You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We need an end-to-end working example for gluing a data reference and recalling it. A binder which enables this will will dictate the API improvements to enable the functionality.
The text was updated successfully, but these errors were encountered:
For reference: #37 was a stab at achieving this change in combination with other changes. While that PR was too complex and large, we want a similar extension of the glue api contract to support remote referenced scraps. This should enable table references and blob store references that can be recalled programmatically.
Some concerns to hash out more:
Does the notebook author need complete reference paths? Likely yes, but custom reference endpoints that get registered could support default prefixes. e.g. ref=s3://bucket/prefix/filename.ext vs ref=filestore://filename.ext
Do we reuse papermill's IO registry? Unclear but would like to not have the large dependency chain
How should the kwargs for glue change? Contract changes here need to be carefully applied to avoid confusion and usability concerns.
We need an end-to-end working example for gluing a data reference and recalling it. A binder which enables this will will dictate the API improvements to enable the functionality.
The text was updated successfully, but these errors were encountered: