mirror of
https://github.com/infiniflow/ragflow.git
synced 2026-07-16 20:57:21 +08:00
74fa54f122
**Addresses the Google Drive integration for #14362**
This PR completely overhauls the Google Drive sync logic to accurately
detect remote deletions, while drastically reducing the memory footprint
during the snapshot phase.
### What changed under the hood:
* **Killed the memory bloat:** Swapped out the massive document
dictionary objects for a lightweight `collections.namedtuple` (`SlimDoc
= namedtuple('SlimDoc', ['id'])`). This prevents RAM spikes during
`retrieve_all_slim_docs_perm_sync` on massive enterprise drives.
* **Flawless downstream integration:** The `SlimDoc` object relies on
simple duck typing. It perfectly delivers the `.id` attribute required
by `ConnectorService.cleanup_stale_documents_for_task`, meaning your
core `hash128` vector cleanup logic runs natively without modification.
* **Fixed the Shared Drive blindspot:** The standard API query was
missing team folders. Injected the `corpora="allDrives"` and
`includeItemsFromAllDrives=True` override flags so the connector now
accurately maps state across both personal workspaces and organizational
Shared Drives.
### Testing:
Isolated the Google API retrieval logic locally to prove the `SlimDoc`
mapping works and correctly registers state drops when a file is trashed
remotely.
### Type of change
- [x] New Feature (non-breaking change which adds functionality)
- [x] Performance Improvement