Install

Get the runtime, the CLI, and (optionally) the data-frame interop extras.

Stable (v1.0.0). The public surface (decorator names, config keys, CLI flags) is under semantic versioning — no breaking changes without a major bump.

Core install

pip install "ematix-flow==0.11.0"

That gives you every backend, the flow CLI binary, and the run_pipeline / run_streaming_pipeline Python entrypoints. No extra services to operate.

Optional extras

ExtraWhat it addsInstall
dfDataFrame interop helpers (polars / pandas) for to_polars() / to_pandas() materialization.pip install "ematix-flow[df]" then pip install polars (or pandas).
sparkPySpark interop helpers (to_pyspark() / from_pyspark()). Heavy — pulls in PySpark + JDBC.pip install "ematix-flow[spark]"
pyarrowRequired for streaming-backend pyclass wrappers when iterating batch-by-batch in Python.pip install pyarrow

The flow binary, run_pipeline, and the typed-Python streaming API work without any extras.

Verify

flow --version
flow connections list