| name | gateway-pulse |
| description | Continuous gateway and LiteLLM monitoring for OpenClaw. Provides real-time health dashboard, alert thresholds (warning/critical), auto-remediation for common failures, metrics export to Prometheus, and WebSocket connection monitoring. Use when monitoring gateway health, LiteLLM status, or managing infrastructure alerts. |
Gateway Pulse
When to use this skill
Use this skill when you need to:
- Monitor OpenClaw Gateway health on port 18789
- Monitor LiteLLM health on port 4000
- Set up real-time health dashboards
- Configure alert thresholds (warning/critical)
- Enable auto-remediation for common failures
- Export metrics to Prometheus
- Monitor WebSocket connection status
When NOT to use this skill
Do NOT use this skill when:
- You need to manage agent lifecycle (use
agent-lifecycle-manager skill)
- You need to manage triad deliberations (use
triad-orchestrator skill)
- You need to perform agent-specific operations
- You need to backup gateway configuration
Inputs required
Before executing, determine:
- Operation type: status, monitor, alert, or metrics
- Target service: gateway, litellm, or all
- Alert thresholds: warning and critical levels
- Monitoring interval: check frequency in seconds
Workflow
1. Check current status
./scripts/gateway-pulse.sh status
./scripts/gateway-pulse.sh status --service litellm
./scripts/gateway-pulse.sh status --service all
2. Start continuous monitoring
./scripts/gateway-pulse.sh monitor
./scripts/gateway-pulse.sh monitor --interval 15
./scripts/gateway-pulse.sh monitor --auto-remediate
3. Configure alerts
./scripts/gateway-pulse.sh alerts --config
./scripts/gateway-pulse.sh alerts --set --warning 5000 --critical 10000
./scripts/gateway-pulse.sh alerts --history
Alert thresholds:
- Latency warning: Response time > 5000ms
- Latency critical: Response time > 10000ms
- Error rate warning: Error rate > 5%
- Error rate critical: Error rate > 20%
4. Auto-remediation
./scripts/gateway-pulse.sh remediate --enable
./scripts/gateway-pulse.sh remediate --run
./scripts/gateway-pulse.sh remediate --history
Auto-remediation actions:
- Restart Gateway service on critical failure
- Restart LiteLLM service on critical failure
- Clear WebSocket connection cache
- Reset connection pools
5. Export metrics
./scripts/gateway-pulse.sh metrics --export
./scripts/gateway-pulse.sh metrics --endpoint
./scripts/gateway-pulse.sh metrics --serve --port 9090
6. Health dashboard
./scripts/gateway-pulse.sh dashboard
./scripts/gateway-pulse.sh status --json
./scripts/gateway-pulse.sh watch
Files
Examples
Example 1: Start monitoring with alerts
./scripts/gateway-pulse.sh monitor --interval 30 --warning 5000 --critical 10000
Example 2: Export Prometheus metrics
./scripts/gateway-pulse.sh metrics --export --format prometheus
Example 3: View health dashboard
./scripts/gateway-pulse.sh dashboard
Troubleshooting
Gateway not responding
- Check Gateway process:
docker ps | grep gateway
- Check Gateway logs:
docker logs heretek-openclaw-core-gateway-1
- Restart Gateway:
./scripts/lifecycle-manager.sh restart --agents gateway
LiteLLM not responding
- Check LiteLLM process:
docker ps | grep litellm
- Check LiteLLM logs:
docker logs heretek-openclaw-core-litellm-1
- Restart LiteLLM:
docker compose -p heretek-openclaw-core restart litellm
WebSocket connection issues
- Check WebSocket endpoint:
curl -i http://127.0.0.1:18789/health
- Verify WebSocket URL in configuration
- Clear connection cache:
./scripts/gateway-pulse.sh remediate --clear-cache
Gateway Endpoints
The OpenClaw Gateway exposes these health endpoints on port 18789:
/health - Overall gateway health/health/<agent> - Specific agent health/metrics - Prometheus metrics/ws - WebSocket RPC endpoint
LiteLLM Endpoints
LiteLLM exposes these endpoints on port 4000:
/health - LiteLLM health check/v1/chat/completions - Chat completions/v1/models - Available models/metrics - Prometheus metrics
