One problem with ChatGPT 3.5 is the lack of context that ChatGPT 3.5 has of your code base, due to the low word limit for the input. GPT-4 is capable of handling over 25,000 words of text, allowing for use cases like long form content creation, extended conversations, and document search and analysis [1]. Enter GPT Feeder.
GPT Feeder is a command-line application that scans the entire codebase, and produces one string consisting of all filenames and file contents that you want included. This string can then be fed into ChatGPT-4, and the model will be able to generate code based on the context of your code base.
...
For more information on how to contribute and run the application, see CONTRIBUTING.md.