init-dataset-repo

Init Dataset Repo

Creates a clean local git repository structured for Hugging Face hosting. Does not push anywhere — hf-setup handles remote creation and push.

Inputs

• Dataset name (kebab-case for the directory and HF repo name).
• Path for the repo. Default: <workspace>/dataset-repo/ if a shape-dataset workspace exists, otherwise ask.
• License (defaults to mit if the user hasn't decided; flag this in the card as a TODO).

Layout produced

<dataset-name>/
├── .gitattributes          # LFS rules
├── .gitignore              # excludes scratch/, .env, __pycache__/
├── LICENSE                 # full license text
├── README.md               # dataset card stub
└── data/                   # empty; prepared splits go here later

.gitattributes contents

*.parquet filter=lfs diff=lfs merge=lfs -text
*.arrow   filter=lfs diff=lfs merge=lfs -text
*.bin     filter=lfs diff=lfs merge=lfs -text
*.h5      filter=lfs diff=lfs merge=lfs -text
*.zip     filter=lfs diff=lfs merge=lfs -text
*.tar.gz  filter=lfs diff=lfs merge=lfs -text
*.jsonl   filter=lfs diff=lfs merge=lfs -text
*.csv     filter=lfs diff=lfs merge=lfs -text
*.png     filter=lfs diff=lfs merge=lfs -text
*.jpg     filter=lfs diff=lfs merge=lfs -text
*.wav     filter=lfs diff=lfs merge=lfs -text
*.mp3     filter=lfs diff=lfs merge=lfs -text

If certain formats are known to be small in this dataset, the user can prune later — better to over-include LFS rules at init time.

README.md stub

Minimal HF dataset card with YAML frontmatter and section headers. Every value is <!-- TODO --> until prep produces real metadata. The full population happens in hf-setup.

Steps

1. mkdir -p <path>/data
2. Write .gitattributes, .gitignore, LICENSE, README.md.
3. git init and git lfs install inside the repo.
4. Initial commit: git add -A && git commit -m "Initialize dataset repo skeleton".
5. Report the path back to the user.

Do not add any HF remote here. hf-setup does that after asking public/private.