Writing changelogs is boring, easy to forget, and always happens at the worst possible time. In a team that ships frequently — like ours at Waveful — it quickly becomes a chore you cannot ignore.
So we automated it.
How it works (it's dumb simple)
ChangelogAction is a GitHub Action that generates a CHANGELOG.md entry on every new tag. It pulls the merged PRs and commits since the previous release, packs them into a structured prompt, and lets an LLM write the actual changelog. The result is committed straight to your main branch.
If you want, it can also send a Telegram notification with a link to the updated file, and annotate the release in Mixpanel so your analytics stay in sync with deployments. Both are optional.
The boring stuff, automated
The real reason we built this is simple: a language model is incredibly good at turning messy git history into readable text. PR titles, commit messages, file diffs — the LLM takes all of that noise and returns something coherent. No manual copy-paste, no forgotten changes, no Friday-evening rush.
It is the kind of text manipulation AI excels at, and it saves us time on every single release.
Dogfooding it since day one
We have been running it on our repositories for months. Push a tag, wait a few seconds, and the changelog is ready.
It is open source under Apache 2.0. If you ship regularly and have an LLM API handy, give it a try.