Improving ChatGPT for Large Projects

ChatGPT Feature requests
,
1

Improving ChatGPT for Large Projects:

Hi Everyone,

I love using ChatGPT, but I’ve noticed that when working on large projects—especially with programming—ChatGPT often forgets earlier parts of the code during long conversations. This means I have to keep copying and pasting code and re-explaining the project, which is frustrating.

The Problem: ChatGPT struggles with large projects because it can’t hold onto big amounts of information like project structures, multiple files, and lots of code. This is a specific challenge for developers but could help others managing large projects too.

Solution:

  • A temporary memory workspace where we can upload project folders, files, and code. ChatGPT could reference these throughout the session, keeping the context intact.
  • Memory Issues: In long chats, ChatGPT forgets earlier details. Expanding the context window or using a system like Redis to store and retrieve relevant parts of the conversation could solve this.

This feature could be available in the paid version and would make ChatGPT way more useful for big projects.

what do you guys think ?

2

If you are not working with a specific function/method then try just using the signature/prototype of that function/method. This often vastly reduces the amount of information that the model needs to hold.

3

The idea of using function or method signatures isn’t just about code; it’s also about managing the information related to an entire project. When dealing with a large project, especially when debugging or addressing issues realted to the code or performing unit testing, the conversation with ChatGPT can become long and complex. As the conversation continues, ChatGPT may start to forget earlier details, whether it’s about the code or even important context like project structure, dependencies, or previous decisions.

4

Have you tried to use OpenAI canvas?

Yes it is in beta and has many issues/wrinkles but some of us are giving it a try and learning how to be more effective with it when coding.