poops

v3.0.0

Published

12 days ago

Straightforward, no-bullshit bundler for the web.

0High
0Medium
0Low

stamat

bundler javascript sass nunjucks liquid esbuild dart-sass tailwindcss toolchain-script static-site-generator

💩 Poops

Straightforward, no-bullshit bundler for the web.

When your day is long
And the night, the night is yours alone
When you're sure you've had enough
Of these bundlers, well hang on
Don't let yourself go
'Cause everybody poops
Everybody poops sometimes

R.E.M. - Everybody Poops :poop:

Intuitive with a minimal learning curve and minimal docs, utilizing the most efficient transpilers and compilers available (like dart-sass and esbuild) Poops aims to be the simplest bundler option there is. If it's not, please do contribute so we can make it so! 🙏 All ideas and contributions are welcome.

It uses a simple config file where you define your input and output paths and it poops out your bundled files. Simple as that.

Features

Bundles SCSS/SASS to CSS
Uses dart-sass for SCSS/SASS bundling
Design token support — import JSON tokens (W3C DTCG & Style Dictionary) as SCSS variables or maps
PostCSS pipeline — use any PostCSS plugin including Tailwind CSS
Bundles JS/TS/JSX/TSX to IIFE/ESM/CJS
Uses esbuild for bundling and transpiling JS/TS/JSX/TSX to IIFE/ESM/CJS
React pre-rendering (Reactor) — renders React components to HTML at build time for static sites with optional hydration
Optional JS and CSS minification using esbuild
Can produce minified code simultaneously with non-minified code! (cause I always forget to minify my code for production)
Supports source maps only for non minified - non production code (optional)
Supports multiple input and output paths
Resolves node modules
Can add a templatable banner to output files (optional)
Static site generation with swappable template engines: Nunjucks (default) or Liquid — with blogging option (optional)
Collections with pagination, and taxonomies — tags/categories as paginated, crawlable landing pages (with localizable labels)
Generates a JSON search index, sitemap.xml, llms.txt, robots.txt and a navigation tree from your pages
RSS and Atom feeds from any collection, no feed template to hand-author
Responsive image processing — resize, WebP/AVIF, crops, EXIF — via the optional poops-images
Shell hooks before and after every pipeline stage, so a generator or a post-processor runs on every rebuild and not just on poops -b
Has a configurable local server (optional)
Rebuilds on file changes (optional)
Live reloads on file changes (optional)

Quick Start

For a superfast start, scaffold a project instead of wiring one up:
npm create poops@latest my-app
create-poops prompts for a template and clones it: base (the clean 💩🌪️Shitstorm starter), sulphuris (+ the sulphuris CSS framework) or hat (+ htmx, Alpine.js, Tailwind). Name it as the second argument to skip the prompt: npm create poops my-app hat.

Poops requires Node.js 22 or newer.

You can install Poops globally:

npm i -g poops

or locally:

npm i -D poops

If you have installed Poops globally, create a poops.json or 💩.json configuration file in the project root (see Configuration on how to configure) and run:

poops or 💩

or pass a custom config. This is useful when you have multiple environments:

poops yourAwesomeConfig.json or 💩 yourAwesomeConfig.json

CLI Options:

| Flag | Short | Description | | ---------------------------- | ----- | ---------------------------------------------------- | | --build | -b | Build the project and exit | | --config <path> | -c | Specify the config file | | --port <number> | -p | Specify the server port, overrides config | | --base-url <path> | -u | Set the base URL prefix for markup, overrides config | | --quiet | -q | Hide the header and the server/livereload info lines |

The --base-url flag is particularly useful for CI/CD pipelines where the deploy path may differ per environment:

poops --build --base-url /blog

The --quiet flag drops the 💩 Poops — vX.Y.Z header (and its terminal bell) plus the Local server / Network / Live reload lines. Handy when you run several Poops instances side by side and only want to see which one is compiling:

poops -q & poops -q -c site/poops.json

Build logs, warnings and errors are unaffected — --quiet only removes the banner.

If you have installed Poops locally you can run it with npx poops or npx 💩 or add a script to your package.json:

{
  "scripts": {
    "build": "npx poops" // or "npx 💩"
  }
}

Configuration

Configuring Poops is simple 😌. Let's presume that we have a example/src/scss and example/src/js directories and we want to bundle the files into example/dist/css and example/dist/js. If you also have markup files, you can use Nunjucks (default) or Liquid templating engine to generate HTML files from your templates. Let's presume that we have a example/src/markup directory and we want to generate HTML files in the root of the your directory.

Just create a poops.json file in the root of your project and add the following (you can see this sample config in this repo's root):

{
  "scripts": [
    {
      "in": "example/src/js/main.ts",
      "out": "example/dist/js/scripts.js",
      "options": {
        "sourcemap": true,
        "minify": true,
        "justMinified": false,
        "format": "iife",
        "target": "es2019"
      }
    }
  ],
  "reactor": [
    {
      "component": "example/src/js/App.jsx",
      "inject": "app_html",
      "in": "example/src/js/app-hydrate.jsx",
      "out": "example/dist/js/app-hydrate.js",
      "options": {
        "minify": true,
        "target": "es2019"
      }
    }
  ],
  "styles": [
    {
      "in": "example/src/scss/index.scss",
      "out": "example/dist/css/styles.css",
      "options": {
        "sourcemap": true,
        "minify": true,
        "justMinified": false
      }
    }
  ],
  "markup": {
    "in": "example/src/markup",
    "out": "/",
    "options": {
      "engine": "nunjucks",
      "site": {
        "title": "Poops",
        "description": "A super simple bundler for simple web projects."
      },
      "data": ["data/links.json", "data/poops.yaml"],
      "includePaths": ["_layouts", "_partials"]
    }
  },
  "copy": [
    {
      "in": "example/src/static",
      "out": "example/dist"
    }
  ],
  "banner": "/* {{ name }} v{{ version }} | {{ homepage }} | {{ license }} License */",
  "serve": {
    "port": 4040,
    "base": "/"
  },
  "livereload": true,
  "watch": ["src"],
  "includePaths": ["node_modules"]
}

Every property is optional, but give Poops nothing to compile — no scripts, styles, postcss, markup, reactor, images or copy — and it exits 0 having written nothing. If you don't have anything to consume, you won't poop. 💩

You can freely remove the properties that you don't need. For example, if you don't want to run a local server, just remove the serve property from the config.

Key checking and editor completion (`$schema`)

A mistyped key is not a build error. A top-level "stlyes", an "inn" in a styles entry, an "engnie" in markup.options — each is read by nothing, and the build stays green with the file it should have written simply missing. You find out when you look.

Poops names every one of them at startup, reading the same schema your editor does:

[info][warn] Unknown key "inn" in styles[0] — ignored. Valid: in, out, options

Key names only, and only in the blocks Poops owns. images belongs to poops-images and site is yours to name, so an unrecognised key in either passes without comment. Types are checked by nobody here: "minify": "yes" reaches the compiler and fails there, loudly.

The schema also drives editor completion and inline docs for every key. Point $schema at the copy in your node_modules:

{
  "$schema": "./node_modules/poops/schema/poops.schema.json",
  "scripts": [{ "in": "src/js/main.ts", "out": "dist/js/app.js" }]
}

Or at the hosted copy, which needs nothing installed:

{
  "$schema": "https://stamat.info/poops/poops.schema.json"
}

VS Code, JetBrains and anything else speaking the language server protocol read it from the file itself. To attach it without touching your config, map it in VS Code's settings.json instead — the same file, matched by name:

{
  "json.schemas": [
    {
      "fileMatch": ["poops.json", "💩.json"],
      "url": "https://stamat.info/poops/poops.schema.json"
    }
  ]
}

The $schema key itself is inert — Poops reads it, recognises it, and does nothing with it. The URL is your editor's business: the startup check reads the copy inside node_modules/poops, so pointing $schema at the hosted file, at a stale one, or leaving it out changes nothing about what the CLI says. Nothing is added to what Poops installs into your project either way.

Blocks belonging to another package. poops.json is shared: septic reads a septic block out of the same file, and Poops has no business calling that a mistake. So an unknown top-level key is accepted in silence when a package by that name is in your dependencies, devDependencies, peerDependencies or optionalDependencies — declared is enough, and Poops never loads it. A key can also arrive one step removed — laxative brings septic, so your package.json says laxative and never septic — and for that a direct dependency vouches for the key in its own manifest: "poops": { "companionKeys": ["septic"] }. Only direct dependencies are read, one directory deep. Nothing declared and nothing vouched, and the key is warned about as before, which is what catches the typo.

Your editor cannot see your node_modules, so the schema cannot make that distinction. It allows an object under any name it does not know and rejects everything else: "stlyes": [ … ] is still flagged, "srve": { … } is not. That is the price of one shared config file, and the CLI still catches what the editor lets through.

A companion that owns a block describes it in its own schema — septic does — and $schema takes one URL, so having both checked means composing them in a local file. Each package's README carries its schema URL and that two-line allOf; this one deliberately does not repeat them, since a URL copied into two repos is a URL that goes stale in one.

The schema is hand-written and version-controlled beside the code, so it can drift from it. Poops' own test suite validates it against the draft-07 meta-schema, then validates poops.json and every complete example in this README and the documentation site against it — so an example that stops being valid config fails the build. Its top-level keys are asserted to be exactly the set poops.js accepts, its exec keys exactly the ones that fire, and its markup.options a superset of the ones the markup engine reads. A per-entry options object — mostly esbuild's and PostCSS's, not Poops' — has no such list, so a wrong type there is caught but a missing option is not. If the editor does not offer an option this page documents, the schema is behind and that is worth reporting.

Scripts

Scripts are bundled with esbuild. Supports .js, .ts, .jsx, and .tsx files out of the box — including React and other JSX frameworks. You can specify multiple scripts to bundle. Each script has the following properties:

in - the input path, can be a file path, an array of file paths, or a glob pattern (e.g. "src/js/*.js", "src/elements/*/index.{js,mjs,cjs,jsx,ts,tsx}"). Globs must use / separators (even on Windows)
out - the output path, can be a directory or a file path. With multiple inputs it must be a directory — entry points from different directories nest their output under the common ancestor (esbuild's outbase). A glob-matched index.* is named after its directory instead, placed relative to the glob's static prefix: "src/elements/*/index.js" gives <out>/accordion.js, while "src/*/accordion/index.js" keeps the differing segment as <out>/blocks/accordion.js. A literal in: "src/index.js" keeps its own basename. To name outputs yourself, out can be a template: {{dir}} is the input's directory relative to the glob's static prefix, {{name}} its basename without extension. "src/elements/*/widget.ts" with out: "dist/js/{{dir}}-{{name}}.js" gives dist/js/accordion-widget.js and dist/js/tabs-widget.js — one bundle per match, named by you instead of by the common ancestor. The template's extension is honoured too, so out: "dist/esm/{{dir}}.mjs" writes .mjs files
options - the options for the bundler. You can apply most of the esbuild options that are not in conflict with Poops. See esbuild's options for more info.

Options:

sourcemap - whether to generate sourcemaps or not, sourcemaps are generated only for non-minified files since they are useful for debugging. Default is false. This is a direct esbuild option
minify - whether to minify the output or not, minification is performed by esbuild and is only applied to non-minified files. Default is false
justMinified - whether you want to have a minified file as output only. Removes the non-minified file from the output. Useful for production builds. Default is false
format - the output format, can be iife or esm or cjs - this is a direct esbuild option
target - the target for the output, can be es2018 or es2019 or es2020 or esnext for instance - this is a direct esbuild option. Default is es2020
jsx - the JSX transform mode, can be transform (default) or automatic. Use automatic for React 17+ JSX runtime which doesn't require importing React in every file - this is a direct esbuild option
nodePaths - extra directories to resolve bare imports from, for this entry only. Merged with the top-level includePaths rather than replacing it - this is a direct esbuild option

scripts property can accept an array of script configurations or just a single script configuration. If you want to bundle multiple scripts, just add them to the scripts array:

{
  "scripts": [
    {
      "in": "src/js/main.ts",
      "out": "dist/js/scripts.js",
      "options": {
        "sourcemap": true,
        "minify": true,
        "justMinified": false,
        "format": "iife",
        "target": "es2019"
      }
    },
    {
      "in": "src/js/other.ts",
      "out": "dist/js/other.js",
      "options": {
        "sourcemap": true,
        "minify": true,
        "justMinified": false,
        "format": "iife",
        "target": "es2019"
      }
    }
  ]
}

JSX/TSX (React) Example

To bundle a React app, just point in to your .jsx or .tsx entry file:

{
  "scripts": [
    {
      "in": "src/js/app.jsx",
      "out": "dist/js/app.js",
      "options": {
        "minify": true,
        "format": "iife",
        "jsx": "automatic"
      }
    }
  ]
}

Setting jsx to automatic uses React's JSX runtime (React 17+), so you don't need import React from 'react' in every file. If you omit jsx or set it to transform, the classic React.createElement transform is used.

As noted earlier, if you don't want to bundle scripts, just remove the scripts property from the config.

Reactor (React Pre-rendering)

The reactor config key defines React components that are pre-rendered to HTML at build time (SSG) and optionally hydrated on the client. This is a separate pipeline from scripts — reactor entries have their own build step, watcher path, and logging tag.

Each reactor entry has the following properties:

component — the file that default-exports a React component (rendered at build time with renderToString)
inject — template global variable name for the rendered HTML (available in both Nunjucks and Liquid)
in (optional) — client entry file for hydration (bundled for the browser)
out (optional) — output path for the client bundle
options (optional) — esbuild options for the client bundle (same as script entries: minify, format, target, sourcemap, etc.)

{
  "reactor": [
    {
      "component": "src/js/App.jsx",
      "inject": "app_html",
      "in": "src/js/app-hydrate.jsx",
      "out": "dist/js/app-hydrate.js",
      "options": {
        "minify": true,
        "target": "es2019"
      }
    }
  ]
}

In your templates, use the inject name to insert the rendered HTML:

<div id="root">{{ app_html | safe }}</div>
<script src="js/app-hydrate.min.js"></script>

If you only need server-side rendering without client hydration, omit in and out:

{
  "reactor": [
    {
      "component": "src/js/App.jsx",
      "inject": "app_html"
    }
  ]
}

How it works:

Poops bundles the component with react-dom/server for Node.js and calls renderToString
The rendered HTML is stored and made available as a template global variable
If in/out are specified, the client entry is bundled for the browser
At runtime, React hydrates the pre-rendered HTML, making it interactive

Poops does not need react or react-dom as its own dependency — they are resolved from your project's node_modules. In watch mode, changes to files in the reactor component's directory trigger re-rendering and client re-bundling. Markup is recompiled only when the rendered output actually changes. Changes to other JS/TS files only trigger the scripts pipeline — the two are independent.

[!NOTE] If you don't need server-side pre-rendering, you can bundle a React app entirely through the regular scripts pipeline — just point in to your .jsx/.tsx entry file and use createRoot on the client. The reactor config is only needed when you want build-time HTML rendering with optional hydration.

Styles

Styles are bundled with Dart Sass. You can specify multiple styles to bundle. Each style has the following properties:

in - the input path, can be a file path, an array of file paths, or a glob pattern (e.g. "src/scss/*.scss", "src/elements/*/index.{scss,sass,css}"). Globs must use / separators (even on Windows) and skip Sass partials (_*.scss). Each matched file is compiled separately
out - the output path, can be a directory or a file path. With multiple inputs it must be a directory — each input compiles to <out>/<basename>.css, so inputs sharing a basename (e.g. a/main.scss and b/main.scss) will overwrite each other. A glob-matched index.* is named after its directory instead, placed relative to the glob's static prefix: "src/elements/*/index.scss" gives <out>/accordion.css, while "src/*/accordion/index.scss" keeps the differing segment as <out>/blocks/accordion.css. A literal in: "src/scss/index.scss" keeps its own basename. To name outputs yourself, out can be a template: {{dir}} is the input's directory relative to the glob's static prefix, {{name}} its basename without extension. "src/elements/*/theme.scss" with out: "dist/{{dir}}-theme.css" gives dist/accordion-theme.css and dist/tabs-theme.css — one output per match, instead of every theme.scss overwriting the same theme.css
options - the options for the bundler.

Options:

sourcemap - whether to generate sourcemaps or not, sourcemaps are generated only for non-minified files since they are useful for debugging. Default is false
minify - whether to minify the output or not, minification is performed by esbuild. Default is false
justMinified - whether you want to have a minified file as output only. Removes the non-minified file from the output. Useful for production builds. Defaults to false.
tokenPaths - a string or array of directory paths containing JSON design token files. Enables the sass-token-importer which lets you @use JSON tokens directly in SCSS. Supports W3C DTCG and Style Dictionary formats with auto-detection.
tokenOutput - output mode for design tokens: "variables" (default) generates flat SCSS variables, "map" generates nested Sass maps.
resolveAliases - whether to resolve {path.to.token} alias references in design tokens. Default is true.

styles property can accept an array of style configurations or just a single style configuration. If you want to bundle multiple styles, just add them to the styles array:

{
  "styles": [
    {
      "in": "src/scss/main.scss",
      "out": "dist/css/styles.css",
      "options": {
        "sourcemap": true,
        "minify": true,
        "justMinified": false
      }
    },
    {
      "in": "src/scss/other.scss",
      "out": "dist/css/other.css",
      "options": {
        "sourcemap": true,
        "minify": true,
        "justMinified": false
      }
    }
  ]
}

Design Tokens

You can import JSON design token files directly into your SCSS using the token: prefix. Define your tokens in JSON once and use them as SCSS variables — no manual variable files to keep in sync.

Given a token file src/tokens/colors.json:

{
  "color": {
    "$type": "color",
    "primary": { "$value": "#0066cc" },
    "secondary": { "$value": "#ff6600" },
    "link": { "$value": "{color.primary}" }
  }
}

Add tokenPaths to your styles config:

{
  "styles": [
    {
      "in": "src/scss/index.scss",
      "out": "dist/css/styles.css",
      "options": {
        "tokenPaths": ["src/tokens"]
      }
    }
  ]
}

Then use the token: prefix in your SCSS:

@use "token:colors" as c;

.btn {
  color: c.$color-primary;
}
.btn:hover {
  color: c.$color-secondary;
}
a {
  color: c.$color-link; // resolved from {color.primary} → #0066cc
}

For Sass maps instead of flat variables, set "tokenOutput": "map":

@use "sass:map";
@use "token:colors" as c;

.btn {
  color: map.get(c.$color, primary);
}

As noted earlier, if you don't want to bundle styles, just remove the styles property from the config.

PostCSS (optional)

Process CSS files with PostCSS and any PostCSS plugins. This is a separate pipeline from Styles (Sass) — use it for tools like Tailwind CSS, Autoprefixer, or any other PostCSS plugin.

PostCSS and its plugins are not bundled with Poops. You need to install them in your project:

npm i -D postcss

Each PostCSS entry has the following properties:

in - the input CSS file path
out - the output path, can be a directory or a file path
options - options for the pipeline

Options:

plugins - an array of PostCSS plugin names to load. Each entry can be a string (plugin name) or a tuple ["plugin-name", { options }] for passing options to the plugin.
minify - whether to minify the output using esbuild. Default is false
justMinified - output only the minified file. Default is false

Source maps: an input CSS ending in a sourceMappingURL has its map composed with the one PostCSS produces, written next to the output. Point in at a Sass output built with "sourcemap": true and a rule still traces to the .scss line it came from, through both passes. An input carrying no map produces none.

postcss property can accept an array of configurations or a single configuration:

{
  "postcss": {
    "in": "src/css/main.css",
    "out": "dist/css/main.css",
    "options": {
      "plugins": ["@tailwindcss/postcss"],
      "minify": true
    }
  }
}

You can also pass options to plugins using the tuple form:

{
  "postcss": {
    "in": "src/css/main.css",
    "out": "dist/css/main.css",
    "options": {
      "plugins": [["autoprefixer", { "grid": true }]]
    }
  }
}

Build order: PostCSS runs after Styles and Markups in the build pipeline. This means PostCSS plugins can reference the compiled markup output (e.g. Tailwind scanning HTML for utility classes). In watch mode, PostCSS is re-triggered after Styles or Markups recompile.

Tailwind CSS Example

Install the deps, then use a config like this:

npm i -D postcss @tailwindcss/postcss tailwindcss

{
  "postcss": {
    "in": "src/css/main.css",
    "out": "dist/css/main.css",
    "options": {
      "plugins": ["@tailwindcss/postcss"],
      "minify": true
    }
  },
  "markup": {
    "in": "src/markup",
    "out": "dist",
    "options": {
      "site": {
        "title": "Poops + Tailwind",
        "description": "A Tailwind CSS example for Poops"
      },
      "includePaths": ["_layouts", "_partials"]
    }
  },
  "serve": { "port": 4040, "base": "/dist" },
  "livereload": true,
  "watch": ["src"]
}

The CSS entry file (src/css/main.css) simply imports Tailwind:

@import "tailwindcss";

Then use Tailwind utility classes directly in your markup templates. Tailwind v4 auto-detects content sources, so no tailwind.config.js is needed.

Using Sass + Tailwind together: If you want both Sass and Tailwind, keep them as separate pipelines writing to separate output files. The Sass pipeline compiles .scss to CSS, while the PostCSS pipeline handles Tailwind independently. They don't need to chain into each other unless you want PostCSS to post-process the Sass output (e.g. with Autoprefixer) — in that case, point postcss.in to the Sass output file and postcss.out to a different file so the original Sass output is preserved for re-processing.

Markups

markup has the same shape as a scripts or styles entry: in, out, and everything else under options.

{
  "markup": {
    "in": "src/markup",
    "out": "dist",
    "options": {
      "engine": "nunjucks",
      "site": { "title": "My Awesome Site" }
    }
  }
}

[!WARNING] Deprecated: Poops 1.x also read these keys directly on markup ("markup": { "site": … }). That still works in 2.x and warns; it stops working in 3.0. Move them into options.

Options:

engine (optional) - the template engine to use. Can be "nunjucks" (default) or "liquid". Nunjucks is a Mozilla template engine inspired by Jinja2. Liquid is a Shopify-compatible template engine. Both engines support the same tags, filters, collections, search index, sitemap, and navigation tree features documented below.
in (entry, not an option) - the input path, can be a directory or a file path, but please just use it as a directory path for now. All files in this directory will be processed and the structure of the directory will be preserved in the output directory with exception to directories that begin with an underscore _ will be ignored.
out (entry, not an option) - the output path, can be only a directory path (for now)
site (optional) - global data that will be available to all templates in the markup directory. Like site title, description, social media links, etc. You can then use this data in your templates {{ site.title }} for instance. Values carry the same package.json tokens a banner does, filled at build time and at any depth: "footer": "v{{ version }}" prints the version your package.json holds, so it cannot drift from the released one. A token your package.json has no field for is left as written rather than becoming the word undefined.
data (optional) - is an array of JSON or YAML data files, that once loaded will be available to all templates in the markup directory. If you provide a path to a file for instance links.json with a facebook property, you can then use this data in your templates {{ links.facebook }}. The base name of the file will be used as the variable name, with spaces, dashes and dots replaced with underscores. So the awesome-links.json will be available as {{ the_awesome_links.facebook }} in your templates. The root directory of the data files is in directory. So if you have a data directory in your in directory, you can specify the data files like this data: ["data/links.json"]. The same goes for the YAML files.
includePaths - an array of paths to directories that will be added to the template engine's include paths. Useful if you want to separate template partials and layouts. For instance, if you have a _includes directory with a header.njk (or header.liquid) partial that you want to include in your markup, you can add it to the include paths and then include the templates like this {% include "header.njk" %}, without specifying the full path to the partial.
baseURL (optional) - a base URL prefix to use instead of relative path prefixes. When set, {{ relativePathPrefix }} will always resolve to this value (with a trailing slash ensured) instead of being computed relative to each page's depth. Useful when deploying under a subdirectory (e.g. "/blog" for domain.com/blog/). When not set, relative prefixes (./, ../, etc.) are used, which work for any deployment location including subdirectories and file:// URLs.
dateFormat (optional) - the default dayjs format the date filter uses when called without an argument. With neither set, date returns the value untouched rather than guessing a format
autoescape (optional) - Nunjucks only. Escape template output by default, so {{ value }} cannot inject HTML and anything meant as markup needs | safe. Defaults to false, since a static site mostly renders content you wrote. Turn it on when templates interpolate anything you did not. The Liquid engine ignores it — liquidjs does not escape by default and Poops does not make it
collections (optional) - the collections to build, if you would rather declare them here than in front matter. See Collections & Pagination
lastUpdated (optional) - keep a "last updated" date per page without hand-maintaining one. true writes the index to .poops-updates.json; a string names the file. See Last updated dates

[!TIP] If, for instance, you are building a simple static onepager for your library, and want to pass a version variable from your package.json, Poops automatically reads your package.json if it exists in your working directory and sets the global variable package to the parsed JSON. So you can use it in your markup files, for example like this: {{ package.version }}.

"Edit this page on GitHub" links. Every page exposes page.filePath — its source file path relative to your project root, with posix separators (e.g. src/markup/docs/index.md). That is exactly the path GitHub's editor expects, so an edit link is one line in your layout:

{% set repoUrl = site.repo or package.homepage %}
{% if page.filePath and repoUrl %}
<a href="{{ repoUrl }}/edit/{{ site.branch or 'main' }}/{{ page.filePath }}">✏️ Edit this page on GitHub</a>
{% endif %}

Put repo and branch in your site data (they fall back to package.homepage and main). Don't rebuild the path from page.url — that is the output URL (.html, and index.md collapses to a directory), so it can't be reversed to the .md source.

Here is a sample markup configuration using the default Nunjucks engine:

{
  "markup": {
    "in": "src/markup",
    "out": "dist",
    "options": {
      "site": {
        "title": "My Awesome Site",
        "description": "This is my awesome site"
      },
      "data": ["data/links.json", "data/other.yaml"],
      "includePaths": ["_includes"],
      "baseURL": "/blog"
    }
  }
}

To use Liquid instead, set the engine property:

{
  "markup": {
    "in": "src/liquid",
    "out": "dist",
    "options": {
      "engine": "liquid",
      "site": {
        "title": "My Awesome Site",
        "description": "This is my awesome site"
      },
      "data": ["_data/links.json", "_data/other.yaml"],
      "includePaths": ["_layouts", "_partials"]
    }
  }
}

If your project doesn't have markups, you can remove the markup property from the config entirely. No code will be executed for this property.

Nunjucks vs Liquid

Both engines support the same feature set (collections, pagination, search index, sitemap, navigation tree, custom tags, and filters). The main differences are in template syntax:

| Feature | Nunjucks | Liquid | | -------------- | ----------------------------- | ------------------------------------- | | File extension | .njk | .liquid | | Inheritance | {% extends "base.html" %} | {% layout "base.liquid" %} | | Default values | {{ x or "fallback" }} | {{ x \| default: "fallback" }} | | Contains check | {% if "x" in items %} | {% if items contains "x" %} | | Safe output | {{ html \| safe }} | {{ html }} (no escaping by default) | | Includes | {% include "partial.njk" %} | {% render "partial.liquid" %} |

Both engines process .html and .md files in addition to their native extension.

Templates from an npm package

Layouts and partials can live in an installed package, so a shared theme ships as a dependency instead of copied files. Reference it by package name — any include/extend name containing a / is resolved from node_modules:

{% extends "my-theme/layout.html" %}
{% block content %}
  <h1>{{ page.title }}</h1>
{% endblock %}

Or from front matter, so the page carries no template syntax:

---
layout: my-theme/layout
---

A theme package must:

Not restrict subpaths with exports — or map its templates explicitly, e.g. "exports": { "./*": "./*" }. Otherwise Node blocks resolving the .html files by path.
Reference its own partials relatively — {% import "./nav.html" as nav %}, not the bare name. A bare name (no /) is always searched in the consumer's project only, never the package.

Bundled filters (toc, breadcrumb, og, canonical, …) are engine-global, so package templates use them with no extra wiring.

Liquid resolves package templates the same way — node_modules is on its include roots, so {% layout "my-theme/layout.liquid" %} and {% render "my-theme/partial.liquid" %} resolve by package name too (a Liquid theme ships .liquid files). The exports/relative-partial rules above apply the same, except containment is by include root rather than the / name gate.

Custom Engines

The engine option also accepts a module specifier — an npm package name or a path relative to your project root. The module's default export must be an engine class:

{
  "markup": {
    "in": "src/markup",
    "out": "dist",
    "options": { "engine": "poops-shopify" }
  }
}

An engine class implements this contract (see lib/markup/engines/ for the two built-in reference implementations):

export default class MyEngine {
  constructor(templatesDir, includePaths, options) {} // options: { autoescape }
  get markupExtensions() {
    return "html|liquid|md";
  } // glob alternation of processed extensions, no dots
  get indexableExtensions() {
    return new Set([".html"]);
  } // extensions eligible for collections, search index and nav
  registerFilters({ dateFormat, markupOut }) {}
  registerTags(getOutputDir) {}
  setGlobal(key, value) {}
  removeGlobal(key) {}
  async render(templatePath, context) {
    return "html";
  } // templatePath is an absolute file path
}

That is the whole required surface. Five more members are feature-detected with a typeof check — implement what your engine can, skip the rest:

| Member | Gives you | Without it | | ---------------------------------- | --------------------------------------------------------------------------------------------------------------- | -------------------------------------------------- | | invalidate(file) | Drop the compiled templates backed by a changed or deleted path (prefix-match, so a deleted directory is covered) | clearCache() is called on every watch compile | | clearCache() | Wipe the whole template cache | No cache management at all | | pagesDependingOn(file) | Re-render only the pages that loaded an edited partial or layout | Any markup edit triggers a full markup compile | | replaceOutExtensions(outputPath) | Map your source extension to a different output one | The default maps .md/.njk/.liquid to .html | | isMarkupSource(absPath) | Claim a file the glob would not call markup, so watch routes it to the markup pipeline | Only markupExtensions matches count |

The built-in engines also carry fileExtension and renderString, but the pipeline never calls either — don't implement them and don't rely on them. The full lifecycle (what Poops calls, in what order, with what) is in the engine API docs.

The easiest starting point is extending a built-in engine — deep imports are intentionally supported for this:

import LiquidEngine from "poops/lib/markup/engines/liquid.js";

export default class MyEngine extends LiquidEngine {
  registerFilters(opts) {
    super.registerFilters(opts);
    this.engine.registerFilter("shout", (str) => String(str).toUpperCase());
  }
}

Collections & Pagination

Collections turn a directory of pages into a sorted, optionally paginated list — blog posts, changelog entries, documentation. A collection maps to a direct subdirectory of your markup in directory: every .html, .njk, .liquid or .md file inside it (except the index.* file) becomes a collection item.

There are two ways to declare a collection:

1. Front matter auto-discovery — add collection to the front matter of the directory's index file:

---
title: Changelog
collection: true
paginate: 10
sort: date
---

collection: true uses the directory name as the collection name; a string (e.g. collection: changelog) names it explicitly. paginate and sort are optional.

2. Config — list collections in the markup config. The name must match a subdirectory of in:

{
  "markup": {
    "in": "src/markup",
    "out": "dist",
    "options": {
      "collections": [
        "changelog",
        {
          "name": "blog",
          "paginate": 5,
          "sort": { "by": "title", "order": "asc" }
        }
      ]
    }
  }
}

Sorting. By default items are sorted by date, newest first. sort can be a field name shorthand ("sort": "title") or an object { "by": "field", "order": "asc" | "desc" }. Sorting by date compares dates (default order desc); any other field compares alphabetically (default order asc).

Items. Each item exposes its own front matter plus properties Poops adds:

url - the item's output path relative to the site root (e.g. changelog/my-post.html)
title - falls back to the file name if not set in front matter
date - falls back to the file's modification time if not set, with a build warning. Set a real date in front matter — mtime is meaningless on CI checkouts (git clone resets it), so undated posts will reshuffle between deploys.
wordcount, excerpt (first paragraph, plain text — a meta-description fallback), fileName, filePath, collection
A collection item is read without a page context, so an item whose first paragraph is built from template tags gets an empty excerpt rather than a guess — listings and feeds fall back to its description. The same page gets its excerpt resolved when it is built on its own.

An item with published: false in its front matter is excluded from the collection and its page is not built.

Using collections in templates. Every collection is available as a global variable named after it, on every page:

{% for post in changelog.items %}
  <a href="{{ relativePathPrefix }}{{ post.url }}">{{ post.title }}</a> — {{ post.date | date }}
{% endfor %}

Pagination. With paginate: N set, the collection's index file is rendered once per page of N items: page 1 to out/changelog/index.html, page 2 to out/changelog/2/index.html, and so on. Inside the index template the collection object carries the page state:

| Variable | Description | | --------------------------- | ------------------------------------------------------- | | pageItems | the items on the current page | | pageNumber / totalPages | current page (1-based) / total page count | | pageUrl | URL of the current page (changelog, changelog/2, …) | | nextPage / nextPageUrl | next page number / URL, null on the last page | | prevPage / prevPageUrl | previous page number / URL, null on the first page |

From the example site's changelog/index.html:

{% for post in changelog.pageItems %}
  <div class="post">
    <h2><a href="{{ relativePathPrefix }}{{ post.url }}">{{ post.title }}</a></h2>
    <div class="date">{{ post.date | date }}</div>
    {{ post.description }}
  </div>
{% endfor %}

{% if changelog.totalPages > 1 %}
  {% if changelog.nextPageUrl %}<a href="{{ relativePathPrefix }}{{ changelog.nextPageUrl }}">Next</a>{% endif %}
  {{ changelog.pageNumber }} of {{ changelog.totalPages }}
  {% if changelog.prevPageUrl %}<a href="{{ relativePathPrefix }}{{ changelog.prevPageUrl }}">Previous</a>{% endif %}
{% endif %}

Or use the {% pagination %} shorthand tag (available in both engines), which renders Previous/Next links and a "page of total" counter — with relativePathPrefix applied — and outputs nothing when there is only one page:

{% pagination changelog %}

Pages 2..N automatically get a distinct <title> — Changelog — Page 2 — so paginated pages don't all share the landing page's title (and its og/jsonld metadata). Page 1 keeps its own title.

Localizing the labels. The — Page N title suffix and the {% pagination %} tag's Previous/Next/of wording default to English. Override them site-wide under site.pagination:

{
  "markup": {
    "options": {
      "site": {
        "pagination": {
          "title": "{title} — Seite {n}",
          "prev": "Zurück",
          "next": "Weiter",
          "of": "von"
        }
      }
    }
  }
}

title accepts {title}, {n} and {total} tokens and applies to pages 2..N (and taxonomy term pages); prev/next/of localize the {% pagination %} tag ({n} of {total} → {n} von {total}).

Item pages themselves are compiled like any other markup file, preserving the directory structure: src/markup/changelog/my-post.md → dist/changelog/my-post.html. A collection directory without an index file still builds its items and exposes the collection to templates — only the paginated listing pages are skipped.

Taxonomies (Tags & Categories)

A taxonomy turns a front-matter field (tags, categories, authors) into its own paginated, crawlable landing page per term — changelog/tag/feature/, blog/category/release/. Declare which fields become taxonomies on the collection, alongside paginate/sort — either in the index front matter or the config entry:

---
title: Changelog
collection: true
paginate: 10
taxonomies:
  - name: tags # front-matter field to group on
    path: tag # URL segment (defaults to name); "tag" for a singular URL
    paginate: 5 # per-term page size (defaults to the collection's paginate)
---

Shorthand: a bare string list (taxonomies: [tags, category]) uses each field name as the URL segment and inherits the collection's paginate. Array-valued fields split per element — a post with tags: [js, css] lands under both tag/js/ and tag/css/. Terms are slugified for the URL (Static Site → static-site).

Term pages render with the collection's own index template — no extra file. On a term page the collection object carries the term context; branch on activeTerm to render a term view:

{% if changelog.activeTerm %}
  <h1>Tagged {{ changelog.activeTerm | humanize }}</h1>
  {% for post in changelog.pageItems %}
    <a href="{{ relativePathPrefix }}{{ post.url }}">{{ post.title }}</a>
  {% endfor %}
  {% pagination changelog %}
{% endif %}

On a term page items/pageItems are scoped to that term (so pagination and groupby narrow to it too); activeTaxonomy holds the URL segment and activeTermSlug the slug. Build tag links anywhere from collection.taxonomies:

{% for tax in changelog.taxonomies %}
  {% for term in tax.terms %}
    <a href="{{ relativePathPrefix }}{{ term.url }}">{{ term.term | humanize }} ({{ term.count }})</a>
  {% endfor %}
{% endfor %}

Each term exposes term, slug, url, count and totalPages.

Term pages get a distinct <title> and og/jsonld metadata (Tag: Feature, paged Tag: Feature — Page 2), and the breadcrumb/jsonld filters resolve them to a Home › Collection › Tag: Term trail automatically (skipping the non-page tag/category URL segment). The Tag:/Category: label comes from path, so it localizes by naming the path in your language (path: etiqueta → Etiqueta: …). Term pages are listed in the sitemap but kept out of the search index and nav.

Custom Tags

image

Poops can generate responsive <img> elements with srcset attributes. Image processing (resize, format conversion) is handled externally — Poops discovers the generated variants on disk and produces the correct HTML markup.

Naming convention: Your image tool should output variants as {name}-{width}w.{ext}. For example, given photo.jpg, the expected variants are: photo-320w.jpg, photo-640w.jpg, photo-320w.webp, photo-640w.webp, etc.

{% image %} tag — generates a full <img> element:

Nunjucks:

{% image 'static/photo.jpg', alt='Hero', class='hero-img', sizes='(max-width: 640px) 100vw, 50vw' %}

Liquid:

{% image 'static/photo.jpg', alt: 'Hero', class: 'hero-img', sizes: '(max-width: 640px) 100vw, 50vw' %}

Output:

<img
  src="static/photo-640w.jpg"
  srcset="
    static/photo-320w.webp 320w,
    static/photo-640w.webp 640w,
    static/photo-960w.webp 960w
  "
  sizes="(max-width: 640px) 100vw, 50vw"
  alt="Hero"
  class="hero-img"
  loading="lazy"
/>

Scans the output directory for files matching {name}-{width}w.{ext}
Groups by format, prefers avif > webp > original format for srcset
Uses the middle-sized variant as src fallback
Prepends relativePathPrefix automatically
Defaults: sizes="100vw", loading="lazy"
Falls back to a plain <img src="..."> if no variants are found

Named crops with size — pass a size kwarg to build the <img> from a named crop/resize group instead of the default responsive widths. The whole group becomes its own srcset (each crop has its own aspect ratio), so a square thumbnail set, a wide banner set, etc. each get correct srcset/width/height:

{% image 'static/photo.jpg', size='thumb', alt='', sizes='240px' %}

<img
  src="static/photo-thumb-480w.webp"
  srcset="static/photo-thumb-480w.webp 480w, static/photo-thumb.webp 960w"
  width="480"
  height="480"
  sizes="240px"
  alt=""
  loading="lazy"
/>

This requires the poops-images compile cache (named-size widths are read from it). The size name matches a named entry in your images.sizes config. The largest member of the group is written without a width suffix (photo-thumb.webp) — poops still srcsets it at its real width from the cache.

poops-images integration: if a .poops-images-cache.json compile cache is found in the output directory (poops-images writes one next to the images it generates), Poops reads variants from it instead of scanning the directory. On top of the scan behavior above, the cache gives you:

width and height attributes on the <img> element (exact dimensions from the cache — prevents layout shift). Pass your own width/height kwargs to override.
Correct src when the source format was converted (e.g. photo.heic → photo.jpg), even when there are no size variants.
By default the srcset is built only from the plain {name}-{width}w.{ext} width variants. Named sizes (photo-thumb-480w.webp) and preprocessed outputs (photo-blurred-640w.jpg) are kept out of it — they are crops and effects with their own aspect ratios. Reach a named crop group on purpose with the size kwarg above (or the srcset filter's second argument).
EXIF metadata via the exif filter (see below).

googleFonts

Generates Google Fonts <link> tags with preconnect hints. Accepts an array of font names (strings) or font objects with weight/italic options.

Nunjucks (supports inline arrays):

{% googleFonts ["Open Sans", "Roboto"] %}

Liquid (pass a variable — inline arrays are not supported in Liquid syntax):

{% googleFonts fonts %}

Where fonts is defined in a data file (e.g. fonts.json):

["Open Sans", "Roboto"]

Output:

<link rel="preconnect" href="https://fonts.googleapis.com" />
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin />
<link
  href="https://fonts.googleapis.com/css2?family=Open+Sans&family=Roboto&display=swap"
  rel="stylesheet"
/>

With specific weights and italics (Nunjucks):

{% googleFonts ["DM Sans", {name: "Poppins", weights: [400, 700], ital: true}] %}

With specific weights and italics (Liquid — via data file):

["DM Sans", { "name": "Poppins", "weights": [400, 700], "ital": true }]

Font object options:

name — font family name
weights — array of weight values (e.g. [400, 700])
ital — set to true to include italic variants
display — font-display strategy, defaults to swap (Nunjucks only, as a keyword argument)

highlight

Syntax-highlights code blocks at build time using highlight.js, eliminating layout shift caused by client-side highlighting. Code is pre-highlighted in the HTML output — you only need the highlight.js CSS theme on the client, not the JS.

{% highlight %} tag — wraps a code block with syntax highlighting (same syntax in both engines):

{% highlight 'javascript' %}
const greet = (name) => {
  return `Hello, ${name}!`;
};
{% endhighlight %}

Output:

<pre><code class="hljs language-javascript"><span class="hljs-keyword">const</span> greet = <span class="hljs-function">...</span></code></pre>

The language argument is optional. If omitted, highlight.js will attempt to auto-detect the language.

Markdown code fences are also highlighted automatically at build time:

```json
{ "name": "poops" }
```

Registered languages: javascript/js, typescript/ts, css, scss, html, xml, json, bash/sh, shell, python/py, ruby/rb, php, java, c, cpp, csharp/cs, go, rust/rs, yaml/yml, markdown/md, sql, diff.

Mermaid fences are not highlighted. highlight.js has no mermaid grammar, so a ```mermaid fence used to fall through to auto-detection and come out wrapped in spans for whatever language it guessed. It now compiles to the markup mermaid documents — <pre class="mermaid">, no <code> element, the diagram source escaped inside:

```mermaid
flowchart TD
  poops[poops] ==>|builds| theme[poops-docs-theme]
  theme -->|documents| poops

  poops ==> boe[book-of-elementals] & hg[hydrargyri] & septic[septic] & sulph[sulphuris] & more[…14 more]
  theme --> boe & hg & septic & sulph
```

<pre class="mermaid">flowchart TD
  poops[poops] ==&gt;|builds| theme[poops-docs-theme]
  theme --&gt;|documents| poops

  poops ==&gt; boe[book-of-elementals] &amp; hg[hydrargyri] &amp; septic[septic] &amp; sulph[sulphuris] &amp; more[…14 more]
  theme --&gt; boe &amp; hg &amp; septic &amp; sulph</pre>

> and & come out escaped, which is what mermaid wants — it reads the element's textContent, and that hands the characters back exactly as they were written.

Poops ships no mermaid and injects no script — loading it is yours to do, which keeps a 700 KB library off every page that has no diagram:

<script type="module">
  import mermaid from 'https://cdn.jsdelivr.net/npm/mermaid@11/dist/mermaid.esm.min.mjs'
</script>

Two honest caveats. Rendering happens in the browser, so a diagram pops in after paint — unlike code blocks, which are highlighted at build time precisely to avoid that; a min-height on pre.mermaid bounds the jump. And with the script absent or JS off, the fence shows its diagram source as text, which reads fine but is not a picture. For zero client JS, render the diagrams to SVG in an exec script before the markup stage and emit those instead.

Fence info strings. Only the first word names the language. Anything after it is carried onto the <code> element instead of being dropped: a bare word becomes a class, a key=value token becomes a data- attribute. This is how a fence marks itself for a later stage — a post-markup exec script that turns code.preview blocks into live demos, for example — without a marker comment in the markdown.

```html preview tab=options widths=375,768
<my-element></my-element>
```

<pre><code class="hljs language-html preview" data-tab="options" data-widths="375,768">…</code></pre>

Values are single tokens — no quotes, no spaces. A trailing = with nothing after it emits a valueless attribute (expanded= → data-expanded=""), for a flag you want to read with hasAttribute rather than as a class. The same applies to the {% highlight %} tag and the highlight filter.

pagination

Renders Previous/Next links and a "page of total" counter for a paginated collection, with relativePathPrefix already applied, and outputs nothing when there is only one page. Same syntax in both engines:

{% pagination changelog %}

The page state it reads, and how to localize its labels, are under Collections & Pagination.

Custom Filters

All filters are available in both engines. The only syntax difference is how arguments are passed: Nunjucks uses parentheses | filter("arg"), Liquid uses a colon | filter: "arg".

slugify — slugifies a string. Usage: {{ "My Awesome Title" | slugify }} will output my-awesome-title
humanize — the inverse of slugify: turns a slug or raw term into a display label. Usage: {{ "static-site" | humanize }} will output Static Site
jsonify — serializes a value to JSON. Usage: {{ myObject | jsonify }}
markdown — renders a markdown string to HTML with GitHub Flavored Markdown extras: emoji shortcodes (e.g. :rocket: → 🚀), alert callouts (> [!NOTE], [!TIP], [!IMPORTANT], [!WARNING], [!CAUTION], [!INFO]) and footnotes ([^1]). Code fences are syntax-highlighted and headings get slug ids plus permalink anchors — a heading built from a template tag is slugged from the words it renders, not from the tag, so # {{ site.title }} and # {{ page.title }} anchor wherever # My Site would. Usage: {{ "**bold** :rocket:" | markdown }}
toc — builds an on-this-page table of contents from rendered HTML: a <nav class="toc" aria-label="On this page"> listing every <h2> and <h3> that has an id, each <li> classed toc-h2/toc-h3 so you indent with CSS rather than nested lists. It reads the same ids the markdown heading renderer emits, so the links always land. Headings classed sr-only are skipped — a visible entry pointing at invisible content only confuses. Returns an empty string when there is nothing to list. Feed it rendered HTML: on a Markdown source, run markdown first, or code fences containing # lines get read as headings. Usage: {{ page.content | markdown | toc }}
date — formats a date string. Uses dayjs format tokens. A default format can be set via the dateFormat config option; with neither, the value is returned untouched.
- Nunjucks: {{ "2024-01-15" | date("MMMM D, YYYY") }}
- Liquid: {{ "2024-01-15" | date: "MMMM D, YYYY" }}
concat — returns a new array with the value appended (does not mutate the original):
- Nunjucks: {{ items | concat("c") }}
- Liquid: {{ items | concat: "c" }}
push — appends a value to an array in place (mutates the original):
- Nunjucks: {{ items | push("c") }}
- Liquid: {{ items | push: "c" }}
svg — reads an SVG file and injects it inline. The path is resolved relative to the project root. Returns empty string if the file doesn't exist or isn't an SVG. Usage: {{ 'src/icons/logo.svg' | svg }}
highlight — syntax-highlights a code string at build time using highlight.js. Takes an optional language argument. If the language is omitted, highlight.js will auto-detect it. Returns a <pre><code class="hljs"> block with highlighted markup.
- Nunjucks: {{ someCodeVariable | highlight('javascript') }}
- Liquid: {{ someCodeVariable | highlight: 'javascript' }}
og — generates Open Graph (and a Twitter card) <meta> tags from a page's front matter and your site data, for link previews on social/chat platforms. Put it in your layout <head>. og:type auto-detects: article when the page has a date, otherwise website.
- Nunjucks: {{ page | og(site) }}
- Liquid: {{ page | og: site }}
Emits og:title, og:description (a missing description falls back to the page's auto-excerpt, then site.description — the excerpt is taken after the template engine has run, so a first paragraph written as {{ site.description }} or supplied by an {% include %} is resolved, and one that resolves to nothing usable falls through rather than shipping the tag's source text), og:type, og:url (made absolute with site.url), og:site_name (from site.title), og:locale (page.lang/site.lang), og:image (page.image/site.image, made absolute), and twitter:card (summary_large_image when there's an image, else summary). For articles it adds article:published_time, article:modified_time and article:author. Attribute values are escaped. Set an og object in front matter to add or override any tag (e.g. og:image:alt, a fixed twitter:card):
```
---
title: My post
date: 2026-01-01
image: static/cover.jpg
og:
  "og:image:alt": Cover illustration
---
```
canonical — generates a <link rel="canonical"> tag pointing at a page's authoritative absolute URL (site.url + the page's url), the dedup signal that stops query-string and duplicate URLs splitting your ranking. Put it in your layout <head>. Front matter canonical overrides the target — an absolute URL as-is, or a path resolved against site.url (for cross-domain or hand-picked canonicals). The homepage canonicals to the site root. Returns nothing without site.url.
- Nunjucks: {{ page | canonical(site) }}
- Liquid: {{ page | canonical: site }}
description — generates the <meta name="description"> tag, from the same chain og and jsonld use: front matter description, then the page's auto-excerpt, then site.description. Put it in your layout <head>. Returns nothing when none of the three is set. Prefer it over writing the tag by hand: Poops renders with autoescape off, so content="{{ page.description }}" ships the front matter verbatim, and one " in a sentence closes the attribute and truncates the description to the words before it.
- Nunjucks: {{ page | description(site) }}
- Liquid: {{ page | description: site }}
jsonld — generates a schema.org JSON-LD <script type="application/ld+json"> block from a page's front matter and your site data, for GEO (Generative Engine Optimization) and structured data. Put it in your layout <head>. The @type auto-detects: BlogPosting when the page has a date, otherwise WebPage.
- Nunjucks: {{ page | jsonld(site) }}
- Liquid: {{ page | jsonld: site }}
It reads these front-matter fields when present: title, description (falls back to the page's auto-excerpt, then site.description), url (made absolute with site.url), date → datePublished, updated → dateModified, author (string or { name }, falls back to site.author), image, lang → inLanguage, and wordcount. publisher comes from site.title; set site.logo to add a publisher.logo ImageObject (made absolute) — Google Article rich results require it. Front-matter values are escaped so they can't break out of the <script> tag.
On the homepage (a page with no url) it also emits a site-level WebSite block with name + url, which declares the site name for search results. On nested pages (a url with at least one folder) it auto-appends a BreadcrumbList block derived from URL depth — a Google breadcrumb rich result, no extra markup (needs site.url for the absolute item URLs). See the breadcrumb filter below for a visible trail from the same data.
For full control, set a jsonld object in front matter — its keys are merged over (and override) the generated defaults, including @type:
```
---
title: How to brew coffee
date: 2026-01-01
jsonld:
  "@type": HowTo
  totalTime: PT5M
---
```
The same jsonld object works in your site data, as a site-wide default — useful when every page on the site is one type.

Published

Vulnerabilities

Links

Maintainers

Keywords

Readme

💩 Poops

Table of Contents

Features

Quick Start

Configuration

Key checking and editor completion ($schema)

Scripts

JSX/TSX (React) Example

Reactor (React Pre-rendering)

Styles

Design Tokens

PostCSS (optional)

Tailwind CSS Example

Markups

Nunjucks vs Liquid

Templates from an npm package

Custom Engines

Collections & Pagination

Taxonomies (Tags & Categories)

Custom Tags

image

googleFonts

highlight

pagination

Custom Filters

Key checking and editor completion (`$schema`)