Replies: 2 comments
|
a common workflow that would benefit from this: I build a feature with Codex These are not the same thread, but should be grouped together, at least optionally so |
0 replies
|
This would be quite cool, the chat is purely a convosation and the claude and sol comunitcate across it big feature but very intresting if done right |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Before submitting
Area
apps/desktop
Problem or use case
I frequently have projects that end up spanning multiple threads. for example, I will review the same project with claude and codex, or I will have a new thread after to monitor a rollout.
I want to be able to group these and filter them together, so I can see them all in one place and see the full history.
Proposed solution
Threads should have metadata based tags, both human and AI generated. At each human input turn a cheap model should check it out and auto assign new tags if needed. They should be done datadog style so a thread tagged branch:greg/my-project can be grouped together with "branch"
New threads should then also inherit the tags from the old thread if you create a new thread using the new thread on branch tool.
Why this matters
Right now I have 15 threads open, one project has two different threads, it would be nice to group them together. Perhaps my usage is a little odd, but I think this is a great solution for the problem of structured search in general. Tags are a great way to do this kind of semi structured search data.
Smallest useful scope
Just add tags and let us search on them. I'd love to be able to group on them too. I think it's actually a fairly big change (sorry theo!).
Alternatives considered
Generally, unstructured search is nice but doesn't allow grouping, and because it's unstructured and (at least for me) I don't always remember what the name of a branch was, it's much harder to find the two or three threads that were all for the same feature or bug.
Risks or tradeoffs
I think the biggest risk is that the AI generated tags may end up being shitty and it would be better to do it in a more structured way, just group by branch and human generated for example.
Examples or references
I have a mostly complete first pass here, I was going to submit a PR but the instructions said not to unless invited, so I submitted an issue instead: https://github.com/gmmeyer/t3code/tree/agent/thread-tags-beta
Contribution
All reactions