> This is the Markdown rendition of https://pyxle.dev/docs/guides/build-optimization, served for AI agents and assistants. > Append `.md` to any pyxle.dev URL to fetch its Markdown; the links below already do. > Index of every page: https://pyxle.dev/llms.txt · Whole docs in one file: https://pyxle.dev/llms-full.txt > Search the docs: https://pyxle.dev/api/docs-search?q=YOUR+QUERY (returns matching pages as Markdown links) # Build Optimization `pyxle build` produces a hashed, minified, code-split client bundle and a matching SSR shell. Most of the optimization is automatic; this guide covers what you get for free, how to inspect it, and how to push image and load performance further. ## What you get automatically Pyxle builds the client with [Vite](https://vitejs.dev/) in production mode, so out of the box: - **Minification** — JavaScript and CSS are minified with esbuild. - **Tree-shaking** — unused exports (including unused `pyxle/client` helpers) are dropped by Rollup. - **Code splitting** — shared code lands in its own chunks, so two pages that import the same component download it once and cache it across navigations. - **Content hashing** — every asset is fingerprinted (`index-DzNTdgZx.js`) for immutable, long-lived caching. - **Per-page CSS** — only the CSS a page actually uses is linked from its shell. ## Preload hints Pyxle's SSR shell injects a `` for the page's entry module **and every chunk it statically imports**, derived from the build manifest's import graph. The browser then fetches those chunks in parallel with HTML parsing instead of discovering them only after parsing the entry — a meaningful first-load win on multi-chunk pages. It's automatic. (Browsers that don't support `modulepreload` simply ignore the hints — the page still loads.) Every hint carries `fetchpriority="low"`: the chunks exist to hydrate a page the server has already painted, so they must not compete with the resources that produce that paint (the document, its CSS, the LCP image). With idle bandwidth — the common case once the shell has arrived — low-priority requests still start immediately, so hydration timing is unchanged; on a contended connection the paint-critical resources win, which is the order you want. For content-first pages you can go further: [`assets.modulePreload: false`](https://pyxle.dev/docs/reference/configuration.md#asset-delivery) drops the hints entirely, and [`assets.hydration: "after-paint"`](https://pyxle.dev/docs/reference/configuration.md#asset-delivery) holds the entry `