* fix(ci): run the release skill-sharing Electron install step in bash
#38130 turned this step into a POSIX path fallback, but the skill sharing
release gate also runs on windows-2022, where the default shell is
PowerShell. The step failed with a ParserError, so release-preflight
skipped every build for the v1.4.250 cut.
* test(e2e): keep the one-time Codex notice out of golden runs
The v1.4.250 golden e2e on Linux timed out clicking New Terminal: the
'Orca now runs Codex without its shared server' toast fires once when the
golden stub (a Codex terminal) appears, and here it landed while the +
menu was open and closed it. Mark it seen in the completed-user fixture
like the other one-time notices, and restore that fixture from the
workflow ref in the release golden job so a recovered tag uses it.
---------
Co-authored-by: m4air <m4air@Mac.localdomain>
* Let a finished worker keep messaging its coordinator
* Tell a finished remote worker why its coordinator is out of reach
* Let a finished worker's ask wait for its answer instead of starting a follow-up
* Share the assignee check, keep the finished lookup in dispatch-lookup, and fix review findings
* Match a finished terminal worker by its process too; cover resume after a follow-up
Claude Code shows its next-prompt suggestion as dim text in the empty
composer. The orchestration draft hold read it as the user's unsent draft
and held wakes for up to 10 minutes after the user last typed. The hold
now asks the typed-only composer reader (the one native chat already uses
for a terminal's unsent input), so dim suggestions and placeholders no
longer count, while real typed drafts still hold delivery.
Fixes#38015. Based on the diagnosis in #38015 and the approach explored
in #38077.
* fix(orchestration): let the task spec decide whether a worker pushes or opens a PR
The worker brief said "Don't post to Slack, GitHub, or other channels during the run." Codex read that as forbidding pushes and pull requests even when the task asked for them. The rule began as a reporting rule (reach the coordinator through the CLI, not Slack or GitHub comments), so fold it back into that sentence.
Fixes#38111
* fix(orchestration): drop the Slack and GitHub clause from the worker brief
* fix(contextual-tours): don't restart the agent sessions tour for existing users
v1.4.249 prepended a "New agents start here" step to the workspace agent
sessions tour. No earlier action completes it, so long-time users who never
closed the old tour (usually because they had already split panes) now see
it from step 1. Mark it seen at startup for profiles that finished
onboarding before that release; new users still get the full tour.
* test(contextual-tours): cover which profiles skip the agent sessions tour
Profiles still in onboarding, or onboarded at or after the v1.4.249 release,
keep the tour; earlier profiles get it marked seen without losing other seen tours.
* fix(contextual-tours): don't reopen the agent sessions tour for users who already split
Replace the startup cutoff backfill with a rule derived from what the user did:
a tour never auto-starts after one of its actions is already done (unless it is
already running this session or replayed), and a start never lands before the
latest done step, so steps added ahead of the user's progress can't rewind it.
* perf(contextual-tours): stop watching the page for a tour progress already blocks
* fix(contextual-tours): decide the progress block inside the tour effect
Selecting the derived flag ran it against partial test stores during render; read
the raw fields and decide in the effect, like the session-consumed check.
---------
Co-authored-by: m4air <m4air@Mac.localdomain>
Co-authored-by: Brennan Benson <79079362+brennanb2025@users.noreply.github.com>
On a host whose temp directory is full or unwritable, every worker_done
failed and the task never settled (#38123). The settlement runs inside
savepoints, and SQLite keeps savepoint journals in a temp file by default.
When that file cannot be written, SQLite rolls back the whole transaction
itself, so our own rollback then failed with 'cannot rollback - no
transaction is active' and hid the real error. Heartbeats and questions
use no savepoints, which is why they kept working.
The orchestration database now keeps temp storage in memory, and the
lifecycle rollback skips when SQLite already ended the transaction so any
remaining failure of this kind reports its own error.
#38325 made the question card's answer field a textarea, so
findNativeChatAnswerField (HTMLInputElement-only) can no longer match.
The bridge's 'origin instanceof HTMLInputElement' branch was already
unreachable since #32035 stopped the context menu passing a paste origin;
the only remaining origin caller (app-menu paste) passes the queued-editor
textarea from findNativeChatPasteField. Remove the branch, the helper, and
the test that only exercised it by calling the bridge directly.
* Highlight the first agent search match that Enter picks
Co-Authored-By: Claude <noreply@anthropic.com>
* Move the search-default highlight into the dropdown menu primitive
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* test: trim redundant notebook and editor tests
* test: remove implementation-budget notebook and PDF tests
---------
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
Tokenize visible diff-pane lines when a diff editor's model, hidden areas or layout change, so outer-scrolled combined-diff sections paint coloured (STA-12770).
Use the shared agent picker for Source Control Customize and AI-author setup. Preserve existing Arguments, templates, Custom commands, model/effort choices and settings ownership; keep nested menus and tooltips within their dialog portal.
Co-authored-by: Jinjing <6427696+AmethystLiang@users.noreply.github.com>
Co-authored-by: Courtenay <1909814+hdsheena@users.noreply.github.com>
Co-authored-by: Andy Zhang <hiandyzhang@gmail.com>
* fix(pi): keep a native chat awake while its background subagents run
The idle sweep stopped a Pi chat 30 quiet minutes after its last turn even
while pi-subagents workers it had started in the background were still
running. On Windows those workers are children of Pi and died with it, so
the coordinator never got their results.
Pi extensions with detached work already declare it through the shared
session-liveness registry that Pi hosts read before idle eviction, and
pi-subagents registers there. Native chat runs Pi as a child process, so
Orca now loads a small bridge extension into native-chat Pi launches that
installs that registry at load and reports, over Pi's RPC stdout, whether
the session has active providers. The Pi adapter keeps the last report and
the idle sweep treats it like provider-held sends: the chat is not stopped
while work is declared, and idles a full window after it ends. Nothing else
reads it, so clear and compaction are unchanged.
Tracked in #28710.
* fix(pi): harden the background-work bridge after review
Keep the bridge ticking after a failed report so the next answer is
re-derived, validate registrations the same way the reference host does,
log a bridge that could not be written, and use type guards in its test.
* fix(pi): confirm Pi background work at idle-stop time
A cached report could be stale when work starts between samples, so the
idle sweep now asks a running child's provider right before stopping it
and stops only on a fresh idle answer. Pi's bridge answers requests that
Orca drops into a private per-child folder, replying on its status line,
so the probe never becomes a dialog or a prompt other extensions or the
model would see. A child launched without the bridge stops as before;
one with the bridge but no ready instance is left for the next sweep.
* test(pi): replay the bridge's recorded probe answers
Recorded on Windows from Pi 1.5.0 with pi-subagents 0.84.0 and the bridge:
ready, active while a worker ran, idle after its wake turn, and shutdown.
* fix(pi): let the Pi child own its probe folder and reject malformed answers
A launch abandoned before Pi started left its prepared folder behind, so
the bridge now creates the folder itself. An answer with no probe pending
could throw while reading Pi's output; replies are now matched exactly and
anything else times out as unknown. Leftover folders are reclaimed only
when their Orca process is proven gone.
* test(pi): reproduce an unasked-for answer from the ready bridge
* fix(native-chat): apply an idle window changed while a provider answered
The idle stop now re-reads the chat idle setting after the provider's
answer, so switching a chat to Never during the probe keeps its agent.
Normalize each Pi launch's private probe folder in the launch parity
snapshots, which now include the background-work bridge.
* fix(session-tabs): send per-client frames without projecting them twice
A frame sent to one client (a tab it opened, picked or reopened) was already
that client's projection, but delivery projected it again. That stacked the
:client-navigation epoch suffix and the selection revision, so the client
treated it as another publication and fenced it. A link clicked in a server
browser page then only brought its new tab forward the first time.
* test(session-tabs): fold the per-client frame publication check into the caller-selection test
* fix(mobile): the chat Send button keeps desktop's arrow while the agent works
#36588 swapped the phone's send icon and label for Queue or Steer while
a turn ran; desktop keeps the plain Send arrow and offers the other way
through Cmd/Ctrl+Enter. The phone now matches, with hold as its
alternate.
* test(mobile): the send button keeps its arrow while the agent works