Mapping inputs and outputs

You can manually map the inputs and outputs of your nodes using the map_input() and map_output() methods. To do this, call map_input() or map_output() in your node and provide the following values:

Show `map_input()` Parameters

string

required

The variable you want to map.

string

required

The expression used to map the input.

string

required

The default value for the mapped input.

Show `map_output()` Parameters

string

required

The variable you want to map.

string

required

The expression used to map the output.

string

required

The default value for the mapped output.

The following example shows how to use map_input() and map_output() to map values for a node:

Python

'''
Build a simple hello world agentic workflow that will combine the result of two tools.
'''

from pydantic import BaseModel
from ibm_watsonx_orchestrate.flow_builder.flows import END, Flow, flow, START

from .get_hello_message import get_hello_message
from .combine_names import combine_names

class Name(BaseModel):
    """
    This class represents a person's name.

    Attributes:
        first_name (str): The person's first name.
        last_name (str): The person's last name.
    """
    first_name: str
    last_name: str

class Message(BaseModel):
    msg: str

@flow(
        name = "hello_message_flow",
        input_schema=Name,
        output_schema=Message
    )
def build_hello_message_flow_datamap(aflow: Flow = None) -> Flow:
    """
    Creates an agentic workflow with two tools: get_hello_message and combine_names.
    Args:
        flow (Flow, optional): The agentic workflow to be built. Defaults to None.
    Returns:
        Flow: The created agentic workflow.
    """
    combine_names_node = aflow.tool(combine_names)
    combine_names_node.map_input(input_variable="first_name", expression="flow.input.first_name")
    combine_names_node.map_input(input_variable="last_name", expression="flow.input.last_name", default_value="default_last_name")

    get_hello_message_node = aflow.tool(get_hello_message, output_schema=Message)

    aflow.edge(START, combine_names_node).edge(combine_names_node, get_hello_message_node).edge(get_hello_message_node, END)

flow.map_output(
    output_variable="Type_Justification",
    expression="flow.Type_Classifier_Node.output.justification_text if flow.Classifier_Node.output.justification_text else \"N/A\"")
             
    return aflow

Nodes use schemas to define the structure of incoming and outgoing data, ensuring consistent validation and predictable behavior. The following parameters describe how schemas and mapping functions control the flow of data through a node.

Show Schema definitions and data mapping Parameters

string

required

Defines the structure and types of the data a node expects to receive, acting as the contract for all required and optional inputs.

string

required

Binds incoming data to the node’s internal parameters using the input schema, optionally transforming or defaulting values.

string

required

Specifies the structured format and types of the data the node will return after execution.

string

required

Constructs the final output object according to the output schema by mapping internal execution results into the defined fields.

The following example shows how a node applies its schema definitions and mapping parameters to process inputs and generate a structured, validated output.

Python

class GEN_AI_NODE(BaseModel):
    description_original: str = Field(description="Formatted description ....", default="")

class JUSTIFICATION_TYPE_RESULT(BaseModel):
    grievance_type_llm: str = Field(description="Type of thing", default="")
    justification_text: str = Field(description="Justification for classified type", default="")

classifier_node = flow.prompt(
      name="Classifier_Node",
      display_name="Classifier_Node",
      description="Type Classifier",
      ...
      input_schema=GEN_AI_NODE,
      output_schema=JUSTIFICATION_TYPE_RESULT)

classifier_node.map_input(input_variable="description_original", expression="flow.Load_Extracted_Data_Node.output.extracted_kvps.Description_Original")

Release Notes

Get Started

Build

Deploy

Analyze

Developer experience

Legal notices