feat(web): keyword research + on-page SEO rewrites (SQLR-33)

[joaoh82]

Summary

Lands the SQLR-33 keyword research + on-page SEO pass for sqlritedb.com. Content/metadata only — no structural redesigns, no new routes, no dep changes.

• New web/seo/keywords.md — head / mid-tail / long-tail keyword tables with target page, intent, priority, and a per-page primary+secondary registry for the P0 pages.
• web/src/app/layout.tsx — site-wide default title + description + keywords array rewritten to lead with "embedded SQL + vector database in Rust".
• web/src/app/page.tsx — landing metadata rewritten; page-level title now explicitly contains the primary head term + product name.
• web/src/components/hero.tsx — H1: "SQLRite — an embedded SQL + vector database in Rust." Lede leads with "SQLite alternative" / "embedded database in Rust" and names HNSW + BM25 inside the 160-char meta-snippet zone.
• web/src/app/docs/page.tsx — H1: "SQLRite docs — getting started with the embedded Rust database." Lede answers "what is SQLRite + how do I use it" in plain language and adds 4 descriptive deep-links into key sections. Docs CTA card now exposes /blog as a second descriptive in-site sibling alongside /.
• web/src/components/benchmarks.tsx — H2: "SQLRite vs SQLite benchmarks — honest numbers, published in public." Sub-copy names HNSW + BM25 as the differentiator workloads.
• web/README.md — points future rewrites at web/seo/keywords.md.

Why

The site has been a beautiful brochure — full metadata surface (OG, JSON-LD, sitemap, RSS, OG images) but no keyword targeting at the wording level. Without explicit terms in the H1 + lede, Google can't surface us for "embedded database in Rust", "SQLite alternative", or "MCP server for SQLite" even though we're a credible answer for all three.

The keyword sheet is the registry — every P0 page now records its primary + secondary keyword so future rewrites stay coordinated rather than drifting.

Internal-linking sweep

Grep confirmed zero non-descriptive anchors in web/src — no "click here" / "read more" / bare "here". The "every doc page links to ≥ 2 sibling pages with descriptive anchors" rule is met via the global <Footer /> (links to /blog, /blog/rss.xml, GitHub) plus the docs CTA card, which now adds an explicit /blog link as a second descriptive in-site sibling.

Verified locally

npm run typecheck   # ✅
npm run lint        # ✅ No ESLint warnings or errors
npm run build       # ✅ 49 static pages generated, no orphans

Test plan

• CI web lint + build steps pass.
• Vercel preview deploy renders the new H1s + lede on /, /docs, and the landing benchmarks section.
• View-source on each P0 route shows the new <title> + <meta name="description">.
• After production promote: optional Screaming Frog re-crawl of https://sqlritedb.com (free up to 500 URLs) to confirm no duplicate H1/title/meta and no orphan pages.

Follow-ups (filed separately)

Long-tail keywords from the sheet that are blog topics rather than rewrites — split into their own tickets:

• SQLR-59 — Blog post: Adding HNSW vector search to a SQLite-style engine.
• SQLR-60 — Blog post: Wiring a SQLite-style engine into Claude Code via MCP.
• SQLR-61 — Blog post: Shipping a Tauri 2 app with an embedded SQL database.
• SQLR-62 — Blog post: Running a SQL engine in a browser tab with WASM.
• SQLR-63 — Docs: dedicated .ask / natural-language-to-SQL section in /docs.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
rust-sqlite	Ready	Preview, Comment	May 12, 2026 3:59pm