Building Workflows

Workflows are a powerful way to orchestrate your agents and teams. They are a series of steps that are executed in a flow that you control.

Building Blocks

1. The Workflow class is the top-level orchestrator that manages the entire execution process.
2. Step is the fundamental unit of work in the workflow system. Each step encapsulates exactly one executor: an Agent, a Team, a custom Python function, or a nested Workflow. This design ensures clarity and maintainability while preserving the individual characteristics of each executor.
3. Loop is a construct that allows you to execute one or more steps multiple times. This is useful when you need to repeat a set of steps until a certain condition is met.
4. Parallel is a construct that allows you to execute one or more steps in parallel. This is useful when you need to execute a set of steps concurrently with the outputs joined together.
5. Condition makes a step conditional based on criteria you specify.
6. Router allows you to specify which step(s) to execute next, effectively creating branching logic in your workflow.

How to make your first workflow?

There are different types of patterns you can use to build your workflows. For example you can combine agents, teams, and functions to build a workflow.

1from kern.workflow import Step, Workflow, StepOutput
2
3def data_preprocessor(step_input):
4    # Custom preprocessing logic
5
6    # Or you can also run any agent/team over here itself
7    # response = some_agent.run(...)
8    return StepOutput(content=f"Processed: {step_input.input}") # <-- Now pass the agent/team response in content here
9
10workflow = Workflow(
11    name="Mixed Execution Pipeline",
12    steps=[
13        research_team,      # Team
14        data_preprocessor,  # Function
15        content_agent,      # Agent
16    ]
17)
18
19workflow.print_response("Analyze the competitive landscape for fintech startups", markdown=True)