config.lima

A site’s config.lima, in Lima, says what the site is and how hadley builds it. A value hadley reads and cannot use stops the build.

site

site:
  origin: https://example.org      # the address the site is published at
  owner: $(persons/ada)            # the person the site belongs to (themes.md, "Persons")
  since: 2026                      # the first year of the copyright line
  email: hello@example.org         # an address templates may show (site.email, site.emailMailto)
  default:
    language: en                   # the site's default language
    title: Ada's notes
  en:
    locale: en_GB
    claim: Notes on type and tools
    description: The site of Ada, who writes about type and tools.
  de:
    locale: de_DE
    claim: Notizen über Schrift und Werkzeuge
    description: Die Site von Ada, die über Schrift und Werkzeuge schreibt.

Each language of the site is a block named after it: its title, claim, tagline, description and locale; what a block leaves out comes from default. origin is the site’s, site.origin, unless a language block gives its own; default gives none. Templates read the values of the page’s language as site.title, site.claim and so on. issn names the site’s ISSN, which llms.txt and the JSON-LD give; repository its source, which llms.txt gives.

Lists and navigation

navigations:
  main: [about]                    # page ids, in their order
  footer: [about, imprint, feed]

profiles: [github, mastodon]       # the owner's profiles the site shows (themes.md, "Persons")

featured_articles:                 # shown above the others on the home page
  - id: first-steps
    label:
      en: Start here
      de: Zum Einstieg

featured_series:                   # the latest article of a collection, above the others
  - id: travel
    label:
      en: Series

A navigation links a page in the language of the page that shows it, or its version in the default language where that language has none; a page that is not listed is left out, and a dev build shows it marked. Templates read the navigations as navigations.main, navigations.footer, the featured articles as featuredList (themes.md).

Building

robots

robots:
  signal: search=yes, ai-input=no, ai-train=no

Adds Content-Signal to the generated robots.txt; it stays off until set. The value is a comma-separated list of name=yes or name=no, each name at most once: search means indexing and search results, excluding AI-generated summaries; ai-input means using content as input to AI answers; ai-train means training or fine-tuning models (Content Signals). These are requests to crawlers, not a technical lock. A site or theme’s own robots.txt takes precedence (deploy.md).

og images

With og: enabled: true, hadley makes og-default.jpg for the site and an og.jpg for every article and essay, from its cover or its first picture, with its title, or without a picture where its og:image is false (content.md). It needs a claim in every language and the token --og-background in the theme’s assets/css/style.css; a missing one stops the build with all of them named. The site’s logo (--og-logo in that stylesheet, else theme/assets/images/logo.svg or logo.png) stands at the bottom of a page’s og image and in the middle of og-default.jpg; a site without one has its title there, in bold. --og-background is the ground of og-default.jpg and of an og image without a picture, and shades the title. An opaque picture fills the og image, cut about its middle or where the page’s ogPosition says; a transparent one lies whole on --og-background. og-default.jpg is set straight and plain; a theme that wants its own look sets --og-accent (#rrggbb), which underlines the claim, or --og-tilt (an angle like -2deg), which turns the claim by it and the logo by half, or both, as teblo.de does. A cover whose style the theme draws through a colour matrix (themes.md, “Covers”) goes through it before it lies on --og-background, so that its og image shows it as its page does. The title is set in Noto Sans where the system has it (fonts-noto-core on Debian and Ubuntu). The JSON-LD of an article whose og image is AI-generated describes that image with what its XMP gives: its creator (dc:creator), its credit (photoshop:Credit, else the creator) and its description (dc:description).

Fields of a page

metadata declares the front matter fields a page may have beyond hadley’s own (content.md), and what a field must be, one of hadley’s too, and the other names it may have:

metadata:
  category:
    allowed: [notes, essays, tools]     # any other value is a problem of the page
  mood:
    required: true                       # a page without it is a problem
    defaultValue: calm
    aliases: [stimmung]                  # another name of the field
  author:
    aliases: [autor, autorin, autor:in]  # other names of one of hadley's fields
  vgwort_pixel:
    template: '<img src="{ value }" width="1" height="1" alt="">'
    publishOnly: true                    # as text in a dev build, as HTML when publishing

A field’s template is a galley template that makes the field’s HTML from its value; templates write it as html.fields.<field>. It escapes what it writes, as a theme’s template does, and cannot include partials.

A page may write a field by any of its names, in another language or a gendered form: größe: 3, 題名: …, autor:in: Noa, autor*in: Noa and autor·in: Noa are Lima keys as they stand, and a name with a space or another character Lima keys do not hold is quoted, "first name": Noa. A name counts as it is spelled (Autor is not autor), in Unicode’s composed form: a ü typed as u and a combining mark is the ü of config.lima. Another name belongs to one field and is no field’s own name, or the build stops. A page that gives one field twice, by two of its names or in two forms of one, keeps the first and has a problem.

Other lists

packages and editors are lists a theme may show as it likes (site.packages, site.editors); hadley reads nothing in them.