@zweer/llms-txt

A CLI tool to generate llms.txt and llms-full.txt files from a docs directory (e.g., VitePress).

Follows the llms.txt specification to make your documentation AI-friendly.

Installation

npm install -g @zweer/llms-txt

Usage

CLI

# Generate from a VitePress docs directory
llms-txt \
  --project-name "My Project" \
  --project-description "A cool project" \
  --site-url "https://example.com"

# Custom directories
llms-txt \
  --project-name "My Project" \
  --project-description "A cool project" \
  --site-url "https://example.com" \
  --docs-dir ./documentation \
  --out-dir ./documentation/public

Options

| Option | Default | Description | | :--- | :--- | :--- | | --project-name <NAME> | required | Project name for the header | | --project-description <DESC> | required | Project description | | --site-url <URL> | required | Base URL of the site | | --docs-dir <PATH> | ./docs | Path to the docs directory | | --out-dir <PATH> | ./docs/public | Output directory |

Programmatic API

import { generateLlmsFullTxt, generateLlmsTxt, scanPages } from '@zweer/llms-txt'

const sections = scanPages('./docs')

const llmsTxt = generateLlmsTxt({
  projectName: 'My Project',
  projectDescription: 'A cool project',
  siteUrl: 'https://example.com',
  sections,
})

const llmsFullTxt = generateLlmsFullTxt({
  projectName: 'My Project',
  projectDescription: 'A cool project',
  siteUrl: 'https://example.com',
  sections,
})

How It Works

1. Scans the docs directory recursively for .md files
2. Extracts title and description from YAML frontmatter (falls back to first # heading)
3. Groups pages by directory (each subdirectory becomes a section)
4. Generates llms.txt with links and descriptions
5. Generates llms-full.txt with the full content of all pages

License

MIT