Overview
The Workflow Editor is a visual node-based tool for designing conversation logic. Drag, drop, and connect nodes to create complex conversation flows.
Node Types
Start Node
The entry point for every workflow. Each workflow has exactly one Start node.
LLM Node
Sends a message to the configured AI model and returns the response.
Settings:
- Temperature (0.0 - 1.0)
- Max tokens
- Custom system prompt override
Condition Node
Branches the workflow based on conditions. Supports:
- Keyword matching
- Sentiment analysis
- Custom expressions
Executes an external function or API call. Results are passed back to the conversation context.
Voice Call Node
Handles voice-specific interactions — speech recognition and synthesis.
End Node
Terminates the conversation with an optional closing message.
Connecting Nodes
Click on a node’s output port and drag to another node’s input port to create a connection. Each connection represents a possible conversation path.
Best Practices
Always build explicit paths between nodes with clear conditions. However, in some cases the AI may autonomously route to a node that has no direct edge — this is called a synapse. Synapses occur when the AI determines that existing routing conditions are unclear or insufficient and decides to reach a node on its own. While this provides flexibility, frequent synapses usually indicate that your edge conditions need to be more specific.
- Keep workflows simple — complex flows are harder to debug
- Use Condition nodes to handle edge cases
- Test each path individually before publishing
