v0.2.1 — CLI help/version, logo & branding

A small polish release on top of v0.2.0.

CLI

• Richer --help — top-level usage now covers both modes (server default, agent), key environment variables, and examples (frugal help).
• Per-mode help on frugal server -h / frugal agent -h and on invalid flags.
• frugal version / --version — the version is now injected at build time, so released binaries and the GHCR image report their tag (e.g. frugal v0.2.1).

Branding & docs

• New raccoon logo and a README header with badges and tagline.

Fixes

• Correct a stray AWSOBS_ reference in the docker-compose.prod.yml example comment.

Install

• Docker: docker pull ghcr.io/tools-plus/frugal:v0.2.1 (multi-arch amd64 + arm64)
• Binary: frugal-linux-amd64 / frugal-linux-arm64 attached below (+ sha256) — run frugal version to confirm.

v0.2.0 — EKS discovery, cost controls, long-range history, drag-zoom

Second release — a big batch across EKS, cost, history, and the dashboard.

EKS

• Auto-discover EKS clusters from AWS credentials (eks:ListClusters + DescribeCluster) and collect node/pod metrics with a minted token per cluster — no kubeconfig, no kubectl, works in a container. Toggle: Admin ▸ Settings ▸ Kubernetes ▸ auto-discover EKS.

Cost controls

• Cost-mode presets — frugal (10m) / balanced (5m) / detailed (1m) set poll interval + period in one click.
• Live monthly CloudWatch cost estimate in Settings, reacting to poll interval, selected services, and the supersede toggle.
• native supersedes CloudWatch — drop the paid namespaces a native poller already covers (ElastiCache/OpenSearch/AmazonMQ) so the same metrics aren't billed twice.
• Daily metrics (S3) auto-throttled to hourly, and the CloudWatch fetch window now tracks the poll interval so wider intervals stay gapless.

History

• Long-range history for every source — CloudWatch from the CloudWatch API, and k8s / agent / native from the persisted SQLite store (downsampled for wide ranges). Time ranges are now 12h / 24h / 3d / 7d; default DB retention raised to 7d.

Dashboard

• Drag-to-zoom on charts and the maximize modal: drag a segment to zoom in, double-click to reset. Zooming re-fetches that slice at finer resolution. Accessible selection styling.
• Header shows a meaningful " series · updated " liveness readout.

Install

• Docker: docker pull ghcr.io/tools-plus/frugal:v0.2.0 (multi-arch amd64 + arm64)
• Binary: frugal-linux-amd64 / frugal-linux-arm64 attached below (+ sha256)

The dashboard has no external auth by default — keep it behind port-forward / VPN / an authenticating ingress.

v0.1.0 — first release

frugal is a single Go binary that gives you cheap AWS + EKS observability — it prefers free, native endpoints over paid CloudWatch, needs no Prometheus and no external database, and serves a live dashboard out of the box.

This first release covers the basic system and core collectors.

Two modes, one binary

• frugal server — web dashboard + data collectors (default mode)
• frugal agent — pushes host metrics + logs from EC2 / EKS nodes to a server

Collectors

• CloudWatch — EC2, RDS, DocumentDB, ElastiCache, AmazonMQ, OpenSearch, S3, ALB, NLB, EKS control plane, Container Insights (auto-discovered via ListMetrics, batched GetMetricData)
• Native free endpoints — Valkey/Redis INFO, OpenSearch _cluster/health + _nodes/stats, RabbitMQ management API (auto-discovered from AWS APIs)
• RDS Performance Insights — DB load / active sessions, free 7-day source
• EKS — live pod/node CPU + memory via metrics-server, plus streamed pod logs
• Host agent — CPU/memory/network/load from /proc + tailed logs, zero AWS API calls

Platform

• Live dashboard with SSE streaming (single embedded HTML file)
• Optional SQLite persistence (hydrates the in-memory stores on restart)
• Multi-user auth with admin/viewer + scoped, per-service roles
• Runtime config edited in Admin ▸ Settings, stored encrypted, hot-applied without restarting the web server

Install

• Docker: docker pull ghcr.io/tools-plus/frugal:v0.1.0 (multi-arch amd64 + arm64)
• Binary: frugal-linux-amd64 / frugal-linux-arm64 attached below (+ sha256)

The dashboard has no external auth by default — keep it behind port-forward / VPN / an authenticating ingress.