Training Your Chatbot with Knowledge

FrontlineHQ uses RAG (Retrieval-Augmented Generation) to power your chatbot. When a customer asks a question, the AI searches your knowledge base for relevant information, then crafts a natural response using that context.

The more knowledge you add, the more accurate your chatbot becomes.

Enter your website URL and click Scrape. We'll crawl your pages and extract text content automatically. This usually captures services, pricing, hours, location, and general business info.

After scraping, review the extracted content. You can edit or delete any chunks that aren't relevant.

Q&A pairs are the most precise way to train your chatbot. Add a question and its exact answer. When a customer asks something similar, the AI will use your answer.

Tips for great Q&A pairs:

• Use the exact wording your customers use (e.g., "How much does it cost?" not "Pricing information")
• Include variations — add the same answer for different phrasings
• Keep answers concise — 2-3 sentences is ideal for chat
• Include specific details like prices, hours, and addresses

Upload PDF, DOCX, or TXT files containing business information. Common uploads include:

• Service menus or price lists
• Employee handbooks or SOPs
• Product catalogs
• FAQ documents

Files are automatically chunked and indexed for search.

There's no minimum, but we recommend at least 20-30 knowledge chunks for a good experience. Most businesses end up with 50-100+ chunks after adding their website plus Q&A pairs.

Check the Knowledge Gaps section in your dashboard to see questions customers asked that your chatbot couldn't answer well — then add Q&A pairs to fill those gaps.