@vertekum/cli

The headless Vertekum runner: the whole pipeline — parse, compose, validate, export — driven from the command line. No browser and no UI load unless asked for.

Get started

npm install --save-dev @vertekum/core @vertekum/cli
npx vertekum init

init scaffolds a config, a seed token collection, and the agent skills. From there the everyday loop is:

vertekum check     # validate: references, compositions, vocabulary, targets
# …edit tokens…
vertekum build     # run the configured export targets

vtk is a shorthand alias for the same binary — vtk check, vtk resolver add -s sem.

vertekum describe --json reports what CAN be configured — registered exporters and their option schemas, validators, compositions, runnable commands; config files only say what IS. That split is what lets an agent (or a script) learn a project's capabilities instead of guessing them.

Verbs

| Verb | What it does | | ---------------------------------------- | ----------------------------------------------------------------------- | | init | scaffold a project: config, seed tokens, agent skills | | check | run every registered validator and report diagnostics | | build | run the configured export targets and write their files | | describe | print the live inventory: extensions, exporters, validators, schemas, commands, compositions | | token / group / set / resolver … | the curation verbs: structural edits that rewrite references and refuse to break the collection | | migrate values | convert stored string values to 2025.10 object notation | | schema eject | copy a schema into the working directory so it can be changed | | dev | launch the Vertekum UI — optional peer; see the contract |

Extensions may contribute further verbs; they appear in --help and describe only when the loaded project actually installs them.

Contract

The short version — the contract page is the full one:

• Exit codes: 1 means the tokens are wrong (diagnostics, refused build); 2 means the invocation was (no config found, usage error).
• Output streams: stdout is data, stderr is logs — --json output stays pipeable.
• --dry-run / --json / --cwd are owned by the runner, uniformly, including for contributed commands.
• Project discovery: the config's directory is the working directory; the CLI walks up from --cwd (default: the current directory) to find vertekum.config.ts.

License

Apache-2.0