core dataset features

[98sean98]

A user should be able to interact with a dataset against a local directory.

Initialise a local directory as a dataset

deploifai dataset init

The entire local directory is linked to the remote dataset. It becomes a mirror of the dataset.

Might need to write some kind of config into the global configuration file: save the absolute path of the local directory and the dataset together?

Push data

Upload files to the linked dataset, given a path to a

• directory (can be the local directory .): upload entire directory
• a file: upload this file

deploifai dataset push [local-path]

The local directory hierarchy should be respected when uploaded data to the remote dataset, i.e. the mirror is upheld. That means careful path prepending is necessary.

This command should only work if the current directory is a directory linked to a dataset, or a subdirectory there of.

Pull data

Downloads files from the linked dataset, given a remote path. A remote path can be

• a directory

deploifai dataset pull [remote-path]

The remote directory hierarchy should be respected when downloading data to the local directory, i.e. the mirror is upheld. That means careful path prepending is necessary.

This command should only work if the current directory is a directory linked to a dataset, or a subdirectory there of.

Get info about the current directory

Checks if the current directory is linked to a dataset. This command should also work if it runs inside a subdirectory of the main directory linked to the dataset.

deploifai dataset info

Expected output:

In root of directory linked to dataset

Linked to dataset-name

In subdirectory of directory linked to dataset

Linked to dataset-name/subdirectory-1/subdirectory-2