Checkpoints Created On Every User Response

[petiteoffense]

I absolutely love Roo Code, but its one remaining unforgivable flaw that makes me want to throw my laptop against the wall is the way that it manages checkpoints. The way Cursor does it was always better, but I swear to God it seems worse in Roo Code right now than it was a month or two ago when I last spent a lot of time with it. I would say it seems almost completely random where checkpoints are created, but it's almost worse than that. It's almost like they're created at the worst and most inconvenient points they could be. And they're way too sparse. It seems crazy to me to for checkpoints to only be created in the middle of successive AI turns and only after files are edited and not even consistently when files are edited. It's driving me crazy, honestly.

I am begging on my hands and knees for checkpoints to work exactly as they do in Cursor--so that I can just go back to any of my prior responses, click into them, edit the prompt, and reissue it, thus resetting the files and the context history. Sometimes all that's needed is just to re-roll the prompt and hope for a different outcome, but it's basically impossible to do that if checkpoints are created only at the points after a file is edited and after the LLM has already planned out what it wants to do. The way checkpoints are currently happening, if I roll back to them, I'm not really resetting the thought process of the LLM that led it to a bad conclusion. Nor can I tweak any problems in my own prompt without having to rollback chat context that followed successful changes, which means that I might have to completely overhaul my prompt given that now-missing context.

[petiteoffense]

Also, to make matters worse, I'm noticing that now checkpoints don't always allow you to cut back the context history to the checkpoint. Sometimes all it allows you to do is restore files. This is so bizarre to me.