Nodes reference
This page describes the node types currently used in the workflow builder.
Input
Use the Input node to define what the workflow needs at runtime.
What you can configure:
- Input parameters (fields users fill in before running)
- A prompt template (shared text you can reuse by copying placeholders)
Common pattern:
- Input → Task → Output
Task
A Task node represents one execution step.
Typical configuration:
- A prompt (supports
{{variable}}placeholders) - Optional selection of sources/tools/actions (depending on your environment)
How execution order works:
- Connect nodes with edges to define the path.
- The first task connected after Input is treated as the starting task.
- Connecting the final task to Output ends the chain.
Output
The Output node is the terminal node.
Use Output to:
- Make the intended end of the workflow explicit
- Keep the graph readable (especially for longer chains)
Sticky note
Sticky notes are for documentation and collaboration.
- They do not execute.
- Use them to explain intent (e.g., “Step 2 formats the response as Markdown”).
Tips for clean graphs
- Use one clear path from Input to Output for predictable execution.
- Name tasks based on outcomes (e.g., “Extract fields”, “Summarize”, “Draft reply”).
- Keep prompts short and specific; move repeated boilerplate into the Input prompt template.