Awee

Metrics

Aggregate reporting, dashboards, and observability for Awee Engine runs - coming soon.

Coming soon

The Metrics dashboard and aggregate reporting API are under active development. Per-step cost and performance data is already captured on every run - this page describes the reporting layer being built on top of it.


What's already there

The engine already captures everything you need. Every workflow step produces an ExposedComponentMetric record with:

  • Full token breakdown - input, output, and thinking tokens
  • Cost in provider currency and your local currency, updated with daily exchange rates
  • Step timing, retry count, and loop iteration
  • Hierarchical step IDs so you can trace cost to any depth in a nested workflow
  • A UUID per step execution for exact correlation

That data is available today via Accounting. The Metrics layer will surface it in a form that doesn't require you to write aggregation code yourself.


What's coming

Run dashboard

Per-run cost and latency overview. See total spend, slowest steps, and cache efficiency at a glance - for any workflow execution.

Aggregate reporting

Roll up cost and performance across many runs. Filter by workflow, user, provider, or model. Export to CSV or query via API.

Token usage trends

Track input, output, and thinking token consumption over time. Spot cost anomalies before they compound.

Model comparison

Run the same workflow against multiple providers and models. Compare cost, latency, and output quality side by side.

Budget alerts

Set per-user, per-workflow, or per-period spend limits. Get notified - or automatically halt - when a threshold is crossed.

Cache efficiency report

See how many expensive calls are being avoided by the cache and how much that's saving you, per workflow and in aggregate.


Stay informed

Metrics reporting is one of the next features on the roadmap. The underlying data infrastructure is already in place - the work is in the reporting and visualisation layer.

If you need aggregate metrics today, the ExposedComponentMetric stream gives you everything required to build your own. See Accounting for the full field reference and example aggregation patterns.

Got a specific reporting requirement? Reach out - knowing what you need to measure helps us prioritise what to build first.

How is this guide?

On this page