🤖 The first AI-assistant in XWiki is here! (BETA)

10 Oct 2024 5 min read

If you've been following the brainstorming and discussions on the XWiki Forum, you know that we are no strangers to the topic of Large Language Models (LLMs).

With 20 years of building open-source software, we’d like to think that we know a thing or two about knowledge management and collaboration needs that organizations have. We, of course, took into consideration both the potential that AI-assistants can have and the emerging challenges — the spread of incorrect information, privacy concerns, biases, copyright issues, and energy consumption — for our Wiki AI Search Engine (in short, WAISE) project, funded by NGI Search.


🎉 Today, we are thrilled to announce that we have recently finished and released the first official BETA version of the application! 🎉


Your open-source ✨ AI-assistant ✨ in XWiki


In recent months, through the WAISE project, we have developed the first version of our LLM application, which integrates AI-powered functionalities directly into your wiki instance.

Today, this extension is ready to be installed on top of XWiki and taken for a spin. Here's what you can do with it (for now):

  • Stop looking for information and, instead, instantly receive the answers you require from your wiki based on your user rights.
    • Example: You use XWiki and OpenProject and would like to know what deadlines are up next week to prepare meeting notes for your team. If this information exists in your tools, you can ask via the chatbot for this information.
  • Transform existing content from your pre-existent knowledge base (summarize conversations and documents, (re)format, rephrase, extract key messages, ask for comparison tables, brainstorming ideas, etc.).
    • Example: A client has asked you a frequently asked question for which you already have a standard answer in your knowledge base. You can ask the LLM to propose a reply to said client based on the existing procedure.
  • Translate content and receive answers in your preferred language from your multilingual company wiki.
  • Ask and receive information from other web applications, integrated with XWiki.
  • Add content directly to XWiki through its editor integration.

Whatever LLM you want in an open-source box. With the content you want and under your control. We even built a multimeter for it. Oh, and a matrix bot.
(edited)
+20:You can customize it
 - Paul Panțiru, Software Engineer for the WAISE project

XWiki-LLM-app

How does it work? 


On the XWiki server, we've created an LLM interface and a search engine that provides you with a chatbot powered by the LLM of your choice.

Through context (pages and documents indexed from your wiki where you installed the extension) and retrieval capabilities based on our Retrieval Augmented Generation (RAG) system, the chatbot serves answers that best fit your question, based on your permission rights.

Additionally, with some development work, it is also possible to add external sources of content to extend the knowledge base available to the chatbot, allowing you to ask questions based on both internal and external sources.

XWiki-LLM-architecture

A video sneak peek into the AI-assistant


How is this LLM Application different from other models?


The application leverages the capabilities of LLMs to enhance content generation, translation, and intelligent knowledge retrieval within your existing workspace, while offering you flexibility and control over your data through the chosen model. This way, you can rest assured that your sensitive information remains private and secure, while you also upscale operations with the AI-assistant.

⚡ Core innovations to check out:

  • Integration capability with other web applications;
  • Ability to control the information the user has access to via the rights' system;
  • Focus on open-source models;
  • Complementary Element Matrix bot that sends questions to the LLM app.

With this unique blend of features, you now have access to an ethical, easy-to-use, and customizable AI-solution that enhances your daily knowledge management and collaboration processes in the workplace.

Our mission is to empower users by creating an AI-driven search server that is open, transparent, and ethical. We're building a conversational layer on top of XWiki that respects users' privacy and content access rights.
Ethical AI is essential because it ensures that AI technology remains trustworthy, respects human rights, and promotes responsible use in a world where data and AI increasingly shape our interactions. Open-source innovation, coupled with a commitment to ethical standards, is the way forward for AI that benefits everyone.
- Ludovic Dubost, Founder and CEO of XWiki SAS

Customize to your heart's content


You probably already know that thanks to XWiki's open-source nature, you can customize the wiki platform so that it fits your needs. Well, we took XWiki's strengths and incorporated the same principles into the LLM application.

What this means is that the application has been built with the same open-source core in mind, offering you extensive customization options to fit your requirements, such as:

  • Control who can make queries, ensuring security and relevance;
  • Choose the LLM that best fits your requirements for optimal performance;
  • Customize the default behavior of the LLM (system prompt) to align with your business needs;
  • Adjust various parameters to fine-tune the application's functionality.

Moreover, our matrix bot adds another layer of customization, making the new application even more user-friendly and effective:

  • Choose from available LLMs to suit different tasks and preferences;
  • Define the LLM's personality to better match your company's tone and style;
  • Select different models per room (works also with encrypted rooms).

Try the AI-assistant for yourself


Do you want to experience the power of the LLM inside XWiki firsthand, and transform the way you manage and collaborate on knowledge?

If you are an enterprise customer seeking to integrate AI into your wiki, you can request a trial by completing the form below.

We would also, as always, greatly appreciate any feedback you could offer us to further make XWiki an even better product. 

We also want to offer a special "Thank you" to NGI Search, without which our progress would not have been as swift or comprehensive. This support has enabled us to innovate and push the boundaries of knowledge management through the integration of LLMs in XWiki, focusing on privacy-preserving, trustworthy search, and discovery. 🙌

You may also be interested in: