ArchMetac4:softwaresystem

A platform for managing architecture documentation as code. Architects describe their organisation in a textual DSL stored in Git; the Kernel reads that repository into an in-memory graph and serves it over REST, the Viewer authors it, and the Publisher turns it into a static site. There is no database — Git is the source of truth, and the same commit always produces the same model.

Git Repositoryc4:externalsystem

The model repository: .arch, .data, .view, .cascade, .metamodel, .bpmn and .md files. This IS the database. The Kernel reads it on load and commits to it on write-back; it never pushes, never pulls and never branches, and is silently inert when the directory is not a Git working tree.

ArchMetaarchimate:applicationcomponent

Architectarchimate:businessactor

Describes the organisation in the DSL, usually by generating .arch files with an LLM and then adjusting them in the Viewer. Reloads the Kernel and reads the diagnostics to find out whether what they meant is what they wrote. The Viewer's edit mode is for curation, not bulk authoring.

Stakeholderc4:person

Reads the published site. Never installs anything, never sees the DSL, and cannot change the model — the site is static HTML with no back end, so there is nothing to give them access to.

ArchMeta Deploymentarchimate:node

How ArchMeta is run. Everything below is a technology-layer reading of the containers above: images, the runtime that executes them, and the volumes that carry state. The product itself is stateless — all state is the mounted Git working tree.