feat(server): API relay gateway backend M0-M4

Gin + GORM + pure-Go SQLite. Users/auth (JWT), API key management with
quotas, proxy gateway with weighted channel failover and health checks,
usage/billing ledger, cross-protocol conversion (Anthropic Messages /
OpenAI Chat Completions / OpenAI Responses), and channel/model admin API.
Channels declare native API formats and auto-convert the rest.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

server/internal/channel/health.go

@@ -0,0 +1,53 @@
+package channel
+
+import (
+	"time"
+
+	"go.uber.org/zap"
+
+	"openteam/server/internal/store"
+)
+
+// StartHealthCheck runs the periodic health-check loop in a goroutine.
+// Only channels that are enabled and have a bound test model are checked.
+func (s *Service) StartHealthCheck() {
+	go func() {
+		ticker := time.NewTicker(s.cfg.HealthCheck.Interval)
+		defer ticker.Stop()
+		s.runHealthCheck()
+		for range ticker.C {
+			s.runHealthCheck()
+		}
+	}()
+}
+
+func (s *Service) runHealthCheck() {
+	var channels []store.Channel
+	if err := s.db.Where("enabled = ?", true).Find(&channels).Error; err != nil {
+		s.log.Warn("health check query failed", zap.Error(err))
+		return
+	}
+	for i := range channels {
+		ch := &channels[i]
+		// Skip channels already in cooldown until the cooldown elapses.
+		if ch.HealthStatus == "cooldown" {
+			continue
+		}
+		ok, _, _, err := s.Test(ch)
+		if err != nil || !ok {
+			s.MarkFailure(ch.ID)
+			continue
+		}
+		s.MarkHealthy(ch.ID, "healthy")
+	}
+}
+
+// RecoverCooldown moves channels back from cooldown to degraded after the
+// cooldown window, giving them another chance to pass the health check.
+// Called periodically; keeps cooldown bounded.
+func (s *Service) RecoverCooldown() {
+	cutoff := time.Now().Add(-s.cfg.HealthCheck.Cooldown)
+	s.db.Model(&store.Channel{}).
+		Where("health_status = ? AND updated_at < ?", "cooldown", cutoff).
+		Updates(map[string]any{"health_status": "degraded", "health_failures": 0})
+}