PaperClipAI/evals/bench/src

TypeScript Runtime Layout

These directories reserve stable homes for the future benchmark CLI and runtime code.

• cli/: user-facing commands
• core/: manifest loading and run orchestration
• datasets/: imported benchmark normalization
• runners/: Paperclip, direct-agent, noop, and other execution backends
• evaluators/: command and benchmark-specific evaluator adapters
• metrics/: metric extraction and failure taxonomy
• reports/: aggregate report generation