WordTrove

Static word-finder site: 3–8 letter finders, word unscrambler, daily 5 letter puzzle, ~290 word-list pages, plus an llms.txt / Markdown layer for AI search. No framework, no database.

Run it locally for testing

The site/ folder is a complete prebuilt site, so you can test without building. Serve it over HTTP — opening index.html directly with file:// will not work, because the pages use absolute paths (/style.css, /data/5.txt) and the tools fetch() their word lists.

cd site
python -m http.server 8788

Then open http://localhost:8788/. Any static server works equally well (npx serve site, php -S localhost:8788 -t site).

What to check:

Page	What should happen
/	Type letters into the green boxes — the list narrows as you type, the meter shows the % ruled out, and a "best next guess" line appears
/word-unscrambler/	PLANETS is prefilled; results group by length. Try ? for a blank tile
/daily-word-puzzle/	Six-row board, on-screen keyboard and physical typing both work; reload mid-game and your guesses come back
/5-letter-words-starting-with-a/	Full list with scores; the "open this pattern in the finder" link prefills the first box
/data/5.txt	Plain space-separated word list (served from /data/, cached a week by _headers)

Two notes on the dev server: unknown URLs return the server's own 404 rather than site/404/index.html (only the real host uses that), and _headers is ignored, so caching and Markdown content types can only be verified after deploy.

Rebuild from source

cd source
python3 build.py              # -> source/out/site   (copy over site/ to promote it)
python3 build.py --preview    # -> source/out/preview

Preview mode writes flat .html files with relative links, so it opens straight from the filesystem without a server — handy for a quick look, but it skips the Markdown layer and uses different URLs than production.

Needs Python 3 and the Hunspell en_US dictionary (sudo apt install hunspell-en-us). The build is Linux-only: it reads /usr/share/hunspell/en_US.{aff,dic} by absolute path and uses a glibc-only date format, so it fails on Windows and macOS. Use the prebuilt site/ folder there, or build inside WSL or a container.

Deploy (free)

Upload the site/ folder to Cloudflare Pages or Netlify (drag-and-drop works). _headers sets caching, Markdown content types and canonical headers. 404.html is picked up automatically.

Change name, domain, analytics, ads

Edit CONFIG at the top of source/build.py and rebuild. The Markdown/llms.txt layer is regenerated only when the SEO suite is available (env var SEO_SUITE); otherwise copy site/llms.txt, site/sitemap.md and site/llm/ from this zip and update the domain inside them.

After launch

1. Confirm the domain is yours and not taken; set SITE_URL.
2. Add the site to Google Search Console and Bing Webmaster Tools; submit /sitemap.xml.
3. Set GA4_ID and rebuild.
4. When traffic grows, apply to AdSense or Ezoic; paste their script into ADS_HEAD and their lines into site/ads.txt, and turn on the .ad slots in style.css.

work/ holds the private research and audit notes (keyword research, keyword map, AEO/GEO findings). Never upload it.