chore(deps): vite 4 → 8 + Rolldown bundler

[dnplkndll]

Summary

Step-wise upgrade of Vite (4 → 5 → 6 → 7 → 8) across the four Vite-using packages (replicad, replicad-evaluator, studio, replicad-app-example). Vite 8 ships with Rolldown as the default bundler — replacing both esbuild (dev transform) and Rollup (production) with a single Rust toolchain. Plugin API stays Rollup-compatible; vite.config files needed no behavior changes across all four majors.

Sending upstream because I've been running this against replicad-based work in another project for a few weeks, and the perf win plus the dev-mode dual-React fix (described below) both apply to your tree exactly as-is. Feel free to take it, split it, or pass.

Commits

Each Vite major lives in its own bisectable commit:

• vite 4 → 5 — also bumps vite-plugin-dts ^3 → ^4, vite-plugin-pwa ^0.16 → ^0.21, @vitejs/plugin-react 3.0.1 → ^4.3.4, vitest ^0.28 → ^1.6
• vite 5 → 6
• vite 6 → 7 — @vitejs/plugin-react ^4 → ^5, vite-plugin-pwa ^0.21 → ^1.0 (which requires workbox-build/workbox-window ^7.4.1 as new peers), vitest ^1 → ^2; .nvmrc 20.12 → 20.19 for the Vite 7 Node floor
• vite 7 → 8 (Rolldown) — @vitejs/plugin-react ^5 → ^6

Plus two follow-up commits:

• chore: restore v-prefix in .nvmrc — cosmetic
• fix(studio): dedupe react under Vite 8 to avoid dual-instance hook errors — see below

The React dedupe fix

In headless-browser smoke-testing the dev server, I hit Invalid hook call. You might have more than one copy of React in the same app when mounting the Workbench tree. Root cause: three React versions live in the workspace — replicad-app-example pins 18.2.0 exact, studio uses ^18.3.1, and replicad-docs pulls 19.x via docusaurus 3.x. Under Vite 8's dep optimizer, studio's pre-bundled chunks for mobx-react and @devbookhq/splitter were resolving react against a different instance than the app's.

Standard fix: resolve.dedupe: ["react", "react-dom"] in packages/studio/vite.config.js. Prod build is unaffected (Rolldown collapses duplicates either way); the fix is dev-mode-load-time. Worth taking even if you choose not to bump Vite, since the same conditions exist on Vite 4 — the optimizer just happens not to trip the resolver the same way today.

Build perf (local, identical machine)

Stage	Before (Vite 4)	After (Vite 8 + Rolldown)	Δ
pnpm prepare-packages wall	7.05 s	5.40 s	1.3× faster
pnpm --filter studio build wall	19.61 s	3.00 s	6.5× faster
pnpm --filter replicad-app-example build wall	3.06 s	1.12 s	2.7× faster
studio index-*.js (gzip)	559 KB	524 KB	-6%
replicad.js (lib, raw)	593 KB	588 KB	-1%

Studio's headline 6.5× win is the most visible: production-mode Workbench/index bundles plus PWA manifest generation, with code-splitting that Rolldown produces more aggressively than Rollup did (Monaco workers + Workbench load lazily as separate chunks). For your publish-studio.yml Cloudflare Pages workflow this is a direct CI-time saving.

Surfaces that did NOT need changes

• OCCT WASM — replicad-opencascadejs and replicad_single.wasm load unchanged.
• Module worker in studio (worker.format: "es") carries across.
• Monaco editor — self-hosts its workers; no library-side change.
• @rollup/browser + rollup-plugin-external-globals as runtime deps of replicad-evaluator — orthogonal to the build-time bundler.
• vite-plugin-dts with rollupTypes: true — uses API Extractor independently of Vite/Rolldown.
• @rollup/plugin-commonjs, -node-resolve, -json, -terser in packages/replicad/package.json — Rolldown's plugin API stays Rollup-compatible.

The chunk-size advisory now reads build.rolldownOptions instead of rollupOptions — informational, expected.

Rolldown surfaces two new informational warnings during the studio build:

• Direct eval use in Monaco editor — known upstream Monaco behavior.
• BatchedMesh import missing from three@0.155.0 via three-mesh-bvh — pre-existing; resolves on a three.js bump.

Verification

• pnpm install clean across all four major bumps.
• pnpm prepare-packages clean at every step.
• pnpm --filter studio build clean; service worker generates.
• Vitest suites: 49/49 pass (39 replicad + 4 replicad-evaluator + 6 replicad-cli) under vitest 2.x.
• Dev-server browser smoke in headless Chromium against http://localhost:5555/workbench: studio renders the Workbench UI byte-identical to a main baseline screenshot. The screenshots in the downstream PR (linked for reference) show the before/with/without-dedupe states.

A note on CI

Your publish-studio.yml currently uses pnpm/action-setup@v2 version: 8 with the pnpm@9.13.2 packageManager and a --no-frozen-lockfile flag, which means CI regenerates the lockfile every run. Not breaking, just wasteful. Worth bumping to pnpm/action-setup@v4 + version: 9 and dropping --no-frozen-lockfile. Happy to do that in a follow-up if you prefer it separate.

Out of scope

• packages/replicad-threejs-helper — uses standalone rollup@3.10.1 (not Vite). Independent migration; deferred.
• Vite Environment API (introduced in v6) — replicad doesn't do SSR; skip.
• rolldownOptions.output.advancedChunks tuning — defaults are fine.

[dnplkndll]

Re: CI run 25765302805 — the build steps all passed (pnpm install, prepare-packages, studio + app-example builds). The failure is on the cloudflare/pages-action@v1 step with Input required and not supplied: apiToken — secrets are not exposed to fork-PR runs by GitHub Actions, so this is expected and unrelated to the upgrade. The publish step would succeed once it runs in your repo's context (e.g. on push to main after merge).

One small heads-up I should have included in the PR body: Rolldown surfaces three new info-level warnings that Rollup hid — Module "path" / "fs" / "crypto" has been externalized for browser compatibility, imported by replicad-opencascadejs/src/replicad_single.js. That's the Emscripten output of opencascade.js pulling node builtins in a path the browser bundle tree-shakes away. Harmless, just noisier than before.

[sgenoud]

Thanks! This looks great, I need to take a bit of time to test it locally and then I'll merge!