fix(mobile): stabilize thread composer and interactions

[juliusmarminge]

What changed

• Keep the floating composer anchored after keyboard dismissal on iOS and Android.
• Stop live-feed updates from pulling users back to the end while they are reading history.
• Add a floating scroll-to-end control that uses the same keyboard-aware LegendList path.
• Bound long pending-input forms to the viewport, keep their actions reachable, and preserve multi-select answers.
• Add focused tests for live-follow, pending-input layout, and answer shaping.

This consolidates #5118 and #5915 into #5986 as the single branch to test and merge. Their original commit authorship is preserved.

Closes #5434.

Why

Several mobile thread interactions shared the same keyboard, overlay, and end-follow boundaries. Testing them together catches the real integration behavior: the composer stays attached to the keyboard, user-driven scrolling remains respected, the return-to-end affordance reflects that same state, and pending input remains usable in the remaining viewport.

Validation

• vp test run apps/mobile/src/features/threads/thread-feed-live-follow.test.ts apps/mobile/src/features/threads/pendingUserInputLayout.test.ts apps/mobile/src/lib/threadActivity.test.ts — 24 tests passed.
• vp run --filter @t3tools/mobile typecheck passed.
• Targeted vp lint --report-unused-disable-directives passed for all changed TypeScript files.
• node scripts/mobile-native-static-check.ts passed SwiftLint, ktlint, and detekt.
• vp fmt --check passed for all changed files.
• React Doctor scoped to the diff scored 88/100; the four remaining findings are non-blocking heuristics for the intentionally bounded question scroller, existing screen size, thread-switch reset, and a tiny normalization pass.

UI evidence

Pending input: before, after, and scroll video.

Scroll-to-end: Android interaction video.

Manual test plan

1. Open and dismiss the keyboard repeatedly; confirm the composer returns to the bottom.
2. Scroll upward while an agent is streaming; confirm new activity does not interrupt reading.
3. Tap the down-arrow during and after momentum; confirm the thread returns to the end and the button disappears.
4. Exercise a long pending-input request with single- and multi-select questions while the keyboard is open and closed.
5. Repeat the keyboard and scrolling cases on both iOS and Android.

Model: GPT-5. Harness: Codex.

Note

Stabilize mobile thread composer, user input card, and settings navigation

• Rewrites PendingUserInputCard as an animated collapsible overlay that replaces the composer while a user-input request is active; supports multi-select options and reports coverage via a SharedValue to drive feed inset adjustments
• Adds event-driven resolveThreadFeedLiveFollow state machine to ThreadFeed that correctly distinguishes user-initiated from programmatic scrolls and exposes a scroll-to-end control pill when not following
• Restructures the settings sheet navigation stack so SettingsContent is nested inside SettingsSheetStack, and removes the ClerkSettingsSheetDetent context throughout
• On iOS, thread settings are now presented as a native context menu on the composer toolbar pill (with haptic feedback), replacing the modal sheet path
• Patches @legendapp/list to 3.3.5 and react-native-screens with fixes for scroll-to-end fallback, anchored end space, MVCP, and header bar button caching to prevent UI glitches
• Patches @react-native-menu/menu to support three-level submenus, keepsMenuPresented on actions, and live in-place menu updates without dismissal
• Risk: buildPendingUserInputAnswers now returns string | ReadonlyArray<string> per question; callers that assume a single string will break

^{Macroscope summarized 401afe7.}

---

Note

Medium Risk
Touches core thread scrolling, keyboard overlays, and settings/Clerk navigation with vendored list patches; behavior is integration-heavy though covered by new unit tests and the PR’s manual plan.

Overview
Thread experience gets the bulk of the work: live-follow now uses an explicit scroll-event resolver so streaming updates stop yanking readers off history, with a floating scroll-to-end control wired to the same follow state. While a user-input request is active, the composer slot is replaced by a collapsible questionnaire (iOS overlay + UI-thread collapse; Android in-flow), with viewport-bounded height, multi-select answers, and combined feed insets so the transcript stays aligned.

Keyboard and composer fixes include iOS safe-area padding keyed on composer focus (not keyboard visibility), Android resume quarantine for stale IME height, KeyboardStickyView on the new-task draft composer, and optional onEditorFocusChange hooks. On iOS, everyday thread settings can change from a native toolbar menu without dismissing the keyboard.

Settings navigation is restructured: settings screens live in a nested SettingsContent stack inside a headerless outer sheet stack so Clerk auth can own chrome without iOS 26 header jumps; ClerkSettingsSheetDetent is removed and deep links navigate through SettingsContent. Clerk Expo is patched so auth/profile back can pop the native stack first.

Dependencies: @legendapp/list moves to the catalog version with expanded patches (composer insets, transparent-header scroll clamps, anchored-end-space). App update checks skip Metro/dev-client and treat known Expo “unavailable” errors as no-ops. ControlPill gains press-in activation for snappier scroll-to-end taps.

^{Reviewed by Cursor Bugbot for commit 401afe7. Bugbot is set up for automated code reviews on this repo. Configure here.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 0a6d9639-03a9-4925-9e0c-110376cd6281

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.3 KiB	−22 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	+3 B (+0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	5.9 KiB	−25 B (−0.4%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.3 KiB	11.3 KiB	+5 B (+0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	−6 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.8 KiB	5.9 KiB	+11 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: f5fce74 · PR result: 401afe7 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 94.6 KiB
• Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Needs human review

5 blocking correctness issues found. This PR introduces substantial new mobile capabilities including collapsible user input cards, scroll-to-end buttons with iOS 26 glass effects, native iOS menus, and complex platform-specific keyboard handling - not a simple stabilization. Five high-severity correctness issues were identified regarding viewport overflow, bar button item loss, keyboard state handling, and environment-scoped thread keys.

^{You can customize Macroscope's approvability policy. Learn more.}

[macroscopeapp]

🟠 High threads/ThreadDetailScreen.tsx:229

derivePendingUserInputMaxHeight clamps the computed available height to a 160px minimum, so on a short or landscape phone with the keyboard open, the card can be 160px tall even when the true available space is far smaller (or negative). Because the card is inside a bottom-anchored overlay that is not itself scrollable, the top of the form is pushed off-screen and its initial questions are unreachable. Additionally, when both activePendingApproval and activePendingUserInput are non-null, the max-height calculation does not subtract the simultaneously rendered PendingApprovalCard, so the overlay holds both cards plus its gap and overflows the viewport, pushing controls or the composer off-screen. Consider allowing the card to shrink below 160px when space is constrained, and subtract the approval card's height from the available space when both are present.

🚀 Reply "fix it for me" or copy this AI Prompt for your agent:

In file @apps/mobile/src/features/threads/ThreadDetailScreen.tsx around line 229:

`derivePendingUserInputMaxHeight` clamps the computed available height to a 160px minimum, so on a short or landscape phone with the keyboard open, the card can be 160px tall even when the true available space is far smaller (or negative). Because the card is inside a bottom-anchored overlay that is not itself scrollable, the top of the form is pushed off-screen and its initial questions are unreachable. Additionally, when both `activePendingApproval` and `activePendingUserInput` are non-null, the max-height calculation does not subtract the simultaneously rendered `PendingApprovalCard`, so the overlay holds both cards plus its gap and overflows the viewport, pushing controls or the composer off-screen. Consider allowing the card to shrink below 160px when space is constrained, and subtract the approval card's height from the available space when both are present.