- “Powered by” badge
- RAG chatbot (text)
- One-click RAG Sitemap
- Custom AI Soul
- BYOK, runs on free API tiers
- 3 Chatbot Themes
- Chat history in the visitor’s browser (localStorage)
- Remove “Powered by” badge
- Visitors upload images, AI sees them
- Role-based chat access
- Multi-language sites (subdirectory)
- Custom RAG Sitemap grouping
- Advanced chatbot styling
- Remove “Powered by” badge
- Visitors upload images, AI sees them
- Role-based chat access
- Multi-language sites (subdirectory)
- Custom RAG Sitemap grouping
- Advanced chatbot styling
Your site is already an organized knowledge base; all it lacks is an entry point an AI can read. Many people assume that installing an AI chatbot means first building a vector database and reorganizing the whole site, and the thought alone is enough to put them off. In fact, the articles and categories you have already written are a ready-made structure, and the entry point is exactly what RAG Chatbot adds. You do not have to redo your content or move data into a vector store; the AI understands the visitor's question and sees the image the visitor pastes in; your site is no longer just a search box that matches keywords.
This assistant can be kept running for as little as $0. AI inference is BYOK, paid directly to the provider, and many providers have free tiers that are more than enough for a site with a few hundred visitors a day. Even if usage grows to where you pay, a thousand answers a month is only about $1.2, less than a single cup of Starbucks.
What makes it this cheap is not which model you use but how it finds the content to read. WordPress AI plugins on the market can also connect cheap models, but they rely on a vector database to compute similarity, pulling back a batch of context-poor fragments at once and dumping the whole batch on the model to piece into a coherent answer; the more scattered the pieces, the stronger the model has to be.
RAG Chatbot computes no similarity; it converges layer by layer along the category hierarchy you arranged by hand, and at each step it only has to judge by meaning which category to go deeper into, with no piecing together, so even the cheapest model can handle it: with Groq's Llama 3 8B, for example, one question costs about $0.0012. This is how the RAG Sitemap saves money: not a stronger model, just the right structure, answering more than 8,000 questions a month for $10. As for the all-in-one AI customer-service platforms that package everything and charge per resolved conversation at $0.29 to $0.99 each, that is another order of magnitude. For individual site owners and small companies, RAG Sitemap for Chatbot is the first WordPress AI assistant they can truly afford.
From the start, this architecture has pushed in one direction: moving compute back onto the user's own device. You play games on your own graphics card and watch video on your own screen; asking an AI will, in the end, also return to inference on your own device. When chip-is-model becomes reality and a small model is etched into a chip's circuits, AI stands by in the background like Wi-Fi, and that is exactly the day RAG Chatbot is waiting for. A small model etched into a chip has limited world knowledge, but edge scenarios never needed world knowledge; what they need is knowledge of the site in front of them. Site knowledge is supplied by the RAG Sitemap, and the model only has to understand and navigate. At that moment, the question is handled locally on the visitor's own device, with no online API needed at all, and the cost of a query returns to the user's hands while the site owner's side goes to zero.
繁體中文 
English