Resetting Staged Files

Sometimes you stage files with crab add and then change your mind — maybe you added the wrong files, or you want to re-chunk with different settings. Resetting unstages files from git's index and cleans up the corresponding chunk data from Crab's local staging area.

What Reset Does

When you crab add a file, two things happen:

1. The file's chunks are written to .crab/staging/
2. A pointer blob is added to git's index (via git add)

Resetting reverses both: it runs git reset HEAD on the pointer files and removes the orphaned chunk data from the staging area.

When to Reset

• Added the wrong files — You ran crab add '*.bin' but meant to exclude some files.
• Want to re-add — You changed chunking settings and want to re-process files.
• Cleaning up after manual git operations — You used git reset or git rm directly, leaving orphaned staging data.

Basic Usage

# Unstage specific file types
crab reset '*.safetensors'

# Unstage files from a directory
crab reset 'models/*'

# Preview what would be unstaged
crab reset --dry-run '*.bin'

Sync Mode

If you've used git reset or git rm directly (bypassing Crab), the staging area may contain chunk data for files that are no longer in git's index. Sync mode cleans these orphans:

crab reset --sync

This scans the staging area, checks each entry against git's current index, and removes data for files that are no longer staged.

Relationship to Git Reset

crab reset is a superset of git reset HEAD -- <paths>. It does everything git reset does (unstage from the index) plus cleans up Crab's staging data. If you only use git reset directly, the staging area retains orphaned chunks until you run crab reset --sync or crab staging clean.

CLI Reference

For complete command syntax and options, see the crab reset reference.