AI code review

This is the one AI feature explicitly allowed an opinion — every explanation feature is told not to have one, and this one is told the opposite: read the diff and flag what deserves a second look. It only ever sees the diff, so its findings are a prompt to check, not a verdict.

Reviewing the working changes

Choosing "Review changes (LLM)" on the working-tree row reads everything you haven't committed yet and reviews it in the same right panel the explanations use. Like the working explanation, nothing is remembered — the tree moves constantly, so every open reviews fresh.

Captured from the running app by this scenario.

Do this

1. Choose Review changes (LLM) from the working-tree row menu

You should see

• The explanation panel shows a finished explanation

Reviewing a branch's range diff

Choosing "Review branch changes (LLM)" reads the whole range between a branch and its base — the same comparison the branch explanation uses — and reviews it before you open a PR. The review is remembered per branch, same as the branch explanation.

Captured from the running app by this scenario.

Do this

1. Choose Review branch changes (LLM) for the feature/login branch (base main)

You should see

• The explanation panel shows a finished explanation

---

This page is generated from apps/e2e/features/ai-code-review.feature, the end-to-end test that drives the feature it describes.