Feature Request: Forking conversations when prompting: Selective Context

As a developer, when using ChatGPT as an aid for developing projects, I often find myself creating a project, prompting ChatGPT for advice, then find a bug. Solving this bug requires me to prompt GPT in the same conversation that I started creating the project in. This creates two problems: I have to use the same space that I started the project in to solve the bug because that space has the entire context of my code, and once the bug is solved, the bug solution context will stay in the conversation limiting the total amount of prompts I can ask GPT in that conversation.

This issue can be avoided if we have a git-like prompting technique where we could fork a conversation, and the GPT would use only the forked context to answer prompts. If we need the forked conversational context back into the “main” context, we can pull it back in.

Here’s an example:


You can sorta do this - you can edit your messages to start a new branch. For integrating stuff you’d need to copy-paste a summary into one of your older messages.

I suspect doing this automagically and reliably requires a level of technological maturity that OpenAI hasn’t achieved yet, but I think it’ll probably come eventually!

Either that, or the models become so good that you won’t even need it anymore.