canvas-network-viz

Canvas Network Visualization

Overview

This skill renders inline network visualizations in the OpenClaw chat interface using the Canvas/A2UI framework. It does NOT establish direct device connections or create new MCP servers. All data is sourced from existing MCP tools.

Visualization Types

Type	Component	Data Sources	Trigger
Topology Map	topology-map	pyATS CDP/LLDP, health metrics	"show network topology"
Dashboard Panel	dashboard-panel	pyATS, Grafana, Prometheus	"show health dashboard for {device}"
Alert Card	alert-card	Grafana alerts, Prometheus alertmanager	"show current alerts"
Change Timeline	change-timeline	ServiceNow	"show change request {CR} status"
Diff View	diff-view	pyATS config/route/ACL	"show config diff for {device}"
Path Trace	path-trace	pyATS, SuzieQ	"trace path from {src} to {dst}"
Health Scorecard	health-scorecard	pyATS, Grafana, Prometheus	"show health scorecard for {site}"

How It Works

1. Operator requests a visualization via natural language
2. Skill identifies visualization type from request pattern
3. Data is fetched from relevant MCP servers via existing tools
4. Data is transformed into the visualization-specific content model
5. A2UI JSON envelope is built with component type, props, and content
6. Canvas framework renders the visualization inline in chat
7. GAIT audit log entry is emitted for the visualization event

Output Format

All visualizations produce A2UI JSON:

{
  "a2ui": {
    "version": "1.0",
    "type": "canvas",
    "component": "<visualization-type>",
    "props": {
      "id": "<uuid>",
      "title": "<title>",
      "timestamp": "<ISO 8601>",
      "dataSources": [...],
      "scope": {...},
      "warnings": [...],
      "content": {...}
    }
  }
}

Graceful Degradation

• If a data source is unavailable, partial data is rendered with warnings
• If no data is available at all, an error-card is returned with guidance
• Health status defaults to "unknown" (gray) when metrics are missing
• Large topologies (200+ nodes) are automatically clustered by site

Integration with Other Skills

Skill	Integration
gait-session-tracking	All visualization events are GAIT-logged
pyats-topology	CDP/LLDP discovery data for topology maps
pyats-health-check	Health metrics for node coloring and dashboards
grafana-observability	Prometheus metrics, alerts, dashboard data
prometheus-monitoring	Direct PromQL queries for metrics
servicenow-change-workflow	Change request lifecycle data

Important Rules

• Read-only: This skill never modifies network devices or MCP server state
• No new credentials: Uses existing MCP server connections only
• Complements Three.js HUD: Does not replace ui/netclaw-visual/
• GAIT mandatory: Every visualization generation is audit-logged
• WCAG 2.1 AA: Color scales meet accessibility contrast requirements