SaaS deployment: DNS records
The canonical hosts live under extractit.dev. Nothing here is created automatically: the
operator adds the records at the DNS provider. Chart values for the EU SaaS are in
deploy/helm/extractit/values-saas-eu.yaml.
| Host | Serves | Record |
|---|---|---|
docs.extractit.dev |
The documentation site (make docs output, site_url in mkdocs.yml) |
CNAME to the static hosting target of the site build |
schemas.extractit.dev |
The static files under schemas/ (JSON Schema $id targets, same paths) |
CNAME to the static hosting target; serve schemas/ as the root |
app.extractit.dev |
Studio and API on the same host (publicUrl and ingress host in the chart) |
A/AAAA or CNAME to the cluster ingress; TLS certificate for this host |
errors.extractit.dev |
Error type URIs (https://errors.extractit.dev/<code>) |
CNAME to the docs host; redirect each path to the matching page in docs.extractit.dev/errors/ |
extractit.dev |
Problem type URIs (https://extractit.dev/problems/<code>) |
A/ALIAS to a redirector; redirect /problems/<code> to docs.extractit.dev/errors/<code> |
Notes:
- Same paths as before; only the host changed. No compatibility aliases are needed because nothing was published under the old hosts.
- Verify with
curl -I https://schemas.extractit.dev/<path>after publishing and check that every$idinschemas/resolves.
Public site on extractit.dev (Cloudflare Pages)
The root domain serves the public site built from site/ (T62: home, pricing, legal pages, in
English at / and Spanish at /es/; the old /en/* URLs redirect to /). It is a static Astro build with no client JavaScript and no
cookies. It also carries the /problems/<code> redirect of the table above, so no separate
redirector is needed.
Not deployed
Nothing here has been deployed and no DNS record has been created. The legal texts are drafts:
publish only after the owner fills every {{…}} placeholder (identity in
site/src/data/identity.json) and marks each legal text of the stage as reviewed in
site/legal.config.json (per stage and per page since T68; see the pre-launch section below).
Build locally
make site-build # draft build: site/dist, legal pages show a draft band
make site-build SITE_ENV=production # fails while reviewed is false or any {{…}} remains
make site-test # unit tests (prices vs plans.yaml, release gate, tokens)
make site-e2e # Playwright: every page in es/en, axe, 360 px, links
Cloudflare Pages project (Workers & Pages → Create → Pages → Connect to Git):
| Setting | Value |
|---|---|
| Production branch | main |
| Root directory | / (the pnpm workspace root, so the lockfile is used) |
| Build command | corepack enable && pnpm install --frozen-lockfile --filter @extractit/site... && SITE_ENV=production pnpm --filter @extractit/site build |
| Build output directory | site/dist |
| Environment variables | NODE_VERSION=22, ASTRO_TELEMETRY_DISABLED=1 |
| Preview deployments | Optional; previews build without SITE_ENV=production, so they show the draft band and send noindex |
site/public/_redirects and site/public/_headers are copied to the output and read by Pages:
/problems/* → https://docs.extractit.dev/errors/:splat (301), plus a restrictive CSP
(default-src 'none'; styles, fonts and images from the site itself only), HSTS and long caching
for /_astro/*.
DNS (zone extractit.dev on Cloudflare):
| Record | Value |
|---|---|
extractit.dev |
Add it as a custom domain of the Pages project; Cloudflare creates a proxied CNAME (flattened at the apex) to <project>.pages.dev |
www.extractit.dev |
Optional: proxied CNAME to <project>.pages.dev, plus a Bulk Redirect www.extractit.dev/* → https://extractit.dev/$1 (301) |
If the zone is not on Cloudflare, use an ALIAS/ANAME record at the apex to <project>.pages.dev
instead (Pages requires the zone on Cloudflare for apex custom domains; otherwise move the
nameservers first).
Check after publishing
curl -sI https://extractit.dev/ | grep -i content-security-policy
curl -sI https://extractit.dev/problems/not-found # 301 → docs.extractit.dev/errors/not-found
curl -s https://extractit.dev/legal/terms/ | grep -c '{{' # 0
The docs build (make docs) writes to build/docs-site/: site/ is this package.
Pre-launch stage and direct upload (T68)
The site has a stage, set in site/launch.config.json and read at build time:
prelaunch (today) |
live |
|
|---|---|---|
| Indexing | <meta name="robots" content="noindex, nofollow"> on every page and X-Robots-Tag: noindex, nofollow on every response (added to _headers at build) |
Indexable once every live legal text is reviewed |
robots.txt |
Allow: /, no Sitemap line |
Allow: / and Sitemap: https://extractit.dev/sitemap.xml |
sitemap.xml |
Not generated | Generated |
| «Start free», «Buy», header button | «Get early access»: mailto: the contact email with a subject (en/es) |
https://app.extractit.dev/signup and the app |
Other links to app.extractit.dev |
None (plain text only); /app redirects to / |
/app redirects to the app |
| Pricing | «Available at launch», no tax, rate or payment notes | Full page with the payments section |
| Legal pages | Legal notice, website-only privacy policy and cookie statement (site/src/content/legal/prelaunch/) |
Terms, privacy, refunds, notice, cookies (site/src/content/legal/<lang>/) |
Crawling is allowed on purpose: a crawler only sees noindex if it may fetch the page, while a
Disallow: / would hide it and leave the URLs indexable (without a snippet) from links elsewhere.
Owner identity. site/src/data/identity.json (legal_name, tax_id, address, email)
is the only source of the legal notice, the privacy policy, the contact page and the
early-access mailto:. The LSSI-CE requires name, tax ID (NIF), address and email.
Release gate. make site-build SITE_ENV=production fails while any generated page contains
{{…}}, while a legal text of the current stage has false under reviewed.<stage> in
site/legal.config.json, or if a legal page of the other stage was generated. Pre-launch needs
reviewed.prelaunch.notice, .privacy and .cookies; the live texts are reviewed separately
under reviewed.live (a website-only privacy policy is not the full one).
SITE_STAGE, SITE_IDENTITY_FILE and SITE_LEGAL_CONFIG override the three files for one build;
they exist for the tests. Never use them for a real deployment.
Cloudflare settings that would contradict the texts. The privacy and cookie texts state that the site sets no cookies, runs no analytics and that its logs are not exported. Keep Cloudflare Web Analytics, Bot Fight Mode and Logpush off for these projects, or update the texts first.
Deploy by direct upload (no Git connection; from the repository root, after wrangler login):
npx wrangler pages project create extractit-site --production-branch main # once
make site-build SITE_ENV=production # fails if not ready
npx wrangler pages deploy site/dist --project-name extractit-site --branch main
Then, in the dashboard: Workers & Pages → extractit-site → Custom domains → Set up a custom
domain → extractit.dev. With the zone on Cloudflare the record is created for you.
Check:
curl -sI https://extractit.dev/ | grep -i x-robots-tag # x-robots-tag: noindex, nofollow
curl -s https://extractit.dev/robots.txt # Allow: /, no Sitemap
curl -sI https://extractit.dev/legal/terms/ | head -1 # 404
curl -s https://extractit.dev/ | grep -c '{{' # 0
Docs on docs.extractit.dev, the same way from build/docs-site (the mkdocs.yml
site_dir):
npx wrangler pages project create extractit-docs --production-branch main # once
make docs # DX_DOCS_SKIP_EXAMPLES=1 skips the examples gallery run (needs Postgres)
npx wrangler pages deploy build/docs-site --project-name extractit-docs --branch main
Then Custom domains → docs.extractit.dev on the extractit-docs project.
From prelaunch to live (when the app is up at app.extractit.dev):
- Set
{"stage": "live"}insite/launch.config.json. - Resolve every
{{DECISION: …}}of the live legal texts (site/src/content/legal/<lang>/) and of the pricing page (site/src/lib/i18n.ts), and set everyreviewed.live.<doc>totrueafter the owner's review. make site-test, thenmake site-build SITE_ENV=productionand deploy as above. The build now generates terms, refunds andsitemap.xml, dropsX-Robots-Tagand links to the app.