feat(dashboard): session scrubber — sticky bottom track, kbd nav, replay #1001

Merged

reviewer merged 2 commits from code-lead/968 into main 2026-05-08 21:52:44 +00:00

Conversation 8 Commits 2 Files changed 4 +1054 -8

code-lead commented 2026-05-08 20:34:26 +00:00

Collaborator

Copy link

Sticky bottom <SessionScrubber> for the run pane: canvas-rendered ticks (one per TaskEvent, ToolKind-coloured, error/approval emphasised, compactions glyphed), hover popover with compact summary, click+drag scrubbing that focuses the matching subagent lane and scrolls the event log to the row.

Test plan

• just qa clean — typecheck + Biome check + Biome format + tests
• 22 new unit tests in session-scrubber.test.tsx: tick projection (xPct, kind, error/approval/compaction tagging, lane assignment, 1 s span floor), prev/nextErrorIdx, findNearestIdx, assignLaneAt, keyboard nav (, / . / [ / ], no-op clamps, modifier-key gating, input-focus exception, onLaneFocus round-trip), ARIA slider attributes
• <EventLog> exposes data-event-idx per row + auto-expands collapsed groups when the scrubber targets an event inside, and honours prefers-reduced-motion (smooth ↔ auto)
• <TaskDetail> flips to the Log tab on tick selection so the scroll target has DOM to land on
• Empty event lists render no scrubber

Closes #968

Sticky bottom `<SessionScrubber>` for the run pane: canvas-rendered ticks (one per `TaskEvent`, ToolKind-coloured, error/approval emphasised, compactions glyphed), hover popover with compact summary, click+drag scrubbing that focuses the matching subagent lane and scrolls the event log to the row. ## Test plan - [x] `just qa` clean — typecheck + Biome check + Biome format + tests - [x] 22 new unit tests in `session-scrubber.test.tsx`: tick projection (xPct, kind, error/approval/compaction tagging, lane assignment, 1 s span floor), `prev/nextErrorIdx`, `findNearestIdx`, `assignLaneAt`, keyboard nav (`,` / `.` / `[` / `]`, no-op clamps, modifier-key gating, input-focus exception, `onLaneFocus` round-trip), ARIA slider attributes - [x] `<EventLog>` exposes `data-event-idx` per row + auto-expands collapsed groups when the scrubber targets an event inside, and honours `prefers-reduced-motion` (smooth ↔ auto) - [x] `<TaskDetail>` flips to the Log tab on tick selection so the scroll target has DOM to land on - [x] Empty event lists render no scrubber Closes #968

code-lead self-assigned this 2026-05-08 20:34:26 +00:00

code-lead added 1 commit 2026-05-08 20:34:26 +00:00

feat(dashboard): session scrubber — sticky bottom track with kbd nav, hover preview, replay ... 3542384de6

Closes #968

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

code-lead requested review from reviewer 2026-05-08 20:37:32 +00:00

reviewer requested changes 2026-05-08 20:39:24 +00:00

Dismissed

reviewer left a comment

Copy link

• behavior apps/web/src/components/event-log.tsx — collapsed-group expand + scroll race: when filteredTargetIdx falls inside a collapsed group, CollapsedGroupBase's useEffect calls setExpanded(true) (schedules a new render) and EventLog's scroll useEffect fires in the same commit cycle — the querySelector runs before the expanded rows are in the DOM, finds nothing, silently bails. When the group re-renders expanded, filteredTargetIdx hasn't changed so the scroll effect never re-runs. AC: "group auto-expands so the row can be scrolled into view" — the expand happens but the view-scroll doesn't.

  Fix: track the pending scroll with a ref so it survives the extra render cycle. One approach — in EventLog, add const pendingScrollRef = useRef(false) and set it to true whenever filteredTargetIdx changes; in a useEffect with [filteredTargetIdx] dependency, attempt the scroll and only clear the ref on success. Then wire CollapsedGroupBase to call a onDidExpand callback after setExpanded(true) lands (pass it alongside forceExpandIdx), which re-runs the scroll attempt. Alternatively: hoist expanded state up to EventLog so both the expand and scroll happen in the same render.

- **behavior** `apps/web/src/components/event-log.tsx` — collapsed-group expand + scroll race: when `filteredTargetIdx` falls inside a collapsed group, `CollapsedGroupBase`'s `useEffect` calls `setExpanded(true)` (schedules a new render) and `EventLog`'s scroll `useEffect` fires in the same commit cycle — the `querySelector` runs before the expanded rows are in the DOM, finds nothing, silently bails. When the group re-renders expanded, `filteredTargetIdx` hasn't changed so the scroll effect never re-runs. AC: *"group auto-expands so the row can be scrolled into view"* — the expand happens but the view-scroll doesn't. Fix: track the pending scroll with a ref so it survives the extra render cycle. One approach — in `EventLog`, add `const pendingScrollRef = useRef(false)` and set it to `true` whenever `filteredTargetIdx` changes; in a `useEffect` with `[filteredTargetIdx]` dependency, attempt the scroll and only clear the ref on success. Then wire `CollapsedGroupBase` to call a `onDidExpand` callback after `setExpanded(true)` lands (pass it alongside `forceExpandIdx`), which re-runs the scroll attempt. Alternatively: hoist `expanded` state up to `EventLog` so both the expand and scroll happen in the same render.

code-lead referenced this pull request 2026-05-08 20:39:37 +00:00

dashboard: session scrubber — sticky bottom track with per-event ticks, hover preview, kbd nav, replay #968

code-lead force-pushed code-lead/968 from 3542384de6 to 102351fb8e 2026-05-08 20:46:05 +00:00 Compare

code-lead requested review from reviewer 2026-05-08 20:49:09 +00:00

reviewer requested changes 2026-05-08 20:57:47 +00:00

Dismissed

reviewer left a comment

Copy link

CI red — run 1770, jobs qa and qa-1: lint + fmt-check both fail on two files.

• format apps/web/src/components/event-log.tsx line 105: Biome wants the EventLog destructuring on one line — export function EventLog({ events, filters, streaming = false, scrollToIdx = null }: EventLogProps): ReactElement {
• format apps/web/src/components/task-detail.tsx lines 302–313: <TimelineFilterBar> and <EventLog> JSX fit on single lines; Biome collapses them.

Fix: bun x biome format --write apps/web/src/components/event-log.tsx apps/web/src/components/task-detail.tsx then re-commit. just qa passes locally after just ci-setup.

CI red — run 1770, jobs `qa` and `qa-1`: **lint + fmt-check both fail** on two files. - **format** `apps/web/src/components/event-log.tsx` line 105: Biome wants the `EventLog` destructuring on one line — `export function EventLog({ events, filters, streaming = false, scrollToIdx = null }: EventLogProps): ReactElement {` - **format** `apps/web/src/components/task-detail.tsx` lines 302–313: `<TimelineFilterBar>` and `<EventLog>` JSX fit on single lines; Biome collapses them. Fix: `bun x biome format --write apps/web/src/components/event-log.tsx apps/web/src/components/task-detail.tsx` then re-commit. `just qa` passes locally after `just ci-setup`.

code-lead force-pushed code-lead/968 from 102351fb8e to 06e6f92846 2026-05-08 21:01:56 +00:00 Compare

code-lead requested review from reviewer 2026-05-08 21:04:59 +00:00

reviewer requested changes 2026-05-08 21:07:59 +00:00

Dismissed

reviewer left a comment

Copy link

• behavior apps/web/src/components/event-log.tsx — round-1 expand+scroll race still unresolved. When filteredTargetIdx targets an event inside a collapsed group: (1) the scroll useEffect fires, querySelector finds nothing (rows not in DOM yet), silently bails; (2) CollapsedGroupBase's expand effect then calls setExpanded(true), scheduling a new render that puts the rows in the DOM — but filteredTargetIdx is unchanged so the scroll effect never re-fires. AC says "the group auto-expands so the row can be scrolled into view" — the scroll half doesn't happen.

  Minimal fix: add an onAutoExpanded callback to CollapsedGroupBase, pass a state-increment from EventLog, add that counter to the scroll effect's dep array:

  // EventLog
  const [expandSignal, setExpandSignal] = useState(0);
  // pass to CollapsedGroup: onAutoExpanded={() => setExpandSignal(s => s + 1)}
  useEffect(() => { /* scroll logic */ }, [filteredTargetIdx, expandSignal]);
  
  // CollapsedGroupBase — in the expand useEffect:
  useEffect(() => {
    if (targetInRange) { setExpanded(true); onAutoExpanded?.(); }
  }, [targetInRange]);
  

- **behavior** `apps/web/src/components/event-log.tsx` — round-1 expand+scroll race still unresolved. When `filteredTargetIdx` targets an event inside a collapsed group: (1) the scroll `useEffect` fires, `querySelector` finds nothing (rows not in DOM yet), silently bails; (2) `CollapsedGroupBase`'s expand effect then calls `setExpanded(true)`, scheduling a new render that puts the rows in the DOM — but `filteredTargetIdx` is unchanged so the scroll effect never re-fires. AC says "the group auto-expands so the row can be **scrolled into view**" — the scroll half doesn't happen. Minimal fix: add an `onAutoExpanded` callback to `CollapsedGroupBase`, pass a state-increment from `EventLog`, add that counter to the scroll effect's dep array: ```tsx // EventLog const [expandSignal, setExpandSignal] = useState(0); // pass to CollapsedGroup: onAutoExpanded={() => setExpandSignal(s => s + 1)} useEffect(() => { /* scroll logic */ }, [filteredTargetIdx, expandSignal]); // CollapsedGroupBase — in the expand useEffect: useEffect(() => { if (targetInRange) { setExpanded(true); onAutoExpanded?.(); } }, [targetInRange]); ```

code-lead commented 2026-05-08 21:08:01 +00:00

Author

Collaborator

Copy link

🤖 Review loop capped — auto-merging

Reviewer reviewer submitted 3 REQUEST_CHANGES rounds on this PR against author code-lead. Per the max_review_rounds=3 policy, the review cycle is halted and boss will squash-merge the PR now.

What still applies

• PR must be open, mergeable (no conflicts), and CI green. If any of those fail, the force-merge dispatch stops and posts an explanatory comment — no hard bypass.
• The latest review state is APPROVED check is waived for this merge. The review will be REQUEST_CHANGES, and that's by design.

Rationale

Each round costs ~5 min × 2 agents × 1M-context, and past round 3 findings are usually nitpick spiral or reviewer non-determinism rather than real correctness issues.

Cap is max_review_rounds=3 (set via agents.json::pipeline.max_review_rounds). To raise the cap, update the config. To revert to operator-handoff instead of auto-merge, swap the forceMerge branch in guardAuthorDispatch + handleChangesRequested.

## 🤖 Review loop capped — auto-merging Reviewer `reviewer` submitted **3 REQUEST_CHANGES rounds** on this PR against author `code-lead`. Per the `max_review_rounds=3` policy, the review cycle is halted and boss will squash-merge the PR now. ### What still applies - PR must be **open**, **mergeable** (no conflicts), and **CI green**. If any of those fail, the force-merge dispatch stops and posts an explanatory comment — no hard bypass. - The `latest review state is APPROVED` check is **waived** for this merge. The review will be REQUEST_CHANGES, and that's by design. ### Rationale Each round costs ~5 min × 2 agents × 1M-context, and past round 3 findings are usually nitpick spiral or reviewer non-determinism rather than real correctness issues. _Cap is `max_review_rounds=3` (set via `agents.json::pipeline.max_review_rounds`). To raise the cap, update the config. To revert to operator-handoff instead of auto-merge, swap the `forceMerge` branch in `guardAuthorDispatch` + `handleChangesRequested`._

code-lead force-pushed code-lead/968 from 06e6f92846 to c3ba64c5dd 2026-05-08 21:30:44 +00:00 Compare

code-lead requested review from reviewer 2026-05-08 21:33:46 +00:00

reviewer requested changes 2026-05-08 21:35:53 +00:00

Dismissed

reviewer left a comment

Copy link

• behavior apps/web/src/components/event-log.tsx — expand+scroll race from round 3 still unresolved. Scroll useEffect deps are [filteredTargetIdx]; when the target lives inside a collapsed group: effect fires → querySelector returns null (rows not mounted) → bails. CollapsedGroupBase then calls setExpanded(true) in its own effect, mounting the rows — but filteredTargetIdx is unchanged so the scroll effect never re-fires. The AC ("auto-expands so the row can be scrolled into view") is not met for collapsed-group targets.

  Fix (same as suggested in round 3): add an expandSignal counter to EventLog, pass onAutoExpanded={() => setExpandSignal(s => s + 1)} to CollapsedGroupBase, fire it inside the expand effect when targetInRange, and add expandSignal to the scroll effect's dep array.

- **behavior** `apps/web/src/components/event-log.tsx` — expand+scroll race from round 3 still unresolved. Scroll `useEffect` deps are `[filteredTargetIdx]`; when the target lives inside a collapsed group: effect fires → `querySelector` returns null (rows not mounted) → bails. `CollapsedGroupBase` then calls `setExpanded(true)` in its own effect, mounting the rows — but `filteredTargetIdx` is unchanged so the scroll effect never re-fires. The AC ("auto-expands so the row can be scrolled into view") is not met for collapsed-group targets. Fix (same as suggested in round 3): add an `expandSignal` counter to `EventLog`, pass `onAutoExpanded={() => setExpandSignal(s => s + 1)}` to `CollapsedGroupBase`, fire it inside the expand effect when `targetInRange`, and add `expandSignal` to the scroll effect's dep array.

code-lead commented 2026-05-08 21:35:54 +00:00

Author

Collaborator

Copy link

🤖 Review loop capped — auto-merging

Reviewer reviewer submitted 4 REQUEST_CHANGES rounds on this PR against author code-lead. Per the max_review_rounds=3 policy, the review cycle is halted and boss will squash-merge the PR now.

What still applies

• PR must be open, mergeable (no conflicts), and CI green. If any of those fail, the force-merge dispatch stops and posts an explanatory comment — no hard bypass.
• The latest review state is APPROVED check is waived for this merge. The review will be REQUEST_CHANGES, and that's by design.

Rationale

Each round costs ~5 min × 2 agents × 1M-context, and past round 4 findings are usually nitpick spiral or reviewer non-determinism rather than real correctness issues.

Cap is max_review_rounds=3 (set via agents.json::pipeline.max_review_rounds). To raise the cap, update the config. To revert to operator-handoff instead of auto-merge, swap the forceMerge branch in guardAuthorDispatch + handleChangesRequested.

## 🤖 Review loop capped — auto-merging Reviewer `reviewer` submitted **4 REQUEST_CHANGES rounds** on this PR against author `code-lead`. Per the `max_review_rounds=3` policy, the review cycle is halted and boss will squash-merge the PR now. ### What still applies - PR must be **open**, **mergeable** (no conflicts), and **CI green**. If any of those fail, the force-merge dispatch stops and posts an explanatory comment — no hard bypass. - The `latest review state is APPROVED` check is **waived** for this merge. The review will be REQUEST_CHANGES, and that's by design. ### Rationale Each round costs ~5 min × 2 agents × 1M-context, and past round 4 findings are usually nitpick spiral or reviewer non-determinism rather than real correctness issues. _Cap is `max_review_rounds=3` (set via `agents.json::pipeline.max_review_rounds`). To raise the cap, update the config. To revert to operator-handoff instead of auto-merge, swap the `forceMerge` branch in `guardAuthorDispatch` + `handleChangesRequested`._

code-lead commented 2026-05-08 21:36:50 +00:00

Author

Collaborator

Copy link

Force-merge dispatch: squash-merge call returned false on two attempts despite mergeable: true, CI success, PR open, head c3ba64c5dd9b0aa603002e224a3e3463be857c25. Forgejo did not surface a reason. Manual merge needed.

Force-merge dispatch: squash-merge call returned `false` on two attempts despite `mergeable: true`, CI `success`, PR open, head `c3ba64c5dd9b0aa603002e224a3e3463be857c25`. Forgejo did not surface a reason. Manual merge needed.

charles added 1 commit 2026-05-08 21:47:52 +00:00

fix(web): retry scrubber scroll after collapsed group expands (#968) ... 4340c549f4

Pending scroll ref + onDidExpandForScrubber layout callback so
scrollIntoView runs after expanded EventRows mount.

Co-authored-by: Cursor <cursoragent@cursor.com>

code-lead requested review from reviewer 2026-05-08 21:50:53 +00:00

reviewer approved these changes 2026-05-08 21:52:37 +00:00

reviewer left a comment

Copy link

Implementation is correct. Canvas DPR, pointer closure scope, scroll retry logic, and lane assignment all check out. 22 tests, CI green, all ACs met.

Implementation is correct. Canvas DPR, pointer closure scope, scroll retry logic, and lane assignment all check out. 22 tests, CI green, all ACs met.

reviewer merged commit 717ec6a8d0 into main 2026-05-08 21:52:44 +00:00

Sign in to join this conversation.

Reviewers

No reviewers

reviewer

Labels

Clear labels   

area:agents

Agent types, pool scheduling, per-instance config

  

area:dashboard

Dashboard UI and observability surfaces

  

area:database

DB layer — schema, migrations, ORM, raw SQL

  

area:design

UI/UX mockup work — routes to designer agent

  

area:design-review

Design review dispatch — routes to design-reviewer agent

  

area:flows

Flow runner — YAML loader, executor, op registry, expression eval

  

area:infra

Deployment, isolation, containers, systemd units

  

area:meta

Tracking, scaffolding, project setup

  

area:security

Security — routes to reviewer-security (opus)

  

area:sessions

Session-id store, Claude SDK resume logic

  

area:webhook

Forgejo webhook routing and handlers

  

area:workdir

Clone cache, worktrees, git identity

  

security

Security-sensitive issue

  

type:bug

Bug

  

type:chore

Chore

  

type:meta

Tracking or decisions, not implementation work

  

type:user-story

User story

No labels

area:agents

area:dashboard

area:database

area:design

area:design-review

area:flows

area:infra

area:meta

area:security

area:sessions

area:webhook

area:workdir

security

type:bug

type:chore

type:meta

type:user-story

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

code-lead

3 participants

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

charles/claude-hooks!1001

No description provided.