Verbseoul — Korean verb conjugator

A free, fast Korean verb conjugator built as a programmatic-SEO site. Every verb gets its own statically-generated page showing the full conjugation table — present / past / future across casual, polite, and formal speech levels — plus extra forms (command, suggestion, question, connectives) and romanization.

The strategy: each page is a genuinely useful interactive reference, not thin content. That's the kind of page AI Overviews summarize but can't replace, and it earns links because people bookmark and share a tool that works.

---

Quick start

npm install
npm run dev        # regenerates conjugations, then starts the dev server
# open http://localhost:3000

Build for production:

npm run build      # runs the generator (prebuild), then next build
npm start

Deploy to Vercel

1. Push this folder to a GitHub repo.
2. In Vercel, "New Project" → import the repo. Framework auto-detects as Next.js.
3. No environment variables are required. Deploy.

The prebuild script regenerates data/conjugations.json on every Vercel build, so adding verbs and pushing is all it takes to publish new pages.

Before going live, set your real domain in lib/site.js (the url field is used for canonical tags, Open Graph, the sitemap, and robots).

---

How it works (architecture)

data/verbs.js              seed list — the single source of truth for what pages exist
scripts/generate.js        plain-Node build step: runs the engine over every verb,
                           writes data/conjugations.json
lib/korean/engine.js       wrapper: curates forms + fixes the polite-present bug
lib/korean/bravender/*     vendored conjugation algorithm (AGPL — see Licensing)
data/conjugations.json     GENERATED. precomputed output the pages import
app/conjugate/[verb]/      the page template, statically rendered per verb (SSG)
app/page.js                homepage + searchable verb index
app/sitemap.js, robots.js  auto-generated from the verb list

Why precompute into JSON instead of calling the engine inside the app? The conjugation library is written in old-style ("sloppy mode") JavaScript that a modern bundler can't compile directly. Running it once in plain Node at build time sidesteps that entirely and keeps the algorithm out of the shipped bundle — smaller, faster, and the pages just read static data.

The polite-present fix

The JavaScript port of the conjugator corrupts the stem of ㄷ-irregular verbs in the polite present form (e.g. 듣다 → 㭤어요 instead of 들어요). Because the polite present is always "casual present + 요" for every verb class, lib/korean/engine.js rebuilds those three forms from the verified casual form. The generator fails the build if the corruption marker ever appears, so a bad form can't ship silently.

---

Adding verbs (this is the growth lever)

Open data/verbs.js and add an entry:

{ hangul: '읽다', slug: 'ikda', en: 'to read', type: 'regular' },

• hangul — dictionary form, must end in 다.
• slug — unique, ASCII, used in the URL (/conjugate/ikda). Revised romanization is the convention here; just keep it unique.
• en — concise meaning. This drives "how to say X in Korean" search intent.
• type — grammatical class label (display only; conjugation is automatic).

Then npm run generate (or just build/deploy). The new page, its sitemap entry, and its internal links appear automatically.

To scale to thousands of verbs, generate data/verbs.js from a source like the Wiktionary Korean verb category plus a TOPIK frequency list, prioritising the words people actually search. Keep the file as plain data — everything downstream is automatic.

---

Suggested next builds

1. A Korean name generator (English → Korean name + meaning). Pure link-bait; the shareable top-of-funnel that earns the backlinks an SEO tool needs.
2. A hand-written grammar hub (speech levels, irregular verbs explained). This is the E-E-A-T / human-expert layer — the right home for authentic voice.
3. Audio — pre-generate TTS clips per form so each page has native pronunciation. Store as static files and reference by form.
4. Monetisation once traffic exists: textbook/app affiliate links, display ads, and a small premium tier (saved verb lists, spaced-repetition practice).

---

Licensing — read before commercial launch

The conjugation algorithm in lib/korean/bravender/ is © Dan Bravender, licensed under AGPL-3.0 (see LICENSE-AGPL.txt and COPYRIGHT in that folder). The two files modified for strict-mode/bundler compatibility are marked inline.

AGPL-3.0 includes a network-use clause (section 13): if users interact with this software over a network (i.e. a public website), you must offer them the complete corresponding source code of the AGPL-covered portion, including any modifications.

Practical compliance, kept simple:

• Keep the conjugator isolated in lib/korean/ (it already is).
• Add a visible link in your site footer to a public repo containing that source and your modifications, e.g. "Conjugation engine source".
• Retain the copyright and license files as-is.

This does not force you to open-source your whole site — only the AGPL'd module and changes to it. If you'd prefer no copyleft obligation at all, replace lib/korean/bravender/ with a clean-room conjugator and update engine.js accordingly (more work, and conjugation correctness is the hard part).

Your own code in this project (everything outside lib/korean/bravender/) is yours to license as you wish.