Overview of problem:

Right now longer chats allow and nearly stop the IDE, which is very frustrating for larger and more complex projects. When is start a new chat Codium has no knowledge of what came before, constant significant time and credits as it refamiliarizes itself with the code and your intentions.

The IDE slows so much that I have to literally wait up to 30 seconds for each keystroke to show. It's an insanely problematic bug that should be easy to fix with the suggestion.

I'm assuming this is because it swells the RAM used internally by the IDE, grabbing a huge amount of memory, to the point that the app is nearly non-functional.

Proposed solution:

Using a scratch disk for chat memory past a few kb would allow for almost infinitely long chats and I'm going project context, without the truly frustrat8ng side effects of slowing the IDE and Codium to nearly a standfill after 10 to 15 minutes of coding with it.

The chat body can be referencer at any time and it will be separate from the software itself, preventing with her basically memory links and incredible slowdowns and functionality. He didn't keep a few minutes of chat before things started to slow down, reducing latency, and after that, any additional latency would be more than made up for in fast ,better functionality, especially in longer chats. We are already accustomed to waiting for the IDE to be "thinking" and say "Generating please wait..."