OOM issues cause auto-completion and chat response to stop

[hazratisulton]

Preconditions:
Run with a large context or little amount of memory

Steps:
Start to chat or trigger an auto-completion request in IDE
Wait for the system to respond

Expected result:
The system should respond correctly

Actual result:
The system may stop responding and not indicating an out-of-memory issue in IDE.
Auto-completion requests or chat responses may be incomplete.

[olegklimov]

What can we do with this?

Memory warning when adding a model.

Clear message un UI about a past OOM event.