When to use
- Classify images by subject and route each class to a different generator
- Categorize customer feedback into support, sales, or bug channels
- Branch a workflow into format-specific tracks (square, vertical, landscape)
- Pick a prompt strategy based on the input asset type
Inputs
Extra context the model may consider when matching cases: text, images, or any mix. This context is not routed. Optional, accepts multiple inputs and an array.
Outputs
Configuration
The list of cases. Each case has a
label (used to name its output port) and a condition (a natural-language rule the model evaluates). Add up to 16 cases. Every case must have a non-empty label and condition.Switch runs on a selectable model (default Gemini 2.5 Flash) to evaluate conditions and consumes credits when it runs.