Performance history.

How to read these results

Help-command startup

—ms

—vs previous — runs (— ms)

Recipe editor first frame

—ms

—vs previous — runs (— ms)

— benchmark runs since March 2026 Latest run ——

01 / STARTUP

Help-command startup

Time to run protostar help init and exit.

Mean execution time / ms

02 / RECIPE EDITOR

The first frame

Time to open the recipe editor and exit after its first frame, including process startup and shutdown.

PROTOSTAR_BENCHMARK_RECIPE_EDITOR=1 .venv/bin/protostar init

Mean execution time / ms

03 / METHODOLOGY

Read the trend, with context.

What each point measures

Each point is the mean of 90 Hyperfine executions after 30 warmups, on GitHub Actions Ubuntu runners with Python 3.14, in the checked-out repository. Neither benchmark measures a completed scaffold or dependency installation.

Runner variability and changes to the runner image or project checkout can affect comparisons. Sustained trends are more useful than an isolated increase.

What the comparisons mean

The summary compares the latest measurement with up to 100 preceding runs of the same benchmark. CI uses a separate check: 30 executions after 5 warmups, failing above 250% of the preceding recorded result. Historical tracking alerts above 200%.

Older wizard measurements remain in the JSON export, but are excluded from the recipe-editor chart and its comparisons.

Recent measurements and commit linksLast 30 per benchmark

Recorded means; dates are shown in your local timezone.
Benchmark	Measured	Mean	Commit

Run the benchmarks locally