Bucket Integrations#

Buckets integrate with MEGA CLI, the Python client, and compatible huggingface_hub workflows. They are suitable for mutable artifacts that do not need repository commits.

1from megatensors._hub import MegaFileSystem2 3fs = MegaFileSystem(token="YOUR_MEGA_TOKEN")4with fs.open("buckets/OWNER/artifacts/metrics.json", "wb") as file:5    file.write(b'{}')

An integration should use the narrowest token scope, validate paths before a copy or delete, and record the repository revision that produced an artifact.

S3 tools and SDKs#

The Bucket S3 Gateway lets applications that already use the AWS S3 API connect without rewriting their storage layer. Use a dedicated Bucket credential, configure path-style addressing, and keep the endpoint, access key, and secret in your application's normal secret store.

S3 clients can also provide a filesystem-style mount when that is the best fit for an existing application. For MEGA's mount extension and platform-managed Space volumes, see Bucket Access Patterns.