@lexiconlang/cli

Command-line tool for working with Lexiconlang outside of a running app.

pnpm add -g @lexiconlang/cli
# or use directly via npx:
npx @lexiconlang/cli ...

Provides the lexiconlang binary with two commands.

build-markov

Train a Markov model from a corpus and save the precomputed JSON model.

lexiconlang build-markov ./corpora/welsh-towns.json --out ./models/welsh.json \
  --order 3 \
  --min-length 4 \
  --max-length 12 \
  --reject-substrings-of-length 5

Corpus formats

• JSON string[]: ["aberffraw", "betws", ...]
• JSON weighted: [{ "word": "betws", "weight": 3 }, ...]
• Newline-delimited text (# lines ignored)

Why pre-train?

Training is deterministic but not free — for large corpora it adds startup time. Pre-compute once, commit the JSON, import as a module.

import welshModel from "./models/welsh.json";
import { markov } from "@lexiconlang/markov";

const townName = markov(welshModel);

scaffold-pack

Generate a new genre-pack package skeleton in your workspace.

lexiconlang scaffold-pack noir --dir ./packages

Creates ./packages/noir/ with:

• package.json (workspace-ready)
• tsconfig.json (extending the workspace base)
• src/index.ts (with example generators)
• src/data.ts (placeholder weighted lists)
• README.md

Edit, fill in your morphemes and templates, and you have a new pack that plays nicely with the rest of the ecosystem.