Docs / CLI vs the web

CLI vs the web

NEMAR has two surfaces: the web app you’re reading right now, and the CLI. They share the same backend, the same accounts, and the same datasets. The difference is workflow style, not capability.

What the web is best for

  • Signing up. The web is the only place to create an account today; the CLI authenticates against an existing account.
  • Browsing and exploring. Discover, dataset detail pages, READMEs, and the citation dashboard live here.
  • One-off uploads. Drop a folder, walk through validation, request publication. No tooling to install.
  • Collaborator management. Inviting and (eventually) removing collaborators happens on the dataset’s collaborators page.
  • Admin review. The admin queue at /admin/publication-requests is web-only.

What the CLI is best for

  • Scripted publishes. CI pipelines that publish a dataset version when a tag is pushed.
  • Parallel uploads. The CLI uses git-annex parallel transfers for very large datasets that would be tedious to drop in a browser.
  • Server-side workflows. Running on a compute node or a shared lab server where the web UI isn’t convenient.
  • Diffing and version checks. The CLI exposes tooling for inspecting dataset state, comparing local vs remote, and so on.
  • Bulk operations. If you’re managing 20+ datasets, the CLI’s JSON output integrates with shell scripts.

You don’t have to pick one

Most researchers use the web for sign-in, discovery, and publication-review interactions, and reach for the CLI when they need to script something. They’re the same account, the same datasets, the same permissions. Switching back and forth is fine.

Installing the CLI

The full install guide lives at docs.nemar.org. Quick start (Bun-based; works on macOS and Linux):

bunx nemar-cli auth login    # opens a browser to sign in
bunx nemar-cli datasets list # shows your datasets
bunx nemar-cli upload <path> # uploads a BIDS folder

Permissions and accounts

The CLI and the web share session state via the same backend. Signing in on the CLI issues a long-lived API token tied to your account; signing in on the web issues a cookie-backed session. Both honor the same admin role, the same dataset ownership, and the same collaborator permissions.