fossil sky
I am writing an app that helps a user to write a large document (120 pages+) and want to be able to have it ask different kinds of questions about the previously written material with granular accuracy of the entire document. I can store the document in progress in my app's dB and am open to generating and storing embeddings or creating fine tuning models for sentences or paragraphs, but I'm not sure how to work with them beyond that.
I also tried creating a training model with the text, it seems to use it as an example to emulate, not a piece of data to "discuss" or expand on. Any tips on using the openAI API combined with application data to ask both broad and specific questions about my document would be appreciated!
Examples question's I'd like to ask:
List all the locations in the story.
Tell me why {character A} is upset with {character B}.
List any unused story callbacks.
Summarize everything that has happened to {character}
What is the sentiment between {character A} and {character B} at the end of chapter 3.
{character A} is in an argument with {character B} Suggest new original dialogue for their conversation.
full gust
solar wyvern