For the complete documentation index, see llms.txt. This page is also available as Markdown.

Create flow

This page describes how to configure the flow settings, build workflow logic, validate the logic and test the interaction. To learn how to deploy the flow to a voice or chat channel see Deploy a flow.

For a detailed step-by-step guide of building a flow, see Build, validate and test a simple flow.

Before you begin

  • Confirm the primary goal of your flow, such as guiding users through a support process or collecting user information.

  • Identify and configure any necessary external systems your flow should interact, like REST APIs.

Step 1: Configure the settings

  1. Sign in to ixHello Customer v2.

  2. In the left navigation menu, go to Flows

  3. Select Create flow to configure the flow settings and click Save to continue to the next step.

Setting
Description

Name

The meaningful name for the Flow that clearly reflects its purpose.

Description

A description to provide further information.

Persona. Text-to-Speech (TTS) language and voice

The language and voice for TTS prompts, used on the voice channel.

Language

This specifies the language and regional variant the AI should use for speaking and writing.

Voice

This is the text‑to‑speech voice the AI uses when generating audio output.

To update settings for an existing flow, click the more icon button more icon and select Settings from the dropdown.

Persona Settings:

Define how the AI behaves within a specific workflow. Consider it as selecting the brain for your automation. These settings influence:

  • Model selection for responses

  • Speed vs. quality balance

  • Cost per message

  • AI's level of intelligence, creativity, or factuality

  • Capability in reasoning, coding, summarization, etc.

Explanation of Each AI Model in the Dropdown

AI Model

Description

Best For

Not Ideal For

GPT‑4o mini

Lightweight, fast, low‑cost model with good baseline intelligence.

- Simple Q&A - Classification tasks - Routing logic - Small summaries - High‑volume, low-cost apps

- Deep reasoning - Multi-step logic - Long document processing

GPT Realtime

Optimized for instantaneous response generation and streaming interactions.

- Real-time chatbots - Voice assistants - Live customer engagement - Instant reply workflows

- Heavy computation - Large-context analytical tasks

GPT‑5

General‑purpose flagship model with strong reasoning, writing, and coding skills.

- Analytical workflows - Long-form content - Coding tasks - Document understanding

- Cases where cost or speed is the top priority

GPT‑5.1

More stable and consistent version of GPT‑5 with improved structure and reliability.

- Business-critical automation - Enterprise-level QA - Reasoning-heavy flows - Reducing hallucinations

- Ultra-low-cost or ultra-fast workflows

GPT‑5.2

Most advanced, most accurate model with top-tier reasoning and long-context handling.

- Technical analysis - Legal/financial reasoning - Complex decision flows - Long document processing

- Situations where low latency or cost matters more

GPT‑5 Mini

Smaller and cheaper variant of GPT‑5, optimized for speed and scale.

- Lightweight chatbots - Fast interactions - Large-scale workloads - Basic NLP tasks

- High reasoning depth - Complex multi-step logic

Language:

The Language dropdown affects the primary language, writing style, and regional formatting of the AI.

It impacts:

  • Spelling (e.g., "colour" vs. "color")

  • Grammar and phrasing

  • Tone and expressions

  • Regional vocabulary (e.g., "lift" vs. "elevator")

  • Pronunciation (if voice is used)

  • Locale-specific formatting (dates, currency, units)

Choosing the Right Language:

  • Provides a localized customer experience

  • Ensures natural conversations

  • Maintains correct spellings for your audience

  • Consistency in your AI flow

  • Accurate voice output

Key Uses:

  • Customer service bots

  • IVR voice flows

  • Global product launches

  • Region-specific content

Voice:

The Voice dropdown lets you choose the text-to-speech (TTS) voice the AI uses when generating audio responses. This controls:

  • Accent

  • Tone (warm, deep, energetic, neutral)

  • Speed & clarity

  • Speaking style

  • Voice model version (V2 / Turbo V2.5 / Flash V2.5)

Voice Type
Voice Models Shown
Key Features
Best For

Multilingual V2

Sarah, Will

Most natural & expressive, multilingual, warm tone

Customer service, multilingual bots, professional voice apps

Turbo V2.5

Eric, Jessica

Fast + high clarity, balanced naturalness

Real-time agents, high-speed workflows

Flash V2.5

Aria, Harry

Fastest, lightweight, efficient

High-performance bots, minimal latency needs

Step 2: Create flow

A flow starts at the Start node. Add nodes from the Available Nodes list to the Flow canvas by dragging and dropping or clicking to place them at the center. Link the nodes using edges to define the logical sequence of operations and decisions. For each node, open the right sidebar and configure its settings to tailor its behavior to meet your needs.

  • For detailed description of the Flow editor page layout see Flow editor.

  • For detailed descriptions of the available nodes click here.

For an existing flow select Design to access the flow editor page.

Step 3: Save and validate the flow

Click Save in the top right corner to store the current configuration to the backend. The save process automatically checks for logic errors or inconsistencies. The result of this save action is visible in the valid/invalid badge, which is next to the flow name.

Alternatively, use the Validate button to perform checks without saving. The result of the validate action is displayed in a feedback toaster.

Validating Flow logic that is different from the stored configuration can result in a discrepancy between the state badge and feedback toaster.

Unsaved Changes Exit Warning:

This feature protects flow builders from losing their in‑progress configuration when they unintentionally navigate away from the Flow Editor. When unsaved changes are detected, the system displays a clear, interruptive warning before allowing the user to move away from the page. This ensures the flow builder has an opportunity to save their work and avoid forced rework.

By proactively alerting users at the moment of potential data loss, this enhancement strengthens reliability, reduces frustration, and supports a smoother, more predictable editing experience.

Data Loss Scenarios:

Be mindful of these actions that can lead to data loss:

  • Closing the browser: Here you see browser-level navigation warning that appears when a flow builder attempts to leave the Flow Editor while there are unsaved changes.

  • Clicking the previous/back button in the browser: In-product confirmation dialog that appears when a user attempts to leave the Flow Editor by clicking the previous button or back button in the browser while there are unsaved changes in the flow.

  • Clicking the logout button in ixHello: In‑product warning dialog that appears when a user tries to log out of the application while there are unsaved changes in the Flow Editor.

  • Refreshing the browser: browser-level reload confirmation dialog that appears when the user attempts to refresh (reload) the page while there are unsaved changes in the Flow Editor.

Preventing Unsaved Changes When Navigating Away

To prevent losing unsaved changes when navigating away from a page, a warning modal is shown. The user can choose to:

  • Continue: Navigate away, and unsaved changes will be lost.

  • Cancel: Stay on the page to manually save changes by clicking the save button.

Step 3.1: Copy, Cut and Paste

The Copy, Cut, and Paste feature allows users to efficiently duplicate or move nodes within the canvas without manually reconfiguring them. This functionality is designed to streamline node management while maintaining clarity and control.

Copy and Paste a Node

If you have a complex node and prefer not to manually configure all its options, such as with a menu node, you can simply copy it. This creates an exact duplicate, preserving all original options and allowing you to attach edges without starting from scratch.

Users can duplicate a selected node along with its configuration.

This also works seamlessly with keyboard shortcuts:

  • Copy: Ctrl + C

  • Paste: Ctrl + V

Behavior

  • The copied node retains its configuration.

  • It appears on the canvas with an offset from the original node for clarity.

  • Edges are not copied.

  • The copied node is automatically selected, allowing easy configuration in the right sidebar.


Cut and Paste a Node

To cut nodes, simply use the keyboard shortcut or click the Cut icon. Once cut, you can paste the node in the desired location.

Users can move a node to a new location by cutting and pasting.

This also works seamlessly with keyboard shortcuts:

  • Cut: Ctrl + X

  • Paste: Ctrl + V

Behavior

  • The node and its configuration are removed from the original position.

  • Edges are removed and not restored.

  • The pasted node is automatically selected for easy configuration.

Step 4: Test the flow

Click Test in the top right corner to chat with the flow logic that is stored in the backend. You'll get a warning to save changes when the flow logic differs from the stored configuration. Interact with the flow as a user would to ensure it behaves as expected. The active node is marked with an orange border during testing.

Next steps

Once your flow is validated and tested, proceed to deploy it to a voice or webchat channel for live user interaction.

Last updated

Was this helpful?