voiceflow
The voiceflow module in DigiSquares lets you design, manage, and run conversational AI workflows powered by OpenAI models. You can create multiple voiceflows, each with its own set of triggers and sub-flows, test them in real time, and hook them into your application.
1. voiceflow List View
The Phrases feature is designed to help users initiate discussions or conversations on a given topic by providing a set of natural language examples. These examples are typically 8 to 9 phrases or sentences generated through advanced language models. The phrases serve as conversational starters, guiding users on how to articulate ideas, questions, or statements relevant to the topic at hand.
This feature is especially useful in scenarios where users may be unsure how to begin or steer a conversation, helping to break the ice and encourage engagement.
On the voiceflow page you can:
- See all voiceflows in the current project, with columns:
- Name: Unique identifier
- Description: Brief summary of its purpose
- Actions:
- 📋 Copy
- ✏️ Edit
- 🗑 Delete
- Search existing voiceflows by name or description.
- Create New voiceflow by clicking the + New voiceflow button.
- Test Spark runs all voiceflows end-to-end in a sandbox environment.
2. Creating a New voiceflow
Click + New voiceflow to open the Properties dialog:
Field | Description |
---|---|
Name* | Enter a unique name for your voiceflow. |
Select Model* | Choose which OpenAI model to drive the flow: |
• GPT-3.5 | |
• GPT-4 | |
Description | (Optional) Explain the voiceflow’s purpose. |
After filling out these fields, click OK to initialize your voiceflow.
3. voiceflow Actions Palette
Once inside a voiceflow, click the + icon to open the Actions palette:
Root Trigger: ChatTrigger
All voiceflows begin with a ChatTrigger node. Under this root you can add:
- Phrases Start the conversation when the user’s input matches one of your example phrases.
- Sub Flow Invoke another voiceflow from within this one, allowing modular, reusable dialogs.
4. Configuring Phrase Triggers
Click Phrases to open the phrases editor:
Section | Description |
---|---|
Model | Shows which model is in use (e.g. gpt-3.5-turbo ). |
Instruction | (Optional) Top-level system prompt or instruction for this trigger. Click ✏️ to edit. |
Add Phrases | Enter one example phrase at a time (e.g. “hello”, “help me”). |
+ Add Phrase | Adds the typed phrase to the trigger list. |
Your Phrases | Displays currently configured phrases as removable tags (e.g. chart , dash ). |
Tip: Provide 8–10 diverse examples per trigger to help the model reliably recognize user intent.
5. Calling Sub-Flows
Click Sub Flow to insert a sub-voiceflow node:
Field | Description |
---|---|
Select voiceflow* | Dropdown of all existing voiceflows in the project. |
OK / Cancel | Insert the sub-flow trigger or cancel. |
Use Sub Flow to break large conversations into smaller, reusable components.
6. Putting It All Together
A simple voiceflow might look like:
├─ Phrases: \[“hello”, “hi there”, “hey”]
│ └─ Message: “Hello! How can I help you today?”
└─ Phrases: \[“goodbye”, “bye”]
└─ Message: “Goodbye! Have a great day.”
Or, to delegate part of the conversation:
└─ Sub Flow: “order-status-check”
7. Best Practices & Tips
- Name Clearly: Use descriptive names (
OrderStatusFlow
,SupportGreeting
). - Limit Phrases: 8–10 well-chosen examples per trigger.
- Leverage Sub-Flows: Modularize repeatable dialog segments.
- System Instructions: Use the Instruction field to give high-level guidance (tone, persona).
- Test Regularly: Click Test Spark after major edits to validate behavior end-to-end.
With voiceflow’s visual editor and OpenAI integration, you can rapidly build, iterate, and deploy sophisticated conversational experiences—without writing a single line of runtime code.