RAG Chatbot

RAG stands for Retrieval-Augmented Generation. Its process relies on two key steps: retrieval and generation. First, your app’s content is converted into embeddings — mathematical representations that make it possible to compare texts. When a user asks a question, it is also transformed into an embedding and compared with your app’s content embeddings to find the most relevant passages through semantic search. These selected excerpts are then passed to the language model, which generates a clear and natural answer — prioritizing information from your app while rephrasing it for better readability.

RAG Chatbot builds its knowledge base from the content you choose within your app. It supports three content types: articles, events, and map points. Sources include both the content you create in the GoodBarber CMS and any external feeds you’ve connected. You stay in full control of the scope — for each section, you define how many items to index and which ones should be included in the chatbot. Whenever content is added or updated, the knowledge base is automatically refreshed, ensuring that answers always reflect your most up-to-date information.

You can customize the chatbot’s appearance to match your app’s visual identity — from colors and fonts to display style. The sources used in answers are highlighted through clear, interactive cards, giving users quick access to the original article, event, or map point. The chatbot thus becomes a seamless, fully integrated experience that fits perfectly within your app’s world.

• 
  Color selection
• 
  Font selection
• 
  Chatbot image
• 
  Style

RAG Chatbot operates on a credit-based system. Each subscription includes a monthly quota of 5,000 credits, automatically renewed. Credits are used for every action — indexing new content, processing user questions, or generating answers. If your quota runs out before the end of the month, you can purchase additional credits to ensure uninterrupted service.

RAG Chatbot lets you choose the AI model used for indexing and generating answers. This choice directly impacts credit consumption: some models are faster and more cost-efficient, while others require more credits but deliver more detailed, nuanced responses. For example, a model like GPT-4o-mini helps you reduce costs, whereas GPT-5 provides maximum quality. You decide based on your priorities — optimizing your budget or focusing on precision. This flexibility allows you to tailor the service to your app’s specific needs.