Automatic compaction of current session to continue.

[greggh]

Pre-submit Checks

• I have searched Warp feature requests and there are no duplicates
• I have searched Warp docs and my feature is not there

Describe the solution you'd like?

The only feature I am missing from competing tools, specifically in the code editing and generation space, is the ability to compact the current session to avoid "Token limit exceeded" errors. In Claude Code it automatically notices that it is running out of context length and runs its compact function. This summarized all of the key context from the entire current session, creates a new session and gives it that compacted session data. This is done automatically, but can also be triggered manually.

Is your feature request related to a problem? Please describe.

I hit the "Token limit exceeded" a lot. Especially when editing large code files. This leads to sessions working on a complex issue only finishing about half of the work and then stopping until I started a new session. Then I have to try to copy and paste in some of the context from that broken session and hope we can pick up from where we left off.

Additional context

In the other tools a long running series of tasks from a plan will reach the end of the context window and compact into a summary of the current session, and a set of next tasks to continue. this will usually include any key insights or instructions you've given the tool, especially instructions for when it is doing something wrong. Without this feature I have to do double the work, which of course isn't as much as if I didn't have Warp of course, but it would really make my life a lot easier and catch up with some other leading tools in the space.

Operating system (OS)

Linux

How important is this feature to you?

4

Warp Internal (ignore) - linear-label:39cc6478-1249-4ee7-950b-c428edfeecd1

None

[dannyneira]

Thanks for this feature request @greggh we may be working on a way to continue using basic models when the tokens are used up for the more advanced models. this way, you can continue to use Agent mode. No eta on release of this, but it may help your use case in addition to other improvements to memory/coding, etc that we are actively working on.

To anyone else interested in this feature, please add a 👍 to the original post at the top to signal that you want this feature, and subscribe if you'd like to be notified.

[greggh]

Thanks for this feature request @greggh we may be working on a way to continue using basic models when the tokens are used up for the more advanced models. this way, you can continue to use Agent mode. No eta on release of this, but it may help your use case in addition to other improvements to memory/coding, etc that we are actively working on.

To anyone else interested in this feature, please add a 👍 to the original post at the top to signal that you want this feature, and subscribe if you'd like to be notified.

I just want to make sure I was clear, I think we're on the same page @dannyneira but lets be sure. I am not talking about my monthly token limit. I am talking about an error that looks more like a context window problem, which is what compaction solves. Right after I get the error I can open another tab and start back up. Thats why I believe this is whats happening.