By name: gzippy-ci-analysis description: Analyze gzippy CI benchmark results from GitHub Actions. Use when checking CI status, reading benchmark results, triaging performance gaps, or comparing runs.
gzippy CI Analysis
Fastest Path to Current Status
gzippy-dev score # Cloud fleet results (authoritative)
gzippy-dev losses # Losses grouped by root cause
gzippy-dev ci triage # CI results (noisier, but faster)
Command Reference
ci triage [--run ID] [--branch NAME]
Groups every CI benchmark scenario into WIN / PARITY / LOSS with root causes.
ci push
All-in-one: pushes current branch, waits for CI (~20-30 min), auto-triages.
ci vs-main [--branch NAME]
Compares current branch against main. Shows improvements and regressions.
ci history [--limit N]
Win rate trend across recent CI runs.
ci compare <run_id_a> <run_id_b>
Direct comparison of gzippy speeds between two CI runs.
ci status
Recent CI run statuses.
ci results [--run ID]
Full benchmark table — every tool, every scenario.
ci gaps [--run ID]
Only scenarios where gzippy is slower than any competitor.
Interpreting Results
- WIN: gzippy is fastest
- PARITY: within measurement noise (don't optimize these)
- LOSS: a competitor is faster
CI Runner Characteristics
- x86_64: ubuntu-latest (shared, 2-4 cores, variable)
- arm64: ubuntu-24.04-arm (shared Graviton, higher variance)
- Small files (~22MB) show CV 10-30% — too noisy for verdicts
- Silesia (211MB) gives clean signal (CV 1-3%)
Red Flags
- CV > 10% on silesia: runner contended, unreliable
- Speed change >5% without touching that path: noise