Reduce API reference prerender memory

[bryantgillespie]

Changes

• Generates API reference nav/meta plus one derived payload per OpenAPI tag before build/dev, keeping generated files out of git.
• Switches the API layout to generated nav/meta only; each tag page loads only its generated payload via import.meta.glob.
• Keeps client hydration while moving ref resolution, schema flattening, response example generation, and route list creation out of page render.
• Keeps the lighter inline markdown renderer for plain API descriptions.
• Deletes the now-orphaned runtime OpenAPI walkers/types/constants so the generator is the single source for API payload shaping.
• Fails loudly if generated API prerender routes are missing/empty instead of silently prerendering zero API pages.

Potential Risks

• Generated API payloads must exist before Nuxt starts; build/dev/generate/postinstall now run scripts/generate-api-reference.ts first.
• The RSS number below is whole-process build RSS, not JS heap. It includes both the inline markdown reduction and generated per-tag payload path.

Review Notes

• Verified pnpm typecheck:scripts.
• Verified /usr/bin/time -l pnpm build; prerender completed 1073 routes with max RSS 6516359168 bytes locally.
• Build output confirms generated API payloads split into separate chunks, e.g. users-BQDrUbI0.mjs 88.7 kB, fields-Bt4Ud968.mjs 47.4 kB, meta-BPhtIzF4.mjs 31.8 kB, plus _tag_ route chunk.

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
docs	Ready	Preview, Comment	Jun 1, 2026 9:43pm