My organisation has 2 types of data. Process and knowledge stores on confluence(wiki)
2nd. A large database with static mapping tables and dynamic data
Im new to this. How do i connect it to database or wiki. I want gpt to access this data dynamically after it has been trained using it
I tried the create gpt but it only accepts attached files
Welcome!
Typically you create a search index on your data, and then retrieve it on demand to answer a question or perform a task; you don’t “train” a model on you data.
If you’re trying to build a cusom gpt (chat.openai.com as opposed to the api platform.openai.com) you could consider exposing your data through an endpoint, and using actions to retrieve it. Then, you’d just have to tell the model how to retrieve the data, and when. In theory, anyways.
Thanks. Is it better to build it with platform?
Depends on what you’re trying to achieve!
Are you a developer or a BA? Do you wanna slap together a prototype to show your boss, or do you have a team to implement and operate a stable enterprise grade product? Do you have any privacy or security concerns? Is your org a managed microsoft customer?
Sounds to me like you are going to want to either embed the data or access via a function/action or some combination of both. As for accessing via function, see this conversation: Private Chat with CSV data - #7 by DevGirl
If you are talking embedding, and NOT using GPT / Assistants API, then you’ll need to basically learn how to put together a RAG system: https://youtu.be/T-D1OfcDW1M?si=Zh08QOongbBXYAhm
In any event, it would probably be wise to follow @Diet advice to put together a prototype which you should be able to do using GPTs. This process will give you an understanding of the elements involved to make your project happen on a larger scale.