As 'Enthusiastic Bee' said, editing the last user input is a must- because so often the LLM gets a simple assumtion wrong and builds a whole response on it. Being able to edit your input and re-roll the response would be a major help.

I am going to add something to this. so there are times where you want to add specific response in the chat, either from your side, or from the ai's side, that is outside of the 'limit'. Currently if we can only increase the limit now to include that specific portion. But it would be kind of nice to be able to select a specific response to be included in the context. I guess right now we could simply turn it into a snip I suppose and the use the context option to select it. So maybe its not needed. Hmm.

Not sure if this is covered already or not, but being able to clone/duplicate chats fits closely with editing. From both a cost standpoint and the limits on length of chats, there are times that a chat will provide feedback on several independent topics. (Think drunk Claude ripping apart a scene.) Being able to retain the initial context while pursuing a separate, more detailed discussion on each subject is desirable.

I connect it with editing chats since most of the time the initial chat will get used to pursue one of the threads. you decide to pursue a second thread, clone the chat, and delete the follow-up conversation specific to the first thread.

This is a must-have. The "Retry" feature on the response rarely gives me anything useful. Being able to edit the prompt before the most recent response would be a knockout feature. I use it all the time with ChatGPT/Claude. You can really home in on the right prompt.

In addition to this, I think it would be useful to be able to name your chats, like with Snippets.

this is very useful, i need to copy the raw content in md format to create the codex cards, the copy button copies the formatted text