Remember prompt context

I’d like to provide ChatGPT a git commit history of a codebase and ask it questions on it (ex: who worked on X?, who fixed the most bugs recently?)

The git commit history is substanticial (~300k tokens). Is there a way to create a model with the data without having to pay the cost for each prompt?

I took a look at fine tuning, but it seems to require you provide it a “prompt” and a “completition”. How can I achieve this?

1 Like

The simplest way to do this would be to use embeddings, you would embed the git commit history and then use simple cosine similarity to compare the query to the git commit history. You might want to check out: openai-cookbook/Question_answering_using_embeddings.ipynb at main · openai/openai-cookbook · GitHub