[solved] Does OpenAI Cache Responses?

Hey all,

I’m putting together a prompt which references some files and describes how they should be modified.

For some context, I’m placing a header section at the beginning of my completion which outlines the document and explains how each section is formatted:

# Listed below are:
# 1. Explanation of how two files need to be changed
# 2. The original files, each one separated by a '---' string
# 3. Files updated with the described changes, with a '---' string in-between each file
# 4. A '####' string, indicating the end of the document

When I run this with test data through OpenAI, however, the updated files are returned without being delimited by ---, causing my program to pick it all up as being the updated contents for the first file.
I’ve attempted to modify the wording for #3, but have received the same response each time.

It seems like after a completion is made for a given prompt, receiving a different response requires a significant amount of difference between prompts. Is there anyway that I could combat this?

Thanks

Never mind, I’ve discovered the problem.
An uninitialized variable wasn’t setting the --- delimiters in the actual prompt, so the API must have viewed the input as being a single file, and came back with updates to the one file.

3 Likes