trac-runtime 0.3.1

TRAC Model Runtime for Python

TRAC Model Runtime for Python

TRAC is a next-generation data and analytics platform for use in highly regulated environments

The TRAC model runtime provides all the APIs needed to write models for the TRAC platform. It includes an implementation of the runtime library that can be used as a development sandbox, so you can run and debug TRAC models right away from your favourite IDE or notebook. A number of tools are included to make it easy to plug in development data and other settings. When your models are ready they can be loaded into a real instance of TRAC for testing and eventual deployment.

Documentation for the TRAC platform is available at trac-platform.readthedocs.io.

Requirements

The TRAC runtime for Python has these requirements:

• Python: 3.7 or later
• Pandas: 1.0 or later
• PySpark 2.4.x or 3.x

Not every combination of versions will work, e.g. PySpark 3 requires Python 3.8.

Installing the runtime

The TRAC runtime package can be installed directly from PyPI:

pip install trac-runtime

The TRAC runtime depends on Pandas and PySpark, so these libraries will be pulled in as dependencies. If you want to target particular versions, install them explicitly first.

Writing a model

Once the runtime is installed you can write your first TRAC model! Start by inheriting the TracModel base class, your IDE should be able to generate stubs for you:

import trac.rt.api as trac

class SampleModel(trac.TracModel):

    def define_parameters(self) -> tp.Dict[str, trac.ModelParameter]:
        pass

    def define_inputs(self) -> tp.Dict[str, trac.ModelInputSchema]:
        pass

    def define_outputs(self) -> tp.Dict[str, trac.ModelOutputSchema]:
        pass

    def run_model(self, ctx: trac.TracContext):
        pass

You can fill in the three define_* methods to declare any parameters, inputs and outputs your model is going to need, then start writing your model code in run_model.

To learn about modelling with TRAC and what is possible, check out the modelling tutorials available in our online documentation. The tutorials are based on example models in the TRAC GitHub repository. We run these examples as part of our CI, so they will always be in sync with the corresponding version of the runtime library.

Building the runtime from source

This is not normally necessary for model development, but if you want to do it here are the commands.

cd trac-runtime/python

# Configure a Python environment

python -m venv ./venv
venv\Scripts\activate              # For Windows platforms
. venv/bin/activate                # For macOS or Linux
pip install -r requirements.txt

# Run the code generator

python ../../dev/codegen/protoc-ctrl.py python_runtime \
    --proto_path trac-api/trac-metadata/src/main/proto \
    --out trac-runtime/python/generated/trac/rt_gen/domain

# Build the Python package files

python ./package-ctrl.py

The package files will appear under build/dist