Skip to main content

​
Overview

The @workflow decorator creates a root trace span. All nested @task, @agent, and @tool spans are captured as children under this workflow. 
from respan_tracing import workflow

​
Parameters

ParameterTypeDefaultDescription
namestr | NoneFunction nameDisplay name for the workflow span.
versionint | NoneNoneVersion number for the workflow.
method_namestr | NoneNoneRequired when decorating a class. Specifies which method to use as the entry point.
processorsstr | List[str] | NoneNoneRoute this span to specific named processors. See add_processor.

​
Function usage

from respan_tracing import RespanTelemetry, workflow, task

telemetry = RespanTelemetry(api_key="your-api-key")

@task(name="extract")
def extract():
    return {"records": [1, 2, 3]}

@task(name="transform")
def transform(data):
    return [x * 2 for x in data["records"]]

@workflow(name="data_pipeline")
def data_pipeline():
    data = extract()
    return transform(data)

print(data_pipeline())  # [2, 4, 6]

​
Class usage

When decorating a class, use method_name to specify the entry point. Calling that method creates the workflow span. 
from respan_tracing import RespanTelemetry, workflow, task
from openai import OpenAI

telemetry = RespanTelemetry(api_key="your-api-key")
client = OpenAI()

@workflow(name="analysis_workflow", method_name="run")
class Analyzer:
    @task(name="analyze")
    def analyze(self, nums):
        return sum(nums)

    def run(self):
        return self.analyze([1, 2, 3])

print(Analyzer().run())  # 6

​
Processor routing

Use processors to send workflow spans to specific exporters only. 
@workflow(name="debug_workflow", processors="debug")
def debug_workflow():
    return "only exported to the 'debug' processor"

@workflow(name="multi_export", processors=["debug", "analytics"])
def multi_export():
    return "exported to both 'debug' and 'analytics' processors"

​
Best practices

  • Use descriptive workflow names for easy navigation in the Traces dashboard
  • Keep workflows coarse-grained — use @task for internal steps
  • One workflow per user request or pipeline run