Vite 6

[sapphi-red]

We're starting a discussion to gather early feedback from the downstream ecosystem and users about their needs for Vite's next Major.

Before commenting, please review current open issues and check the previous Vite Major discussions:

• Vite 3
• Vite 4
• Vite 5

Feel free to bring back proposals if you want to expand on them. If possible, create an issue for your feature and link it to this discussion so we can better track the proposal.

Timeline

We've been aligning Vite's major releases with Node EOL. In Vite 5, we dropped both Node 14 and 16. The next EOL of Node's LTS version (Node 18) is in May 2025 which is far away from now. We discussed with the Vite team so far is that Vite 6 will be released as needed this year.

Node.js versions EOL

Initial thoughts

The changes that would trigger this new major could be:

• The Vite Runtime API is ready to go out of experimental and ssrLoadModule can be implemented on top of it
• Rolldown is ready to replace esbuild for prebundling dependencies. This may actually be for Vite 7, and in Vite 6 it could end up being opt-in to start testing the waters.

Check out the Vite 6 milestone, we'll keep it updated with new issues and PRs as we work towards the next Major. We're eager to hear your feedback and ideas.

Note: We started work on Vite 6 prompted by the new Environment API. Check out the rationale here.

[ghiscoding]

For me the following 2 PRs is what I'm really waiting for. I'm assuming that they should be merged in v5 but since they've been opened for a while now, are they better suited for v6? Maybe Parcel Watcher could be on v6.

• CSP: feat: csp nonce support #14653 (rebased in PR feat: csp nonce support #16052 and available in v5.2.x)
• Replace Chokidar with Parcel Watcher: feat(server)!: use parcel watcher #14731

CSP (Content Security Policy) is most probably what a lot of users are waiting for. I would really like to see Vite supporting CSP out of the box for both Dev and Prod environments with the same code base.

Another subject that I'm having difficulties also is highlighted in the Discussion below in HMR/Watch in Library Mode, it would be great if Vite could implement this too

• Vite "project references" for running the dev server with on-demand build & HMR for dependency Vite library projects #10041

ahh wait, there's also this old issue which was opened by Evan

• Lib mode format: es.min #6555

[rtsao]

I would love to see #15395 included in an upcoming major or minor release. It fixes one of the major underlying causes of an infinite hang (#11468) which is currently the second-most discussed issue of all time (with 158+ replies since 2022). I know this silent hang can be a significant source of frustration because there's no actionable feedback or error messages and it has proven especially challenging for developers to debug. Working around this problem can be nearly impossible if import graphs are large and difficult to refactor, so an upstream fix would be super helpful.

[lucsky]

I would love to see import attributes being supported and provided to plugins. #12140 was closed on the premise that import assertions were being phased out, but import attributes (their replacement) are now both in TC39 stage 3 and fully supported by Typescript.

[korkorran]

What could be great is a custom error modal.
Handling compilation error with a custom modal, letting you display component not affected by the error (header, footer, sidebar).
It would enable great collaboration on vite apps.

[emosheeep]

Can Vite support bundleless transform? Preserve directory structure and only transpile code. It helps to customize export field with match rules in package.json and may have a better building or tree-shaking performance.

[leookun]

build.rollupOptions.output.manualChunks accept an original file path, return a chunk path, might be able to meet this

[aarjithn]

It is not super straightforward with rollup - here is a long discussion rollup/rollup#2756

[patak-dev]

Hey! We decided to release the Environment API branch as [email protected]. The changes started by the Runtime API are now prompting a complete review of the way Vite handles environments (client, SSR, workerd, etc). We were originally going to release this branch as [email protected] but the changes are now too big for a minor so we're going to start working on the next major directly with it. We're probably going to have ~2 months of alpha period and once we feel confident with the stability of the new APIs, we'll kick the beta.

If you are building a framework or are working in a Runtime (like the Cloudflare team), the next months are going to be very important to get the APIs right and serve as the base for the next iteration. We're thinking on taking a lifting responsibilities to be able to share more in the ecosystem. For example, we're introducing the concept of a ViteBuilder that lets frameworks orchestrate all the environment builds they need to get a single vite build command to build the whole app. This means changing our stance on the "framework-as-a-plugin" pattern (SvelteKit, Remix, QwikCity, etc), and making it a first class pattern from Vite 6 forward.

Please check out the current Environment API docs and give us feedback. We created a new discussion to gather feedback about the new APIs. And if you think someone you know would be interested in helping us testing (Ecosystem CI is failing for 1/3 of downstream projects now) and creating prototypes (Workerd framework-agnostic integrations, RSC, etc), please contact them and send them the PR and the discussion.

[panstromek]

I'm waiting for this one of mine: #16184, it's already in 6.0.0-alpha.1, which contains also some other bug fixes.

[patak-dev]

#16184 has been released now in [email protected]

[panstromek]

Thanks a lot!

[patak-dev]

Thanks for the heads up @larouxn. This has been resolved now. I'll hide the message because it is unrelated to this discussion.

[leookun]

Seems there's no way to add build cache for vite project

[DrGor]

No hope for #1579 ?

[Lyumih]

I hope this problems with loading 3000+ components in development mode of large project will be solved. Maybe batch, chanks or another algorithm.

Issue: #13570
Example: https://github.com/Lyumih/vite-check-components-loading/tree/main

[sanbei101]

This is a great feat, look forward to the birth of the rolldown!

[joebochill]

Any plans to support custom project templates the way you could do with CRA?

#2155

[rtritto]

What do you think to use ./.vite directory instead of ./node_modules as default directory for all temporary files (cache, dist etc)?

Some advantages/benefits:

• it's cleaner, explicit and clear (e.g. in the project, to check/delete it, you don't have to navigate inside the directory)
• support (with 0 config) every package manager (e.g. yarn with Plug'n'Play) that doesn't use node-modules directory

My current settings:

// vite.config.ts
export default {
  cacheDir: '.vite',
  build: {
    outDir: '.vite/dist'
  }
}

[itamar-smirra-port]

I would really appreciate to see this issue fixed - 13672. Currently I'm using Vite 5 and my chunks gets different hashes even when the code wasn't changed. All the suggested workarounds didn't work for me.
Thanks!

[rtritto]

What about to replace ws with uWebSockets.js?
Related #16579

FYI @privatenumber

[sai-cb]

For visibility, the beta.3 seems to have introduced a regression, producing larger chunks during build:

#18523

[ghiscoding]

Out of curiosity, I wonder what is the rough estimate of Vite 6 release? I'm asking because it was mentioned that it would probably happen this year but the EOY is not that far away anymore. Also, if a release does happen this year then will it keep current Node 18 requirement or will that bump to Node 20? Because that too is not that far away anymore. Cheers

[patak-dev]

We only drop Node versions after their EOL. So we aren't dropping Node 18 in Vite 6. We'll release Vite 7 around May to do that.

About Vite 6 release date, we're going to try to release it in around two weeks. It was delayed once more because we introduced several changes to the Environment API. It should be worth the wait.