Author, validate, and render

Install the specification package:

npm install @architecture-tokens/spec

Create payments.yaml with kind: architecture-model, an id, exact library versions, and tokens arrays. Validate it, then generate a normalized renderer input for draw.io. The smallest useful command sequence is:

npm install github:architecture-tokens/drawio-prototype#f9e00c7c12750fe3c6fd576c48ca01b18ffcde62
npx archtokens validate payments.yaml --format human
npx archtokens generate payments.yaml --out payments.drawio --ai-model gpt-5.6 --format human

The CLI loads its reference libraries automatically. Add repeatable --library and --policy flags only when you have created additional local libraries or policy sets. Pin released package versions in automation so a build cannot silently change its vocabulary or layout implementation. The validator checks schema, token applicability, required values, conflicts, and policy rules before generation. --ai-model takes precedence over ARCHTOKENS_OPENAI_MODEL, then the default gpt-5.6; OPENAI_API_KEY is only needed by generate. No live API call is part of this documentation check.