Built whatbroke, a CLI and GitHub Action comparing agent runs by tool calls, arguments, outputs, latency and cost. Importers support JSONL, OpenTelemetry, Langfuse and LangSmith, with configurable CI failure gates. https://github.com/arthi-arumugam-git/whatbroke
0 likes · 0 views · Nov 12, 2026
Agent-run diffing CLI and GitHub Action for evaluation correctness, latency and cost.