I run each task in one chat and one worktree. After the task is finished (sometimes 15,) I do not want to archive the worktree and create a new one, I want to reuse it to save time/space/etc. To start a new task in the same chat it’s best to clear the context, but I am not able to do this in the Desktop app??!? Compacting sometimes keeps 1/3 of the context and it keeps previous task’s stuff in context.

Am I missing a way? Is this planned?

I just say forget everything before this prompt and focus on this prompt. I also add and remove data by section or topic. It’s like editing. (I am not implying the model can literally clear its context, which it can’t.)

What do you mean? Are you saying you can’t start a new chat or are you asking about memory or same chat context?

I read it as same chat context, clearing chat context window.

I’d probably just start a new chat for unrelated tasks since that’s the only way to guarantee a completely fresh context. If you’re trying to stay organized, keeping separate chats or worktrees by project or topic tends to work well too. As far as I know, the ChatGPT Desktop app doesn’t currently have a way to wipe only the model’s context while keeping the same conversation intact.

A possible workaround is to explicitly compress completed sections yourself before they become old context; for example, tell the model to treat everything above as FEATURE_X_DONE, with a short note defining what that tag represents, and to use only that abstraction going forward unless you ask to expand it. It won’t actually remove or replace the existing context, and there’s no guarantee the app’s context compaction will preserve only the tag, but it seems like a reasonable way to encourage later summaries to carry a concise representation instead of repeating the full discussion.

If OpenAI adds a ‘Clear Context’ feature, that would address this use case.

As far as I know a “in chat clear context feature “ does not exist on any ChatGPT apps.