TickerLog docs

The source of docs.tickerlog.io, built with VitePress.

npm install
npm run dev      # http://localhost:5199
npm run build    # docs/.vitepress/dist
npm run preview

Layout

Everything the site is made of lives in docs/:

Path	What it is
docs/**/*.md	The pages. Plain Markdown, edited directly.
docs/index.md	Landing page.
docs/.vitepress/config.mjs	Navigation, sidebar, navbar, footer, head tags.
docs/.vitepress/theme/style.css	The palette, fonts and layout overrides.
docs/.vitepress/theme/components/	Card and CardGroup, usable in any page.
docs/public/	Favicon and logos, copied to the site root as-is. favicon.svg is the
shared artwork, byte for byte the same file the app, the marketing site and the admin panel
serve; logo/{light,dark}.svg nest that same file as the navbar lockup's mark and differ only
in their wordmark and badge colours. The wordmark is Figtree 900 converted to outlines, so
changing the words or the weight means regenerating the paths.

Adding a page means creating the .md file and adding it to the sidebar in config.mjs; nothing else generates or rewrites the content.

Give every page a title in the frontmatter and an # h1 in the body — VitePress does not print the frontmatter title on the page, so without the heading the page opens on its first paragraph.

Card and CardGroup are Vue components registered globally, so they work inside Markdown without an import. Card takes title, href, horizontal and an optional icon — five icons are drawn (rocket, upload, chart-line, gear, book-open), and an unrecognised name renders no icon rather than failing the build.

Worth knowing

• The font comes from Google Fonts — Figtree alone, at 400/500/600/700/800/900: h1/h2 are 900 and h3/h4 are 800, matching the app and the marketing site, so the request has to carry 900 or the browser synthesises it. Inter was the body face and is gone. It is a third-party request on a hostname the privacy policy currently says makes none. Either self-host Figtree in docs/public/ or check the policy's wording. Still open.
• Search is a local index built at compile time — no service, no network call, no key.
• No telemetry. Nothing writes to localStorage except VitePress's own colour-scheme preference, which is necessary rather than analytical.
• No "copy page as Markdown" or LLM hand-off menu. If it is wanted, a llms.txt or a per-page copy button has to be built.

Deployment — Cloudflare Workers

Pushing to main deploys. .github/workflows/deploy.yml builds and uploads on every push, and can be re-run by hand from the Actions tab. Deploying locally is the fallback, for when the site has to go out without a commit:

npm run deploy   # builds, then uploads docs/.vitepress/dist

The workflow needs one repository secret, CLOUDFLARE_API_TOKEN — created at Cloudflare Dashboard → My Profile → API Tokens → Create Token, from the Edit Cloudflare Workers template, scoped to the TickerLog account. Nothing else is secret: the account id is in the workflow file, and there are no build time variables. Rotating the token is a Dashboard change plus a secret update, with no code change.

wrangler.toml is the source of truth, including the docs.tickerlog.io custom domain — a route set in the Dashboard alone is dropped by the next deploy. There is no main and no Worker script: an assets-only Worker is served straight from Cloudflare's asset storage, so no request is billed as an invocation.

Two settings there are load-bearing rather than taste:

• html_handling = "drop-trailing-slash" matches cleanUrls: true in config.mjs. The build writes flat start/first-trade.html files and links to them without the extension, so /start/first-trade must serve the file and /start/first-trade/ must redirect to it. Cloudflare's default serves both, which is two URLs for one page.
• not_found_handling = "404-page" serves the themed 404.html. Without it an unknown path gets Cloudflare's generic 404, with no nav and no way back in.

Nothing runs server-side, so a rollback is a redeploy of the previous build. Note the free plan permits this: Cloudflare has no non-commercial restriction on static assets, where Vercel's Hobby plan does and would rule out a paid product's documentation.

---

The docs rule from the main repo still applies: what the docs claim is what the app does, and no page walks through connecting a broker until sync ships.