dandi upload#

dandi [<global options>] upload [<options>] [<path> ...]

Upload Dandiset files to DANDI Archive.

The target Dandiset to upload to must already be registered in the archive, and a dandiset.yaml file must exist in the common ancestor of the given paths (or the current directory, if no paths are specified) or a parent directory thereof.

Local Dandisets should pass validation. For that, the assets should first be organized using the dandi organize command.

By default, all *.nwb, *.zarr, and *.ngff assets in the Dandiset (ignoring directories starting with a period) will be considered for the upload. You can point to specific files you would like to validate and have uploaded.


-e, --existing [error|skip|force|overwrite|refresh]#

How to handle files that already exist on the server:

  • error — raise an error

  • skip — skip the file

  • force — force reupload

  • overwrite — force upload if either size or modification time differs

  • refresh [default] — upload only if local modification time is ahead of the remote

-i, --dandi-instance <instance>#

DANDI instance (either a base URL or a known instance name) to upload to [default: dandi]

-J, --jobs N[:M]#

Number of assets to upload in parallel and, optionally, number of upload threads per asset [default: 5:5]


Delete assets on the server that do not exist locally after uploading

--validation [require|skip|ignore]#

How to handle invalid assets:

  • require [default] — Do not upload any invalid assets

  • skip — Do not check assets for validity

  • ignore — Emit an error message for invalid assets but upload them anyway

Data should pass validation before uploading. Use of this option is highly discouraged.

Development Options#

The following options are intended only for development & testing purposes. They are only available if the DANDI_DEVEL environment variable is set to a nonempty value.


Upload all file types, not just NWBs and Zarrs


Do not use pyout callbacks, do not swallow exceptions, do not parallelize.


Update Dandiset metadata based on the local dandiset.yaml file