Datasets and browser commits

Create a dataset

Select New dataset from the main Datasets page, or open a group and select New dataset there. Choose the owning group or personal account, then enter a name, path slug, and optional description.

The button appears only where your role permits dataset creation. A newly created dataset is an empty Git repository.

Upload and commit files

Open the dataset's Files tab and select Upload files. Choose one or more files and select a storage mode:

• Automatic sends binary files and files larger than 10 MiB to Git LFS.
• Ordinary Git forces the selected files into Git.
• Git LFS forces the selected files into S3-backed LFS storage.

Uploads are staged in a private browser draft. The published branch does not change until you enter a commit message and select Commit changes. You may discard the draft instead. File deletions use the same draft-and-commit flow.

Large LFS content travels directly between the browser and S3-compatible storage, including multipart uploads. Django authorizes the operation but does not proxy the file body.

Browse a dataset

Dataset pages provide separate views for:

• Files, including directory navigation, README rendering, and individual downloads;
• History, containing commits reachable from the selected revision;
• Branches and Tags;
• Access, for users allowed to manage explicit grants; and
• Settings, for editing the name, slug, or description and, when authorized, deleting the dataset.

The clone popover provides the Niyān command and repository URL for opening the same dataset locally.

When to use the CLI instead

Browser commits are useful for manual uploads and small sets of changes. Prefer the CLI when preserving a directory tree, working across branches, reviewing local changes, automating a repeatable process, or maintaining a dataset over time.