Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.definable.ai/llms.txt

Use this file to discover all available pages before exploring further.

Branch workflow execution based on a condition function that evaluates step context. 
from definable.agent import Agent
from definable.agent.workflow import Workflow, Step, Condition

reviewer = Agent(model="gpt-4o", instructions="Review content. End with APPROVED or REJECTED.")
publisher = Agent(model="gpt-4o", instructions="Format and publish the approved content.")
rewriter = Agent(model="gpt-4o", instructions="Rewrite and improve the rejected content.")

workflow = Workflow(
    name="review-pipeline",
    steps=[
        Step(name="draft", agent=drafter),
        Step(name="review", agent=reviewer),
        Condition(
            name="quality-gate",
            condition=lambda ctx: "APPROVED" in (ctx.get_last_step_content() or ""),
            true_steps=Step(name="publish", agent=publisher),
            false_steps=Step(name="rewrite", agent=rewriter),
        ),
    ],
)

​
How It Works

  1. draft and review steps execute sequentially.
  2. The condition function receives the step context.
  3. If it returns True, the true_steps branch executes.
  4. If it returns False, the false_steps branch executes.

​
Multiple True/False Steps

Both branches accept a single step or a list: 
Condition(
    name="gate",
    condition=lambda ctx: score > 7,
    true_steps=[
        Step(name="polish", agent=editor),
        Step(name="publish", agent=publisher),
    ],
    false_steps=Step(name="rewrite", agent=writer),
)

​
Async Conditions

async def check_quality(ctx):
    content = ctx.get_last_step_content()
    # Call an external API, run evaluation, etc.
    return "APPROVED" in content

Condition(name="gate", condition=check_quality, ...)