Grounded question format, scoring for recall, precision, evidence and groundedness, and a text-search baseline.

Command-line interface for inspecting, querying, visualizing, enriching, and evaluating a repository graph.