Skip to content

Self-contained outputs

kube-saver produces portable report and review files without a maintainer-operated service. Reading an HTML report works offline; scanning, serving fresh data, or executing patch commands requires Kubernetes API access. Review internal cluster names before sharing or committing generated files.

Design guarantee

Generated HTML uses inline assets and does not call a hosted service. Local Markdown and JSON remain readable with standard tools. Running the application still requires its Python dependencies, credentials, and cluster access; release artifacts are not a guarantee of compatibility with future environments.

Output inventory

HTML report (kube-saver report)

  • Fully self-contained: inline CSS, no JavaScript, no CDN, no hosted assets
  • Works offline in any browser
  • Review for internal cluster names before emailing or committing
  • Contains the full waste breakdown, cost table, and recommendation list as of the generation time
  • The report is a snapshot, it does not fetch live data

PR plan files (kube-saver pr-plan)

All files are written to a local directory you specify:

File Format Purpose
summary.md Markdown One-page summary for human reviewers
review.txt Plain text Every recommended change with current vs. suggested value and reasoning
apply-patches.sh Bash script Ready-to-run commands to apply the recommended resource changes
README.md Markdown Context and instructions for the reviewer

The apply-patches.sh script is not auto-executed. Set KUBE_SAVER_APPLY_CONTEXT to an explicitly reviewed target before running it; the script stops on the first failed command. It is a file you review, test, and run yourself. kube-saver will never mutate your cluster without your explicit action.

Notifications (kube-saver notify)

A daily summary and, when eligible, a spike alert:

File Content
daily-summary-{date}.md Namespace-level waste, cost, and efficiency table
spike-alert-{date}.md Written above notify --threshold (default $100/month), with complete measured metrics and no partial scan

These files are designed to be consumed by:

  • A cron job that emails them to a team
  • A CI pipeline that archives them as artifacts
  • A Git-based ops workflow that commits them for audit

JSON

Standard JSON, no hosted dependency:

  • --json PATH on the report command writes a JSON summary

HTTP API (kube-saver serve)

  • Loopback-only by default (127.0.0.1)
  • Read-only endpoints
  • No authentication (because it is not designed to be public)
  • See Safety & trust if you need to expose it behind a proxy

Why no hosted service

There is no required kube-saver account, hosted dashboard, or CDN for reports. Install the package, grant Kubernetes read permissions, and use metrics-server for measured usage. Air-gapped installations need the package and dependencies available locally; credential acquisition depends on the kubeconfig.

The report is a snapshot with modeled CPU/memory costs. It does not replace continuous monitoring, invoice reconciliation, or autoscaling. See the cost model and FAQ for what the numbers mean.

See also