Branching and Conditional Routing

Where we left off

Every flow so far has been one straight line, or one line with an Agent deciding whether to make a side trip to a tool. The Conditional Router component branches the line itself: one input, a condition, two possible paths, and only one of them actually runs. If you've done langgraph, this is the direct Langflow equivalent of a conditional edge.

Do this yourself

1. Add Chat Input, a Conditional Router (search "conditional" in the sidebar), and two Chat Output components.
2. Wire Chat Input into the router's Text Input field. Set Operator to contains, Match Text to urgent, leave Case Sensitive off.
3. Fill in True Case Message ("Escalating to a human, priority queue.") and False Case Message ("Logged, we'll get to it in the usual order.").

4. Wire the router's True Result output into one Chat Output, its False Result output into the other. Compare to this screenshot:

   
5. In the Playground, send "this is urgent, my payment failed" and watch only the True-branch Chat Output produce a message, run it again with "just checking in, no rush" and watch the other branch fire instead. The path that doesn't match never runs at all, it's not that its message is empty, it's skipped entirely.

The code, piece by piece

router.set(
    input_text=chat_input.message_response,
    operator="contains",
    match_text="urgent",
    true_case_message=Message(text="Escalating to a human, priority queue."),
    false_case_message=Message(text="Logged, we'll get to it in the usual order."),
)

operator supports more than contains, equals, starts/ends with, regex, and numeric comparisons are all options, useful for routing on things other than free text. The two *_case_message fields are what get forwarded down whichever path is taken.

graph = Graph()
graph.add_component(chat_input)
graph.add_component(router)
graph.add_component(urgent_output)
graph.add_component(normal_output)

Every earlier lesson used Graph(start=..., end=...), a shortcut that only works for a single start and a single end. A branching graph has two possible ends, so this lesson builds it the more general way, adding each component individually, the edges themselves are still inferred from the .set() calls above, exactly like every other lesson.

for run_output in results:
    for out in run_output.outputs:
        print(f"  -> {out.component_display_name}: {out.messages[0].message}")

Only the taken branch's Chat Output shows up in results at all, run this and count the lines, one "Input" line, then exactly one -> line underneath it, never two.