What is an Edge?
An edge is a connection between two nodes that defines how data flows and when nodes execute. Edges are the arrows you see on the canvas connecting one node to another.Execution Flow
Edges determine the order in which nodes execute
Data Transfer
Output from one node becomes input to the next
Conditional Routing
Switch nodes use edges to route to different paths
Special Connections
Tool and memory edges create specialized relationships
Edge Types
Parallel Edges
Default execution flowStandard connections for concurrent node processing
Tool Edges
LLM-to-Tool connectionsEnable AI function calling and tool use
Memory Edges
Stateful conversationsConnect memory nodes to LLMs for chat history
Conditional Edges
IF/ELIF/ELSE branchingRoute execution based on switch node conditions
Error Edges
Failure handlingGracefully handle node execution errors
Edge Anatomy
Handles
Nodes connect through handles - connection points that define how edges attach:- Input Handles
- Output Handles
- Special Handles
Left side of nodes
- Receive incoming execution flow
- Accept data from previous nodes
- Trigger node execution when activated
Edge Properties
Each edge has:- Source Node + Handle: Where the edge starts
- Target Node + Handle: Where the edge ends
- Edge Type: Parallel, tool, memory, conditional, or error
- Visual Style: Different colors/patterns for different types
How Edges Work
1
Execution Order
Edges define which nodes run first and what runs next
2
Data Flow
Output from source node is passed as input to target node
3
Parallel Processing
Multiple edges from one node trigger concurrent execution
4
Conditional Routing
Switch nodes evaluate conditions and activate specific edges
5
Error Handling
Failed nodes can route to error edges instead of parallel paths
Creating Edges
Via Canvas
Via Canvas
Drag from output handle to input handle
- Click and drag from a source node’s output handle
- Drag to the target node’s input handle
- Release to create the connection
Special Edges
Special Edges
Tool and memory edges require specific nodes
- Tool Edges: Connect LLM’s TOOLS handle to Tool node input
- Memory Edges: Connect Memory node (top) to LLM’s MEMORY handle (bottom)
Validation
Validation
Splox prevents invalid connections
- Can’t create cycles (except in subflows)
- Can’t connect incompatible handle types
- Some nodes have handle restrictions
Best Practices
Parallel When Possible
Use parallel edges for concurrent execution to improve speed
Always Handle Errors
Connect error edges to prevent silent failures
Keep Flows Simple
Avoid overly complex branching - use subflows instead
Use Merge Nodes
Combine parallel paths before continuing workflow
Common Patterns
- Fan-Out, Fan-In
- Conditional Processing
- AI Agent with Tools
- Error Recovery
Split execution, then recombineUse parallel edges to fan out, merge node to fan in.
Detailed Documentation
Parallel Edges
Standard execution flow for concurrent processing
Tool Edges
Connect LLMs to tools for function calling
Memory Edges
Enable stateful conversations with chat history
Conditional Edges
Route execution based on conditions
Error Edges
Handle node execution failures gracefully