This variation keeps the Rust exporter and S3 publish path as the source of truth, then lets Databricks consume the published parquet snapshots.
flowchart LR
C[Convex]
X[convex-sync run]
S3[staging/current in S3]
UC[Unity Catalog external location]
V[stable SQL views]
C --> X
X --> S3
S3 --> UC
UC --> V
terraform/unity_catalog_s3_external_location/: storage credential, external location, grantssql/register_staging_views.sql.tmpl: stableread_files(...)view template
The published staging/current prefix must be covered by a Unity Catalog
external location before the view sync is applied.
Recommended entrypoint:
just databricks-sync-staging-views --warehouse-id <warehouse-id> --bucket <bucket> --prefix <prefix>