Skip to main content
Once a Test Run is complete, the Portal gives you three layers of information:
  • Details: what configuration was used + timestamps
  • Results: macro metrics and score distribution
  • Requests: per-request scores, criteria breakdowns, and response comparisons

​
Run details (what was executed)

On the Test Run page, the Details section includes:
  • Date created / completed
  • Status (e.g. COMPLETED, ERROR, etc.)
  • Description
  • Configuration
    • The UI lets you open a read-only config panel to inspect the run’s config (including the model).

​
Results (macro view)

The Results section includes:
  • # Requests: total requests in the run
  • Completed % / Error %: how many requests completed successfully vs errored
  • Response time percentiles
    • The UI shows a headline percentile and exposes additional percentiles (e.g. p50/p90/p95) in a tooltip.

​
Score distribution (1–5)

The score distribution shows the run broken into five buckets:
  • 1 — Poor
  • 2 — Fair
  • 3 — Good
  • 4 — Great
  • 5 — Perfect
You can click buckets to filter the requests table to those scores. Use this to quickly focus on the lowest-performing cases.
The Test Runs table on the Test Set page shows Pass Rate, computed as the percentage of completed requests with a score of 3 (Good) or better.

​
Requests table (debugging view)

The requests table is where you debug individual cases. You’ll see:
  • Status for each request execution
  • Tags associated with the underlying test request
  • Score
    • Hover to see the criteria breakdown (criteria name + 1–5 star value + optional description)
  • Time (per-request response time, when available)
  • Actions
    • View: open the original request for deeper inspection
    • Responses: compare the baseline/original response against the test-run response side-by-side
    • Compare: open a comparison modal for the same request across multiple runs

​
Comparing runs

Maitai supports two comparison workflows in the Portal:
  • Compare Runs (test-set level): select multiple completed runs from the Test Set page and open a table that shows request-by-request scores across runs. You can also open a “Responses” comparison to view multiple run outputs side-by-side.
  • Compare (single request across runs): from a Test Run’s requests table, open a modal that lists how that specific request performed across different runs.