How to feed a large context?

[crystal drum]

I'm trying to feed the content of a book to the gpt-3.5 turbo API, but it has a word count limit of about 4000. I want to create a chatbot which has context of all the content of the book which is about 700,000 characters. How can I do it ? Is it possible ?

[worthy blade]

You can only fit certain amount of tokens into each result but you can build a vector DB, and then search the correct context from there with each query

[crystal drum]

hm right thanks

[devout igloo]

Check out embeddings.