Skip to content

Releases: pallaprolus/kube-foresight

v0.3.1

14 Jun 02:21
@pallaprolus pallaprolus
v0.3.1
2b3d9f1

Choose a tag to compare

View all tags
v0.3.1 Latest
Latest

Patch release — fixes the dashboard on PyPI installs.

Fixed

  • Packaging: the wheel and sdist now include the dashboard's Jinja templates and static assets (kube_foresight/dashboard/{templates,static}). On a non-editable install, kube-foresight dashboard previously crashed at startup (StaticFiles: directory does not exist) because those files were missing from the published distributions (since 0.2.0). The CLI was unaffected.

pip install "kube-foresight[dashboard]" → kube-foresight dashboard --demo now works.

Full changelog: v0.3.0...v0.3.1

Assets 2
Loading

v0.3.0

14 Jun 02:09
@pallaprolus pallaprolus
v0.3.0
9fb82c8

Choose a tag to compare

View all tags
v0.3.0

Alpha release. Analysis is read-only — the CLI never changes your cluster.

Recommendation quality & safety

  • Per-resource sizing — CPU and memory are sized independently, so a CPU-wasteful workload pinned at its memory limit still gets its CPU cut.
  • Default strategy is now p99 — a safer savings/violation trade-off.
  • Sizes on raw usage — removed pre-percentile outlier filtering that could under-provision against real demand spikes.

Validation

  • New backtest harness (benchmarks/) that validates recommendations against a public production trace (Alibaba 2018) with a held-out train/test split, plus a one-command fetch script.

Distribution

  • Docker images are now published to GitHub Container Registry: ghcr.io/pallaprolus/kube-foresight.

Docs

  • Operator-first README rewrite, accurate cost framing (reclaimable capacity, not a billing forecast), and durable positioning.

Full changelog: v0.2.0...v0.3.0

Loading