[TanStack Start] Server functions: Could we add first-class tRPC support until server functions reach tRPC feature parity?

[alexwork1611]

Hello,

Honest question here. Why is TanStack Start competing against tRPC by making their own custom RPC implementation? Doesn't tRPC work great already? I've got a working (production-grade) TSS + tRPC project going and it's a lot of fun. From batch queries to type-safe automatic query key generation for invalidating queries.

I tried working with TSS's server functions and they felt very primitive and basic.

And I also saw that TanStack Start removed the tRPC examples, which tells me that TSS is trying to replace tRPC altogether, when tRPC is clearly the educated choice thanks to its maturity, stability, API feature richness, DX, and so on.

@KATT and @tannerlinsley, didn't you discuss a possible first-class solution for TanStack Start and tRPC? At least for TSS's inception (maybe up to v2, when TSS could have their own mature RPC implementation). This would make tRPC a transitionary helper so that TSS has more features out of the box from very early stages, preceding a more robust, proprietary RPC system.

Please consider adding first-class tRPC support, at least as an officially supported alternative to server functions.

Hope this makes sense and can be a productive conversation starter.

Thank you!

[ivawzh]

I'm genuinely interested in understanding TanStack's server function roadmap. If server functions don't plan to support features like batching and contract export in the near future, and considering server function's incompatibility with other frameworks compared to libs like tRPC, I'm struggling to see compelling reasons to choose server functions over alternatives like oRPC or tRPC.

Is the primary benefit of server functions to have server code defined within frontend files?
I'd appreciate clarification on what makes server functions a worthwhile choice given these current limitations.

[alexwork1611]

Your thinking is spot on. Unfortunately, it looks like the TSS team has an idealistic approach to framework development and want to reinvent the wheel, when they should have leveraged the work done by guys like @KATT. Sad.

[AhmedBaset]

I'm curious to know what you mean by first-class support more than what's already available?

And I also saw that TanStack Start removed the tRPC examples, which tells me that TSS is trying to replace tRPC altogether

https://tanstack.com/start/latest/docs/framework/react/examples/with-trpc

[tannerlinsley]

Let me clear some things up:

• We consulted with TRPC team on server functions they were extremely helpful
• when faced with the option to build server functions we did so with a lightweight but feature full approach embracing type safety and a few features from trpc we found critical. But not all of them on purpose.
• When faced with the decision to enhance server/api routes we decided not to and instead will be recommending (and offering examples and integration documentation for) TRPC, hono and many others.

server functions are not a reinvention of the wheel since they fundamentally do something that none of these tools do with regards to isomorphism, code splitting and colocation.

We have no plans to replicate any full stack public api design tools including TRPC.