Parallel Workflow

Example Use-Cases: Multi-source research, parallel analysis, concurrent data processing

Parallel workflows maintain deterministic results while dramatically reducing execution time for independent operations.

Example

1from kern.workflow import Parallel, Step, Workflow
2
3workflow = Workflow(
4    name="Parallel Research Pipeline",
5    steps=[
6        Parallel(
7            Step(name="HackerNews Research", agent=hn_researcher),
8            Step(name="Web Research", agent=web_researcher),
9            Step(name="Academic Research", agent=academic_researcher),
10            name="Research Step"
11        ),
12        Step(name="Synthesis", agent=synthesizer),  # Combines the results and produces a report
13    ]
14)
15
16workflow.print_response("Write about the latest AI developments", markdown=True)

Handling Session State Data in Parallel Steps

When using custom Python functions in your steps, you can access and update the Worfklow session state via the run_context parameter.

If you are performing session state updates in Parallel Steps, be aware that concurrent access to shared state will require coordination to avoid race conditions.

Developer Resources

• Parallel Steps Workflow

Reference

For complete API documentation, see Parallel Steps Reference.