Skip to content

feat(sidebar): add title-only session cards - #23512

Open
rio-evrard wants to merge 1 commit into
kirodotdev:mainfrom
rio-evrard:feat/title-only-session-cards
Open

rio-evrard wants to merge 1 commit into
kirodotdev:mainfrom
rio-evrard:feat/title-only-session-cards

Conversation

@rio-evrard

@rio-evrard rio-evrard commented Oct 23, 2026 •

Copy link
Copy Markdown

Problem / Motivation

Goal: Let people reduce sidebar session cards to their titles with one optional setting.

The sidebar has separate controls for some card details, but no single title-only mode. People who organize many sessions by title still see repeated agent metadata, timestamps, activity previews, and linked-source chips.

Why it matters

A title-only option makes dense session lists faster to scan and keeps unrelated issue or pull-request references out of screenshots and screen shares.

Not a goal

What changed (motivation → approach → change)

A browser-local Title-only session cards switch now appears under Settings → Chat → Sessions, after Compact empty folders and Full titles on scroll, the other two settings that change how sidebar rows look. It defaults off, updates the mounted sidebar immediately, and persists with the existing chat display preferences.

When enabled, the shared session row keeps its title and hides the agent/timestamp metadata line, the secondary activity or message line, and linked source chips. A session that needs the user (an owed approval, an unanswered question, a waiting loop, an interrupted turn, a blocked or budget-stopped monitor, or an unread result) keeps one icon in front of its title, with the same glyph and label the full card uses, so it does not look idle. A goal loop, workflow run or monitor action lights the title-line live dot; a goal loop whose last turn was interrupted shows the danger loop icon instead. A collapsed conductor whose hidden sessions need the user shows the needs-you icon, and one whose hidden sessions are running shows the live dot. A Crew Mode root whose crew holds something for the user (a local or remote worker waiting on a tool approval, or a ledger item waiting on a person) shows the needs-you icon with the count and label of the full card's crew summary line, which it reads from the same hook, so the two cannot disagree. While a crew member works, local or remote (the worker list's working group, as the summary line counts it), the root lights the live dot. If the worker list cannot be read, the title line shows one fixed-size warning icon labelled with the summary line's error text, instead of stale counts. The full notice and its "Ask the agent" hand-off would push the title out of a narrow row, so they stay on the full card and in the Crew panel. Every Crew Mode root keeps its bolt on the title line, labelled with the mode's name, because the agent line that names the mode is hidden. A paused conductor mode crew keeps its Paused icon there, and a crew whose open work all waits in the workers' own loops keeps the running-crew icon. A Slack code channel conductor keeps its folded crew's state icon, its channel badge, its worker fold and its folded workers' most pressing want on the title line, since all of them sat on the hidden agent and status lines. A muted session keeps its labelled mute icon on the title line, because muting also suppresses chimes, toasts and unread markers. The other metadata-line markers (pinned, incognito or temporary memory, process released, location and connected-channel chips) are hidden with that line in title-only mode. The title-line marks live in chat-sidebar/compactTitleMarks.tsx and render beside the title element, not inside it, so the row's double-click rename and the scroll peek read only the title text; and the crew summary line moves unchanged into chat-sidebar/CrewRootSummaryLine.tsx with its needs-you count as a shared hook, which keeps SessionRow.tsx under the 2,000-line cap. The same row component supplies list and board views, so both receive the same presentation without changing stored session data or provider refresh behavior.

The generated documentation registry (src/kiro_crew/docs/settings-registry.generated.json), command-palette registry (website/src/components/commandPalette/settingsRegistry.gen.ts), generated UI index, supported locale catalogs under website/src/i18n, and focused tests describe and lock in the preference. This is broader than #21925, which hides only ordinary message previews while preserving the remaining card status and metadata.

Backwards compatibility

Compatible: the new browser-local preference defaults to false, so existing installations retain detailed session cards.

Reader: website/src/pages/ChatSidebar.tsx:ChatSidebar -- chat -- reads the preference and passes it through the shared row shell.
Reader: website/src/pages/chat-sidebar/SessionRow.tsx:SessionRow -- chat -- hides only rendered card details when the preference is enabled; slot payloads and row actions are unchanged.
Reader: website/src/pages/settings/ChatPanel.tsx:ChatPanel -- settings -- writes the preference through the existing mc-chat-config path.

Tests

  • ChatSettings.test.tsx covers the detailed default, stored opt-in, and repair of invalid stored values.
  • ChatSidebar.sourceLinkChip.test.tsx verifies that titles remain while the activity preview, timestamp, and source chips disappear; that a pending approval, an unread result and a muted session each keep a labelled title-line marker; that a workflow run and a goal loop light the live dot; that a stalled goal loop shows its interrupted marker and no live dot; that a collapsed conductor with a hidden session needing approval shows the needs-you icon, alongside the live dot when the conductor itself is working; that a Crew Mode root shows its crew's needs-you count for a local worker's approval, a remote worker's approval and a ledger item waiting on a person, and its Crew Mode mark; that a working crew member, local or remote only, lights the root's live dot; that a failed worker-list read shows the warning icon and keeps the title; that a paused conductor mode crew keeps its Paused mark; that a running crew keeps its running-crew marker; and that full cards get none of these extra cues. The mute and Crew Mode tests fail with their SessionRow.tsx branches removed.
  • ChatSidebar.codeChannelCrew.test.tsx verifies that a title-only code channel conductor keeps its crew state, channel badge and worker fold on the title line. This test and the paused and running-crew tests fail with the fix removed.
  • ChatSidebar.codeChannelCrew.test.tsx also checks that double-clicking a title-only card's worker fold does not start a rename; that test fails with the fix removed.
  • The three files on the final commit: 125 tests passed, and Vitest exits 0 (the list mocks now return lists). The remote-worker, ledger, live-dot, remote-working and failed-read tests fail with the fix removed. The seven sidebar and crew test files that cover these rows: 277 passed.
  • scripts/check_file_line_cap.py passes.
  • TypeScript, targeted ESLint, i18n:check and git diff --check passed; regenerating the settings registry and UI index produced no changes.

Manual verification

Verified in an isolated local dashboard profile:

  1. Detailed cards showed agent metadata, timestamps, message previews, and source chips.
  2. Enabling Settings → Chat → Sessions → Title-only session cards reduced each card to its title.
  3. Refreshing preserved the preference.
  4. Disabling the setting restored detailed cards.

The screenshots use synthetic sessions and links. They show the title-only layout but not the title-line icons, which were added later and are covered by the tests above.

Screenshots / video

Detailed cards Title-only cards
Detailed session cards Title-only session cards

Title-only session cards setting enabled

Related Issues

Part of #21239

Follow-ups left out of this PR: #29861 (collapsed-parent live-dot test), #29862 (pinned, incognito/temporary, popped-out and released icons), #29863 (setting description copy), #29864 (history.md spec paragraph).

Related overlap: #21925 provides a narrower preview-only setting and intentionally keeps status indicators and other card details visible.

Checklist

  • One commit with a Conventional Commits title
  • Existing focused tests pass and new tests cover the preference
  • Self-review completed; code follows project style guidelines
  • Documentation and generated settings metadata updated
  • No secrets, credentials, or internal references in the diff

Contribution License Agreement

Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.

@rio-evrard
rio-evrard marked this pull request as ready for review October 23, 2026 18:04
@rio-evrard
rio-evrard requested review from a team and KiroCrewCaptain as code owners October 23, 2026 18:04
@rio-evrard
rio-evrard requested a review from buluoray October 23, 2026 18:04
@github-actions github-actions Bot added fork Pull request from a fork (external contributor) readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Oct 23, 2026
@github-actions

github-actions Bot commented Oct 23, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5, fork) — ✅ PASS (a BLOCK with no Trigger and no Outcome does not block)

UX-level review of 41d34ec9ee29434fa1ddf4991734e7666f86db21 via the fork AI-review pipeline — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

UX-Verdict: PASS

cannot evaluate: missing the title-line marks. The description says the screenshots "show the title-only layout but not the title-line icons." Those icons are the only cue left that a title-only session needs the user or is running.

Evidence gaps

  • Needs-you marks (compactAttentionMark: approval shield, question, loop-waiting, interrupted triangle, stalled goal, blocked monitor, unread dot) beside a title. No screenshot shows them. To close: a title-only sidebar with at least one session in a needs-you state and one unread.
  • Live dot on a running goal loop, workflow, or collapsed conductor in title-only mode. To close: a title-only sidebar with one running session.
  • Crew Mode root marks (CompactCrewMarks: Zap bolt, Paused, Hand "N need you" count, worker-list error triangle). No screenshot shows them. To close: a title-only Crew Mode root, with the needs-you and paused states.
  • Code channel conductor in title-only mode (crew state, channel badge, worker fold on the title line), and a muted session's mute icon. To close: one screenshot of each.
  • How several marks line up on a narrow row (bolt + paused + hand + live dot ahead of a long title). To close: a narrow-width title-only screenshot with a long, truncated title.

Notes

  • compact_session_cards_desc says different things in different languages: English lists "metadata, activity previews, and linked source chips", while bn and de mention only previews and chips. Make every locale name the same three hidden items.
  • The always-on Zap bolt on every Crew Mode root is an icon with no visible text (its label "components.crewMode.eyebrow" is only a hover tooltip). Show its meaning in a screenshot, or confirm it matches the bolt the full card uses, so a newcomer can identify it.

[UX-REVIEWED] 41d34ec

Description read

[DESCRIPTION-READ] bb1df052b12a68dc9806e7d3a6e0002ddfea023884ea9969707c3e952b49731a

That sha256 names everything this verdict read: the PR description, plus the 6 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 23, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5, fork) — ✅ PASS

Premise-level review of 41d34ec9ee29434fa1ddf4991734e7666f86db21 via the fork AI-review pipeline — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

First-Principles-Verdict: PASS

The opt-in setting has a linked issue (#21239) and changes no default. The one weakness is that the title-line marks spell out each row state's glyph and label a second time, apart from the full card's resolver.

Notes

  • Item 3 / Item 7 — duplicate of website/src/pages/chat-sidebar/SessionRow.tsx resolver entries and CodeChannelCrewLine: compactAttentionMark restates glyph and label for 9 resolver keys (I counted code_channel_crew, pending_approval, subagent_awaiting, needs_input, loop_waiting, goal_loop stalled, interrupted, structured_monitor_passive, crew_running). Each of those already has a build: entry in the row-state resolver. codeChannelCrewMark restates CodeChannelCrewLine's 4-state table in the same file. The description promises "the same glyph and label the full card uses", but nothing ties the two copies together, so they can drift apart. Subtraction: delete the per-key table in compactAttentionMark and codeChannelCrewMark, and take each mark's glyph and label from the resolver entry or line that already defines it.

What this change ships

Inventory (10 items) — 8 justified

Intent: let people shrink sidebar session cards to their titles with one optional setting. ADDITION.

  1. New Settings → Chat → Sessions toggle "Title-only session cards", off by default — justified
  2. With it on, rows hide the agent/timestamp line, the activity line and the source chips — justified
  3. An owed, blocked or unread session shows one labelled icon in front of its title — duplicate of SessionRow resolver glyph/label entries
  4. A goal loop, workflow run or monitor action lights the title live dot in this mode — justified
  5. A collapsed conductor shows its hidden sessions' needs-you icon or running dot — justified
  6. A Crew Mode root shows bolt, Paused, needs-you count, live dot and read-error icon — justified
  7. A code channel conductor keeps its crew state, badge and worker fold on the title line — duplicate of CodeChannelCrewLine state table
  8. A muted session keeps its mute icon on the title line — justified
  9. The crew summary line moves to its own module with a shared needs-you hook — justified
  10. Settings/palette/UI-index registries, 12 locales, a feature-map row and committed screenshots — justified

[FIRST-PRINCIPLES-REVIEWED] 41d34ec

Description read

[DESCRIPTION-READ] 6a4827280694695253e394bd964ee678936860567e9956c98cd15904de05fea4

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions

github-actions Bot commented Oct 23, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5, fork) — ✅ PASS

Design-level review of 41d34ec9ee29434fa1ddf4991734e7666f86db21 via the fork AI-review pipeline — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks: the preference defaults off and stays in the browser, and no reader outside the patch loses anything.

Notes

  • compactAttentionMark is a second copy of how the row-state resolver maps each state to a glyph and label. The description promises the title-line mark uses "the same glyph and label the full card uses", but nothing ties the two together. When the resolver changes a state or adds one, the title-only card will quietly drift or show no mark. Have the resolver return its marker and label once, and let both card modes read that.
  • The committed screenshots show the title-only layout but none of the title-line marks, which the description says came later. A screenshot of the mark states (needs-you, live dot, crew bolt, muted) would let a reviewer judge how the title line looks.

[DESIGN-REVIEWED] 41d34ec

Description read

[DESCRIPTION-READ] 3a4cf0dd46dd561dc1f547afd514177eb627b5c9cac1e5ac37530f09b432d7a5

That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 23, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review (fork) — 🔴 changes requested (blocking)

Reviewed 41d34ec9ee29434fa1ddf4991734e7666f86db21 via the fork AI-review pipeline; updated in place on each push.

Adjudication upheld 1 of 1 blocking finding(s); those must be resolved before merging.

BLOCKING -- website/src/pages/chat-sidebar/compactTitleMarks.tsx:99 -- Compact crew errors lose readable notice and recovery action
<span className="inline-flex items-center align-middle mr-1.5" data-testid="session-compact-crew-error" role="img" aria-label={i18nT('commandCenter.workers_error')} title={i18nT('commandCenter.workers_error')}>
Trigger: With title-only cards enabled, a Crew Mode root’s worker-list request returns HTTP 500; readCrewWorkers → useCrewWorkerSummary sets failed, and SessionRow renders CompactCrewMarks instead of CrewRootSummaryLine.
Outcome: The sidebar replaces its readable failure notice and “Ask the agent” action with a noninteractive icon whose explanation exists only in attributes, leaving touch users without readable failure text or recovery on this surface.
Anchor: errors-use-error-notice
Fix: Render the failed-read branch through an inline ErrorNotice with its hand-off decision, while keeping ordinary metadata and summaries hidden.

[BLOCK-MERGE] 41d34ec
[GPT-REVIEWED] 41d34ec

Adjudication (Opus 5.5) — is blocking on each finding proportionate?

F1: the rule this finding is anchored to, errors-use-error-notice in .review-base-rules/website-AUTOSDE.yaml:675-676, has blocking: true, so the verdict is UPHOLD without weighing it further. The diff's compactTitleMarks.tsx:99 (the patch hunk at authentic.patch:919-923) shows a failed worker-list read, useCrewWorkerSummary's failed, as a hand-built role="img" TriangleAlert. It does not use ErrorNotice/askAgent, unlike the full-card path at CrewRootSummaryLine.tsx:499. The real fix is small: a compact inline ErrorNotice in that branch.

[ADJUDICATION] 41d34ec9ee29434fa1ddf4991734e7666f86db21 total=1 uphold=1 downgrade=0
UPHOLD F1 website/src/pages/chat-sidebar/compactTitleMarks.tsx:99 reason=autosde-blocking-rule
[GPT-ADJUDICATED] 41d34ec9ee29434fa1ddf4991734e7666f86db21

@github-actions

github-actions Bot commented Oct 23, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review (fork) — ✅ no blocking findings

Reviewed 41d34ec9ee29434fa1ddf4991734e7666f86db21 via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 41d34ec

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 23, 2026
@rio-evrard
rio-evrard force-pushed the feat/title-only-session-cards branch from 37dd33a to 40a3f81 Compare October 23, 2026 21:41
@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 24, 2026
@rio-evrard
rio-evrard force-pushed the feat/title-only-session-cards branch from 40a3f81 to 6b8c6fb Compare October 24, 2026 08:10
@github-actions github-actions Bot removed the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 24, 2026
@rio-evrard
rio-evrard force-pushed the feat/title-only-session-cards branch from 6b8c6fb to 029c8c9 Compare October 24, 2026 09:42
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Oct 24, 2026
@rio-evrard
rio-evrard force-pushed the feat/title-only-session-cards branch from 029c8c9 to 206f141 Compare October 24, 2026 11:52
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Oct 24, 2026
buluoray
buluoray previously approved these changes Oct 24, 2026

@buluoray buluoray left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewed at 206f141dde9f694450069b8a925ca1d034708cb0.

Approving. Title-only cards are opt-in and off by default. A bad stored value is repaired to false (ChatSettings.tsx:178). Re-running npm run gen:ui and npm run gen:settings leaves the tree clean. i18n:check passes. Committing the evidence under temp-screenshots/ is the documented path for a fork PR, and MANIFEST.in prunes it from packages.

Verified locally:

  • vitest run src/test/ChatSidebar.sourceLinkChip.test.tsx src/test/ChatSettings.test.tsx: 86/86 pass.
  • tsc -p tsconfig.app.json --noEmit: clean.
  • Mutation checks, one at a time:
    • Removing the repair turns repairs a non-boolean compactSessionCards value… red.
    • Re-showing the meta line or the secondary line turns shows only the title… red (and the approval and unread tests for the secondary line).
    • Disabling the title-line marker turns 5 tests red.
    • Dropping the loop, workflow and monitor live keys turns 3 red.
    • Dropping the stalled-loop exception turns its test red.

Non-blocking follow-ups:

  1. Test gap. The live dot for a collapsed conductor whose hidden sessions are running and that has no marker of its own (showLiveDot's second clause) has no test. Replacing that clause with false keeps every test green. A case with conductor.aggregate.running > 0, needsYou = 0 and title-only on would cover it.
  2. Cues that disappear in title-only mode. Muted, process released, incognito / temporary memory mode, popped out and pinned all live in the meta line, which this mode hides. Muted was deliberately at rest (role="img", labelled "Muted"), so a muted session now reads the same as any other to sighted users and to screen readers. The memory-mode glyphs are privacy cues. Consider keeping these small glyphs on the title line, the same way the attention marker is kept.
  3. Toggle placement. The new toggle sits between "Split view session grid" and "Dim inactive panes", which depends on split view (ChatPanel.tsx:1032-1034). It fits better next to "PR and Issue Chips on Session Cards" (ChatPanel.tsx:1100), the other session-card setting.
  4. Description copy. "Show only each session title" isn't quite true now that a status marker can appear. All ten translated descriptions also drop "metadata", which the English keeps. The UX lane suggested replacement wording; the translations need redoing once the English is settled.
  5. Spec and evidence. docs/system-specs/modules/history.md owns SessionRow.tsx but doesn't mention the title-only mode or its live-dot and marker rule. The UX lane also asked for one title-only screenshot with the approval, unread or interrupted marker and one with a running loop or workflow; the current shot shows only idle titles.

@rio-evrard
rio-evrard force-pushed the feat/title-only-session-cards branch 2 times, most recently from 25f180c to c22d556 Compare October 29, 2026 10:32
@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge and removed readiness: checking Automated validation is still running merge conflict Branch has merge conflicts with its base — author must resolve before merge labels Oct 29, 2026

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @rio-evrard for narrowing this. I reviewed head 550320d82b. Three things still block merge:

  1. Crew Mode approval wait is hidden in title-only mode. compactAttention in website/src/pages/chat-sidebar/SessionRow.tsx has no branch for the crew_approval row state (main sets it at SessionRow.tsx:1146). With {!compactSessionCards && ( wrapping the meta line, the <CrewEyebrow /> is hidden too. Result: a conductor whose crew member waits for approval shows a bare title, and the user never sees the wait. Fix: add a rowStateEntry?.key === 'crew_approval' branch (warn glyph, label components.crewMode.summary_approval) before subtree_needs_you, keep a crew marker on the title line when crewRoot, and add a test for both. This is the open Opus finding from 10-30.
  2. Merge conflict with main. Please rebase onto current main.
  3. Follow-ups need tracking. @iamwhatever's change request asked for each remaining item to be fixed here or filed as an issue linked in the description. Please open one issue per item you listed (collapsed-parent live-dot test; pinned / incognito / temporary / popped-out / released icons; setting description copy; history.md spec paragraph) and link them in the PR description.

After that, CI and the AI lanes will run on the new head.

AI-assisted review (code-review-sage), sent by CommunityCrew

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @rio-evrard. I re-reviewed head f36df8e29b against my change request from 11-04.

Fixed:

  1. Crew Mode approval wait. compactAttention now has a crew_approval branch (warn glyph, components.crewMode.summary_approval), and a Crew Mode root keeps its bolt on the title line. The new test keeps a crew member approval wait and the Crew Mode mark on a title-only card covers both.
  2. Follow-ups. #29861, #29862, #29863 and #29864 exist and are linked in the description.

One thing is left:

  1. Merge conflict with main again. main moved after your 01:32 UTC rebase. The only conflict is the generated file src/kiro_crew/docs/ui-index.generated.json. Please rebase onto current main (or merge it), then regenerate that file with npm run gen:ui instead of hand-merging it, and push.

Once that is pushed, I'll approve on the new head, which queues its fork CI.

AI-assisted review, sent by CommunityCrew

@bolichen97

Copy link
Copy Markdown
Contributor

[operator: bolichen97] Maintainer decision: @rio-evrard, this PR stays yours until 2026-11-12 12:30 UTC. If there is no new push or reply here by then, a maintainer will continue it in a new PR that builds on your commit and credits you as co-author.

main has moved since the 2026-11-05 review: the rebase now conflicts in five files, and title-only mode has to cover what main added to session rows, including: the code_channel_crew row state has no title-line marker, the code-channel fold toggle and worker marks (CodeChannelRowMarks) sit on the agent line that title-only hides, crew_running has no title-line cue, and an Ultra Mode crew's Paused line is hidden.

There has been no activity here for a while. If you are still working on this, please leave a comment; otherwise a maintainer may pick it up so that it does not stall.

@rio-evrard

Copy link
Copy Markdown
Author

@bolichen97 still mine, thanks. Rebased onto main and pushed ba0ac25d. Title-only mode now covers the four cases you listed: the code_channel_crew state, the code-channel badge, fold and worker marks, crew_running, and an Ultra Mode crew's Paused state. Each has a test.

@KiroCrewCaptain

KiroCrewCaptain commented Nov 11, 2026 •

Copy link
Copy Markdown
Collaborator

Code change approver recommendation: A person should decide

Why: the decider thinks the change is probably fine but estimates a 15% chance that this change introduces a defect, above the 12% limit for approving on its own.

Opt-in, off-by-default title-only session cards. The value is stored in mc-chat-config (non-booleans are reset to false), propagated via the existing mc-config-changed listener in useBoardColumns, and carried through RowShell to list and board rows. Every new SessionRow branch is gated on compactSessionCards, so full cards render as on main. The CrewRootSummaryLine extraction is a straight move with its needs-you count factored into a shared hook (same inputs), and the line keeps ROW_STATUS_LINE_CLS. The earlier blocking finding (marks inside [data-session-title], so a double-click started a rename and the scroll peek copied mark text) is fixed by CompactTitleRow placing marks beside the title, with a dblclick test. Owed, unread, muted, goal-loop/workflow/monitor-action, hidden-subtree, Crew Mode and code-channel states each get a labelled title-line mark, with focused tests. Failed CI checks are rate-limit or infrastructure failures, not code verdicts.

An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit a48f2c7194 · 797 changed lines. If this recommendation is wrong, reply here: replies are how the approver is evaluated.

@KiroCrewCaptain KiroCrewCaptain left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  • Blocking: title-only mode still hides two Crew Mode approval states.

Trigger: a remote crew worker reports group: "approval" and status: "needs_approval" without a local slot, or the root is running with crew_work.person > 0.

Outcome: the full card's CrewRootSummaryLine reports “1 needs you,” but compactAttention only checks local crewWaiting and the row resolver. The title-only card therefore shows no marker, or only a running dot. A required approval looks idle or still working, which breaks this PR's stated goal.

Please derive the title marker from the same needs-you aggregate used by CrewRootSummaryLine, and cover both remote-worker and root-ledger cases.

@KiroCrewCaptain

Copy link
Copy Markdown
Collaborator
  • Blocking: the new Paused-mark test leaves Vitest with an unhandled error.

ChatSidebar.sourceLinkChip.test.tsx mocks chatTags with {}. The new async findByTestId lets that query resolve while mounted, then tags.ts:36 tries to iterate the object.

I reproduced this with npx vitest run src/test/ChatSidebar.sourceLinkChip.test.tsx: all 56 assertions pass, but Vitest reports TypeError: tags is not iterable and exits 1.

Please return a valid tag list from this mock, then verify the test command exits 0.

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Requesting changes on 4c53c4f982, @rio-evrard. Thanks for the rebase. Both of Crew Captain's blockers are fixed on this head:

  • Needs-you on title-only cards: CompactCrewMarks and CrewRootSummaryLine now read one count, useCrewRootNeedsYou. A remote worker in the approval group and a ledger item on a person both show the crew_needs_you marker, and the new tests cover both cases.
  • Vitest exit code: the chatTags and chatFolders mocks now resolve to lists. Here, the sidebar, crew and ChatSettings suites pass with exit 0 (197 files, 2515 tests). tsc --noEmit and the line cap pass too, and SessionRow.tsx is at 1,985 lines.

I approved the 17 held fork runs on this head.

One blocker, from CI: Fast Gate's Source Gates job fails the feature-map gate (scripts/check_feature_map.py):

feature-map gate: this change adds or removes a dashboard feature but leaves docs/feature-map/README.md untouched.
  added    website/src/pages/chat-sidebar/CrewRootSummaryLine.tsx
  added    website/src/pages/chat-sidebar/compactTitleMarks.tsx

It reproduces here with FEATURE_MAP_BASE_REF=origin/main python3 scripts/check_feature_map.py. The fix is one row in docs/feature-map/README.md for Title-only session cards: what it does, the path (Settings → Chat → Sessions → Title-only session cards), and the frontend files (pages/chat-sidebar/SessionRow.tsx, pages/chat-sidebar/compactTitleMarks.tsx, pages/chat-sidebar/CrewRootSummaryLine.tsx, pages/settings/ChatPanel.tsx, pages/chat/ChatSettings.tsx). There is no backend handler. The fork AI lanes start only after Fast Gate passes, so they have not run on this head yet.

AI-assisted, sent by CommunityCrew

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Requesting changes on 45d848f4dc, @rio-evrard. The feature-map row is in, and Fast Gate passes now. Opus 5.5, First Principles, Design, UX, Security Scope and Internal Content Scan pass on this head. I approved the 17 held fork runs on it.

Three things are left:

  1. Merge conflict with main. GitHub reports the PR as conflicting. The only conflict is in src/kiro_crew/docs/ui-index.generated.json, because main moved after your rebase. Rebase onto main and run npm run gen:ui in website/ to regenerate the file; don't hand-merge it.

  2. GPT 6.1 blocker, upheld: a failed worker-list read is silent on a title-only card (compactTitleMarks.tsx:69). CompactCrewMarks takes only lastWorkers and paused from useCrewWorkerSummary. When the read fails (failed: true), the card shows stale counts or nothing. The full card's ErrorNotice lives in CrewRootSummaryLine, which is not mounted in title-only mode. That breaks the blocking errors-use-error-notice rule. Read failed and show the existing inline worker-list ErrorNotice (with its hand-off), as CrewRootSummaryLine does.

  3. GPT 6.1 blocker, upheld: a running remote worker shows no activity (compactTitleMarks.tsx:77). The working prop comes from crewWorking in SessionRow.tsx (line 399), which looks only at local slots. A root whose only running worker is remote (group: 'working' in the worker list) shows neither the live dot nor a running mark, while the hidden full card counts it as running. Take the running count from the merged list inside the component, as CrewRootSummaryLine does: crewRowCounts(summarizeCrewWorkers(lastWorkers)). Keep the current precedence (pause, then the needs-you mark, then a single dot).

Please add a test for each blocker: a failed crewWorkers read shows the notice on a title-only root, and a remote worker in the working group lights the root's dot. The frontend and backend test jobs were still running when I wrote this.

AI-assisted, sent by CommunityCrew

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Requesting changes on 2cdd30095f, @rio-evrard. Against my last change request (5893572484):

  • Failed worker-list read: fixed. CompactCrewMarks reads failed, drops the stale needs-you count and shows the inline worker-list ErrorNotice. Covered by "says a failed crew read on a title-only card, with no stale count".
  • Running remote worker: fixed. The dot also lights from the merged list's working count (crewRowCounts(summarizeCrewWorkers(lastWorkers))). Covered by "lights the live dot on a title-only card while only a remote crew worker works".
  • Conflict with main: still open. Main moved again, and src/kiro_crew/docs/ui-index.generated.json conflicts once more. Rebase and run npm run gen:ui again.

GPT 6.1, First Principles, Design, UX, Security Scope and Internal Content Scan pass on this head. I approved the 17 held fork runs on it.

New blocker, from Opus 5.5, which I checked on this head: the title-only marks render inside the [data-session-title] element (opened at SessionRow.tsx:1820). That covers CodeChannelRowMarks at line 1869 and CompactCrewMarks at line 1851, including its new ErrorNotice.

  • The row's onDoubleClick (line 1492) starts a rename when the target is inside [data-session-title]. CodeChannelCrewToggle stops mousedown, pointerdown, click and auxclick, but not dblclick. So double-clicking the worker fold, the channel link or "Ask agent" toggles the fold twice and opens rename.
  • The scroll peek reads the title element's textContent, so a clipped title would show the worker count, the channel state word or the error text glued to it.
  • Fix: put the compact marks in a sibling wrapper next to the title element (for example a flex row holding the marks and then the title div), so [data-session-title] holds only the title text. A test that double-clicks a compact mark and checks that rename does not start would pin it.

The frontend, backend and E2E jobs were still running when I wrote this.

AI-assisted, sent by CommunityCrew

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Requesting changes on a48f2c7194, @rio-evrard. Against my last change request (5896014424):

  • Marks inside the title element: fixed. CompactTitleRow now puts the marks in their own span next to [data-session-title], so double-clicking a mark no longer starts a rename and the scroll peek reads only the title.
  • ui-index.generated.json conflict: fixed. The head merges cleanly into current main.

One blocker left, from GPT 6.1 and Opus 5.5 (both upheld), which I checked on this head: a failed crew read squeezes the title out of a title-only card.

  • CompactCrewMarks renders the full inline ErrorNotice ("Could not load the crew." plus "Ask the agent") when failed is true (compactTitleMarks.tsx:97).
  • It sits in the marks span, which is inline-flex shrink-0 (compactTitleMarks.tsx:110), so the span keeps the notice's full width. The notice's own min-w-0 and truncate cannot narrow it.
  • The title next to it is flex-1 min-w-0, so at the 180px minimum sidebar width it gets little or no room, and the row's overflow-hidden clips the button. This lasts as long as the read keeps failing, on every Ultra Mode root row.
  • Fix: in the marks, show the failure as a fixed-size glyph (for example TriangleAlert with role="img" and the commandCenter.workers_error text as its label and tooltip), and offer the hand-off somewhere off the title line. Please add a test that a failed read renders that glyph and keeps the title.

The other red checks on this head are not your code. PR Policy (it could not read the PR body), Dependency License Gate and Await Fast Gate hit the GitHub API rate limit, and Coverage Gate failed closed because the frontend tests were skipped after Await Fast Gate. They will run again on your next push.

AI-assisted, sent by CommunityCrew

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approving 41d34ec9ee, @rio-evrard. Against my last change request on a48f2c7194:

  • Failed crew read squeezing the title: fixed. CompactCrewMarks now shows one fixed-size TriangleAlert glyph (session-compact-crew-error, labelled "Could not load the crew.") instead of the full inline ErrorNotice, so the marks span stays narrow and the title keeps its room. The test checks the label, that the glyph sits in the marks span, that the notice text is gone, and that [data-session-title] still reads "Crew root".
  • The rest of the diff since then is the rebase: the Ultra Mode → conductor mode renames, the work.run.settled verdict for allDone, and the regenerated ui-index.generated.json.

I approved the 16 held fork runs on this head. Auto-merge is on. If E2E, Render Benchmark or i18n Render Gate go red with Playwright does not support chromium on ubuntu26.04-x64, that is the fork runner image, not this PR; it is tracked in #37647 and fixed by #37653.

AI-assisted, sent by Bolin Review bot

@bolichen97

Copy link
Copy Markdown
Contributor

Two older change requests still hold this PR, and both look addressed on 41d34ec9ee:

  • @KiroCrewCaptain (11-11, on ba0ac25d24): the chatTags mock returning {}. ChatSidebar.sourceLinkChip.test.tsx:47 now mocks chatTags and chatFolders with mockResolvedValue([]), so tags.ts gets a list.
  • @iamwhatever (10-24, on 206f141dde): could you re-review on the current head, or dismiss if your points are covered? My approval is on 41d34ec9ee and auto-merge is on.

AI-assisted, sent by Bolin Review bot

@bolichen97

Copy link
Copy Markdown
Contributor

@rio-evrard answering your question from Slack: I won't override GPT 6.1 here as it stands, but you don't need to bring the full inline ErrorNotice back either.

GPT's point holds. On 41d34ec9ee the failed-read state is a role="img" span whose text exists only in aria-label/title. On a touch screen there is no hover, so the person sees an icon with no text and has no way to "Ask the agent", and that breaks the errors-use-error-notice rule. Your point holds too: an inline notice on a title-only card defeats the mode.

Please do both: keep the small warning icon on the title line, and make it a real button (aria-haspopup, keyboard and tap reachable) that opens the existing ErrorNotice with its hand-off action in a popover. The title stays one line, and the error text and recovery are one tap away on every device. Add a test that taps the icon and finds the notice and its action. Then reply on GPT's finding saying the notice is behind the button, so the adjudicator can rule on the new head.

If you'd rather keep the plain icon, the alternative is to show the full card for that one row while its read has failed. That is also fine with me.

AI-assisted, sent by Bolin Review bot

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

fork Pull request from a fork (external contributor) merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: action required A blocking check or review needs attention

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants