Benchmark analytics.

seshat bench agents \
  --agents codex,claude,ollama \
  --fixtures rust,python,typescript \
  --iterations 5 \
  --model seshat-commit \
  --format text \
  --pt-br \
  --keep-temp \
  --show-samples 3 \
  --report bench.html \
  --json bench.json \
  --codex-bin codex \
  --codex-home ~/.codex \
  --codex-model gpt-5.3-codex \
  --claude-bin claude \
  --claude-config-dir ~/.claude \
  --claude-model claude-sonnet-4-6 \
  --ollama-model juniormartinxo/seshat-commit \
  --profile amjr