Skip to content

fix(web,mobile): queued messages get a simpler Send now row, no reordering - #25846

Merged
t3dotgg merged 3 commits into
mainfrom
t3/clean-up-queued-message-controls
Oct 27, 2026
Merged

t3dotgg merged 3 commits into
mainfrom
t3/clean-up-queued-message-controls

Conversation

@t3dotgg

@t3dotgg t3dotgg commented Oct 27, 2026

Copy link
Copy Markdown
Member

The queued message row had too much noise: an "Up next · 1 queued" header that repeated itself, position numbers, a "Steer" label with a right-pointing arrow, and edit sitting between Steer and remove. Reordering queued messages was also not worth its weight.

Now each queued message shows ↑ Send now | ✎ ✕. Send now comes first and uses the same up arrow as the composer's send button. A divider keeps edit and remove apart from it. Send now only shows when it can actually send.

Reordering is gone from web (drag and arrow keys) and mobile (drag handle, Move up/down). The server queued-run.reorder command and the MCP t3_queue reorder action stay for now: old event logs must still replay, and changing MCP tool text needs an eval pass. Mobile also says "Send now" everywhere, and the user docs no longer mention Up next or reordering.

Before After
before after

Proof: vp run check:changed passes. In a real dev app with a running turn, I queued two messages, removed one, and used Send now on the other. Both worked. Mobile is covered by typecheck and unit tests only. A device capture was blocked: this checkout has no cached native dev client, and building one needs Xcode or EAS.

Council review

Rounds: 2

Delegate Round 1 Round 2
Claude Opus 5.5 8/10 9/10
Muse Wonder Full 8/10 9/10
Codex GPT-6.1 Sol 9/10 9/10

Created with Claude Opus 5.5 in Claude Code.

🤖 Generated with Claude Code

@github-actions github-actions Bot added vouch:trusted Author is trusted by repo permissions or the VOUCHED list. size:XL 500-999 changed lines (additions + deletions). labels Oct 27, 2026
@macroscopeapp

macroscopeapp Bot commented Oct 27, 2026 •

Copy link
Copy Markdown
Contributor

Approvability

Verdict: Needs Human Approval

Macroscope's review found this PR requires human approval — This PR changes production queue behavior across web and mobile by removing queued-message reordering and its client command plumbing, while reshaping the Send now/edit/remove controls. Because an existing user-facing capability is being removed across platforms, the behavioral change merits human review.

No code changes detected at 250f00f. Prior analysis still applies.

You can add or adjust custom eligibility rules. Learn more.

t3dotgg and others added 3 commits October 27, 2026 02:12
Drop the queue header and reordering. Send now is first, with an up arrow,
and sits apart from edit and remove.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Hide Send now when it can never send, match the mobile accessibility
label to the visible text, and drop reorder from the user docs.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
@t3dotgg
t3dotgg force-pushed the t3/clean-up-queued-message-controls branch from 36bb3b1 to 250f00f Compare October 27, 2026 09:12
@coderabbitai

coderabbitai Bot commented Oct 27, 2026 •

Copy link
Copy Markdown
Contributor

Review in Change Stack →

No actionable comments were generated in the recent review.

Recent review info
Run configuration
Commits

Reviewing files that changed from the base of the PR and between a24b109 and 250f00f.

Files selected for processing: 11
  • apps/mobile/src/features/threads/ThreadQueueControl.tsx
  • apps/mobile/src/features/threads/floating-working-control.tsx
  • apps/mobile/src/features/threads/threadQueueControlPresentation.test.ts
  • apps/mobile/src/features/threads/threadQueueControlPresentation.ts
  • apps/web/src/components/chat/QueuedRunsControl.browser.test.tsx
  • apps/web/src/components/chat/QueuedRunsControl.tsx
  • docs/user/composer.md
  • packages/client-runtime/src/operations/commands.ts
  • packages/client-runtime/src/state/threadCommands.ts
  • packages/client-runtime/src/state/threadWorkflows.test.ts
  • packages/client-runtime/src/state/threadWorkflows.ts
Files with no reviewable changes: 5
  • apps/web/src/components/chat/QueuedRunsControl.browser.test.tsx
  • packages/client-runtime/src/operations/commands.ts
  • packages/client-runtime/src/state/threadWorkflows.ts
  • packages/client-runtime/src/state/threadCommands.ts
  • apps/mobile/src/features/threads/threadQueueControlPresentation.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 2 remain after this review.

Walkthrough

Queue reordering was removed from the runtime command API and from the web and mobile queue controls. The web control adds a per-item “Send now” action. The documentation and tests were updated to reflect the changed queue controls.

Changes: Queued Message Controls

Layer / File(s) Summary
Remove reordering from runtime contracts
packages/client-runtime/src/operations/commands.ts, packages/client-runtime/src/state/threadCommands.ts, packages/client-runtime/src/state/threadWorkflows.ts, packages/client-runtime/src/state/threadWorkflows.test.ts
The workflow state and command environment no longer expose queue reordering. Workflow tests no longer check reorder capability.
Update web queue controls
apps/web/src/components/chat/QueuedRunsControl.tsx, apps/web/src/components/chat/QueuedRunsControl.browser.test.tsx, docs/user/composer.md
The web queue removes drag-and-drop reordering and adds a per-item “Send now” action. The browser test and composer documentation reflect the updated controls.
Update mobile queue controls
apps/mobile/src/features/threads/ThreadQueueControl.tsx, apps/mobile/src/features/threads/threadQueueControlPresentation.ts, apps/mobile/src/features/threads/threadQueueControlPresentation.test.ts, apps/mobile/src/features/threads/floating-working-control.tsx
Mobile queue rows remove drag and move controls. The row controls, presentation helpers, tests, and accessibility hint reflect the remaining actions.

Priority: Normal

Estimated code review effort: 3 (Moderate) | ~20 minutes

Change: Bug fix

Possibly related issues

Suggested labels: bugfix

Suggested reviewers: juliusmarminge, yash-singh1, macroscopeapp

Merge Risk: Minimal · up to 250f0

The updated queue controls appear ready to merge after normal checks. Mobile behavior has not been validated on a device.

Pre-merge checks: Passed 4
Check name Status Explanation
Title check Passed The title clearly summarizes the main changes: a simpler “Send now” row and removal of queue reordering for web and mobile.
Description check Passed The description explains the problem, the web and mobile changes, retained server and MCP behavior, verification results, limitations, and includes before/after screenshots. It does not explicitly lin…
Linked Issues check Passed Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check Passed Check skipped because no linked issues were found for this pull request.
Full details: Description check

Explanation

The description explains the problem, the web and mobile changes, retained server and MCP behavior, verification results, limitations, and includes before/after screenshots. It does not explicitly link an issue or maintainer approval, but it provides a focused rationale for the scope.

Finishing Touches
📝 Generate docstrings
  • Commit to this branch
  • Create a new PR
🧪 Generate unit tests (beta)
  • Commit to this branch
  • Create a new PR
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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

@t3dotgg
t3dotgg merged commit 92e6450 into main Oct 27, 2026
32 checks passed
@t3dotgg
t3dotgg deleted the t3/clean-up-queued-message-controls branch October 27, 2026 09:26
adampeterhiggins added a commit to adampeterhiggins/t3code that referenced this pull request Oct 27, 2026
* refactor(client): composers share one workspace snapshot refresh per provider and cwd (pingdotgg#25370)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): composer pastes and picks are cancelled by their draft, not by the composer (pingdotgg#25376)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* chore(mobile): e2e poses for current devices, Duo and landscape fixes

Poses now cover iPhone SE, iPhone 18 Pro, iPhone Duo folded and unfolded,
iPad Pro 13" in both orientations, Pixel 10 Pro, Pixel 10 Pro Fold folded,
unfolded and unfolded landscape, and Pixel Tablet in both orientations.
Landscape poses rotate once the app is up (a rotation during launch is lost)
and check that the screen really is wider than tall. Taps fall back to the
element's centre when agent-device refuses a covered ref, and a gateway
whose workers briefly reconnect is retried.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

* fix(mobile): select the correct draft after adding a project (pingdotgg#24269)

* fix(mobile): mobile:install no longer requires --prepare

The --prepare flag from pingdotgg#25390 had no default, so effect/cli made it
required and every mobile:install failed with "Missing required flag".

Co-Authored-By: Claude Opus 5.5 <[email protected]>

* fix(mobile): e2e landscape poses rotate for real on tablets and folds

agent-device names rotations relative to the device's natural orientation,
which is landscape on the Pixel Tablet and an unfolded Fold, so
"landscape-left" turned them portrait. The runner now asks for a shape and
tries each rotation until the screen has it. It also prepares each platform's
dev client before leasing, rides out longer worker drops, and skips the
Duo poses by default until agent-device can fold Simlock's simulators.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

* fix(server): goals stop the running turn, clear while working, no echoes (pingdotgg#25426)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): pull request list remembers each answer for the next start (pingdotgg#25364)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): HTML previews handle frame messages in one listener (pingdotgg#25456)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): review guide runs are watched from their answers, not from mounted components (pingdotgg#25459)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): resize drags, panel shells and the browser host own their listeners (pingdotgg#25464)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): find shows its match after a thread switch, and the timeline hears scrolling through one viewport signal (pingdotgg#25463)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): a new thread's first send no longer sticks on Starting (pingdotgg#25477)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(client): drafts for answered questions are cleared even when their thread is not open (pingdotgg#25457)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): terminal output no longer re-renders the drawer, theme search runs from handlers (pingdotgg#25458)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): stable callbacks refresh before layout effects read them (pingdotgg#25489)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* test(web): say why the useStableCallback test renders without a router (pingdotgg#25493)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): queued edits end when their run starts, wherever you are (pingdotgg#25468)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): root theme, toast and chrome sync start with the app (pingdotgg#25461)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): root theme, toast and chrome sync start with the app (pingdotgg#25461)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): editing a queued message while a question is open no longer throws (pingdotgg#25503)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): browser viewport resizes work without an open preview panel (pingdotgg#25462)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* test(web): attachment preview omission tests no longer time out on CI (pingdotgg#25515)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): describe when composer view commands can reject (pingdotgg#25516)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): browser history and titles show up again (pingdotgg#25491)

* fix(web): switching back to a thread no longer restores its tab's older title (pingdotgg#25521)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): pull request page reacts to turns and sweep Escape without effects (pingdotgg#25523)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): an armed pull request confirm no longer outlives its button (pingdotgg#25530)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): keep expanding composer text above its toolbar (pingdotgg#25396)

* fix(server): release completed PR sync sweep data from the read cache (pingdotgg#25285)

* fix(mobile): drop the unused useLayoutEffect import that fails lint on main (pingdotgg#25575)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): provider settings and Codex setup run as explicit flows (pingdotgg#25524)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* test(web): attachment preview omission cases check the prepared document instead of rendering each one (pingdotgg#25562)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): pull request screens use a file-jump store and handler work (pingdotgg#25525)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): label the pull request toast action View (pingdotgg#25581)

* refactor(mobile): media previews share file leases and close on blur in one place (pingdotgg#25527)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(shared): pin T3 Connect cloudflared to 2026.10.4, which drops the compromised eMudhra intermediates (pingdotgg#25572)

Co-authored-by: Theo Browne <[email protected]>

* fix(web): skill chips in chat open the same details popover as the composer (pingdotgg#25432)

* refactor(web): chat route shortcuts register once instead of rebuilding on every store change (pingdotgg#25366)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(clients): queued messages can't be edited while a question is open (pingdotgg#25593)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* docs: add an FAQ that answers what people ask every day (pingdotgg#25451)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* feat(web): warn before adding a second Claude or Codex account (pingdotgg#25475)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(server): expire new-model badges after seven days (pingdotgg#25609)

* refactor(web): running timers share one clock instead of an interval each (pingdotgg#25616)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): simplify the created pull request success row (pingdotgg#25613)

* fix(web): thread jump hints wait for their delay on every hold (pingdotgg#25636)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): open threads from the local cache instead of "Thread unavailable" (pingdotgg#25627)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(infra): task-cache telemetry no longer strands an event queued during a flush (pingdotgg#25653)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): video previews attach listeners on their element and share one visibility observer (pingdotgg#25630)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): voice input session owns recording, metering and keep-awake (pingdotgg#25631)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): screen readers can wrap to the first and last message of a long thread (pingdotgg#23780)

Co-authored-by: PR Batch Tester <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>
Co-authored-by: nibarAhmed <[email protected]>
Co-authored-by: maria-rcks <[email protected]>

* fix(client-runtime): a slow host no longer shows settings as read-only (pingdotgg#25526)

Co-authored-by: Julius Marminge <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>
Signed-off-by: Yordis Prieto <[email protected]>

* refactor(mobile): thread rows share one clock and cache SVGs instead of per-row effects (pingdotgg#25612)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(server): a thread whose worktree was deleted recovers on the next message (pingdotgg#18359)

Co-authored-by: Julius Marminge <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>
Signed-off-by: Yordis Prieto <[email protected]>

* refactor(mobile): pull request list records answers where they land (pingdotgg#25373)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): keep floating status within the composer width (pingdotgg#25671)

* refactor(web): settings group by concern, with General split into focused pages (pingdotgg#25651)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(server): telemetry keeps a failed batch for 30 minutes, not 1.5 (pingdotgg#25474)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): a caret move in one composer no longer moves another draft's insert point (pingdotgg#25681)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* test(web): attachment preview Word and large-text tests no longer flake on CI (pingdotgg#25664)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* feat(evals): shared evals package and an eval for T3 tool-use mistakes (pingdotgg#24635)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): Cmd+O presses the Open in editor button instead of a per-picker listener (pingdotgg#25667)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): queued messages are delivered by an app service, not a render loop (pingdotgg#25678)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* test(web): PDF viewer behavior is covered by fast tests (pingdotgg#25699)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): drop the unreachable pending-task editor from new-task drafts (pingdotgg#25684)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* feat(web): thread state lives at the end of the timeline, not stacked on the composer (pingdotgg#25401)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): drafts promote from the draft store, not from route effects (pingdotgg#25615)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): settings use web's pages and groups (pingdotgg#25686)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): thread inspector mode is derived, not synced (pingdotgg#25682)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): pull request list, filters and viewed files without synchronizing effects (pingdotgg#25614)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): PDF search finds phrases that wrap across lines (pingdotgg#25702)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* feat(server): cut T3's MCP tool schemas by a third and merge 18 tools into action tools (pingdotgg#25423)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): pull request rows refresh their diff stats after a new push (pingdotgg#25713)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(workflows): keep completed workflows reachable (pingdotgg#24587)

Co-authored-by: maria-rcks <[email protected]>

* fix(web): devices agents open in background threads now show up (pingdotgg#25714)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): optimistic chat messages leave without a timer (pingdotgg#25710)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* feat(acp): ACP session notices show as notices, not reply text (pingdotgg#25704)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): scale and align thread list icons (pingdotgg#25668)

* refactor(mobile): new-task branch defaults are derived, not written after render (pingdotgg#25706)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* feat(provider): add Vercel fx as a provider (pingdotgg#18277)

Co-authored-by: Julius Marminge <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): usage chart bars round at the top instead of the bottom (pingdotgg#21383)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): a draft route's branch checkout runs once, even across a reconnect (pingdotgg#25712)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): thread details popover closes when the card moves inline (pingdotgg#25731)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): diff and file reveals are requests, and workspace edits refetch by construction (pingdotgg#25732)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* feat(web): Add provider leads with first-class providers (pingdotgg#25753)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): app-lifetime services install once, and the usage widget stops re-rendering on every presentation change (pingdotgg#25740)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(client-runtime): searches and refreshes no longer wait on an unreachable environment (pingdotgg#25703)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): command palette state lives in a store, git and branch pickers drop timer and sync effects (pingdotgg#25733)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): restore Android chat content corners (pingdotgg#23989)

Co-authored-by: maria-rcks <[email protected]>

* fix(server): agents show trends and comparisons as inline charts again (pingdotgg#25766)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): isolate file tree state by workspace (pingdotgg#24270)

* fix(mobile): finished subagents stay reachable from the Agents pill (pingdotgg#25759)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): keep segmented controls readable on cards and screens (pingdotgg#25729)

* fix(web): cancelling update everything no longer flashes the update button's focus ring and tooltip (pingdotgg#25774)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): lists end at the search field, and HTML previews start below the header (pingdotgg#25765)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(provider-cursor): bump @cursor/sdk to 1.0.38 so long-context tiers send Max Mode (pingdotgg#21074)

Co-authored-by: Julius Marminge <[email protected]>

* fix(server): Sonnet also answers trend and comparison data with inline charts (pingdotgg#25776)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(server): Claude MCP timeouts preserve the native session (pingdotgg#25660)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(server): preserve settled runs during delegated acknowledgements (pingdotgg#25672)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): align the iPad composer with workspace navigation (pingdotgg#24942)

Co-authored-by: maria-rcks <[email protected]>

* chore(evals): store eval cases as .ndjson so review tools skip them (pingdotgg#25782)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* test(server): CI fails when agent context grows past main (pingdotgg#25785)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): theme preferences reach the native runtime without a layout effect (pingdotgg#25767)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): the share inbox is a module store, presented from navigation state (pingdotgg#25744)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): drop the unreachable paint-only thread timeline path (pingdotgg#25787)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* chore(release): prepare v0.1.3

* refactor(mobile): notifications, quick actions and Connect onboarding navigate through the root ref (pingdotgg#25769)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* test(server): data regression test now fails when recent fetch savings regress (pingdotgg#25795)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(server): all MCP tools now send compact schemas, cutting tool context by 23% (pingdotgg#25790)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(web): the new design is the only design, with no toggle (pingdotgg#25778)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* feat(mobile): add a "Use last project" preference for new threads (pingdotgg#24741)

Co-authored-by: none23 <[email protected]>
Co-authored-by: Julius Marminge <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): round shared Android sheets and Chat settings (pingdotgg#18547)

* feat(mobile): split wide Android Settings into two panes (pingdotgg#24298)

* fix(web): keep right panel maximize available in narrow windows (pingdotgg#25389)

* fix(server): Macs without Android no longer show avdmanager discovery warnings (pingdotgg#25815)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): onboarding import no longer hangs when the project never arrives (pingdotgg#25805)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): Windows image paths and non-ASCII file links open like desktop (pingdotgg#25802)

Co-authored-by: Claude Opus 5.5 <[email protected]>
Co-authored-by: amv146 <[email protected]>

* fix(web,mobile): markdown previews no longer show YAML frontmatter as a giant heading (pingdotgg#25804)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): compact toggle shows the token count again (pingdotgg#25818)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(server): preserve interrupted Claude context after revert (pingdotgg#25695)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web,mobile): bot approvals are visible on the PR summary (pingdotgg#25829)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web,mobile): stopped runs no longer look like errors (pingdotgg#25830)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): share imports run outside the new-task screen (pingdotgg#25793)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(build): TEMPORARY workaround so linux-arm64 builds run on 16/64 KiB page kernels (pingdotgg#25825)

TEMPORARY workaround for an upstream Node bug: nodejs/node#67048
Node 26.9+ --build-sea writes linux-arm64 executables that 16/64 KiB page kernels cannot load.
linux-arm64 now injects with postject, and the archive build rejects incompatible ELF layouts.
One-time exception. Revert once Node ships a fix: grep for the Node issue URL and drop postject
from scripts/package.json.

Fixes pingdotgg#23433. Replaces pingdotgg#23443.

Co-authored-by: Lakshmi Tanmay <[email protected]>
Co-authored-by: maria-rcks <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>

* fix(server): preserve completed runs when accepting child deliveries (pingdotgg#25666)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(client-runtime): iOS resume keeps live sockets and drops dead ones 4x faster (pingdotgg#25824)

Co-authored-by: Omar <[email protected]>
Co-authored-by: Carlos D. Escobar-Valbuena <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>

* feat: model labels show the name you'd say out loud (pingdotgg#24205)

Co-authored-by: Claude Opus 5.5 (Claude Code) <[email protected]>
Co-authored-by: maria-rcks <[email protected]>
Co-authored-by: Julius Marminge <[email protected]>

* chore(review): CodeRabbit reviews tests against docs/internals/testing.md (pingdotgg#24204)

* refactor(mobile): showcase capture readiness is a tested pure model (pingdotgg#25816)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): a held thread move retires when a snooze wakes, even with no list open (pingdotgg#25809)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): switching a file between markdown and the editor no longer saves a draft twice (pingdotgg#25768)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web,mobile): queued messages get a simpler Send now row, no reordering (pingdotgg#25846)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): Add project path survives its host dropping out and returning (pingdotgg#25819)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* refactor(mobile): opening Review or Files shows the inspector in the first commit (pingdotgg#25817)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(web): Diagnostics keeps its own settings sidebar entry

Upstream's settings regrouping made Diagnostics a detail page under
Maintenance, dropping the fork's sidebar entry.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

---------

Signed-off-by: Yordis Prieto <[email protected]>
Co-authored-by: Julius Marminge <[email protected]>
Co-authored-by: Claude Opus 5.5 <[email protected]>
Co-authored-by: Theo Browne <[email protected]>
Co-authored-by: Alex Southwell <[email protected]>
Co-authored-by: Alex <[email protected]>
Co-authored-by: project516 <[email protected]>
Co-authored-by: oliver <[email protected]>
Co-authored-by: Yash Singh <[email protected]>
Co-authored-by: shivam <[email protected]>
Co-authored-by: PR Batch Tester <[email protected]>
Co-authored-by: nibarAhmed <[email protected]>
Co-authored-by: maria-rcks <[email protected]>
Co-authored-by: Yordis Prieto <[email protected]>
Co-authored-by: Bilal Bakr <[email protected]>
Co-authored-by: Andrew Gadzik <[email protected]>
Co-authored-by: Julius Marminge <[email protected]>
Co-authored-by: Aditya Vardhan Sharma <[email protected]>
Co-authored-by: gcrft123 <[email protected]>
Co-authored-by: t3-code[bot] <269035359+t3-code[bot]@users.noreply.github.com>
Co-authored-by: none23 <[email protected]>
Co-authored-by: amv146 <[email protected]>
Co-authored-by: Lakshmi Tanmay <[email protected]>
Co-authored-by: Omar <[email protected]>
Co-authored-by: Carlos D. Escobar-Valbuena <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Oct 29, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard passes 'Queued behind a pending request' and stays after the compaction and /goal branches. pingdotgg#25846 renamed the queued message action to Send now, and composer.md names it.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Oct 30, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard passes 'Queued behind a pending request' and stays after the compaction and /goal branches. pingdotgg#25846 renamed the queued message action to Send now, and composer.md names it.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Oct 30, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard passes 'Queued behind a pending request' and stays after the compaction and /goal branches. pingdotgg#25846 renamed the queued message action to Send now, and composer.md names it.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Oct 30, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Oct 30, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Oct 30, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

Copy link
Copy Markdown

I was relying on queued-message reordering to adjust the order of requests as priorities changed, so it’s unfortunate to see it removed. Would you consider bringing it back? Drag-and-drop reordering, like in Codex, would be really useful and intuitive.

Thanks!

immakermatty added a commit to Lazurio/t3code that referenced this pull request Oct 30, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

On v0.1.5 pingdotgg#26715 queues a steer for a provider that can neither steer
its running turn nor restart it. The overlay's pending-request check
runs before that branch and still holds; the row's
ThreadManagementService test gives its projection empty providerThreads
and providerSessions, which the new branch reads.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Oct 31, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

On v0.1.5 pingdotgg#26715 queues a steer for a provider that can neither steer
its running turn nor restart it. The overlay's pending-request check
runs before that branch and still holds; the row's
ThreadManagementService test gives its projection empty providerThreads
and providerSessions, which the new branch reads.

On v0.1.6 upstream still steers into a turn that waits on a pending
runtime request, and the guard still reads the thread's runtimeRequests on
every message.dispatch. Claude's MCP elicitations now wait on the approval
card (pingdotgg#21903) as pending mcp-elicitation runtime requests
(ClaudeAdapterV2.ts), and a steer would cancel them like any other request,
so CommandPolicy.test.ts covers that kind as well. Upstream's Force stop
cases (pingdotgg#27252) and this row's sendToThread cases are two separate
it.effect.each blocks in ThreadManagementService.test.ts.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
kimkhoi2202 added a commit to kimkhoi2202/t3code that referenced this pull request Nov 3, 2026
Upstream removed queue reordering in pingdotgg#25846 as not worth its weight. This
fork wants it back. Each queued message shows its position again, which
turns into a grip on hover: drag it to a new spot, or focus it and press
the up or down arrow. The server never dropped the queued-run.reorder
command, so only the client command and the web controls return.

Unlike before, a held queue with no provider session left can be
reordered too. The rest of upstream's simpler row stays: Send now first,
and no Up next header. Mobile keeps its upstream controls.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
kimkhoi2202 added a commit to kimkhoi2202/t3code that referenced this pull request Nov 4, 2026
Upstream removed queue reordering in pingdotgg#25846 as not worth its weight. This
fork wants it back. Each queued message shows its position again, which
turns into a grip on hover: drag it to a new spot, or focus it and press
the up or down arrow. The server never dropped the queued-run.reorder
command, so only the client command and the web controls return.

Unlike before, a held queue with no provider session left can be
reordered too. The rest of upstream's simpler row stays: Send now first,
and no Up next header. Mobile keeps its upstream controls.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Nov 4, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

On v0.1.5 pingdotgg#26715 queues a steer for a provider that can neither steer
its running turn nor restart it. The overlay's pending-request check
runs before that branch and still holds; the row's
ThreadManagementService test gives its projection empty providerThreads
and providerSessions, which the new branch reads.

On v0.1.6 upstream still steers into a turn that waits on a pending
runtime request, and the guard still reads the thread's runtimeRequests on
every message.dispatch. Claude's MCP elicitations now wait on the approval
card (pingdotgg#21903) as pending mcp-elicitation runtime requests
(ClaudeAdapterV2.ts), and a steer would cancel them like any other request,
so CommandPolicy.test.ts covers that kind as well. Upstream's Force stop
cases (pingdotgg#27252) and this row's sendToThread cases are two separate
it.effect.each blocks in ThreadManagementService.test.ts.

On v0.1.7: upstream lets a delegated task's question steer a busy parent
turn as well (pingdotgg#29909), on the same path as a completion
wake. A steer into a turn that waits on the person would cancel that
person's question, so the guard now covers both: `waitsOnPendingRequest`
comes before the completion-wake and task-question conditions. The
pending-request test also sends a task question and checks that it stays
queued; without the guard it steers.

Also on v0.1.7, a follow-up sent before the run's provider turn started
waited for a turn of its own (#170). resolveDeliveryIntent
queues automatic input while the active run is preparing or starting, or
has no running provider turn of its attempt, and the composer sends auto
while its phase is connecting. A cold start, such as Buddy's agent over
ACP, leaves that gap open for seconds, so the message ran after the turn
instead of steering it. Such a queued run now notes the run it was sent
during (steersRunId, optional in orchestrationV2.ts, from
runAwaitingProviderTurn in CommandPolicy.ts). When that run's provider
turn reports running, the orchestrator promotes those runs to a steer
under the thread lock, only if upstream's resolveMessageDispatchIntent
would steer auto input into that run now (pending request, maintenance
and compaction targets, steering capability) and stopReachedRun allows
it; held runs, /goal commands, notifications and delegation wakes are
left alone. A provider that cannot steer, or a turn that never starts,
keeps upstream's queue. SteeringCompletion.integration.test.ts sends a
follow-up before the turn starts and expects it steered in; without the
reaction it waits in the queue. composer.md and the allowlist follow.

A running provider-turn update repeats with every usage tick, so the
reaction first reads the thread's queued runs without the lock and takes
the lock only while a follow-up waits for its turn. The web composer sends
auto while the turn starts, which the server now steers in; a send the
person makes as Queue (Queue as the default action, or the alternate key
over a steer default) stays queued. resolveComposerDispatchMode takes
turnStarting, which ChatComposer.tsx and composerCommandKey.ts set while
the phase is connecting, and composerDispatch.test.ts fails without it.
The mobile composer already sends queue while a turn starts, as upstream.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Nov 4, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

On v0.1.5 pingdotgg#26715 queues a steer for a provider that can neither steer
its running turn nor restart it. The overlay's pending-request check
runs before that branch and still holds; the row's
ThreadManagementService test gives its projection empty providerThreads
and providerSessions, which the new branch reads.

On v0.1.6 upstream still steers into a turn that waits on a pending
runtime request, and the guard still reads the thread's runtimeRequests on
every message.dispatch. Claude's MCP elicitations now wait on the approval
card (pingdotgg#21903) as pending mcp-elicitation runtime requests
(ClaudeAdapterV2.ts), and a steer would cancel them like any other request,
so CommandPolicy.test.ts covers that kind as well. Upstream's Force stop
cases (pingdotgg#27252) and this row's sendToThread cases are two separate
it.effect.each blocks in ThreadManagementService.test.ts.

On v0.1.7: upstream lets a delegated task's question steer a busy parent
turn as well (pingdotgg#29909), on the same path as a completion
wake. A steer into a turn that waits on the person would cancel that
person's question, so the guard now covers both: `waitsOnPendingRequest`
comes before the completion-wake and task-question conditions. The
pending-request test also sends a task question and checks that it stays
queued; without the guard it steers.

Also on v0.1.7, a follow-up sent before the run's provider turn started
waited for a turn of its own (#170). resolveDeliveryIntent
queues automatic input while the active run is preparing or starting, or
has no running provider turn of its attempt, and upstream's web composer
turns a send made while the run is preparing or starting into Queue
(planComposerTurn in ChatView.logic.ts). A cold start, such as Buddy's
agent over ACP, leaves that gap open for seconds, so the message ran after
the turn instead of steering it. Such a queued run now notes the run it
was sent during (steersRunId, optional in orchestrationV2.ts, from
runAwaitingProviderTurn in CommandPolicy.ts). When that run's provider
turn reports running, the orchestrator promotes those runs to a steer
under the thread lock, only if upstream's resolveMessageDispatchIntent
would steer auto input into that run now (pending request, maintenance and
compaction targets, steering capability) and stopReachedRun allows it;
held runs, /goal commands, notifications and delegation wakes are left
alone. A provider that cannot steer, or a turn that never starts, keeps
upstream's queue. SteeringCompletion.integration.test.ts sends a follow-up
before the turn starts and expects it steered in; without the reaction it
waits in the queue. composer.md and the allowlist follow.

A running provider-turn update repeats with every usage tick. The first
running update of each provider turn takes the thread lock
(makeFirstRunningUpdate in CommandPolicy.ts), because a follow-up committed
alongside it is visible only under the lock; a later one first reads the
thread's queued runs without the lock and takes it only while a follow-up
waits for its turn. Under the lock the reaction reads the queued runs again
and reads the whole projection only while a follow-up waits.
CommandPolicy.test.ts covers runAwaitingProviderTurn and
makeFirstRunningUpdate, and the contract test pins the lock's order; no
test drives the race itself, because nothing between message.dispatch's
read and its commit can be held from a test.

The web composer now keeps auto for a send made while the run is set up,
which the server steers in; planComposerTurn still turns the other choices
and a send that compacts first into Queue, and ChatView.logic.test.ts fails
without the change. A send the person makes as Queue (Queue as the default
action, or the alternate key over a steer default) stays queued.
resolveComposerDispatchMode takes turnStarting, which ChatComposer.tsx and
composerCommandKey.ts set while the phase is connecting, and
composerDispatch.test.ts fails without it. The allowlist adds
ChatView.logic.ts and its test. The official mobile app is upstream's and
still sends queue while a turn starts.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
kimkhoi2202 added a commit to kimkhoi2202/t3code that referenced this pull request Nov 4, 2026
Upstream removed queue reordering in pingdotgg#25846 as not worth its weight. This
fork wants it back. Each queued message shows its position again, which
turns into a grip on hover: drag it to a new spot, or focus it and press
the up or down arrow. The server never dropped the queued-run.reorder
command, so only the client command and the web controls return.

Unlike before, a held queue with no provider session left can be
reordered too. The rest of upstream's simpler row stays: Send now first,
and no Up next header. Mobile keeps its upstream controls.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Nov 5, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

On v0.1.5 pingdotgg#26715 queues a steer for a provider that can neither steer
its running turn nor restart it. The overlay's pending-request check
runs before that branch and still holds; the row's
ThreadManagementService test gives its projection empty providerThreads
and providerSessions, which the new branch reads.

On v0.1.6 upstream still steers into a turn that waits on a pending
runtime request, and the guard still reads the thread's runtimeRequests on
every message.dispatch. Claude's MCP elicitations now wait on the approval
card (pingdotgg#21903) as pending mcp-elicitation runtime requests
(ClaudeAdapterV2.ts), and a steer would cancel them like any other request,
so CommandPolicy.test.ts covers that kind as well. Upstream's Force stop
cases (pingdotgg#27252) and this row's sendToThread cases are two separate
it.effect.each blocks in ThreadManagementService.test.ts.

On v0.1.7: upstream lets a delegated task's question steer a busy parent
turn as well (pingdotgg#29909), on the same path as a completion
wake. A steer into a turn that waits on the person would cancel that
person's question, so the guard now covers both: `waitsOnPendingRequest`
comes before the completion-wake and task-question conditions. The
pending-request test also sends a task question and checks that it stays
queued; without the guard it steers.

Also on v0.1.7, a follow-up sent before the run's provider turn started
waited for a turn of its own (#170). resolveDeliveryIntent
queues automatic input while the active run is preparing or starting, or
has no running provider turn of its attempt, and upstream's web composer
turns a send made while the run is preparing or starting into Queue
(planComposerTurn in ChatView.logic.ts). A cold start, such as Buddy's
agent over ACP, leaves that gap open for seconds, so the message ran after
the turn instead of steering it. Such a queued run now notes the run it
was sent during (steersRunId, optional in orchestrationV2.ts, from
runAwaitingProviderTurn in CommandPolicy.ts). When that run's provider
turn reports running, the orchestrator promotes those runs to a steer
under the thread lock, only if upstream's resolveMessageDispatchIntent
would steer auto input into that run now (pending request, maintenance and
compaction targets, steering capability) and stopReachedRun allows it;
held runs, /goal commands, notifications and delegation wakes are left
alone. A provider that cannot steer, or a turn that never starts, keeps
upstream's queue. SteeringCompletion.integration.test.ts sends a follow-up
before the turn starts and expects it steered in; without the reaction it
waits in the queue. composer.md follows.

A running provider-turn update repeats with every usage tick. The first
running update of each provider turn takes the thread lock
(makeFirstRunningUpdate in CommandPolicy.ts), because a follow-up committed
alongside it is visible only under the lock; a later one first reads the
thread's queued runs without the lock and takes it only while a follow-up
waits for its turn. Under the lock the reaction reads the queued runs again
and reads the whole projection only while a follow-up waits.
CommandPolicy.test.ts covers runAwaitingProviderTurn and
makeFirstRunningUpdate, and the contract test pins the lock's order; no
test drives the race itself, because nothing between message.dispatch's
read and its commit can be held from a test.

The web composer now keeps auto for a send made while the run is set up,
which the server steers in; planComposerTurn still turns the other choices
and a send that compacts first into Queue, and ChatView.logic.test.ts fails
without the change. A send the person makes as Queue (Queue as the default
action, or the alternate key over a steer default) stays queued.
resolveComposerDispatchMode takes turnStarting, which ChatComposer.tsx and
composerCommandKey.ts set while the phase is connecting, and
composerDispatch.test.ts fails without it. Row 5's allowlist names
ChatView.logic.ts and its test, composerCommandKey.ts, composerDispatch.ts
and its test, and orchestrationV2.ts. The official mobile app is upstream's and
still sends queue while a turn starts.

Each queued follow-up gets one promotion, under its own command id
(command:steer-awaiting-turn:<runId>). A refused promotion leaves its
message queued, logs a warning and does not hold back the next follow-up,
and a later usage tick does not retry it. Two integration tests cover
this: a /compact sent during the start no longer blocks the follow-up
after it, and a refused promotion is tried once and the next follow-up
still steers in.

A /goal the person sends into a turn that waits on a question still
interrupts that turn, as on main and upstream; #173 carries
the fix to upstream's pull request. A follow-up sent to Buddy during a
cold start now steers, so row 33 gives the steer Buddy's page as the
Czech block (#150).

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Nov 5, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

On v0.1.5 pingdotgg#26715 queues a steer for a provider that can neither steer
its running turn nor restart it. The overlay's pending-request check
runs before that branch and still holds; the row's
ThreadManagementService test gives its projection empty providerThreads
and providerSessions, which the new branch reads.

On v0.1.6 upstream still steers into a turn that waits on a pending
runtime request, and the guard still reads the thread's runtimeRequests on
every message.dispatch. Claude's MCP elicitations now wait on the approval
card (pingdotgg#21903) as pending mcp-elicitation runtime requests
(ClaudeAdapterV2.ts), and a steer would cancel them like any other request,
so CommandPolicy.test.ts covers that kind as well. Upstream's Force stop
cases (pingdotgg#27252) and this row's sendToThread cases are two separate
it.effect.each blocks in ThreadManagementService.test.ts.

On v0.1.7: upstream lets a delegated task's question steer a busy parent
turn as well (pingdotgg#29909), on the same path as a completion
wake. A steer into a turn that waits on the person would cancel that
person's question, so the guard now covers both: `waitsOnPendingRequest`
comes before the completion-wake and task-question conditions. The
pending-request test also sends a task question and checks that it stays
queued; without the guard it steers.

Also on v0.1.7, a follow-up sent before the run's provider turn started
waited for a turn of its own (#170). resolveDeliveryIntent
queues automatic input while the active run is preparing or starting, or
has no running provider turn of its attempt, and upstream's web composer
turns a send made while the run is preparing or starting into Queue
(planComposerTurn in ChatView.logic.ts). A cold start, such as Buddy's
agent over ACP, leaves that gap open for seconds, so the message ran after
the turn instead of steering it. Such a queued run now notes the run it
was sent during (steersRunId, optional in orchestrationV2.ts, from
runAwaitingProviderTurn in CommandPolicy.ts). When that run's provider
turn reports running, the orchestrator promotes those runs to a steer
under the thread lock, only if upstream's resolveMessageDispatchIntent
would steer auto input into that run now (pending request, maintenance and
compaction targets, steering capability) and stopReachedRun allows it;
held runs, /goal commands, notifications and delegation wakes are left
alone. A provider that cannot steer, or a turn that never starts, keeps
upstream's queue. SteeringCompletion.integration.test.ts sends a follow-up
before the turn starts and expects it steered in; without the reaction it
waits in the queue. composer.md follows.

A running provider-turn update repeats with every usage tick. The first
running update of each provider turn takes the thread lock
(makeFirstRunningUpdate in CommandPolicy.ts), because a follow-up committed
alongside it is visible only under the lock; a later one first reads the
thread's queued runs without the lock and takes it only while a follow-up
waits for its turn. Under the lock the reaction reads the queued runs again
and reads the whole projection only while a follow-up waits.
CommandPolicy.test.ts covers runAwaitingProviderTurn and
makeFirstRunningUpdate, and the contract test pins the lock's order; no
test drives the race itself, because nothing between message.dispatch's
read and its commit can be held from a test.

The web composer now keeps auto for a send made while the run is set up,
which the server steers in; planComposerTurn still turns the other choices
and a send that compacts first into Queue, and ChatView.logic.test.ts fails
without the change. A send the person makes as Queue (Queue as the default
action, or the alternate key over a steer default) stays queued.
resolveComposerDispatchMode takes turnStarting, which ChatComposer.tsx and
composerCommandKey.ts set while the phase is connecting, and
composerDispatch.test.ts fails without it. Row 5's allowlist names
ChatView.logic.ts and its test, composerCommandKey.ts, composerDispatch.ts
and its test, and orchestrationV2.ts. The official mobile app is upstream's and
still sends queue while a turn starts.

Each queued follow-up gets one promotion, under its own command id
(command:steer-awaiting-turn:<runId>). A refused promotion leaves its
message queued, logs a warning and does not hold back the next follow-up,
and a later usage tick does not retry it. Two integration tests cover
this: a /compact sent during the start no longer blocks the follow-up
after it, and a refused promotion is tried once and the next follow-up
still steers in.

A /goal the person sends into a turn that waits on a question still
interrupts that turn, as on main and upstream; #173 carries
the fix to upstream's pull request. A follow-up sent to Buddy during a
cold start now steers, so row 33 gives the steer Buddy's page as the
Czech block (#150).

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Nov 5, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

On v0.1.5 pingdotgg#26715 queues a steer for a provider that can neither steer
its running turn nor restart it. The overlay's pending-request check
runs before that branch and still holds; the row's
ThreadManagementService test gives its projection empty providerThreads
and providerSessions, which the new branch reads.

On v0.1.6 upstream still steers into a turn that waits on a pending
runtime request, and the guard still reads the thread's runtimeRequests on
every message.dispatch. Claude's MCP elicitations now wait on the approval
card (pingdotgg#21903) as pending mcp-elicitation runtime requests
(ClaudeAdapterV2.ts), and a steer would cancel them like any other request,
so CommandPolicy.test.ts covers that kind as well. Upstream's Force stop
cases (pingdotgg#27252) and this row's sendToThread cases are two separate
it.effect.each blocks in ThreadManagementService.test.ts.

On v0.1.7: upstream lets a delegated task's question steer a busy parent
turn as well (pingdotgg#29909), on the same path as a completion
wake. A steer into a turn that waits on the person would cancel that
person's question, so the guard now covers both: `waitsOnPendingRequest`
comes before the completion-wake and task-question conditions. The
pending-request test also sends a task question and checks that it stays
queued; without the guard it steers.

Also on v0.1.7, a follow-up sent before the run's provider turn started
waited for a turn of its own (#170). resolveDeliveryIntent
queues automatic input while the active run is preparing or starting, or
has no running provider turn of its attempt, and upstream's web composer
turns a send made while the run is preparing or starting into Queue
(planComposerTurn in ChatView.logic.ts). A cold start, such as Buddy's
agent over ACP, leaves that gap open for seconds, so the message ran after
the turn instead of steering it. Such a queued run now notes the run it
was sent during (steersRunId, optional in orchestrationV2.ts, from
runAwaitingProviderTurn in CommandPolicy.ts). When that run's provider
turn reports running, the orchestrator promotes those runs to a steer
under the thread lock, only if upstream's resolveMessageDispatchIntent
would steer auto input into that run now (pending request, maintenance and
compaction targets, steering capability) and stopReachedRun allows it;
held runs, /goal commands, notifications and delegation wakes are left
alone. A provider that cannot steer, or a turn that never starts, keeps
upstream's queue. SteeringCompletion.integration.test.ts sends a follow-up
before the turn starts and expects it steered in; without the reaction it
waits in the queue. composer.md follows.

A running provider-turn update repeats with every usage tick. The first
running update of each provider turn takes the thread lock
(makeFirstRunningUpdate in CommandPolicy.ts), because a follow-up committed
alongside it is visible only under the lock; a later one first reads the
thread's queued runs without the lock and takes it only while a follow-up
waits for its turn. Under the lock the reaction reads the queued runs again
and reads the whole projection only while a follow-up waits.
CommandPolicy.test.ts covers runAwaitingProviderTurn and
makeFirstRunningUpdate, and the contract test pins the lock's order; no
test drives the race itself, because nothing between message.dispatch's
read and its commit can be held from a test.

The web composer now keeps auto for a send made while the run is set up,
which the server steers in; planComposerTurn still turns the other choices
and a send that compacts first into Queue, and ChatView.logic.test.ts fails
without the change. A send the person makes as Queue (Queue as the default
action, or the alternate key over a steer default) stays queued.
resolveComposerDispatchMode takes turnStarting, which ChatComposer.tsx and
composerCommandKey.ts set while the phase is connecting, and
composerDispatch.test.ts fails without it. Row 5's allowlist names
ChatView.logic.ts and its test, composerCommandKey.ts, composerDispatch.ts
and its test, and orchestrationV2.ts. The official mobile app is upstream's and
still sends queue while a turn starts.

Each queued follow-up gets one promotion, under its own command id
(command:steer-awaiting-turn:<runId>). A refused promotion leaves its
message queued, logs a warning and does not hold back the next follow-up,
and a later usage tick does not retry it. Two integration tests cover
this: a /compact sent during the start no longer blocks the follow-up
after it, and a refused promotion is tried once and the next follow-up
still steers in.

A /goal the person sends into a turn that waits on a question still
interrupts that turn, as on main and upstream; #173 carries
the fix to upstream's pull request. A follow-up sent to Buddy during a
cold start now steers, so row 33 gives the steer Buddy's page as the
Czech block (#150).

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Nov 5, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

On v0.1.5 pingdotgg#26715 queues a steer for a provider that can neither steer
its running turn nor restart it. The overlay's pending-request check
runs before that branch and still holds; the row's
ThreadManagementService test gives its projection empty providerThreads
and providerSessions, which the new branch reads.

On v0.1.6 upstream still steers into a turn that waits on a pending
runtime request, and the guard still reads the thread's runtimeRequests on
every message.dispatch. Claude's MCP elicitations now wait on the approval
card (pingdotgg#21903) as pending mcp-elicitation runtime requests
(ClaudeAdapterV2.ts), and a steer would cancel them like any other request,
so CommandPolicy.test.ts covers that kind as well. Upstream's Force stop
cases (pingdotgg#27252) and this row's sendToThread cases are two separate
it.effect.each blocks in ThreadManagementService.test.ts.

On v0.1.7: upstream lets a delegated task's question steer a busy parent
turn as well (pingdotgg#29909), on the same path as a completion
wake. A steer into a turn that waits on the person would cancel that
person's question, so the guard now covers both: `waitsOnPendingRequest`
comes before the completion-wake and task-question conditions. The
pending-request test also sends a task question and checks that it stays
queued; without the guard it steers.

Also on v0.1.7, a follow-up sent before the run's provider turn started
waited for a turn of its own (#170). resolveDeliveryIntent
queues automatic input while the active run is preparing or starting, or
has no running provider turn of its attempt, and upstream's web composer
turns a send made while the run is preparing or starting into Queue
(planComposerTurn in ChatView.logic.ts). A cold start, such as Buddy's
agent over ACP, leaves that gap open for seconds, so the message ran after
the turn instead of steering it. Such a queued run now notes the run it
was sent during (steersRunId, optional in orchestrationV2.ts, from
runAwaitingProviderTurn in CommandPolicy.ts). When that run's provider
turn reports running, the orchestrator promotes those runs to a steer
under the thread lock, only if upstream's resolveMessageDispatchIntent
would steer auto input into that run now (pending request, maintenance and
compaction targets, steering capability) and stopReachedRun allows it;
held runs, /goal commands, notifications and delegation wakes are left
alone. A provider that cannot steer, or a turn that never starts, keeps
upstream's queue. SteeringCompletion.integration.test.ts sends a follow-up
before the turn starts and expects it steered in; without the reaction it
waits in the queue. composer.md follows.

A running provider-turn update repeats with every usage tick. The first
running update of each provider turn takes the thread lock
(makeFirstRunningUpdate in CommandPolicy.ts), because a follow-up committed
alongside it is visible only under the lock; a later one first reads the
thread's queued runs without the lock and takes it only while a follow-up
waits for its turn. Under the lock the reaction reads the queued runs again
and reads the whole projection only while a follow-up waits.
CommandPolicy.test.ts covers runAwaitingProviderTurn and
makeFirstRunningUpdate, and the contract test pins the lock's order; no
test drives the race itself, because nothing between message.dispatch's
read and its commit can be held from a test.

The web composer now keeps auto for a send made while the run is set up,
which the server steers in; planComposerTurn still turns the other choices
and a send that compacts first into Queue, and ChatView.logic.test.ts fails
without the change. A send the person makes as Queue (Queue as the default
action, or the alternate key over a steer default) stays queued.
resolveComposerDispatchMode takes turnStarting, which ChatComposer.tsx and
composerCommandKey.ts set while the phase is connecting, and
composerDispatch.test.ts fails without it. Row 5's allowlist names
ChatView.logic.ts and its test, composerCommandKey.ts, composerDispatch.ts
and its test, and orchestrationV2.ts. The official mobile app is upstream's and
still sends queue while a turn starts.

Each queued follow-up gets one promotion, under its own command id
(command:steer-awaiting-turn:<runId>). A refused promotion leaves its
message queued, logs a warning and does not hold back the next follow-up,
and a later usage tick does not retry it. Two integration tests cover
this: a /compact sent during the start no longer blocks the follow-up
after it, and a refused promotion is tried once and the next follow-up
still steers in.

A /goal the person sends into a turn that waits on a question still
interrupts that turn, as on main and upstream; #173 carries
the fix to upstream's pull request. A follow-up sent to Buddy during a
cold start now steers, so row 33 gives the steer Buddy's page as the
Czech block (#150).

Co-Authored-By: Claude Opus 5.5 <[email protected]>
immakermatty added a commit to Lazurio/t3code that referenced this pull request Nov 5, 2026
From v0.1.3 a follow-up steers the running turn by default
(pingdotgg#24718). A steer into a turn that waits on a pending
runtime request aborts the provider's pending callback: Claude steers with
priority "now", a pending AskUserQuestion resolves as cancelled and an
approval as cancel, and the person never sees the question
(#139). #119 guarded only sendToThread's auto
mode; the composer's explicit steer, the auto branch of the delivery intent
and promoting a queued message still steered through it.

The guard now lives in one place, CommandPolicy.resolveMessageDispatchIntent:
while the thread has a pending runtime request, any steer_active becomes
queue_after_active. It covers a steer requested as a delivery intent (auto or
steer) and one requested as a dispatch mode (sendToThread, older clients).
The orchestrator reads the thread with its runtime requests on every
message.dispatch, so no caller has to load them. Promoting a queued message
to a steer checks the same predicate (waitsOnPendingRequest) and puts the
message first in the queue instead, so it runs right after the answer.
Restart is unchanged.

sendToThread's own check from #119 is gone, and
ThreadManagementService.ts is upstream's again; its test resolves the
dispatched mode through the guard, and an explicit steer with a pending
request now queues too. The allowlist trades ThreadManagementService.ts for
CommandPolicy.ts, Orchestrator.ts and their tests, and the user guides for
the composer and the agent CLI say a steer queues while a question waits.

On v0.1.4: upstream renamed stopRunForGoalCommand to
stopRunForQueuedCommand with a required reason (pingdotgg#25837); the guard
passes 'Queued behind a pending request' and stays after the compaction
and /goal branches. pingdotgg#25846 renamed the queued message action to Send
now, and composer.md names it. A delegated task's completion wake, which
v0.1.4's remote delegation can also produce (RemoteDelegation.ts), set
steer_active after the guard, as it did on the v0.1.3 line; it now
steers only while waitsOnPendingRequest is false, and
SteeringCompletion.integration.test.ts queues such a wake behind a
pending request.

On v0.1.5 pingdotgg#26715 queues a steer for a provider that can neither steer
its running turn nor restart it. The overlay's pending-request check
runs before that branch and still holds; the row's
ThreadManagementService test gives its projection empty providerThreads
and providerSessions, which the new branch reads.

On v0.1.6 upstream still steers into a turn that waits on a pending
runtime request, and the guard still reads the thread's runtimeRequests on
every message.dispatch. Claude's MCP elicitations now wait on the approval
card (pingdotgg#21903) as pending mcp-elicitation runtime requests
(ClaudeAdapterV2.ts), and a steer would cancel them like any other request,
so CommandPolicy.test.ts covers that kind as well. Upstream's Force stop
cases (pingdotgg#27252) and this row's sendToThread cases are two separate
it.effect.each blocks in ThreadManagementService.test.ts.

On v0.1.7: upstream lets a delegated task's question steer a busy parent
turn as well (pingdotgg#29909), on the same path as a completion
wake. A steer into a turn that waits on the person would cancel that
person's question, so the guard now covers both: `waitsOnPendingRequest`
comes before the completion-wake and task-question conditions. The
pending-request test also sends a task question and checks that it stays
queued; without the guard it steers.

Also on v0.1.7, a follow-up sent before the run's provider turn started
waited for a turn of its own (#170). resolveDeliveryIntent
queues automatic input while the active run is preparing or starting, or
has no running provider turn of its attempt, and upstream's web composer
turns a send made while the run is preparing or starting into Queue
(planComposerTurn in ChatView.logic.ts). A cold start, such as Buddy's
agent over ACP, leaves that gap open for seconds, so the message ran after
the turn instead of steering it. Such a queued run now notes the run it
was sent during (steersRunId, optional in orchestrationV2.ts, from
runAwaitingProviderTurn in CommandPolicy.ts). When that run's provider
turn reports running, the orchestrator promotes those runs to a steer
under the thread lock, only if upstream's resolveMessageDispatchIntent
would steer auto input into that run now (pending request, maintenance and
compaction targets, steering capability) and stopReachedRun allows it;
held runs, /goal commands, notifications and delegation wakes are left
alone. A provider that cannot steer, or a turn that never starts, keeps
upstream's queue. SteeringCompletion.integration.test.ts sends a follow-up
before the turn starts and expects it steered in; without the reaction it
waits in the queue. composer.md follows.

A running provider-turn update repeats with every usage tick. The first
running update of each provider turn takes the thread lock
(makeFirstRunningUpdate in CommandPolicy.ts), because a follow-up committed
alongside it is visible only under the lock; a later one first reads the
thread's queued runs without the lock and takes it only while a follow-up
waits for its turn. Under the lock the reaction reads the queued runs again
and reads the whole projection only while a follow-up waits.
CommandPolicy.test.ts covers runAwaitingProviderTurn and
makeFirstRunningUpdate, and the contract test pins the lock's order; no
test drives the race itself, because nothing between message.dispatch's
read and its commit can be held from a test.

The web composer now keeps auto for a send made while the run is set up,
which the server steers in; planComposerTurn still turns the other choices
and a send that compacts first into Queue, and ChatView.logic.test.ts fails
without the change. A send the person makes as Queue (Queue as the default
action, or the alternate key over a steer default) stays queued.
resolveComposerDispatchMode takes turnStarting, which ChatComposer.tsx and
composerCommandKey.ts set while the phase is connecting, and
composerDispatch.test.ts fails without it. Row 5's allowlist names
ChatView.logic.ts and its test, composerCommandKey.ts, composerDispatch.ts
and its test, and orchestrationV2.ts. The official mobile app is upstream's and
still sends queue while a turn starts.

Each queued follow-up gets one promotion, under its own command id
(command:steer-awaiting-turn:<runId>). A refused promotion leaves its
message queued, logs a warning and does not hold back the next follow-up,
and a later usage tick does not retry it. Two integration tests cover
this: a /compact sent during the start no longer blocks the follow-up
after it, and a refused promotion is tried once and the next follow-up
still steers in.

A /goal the person sends into a turn that waits on a question still
interrupts that turn, as on main and upstream; #173 carries
the fix to upstream's pull request. A follow-up sent to Buddy during a
cold start now steers, so row 33 gives the steer Buddy's page as the
Czech block (#150).

Co-Authored-By: Claude Opus 5.5 <[email protected]>
kimkhoi2202 added a commit to kimkhoi2202/t3code that referenced this pull request Nov 6, 2026
Upstream removed queue reordering in pingdotgg#25846 as not worth its weight. This
fork wants it back. Each queued message shows its position again, which
turns into a grip on hover: drag it to a new spot, or focus it and press
the up or down arrow. The server never dropped the queued-run.reorder
command, so only the client command and the web controls return.

Unlike before, a held queue with no provider session left can be
reordered too. The rest of upstream's simpler row stays: Send now first,
and no Up next header. Mobile keeps its upstream controls.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

size:XL 500-999 changed lines (additions + deletions). vouch:trusted Author is trusted by repo permissions or the VOUCHED list.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants