Testing & Preview

The flow builder includes a built-in conversation simulator that lets you test your bot without deploying it. You can have a live conversation with your flow, watch execution move through nodes on the canvas, inspect variables in real time, and debug issues before going to production.

Opening the Test Panel

Open the conversation simulator in any of these ways:

• Click the Test button (play icon) in the top toolbar.
• Press Ctrl+T on your keyboard.
• Right-click the canvas and select Test Flow.

The test panel opens on the right side of the flow builder with a chat interface where you interact with your bot exactly as an end user would.

imageTest panel open on the right side of the flow builder showing a chat conversation with user messages and bot responses

The test panel with a live conversation

WARNING

Test conversations consume API calls against your configured AI provider. If your provider has rate limits or usage caps, test conversations count toward those limits.

Execution Path Highlighting

As the test conversation progresses, the canvas highlights the path that execution takes through your flow:

Visual Indicator	Meaning
Green border / dot	Node executed successfully
Red border / dot	Node encountered an error during execution
Animated edge	The connection currently being traversed
Dimmed nodes	Nodes that were not reached in this conversation

This visual trace makes it easy to confirm that your flow is following the expected path. If execution takes an unexpected branch, you can immediately see which node made the routing decision.

TIP

After a test conversation, the execution path remains highlighted on the canvas. Click Reset in the test panel to clear the highlights and start fresh.

imageFlow builder canvas with green-highlighted nodes and animated edges showing the execution path taken during a test conversation, with dimmed unreached nodes

Execution path highlighting on the canvas during testing

Variable Inspector

The Variables tab in the test panel shows all session variables and their current values, updating in real time as nodes execute.

Column	Description
Name	The variable name (e.g., user_name, llm_response)
Value	The current value, formatted as a string, number, or expandable JSON object
Set By	Which node last set this variable
Type	The data type (string, number, boolean, array, object)

System variables (prefixed with sys_) are also visible. During testing, sys_channel is set to "test" and sys_conversation_id is generated fresh for each session. For variables containing objects or arrays, click the row to expand it and view the full JSON structure with syntax highlighting.

Debug Mode

Toggle Debug Mode in the test panel header to see detailed execution information beneath each bot response.

Node Execution Log

A collapsible section under each response lists every node that executed, in order:

1. start_1 (Start)           - 0ms
2. llm_1  (LLM Response)     - 1,243ms
3. cond_1 (Condition)         - 2ms    → matched rule: "tier = premium"
4. msg_2  (Send Message)      - 1ms

The log shows the node ID and type, execution time in milliseconds, routing decisions for conditional nodes, and error details if a node failed.

imageDebug mode panel showing a collapsible node execution log with node IDs, types, execution times, and routing decisions listed in order

Debug mode with the node execution log expanded

LLM Debug Details

For LLM Response and KB Search nodes, debug mode shows the full system prompt (with variables resolved), model name and parameters, token usage (prompt, completion, total), knowledge base results with relevance scores, and tool calls made in agentic mode.

API Call Debug Details

For API Call nodes, debug mode shows the request URL, method, headers, request body (with variables resolved), response status code and body, and which variables were set via response mapping.

Resetting the Conversation

Click the Reset button (circular arrow icon) or press Ctrl+Shift+R to clear all chat messages, reset session variables to defaults, clear execution path highlights, and start a new session.

TIP

Reset frequently between test scenarios. Variables from a previous test can carry over and cause unexpected behavior if you do not clear them.

Testing Different Scenarios

To thoroughly test your flow, cover these areas:

• Happy path: Send messages that follow the intended conversation flow and verify each node executes correctly.
• Edge cases: Send empty input, very long text, special characters, and unrelated messages.
• Conditional branches: For every Condition, Intent Router, and IVR Menu node, send inputs that exercise each branch and confirm routing via the debug log.
• Multi-turn conversations: Verify that variables persist across turns, loops exit correctly, and the bot maintains context.
• Error scenarios: Test with a missing integration or unreachable API endpoint to verify error handling.

Channel Simulation

The test panel simulates a web chat channel. Voice-specific nodes (Voice Prompt, Voice Input, IVR Menu) fall back to text-based interaction -- DTMF options appear as clickable buttons. Rich content elements render in a simplified preview, and audio input/output uses text equivalents.

Keyboard Shortcuts

Shortcut	Action
Ctrl+T	Toggle test panel open/closed
Ctrl+Shift+R	Reset test conversation
Enter	Send message
Shift+Enter	New line in message input
Escape	Close test panel

Best Practices

• Test early and often: Run the simulator after every significant change to catch issues before they compound.
• Enable debug mode for new flows to understand exactly how execution proceeds.
• Check variable values after each node to verify data flows correctly between nodes.
• Test all branches: Do not assume conditional paths work -- verify each one with the appropriate input.
• Save before testing: Saving first ensures you have a checkpoint to return to if you make changes during testing.