General Assistant Settings
Steps to tailor the behavior and functionalities of your Refly AI assistant using the options available in the Assistant Settings page.
Introduction
This guide will help you configure the general settings of your Refly AI assistant. Tailor the behavior and functionalities of your assistant using the options available in the Assistant Settings page.
Configuring System Instructions
Instructions
Modify the system instructions that the assistant uses for each message. This allows you to customize how the assistant responds to user inquiries, ensuring it aligns with your company’s tone and information.
The maximum length for instructions is 64,000 characters.
Why It Matters: Detailed and specific instructions ensure that the assistant provides accurate, consistent, and relevant responses, enhancing user satisfaction and trust in your service.
Setting the Primary Language
Primary Language
Select the language most commonly spoken by your customers to ensure that communications are tailored appropriately.
- Use the dropdown menu to select the desired language.
Why It Matters: Setting the correct primary language ensures that your assistant can communicate effectively with your audience, reducing misunderstandings and improving the overall user experience.
Adjusting Top P Parameter
Top P
Adjust the Top P parameter for sampling with temperature, known as nucleus sampling. This setting controls the probability mass considered by the model.
- Use the slider to set the desired Top P value. The default value is 0.5.
We recommend altering either top P parameter or the temperature setting, but not both.
Why It Matters: Adjusting the Top P parameter helps in fine-tuning the assistant’s responses. A lower value makes the assistant more deterministic, while a higher value introduces more variability. This helps in balancing between creativity and accuracy in responses.
Managing Allowed Domains
Allowed Domains
Manage the domains where the assistant is allowed to operate.
- Add new domains by entering the URL and clicking the Add button.
- Delete existing domains by clicking the Delete button next to the domain URL.
Why It Matters: Restricting the assistant to specific domains ensures it only interacts with users in approved contexts, enhancing security and relevance. This prevents the assistant from being used in unintended or inappropriate environments.
Choosing the AI Model
Model Selection
You can select from multiple AI models for your assistant to optimize performance according to your needs:
- GPT-4o: The top-of-the-line model designed for complex and detailed tasks, offering advanced intelligence and performance.
- GPT-4o mini: A cost-effective, smart model perfect for quick and simple tasks, offering fast and efficient performance.
- GPT-3.5 Turbo: A fast and budget-friendly model ideal for basic tasks, providing quick and reliable responses.
Why It Matters: Different models offer various balances between capability and efficiency. Selecting the appropriate model allows you to tailor the assistant's performance to meet your specific requirements, whether that involves maximizing processing power or optimizing for speed and resource use.
Pausing the Assistant
Pause Assistant
This option allows you to temporarily pause your assistant. You can reactivate the assistant at any moment by toggling the switch.
Why It Matters: Being able to pause the assistant provides control over its availability, which is useful during maintenance periods, updates, or when addressing unforeseen issues. This ensures that users always have a reliable interaction experience.
Saving Changes
Save Changes
After configuring the desired settings, make sure to click the Save button at the bottom right to apply the changes.
By configuring these settings, you can optimize the performance and behavior of your AI assistant to better meet the needs of your customers.