Skip to main content
Evaluations stores the history of all executions performed for a selected release. Each execution is recorded together with its inputs, outputs, execution metadata, and release information. The Evaluations tab makes it possible to review test results, inspect individual executions, analyze prompt behavior, and review results associated with selected releases.

​
Selecting a Release

The release selector is located at the top of the page. Evaluations always displays data for the currently selected release. When a different release is selected, the execution history is automatically updated.

​
Page Structure

The Evaluations page consists of two primary areas:

Release Information

Shows release metadata, input variables, output variables, and release history access.

Executions Table

Lists executions recorded for the selected release.
The left panel displays information about the selected release. The main area contains the execution history for that release.

​
Release Information

The release panel provides information about the currently selected release, including:
  • Label
  • Name
  • Release ID
  • Deployment Date
  • Input Variables
  • Output Variables
The panel also provides access to Release History.

​
Executions Table

Each row in the table represents a single execution. For every execution, Evaluations displays:
  • Execution ID
  • Timestamp
  • Input Variables
  • Output
  • Estimated Cost
  • Status
This allows users to quickly review execution results and identify executions that require further investigation.

​
Viewing Execution Details

Select any execution to open the Execution Details dialog. Execution Details provides access to:
  • Execution metadata
  • Release information
  • Provider and model settings
  • Input variables
  • Output variables
  • Prompt Details
  • Raw Request
  • Raw Response
Detailed execution analysis is covered in the Execution Details section.

​
Release History

Each release is stored as part of the release history. Release History allows users to:
  • Review previous releases
  • Open a release in Sandbox
  • Delete releases
Additional information is available in the Release History section.

​
Exporting Results

Execution results can be exported for external analysis. Supported export formats include:
  • CSV
  • JSON
Exports contain execution data for the selected release.

​
Analytics

The Analytics action opens a dedicated analytics dialog. Analytics provides aggregated statistics for a selected release and time period, including:
  • Total Runs
  • Failures
  • Estimated Total Cost
  • Provider-level cost breakdown
  • Cost trends over time
Detailed analytics functionality is covered in the Analytics section.
Analytics values are intended for operational visibility and may differ from provider billing records.