Not possible to have a queued up prompt

[unphased]

1. If I type a prompt while LLM is thinking and responding, currently it is impossible to submit while the LLM is responding. I'd like to have a choice between submitting what I typed for sending at the earliest opportunity (e.g. after the completion of the next tool call, compare to hitting enter in Codex), or to queue it for the completion of the current task (compare to hitting tab in Codex)
2. If I type a prompt while LLM is thinking and responding, I am unable to submit, and then when Dirac needs to approve a tool call, I am in a terrible situation in which I actually need to nuke what I spent time typing up already just to be able to type a y to approve the permission.

[dirac-run]

good point. i had once implemented something like this to queue the next message, the thing was causing sync issues so i never pushed it.

i will rethink this with the currently in-progress refactor. keeping the task open in case others also want to chime in

[unphased]

Best of luck with the refactor. Looking forward to be able to use this harness soon.