interFace
Prompt customiztion

Specific Prompt

9min

This guide will help you understand the Knowledge Base Retrieval Prompt and provide instructions on how to customize it.

The default prompt



Document image


Use context_search tool to retrieve the response. If the information IS NOT CONTAINED IN THE CONTEXT respond with a fallback message that lets the user know you could not find the information. Keep the responses of max 40 words length and keep the answer as concise as possible. If asked about any specific people (by name), places, or telephone numbers not mentioned in the document, state that you are not allowed to speak about people due to privacy. In case a user poses a question not related to the document address the query accurately, then smoothly redirect the conversation back to context.

This prompt is designed for an AI or digital assistant tasked with providing concise, context-specific answers based on a given document or dataset. Here’s how it works:

The Breakdown

1.Document Retrieval

Please note: Use context_search tool to retrieve the response is part of the prompt that uses the knowledge base as a source of information and must not be changed unless you will not use the knowledge base.

2. The Fallback message

“If the information IS NOT CONTAINED IN THE CONTEXT respond with a fallback message that lets the user know you could not find the information.”

This is the fallback message displayed when the requested information isn't available in the knowledge base. You can customize it, for instance: "Sorry, I couldn't find the information you're looking for. Please reach out to our support team for further assistance."

3. Response Length

Keep the responses of max 40 words in length and keep the answer as concise as possible.



By default, the model delivers concise, high-level answers with a 30-word limit. Increasing the word limit allows for more detailed, comprehensive responses, providing deeper insights.

Please note: Extending word limits can lead to longer response times, but it also allows for more detailed and nuanced conversations

Document image


4. Data Privacy

"If asked about any specific individuals, locations, or contact details not mentioned in the document, indicate that you are unable to discuss such matters due to privacy concerns."

This sentence covers the handling of third-party data privacy. Removing it may have privacy implications, as it ensures the model refrains from discussing individuals, locations, or contact details not mentioned in the document.

5. Conversational focus

“In case a user poses a question not related to the document address the query accurately, then smoothly redirect the conversation back to context.”

When a user asks an unrelated question, the model should answer briefly and then redirect to the original topic, keeping the conversation relevant and focused.

Example:

User: "What do you think about the new superhero movie coming out next week?"

Digital Human:

"The new superhero movie looks promising! Now, back to the financial report—how might Q2 revenue trends affect next quarter's projections?"