Your website AI assistant
Understand what Bulbul knows, how RAG works and what comes next.
An assistant for your own website
Bulbul uses an existing AI model with your website's approved knowledge. Visitors can ask about your products, services and policies in their own words. RAG retrieves relevant document passages before the model writes its answer.
This is not a newly trained model or an assistant that automatically knows every page. Each website agent has its own domain, documents, appearance and conversations.
Give visitors a useful next step
Add helpful product or service explanations to your source documents. Include clear names, suitability, limitations and page links. A cited document passage can supply a clickable link on that agent's registered domain.
For example, a bicycle shop document could explain a tube repair kit and link to its product page. Bulbul can show that supplied link alongside a sourced answer. Live stock, prices, local service availability and checkout are not connected by a document upload.
Keep the boundary clear
Owners review and approve document sections. Uploading knowledge does not permanently train the model; changes happen through source re-upload and approval. Private account information is not made accessible by installing the public widget.
Connected conversations keep recent context for follow-up questions while checking every fact against current approved documents. Live catalog integrations remain planned. The public demo answers live questions about Bulbul from its own public product guide. It cannot read customer knowledge. Your signed-in workspace uses its own approved documents and scoped database retrieval.