> > > SpecAdr003TechnologyStackPythonGit

ADR-003: Technology Stack: Python, Git, and Textual DSL

Status

Accepted

Context

The early design considered extending Java-based diagramming tools, including PlantUML, to support additional BPMN behavior.

That path would couple SOTER to a heavy diagramming stack and make the modeling language secondary to rendering.

Decision

SOTER uses a dedicated Python implementation for parsing, analysis, compilation, and projection.

The canonical source is a textual DSL stored in Git. Rendering is performed by separate projectors and browser-compatible visualization components.

Rationale

Python gives SOTER a practical implementation environment for parser work, static analysis, test automation, Django-based UI work, and local development.

Git makes the model reviewable and auditable. Textual source keeps SOTER closer to infrastructure-as-code and documentation-as-code practices than to manual diagram editing.

Consequences

Positive

CI/CD automation is straightforward.
Textual model changes can be reviewed and diffed.
Rendering engines can be swapped behind projector boundaries.

Negative

SOTER must own more graph-analysis and projection code.
The Python stack must be kept disciplined with typing and tests.

Neutral or Operational

PlantUML, D2, BPMN, Mermaid, and browser renderers remain projection targets, not the core modeling layer.

Alternatives Considered

Extend Java / PlantUML Directly

Rejected because it would make rendering infrastructure drive the modeling architecture.

Visual-First Modeling Tool

Rejected because it conflicts with Git-first model governance.

Migration or Adoption

Legacy prototype concepts should be reimplemented as Python modules and tested through the normal SOTER pipeline.