The Custom Prompts feature allows you to define specific instructions for your chatbot at a node level. This enables greater control over how the chatbot interacts with users by tailoring its responses to specific scenarios or tasks.

​
What Are Custom Prompts?

Custom prompts are node-specific instructions that modify the chatbot’s behavior and responses. By creating custom prompts, you can:
  • Define specific tones, styles, or objectives for each node.
  • Tailor responses to match the context of a conversation.
  • Combine prompts with tools and variables to make interactions dynamic.

​
Creating Custom Prompts

​
1. Using the Text Editor

Custom prompts are written directly in the Text Editor of a node. The Text Editor allows you to:
  • Write instructions for the chatbot in plain text.
  • Insert variables and tools dynamically using the { shortcut.

​
Steps:

  1. Open the node’s settings and go to the Overview tab.
  2. In the Instructions field, write the prompt for that node.
  3. Use { to insert variables or tools dynamically.
Example: 
Hello {user_name}, I see you’re interested in {product_name}. How can I assist you further?

​
2. Combining with LLM Configuration

Enhance your custom prompts by adjusting the LLM Configuration for the node:
  • Temperature: Adjust creativity or randomness in responses.
  • Max Tokens: Limit the length of responses.
  • Rewind Level: Allow the chatbot to recall context from previous nodes.

​
Example:

  • Set a custom prompt: “Provide concise and fact-based responses.”
  • Adjust the Temperature to 0.3 for precise, non-creative answers.
Visual Placeholder:
Custom Prompt with LLM Configuration
Image showing LLM settings paired with a custom prompt.

​
Examples of Custom Prompts

  1. Customer Support: 
    Hi {user_name}, how can I assist you today? Please be as specific as possible.
  2. FAQ Assistance: 
    Based on our knowledge base, here is what I found for {query_topic}.
  3. Appointment Scheduling: 
    Let’s schedule your appointment. What date and time works best for you?
  4. E-Commerce Recommendation: 
    I recommend {product_name} for your needs. Would you like to add it to your cart?

​
Testing Custom Prompts

Once created, test your custom prompts to ensure they generate the desired responses:
  1. Use the Test Tool in the Canvas Workspace.
  2. Input sample user queries and observe the chatbot’s response.
  3. Adjust prompts or configurations as needed for better results.

​
Best Practices for Custom Prompts

  • Be Specific: Write clear and concise instructions for the chatbot.
  • Use Variables: Incorporate dynamic data for personalized responses.
  • Adjust LLM Settings: Pair prompts with appropriate temperature and token limits for optimal behavior.
  • Test Thoroughly: Simulate various user scenarios to refine prompts.

​
Example Flow with Custom Prompts

Scenario: An e-commerce chatbot flow:
  1. Start Node: Greets the user with a custom welcome prompt.
  2. Product Node: Recommends products based on user input.
  3. FAQ Node: Provides detailed answers from the knowledge base.
  4. End Node: Thanks the user and summarizes the session.
Ready to optimize your chatbot further? Explore Models Configuration to fine-tune your AI’s performance.