Cloud storage

Octue SDK is currently integrated with Google Cloud Storage. Later on, we intend to integrate with other cloud providers, too. If you need a certain cloud provider and are interested in contributing to or sponsoring its development in Octue SDK, please join the discussion in this issue.

Data container classes

All of the data container classes in the SDK have a to_cloud and a from_cloud method, which handles their upload/download to/from the cloud, including all relevant metadata from the instance (e.g. labels, ID). Data integrity is checked before and after upload and download to ensure any data corruption is avoided.


Assuming you have an instance of Datafile called my_datafile:

my_datafile.to_cloud(project_name=<project-name>, bucket_name=<bucket-name>, path_in_bucket=<path/in/bucket>)
>>> gs://<bucket-name>/<path/in/bucket>

downloaded_datafile = Datafile(path="gs://<bucket-name>/<path/in/bucket>, project_name=<project-name>)


Datasets are uploaded into an output directory. Within, a directory named after the dataset is created containing the dataset’s datafiles:

|--- <>
     |--- <file-0>
     |--- <file-1>
     |--- <file-2>

Datasets are downloaded from their directory. Assuming you have an instance of Dataset called my_dataset:

my_dataset.to_cloud(project_name=<project-name>, bucket_name=<bucket-name>, output_directory=<output-directory>)
>>> gs://<bucket-name>/<output-directory>/<>

downloaded_dataset = Dataset.from_cloud(project_name=<project-name>, bucket_name=<bucket-name>, path_to_dataset_directory=<output-directory>/<>)


Manifests are uploaded as a file to the given path. The manifest’s datasets are uploaded by default into the same directory as the manifest file, but this can be disabled by specifying store_datasets=False. Manifests are downloaded by providing the path to the manifest file. Assuming you have an instance of Manifest called my_manifest:

my_manifest.to_cloud(project_name=<project-name>, bucket_name=<bucket-name>, path_to_manifest_file=<path/to/manifest/file.json>)
>>> gs://<bucket-name>/<path/to/manifest/file.json>

downloaded_manifest = Manifest.from_cloud(project_name=<project-name>, bucket_name=<bucket-name>, path_to_manifest_file=<path/to/manifest/file.json>)

The datasets specified in a manifest can be located anywhere in any bucket as long as their paths are correct in their Dataset instances. A manifest that takes advantage of this can be created locally with datasets whose paths are set to cloud locations; it can be uploaded by supplying store_datasets=False as a parameter to Manifest.to_cloud.