Add chatbox plugin with GPT-based response system
closed
mostafa .H
Wiki.js is a powerful and flexible wiki software that allows organizations to create and maintain collaborative documentation and knowledge bases. However, users may sometimes have specific questions that are not covered in the existing documentation, and it can be difficult to find the information they need. To address this issue, we propose adding a chatbox plugin with a GPT-based response system that can provide personalized and relevant answers to user questions.
The chatbox would be a simple interface that allows users to ask questions in natural language. The GPT-based response system would then analyze the question using OpenAI's embedding technology and find the most relevant section of the wiki page. It would then provide a short, concise answer to the user's question, based on the content of that section. If the answer is not satisfactory or the user needs more information, the chatbox can provide a link to the relevant wiki page or allow the user to ask another question.
Implementation:
To implement this feature, we will need to develop a plugin that integrates with OpenAI's GPT models and the existing Wiki.js codebase. The plugin will need to provide a user interface for the chatbox and integrate with the existing search functionality to find the most relevant sections of the wiki page. We will also need to implement the logic for analyzing the user's question using OpenAI's embedding technology and generating an appropriate response.
Testing:
To ensure the quality and effectiveness of this feature, we will need to conduct extensive testing and validation. We will need to test the chatbox plugin in various scenarios and use cases to ensure that it provides accurate and relevant responses. We will also need to test the integration with OpenAI's GPT models to ensure that it can handle a wide range of user questions and provide appropriate responses.
Conclusion:
Adding a chatbox plugin with a GPT-based response system can greatly enhance the usability and accessibility of Wiki.js for users. It can provide personalized and relevant answers to user questions and help users find the information they need quickly and easily. We believe that this feature will be a valuable addition to Wiki.js and will help organizations create and maintain high-quality documentation and knowledge bases.
Joshua Krüger
I am also surprised by the answer.
I agree that it is not the idea to add ChatGPT everywhere, but it's making total sense in a Wiki / Knowledge Base. I think NextCloud has shown some nice examples how this could look (in an ethical way) and the Context Chat in their Collectives Integration could be a good inspiration (and the code, and the models used are open source and the frontend is also vue)
https://github.com/nextcloud/context_chat
https://apps.nextcloud.com/apps/context_chat
https://apps.nextcloud.com/apps/context_chat_backend
A
Aditunoe
Wow, I'm surprised about this reaction. I think it makes total sense.
- A chatbot is not made for facts, but for generating language. It makes total sense to give it access to a knowledge base like a wiki, to provide accurate information about content.
- Mostly people are to lazy to go trough the whole wiki searching for what they want, they can use keywords to search, but maybe they use the wrong ones. Giving them the possibility to describe the problem and get an answer with a link to the original source, for me thats a huge advantage.
Another question is, if it has to be GPT from OpenAI or just any powerfull LLM. Though is probably easier to just implented with an API like OpenAI.
Nicolas Giard
closed
The goal is not to add chatGPT to every single product on the internet... this is not useful. The day this tech will have matured a bit, it can be revisited. But in its current state, it just doesn't make sense.
Andreas Scheucher
Nicolas GiardI would say, now is the time. If not now, then yesterday.
Even Obsidan has it for a year now: https://forum.obsidian.md/t/introducing-obsidian-ai-assistant-your-gateway-to-ai-models-in-obsidian-notes/59437
Considering to switch for 2 years now, this might be the tipping point.
C
Cenk CB Bonke
Nicolas Giard, would now not be a good time? As the other comments mention, a text generating llm can be a very handy. Especially in the context of knowledge bases.
I also want to mention that ChatGPT is not the only option. Certainly an option i personally would never use. Luckily for us, there are a ton of way's to run a local LLM. So a simple integration to support a basic llm model to read and get context would be outstanding and a game changer for many people!
D
Daniel
This text ist definitely not generated by GPT.
:/