Feldera Python is the Feldera SDK for Python developers.

uv pip install feldera

uv pip install git+https://github.com/feldera/feldera#subdirectory=python

Similarly, to install from a specific branch:

uv pip install git+https://github.com/feldera/feldera@{BRANCH_NAME}#subdirectory=python

Replace {BRANCH_NAME} with the name of the branch you want to install from.

If you have cloned the Feldera repo, you can install the python SDK as follows:

# the Feldera Python SDK is present inside the python/ directory
cd python
# If you don't have a virtual environment, create one
uv venv
source .venv/activate
# Install the SDK in editable mode
uv pip install .

The Python SDK documentation is available at Feldera Python SDK Docs.

To build the html documentation run:

Ensure that you have sphinx installed. If not, install it using uv pip install sphinx.

Then run the following commands:

cd docs
sphinx-apidoc -o . ../feldera
make html

To clean the build, run make clean.

To run unit tests:

cd python && python3 -m pytest tests/

• This will detect and run all test files that match the pattern test_*.py or *_test.py.
• By default, the tests expect a running Feldera instance at http://localhost:8080. To override the default endpoint, set the FELDERA_HOST environment variable.

To run tests from a specific file:

(cd python && python3 -m pytest ./tests/path-to-file.py)

To run a specific test:

uv run python -m pytest tests/test_shared_pipeline.py::TestPipeline::test_adhoc_query_hash -v

The tests validate end-to-end correctness of SQL functionality. To run the tests use:

cd python
PYTHONPATH=`pwd` ./tests/run-all-tests.sh

To reduce redundant compilation cycles during testing:

• Inherit from SharedTestPipeline instead of unittest.TestCase.
• Define DDLs (e.g., CREATE TABLE, CREATE VIEW) in the docstring of each test method.
  • All DDLs from all test functions in the class are combined and compiled into a single pipeline.
  • If a table or view is already defined in one test, it can be used directly in others without redefinition.
  • Ensure that all table and view names are unique within the class.
• Use @enterprise_only on tests that require Enterprise features. Their DDLs will be skipped on OSS builds.
• Use self.set_runtime_config(...) to override the default pipeline config.
  • Reset it at the end using self.reset_runtime_config().
• Access the shared pipeline via self.pipeline.

from tests.shared_test_pipeline import SharedTestPipeline

class TestAverage(SharedTestPipeline):
    def test_average(self):
        """
        CREATE TABLE students(id INT, name STRING);
        CREATE MATERIALIZED VIEW v AS SELECT * FROM students;
        """
        ...
        self.pipeline.start()
        self.pipeline.input_pandas("students", df)
        self.pipeline.wait_for_completion(True)
        ...

Use Ruff to run the lint checks that will be executed by the precommit hook when a PR is submitted:

ruff check python/

To reformat the code in the same way as the precommit hook:

ruff format