@tfks/lume-plugins

Lume plugins published by TFKS. Each plugin lives under src/<name>/ and is exported from deno.json.

Setup

just setup     # check prereqs, configure git hooks, cache deps
just           # list available recipes
just test      # run the test suite
just validate  # run the full CI suite locally

Layout

• src/<name>/mod.ts — plugin entry point (one directory per plugin)
• src/<name>/mod.test.ts — tests
• Justfile — task runner
• wiki/ — repo-local memory (agent protocol, prompts)
• AGENTS.md — operating manual for AI agents (and humans)
• .forgejo/workflows/ — CI/CD

Releasing

just release # patch bump (default); pass `minor` or `major` to override

Plugins

llms — llms.txt generator

Emits a llms.txt file for your Lume site. This plugin renders the page title, description, and tags on each line so an LLM can decide whether a link is worth drilling into before fetching it.

// _config.ts
import lume from "lume/mod.ts"
import llms from "jsr:@tfks/lume-plugins/llms"

const site = lume()
site.use(llms({
  title: "My Site",
  description: "Notes on distributed systems.",
}))
export default site

Each emitted line looks like:

- [Quorum reads explained](/posts/quorum-reads/): when leader leases fail you — #consensus #postgres

Scoping which pages to include

The plugin follows the same convention as the built-in sitemap and feed plugins: pass a Lume search query via the query option. The query runs against page frontmatter, so a page opts in by setting the matching field.

Opt-in pattern — only pages with llms: true in their frontmatter:

site.use(llms({
  query: "llms=true",
}))

---
title: Quorum reads explained
llms: true
tags: [consensus, postgres]
---

Opt-out pattern — include everything except pages flagged llms: false:

site.use(llms({
  query: "llms!=false",
}))

Combined filters — queries support =, !=, ^=, $=, *=, <, >, <=, >=, space-separated AND, and | OR. So you can write query: "type=post llms!=false" to grab posts that haven't opted out.

For predicate logic the query syntax can't express, fall through to filter:

site.use(llms({
  query: "type=post",
  filter: (data) => !data.draft && (data.tags ?? []).length > 0,
}))

Options

Option	Type	Default	Notes
filename	string	"llms.txt"	Output filename, written to site.options.dest.
title	string	"Site"	H1 heading.
description	string	""	Blockquote summary under the H1.
body	string	""	Free-form markdown inserted before the page list.
query	string	all pages	Lume search query, e.g. "llms=true". Same syntax as sitemap/feed.
filter	(data) => boolean	include all	Escape hatch applied after query; receives page.data.
sort	(a, b) => number	by URL	Custom ordering, receives whole pages.