Skip to main content

FAQ

Why does gji need shell integration?​

Normal CLI processes cannot change the working directory of the parent shell. The shell integration bridges that gap so commands like gji go can move your shell directly.

Can I use gji without shell integration?​

Yes. Use --print and consume the returned path in scripts or wrappers.

Why are worktrees stored outside the main repository?​

That keeps branch-specific files isolated and avoids mixing generated outputs, dependency folders, and editor state into a single checkout.

Why do worktrees matter more in AI-assisted workflows?​

Because AI tends to increase the number of parallel tasks around one repository, not decrease it.

You may want your main feature open, a PR review open, and a scratch branch for agent-generated changes at the same time. Worktrees give each task its own directory and state so they do not step on each other.

gji makes that style practical enough to use constantly.

What path layout does gji use?​

Worktrees are stored at:

../worktrees/<repo>/<branch>

That deterministic layout makes shell aliases, editor bookmarks, and scripts easier to keep stable.

Do I need to reinstall dependencies in every worktree?​

Usually yes, because each worktree is a separate directory with its own dependency state. Hooks make that painless by running your install command automatically.

How do I copy gitignored files like .env.local into new worktrees?​

Use gji sync-files from the main repository:

gji sync-files add .env.local .npmrc

New worktrees receive those files before install hooks run. The setting is stored in your global per-repo config so private file names do not need to be committed.

Can I use URL and PR number formats together?​

Yes. gji pr accepts a numeric PR reference or a full GitHub pull request URL.

Is configuration global or repo-specific?​

Both. Global defaults live in ~/.config/gji/config.json, and .gji.json can override them per repository.

Can I re-run setup without recreating the worktree?​

Yes. Use:

gji trigger-hook afterCreate

That is useful when the worktree already exists but you need to rerun install or bootstrap steps.