I have a book, I want OpenAI to be trained on the book. I want it teach the general public based on the text from the book. It is a very important book, I believe it will help everyone. How can I do it?
Either wait for the 32k token model or use embedding API
I agree that embeddings are the way to go. There are full databases for holding and working with embeddings, and I would recommend that approach if you think it will get a lot of traffic. For lower-traffic usage, I’ve seen embeddings stored in a Google Sheet. Really depends on how granular you are going to get, how big the book is, and how much traffic. The more granular you want your results, the more lines of embeddings you will have.
Please do you have any tutorial on how to get started with embeddings API or a link?
What is this book about?
What is an average chapter size?
What is your final goal to have the trained model do?
Is this your book, or someone else’s? I think it definitely helps if you know the subject matter inside and out.