How to make a local open source AI chatbot who has access to Fedora documentation
Article Description:
The article walks you through setting up an AI chatbot which uses a RAG database of Fedora documentation to answer questions about Fedora. The article shows you how to use the open source RAG database application Docs2DB to ingest a collection of documents and use them for enhancing an LLM with RAG (Retrieval Augmented Generation), effectively giving the LLM the knowledge contained in the ingested documents.
@ellis +1 from me, as well.
I’ve created Pagure ticket #428 to track your article. Please use it to communicate with the editors about the article by leaving comments on it.
Articles are created in the the Fedora Magazine WordPress instance. You should be able to use your FAS account login to access that site.
@hricky Contributions to the Fedora Magazine in content or editing are appreciated but we would ask that you give the authors the opportunity to create the content. By following the ticket number you will know when they have determined it is ready for review. That is a more appropriate time to ask for clarification or more information about it.
I suggest this not in an attempt to restrict you but to NOT dampen the interest of potential contributors. Thank you for your understanding.
How could I take away the opportunity of authors to create content? Even if I could, it wouldn’t even cross my mind to do so.
Since the article will be based on the blog post mentioned in the initial proposal post here, and if AI is used to write it, then the Fedora Magazine article will also inherit this potential AI assistance. In such a possible case, I would suggest that the author read (if they hasn’t already) the Fedora AI-Assisted Contributions Policy. That’s why I asked.
I tried to read the articles several times when they were ready for review by your team, but I didn’t have access. That’s why I’m writing this now, in advance.
I’m sorry if this in any way bothers the autors or disrupts the order and workflow in your team. Please accept my apologies!
@hricky Thanks for the link! I’ll be sure to annotate according to the policy. I designed the script and used ChatGPT to write it, and I tested and tweaked it by hand myself, and iterated as needed. (“Assisted-by: ChatGPT”.) The graphic is from Gemini. I wrote all the copy unassisted.
