Manual:AI integrations - Overview/FAQ

What is the difference between the AI Editing Assistant and the Chatbot?

The AI Editing Assistant integrates an LLM (with "world knowledge," without "expertise" from the wiki) into the editing mode, allowing the content of a page to be edited directly. The Chatbot integration makes the expertise from the wiki available to a chatbot (RAG data export) and is primarily used to find and use existing information.

How does the integrated Chatbot work?

The following process takes wiki content and transfers it to the chat interface:

Chatbot system

1. Provision of raw data from the wiki for processing in the RAG pipeline.
2. Request to the chatbot via the chat interface.
3. Transfer of the request to the RAG pipeline. There, the data is processed accordingly.
4. Transfer to the LLM. Text generation by the language model.
5. Transfer to the chatbot.
6. Transfer to the chat interface. Output there.

 

What are the LLM requirements?

• BlueSpice does not provide the LLM that is a prerequisite for the chatbot feature. The LLM must be provided by the customer and the wiki must be able to communicate with it by exchanging API keys.
• Currently, the following LLMs are supported:
  • OpenAI and OpenAI-compatible systems (including, for example, Microsoft Azure, DeepSeek, Google Gemini)
  • Ollama, Mistral AI (open source, e.g., via IONOS)
• Cloud hosting requirement: Internet access for the LLM is required (common encryption).

How does the RAG pipeline work?

• The wiki's content must be processed in a RAG pipeline so that the LLM can learn from it.
• Functions:
  • Retrieval: Relevant information related to a user query is retrieved from the wiki (= document collections) and provided in context.
  • Generation: The LLM uses the retrieved information as context to formulate a content-based answer.
• Integrated RAG pipeline (extension WikiRAG):
  • Prepares the raw data for further processing in the LLM.
  • The RAG pipeline also handles access control (ACL) and metadata management.
• Customers can operate their own RAG pipeline. An external RAG pipeline must independently manage ACL and metadata management. BlueSpice only provides the raw data.
• Limitations in the cloud: A dedicated RAG pipeline has to be accessible via the internet. However, this is not unprotected public access; rather, access is achieved using common security and encryption mechanisms.

Can some wiki content be excluded as data source?

Yes,

• the user namespace and non-content namespaces are fundamentally excluded from output in the chatbot.
• there is a behavior switch in the page options that specifies whether a page should be indexed for the chatbot or not.

PDF exclude - start

---

To submit feedback about this documentation, visit our community forum.

PDF exclude - end

A large language model (LLM) is a language model trained with self-supervised machine learning on a vast amount of text, designed for natural language processing tasks, especially language generation. The largest and most capable LLMs are generative pre-trained transformers (GPTs) and provide the core capabilities of chatbots such as ChatGPT, Gemini and Claude. wikipedia

Retrieval-augmented generation (RAG) is a technique that enables large language models (LLMs) to retrieve and incorporate new information. With RAG, LLMs do not respond to user queries until they refer to a specified set of documents. These documents supplement information from the LLM's pre-existing training data. This allows LLMs to use domain-specific and/or updated information that is not available in the training data. For example, this helps LLM-based chatbots access internal company data or generate responses based on authoritative sources. wikipedia