Monitoring Dashboard

Building custom monitoring dashboards for LeafMesh systems.

LeafCraft Dashboard

The LeafCraft dashboard is the primary way to monitor your agents — it provides full traceability, session explorer, agent performance, LLM cost analytics, and more out of the box (see Traceability API for details).

This page covers building additional custom dashboards using the LeafMesh analytics APIs for use cases like internal tooling or custom monitoring.

Data Sources for Custom Dashboards

1. Health Check Endpoint

Poll the API server health endpoint:

GET /health

{
  "status": "healthy",
  "agents": {
    "total_registered": 8,
    "healthy": 7,
    "degraded": 1,
    "failed": 0
  }
}

2. LeafMesh Analytics API

# Usage analytics
analytics = Studio's Dashboard (HTTP)

# Agent statistics
agent_stats = leafmesh.get_agent_stats()

# Cache statistics
cache_stats = Studio's Dashboard (HTTP)

Building a Custom Dashboard

FastAPI Dashboard Endpoint

from fastapi import FastAPI

app = FastAPI()

@app.get("/dashboard/metrics")
async def dashboard_metrics():
    return {
        "analytics": Studio's Dashboard (HTTP),
        "agents": leafmesh.get_agent_stats(),
        "cache": Studio's Dashboard (HTTP)
    }
    # For component health status, use the built-in health endpoint:
    # GET http://localhost:18820/health

Key Metrics to Display

Panel	Metrics	Source
System Health	Agent health	Health endpoint
Agent Performance	Response times, error rates	`leafmesh.get_agent_stats()`
LLM Costs	Total cost, cost by model, cache savings	Studio's Dashboard
Event Flow	Events/second	Studio's Dashboard

Integration with Grafana

Export metrics to Prometheus for Grafana dashboards:

from prometheus_client import Gauge, Counter, start_http_server

agent_errors = Counter("swarm_agent_errors_total", "Agent errors", ["agent"])
agent_latency = Gauge("swarm_agent_latency_ms", "Agent response time", ["agent"])
cache_hit_rate = Gauge("swarm_cache_hit_rate", "LLM cache hit rate")

async def export_metrics():
    """Periodically export metrics to Prometheus"""
    agent_stats = leafmesh.get_agent_stats()
    for agent, stats in agent_stats.items():
        agent_latency.labels(agent=agent).set(stats.get("avg_response_ms", 0))

    cache = Studio's Dashboard (HTTP)
    cache_hit_rate.set(cache.get("hit_rate", 0))

Next Steps

Traceability API — REST API for session/agent/span hierarchy
Alerting — Alert configuration
Monitoring — System monitoring
Metrics — Metrics reference

LeafMesh — Building monitoring dashboards