feros
Evaluation

Run Lifecycle

Transition model for creating, executing, and finalizing evaluation runs.

Run Lifecycle

Typical Flow

  1. Create or update evaluation config.
  2. Trigger run from config version.
  3. Run transitions from queued to running.
  4. Run ends as completed, failed, or cancelled.
  5. Inspect run detail, events, and judgments.

Transition Rules

  • queued -> running or cancelled
  • running -> completed, failed, or cancelled
  • Terminal states do not transition further

On this page