It's possible to restrict the AI Agent's responses to information provided from a specific source. This ensures that the AI Agent only responds from that source.
AI software will always attempt to generate an answer to questions asked in the Chat Widget. To do this, the AI Agent will use the provided knowledge as well as OpenAI's general knowledge. This allows the AI Agent to formulate responses to all types of questions.
In some cases, you prefer the AI Agent to limit itself to the added knowledge. In this article, you will learn different ways to achieve this.
- Limit your AI Agent: It is important to set boundaries for your AI Agent so that it does not answer questions about topics outside its intended scope. In this Help Center article, you can find the best way to configure this.
- Add extra instructions to the AI Agent: In addition to setting boundaries, you can further restrict knowledge with an extra instruction in the Domain Knowledge section
Example instruction:For all questions, only use knowledge from source: https://www.examplewebshop.com. If the information is not available there, respond with:
"I'm sorry, but I can’t provide an answer to that."Example:
- How does the return process work?
- When will I receive a refund?
Restrict to a specific URL for specific topics
It is also possible to write an instruction for specific topics. For example, if someone asks a question about Topic A, the AI Agent should only use a specific URL. This instruction works best in English. You can customize this instruction to fit your needs, such as specifying how the AI Agent should respond when it cannot find the requested information.
Example instruction:
Use the provided information delimited by triple quotes to answer questions about topic A. If the answer is not in the information, please say that you can’t answer this.
"""Source URL: [https://example.com/faq]. Only use the information from this link.
"""
Restrict to a specific document for specific topics
If you want to limit the AI Agent's knowledge to a specific document, you need to add the document's URL after Source URL. To find the Source URL of a document, download it from the Document Scraper and copy the URL from the browser. Then, paste this URL after Source URL.
Example instruction:
Use the provided information delimited by triple quotes to answer questions about topic A. If the answer is not in the information, please say that you can’t answer this. Do not mention it is not in the document.
"""
Source URL: [https://document.url]. Only use the information from this link.
"""