soev-evals
soev_evals
Evaluate agents: run the cases of a project against an agent, score the answers, and archive each run.
soev_evals.core holds the models of a project and its cases, and the scoring. soev_evals.criteria holds the
built-in criteria, and soev_evals.judges holds the model judges. soev_evals.run runs, archives and rescores
evaluation runs. soev_evals.arms sends the cases to an agent, and soev_evals.track records runs as
OpenTelemetry spans. soev_evals.cli loads a project for the command soev-evals.