SolidStart: Public Roadmap - DeVinxi and Beyond

[atilafassina]

Hey, folks!

Here's a cheeky video version if you don't feel like reading.

Today marks a special day for SolidStart. Thanks to our partner Anomaly we now have formed a dedicated task force to raise the bar for our framework and maintenance. It has been some busy and amazing couple of weeks working together with Dax and his team, collecting feedback on their experience and shaping our roadmap. Which I'm happy to share with you below!

As my main task on the Solid's Core team I'll take responsibility in managing feedback and addressing them with the team. But that's just part of our renewed commitment.

Our main priority is replacing Vinxi with a pure Vite-based system, feel free to follow the ongoing work at devinxi branch. Our release plan has 3 steps.

Release	Milestone
2.0.0-alpha	Feature parity reached with current 1.x.x release
2.0.0-beta	Support for Solid 2.x.x is achieved
2.0.0	Battle-tested for edge cases and regressions

We will also work closely with the Docs team to make sure every release is properly documented for migrations.

• The 2.0.0-alpha effort is being carried by @Brendonovich and @katywings .
• Docs and Day-to-Day will be led by @brenelz and myself.

Going forward, we are moving current version of SolidStart to 1.x branch.
DeVinxi work will soon be merged in main.

New features and changes to v1 will require more thorough testing so we can port them to the new architecture as we map and perform 2.0 work.

What about features?

Work has not stopped in improving the developer experience. As an immediate follow-up to DeVinxi work we intend to make Server-Functions easier to use. We will share more on that progress soon enough, meanwhile you can join us in Solid Discord at #solidstart to share feedback and keep tabs on our progress.

It's important to say that with DeVinxi efforts, we're also bound to an update on our infrastructure, before SolidStart 2.0.0 reaches stability, we're aiming to have Nitro updated to v3 and with a more straightforward integration.

Keeping the lights on

At the moment we have many issues that lack triage and a few opened PRs that must be reviewed. We also reckon our documentation could use a touch up and some bulk. This will be added to our ongoing efforts in parallel to the DeVinxi initiative.

@brenelz and I are working with the Docs team to improve your experience using SolidStart, but we could use your help either reaching out to Docs team directly and contributing, or letting us know what are missing spots you're more eager to see covered!

And then what?

With SolidStart 2.0 we are looking into improving the developer experience for streaming and realtime experiences towards the end of 2025 and beginning of the next.

Primitives, not Frameworks

As topsy-turvy as the above quote may sound for a meta-framework, we want to provide you with flexible ergonomics and strong defaults. That has been the cornerstone of everything inside the Solid ecosystem and will remain true for SolidStart. We will remain unopionated but we will still provide users with great features for build your fullstack apps towards the future, bridging the gap and reducing the tradeoffs between Multi-Page Apps (MPAs) and Single-Page Apps (SPAs).

As always, you can find us around here or on discord.gg/solidjs

🤘 All the best,

@atilafassina, @Brendonovich , @brenelz , @katywings , and the Solid team

[doeixd]

Exciting stuff. Love the vision and momentum!

[atilafassina]

Thanks a lot, @doeixd !
Appreciate all support you've always given us! Let's go!! 🔥

[genude]

G stuff

[atilafassina]

Thanks! :)

[aarjithn]

Curious why you wouldn’t consolidate this effort under TanStack start, now that it’s ‘framework-agnostic’ especially since you seem to be following a similar path with Devinxi.

[huseeiin]

Curious why you wouldn’t consolidate this effort under TanStack start, now that it’s ‘framework-agnostic’ especially since you seem to be following a similar path with Devinxi.

the main difference is solid router, you can't use it in tanstack start

[aarjithn]

@huseeiin well there is solid version of tanstack router https://tanstack.com/router/latest/docs/framework/solid/overview

[huseeiin]

@huseeiin well there is solid version of tanstack router https://tanstack.com/router/latest/docs/framework/solid/overview

they're not the same router

[aarjithn]

True, it's not THE solid router but it's a solid router

[Brendonovich]

@aarjithn It is a router that works with solid, but the goals of SS and TSS aren't quite the same:

• TSS builds on top of Tanstack Router as a more encompassing solution, if those are the technologies you want to use
• SS doesn't build on top of Solid Router, it's more intended to be server/full-stack primitives for you to expand on, whether that's using Solid Router, Tanstack Router, or something else.

[saurabhp75]

Excited for Solid Start. It's been in the trenches for quite some time. Time to rise and shine.

[atilafassina]

Appreciate you saying that!
We have quite a few other tweaks we want to do to empower the ecosystem going forward

[LiMao00]

"replacing Vinxi with a pure Vite-based system" is based on what reasons and considerations?

[atilafassina]

Vinxi was created to circumvent shortcomings on the Vite setup - Vite 6 Environments API was their response to most of those shortcomings and now Tanstack Start, us, and even Nitro are reevaluating the best way to integrate.

The endgame (for us) is to provide our users with the best primitives to build great apps. And sometimes that means observing the ecosystem growth and reevaluating things from first principles.

[f1shy-dev]

very cool!

[everton-dgn]

@atilafassina I’m a big fan of Solid and especially SolidStart – they’re my favorite frameworks, and I can’t overstate how excited this news makes me. That said, I’d like to know if it would be possible to share, even in a very general sense, a minimum time estimate for when 2.0.0 might reach a stage considered “battle-tested” for extreme cases and regressions. Are we talking at least 2 years, maybe 3, or more? And additionally, if SolidJS itself were to reach a 2.0 version, what would be the rough idea of the timeline for that?

[atilafassina]

That said, I’d like to know if it would be possible to share, even in a very general sense, a minimum time estimate for when 2.0.0 might reach a stage considered “battle-tested” for extreme cases and regressions.

That's our goal for the stable release. While we don't work with time-based deadlines, I expect to measure it in weeks/months and not in years. Maybe I'm an optimistic, but I believe the road towards alpha is the trickiest one because there's a lot of architecture thinking and refactoring, once we get the right abstractions in place, I believe landing on the good APIs and integrations will be quicker.

[debpalash]

Been tracking solid since 2021, this sounds so exciting <3

[cbetori-hgr]

Why is the devinxi branch using tanstack Server-Functions plugin. Is this something SS will replace at some point? It feels like this is something SS should own.

[aarjithn]

I think an advantage of Tanstack is that router / start is just one part of it, but it is also supplemented and well integrated with Tanstack query / db etc. Sure, it can work with others too like react-router or solid-router but the ecosystem gets first class (nicer) support. So while people can translate from TanstackStart(React) -> TanstackStart(Solid) with relative ease, I think wiring up everything to a framework outside that could be bit more challenging.

[cbetori]

@aarjithn Im not really sure how that addresses my concern.

I don't want to use Tanstack. I can appreciate it but it's not for me.

"SolidStart remains open to all libraries in the Solid ecosystem(not just the amazing ones created by Tanstack) and help push Solid-first patterns". But does using Tanstack server functions align with that?

Maybe this isn't a big deal. I actually don't know.

[atilafassina]

Hey @cbetori , first of all I'd like to take responsibility and apologize for how we handled adopting Tanstack Server Functions plugin

Only as of SS 1.1 (which caused breaking changes).

That was a fumble on our part, there were a few regressions that (to this day) haven't been fully addressed. I hope you see that this public roadmap and the release plan are a direct reflection on our learnings from that bumpy migration start.

---

Now, fast forward to your latest point.

I don't want to use Tanstack. I can appreciate it but it's not for me.

"SolidStart remains open to all libraries in the Solid ecosystem(not just the amazing ones created by Tanstack) and help push Solid-first patterns". But does using Tanstack server functions align with that?

Very much so. SolidStart with a different mindset than most other fullstack framework. We take very seriously the meta aspect of our framework. Our focus is to the deliver a collection of primitives to help you get started (hence the name of the framework) - we don't intend to print opinions on it and we definitely don't want to reinvent wheels. We want Start to be a collection of tools that you can pick apart if you want to and the Start package will provide you with only the small set of requirements to have Solid working properly on the server and wire it to the client. All the rest, as Ryan said, is "plumbing".

The risk of tanstack abandoning their plugin exists, of course. There's also a risk that Vite stops being the best solution, or that Seroval stops being the best serializer for us, or Nitro not being the right fit for our backend. Creating a fullstack framework with zero external dependencies would be too big of an undertaking for our small team (and tbh, probably unnecessary).

So, now and today, the Server-Functions plugin fits the bill on what we need (same as the other tools I mentioned above). If something changes, we'll react and adapt as we did with Vinxi - we could've jumped into Vinxi and help maintenance. It took us away from Vinxi was not that, it was the fact that Vite/Nitro saw value in what Vinxi brought to the table and together they superseded its necessity in ~75% . Now @Brendonovich and @katywings are leading the effort on closing the gap in that other 25% and that's what SolidStart 2.0.0-alpha will be.

---

Lastly,

My concern is that SolidStart will have to rely on tools where its considered a second class citizen

That's not true. Solid's influence is growing inside Tanstack - all their UI tools are built with Solid and we collaborate a lot with the Tanstack team. SolidStart depending on Tanstack software is just the last (and most public) symptom of this symbiosis.

[cbetori-hgr]

@atilafassina Thanks for the clear and well articulated response. It helps give me a lot more confidence in choosing Solid Start for my companies next project.

[atilafassina]

thanks a lot for being so understanding and for offering candid feedback.
This is what will make our community and framework great! 🔥

[bjesuiter]

I love to see that effort. I' currently deep into exploring solidstart with tanstack router and many things look very magic to me.
(I'm a senior webdev dev btw and follow Ryan's streams, but still...)

One thing is that the boundaries between the technologies are often very unclear.

For example:
the h3 server settings can be found under the "server" key in app.config.ts => these are extremely important for using websockets in solidstart.

Or:
I used the template for tanstack router in solidstart init. This worked, but I still have confusions:

• I do have the tanstack router now completely controlling my routes dir, this is clear. But, does solidstart still control my server functions and actions?

• one issue arose with "adding a title to the head of my html document": the normal html doc is generated by the tag in solidstart in entry-server. But to be able to use "head tag merging" with tanstack router I need to render the component from tanstack router into my head. But this tag is only valid below the tanstack router provider 😃

my solution was to use a portal in the tanstack router __root file to portal this into the tag, but that feels kinda unergonomic.

enough rambling

These are only some examples which are kind of confusing. I'll try to put my findings into more actionable discussion thoughts and issues here in GitHub / discord to help drive this forward.

Today I found, that TanStack Start has a solid variant, which is basically solidstart + tanstack router + tanstack start functions.
I read the docs today and found the DX of tanstack solid start (basically tanstack start with solid) to be a huge leap over solidstart with solidstart router and even solidstart with tanstack router.
Everything feels more coherent in tanstack solid start.

I hope we can reach this level of cohesion for solidstart and make it clear where solidstart ends and tanstack router / extra tanstack start functions begin, so that it truly feels like I'm owning the pieces, like it does in solidJS.