Codex Renderer is consuming excessive memory

What version of the Codex App are you using (From “About Codex” dialog)?

26.803.41515

What subscription do you have?

ChatGPT Pro 20x

What platform is your computer?

arm64

What issue are you seeing?

The Codex Renderer processes are consuming an excessive amount of memory.

This may be related to conversations becoming very large over time. A long conversation contains a significant amount of scrollable content, and it appears that too much of this content may remain loaded or rendered in memory. Implementing more efficient list virtualization, lazy rendering, or pagination could potentially reduce the memory usage of long conversations.

I am also not sure why two separate Codex Renderer processes are running at the same time, or whether both processes are contributing to the excessive memory consumption.

What steps can reproduce the bug?

1. Open or continue using a conversation that contains a large number of messages.
2. Scroll through the conversation or continue adding messages.
3. Check the application's memory usage in the system process monitor.
4. Observe that the Codex Renderer processes consume an excessive amount of memory.

The exact conditions that trigger the issue are currently unclear.

What is the expected behavior?

Memory usage should remain within reasonable limits, even when a conversation contains a large number of messages. Content that is not currently visible should not remain fully rendered in memory unnecessarily.

Additionally, the application should not run multiple renderer processes unless they are required.

Additional information

The issue appears to become more noticeable as the conversation grows longer.

Start by reproducing the issue with a conversation containing many messages, then monitor Codex Renderer processes and memory usage in the system process monitor while scrolling and adding messages. The exact renderer entry point and tests are not identified in the issue. Done means memory remains within reasonable limits for long conversations and unnecessary renderer processes are not running.