Repository navigation
Replies: 5 comments
|
Once the sidebar is a trustworthy attention queue, I also need to traverse that queue without scanning and clicking it. I often have agents across several hosts in either Needs input or Ready to review. Today I visually scan the sidebar, open one, return, and repeat; when the sidebar is hidden the workflow disappears entirely. The flow I want is a rebindable “next workspace needing attention” shortcut that walks the existing status order, wraps, and opens the relevant session. Its target set should stay current across connected hosts even when the sidebar is not mounted. This is desktop/web keyboard behavior and provider-neutral: it should consume the same workspace status model this discussion is trying to make reliable. It belongs in core navigation/focus handling; a plugin could render another queue, but should not have to duplicate the authoritative target ordering. This adds my concrete high-concurrency use case, but I cannot claim demand beyond the participants already here. PR #2521 is a prior exploration of the shortcut. |
|
Strong +1 on the inbox model. I routinely run ~100 sessions, and at that scale the usual workarounds — labels-as-read-state and archive-as-not-now — fall apart completely, so a real triage layer is the difference between the sidebar being useful and being noise. Concrete prior art worth studying: T3 Code's Sidebar V2 ships almost exactly this, and it's the same product category (agent-orchestration sidebar), so the design decisions transfer directly. The key thing it gets right is using two complementary axes rather than a single snooze:
That second axis is a direct answer to the "let genuinely new activity break through the snooze" point in the original post — it's a separate mechanism, not a flag on snooze. It also sidesteps a failure mode that bites hard at high session counts: if you only have time-based snooze, batches that expire together ("tomorrow 9am") recreate the exact overload you snoozed to escape. Activity/inactivity-based settle staggers re-entry naturally. Suggested state model, borrowing what works there: active / snoozed / settled / restored, with settled items re-entering as unread / needs-review so you actually notice they came back. And strongly agreeing with @colonelpanic8 — at ~100 sessions none of this is usable without a keyboard "next needs-attention" traversal and bulk actions; scanning-and-clicking doesn't scale. On the plugin/core boundary, since it'll come up: I looked at what the current plugin API (v0.8) can do. A plugin can render a parallel triage panel and keep its own snooze/unread/settle state via daemon-side RPCs — but it can't integrate into the native sidebar's ordering, can't control read/unread on native rows ( Reference write-up of the T3 Code behavior: https://www.volanea.com/blog/t3-code-sidebar-v2 (can add screenshots of my own sidebar — the snooze menu and a restored-on-activity settled thread — if that'd help.)
|
|
The Mark as unread half of this is shipped. #3603 is in v0.8.0: open the workspace menu on a finished workspace, choose Mark as unread, and the attention dot comes back so a quick glance no longer drops it out of your queue. Reopening the workspace or interacting with the chat clears the mark again, so it behaves like the completion attention you already have. It needs the app and the daemon both on 0.8.0. Snooze is not built, and neither is colonelpanic8's "next workspace needing attention" shortcut. Both are still open here. |
|
For the keyboard traversal raised here, #6586 removes the nine-workspace limit from Previous workspace and Next workspace. After the next release, you can use those actions to walk the full visible sidebar in its displayed order, including when grouped by status. They still visit every visible workspace, so this is an alternative to clicking each row, not the requested shortcut that selects only workspaces needing attention. Snooze is also unaffected. Thanks for describing how you review many concurrent sessions. The fix is merged and will be in the next release. The current release, v0.11.2, does not contain it, and the next version number is not decided yet. Watch the Releases page if you want it sooner; changes often appear in a beta before the stable release. |
|
The mark-unread part of your inbox workflow gets a reopening fix in #7608: after marking a workspace unread and restarting the app, returning to it will load the chat and current history instead of showing extra Files and Changes tabs with the chat stuck loading. This addresses returning to deferred review; the merge adds no snooze or attention-only navigation. Thanks for describing how you use the review queue. The fix is merged and will be in the next release. The current stable, v0.12.0, and latest beta, v0.13.0-beta.1, do not contain it; the next version number is not decided yet. Watch Releases if you want it sooner, since changes often appear in a beta before the stable release. |

Uh oh!
There was an error while loading. Please reload this page.
I often use Paseo with many workspaces/agents running in parallel, and I increasingly treat the workspace sidebar as an attention queue / inbox, not just a list of sessions.
That creates two closely related workflow problems for me.
1. Opening a workspace is not the same as reviewing it
Sometimes an agent finishes and the workspace correctly shows that it needs my attention. I click it only to quickly glance at the result, but opening it clears that attention/read state.
At that point the workspace visually looks as if I have already reviewed it, even though I may have only looked at it for a few seconds and still want to come back to it later.
What I would like is the equivalent of Mark as unread / Mark as ready to review: a way to put that finished workspace back into my review queue.
I currently approximate this with labels, but I think labels and read state represent different things:
The Gmail mental model is useful here: labels and read/unread are intentionally separate concepts.
This seems closely related to previous work in Paseo:
The workflow described in #3603 is almost exactly the problem I am seeing in daily use.
2. Sometimes I intentionally do not want to see a workspace until later
A second case is when I consciously decide: “I will review this tomorrow” or “come back to this later today.”
I do not want to archive the workspace because it is not finished from my perspective, but leaving several such workspaces in the main sidebar makes the active attention queue noisy.
For this, Gmail's Snooze is a very natural model:
There has already been substantial work exploring this exact direction:
I especially like the workspace-level semantics described in #2525: snooze should not stop agents or hide data; it should primarily change where the workspace appears in the attention UI, with new important activity able to surface it again.
Why I think these belong together
To me these are both parts of the same workspace triage / inbox workflow rather than two unrelated features.
There are several orthogonal states:
Keeping these concepts separate would make the sidebar much easier to use when orchestrating many concurrent agents.
If scope needs to stay small, Mark unread / Ready to review feels like the smallest and highest-value first step because it directly complements the existing Mark as read behavior. Snooze could then be a broader follow-up built on the same inbox/triage mental model.
I am mainly interested in discussing the workflow and shared need here rather than prescribing the implementation. The existing PRs above already contain useful implementation experiments, but the user problem for me is simply: help me maintain a trustworthy queue of workspaces that still need my attention, without using labels as a substitute for read state or archive as a substitute for “not now.”
All reactions