Yantrix

Code Generation ​

Why generate code from diagrams? ​

State machine logic is one of the most diagram-friendly kinds of software: a box-and-arrow picture of states and transitions IS the specification. When that picture is also the source of truth for the running code, several things follow naturally:

No drift. The diagram in your architecture docs is the same artifact that produces the runtime. They cannot contradict each other because they are the same file.

Polyglot from one source. A single .mermaid diagram generates a TypeScript class, a zero-dependency JS factory, a Python module, or a Java class. The business logic lives once; the target language is a flag.

Review-friendly. FSM changes appear in diffs as Mermaid syntax, which non-engineers can read. The generated output changes are a consequence, not the subject of review.

Living documentation. Embedding state diagrams in architectural decision records (ADRs) or wikis is standard practice. With Yantrix, those diagrams are not illustrations -- they are the implementation.

Codegen as a tool ​

Yantrix codegen can be invoked two ways:

CLI -- suitable for build scripts, CI, and one-off generation:

bash
yantrix codegen ./diagram.mermaid -l typescript -o src/Automata.ts -c MyAutomata

See CLI Reference for all options.

API -- embed generation in Node.js tooling:

typescript
import { generateAutomataFromStateDiagram } from '@yantrix/codegen';
import { parseStateDiagram, createStateDiagram } from '@yantrix/mermaid-parser';

const structure = await parseStateDiagram(diagramText);
const matrix = await createStateDiagram(structure);
const code = await generateAutomataFromStateDiagram(matrix, {
  outLang: 'typescript',
  className: 'MyAutomata',
});

Supported languages ​

See Language Support for the full feature matrix across all dialects.

Implementation internals ​

If you are contributing to codegen or adding a dialect, see: