Technical deep-dives about OndatraSQL. Learn about SQL data pipelines, DuckDB execution, DuckLake storage, and pipeline architecture.
OndatraSQL's tracked model kind uses content hashing to detect row changes and sync only modified data to external APIs for outbound sync.
What happens when you run ondatrasql run. The full sequence from model discovery and dependency resolution to DuckLake atomic commit.
How OndatraSQL combines SQL for transformation with Starlark for transport. Lib functions become table sources in FROM. One language for queries, one for the network.
OndatraSQL stores pipeline output in DuckLake — a catalog layer on top of Parquet files with snapshots, time-travel, and schema evolution.
OndatraSQL