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?