Repository navigation
feat(vscode): show todo progress in the session dock - #14813
Conversation
Code Review SummaryStatus: No Issues Found | Recommendation: Merge Incremental review of Verified the previous inline findings against current HEAD: all five are already resolved in earlier commits (the No new correctness, security, performance, or resource-leak issues found. The Files Reviewed (5 files)
Previous Review Summaries (6 snapshots, latest commit 3518613)Current summary above is authoritative. Previous snapshots are kept for context only. Previous review (commit 3518613)Status: No Issues Found | Recommendation: Merge Incremental review of commit The prior inline findings are outdated at HEAD (fresh GitHub state shows No new correctness, security, performance, or resource-leak issues found. Files Reviewed (3 files)
Previous review (commit 710506a)Status: No Issues Found | Recommendation: Merge Incremental review of commit The refactor itself is clean:
No new correctness, security, performance, or resource-leak issues found. Files Reviewed (7 files)
Previous review (commit e99ecbc)Status: No Issues Found | Recommendation: Merge All five findings from the previous review are still resolved at HEAD ( Verification of the previously-flagged lines against current code:
Files Reviewed (6 files)
Previous review (commit 1125890)Status: No Issues Found | Recommendation: Merge All five findings from the previous review are resolved at HEAD (
No new issues were found in the incremental changes. Files Reviewed (10 files)
Previous review (commit c74c8e2)Status: 5 Issues Found | Recommendation: Address before merge Overview
Issue Details (click to expand)WARNING
SUGGESTION
Files Reviewed (7 files)
Incremental diff since Fix these issues in Kilo Cloud Previous review (commit 8b144bb)Status: 5 Issues Found | Recommendation: Address before merge Overview
Issue Details (click to expand)WARNING
SUGGESTION
Files Reviewed (38 files)
Reviewed by deepseek-v4.1-flash · Input: 79.8K · Output: 21.4K · Cached: 715.6K Review guidance: REVIEW.md from base branch |
|
@marius-kilocode what is this file about? packages/kilo-docs/public/img/screenshot-tests/kilo-vscode/visual-regression/chat/chat-view-session-dock-todo-states-chromium-linux.png not listed in the summary but is that Review & Support overlapping with details meant to be like that? |
WebReflection
left a comment
There was a problem hiding this comment.
Love the feature but I've disabled auto-merge since this file in particular doesn't look right to me https://github.com/Kilo-Org/kilocode/pull/14813/changes#diff-0cb3d75767d60566de4125ea477da9f10acce020d47085824da73767c2a52c1e ... is that's meant to be like that though, not blocking, feel free to merge
|
@WebReflection this is primarily due to the environment of the screenshot tests. Primarily I look in those for the todo ui being correctly rendered, but your right it could be less noisy. |
- The finished label no longer carries a strike when the chip settles. - Remove the unused session prop and data-session attribute. - Share one run-boundary helper between the agent stack and the todo chip. - Plan the dock fit once per frame, reading before writing layout. - The todo-states story renders only the session dock, so the docs screenshot drops the welcome chrome and the busy header skeletons.
|
@WebReflection I simplified the test to only render todo related ui |
What Problem This Solves
The to-do list showed as a bar under the session title, away from the prompt. Background agents and goals already moved to the session dock above the prompt, so the to-do progress was the only run status left at the top.
Why This Change Was Made
The dock now shows one row for the run: background agents on the left of the spinner ("who works") and progress on the right ("how far"), first the to-do chip and then the goal. The to-do bar under the title and its revert-to-checkpoint behavior are removed.
The chip shows a ring with one segment per item, the done count, and the current item. While a turn runs it trails the spinner. When the session is idle it sits next to the session actions, so an unfinished list stays visible. Click the chip to open the full list, which is read-only.
Behavior details:
The list is read-only. The old header list reverted to a checkpoint when a done item was clicked, but it matched items by position and reverted to the to-do write before the one that completed the item, so it often landed on the wrong item when an agent marked several items done at once. Checkpoints stay available in the transcript.
User Impact
To-do progress is next to the prompt, together with the spinner, background agents, and the goal. The session header is shorter.
Evidence
Real VS Code self-test with DeepSeek V4.1 Flash in Agent Manager, with a background subagent running and a 4-item list. While working, the chip trails the spinner and shows the current item:
Mid-run progress:
The last item finishes, the chip shows the finished state:
Idle, next to the session actions:
The full read-only list:
Checks on the head commit: test, test-vscode, typecheck, docs-build, Check Links, Check Source Links, Check markdown table padding, Check forbidden strings, Check shared upstream annotations, CodeQL, and Visual Regression Tests all pass.