The official documentation for Market Data — covering the REST API, SDKs, and Google Sheets Add-On. Built with Docusaurus 3.

Production: www.marketdata.app/docs/ Staging: www-staging.marketdata.app/docs/

pnpm install    # Install dependencies
pnpm start      # Start dev server at localhost:3000
pnpm build      # Production build

This project uses pnpm, pinned by the packageManager field in package.json so CI and every machine provision the same version. Both halves of the origin are now on pnpm: MarketDataApp/website has been since before this repo moved.

Two pnpm behaviours are worth knowing before your first install, both recorded in pnpm-workspace.yaml:

• pnpm refuses to run a dependency's install scripts until you say so, and fails the install rather than skipping them quietly. allowBuilds there is the list of decisions. When pnpm meets a new one it writes a placeholder line for you to fill in; set it to true or false rather than deleting it.
• pnpm does not hoist, so a package you import must be one you declared. That is a feature: it caught @docusaurus/theme-common, which ten swizzled components in src/theme/ had imported for as long as they have existed without it ever appearing in package.json. Under yarn it resolved by accident, at whatever version preset-classic happened to pull in.

The site is hosted on Cloudflare Pages. Both production and staging use the same /docs/ base path — routing is determined by hostname, not path prefix. Deployment is handled by a separate orchestrator repo (MarketDataApp/www-marketdata-app) that merges build artifacts from R2 and deploys to unified Pages projects.

Browser → Cloudflare DNS → Cloudflare Pages → Response

1. DNS resolves the hostname (both are proxied CNAMEs in Cloudflare)
2. Pages serves the file from its docs/ directory (built and nested there by CI)

There used to be a Worker in front of Pages. It was retired on 2026-09-01 (MarketData-App/www-marketdata-app#15) and nothing proxies /docs/* any more. Everything it did is now served by Cloudflare or by the build — including the canonical Link header on Markdown responses, which _headers rules in the orchestrator repo took over (#16). See CLAUDE.md for the full mapping and for why those rules' order is load-bearing.

Deployment is fully automated via GitHub Actions across two repos:

1. This repo (.github/workflows/deploy-docs.yml) — builds Docusaurus, uploads to R2, triggers orchestrator
2. Orchestrator (MarketDataApp/www-marketdata-app) — downloads all sources from R2, merges into unified build, deploys to CF Pages, runs post-deploy tests

Push to staging/main → Build → Upload to R2 → Trigger orchestrator → Deploy to CF Pages → Tests

Workflow:

1. Push to staging — builds and deploys to staging
2. Verify changes at www-staging.marketdata.app/docs/
3. Open a PR from staging → main and merge — deploys to production

# E2E tests (Playwright — Context7 widget, Markdown actions row)
TEST_ENV=staging pnpm run test:e2e

# Script tests (option-symbol checker, Chromium resolver)
pnpm run test:scripts

The e2e suite runs against whatever Chromium the machine already has — /usr/bin/chromium on Linux, Chromium.app/Google Chrome.app on macOS — not against a build this repository pins. scripts/resolve-chromium.js picks it, and playwright.config.js uses that one resolver.

This is deliberate. browserName: 'chromium' is an invisible version pin: it resolves to the single revision bundled with the installed @playwright/test, so the browser under test only moves when someone bumps a devDependency and re-runs playwright install. That would make this repository the gate on every Chromium update. These specs load third-party script into a real page, so they should see the browser our readers run, the day their OS updates it, with no commit here.

• Every run prints the binary it chose (Chromium: /usr/bin/chromium (system)). Read that line first when an e2e test fails.
• Set CHROMIUM_PATH=/path/to/binary to force a specific browser.
• With no system browser found, it falls back to Playwright's bundled build, so a fresh clone still works after npx playwright install chromium.
• CI runs node scripts/resolve-chromium.js || npx playwright install --with-deps chromium, so the runner downloads a browser only when it has none.
• Trade-off: Playwright only guarantees the revision it bundles, so a system browser far ahead of it can drift on CDP behaviour. That is the price of not holding updates back, and it fails loudly rather than silently.

api/              # API reference docs (MDX)
sdk/              # SDK docs — Go, Python, PHP (MDX)
sheets/           # Google Sheets Add-On docs (MDX)
account/          # Account & billing docs (MDX)
src/
  theme/          # Swizzled Docusaurus theme components
  css/            # Custom styles
lib/              # MDX→Markdown and HTML→Markdown converters (the twin converters)
plugins/          # markdown-twins, llms.txt, redirects
scripts/          # Build-time checks — see Testing
e2e/              # Playwright end-to-end tests
.github/workflows # CI/CD pipeline

Search is powered by Algolia DocSearch. The crawler configuration is managed in the Algolia dashboard, not in this repository.

MIT