perf(landing): defer Features preview, memoize integration grid, trim dead weight

[waleedlatif1]

Summary

Landing performance wins from a verified audit (each finding adversarially confirmed before implementing). No behavior, SEO, or render-mode change — all confirmed empirically.

Defer the Features workflow preview (biggest win)

landing-preview-mount.tsx is code-split (ssr:false) but was mounted eagerly, so its reactflow + framer-motion chunk (~85–100 KB gz) downloaded right after hydration despite sitting below the fold. It's now gated on an IntersectionObserver (400px preload margin, loads once), reusing the existing dimension-stable aspect-[1116/615] placeholder → zero CLS. SEO-neutral (the preview was already ssr:false, so SSR only ever emitted the placeholder).

Memoize the integration grid

integration-grid.tsx rebuilt category facets and re-lowercased ~3,480 operation/trigger strings on every keystroke, and re-rendered all ~220 rows. Now: facets + a per-integration lowercased search index are derived once (useMemo over the stable list), and IntegrationRow is memo'd (both props are stable references). Match semantics are identical (per-field includes, not a joined string — verified: ?q=slack returns the same 6 results).

Hero resize listener

Subscribes once via a phaseRef instead of re-subscribing on every animation phase (~30×/loop).

Delete dead weight

Orphaned public/static/mothership.gif (1.5 MB) + public/landing/sim-mothership.webp (87 KB), the unused static Season font weights (only the loaded variable .woff2 is kept), and the unused CtaChat component — all re-verified unreferenced across the repo.

Not included (deliberately)

• Hero per-character re-render isolation + off-screen animation gating — real, but the fix touches a 1,191-line animation component (synchronous card-height coupling; restart-from-home gating) and needs a visual-QA pass; deferred to its own change rather than risk an animation regression.
• Blog next/image unoptimized — left as-is.

Testing

• tsc --noEmit 0 errors; biome check clean
• next build: compiled successfully, 712/712 pages, 0 errors; font deletions safe (no next/font errors); render modes unchanged
• Runtime: /integrations?q=slack returns the identical filtered set (223 → 6); /workflows (a gated-preview consumer) → 200
• Every change adversarially reviewed (effect/observer lifecycle, filter equivalence, memo prop stability, deletion safety) — no issues

Type of Change

• Performance / improvement (no behavior change)

[vercel]

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

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Skipped		Jul 1, 2026 12:10am

[cursor]

PR Summary

Low Risk
Performance-only landing changes with preserved filter semantics and CLS-safe preview gating; main risk is edge cases around IntersectionObserver lifecycle or memoized row props, not security or data paths.

Overview
Landing load and interaction performance without intended behavior changes: below-the-fold workflow previews no longer pull the heavy reactflow + framer-motion chunk on first paint, the integrations search UI does less work per keystroke, and the hero resize handler stops re-binding on every animation phase.

Features preview (landing-preview-mount.tsx): the dynamic preview still uses ssr: false and the same aspect-[1116/615] placeholder, but the chunk now loads only after an IntersectionObserver sees the mount (400px preload margin, one-way flip to inView). Without IntersectionObserver, it loads eagerly so the UI does not stay on a blank placeholder.

Integrations grid: category facets and a per-integration lowercased search index are built once via useMemo; filtering uses that index with the same per-field includes semantics as before. IntegrationRow is wrapped in memo to cut unnecessary row re-renders while typing.

Hero visual: window resize uses a phaseRef so the listener is registered once and still snaps the cursor for the current phase.

Cleanup: removes the unused CtaChat component (no remaining references in the repo per this change set).

^{Reviewed by Cursor Bugbot for commit 3bfabc2. Configure here.}

[greptile-apps]

Greptile Summary

This PR improves landing-page performance without changing the visible page behavior. The main changes are:

• Defers the heavy landing preview until it nears the viewport.
• Adds a fallback for browsers without IntersectionObserver.
• Memoizes integration filtering and row rendering.
• Keeps the hero resize listener subscribed once.
• Removes unused landing assets, fonts, and the unused CTA chat component.

Confidence Score: 5/5

This looks safe to merge.

• No blocking issues found in the changed code.

Important Files Changed

Filename	Overview
apps/sim/app/(landing)/components/landing-preview/landing-preview-mount.tsx	Adds viewport-gated loading for the landing preview and eagerly loads it when IntersectionObserver is unavailable.
apps/sim/app/(landing)/integrations/components/integration-grid.tsx	Memoizes category facets, search indexing, and filtered integration results.
apps/sim/app/(landing)/integrations/components/integration-card.tsx	Memoizes the integration row component while keeping the same props and named export.
apps/sim/app/(landing)/components/hero/components/hero-visual/hero-visual.tsx	Uses a ref so the resize handler reads the current animation phase without resubscribing.

_{Reviews (2): Last reviewed commit: "perf(landing): load preview eagerly when..." | Re-trigger Greptile}

[waleedlatif1]

@greptile review

[waleedlatif1]

@cursor review

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 3bfabc2. Configure here.}