)
 				}
 				prevAvgLatency := 0.0
-				if prevStats.Requests > 0 {
+				if hasBaseline {
 					prevAvgLatency = float64(prevStats.LatencySumMs) / float64(prevStats.Requests)
 				}
 
 				trends = &DashboardTrends{
 					Requests:  CalculateTrend(totalRequests, prevStats.Requests),
 					Tokens:    CalculateTrend(ts.TokensIn+ts.TokensOut, prevStats.TokensIn+prevStats.TokensOut),
-					ErrorRate: CalculateTrendFloat(errorRate, prevErrorRate),
-					Latency:   CalculateTrendFloat(ts.AvgLatency, prevAvgLatency),
+					ErrorRate: CalculateTrendFloatWithBaseline(errorRate, prevErrorRate, hasBaseline),
+					Latency:   CalculateTrendFloatWithBaseline(ts.AvgLatency, prevAvgLatency, hasBaseline),
 				}
 			}
 		}