The code is on GitHub.
Tools we built that are useful on their own. Issues and pull requests welcome.
A notebook your coding agent can drive, and that keeps everything it computes.
Cells are cached by their source, inputs and environment, so identical work is a cache hit. Python, R, SQL and prompt cells share one graph, and a notebook is plain .py files and a TOML manifest.
$ uv tool install "strata-notebook[mcp,tui]"Run the tests your change can affect, and see why each was picked.
Follows changed functions and classes back to the pytest tests and ASV benchmarks that can observe them, without running your code.
$ uv tool install diffconeCompile-time Spark encoders for Scala 3.
Derives Spark's own encoders at compile time without scala.reflect.runtime, a step toward running Spark on Scala 3.
A Python-to-Java bridge over shared memory.
Zero-copy Apache Arrow transfer, sync and async clients, and a PySpark adapter.
Lazy loading and version checks for optional dependencies.
A decorator declares what a class or function needs, imports it on first use and checks the installed version.
$ pip install optional-dependency-manager