How to provide more context to the API to influence results?

Hello all, I’m attempting to create a small project involving openAI’s API that involves effectively offering suggestions to supplement some worldbuilding articles.

Effectively, in order to do what I need, I need to somehow provide the AI with the context of a world like its settlements, characters, geography, or other texts written out so it can compare against them and generate suggestions for more content. Is there a way to provide that as some sort of memory object, variable, or training data?

Hi @memelypepeartly ,

welcome to the community!

Have you allready taken a look into the topic of embeddings? Embeddings - OpenAI API By combining this with another LLM to get the output you should have a working solution. However this might be a bit advanced and not the best way to start with as your first use case depending on your prior (programming) experience.

Hi @memelypepeartly, and welcome to the community!

As @linus indicates, embeddings might be ideal for your project. One avenue that cannot be understated, especially when all these new technical concepts are swirling around, is to lean on some frameworks to get you started.

A good understanding of embeddings is shared by @wfhbrian.

CustomGPT helps you lean into a solution without learning every aspect of the underlying technology. With a few PDF documents, you are off to the races with something you can test and play with. In just minutes, you can develop a working example. And this platform also has an API and widget plugins, making it possible to embed its capabilities into any web or mobile app.

2 Likes