pnwinsects-landing

The landing page for pnwinsects.org — the front door to the PNW Insects family of sites. It introduces the project and links out to each group site as it comes online.

The catalogue itself lives elsewhere: pnwinsects/pnwmoths builds moths.pnwinsects.org. This repo is deliberately small — an Eleventy build over a handful of templates, with the same deploy machinery as pnwmoths so the two are diffable, but none of its species data pipeline.

Layout

Path	Contents
src/index.njk	The page itself — banner, Focus, Resources
src/_includes/base.njk	HTML shell, partner banner, footer
src/_data/resources.ts	The Resources cards. Adding a site is an edit here.
src/_data/partners.ts	Institutional partners and funders, hand-synced with pnwmoths
public/	Copied verbatim to the site root: /images, /styles
scripts/upload-site.ts	Additive uploader to bunny.net Storage
scripts/deploy-smoke.ts	Post-deploy freshness check against the live origin

Adding a resource card

Append an entry to resources in src/_data/resources.ts and drop its image in public/images/. The grid in src/index.njk loops the array, so no template change is needed. Cards render in array order, and the layout is designed to look right with one card or with a dozen.

Running locally

nvm use          # Node 24, see .nvmrc
npm install
npm run dev      # Eleventy dev server with live reload
npm run build    # build + blocking link check, output in _site/
npm test         # node --test, built-ins only
npm run typecheck

No lockfile yet

package-lock.json is not committed. The machine this repo was scaffolded on cannot reach registry.npmjs.org — the TLS handshake is refused for npm and curl alike — so a lockfile could not be generated. Until one is committed, CI runs npm install rather than npm ci, and actions/setup-node's npm cache is disabled (it derives its key from a lockfile and hard-fails without one).

Fixing this is a three-line change from any machine with working registry access: run npm install, commit the resulting package-lock.json, then restore npm ci and cache: 'npm' in both workflows.

Hosting

Static files only — no server, no database, matching the pnwmoths architecture.

• Storage zone: pnwinsects on bunny.net (la.storage.bunnycdn.com)
• Pull zone: https://pnwinsects.b-cdn.net/
• Public domain: pnwinsects.org and www.pnwinsects.org, pointed at the pull zone by CNAME. Both hostnames also need to be added as Custom Hostnames on the pull zone in the bunny.net dashboard, with a certificate issued, before HTTPS will work. Not live yet — see "Domain cutover" below.

Deploying

Push to main. .github/workflows/production.yml typechecks, tests, builds, uploads to Bunny Storage, and then smoke-checks the CDN.

The upload is additive: scripts/upload-site.ts PUTs new and changed files and never deletes. It keeps a content-hash manifest at _site-manifest.json in the zone root and reads it back from the Storage API rather than the CDN, so a stale edge cache cannot cause a file to be skipped. FORCE_FULL=1 re-uploads everything; DRY_RUN=1 prints the plan and makes no network calls.

Deletion is therefore manual, via the bunny.net dashboard. That is the same trade the moths site makes — an uploader that cannot delete cannot destroy the site through a build bug.

Required pull-zone cache configuration

The pull zone must serve HTML with Cache-Control: no-cache. Deploys are additive and never purge, so cache correctness comes from headers alone: index.html changes in place on every deploy, and a long TTL would pin a month-old copy of the site at the edge. Static assets should keep their long TTL — they are safe to cache.

This is dashboard configuration, not something the repo can enforce, so npm run deploy:smoke asserts it after every deploy and fails the workflow if it regresses.

Two settings under Pull Zone → Caching → General produce the correct behaviour, and they are the whole story — no edge rule is involved:

Setting	Required value
Smart Cache	ON
Cache expiration time	Respect origin Cache-Control
Browser cache expiration time	Match server cache expiration

The origin (Bunny Storage) sends no Cache-Control of its own — scripts/upload-site.ts sets none. Smart Cache is what classifies responses by extension and MIME type: HTML is non-cacheable and gets no-cache, while CSS/JS/images are cacheable and get Bunny's max-age=25600000. "Respect origin Cache-Control" keeps Bunny from overriding that classification with a fixed TTL.

A zone created with Bunny's defaults has Smart Cache off and Cache expiration set to "Override: 1 month", which is exactly where a public, max-age=2592000 on HTML comes from. After changing these, purge the zone — the previously cached copy is served with its original header until it is evicted.

The moths pull zone additionally carries one edge rule forcing no-cache on *.csv, because it serves mutable CSV data that Smart Cache would otherwise treat as a cacheable asset. This site serves no CSVs and deliberately omits that rule. See ADR 0009 in pnwmoths for the reasoning and the full settings.

Domain cutover

Until pnwinsects.org is repointed, it resolves to the registrar's parking page, so the smoke check targets https://pnwinsects.b-cdn.net — the hostname the site is really served from. The canonical URLs in the HTML already say pnwinsects.org; that is intentional.

To finish the cutover: add both hostnames as Custom Hostnames on the pull zone and issue certificates, have the CNAMEs pointed at pnwinsects.b-cdn.net, then change DEFAULT_CDN_ORIGIN in scripts/deploy-smoke.ts to https://pnwinsects.org so the check starts guarding the domain visitors actually use.

Required secret

BUNNY_STORAGE_PASSWORD — the storage zone's password, set as a repository secret and scoped to the production environment. Find it under Storage → pnwinsects → FTP & API Access in the bunny.net dashboard.