Skip to content

Cleanup and backups

Cleanup is the third tab of the repository, for freeing space in the open project. It scans in the background and fills in as it reads.

Ready to clean lists what can go, each row with its size and action:

  • Finished worktrees — worktrees whose work is already in the default branch, squash merges included. Their build output goes to the Trash first, and uncommitted and other ignored files, such as .env, go into a recovery copy.
  • Merged branches — merged and not checked out. Squash-merged branches — their remote branch is gone and main holds their changes; a recovery copy pins their commits first.
  • Build output and packages — ignored folders the next build or install writes again, such as build, node_modules and Pods, in every worktree of the project. Each folder names its rule and the command that restores it (Restore: npm install). Tracked files, .env and folders holding a release build stay unticked.
  • Chats — the Claude Code and Codex chats that ran in the project’s folders; Clean up… deletes old ones or frees only what their next run writes again.
  • Temporary Git packs — leftover tmp_pack_* files older than an hour.
  • Backups, stale worktree entries and Git’s object store, each with its own dialog.

Check first lists worktrees and branches that hold work the default branch lacks; each opens the usual confirmation. Worktrees someone works in — this window, an agent, a terminal — are never offered.

Suggestions read the repository itself. An action stages its change for you to commit in Changes; the files stay on disk.

  • Build output in Git, Likely secrets in Git (by name only) and System and editor files in Git such as .DS_Store offer Stop tracking….
  • Build folders .gitignore misses offers Add to .gitignore….
  • Names that break on other systems lists paths that differ only in case, names Windows forbids and paths over 260 characters.
  • What makes the repository grow names the files whose versions take the most space over the history.
  • Large files without Git LFS offers Track with LFS… when Git LFS is installed and set up for the repository; ffgit never runs git lfs install.
  • How pull requests are merged appears when merge commits brought many wip and fix commits, and links to the project’s settings.
  • Size limit sets the size over which Changes marks a staged file.

Off until you pick a rule, kept per project: once an hour ffgit removes a worktree an agent tool made for chats when it was idle for 3, 7, 14 or 30 days, its chats are settled, its pull or merge request merged, or nothing of its own differs from the default branch. Never while a chat runs there, and always through the same recovery copy as Remove worktree…. The bell names what went.

A clone can be partial: earlier file versions, such as golden images, stay on the remote until a diff or image comparison opens them, and ffgit downloads them in one batch. Make lighter… turns a clone already on disk into a partial clone in place, keeping branches, stashes, unpushed commits, worktrees and backups. It deletes old versions only after Git has checked the result and the remote has served one of them.

With Automatic backups on, ffgit saves a recovery copy before switching branches, discarding, stash, merge, rebase, pull, editing history and other commands that rewrite files. Copies store only changed contents and are kept in the project’s .ffgit/backups folder, ignored by Git, until you delete them.

From Backups you can:

  • Verify a copy;
  • Restore to new folder… — saved files and staged changes, never over your work;
  • Export… a copy to move it to another computer, or Import backup…;
  • Create backup on demand, and Clean up… old copies by age, with the space freed named first.

The folder and its disk use show in Backups, and the location and retention can be changed.