Upcoming Feature: Task Queue for Chat - What use cases are you excited about?

Discussions
1

Hey everyone!

@ericzakariasson shared a sneak peek of an upcoming feature with Queues in Chat:

I’ve had a chance to test this out and it’s really practical for managing multiple requests in Chat where you know upfront things will need to be done. Once the first task finishes the next prompt starts and I remember many on the forum asked for this :slight_smile:

What use cases are you most excited about? I’m thinking this could be great for:

  • Breaking down large refactoring tasks
  • Managing multiple file changes across a project
  • …

What workflows do you think this would help with? Share your thoughts on how you’d use a feature like this!

4 Likes
3

Hopefully we can queue our 4 requests every 5 hours so we don’t have to be at our computer every time the rate limit ends. This way, we might be able to squeeze in 12 requests a day

3 Likes
4

AI generated code is too unreliable to use this feature effectively. I have to test the edits that Sonnet produces after each request and 70% of the time it needs to be reworked or fixed. I can’t imagine refactoring and testing such huge changes (if all tasks in the queue are related to the same feature)

5

One thing I’d love to have is something akin to being able to write a “manager” prompt that can respond on my behalf.

There are a lot of times the agent will pause and say “Hey now you can run the tests”

And I have to say “NO YOU DO IT!”
Or it will say “The next step is to delete the unused code”
And I have to reply “Ok go”

If I could write a manager prompt with a few bullet points on how to keep it moving along vs. when to interrupt me, that would go a long way!

6

I will create a progress track file and a task file.
Then, I spam “Read the docs and do the tasks” and hope it works :skull:

7

@venkatd you can add that into your user rules with a simple statement, instructing Agent to Run the tests directly without asking, adding the test command to auto-run allowlist etc.

Same with not finishing tasks, instruct Agent to plan steps to complete the task and work on it until completion.

@thepKz thats great, yes its good to have the task and progress/track file for AI to keep track what it did. I’m doing the same!

@DevSalem most of that can be prevented, but it depends also on the language, project code so far, separation of concerns etc… On well structured projects I dont see this issue anymore.

The tasks in the queue do not have to be related to same feature, that depends on your anticipation of tasks :slight_smile: so e.g. after finishing main task you could ask for a cleanup of another area, or to run tests and ensure its all working.

1 Like