* chore: use 30-day benchmark history instead of 60-day
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore: regenerate history plots for 30-day window
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* fix: scale history plot y-axis from in-window data only
The plots drew the full 365-day history but only restricted xlim to the
HISTORY_DAYS_SIZE window, so matplotlib autoscaled ymax from data outside
the visible range. Filter points to the window and set ymax from in-window
values (5% headroom).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
The pytest parameter id was renamed `waku_light_client` -> `wakuV2LightClient`
on 2026-06-06, which is embedded in the result filenames. update_readme.py
matched results against hardcoded old-name strings, so from 2026-06-06 onward it
found zero matches: every metric cell rendered N/A and the 60-day history charts
went empty. CI stayed green because the benchmark JSON is still produced and
committed correctly — only the README-generation matching step silently found
nothing.
Add normalize_test_name() to map the renamed variant back to the canonical
`waku_light_client` key, applied both when keying loaded JSON and when resolving
per-test performance chart PNGs. This restores data for new runs while keeping
historical (old-name) runs on the same series, and is resilient to the rename.
Regenerated README.md and docs/*.png reflect the restored data.
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* feat: increase history to 60 days
* feat: added test idle waku_light_client_False
* feat: added goorutines and thread count
* fix: bytes always in mb
* chore: update images and README
