Guidelines
Set up behavioral guidelines for your model to customize how the chatbot responds, ensuring it aligns with your intended tone, style, and interaction preferences.
Guidelines are submitted to your model on each request and can contain variables referenced from the current conversation.
Keep in mind that implementing these guidelines can increase the token usage when interacting with the language model. While the increase may not be substantial, it's important to consider this impact on efficiency and performance.
Here are few examples of guidelines that you might want to use:
Topic Focus: To specify that the chatbot should only discuss particular topics, such as customer support, product details, or specific services. You can also restrict it from discussing unrelated subjects to keep conversations on track.
Tone and Formality: Define the desired tone for the chatbot, whether it should be formal, casual, friendly, or professional. You can guide the chatbot to adjust its language style depending on the audience or context.
Sensitive Content Restrictions: Set guidelines to prevent the chatbot from engaging in sensitive or inappropriate discussions, such as political opinions, controversial topics, or personal advice.
Provide Additional Information: To include supplementary knowledge that may not be part of the model's training, like company-specific policies, product features, or frequently asked questions (FAQs), so the chatbot can respond with more accurate and relevant information.
Response Length: Specify how detailed or concise the chatbot’s responses should be, whether you prefer short, to-the-point answers or more thorough, descriptive explanations.
Personality and Empathy: Direct the chatbot to exhibit certain behaviors, like showing empathy or being more solution-focused when addressing customer concerns, ensuring that the chatbot aligns with your brand’s values.
Language Use: If your chatbot should only use certain languages or dialects, or avoid jargon, this can be configured to ensure clear and accessible communication.
Last updated