I connected my AI application to a Google Drive data store with the Drive source set to ‘All’.
Does this mean each user that talks to the agent will be able to access
- All their own drives?
- Only those drives they have in common with the user who created the data store?
I tried to query the agent via a second user for a file that the original data store creator did not have access to - and it worked.
Does that mean this answer about ''drives in common" Agent Builder data store with Google Drive: the whole Drive!? is outdated?
The tutorial panel at the time of data store creation also seems to be misleading
“Specify the drive source for your data store: All: To add your entire drive to the data store.”
Yes that’s exactly correct. At least right now when you select “All” it is the whole organization drive and all files created by any user of the organization, this puts the pressure on the useres to handle permissions carefully to prevent their results from appearing to unintended searchers
Based on your test, it sounds like the Google Drive data store is resolving access at query time against the current end user’s Drive permissions, not only against the files shared with the original data store creator. That would also fit the data-federation model, where Drive content is retrieved from the source rather than copied into a static index. If that is correct, the UI text “your entire drive” is misleading, because it can be read as the creator’s Drive, while the runtime behavior appears to be user-permission scoped. It would be helpful if Google clarified whether “All” means “all Drive content accessible to the querying user,” and whether the older “drives in common” answer is outdated.