Currently a user can delete an entire conversation thread. We want to add the ability to edit thread history to delete any particular compbination of a user prompt and assistant response.
Context
For context, one use case here is when a model goes seriously off the rails and you have to rein it back in, but don't need or want that side quest polluting the context.