Analytics¶

I keep track of everything. Not because I'm nosy, but because you can't improve what you don't measure.

What I Track¶

The Analytics tab shows you:

Change velocity -- How many beads are getting closed per day, per week, per sprint. Is work flowing or stalling?
Cost tracking -- How many tokens I'm burning, what that costs per provider, and whether the trend is up or down
Agent performance -- Iterations per bead, success rates, how long things take. Which agents are efficient? Which ones struggle?
Workflow metrics -- Throughput, latency, completion rates. Are my workflows well-designed or do they have bottlenecks?

API¶

# How fast is work moving?
curl "http://localhost:8080/api/v1/analytics/change-velocity?project_id=my-project&time_window=7d"

# How are workflows performing?
curl http://localhost:8080/api/v1/workflows/analytics

# What are providers costing me?
curl http://localhost:8080/api/v1/providers/stats

Grafana¶

For the deep analysis, I ship with pre-configured Grafana dashboards at http://localhost:3000 (default: admin/admin):

Loom Overview -- System health, bead throughput, agent utilization
Provider Metrics -- Token usage, latency, error rates per provider
Agent Performance -- Iteration counts, execution times, success rates

These dashboards talk to Prometheus (where I store metrics) and Loki (where I store logs). I also have Jaeger for distributed tracing, if you want to follow a single request through my entire system.

Custom Metrics¶

Everything I report is backed by OpenTelemetry. Here's what I export:

Metric	Type	What It Tells You
`loom.beads.total`	UpDownCounter	Total beads in the system
`loom.beads.completed`	Counter	Beads I've finished
`loom.beads.active`	UpDownCounter	Beads currently being worked
`loom.agent.iterations`	Counter	How many action loop turns I've run
`loom.workflows.started`	Counter	Workflows kicked off
`loom.workflows.completed`	Counter	Workflows finished
`loom.dispatch.latency`	Histogram	How long dispatch takes (ms)
`loom.agent.execution_time`	Histogram	How long agents spend on beads (ms)

If you're building your own dashboards, these are the metrics to query.