I have a feature request to keep a petsistent chat open instead of making new chats per request which doesn’t seem to work as well for me.
We should be able to have control over context summarization for example once it gets 80-85% it could summarize or condense the context in the same chat instead of wiping it out to 5% like it does now. This would have tunables or some rules of what to cut out like log files, etc. It would prioritized learned (like learned about project or learned from user) and would add to rules/memories.
Potentially this would allow the chat window to be open forever like in most chat applicaitons and would automatically archive it after summarization (like export it to a chatlog and keep it and then remove it from memory because it was summarized thus keeping the chat scrollback low enough that it doesn’t cause lag)
The idea is so i can have multiple chat tabs open with different models. I’m not sure how the editing would work if they both edited a file at the same time. I have had another feature request in for a while to have a control model that presents the user input to multiple models and looks at all their answers and picks the best one (would be cool! )