Context
Storage/tree adapters are likely candidates for lazy materialization, but their economics are different from issue/chat/content records. Before adding config, decide whether the shared materialization policy from #194 maps cleanly to storage trees or whether they need a smaller provider-specific sync scope model.
Umbrella: #195
Candidate adapters
- Google Drive
- OneDrive
- SharePoint
- Box
- Dropbox
- S3
- GCS
- Azure Blob
Design questions
- What is the target key: account, drive, site, bucket, container, folder prefix, or some combination?
- What are the resources: metadata, file content, folders, permissions, comments, versions?
- Should lazy mean metadata-only, index-only, no content bodies, or no child traversal?
- Can provider APIs apply prefix/path, modified_since, MIME type, size, or folder filters cheaply?
- How should webhook/delta events behave for lazy targets?
Acceptance criteria
- Produce a short design note or adapter README update explaining the chosen model.
- If implemented, use adapter-core shared materialization utilities for target/resource rule resolution.
- Preserve current eager/default behavior.
- Add tests for folder/prefix target matching, metadata-only/content lazy behavior, modified_since cursor behavior, and webhook/delta handling.
- If not implemented for a provider, document why materialization policy is not a good fit.
Context
Storage/tree adapters are likely candidates for lazy materialization, but their economics are different from issue/chat/content records. Before adding config, decide whether the shared materialization policy from #194 maps cleanly to storage trees or whether they need a smaller provider-specific sync scope model.
Umbrella: #195
Candidate adapters
Design questions
Acceptance criteria