Branching logic and conditions

What you will do

Add branching logic to a workflow so it takes different paths depending on what the user provides or what the AI returns.

Why branching matters

A linear workflow (input → process → output) handles one scenario. Real tasks have conditions: if the email is urgent, escalate it; if the lead score is high, send a personalised response; if the input is unclear, ask a follow-up question.

Add a routing block

MindStudio provides routing blocks that evaluate a condition and send the workflow down one of multiple paths.

1. After a Generate Text block that classifies the input, add a Router block.
2. Define the conditions. For example: if the classification variable equals "bug", go to Path A. If it equals "feature_request", go to Path B. Otherwise, go to a default path.
3. Each path continues with its own blocks (different prompts, different tools, different outputs).

Use variables for conditions

Variables created in earlier blocks are available in later blocks. A common pattern:

1. User Input block creates email_text.
2. Generate Text block classifies the email and stores the result in category.
3. Router block checks category and branches.
4. Each branch has a Generate Text block with a prompt specific to that category.

Add a loop for unclear inputs

If the classification is "unclear," you can route back to a User Input block that asks the user to provide more context. This creates a loop that only exits when the input is clear enough to classify.

Keep loops simple. A workflow that can loop indefinitely is a workflow that will eventually confuse the user. Add a maximum iteration count or a fallback path that exits after two attempts.

Test each path

When testing, provide inputs that should trigger each branch. Verify that the workflow follows the correct path and produces the right output for each scenario.

What you should see

A workflow with multiple paths that respond differently to different inputs. The canvas should visually show the branches diverging and (optionally) reconverging at the end.