gpui-archipelago (GPUI-lago)

gpui-archipelago

(aka GPUI-lago — “jee pyu ay la go”)

Every fork is an island. There’s no mainland and no one’s building one, so the islands have to find each other. Here they can.

Ideally, Zed would have published GPUI as a stable, versioned crate, but that didn’t happen. Various people forked it instead, and each fork ended up solving a different part of the problem:

  • gpui-unofficial — publishes up-to-date crates on crates.io, which covers distribution. Currently needs a maintainer.
  • gpui-kit — Longbridge's shadow distribution: republishes upstream Zed crates to crates.io as gpui-pre under its own namespace, fronted by a gpui-kit facade over gpui-base, gpui-component, and gpui-shell — Jason Lee's answer to 315 days without an official release.
  • gpui-ce — a community fork working to accelerate testing of upstream features and maintain a faster-moving community version, publishing crates for standalone desktop apps.
  • Kael — forked from GPUI direct; selectively tracks upstream while building out general-purpose features (gradients, webviews, form controls, component library).
  • WGPUI — GPUI ported to a single cross-platform backend (wgpu + winit) instead of per-platform code — leaner, and features land across Windows, macOS, Linux, and WebAssembly in one code path. WgpuSurface lets you embed raw wgpu render passes (3D, game engines, custom shaders, anything) directly in your UI tree, making WGPUI suitable as a shell for 3D applications and mixed-mode interfaces.

Meanwhile people keep turning up across GitHub issues, Discord, and Reddit threads with similar frustrations across these multiple projects.

What this is

gpui-archipelago is a community-maintained space for anyone building with GPUI outside of Zed. Maybe you run your own fork, maybe you depend on someone else’s, maybe you want to ship something on top of GPUI and minimize your chances of encountering unnecessary instability and breakage.

The goal is not to pick a winner, but to understand the landscape: what each approach solves, what it doesn’t, and where common ground might exist across forks.

Forks at a glance

Fork Stars Forks Crate
Zed (upstream GPUI) 88,000 10,000 ◑ Outdated †
gpui-ce governance still forming 925 105 ● Published
gpui-unofficial needs a maintainer upstream mirror 44 7 ● Published
gpui-pre — — ● Published
WGPUI 64 12 ○ Unpublished
Kael 34 2 ● Published
GPUI Box — — ● Published

† The official gpui crate on crates.io is months behind upstream. For current upstream crates, use gpui-unofficial — an automated mirror of upstream Zed.

Star and fork counts are manually maintained in the Manifest — help keep them current.

Two kinds of problems

Ecosystem problems

Problems with how GPUI presents itself as a software project: distribution, stability, versioning, docs, guides for getting started at all. These are usually fixable without touching GPUI source code.

Capability problems

Problems with what GPUI can do: missing elements, missing escape hatches, missing points where the API can be extended. These need code changes, and different forks may make those changes in incompatible ways. This is where a shared spec is useful.

How issues are structured

Each issue lives in issues/. The format:

issue template 
## The problem
Written from the perspective of someone who just hit this wall.
What did they try to do, and what went wrong?

## Why it exists upstream
A plain explanation of the Zed priorities or constraints behind
the gap. Give context, don't assign blame.

## How people currently solve it
Known approaches: forks, crates, workarounds, conventions.
For each one: what it solves, what it doesn't, who maintains it.

## What nobody has solved yet
The part that's still genuinely open.

## Common ground opportunity
Optional. A convention, interface, or spec that several forks
could adopt to stay compatible with each other.

How to contribute

  • Add an issue — open a PR with a new file in issues/. Partial drafts and LLM-generated text are fine, just keep it reasonable and concise.
  • Add an approach — know how a fork handles an existing issue? Add it.
  • Correct something — wrong, stale, or unfair? Open an issue or PR.
  • Add your fork — see the Forks table above. Any fork is welcome, however immature.
  • Add a project — the Manifest lists forks, apps, libraries, tooling, and learning material built on GPUI. It isn’t limited to crates.
  • Share your experience — started a fork, abandoned one, got burned and learned from it? Write it down here, or come tell us about it on Discord.

Where this comes from

I started this repo after trying not to fork, failing to use upstream as-is, and picking up a few things along the way. So yes, there’s a bias. Contributions from the other islands are what would keep this from being one person’s opinion.

Status

Still extremely alpha.

News