File Status

Every Crab-tracked file in your working tree is in one of three states: hydrated (full content on disk), pointer (dehydrated, content in cloud), or modified (hydrated with uncommitted changes). Understanding these states is key to working effectively with Crab.

File States

Checking Status

crab status

Tracked files (42 total):
  H  models/weights.bin       1.2 GB  (hydrated)
  H  models/embeddings.bin    800 MB  (hydrated)
  P  data/eval.safetensors    300 MB  (pointer)
  P  data/test.safetensors    150 MB  (pointer)
  M  config/vocab.bin         2.1 MB  (modified)

Why "Hydrated Files Show as Modified"

This is the most common source of confusion for new Crab users. Here's what happens:

1. Git's index stores the pointer blob (what was committed)
2. Your working tree has the full content (after hydration)
3. git status compares them and sees a difference → "modified"

This is expected behavior, not a bug. The file content is correct — it's just different from the pointer in git's index.

To make git status clean: dehydrate before committing or pulling:

crab dehydrate --all
git status              # clean
git pull origin main    # fast, pointer-only diffs
crab hydrate --all      # restore content

Filtering by Pattern

# Check only model files
crab status '*.safetensors'

# Machine-readable output for scripting
crab status --porcelain

Porcelain output (one line per file):

p models/weights.bin
h models/embeddings.bin
m data/train.safetensors

Common Workflows

"Which files can I work with right now?"

crab status | grep "hydrated"

"What do I need to hydrate?"

crab status | grep "pointer"

"Do I have uncommitted changes in tracked files?"

crab status | grep "modified"

CLI Reference

For complete command syntax, porcelain format, and JSON output, see the crab status reference.