retrack 1.0.0a1

A business rules engine

A business rules engine

Installation

pip install retrack

Usage

import retrack

runner = retrack.Runner.from_json("your-rule.json")

response = runner.execute(input_data)

Or, if you want to create the parser and runner manually:

import retrack

# Parse the rule/model
parser = retrack.Parser(rule)

# Create a runner
runner = retrack.Runner(parser, name="your-rule")

# Run the rule/model passing the data
runner.execute(data)

The Parser class parses the rule/model and creates a graph of nodes. The Runner class runs the rule/model using the data passed to the runner. The data is a dictionary or a list of dictionaries containing the data that will be used to evaluate the conditions and execute the actions. To see wich data is required for the given rule/model, check the runner.request_model property that is a pydantic model used to validate the data.

Optionally you can name the rule by passing the name field to the retrack.Runner constructor. This is useful to identify the rule when exceptions are raised.

Creating a rule/model

A rule is a set of conditions and actions that are executed when the conditions are met. The conditions are evaluated using the data passed to the runner. The actions are executed when the conditions are met.

Each rule is composed of many nodes. To see each node type, check the nodes folder.

To create a rule, you need to create a JSON file with the following structure:

{
  "nodes": {
		"node id": {
			"id": "node id",
			"data": {},
			"inputs": {},
			"outputs": {},
			"name": "node name",
		},
    // ... more nodes
  }
}

The nodes key is a dictionary of nodes. Each node has the following properties:

• id: The node id. This is used to reference the node in the inputs and outputs properties.
• data: The node data. This is used as a metadata for the node.
• inputs: The node inputs. This is used to reference the node inputs.
• outputs: The node outputs. This is used to reference the node outputs.
• name: The node name. This is used to define the node type.

The inputs and outputs properties are dictionaries of node connections. Each connection has the following properties:

• node: The node id that is connected to the current node.
• input: The input name of the connection that is connected to the current node. This is only used in the inputs property.
• output: The output name of the connection that is connected to the current node. This is only used in the outputs property.

To see some examples, check the examples folder.

Creating a custom node

To create a custom node, you need to create a class that inherits from the BaseNode class. Each node is a pydantic model, so you can use pydantic features to create your custom node. To see the available features, check the pydantic documentation.

To create a custom node you need to define the inputs and outputs of the node. To do this, you need to define the inputs and outputs class attributes. Let's see an example of a custom node that has two inputs, sum them and return the result:

import retrack
import pydantic
import pandas as pd
import typing


class SumInputsModel(pydantic.BaseModel):
    input_value_0: retrack.InputConnectionModel
    input_value_1: retrack.InputConnectionModel


class SumOutputsModel(pydantic.BaseModel):
    output_value: retrack.OutputConnectionModel


class SumNode(retrack.BaseNode):
    inputs: SumInputsModel
    outputs: SumOutputsModel

    def run(self, input_value_0: pd.Series,
        input_value_1: pd.Series,
    ) -> typing.Dict[str, pd.Series]:
        output_value = input_value_0.astype(float) + input_value_1.astype(float)
        return {
            "output_value": output_value,
        }

After creating the custom node, you need to register it in the nodes registry and pass the registry to the parser. Let's see an example:

import retrack

# Register the custom node
retrack.component_registry.register_node("sum", SumNode)

# Parse the rule/model
parser = Parser(rule, component_registry=retrack.component_registry)

Contributing

Contributions are welcome! Please read the contributing guidelines first.