Document the frontend migration options

[playforge-coding]

Stacked on #27. Documentation only.

Renames sveltekit-migration.md → frontend-migration.md and restructures it as a decision record rather than a plan for one framework: what phase 1 established, the SSR boundary (framework-independent), four options with their real costs, and what two spikes actually showed.

The finding that changed the recommendation

I'd claimed Solid's advantage over SvelteKit was that you keep Rsbuild — one package, two entries, no toolchain churn. The spike disproved it.

@rsbuild/plugin-solid 1.2.2 did not transform JSX under @rsbuild/core 2.1.10, in a shared config or a dedicated one. Output was React's JSX runtime both times:

(0,n.jsxs)(g,{onClick:()=>i(r()+1),children:["clicked ",r()]})   // shared config
React.createElement(v,{onClick:()=>t(e()+1)},"clicked ",e())      // solid-only config

rsbuild inspect showed no babel loader in the resolved Rspack config — the plugin is silently a no-op. Scoping pluginReact with exclude did not hand the file over. Peer ranges are compatible, so this looks like a v1-era plugin against the v2 plugin API rather than a hard incompatibility, but I could not resolve it.

vite 8.2.0 + vite-plugin-solid 2.11.14 compiled the identical source correctly — Solid's template/insert output with getter-based props.

Consequence: a Solid app needs Vite today, so it lands on the same two-package, two-build-tool, route-split shape as SvelteKit. oxfmt/oxlint are still preserved, since .jsx stays .jsx — that advantage survives.

Kept from the SvelteKit spike

A +page.server.js importing @spelling-creator/core/users server-rendered real data with no JavaScript executed, and no browser-tier module leaked into the server bundle — so the core/browser/* split holds under a real SSR build, not just under the lint rule.

Also recorded

• The SSR boundary, with the finding that / has nothing data-driven to render (all four fetches auth-gated or DOMParser-based).
• SSR replaces the crawler prerender but not the /og-image screenshots; Satori + @resvg/resvg-wasm is the replacement, at the cost of embedded fonts and a designed card rather than a screenshot.
• The measured Solid conversion surface: 59 files, 189 useState, 107 destructured props, 70 .map(), 44 router sites, and 58 forwardRefs that disappear.
• The failure mode that needs a review rule: destructuring a prop in Solid doesn't error, it silently stops updating.

Verification

pnpm lint, pnpm build:docs pass; the page builds and the overview link is updated. No stale references to the old filename. Lockfile untouched.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 1de0761c-3039-44dc-90d7-9bc5e90fad95

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}