GreenKube

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

CLI recommend command now uses the unified recommendation engine (all 9 types) instead of legacy 2-type API
CLI recommend reads from database by default (consistent with API); added --live flag for real-time mode
read_combined_metrics_from_database() called with correct parameter names (start_time/end_time)
Cost normalization in run_range() now divides range total by number of time steps
USER_AGENT header dynamically reflects the actual package version
Removed duplicate DEFAULT_COST class attribute from Config
Helm recommendSystemNamespaces moved inside recommendations scope in values.yaml
PostgreSQL credentials no longer shipped as plain text in Helm defaults
DB connection string sourced from Secret instead of inline env var in deployment
collect_detailed_info() now delegates to collect() to avoid inconsistent results
Expanded test fixture env patching to prevent production defaults leaking into tests

Namespace input validation on all API endpoints (Kubernetes naming rules)
Contributing guide (CONTRIBUTING.md)
This changelog
Architecture diagram in docs/architecture.md
API curl examples in README
API security: Optional bearer-token authentication (GREENKUBE_API_KEY), configurable CORS origins, rate limiting via slowapi
Pagination: GET /api/v1/metrics now supports offset and limit query parameters
Docker healthcheck: Built-in HEALTHCHECK instruction for standalone usage
Helm chart tests: helm test connectivity validation via test-connection.yaml
Graceful shutdown: preStop lifecycle hook on the API container
Integration tests: End-to-end API tests with real SQLite backend and migration tests
Methodology section in README explaining how energy and CO₂e are estimated
Shared parse_duration() utility used by both CLI and API
Config.reload() for clean test isolation

Minimum Python version raised from 3.9 to 3.10 (3.9 reached EOL October 2025)
Helm chart generates a random PostgreSQL password when none is provided
Replaced f-string logging with lazy %-formatting throughout the codebase
Recommendation model uses typed scope field instead of sentinel pod_name="*"

SvelteKit web dashboard with real-time charts, node inventory, and recommendations
Full REST API (FastAPI) with metrics, nodes, namespaces, recommendations, and timeseries endpoints
Multi-resource metrics: CPU, memory, network I/O, disk I/O, ephemeral storage, restarts
9-type recommendation engine: zombie, CPU/memory rightsizing, autoscaling, off-peak, idle namespace, carbon-aware, overprovisioned/underutilized node
HPA-aware filtering for autoscaling recommendations
Recommendation history storage and API endpoint
PostgreSQL, SQLite, and Elasticsearch storage backends
Prometheus metrics exposition (/metrics)
Historical range reports with daily/monthly/yearly aggregation
Node snapshot history for accurate time-range analysis
Embodied emissions via Boavizta API integration
Estimation transparency (flags and reasons for estimated values)
Helm chart with PostgreSQL StatefulSet, RBAC, health probes, and auto-discovery
Pre-commit hooks (Ruff, Gitleaks)
CI/CD pipeline (GitHub Actions): lint, test, multi-arch Docker build, Helm publish

This site is open source. Improve this page.