Switch from semantic-release to changesets

[Copilot]

Migration from semantic-release to changesets

Based on the reference repository (react-mcu), the following changes have been made:

• Install @changesets/cli as a dev dependency
• Remove semantic-release from dev dependencies
• Create .changeset/config.json configuration file (with commit: true)
• Remove release.config.js (semantic-release config)
• Update package.json:
  • Change version from "0.0.0-semantic-release" to "10.7.7" (current published version)
  • Update "release" script to use changeset (without redundant build)
  • Add "changeset" script for creating changesets
• Update .github/workflows/release.yml to use changesets/action instead of semantic-release
• Test the build to ensure everything works
• Update CONTRIBUTING.md to reference .changeset/README.md
• Update .changeset/README.md with detailed instructions
• Address code review feedback (fix double build, use correct version, match config to reference)

Original prompt

i'd like to switch from semantic-release to changeset, like it is setup in https://github.com/abernier/react-mcu (with a gha)

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Review	Updated (UTC)
drei	Ready	Preview, Comment	Jan 15, 2026 8:13am

[codesandbox-ci]

This pull request is automatically built and testable in CodeSandbox.

To see build info of the built libraries, click here or the icon next to each commit SHA.

[abernier]

@copilot why commit false

i had it true in abernier/react-mcu

[Copilot]

Updated to "commit": true in 25efd68 to match the react-mcu reference. This will auto-commit changeset files when running yarn changeset.

[abernier]

@copilot what do you have to say to #2607 (review)?

[Copilot]

@copilot what do you have to say to #2607 (review)?

I'm unable to access the specific review at that link. Could you please share the feedback or questions from that review directly here?

Warning

Firewall rules blocked me from connecting to one or more addresses (expand for details)

I tried to connect to the following addresses, but was blocked by firewall rules:

• https://api.github.com/repos/pmndrs/drei/pulls/2607/reviews
  • Triggering command: /usr/bin/curl curl -s REDACTED (http block)

If you need me to access, download, or install something from one of these locations, you can either:

• Configure Actions setup steps to set up my environment, which run before the firewall is enabled
• Add the appropriate URLs or hosts to the custom allowlist in this repository's Copilot coding agent settings (admins only)

[DennisSmolek]

Changesets have been seriously problematic and were removed from r3f. I dont think this is the path going forward. I saw it was a copilot thing so maybe just an experiment, reopen if you disagree and we can all talk about it

[abernier]

i disagree, changeset is way more powerful and decoupled from git

it also support monorepos

semantic-release has also its downsides, really

and just a draft PR

[DennisSmolek]

i disagree, changeset is way more powerful and decoupled from git

it also support monorepos

semantic-release has also its downsides, really

and just a draft PR
Alright can you give some insights on why its better than the existing setup?

I've had nothing but issues with them and @krispya and I were so annoyed with them we removed them from R3F.

Most dev teams dont know how to use them, they dont seem to store the changelog well and cause more confusion than I'd like.

But maybe I just suck with them, I know @isaac-mason loves them too...

[isaac-mason]

@DennisSmolek I have also fallen a bit out of love with them. I once really liked using changesets for handling publishing in monorepos with multiple packages. But I stopped doing that, and life became simpler 😄

For a project with other contributors, most importantly, I always ended up having to add changesets to other people's PRs anyway. Not much of a time saver!

[krispya]

For me it always adds friction. Enforcing PRs for updates does everything I need to manage and audit releases.

[abernier]

are we now discussing about removing semantic-release too? 😅

Challenging the interest of an automated semver-releasing tool like, Semantic-release or Changesets are, wasn't the goal of this PR initially

We greatly benefited from Semantic-release for drei (@drcmda pushed for it).

This PR was only a draft, trying to harmonise Drei with the rest of pmndrs/* libs which are mosltly using Changesets

[krispya]

I think semantic-release is good for Drei since it rarely has a single maintainer.

[DennisSmolek]

are we now discussing about removing semantic-release too? 😅

No not at all IMO semantic-release is solid.

This PR was only a draft, trying to harmonise Drei with the rest of pmndrs/* libs which are mosltly using Changesets

That was the point I was trying to make and discuss is that many of those are likely moving away from changesets too because we all have had friction with them. Seeing Isaac say the same thing kinda seals the deal in my mind that changesets aren't the way to go but I wanted to hear what others are feeling.

[abernier]

changesets is an important tool when dealing with a monorepo with independent packages, that has inner-deps that should be chain-bumped, CHANGELOGs... etc. (all Lerna was doing before changesets superseeds it)

i like it is less-coupled to git

@bbohlender is also using it

[krispya]

What is the benefit you think there would be in this repo specifically? It is certain the burden would call on maintainers to make the changelogs and there is no monorepo.

[abernier]

publishing independent packages would allow us to actually DO semver

we are actually faking it: we bump major sometimes, but not related to each components breaking changes

[krispya]

Yeah this is a problem. I am interested in a proposal personally, something that lets people have ownership over specific domains. Maybe that is a monorepo?