Ryan Laird — CV

Source for cv.rjmlaird.co.uk. Astro 7, static output, content fetched at build time from api.rjmlaird.co.uk and validated with Zod.

Identity and structured data

rjmlaird.co.uk owns the canonical Person (https://rjmlaird.co.uk/#person). This site does not define a competing entity. src/lib/structuredData.ts emits one @graph containing:

Node	@id	Notes
WebSite	https://cv.rjmlaird.co.uk/#website	isPartOf the main site's #website; publisher is the canonical Person
ProfilePage	https://cv.rjmlaird.co.uk/#webpage	mainEntity/about → canonical Person; dateCreated, dateModified
Person	https://rjmlaird.co.uk/#person	Slim stub (name, url, jobTitle, image, current worksFor, sameAs)

The stub is deliberate: Google evaluates each page's JSON-LD on its own and does not reliably resolve @ids across domains, so the ProfilePage needs a Person node on the page. Full detail (occupation history, credentials, awards, memberships, languages, knowsAbout) belongs on the main site, built from the API's /v1/schema/person endpoint.

Rules that are enforced by pnpm check:dist:

• no email address in JSON-LD (it stays visible on the page)
• worksFor is the current employer only
• sameAs is https-only, deduped, and excludes messaging links (wa.me, signal.me, t.me)
• the CV is self-canonical; it is not canonicalised to the main site

All identity constants live in src/config/site.ts. Change them there, nowhere else. CV_FIRST_PUBLISHED should be set to the real publish date; until it is, dateCreated falls back to the build date.

Structure

src/
├── config/site.ts          # Origins, canonical @ids, current employer, contact email
├── data/
│   ├── nav.ts              # Section nav links
│   └── skills.yaml         # Skills taxonomy (see "Skills" below)
├── lib/
│   ├── api.ts              # Fetch + validate every collection; memoised; fail-fast
│   ├── cdn.ts              # Resolves CDN document ids to URLs
│   ├── sameAs.ts           # Normalise / filter / dedupe profile URLs
│   ├── skills.ts           # Taxonomy loader + tag aggregation
│   ├── structuredData.ts   # JSON-LD graph (see above)
│   ├── format.ts           # Date/text helpers
│   └── schemas/            # One Zod schema per API collection
├── layouts/Layout.astro    # <head>, meta, JSON-LD, fonts, global CSS
├── components/             # One component per CV section, plus Nav and AtsView
├── pages/
│   ├── index.astro         # The visual CV
│   └── ats.astro           # Plain-text CV for ATS parsers (noindex)
└── styles/global.css       # Design tokens and styles
scripts/check-dist.mjs      # Post-build checks

Fonts (Space Grotesk, Inter, JetBrains Mono) are self-hosted via @fontsource-variable/*. There is no request to Google Fonts.

Content and the API

Almost everything comes from api.rjmlaird.co.uk at build time (src/lib/api.ts). Each endpoint is requested once per build and shared by every component, so sections can't disagree.

The build fails if profile, experience or education can't be fetched or don't validate. A blank CV is never deployed. Other collections degrade to empty sections with a logged warning.

Variable	Effect
ALLOW_PARTIAL=1	Let the build continue when critical data is missing (local work only)
CV_API_BASE	Override the API base URL, e.g. http://127.0.0.1:4599/api to build against a mock

Skills

The Skills section is built from the skills tags on experience and education entries (aggregateSkills), not curated by hand. src/data/skills.yaml supplies categories and canonical names only; it never adds a skill on its own. Tags that match nothing in the taxonomy still appear, under "Other".

skills.yaml accepts either a map of category_key: [skill names] (current) or an array of { name, category, description, proficiency, relatedSkills } entries. Any other shape fails the build. New category keys need a label in CATEGORY_LABELS in src/lib/skills.ts; unknown keys fall back to a Title Case of the key.

ATS / print

• /ats/ is a plain, single-column version for applicant tracking systems and clean PDFs. It is noindex, canonicalised to /, and excluded from the sitemap.
• "Download / Print" on the main page prints the visual layout.

Commands

Command	Action
pnpm install	Install dependencies (pnpm only; version pinned in packageManager)
pnpm dev	Dev server at localhost:4321
pnpm build	astro check then production build to ./dist/
pnpm preview	Preview the production build
pnpm typecheck	Type-check only
pnpm lint	ESLint (eslint.config.mjs)
pnpm format	Prettier (writes; run as its own commit)
pnpm check:dist	Post-build checks on ./dist
pnpm verify	Typecheck, lint, build, then dist checks
pnpm check:live	After deploy: check CV and main site agree on the Person (needs network)

Node version is in .nvmrc (24). TypeScript is pinned to 5.9 because astro check doesn't work with TypeScript 7 yet.

Deploying

.github/workflows/deploy-cv.yml builds and deploys to GitHub Pages on:

• push to master
• repository_dispatch with type content-updated. Fire it from the API repo when CV data changes: gh api repos/rjmlaird/my-cv/dispatches -f event_type=content-updated
• a nightly schedule (04:17 UTC) as a backstop
• manual workflow_dispatch

CI runs lint, build and check:dist; a failure stops the deploy. Update SITE_URL in src/config/site.ts and site in astro.config.mjs if the domain ever changes.

After deploying

See POST-DEPLOY-CHECKLIST.md for the live checks (Rich Results Test, Search Console, pnpm check:live).

SEO notes

• Canonical URL, Open Graph and Twitter meta are generated from the page title and description props.
• <meta name="keywords"> is intentionally absent (search engines ignore it).
• The sitemap deliberately has no lastmod: nightly rebuilds would bump it daily with no real content change.