Hi I am currently developing a wordpress website for tourism and have added the WPbot that can use chat GPT through an API.
The website information is professionally curated for high end clients and I would like the chat GPT bot to only use data from the website, or at least prioritize the data over the global GPT database as there is information I have that it doesn’t or information that I do not want to be given to the user.
The fine tuning function using questions/answers (jsonl files) seems to only guide the format/ personality of the chatbot but it still uses the global database, however I want to incorporate my own database preferably through website pages/database, alternatively through a file like using LangChain with PDF/txt files.
How can I train GPT 3.5 to use my own database in WordPress and is there a way I can modify the WPbot pluggin?
What you want is a knowledge graph (vector database) (I recommend Weaviate for general-use, but there are plenty). You embed your information into this database and then use some sort of distance function to augment the LLMs generation by injecting the (hopefully) information into the prompt and instructing the LLM that the injected information takes precedence.
This is called “RAG” (Retrieval Augmented Generation)
Your swift reply is much appreciated.
This gives me a bit to look into. I’m still new to the AI technology so may take some time but thanks for pointing me in the right direction, this will definitely help.
So if I had to set this up and implement it into the website, the data from my database will be used when users ask questions about the data.
Overall the chatbot is meant to act as an AI concierge/ travel assistant on the tourism data in my website.
Yes. There will be a plethora of additional questions once you’ve set it up but checking out Weaviate, reading their documentation, and then tinkering is a great start
wonderful, thanks I appreciated the assistance