Feature request: Context filters.

[agrimes]

I'm trying a micro model, Llama3.2 3B, it basically works but spews some junk into the context that would be easy to grep / sed out of the context window that aren't being caught by lm studio. So the user should be able to set up "MUTE" strings which are simply removed from the stream, and "TERMINATE" strings that cause processing to stop.

[Harald B]

The new update (currently on the beta branch, live in a few days) will let you select llama3 from the chat/instruct template dropdown. That should fix this.