How does the Knowledge Base function?
The Knowledge Base acts as a reference library for your agent, allowing it to access and use stored information when responding to customer questions. It’s best suited for holding larger sets of data that would be cumbersome to include directly in the agent’s prompt.
When a user asks a question, the agent dynamically searches for the most relevant information in the Knowledge Base and uses that to form its response. It doesn’t memorize the entire content permanently.
What is RAG?
Retrieval-Augmented Generation (RAG) enhances the agent’s accuracy by allowing it to pull data from external sources. Instead of relying solely on pre-trained models, the agent taps into the Knowledge Base to generate more precise and context-aware answers.
How does the agent choose which entry from the Knowledge Base to respond with?
The agent analyzes the user’s question and matches it with the most relevant content in the Knowledge Base. It then selects and presents the best possible answer based on that analysis.
Does the agent copy responses verbatim or create summaries?
Typically, the agent will craft responses in its own words. However, if the information in the Knowledge Base is concise and well-written, the agent may include exact phrases where it makes sense.
Q: What if my Knowledge Base includes conflicting information?
A: When different documents provide conflicting details (e.g., varying numbers or facts), the agent’s response may vary depending on which source it retrieves.
To avoid this:
Regularly review and remove outdated content.
Maintain consistency across all documents.
Test the agent’s answers to ensure they are accurate.
How do I add content to the Knowledge Base?
You can upload documents (PDFs), import content from specific URLs, or create entries directly within the AiGent interface.
Can I upload multiple content sources?
Yes. The Knowledge Base can include a mix of PDFs, web URLs, and documents you add directly. The agent will search across all available content regardless of its source.
Does adding a URL mean the agent will scan my entire website?
No, it will only process the specific page linked. If you need multiple pages indexed, each page’s URL must be added individually.
Will every PDF file work when uploaded?
Not necessarily. Complex PDFs with columns, images, or heavy formatting may cause issues during processing. If problems arise, consider converting the file to a plain-text version before uploading.
How does the agent handle lengthy documents?
The agent doesn’t read every word of large files at once. It searches for the sections most relevant to the user’s query.
For better results:
Use clear headings and structured subheadings.
Break down large files into smaller, topic-focused documents.
Can I control which document the agent references first?
No, prioritizing specific documents isn’t possible. However, you can improve accuracy by:
Refining your content for better clarity.
Grouping related topics together.
Removing unnecessary or outdated information.
What if the agent isn’t pulling the correct information?
If you notice inaccuracies, you can:
Test alternate phrasing of questions.
Rework sections of content for better clarity.
Organize documents with clear, descriptive headers.
