Skip to main content

​
Overview

The @task decorator creates a child span for a single unit of work — an LLM call, data processing step, validation, or any discrete operation within a workflow. 
from respan_tracing import task

​
Parameters

ParameterTypeDefaultDescription
namestr | NoneFunction nameDisplay name for the task span.
versionint | NoneNoneVersion number for the task.
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
from openai import OpenAI

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

@task(name="generate_text")
def generate_text(prompt: str):
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": prompt}],
    )
    return response.choices[0].message.content

@task(name="summarize")
def summarize(text: str):
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": f"Summarize: {text}"}],
    )
    return response.choices[0].message.content

@workflow(name="content_pipeline")
def content_pipeline(topic: str):
    text = generate_text(f"Write about {topic}")
    return summarize(text)

print(content_pipeline("Python"))

​
Class usage

from respan_tracing import RespanTelemetry, workflow, task

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

@workflow(name="agent_run", method_name="run")
class Agent:
    @task(name="respond")
    def respond(self, prompt: str):
        return f"Echo: {prompt}"

    def run(self):
        return self.respond("Hello")

print(Agent().run())

​
Features

  • Automatic I/O capture — Function arguments and return values are serialized as span input/output (up to 1MB).
  • Exception recording — Exceptions are automatically recorded on the span with error status.
  • Async support — Works with async def functions and generators.

​
Best practices

  • Keep tasks small and focused on a single operation
  • Use task names that reflect intent, not implementation details
  • Nest tasks inside workflows for proper trace hierarchy