setup

An autoresearch experiment configuration requires a domain, target file, evaluation command, metric, direction, and optional evaluator. The setup script validates these inputs to create a runnable benchmarking workflow for prompt optimization tasks.

Reproducible benchmarking workflows require a domain, target file, evaluation command, metric, and direction. The setup process validates target files and executes the setup script to report baseline metrics for reliable experiment initialization.

Yes, you can list existing experiments and available built-in evaluators. This discovery feature helps you review current configurations and choose the right evaluator path before initializing a new experiment.

Yes, the experiment setup supports an interactive mode that guides you through the configuration. It collects the domain, experiment name, target file, evaluation command, metric, direction, evaluator, and storage scope to build the setup.

Validating target files ensures the setup script reliably reports baseline metrics. This validation step is required to guarantee reliable experiment initialization and prevent configuration errors during benchmarking.