Future Features
Planned features that are out of current scope.
Cloud storage for DuckLake
- Azure Blob Storage: configure DuckDB's
azureextension to point DuckLake at an Azure Blob container - S3-compatible storage: configure DuckDB's
httpfsextension for any S3-compatible endpoint
Large dataset writes
- DuckDBPyRelation support: allow assets to return a lazy DuckDB relation instead of a materialized
pl.DataFrame, enabling zero-copy writes of arbitrary size
Security
- Fine-grained access control: authentication and authorization at the transaction level (per-query, per-write)
- Activity logging: audit trail of all data access and mutations for compliance and debugging
Observability
- Prometheus metrics:
/metricsendpoint with Grafana dashboards for multi-station deployments and alerting
Data integrity
- Catalog consistency check: verify every referenced Parquet file exists in storage (fsck for DuckLake)
- Stale asset reconciliation: mark Dagster assets as stale when backing files are missing or corrupted
Data formats
- Binary files (DICOM, NIfTI): file-reference pattern with metadata in catalog, binary in object storage
- ML artifacts (pickle, safetensor): same pattern, potentially with versioning