Creating and Connecting Nodes

​

Overview

​

Creating nodes

​

Method 1: Ask Triton

• “Add a validation Action to this Flow”
• “Insert a sentiment analysis Agent after the input”
• “Add three Actions: fetch_data, process_data, and store_results”

• Triton creates the node
• Suggests appropriate connections
• Configures basic settings
• Maintains Flow structure

​

Method 2: Right-click menu

1. Right-click on empty Canvas space
2. Select Add Node
3. Choose node type:
   • Action — Select from your existing Actions or create new
   • Agent — Select from existing Agents or create new
   • Flow — Add a sub-Flow
   • Control — Router, Merge, Loop
4. Click to place on Canvas

​

Method 3: Drag from Agent Toolbox

1. Open Agent Toolbox panel (right side)
2. Browse or search for component
3. Drag onto Canvas
4. Drop where you want it

​

Method 4: Copy and paste

1. Select the node(s) you want to copy
2. Copy: Cmd+C (Mac) / Ctrl+C (Windows)
3. Paste: Cmd+V (Mac) / Ctrl+V (Windows)
4. The duplicate appears offset slightly

​

Method 5: Quick add (keyboard)

1. Press A (Add) or / (Command)
2. Type to search for component
3. Arrow keys to select
4. Enter to add to Canvas at current view center

​

Connecting nodes

​

Method 1: Click and drag

1. Hover over the source node’s output port (right side)
2. Click and hold on the port
3. Drag toward the destination node
4. Drop on the destination’s input port (left side)
5. Edge appears, connecting the nodes

• Compatible ports highlight green when hovering
• Incompatible ports highlight red
• Edge follows cursor while dragging

​

Method 2: Select and connect

1. Select the source node
2. Right-click → Connect to…
3. Choose the destination node from list
4. Select which output port → which input port
5. Click Connect

​

Method 3: Ask Triton

• “Connect validate_input to process_data”
• “Wire the output of fetch_user to the input of notify_user”
• “Connect all three fetch Actions to the merge node”

• Automatic port mapping
• Handles type conversions if needed
• Maintains data flow logic

​

Method 4: Auto-connect

1. Drag a new node onto an existing edge
2. Drop it on the edge line
3. The edge automatically splits and connects through the new node

Before: A → C
After:  A → B → C (B is the new node)

​

Connection validation

​

Compatible connections

• Type match: Output type equals input type
• Schema compatible: Fields align (or can be mapped)
• No cycles: Would not create an infinite loop (unless intentional)

​

Incompatible connections

• Type mismatch: String output → Number input
• Missing required fields: Output lacks required input fields
• Cycle detected: Would create unintended loop

• Port highlights red when hovering
• Cannot drop to create connection
• Error message explains why

​

Type conversion

• Source outputs: { "value": 123 }
• Destination expects: { "value": "string" }
• Triton offers to add a type conversion step

​

Mapping fields

​

Automatic mapping

{
  "user_id": "123",
  "user_name": "Alice"
}

{
  "user_id": "string",
  "name": "string"
}

• user_id → user_id ✓ (exact match)
• user_name → name (no match, needs manual mapping)

​

Manual mapping

1. Click the edge
2. Properties Panel shows Field Mapping
3. Map each destination field:
   • Select source field from dropdown
   • Or write a transformation expression
4. Click Save

• user_name → name (simple rename)
• first_name + " " + last_name → full_name (combine fields)
• price * 1.1 → price_with_tax (calculation)

​

Disconnecting nodes

​

Remove a single connection

1. Click the edge to select it
2. Delete: Press Delete or Backspace
3. Or right-click edge → Delete

​

Remove all connections from a node

1. Select the node
2. Right-click → Disconnect All
3. Confirm

​

Remove node and reconnect

1. Select the node
2. Right-click → Delete and Bridge
3. The node is removed and edges reconnect around it

Before: A → B → C
After:  A → C (B deleted, A now connects directly to C)

​

Advanced connection techniques

​

Conditional connections

1. Connect nodes normally
2. Click the edge
3. Properties Panel → Add Condition
4. Define condition: status == "approved"
5. Edge now labeled with condition

​

Named output ports

Process Node
├── success (output port)
├── error (output port)
└── warning (output port)

• success → Continue normal flow
• error → Error handling
• warning → Log and continue

​

Multiple inputs

Source A ↘
Source B → Merge → Destination
Source C ↗

• Wait for all: Waits for A, B, and C before proceeding
• First wins: Proceeds as soon as any input arrives
• Custom logic: Combine inputs with a function

​

Connection best practices

Connect left to right — Makes flow easy to follow

Avoid crossing edges — Rearrange nodes to minimize crossings

Use named ports — Clarifies purpose of each connection

Validate schemas — Ensure type compatibility before connecting

Test connections — Run executions to verify data flows correctly

​

Common connection patterns

​

Pattern 1: Linear connection

A → B → C → D

​

Pattern 2: Fan-out

A → B ↘
      C
  → D ↗

​

Pattern 3: Fan-in

A ↘
B → D
C ↗

​

Pattern 4: Error handling

A → [success] → B
  → [error] → Error Handler

​

Pattern 5: Loop

Input → Process → [not done] → Process (loop back)
                → [done] → Output

​

Troubleshooting connections

​

Problem: Can’t connect two nodes

• Type mismatch between ports
• Schema incompatibility
• Would create invalid cycle

• Check output/input types (hover over ports)
• Add transformation node between them
• Ask Triton: “Why can’t I connect A to B?”

​

Problem: Connection exists but data doesn’t flow

• Condition on edge isn’t met
• Upstream node failed
• Field mapping incorrect

• Check execution trace
• Verify edge condition
• Review field mapping

​

Problem: Too many crossing edges

• Rearrange nodes to reduce crossings
• Use auto-layout: Right-click → Auto Layout
• Split into sub-Flows

​

Next steps