Pipeline Without Agent

This example demonstrates that a Digital Employee pipeline can contain any kind of steps — not just agent steps. A pipeline can be composed of simple components such as a log step, custom processing steps, or any other step from the gllm_pipeline ecosystem.

In this example, we build a pipeline that uses only a HelloWorldComponent step. No DigitalEmployeeAgentBuilder is involved.

Key takeaway: A pipeline is flexible. You can use add_step() to add any step type — from simple components to full agent steps. This example shows the minimal case.

Prerequisites

• Install the digital-employee-core package

• Set any required environment variables (e.g., for load_dotenv())

Example: Simple Hello World Pipeline

This example demonstrates a simple custom component that prepends "Hello World" to any input message. It shows how to create a basic pipeline step without using an agent, LLM, or MCP integration.

Import the Package

import asyncio
from typing import Any

from dotenv import load_dotenv
from gllm_core.schema import Component
from gllm_pipeline.steps import step

from digital_employee_core import DigitalEmployeeBuilder, DigitalEmployeePipelineBuilder

load_dotenv()

Define a Simple Component

class HelloWorldComponent(Component):
    """Component that adds 'Hello World' prefix to the input message."""

    def __init__(self) -> None:
        super().__init__()

    async def _run(self, message: str, **kwargs: Any) -> str:
        return "Hello World " + message

Build the Pipeline Without Agent

digital_employee = (
    DigitalEmployeeBuilder()
    .set_name("DE Hello World")
    .set_email("[e-mail]")
    .set_job(
        title="Digital Assistant",
        description="A helpful digital employee assistant",
        instruction="You are a helpful digital employee assistant. Answer questions clearly and concisely.",
    )
    .set_pipeline(
        DigitalEmployeePipelineBuilder()
        .add_step(
            step(
                component=HelloWorldComponent(),
                input_state_map={"message": "message"},
                output_state="message",
                name="HelloWorld",
            )
        )
    )
    .build()
)

Notice that the pipeline uses add_step() with a step() from gllm_pipeline.steps — no DigitalEmployeeAgentBuilder is used. The pipeline is composed of a single, simple component step.

Invoke the Pipeline

message = "how are you?"

response = asyncio.run(digital_employee.invoke(message=message))

print(response)

Summary

• A pipeline can contain any kind of steps — not limited to agent steps.

• Use step() from gllm_pipeline.steps to wrap components.

• Use add_step() on DigitalEmployeePipelineBuilder to add steps.

• For agent-based pipelines with MCPs and LLMs, see Pipeline With Agent.