Prerender SPA at build time so LLM/search crawlers see the full page

[ampedraszewska]

Why

Builds on PR #1 (meta + JSON-LD + llms.txt). PR #1 gave LLM and search crawlers rich metadata in the initial HTML response — but the actual page content (team, thesis, portfolio news ticker) still lived behind the React bundle. Crawlers that don't reliably run JS (GPTBot, ClaudeBot, PerplexityBot, and the indexing pass for several others) only saw <div id="root"></div> for the body.

This PR finishes the fix by prerendering the page at build time: after vite build, a Puppeteer headless browser loads the built SPA, captures the fully-rendered DOM, and writes it back to dist/index.html (and dist/404.html for the GitHub Pages SPA fallback).

What that gives us

LLM crawlers and search engines now receive the complete Hero component as static HTML:

• The full headline and sub-thesis ("early-stage fund backing Poland's boldest tech founders…")
• All three partners with names, roles, and LinkedIn links (Aleksandra Pedraszewska, Karolina Kukielka, Zuzanna Brzosko)
• The news ticker carrying the Replenit pre-seed announcement
• Locations, contact, partner logos

The React bundle still hydrates and runs client-side for interactive features (mouse-tracking effect, mobile menu).

What changed

• scripts/prerender.mjs (new): post-build script that uses Vite's preview() API and Puppeteer
• package.json:
  • build is now vite build && node scripts/prerender.mjs
  • build:nossg added as an escape hatch (just vite build, in case the prerender step ever needs to be skipped)
  • puppeteer added to devDependencies

No changes to React components, routing, or build configuration beyond the script orchestration.

Local test

$ npm run build
> vite build && node scripts/prerender.mjs

vite v5.4.10 building for production...
✓ 1665 modules transformed.
dist/index.html                   4.96 kB │ gzip:  1.47 kB
✓ built in 2.10s
Created 404.html for GitHub Pages SPA fallback
[prerender] starting Vite preview server on port 4173
[prerender] rendering http://127.0.0.1:4173/
[prerender] wrote /private/tmp/vastpoint-work/vastpoint-test/dist/index.html (25730 bytes)
[prerender] wrote dist/404.html (mirror of index.html)
[prerender] done

$ grep -oE "Aleksandra Pedraszewska|Karolina Kukielka|Zuzanna Brzosko|Replenit" dist/index.html | sort -u
Aleksandra Pedraszewska
Karolina Kukielka
Replenit
Zuzanna Brzosko

Verify after merge

GitHub Actions runs the same npm run build. About 30-60 sec extra build time for Chromium download (cached on subsequent runs).

After deploy:

curl -A "GPTBot/1.0" https://vastpoint.vc/ | grep -E "Pedraszewska|Replenit"

Should print both. Currently (after PR #1 only) prints neither.

What is not in this PR

• No content changes
• No design changes
• No routing changes
• The CI build will be slightly slower because Puppeteer downloads Chromium. We can tighten this later with puppeteer-core plus setup-chrome, but the simple path works first.