Main Instructions
This is the most critical section for defining your assistant's personality and core behavior. Here, you set its role, tone, and the fundamental rules it must follow when interacting with users.
Instructions Choice
The platform offers two distinct modes for setting up instructions, which you can select from the dropdown menu:
- Use default + custom prompt: (Recommended) This mode combines a pre-configured, optimized system prompt with your own specific instructions. It's the best way to get started quickly while still being able to customize the AI's behavior.
- Use custom prompt only: This mode gives you complete control by disabling the default prompt entirely. Your instructions will be the only ones the AI follows. This is intended for advanced users who want to build a highly specific personality from scratch.

Configuration Options
Custom prompt for Answer generation: This is where you add your primary instructions. Define the assistant's role, personality, tone, and any specific rules it must always follow.
- Example: "You are a helpful and friendly support assistant for the Smartly brand. Always answer in a positive and encouraging tone. Do not mention competitors."
Custom prompt if no chunk is found: These instructions are triggered only when the assistant cannot find any relevant information in the Knowledge Base to answer a user's question.
- Example: "I couldn't find the information in my knowledge base. Please tell the user that you don't have the answer and suggest they rephrase their question or contact support at [email protected]."
Custom prompt for response formatting: Use this field to specify how the assistant should structure its answers. You can ask for Markdown, bullet points, JSON, etc.
- Example: "Format your answers using Markdown. Use bold for key terms and bullet points for lists."
Max answer length (in words): This setting allows you to guide the length of the AI's responses.
- The assistant will try to stay within this limit, but the final length can vary depending on the model and the complexity of the question.
Updated 2 days ago