Dashboard

What is the Dashboard?

Dashboard home brings your products, operational health, and quality checks into one workspace overview.

Your products

Your three most recently created products appear first, with their descriptions and capability and surface counts. Create product opens the creation panel; View all opens the products list. Each product offers a next action: Add capabilities, Add a surface, or Open product. Configured surfaces describe setup, not deployment status or production health.

Product updates shows the latest saved product details for those same products. These timestamps do not imply a deployment or a change to an underlying agent or flow.

Production health

Choose Last 7 days, Last 30 days, or Last 90 days. The selection applies to production health, execution costs, and Activity. Refresh production metrics requests a fresh snapshot.

  • Runs includes agent and flow root runs, including unfinished runs. Historical runs with an unknown type remain visible in the total.
  • Success rate measures successful runs as a share of finished runs.
  • p95 latency measures the slowest five percent of finished runs.
  • Model cost per success divides the period’s model costs by its successful runs. It is unavailable when model costs are incomplete.

Select a metric to explore logs in the same time window. Success rate opens failed-run logs. The Attention strip highlights operational regressions; expand it to see additional notices.

Execution costs

The cost breakdown shows estimated Model, Tool, and Total execution cost. Nested work is included once, and estimates exclude billing margin.

When charges are unreported or unpriced, the dashboard shows a known subtotal and the number of runs with incomplete cost data. A Partial cost data note beside the cost breakdown explains that the totals include known costs only. Missing cost data does not necessarily mean a run is still running. An unavailable model cost per success does not mean the run was free.

Activity and product quality

Activity charts daily root-run counts from the same ledger and window as production health. Switch between Total, Agents, and Flows. View daily counts provides an accessible table; each date opens the matching logs.

Product quality shows the latest scored eval suite in your workspace, its pass rate, scoring date, and comparison with the previous scored run of the same suite. It is evidence from that suite, not a combined quality score for every product. The scoring date can predate the selected production window. View evals opens your suites and runs.

A workspace without traffic shows Nothing has run yet. Products remain accessible when production metrics fail to load. Previously loaded metrics remain visible with a warning if a refresh fails.

Next steps