›_ Codex Issues
Codex Issues / #2617

Task Queuing

Resolved 💬 2 comments Opened Aug 23, 2025 by mpelteshki Closed Sep 3, 2025
💡 Likely answer: A maintainer (davidgilbertson, contributor) responded on this thread — see the highlighted reply below.

What feature would you like to see?

When working with codex, I often come up with new ideas/instructions while codex is still working. It would be great to have it as an option to not interrupt it, but just add another task to the back of its queue/todo list.

Are you interested in implementing this feature?

_No response_

Additional information

_No response_

View original on GitHub ↗

2 Comments

davidgilbertson contributor · 1 year ago

Love it, I came here to log exactly this.

My analogy: with a messaging app, if I ask a colleague to do something, I don't need to wait for them to finish the task and return control to me in the chat window; I can easily add another message like "oh also, make the play button left aligned".

I can imagine two levels of implementation:

  • Easier: the model always finishes what it's working on, then checks to see if anything was added to the TODO list while it was working. If so, it does it before returning to the user.
  • Harder (more like a human interaction): if the user adds a new message while the model is working, it can change the course of what it's currently doing ("actually, leave the button as is for now")

Perhaps the harder one is best handled by the current 'interrupt' UX, and the first one can be handled by a new 'TODO queue' mechanism.

tibo-openai collaborator · 11 months ago

This was implemented around 0.25 or 0.26, please install the latest version and file another issue if it doesn't work well, thanks for the feature request!