Auto-diagrams from Kubernetes manifests

Kubernetes is the software many teams use to run containers across clusters, but its YAML files read more like machine instructions than system maps. KubeDiagrams turns those manifests, Helm charts, and live cluster state into architecture diagrams automatically. (github.com) The project’s GitHub repository says it can ingest Kubernetes manifest files, Kustomization files, Helm charts, helmfile descriptors, and “actual cluster state,” then export diagrams in formats including DOT, draw.io, PDF, PNG, SVG, and TIFF. The repository showed about 2.3 million stars? No — about 2.3 thousand stars and 135 forks on April 22, 2026. (github.com) The latest PyPI release, version 0.7.0, was published on February 18, 2026. The package page says KubeDiagrams is also distributed as a Docker image, a `kubectl` plugin, a Nix flake, and a GitHub Action. (pypi.org) In Kubernetes, a manifest is a text file that declares objects like Deployments, Services, and Ingress rules; Helm charts bundle many of those files into reusable application packages. Raw manifests tell the cluster what to run, but they do not show the full shape of an application at a glance. (arxiv.org) That gap has become a documentation problem as teams push frequent changes into distributed systems. A May 28, 2025 research paper on KubeDiagrams said manual diagrams made in tools like Draw.io or Lucidchart become “quickly outdated and error-prone” as Kubernetes environments evolve. (arxiv.org) KubeDiagrams’ pitch is that the diagram should come from the same source as the deployment itself. The GitHub README says the tool supports “most of all Kubernetes built-in resources,” custom resources, resource clustering, and relationship mapping, which lets one rendering step reflect what is defined in code or running in the cluster. (github.com) The project also ships examples for systems that are hard to explain from YAML alone, including WordPress, ZooKeeper, Cassandra, Argo CD, cert-manager, Kube Prometheus Stack, and the Train Ticket microservice benchmark. Those examples show the tool aimed at multi-service environments where onboarding and review often depend on someone translating manifests into a picture by hand. (pypi.org) The paper’s authors, Philippe Merle of Inria and Fabio Petrillo of École de technologie supérieure, wrote that developers, operators, and architects struggle to build “accurate mental models” from raw manifests and cluster descriptions. Their case studies argued that auto-generated diagrams improved system comprehension and architectural reasoning in cloud-native applications. (arxiv.org) KubeDiagrams does not replace Kubernetes configuration; it renders that configuration into something a human can scan in seconds. For teams buried in YAML, that means the diagram can change when the cluster changes, instead of after someone remembers to update a slide. (github.com)