Train your AI with your business knowledge

The Dunefox knowledge base accepts PDFs, Word documents, individual FAQ entries (question and answer pairs), and website URLs. When a URL is added, Dunefox reads and indexes the content from that page automatically. All uploaded content is chunked, vectorized, and indexed so the AI can retrieve precise answers from it.

Content is organized into named Knowledge Collections, for example, 'Product FAQs', 'Refund Policy', 'Technical Support'. Each collection can be assigned to specific channels (e.g., only WhatsApp, or all channels) and toggled active or inactive without deleting content. This allows seasonal content to be switched on and off without losing data.

Dunefox uses RAG (Retrieval-Augmented Generation): the AI retrieves from your exact business content before generating a response, rather than generating from general training data. This eliminates hallucination. Every answer is grounded in content you have uploaded. The indexed chunk count shows exactly how many pieces of content are in the AI's library at any time.

Yes. Each Knowledge Collection in Dunefox can be assigned to specific channels (WhatsApp only, website chat only, or all channels). This means your WhatsApp AI agent can draw from a different knowledge set than your website widget if your products or audiences differ across channels.

The Dunefox knowledge base dashboard shows real-time stats: total number of documents uploaded, total Knowledge Collections, and total indexed chunks. Each collection shows how many content chunks have been processed from its documents. You can also test the AI directly from the knowledge base to verify it answers correctly before going live.

Training is as simple as uploading your content. Add PDFs, Word documents, FAQ pairs, or website URLs to a Knowledge Collection. Dunefox automatically chunks, vectorizes, and indexes the content so the AI can retrieve precise answers. There is no model fine-tuning, no code, and no waiting. The AI starts answering from your content within minutes of upload. You can test answers directly from the dashboard before making the collection live on any channel.

When the AI cannot find a relevant answer in your knowledge base, it does not guess or hallucinate. Instead, it acknowledges that it cannot answer and offers to connect the customer with a human agent. The conversation is escalated to your shared inbox with the full chat history and the unanswered question highlighted, so your team can respond with context. You can also configure a custom fallback message for these situations.

Update your knowledge base whenever your business information changes, new products, updated pricing, revised policies, or seasonal FAQs. Dunefox makes this easy: upload a new document or edit an FAQ pair, and the AI starts using the updated content immediately. For businesses with frequently changing information such as inventory, pricing tiers, or event schedules, we recommend a weekly review to ensure the AI always reflects your latest details.

Yes. Dunefox indexes content at the chunk level, which means it can handle large volumes of product descriptions, SKU details, specifications, and pricing information. The AI retrieves the specific chunk that matches the customer query rather than scanning the entire catalog, so response speed and accuracy are maintained regardless of catalog size. E-commerce businesses with thousands of SKUs use this to power instant product Q&A across WhatsApp and website chat.