Lance has metadata and index caches, but right now we just use the default LanceSession.
We should expose config for cache sizes so we can tune remote LanceDB tables better, especially when tables are on S3 / GCS.
Probably useful settings:
- index cache size
- metadata cache size
- whether cache is shared across execution runs
- cache scope: app_id / user_id / tenant_id / none
Defaults should be safe for multi-tenant usage. App level isolation seems like the best default.
Lance has metadata and index caches, but right now we just use the default LanceSession.
We should expose config for cache sizes so we can tune remote LanceDB tables better, especially when tables are on S3 / GCS.
Probably useful settings:
Defaults should be safe for multi-tenant usage. App level isolation seems like the best default.