Commit Graph
4974 Commits
Author SHA1 Message Date
Brennan Benson 80937f7eed Show a Fast-only speed as the Fast mode switch on mobile (#40523)
On the phone, Codex's Standard/Fast speed was a Speed row opening a
list while Claude's Fast mode was a switch. A Fast-only speed now uses
the same switch row; more speeds, or a lone Flex, keep the list. The
lone-Fast rule moves to shared code so desktop and mobile follow one
rule.
2026-11-09 18:11:52 -08:00
Brennan Benson 1e967661a8 Fix false Codex worker readiness after startup (#40246)
* Keep Codex session startup separate from prompt submission

* Use normalized hook timestamps in CLI regression fixture

* Preserve root evidence across cached starts and relay child hooks

* Check cached starts cannot refresh expired root evidence

* Preserve Codex child ownership through cached session starts

* Keep session startup context within its terminal and session

* Narrow the eligible Codex context directly for typechecking

* Keep Codex turn facts scoped to their root session
2026-11-09 17:35:42 -08:00
Neil 31eff247fa refactor(renderer): organize work-item and review defining owners (#40562) 2026-11-09 17:08:01 -08:00
Jinjing d9e2f485f3 fix(agent-picker): preserve model discovery across queued waits (#40545)
* Keep saved agent models searchable while catalogs refresh

- Load saved catalogs independently of pending refreshes.
- Preserve edited presets until a completed catalog confirms removal.

* wip

* wip

* Preserve model discovery status when a read times out

Keep catalog responses provisional while discovery continues, seed account
catalogs when late results arrive, and retire queued discovery on shutdown.

* Simplify model catalog deadlines and picker state
2026-11-09 16:40:51 -08:00
Kelvin Amoaba 66dcf33b54 fix(native-chat): Continue in Chat UI on an unsent Claude fork brings back the chat it came from (#40551)
* fix(native-chat): bring back the chat an unsaved terminal fork came from on Continue in Chat UI

* fix(agent-hooks): name a forked Claude terminal's session from its fork SessionStart

* fix(native-chat): bind a fork's origin chat to the fork's own conversation

* fix(native-chat): bind a fork to its launched pane and keep the stand-in chat's draft
2026-11-10 00:28:54 +00:00
Brennan Benson 55db56d989 Make Settings easier to scan and organize General (#40249)
* Simplify Settings chrome and organize General into focused views

* Preserve search and skill actions across Settings views

* Correct Settings validation leftovers

* Keep skill conflict details visible in compact Settings rows

* Refresh Settings search and protected copy catalogs

* Use Settings design tokens and button variants

* Keep Settings test setup and translation catalogs current

* Keep Settings search metadata independent of display labels

* Build repository search groups only for the selected pane

* Keep pending skill installs and server menu labels accurate

* Use the supported role locator in the skill regression

* Keep absent skill installation actions on their supported path

* Restore the Settings sidebar to its existing appearance

* Restore shared navigation title keys in pane headings

* Preserve the Settings shell entry animation

* Drop unused Voice labels after the main sync

* Preserve main Voice translations and the moved skill module import
2026-11-09 16:12:36 -08:00
Jinwoo Hong 533a31b742 fix(agent-launch): restarted agents keep the launch they ran with (#38998)
* feat(agent-session): keep the permission mode a Claude session reports

Claude names its live permission mode on prompt, tool and Stop hooks. Keep the
root agent's latest value on the provider session, next to effort, so a resume
without a recorded launch can tell whether the session ran in bypass.

Ports the capture from #36359 by Nelef.

* fix(agent-launch): resume an unrecorded launch bare unless it reported bypass

Without a recorded launch, every restart path now resumes with no settings
Arguments, env or permission default; the configured Command still runs. A
session whose root agent last reported bypassPermissions keeps its bypass flag,
so a hand-typed claude --dangerously-skip-permissions keeps it while a bare
hand-typed claude never gains one (#29898).

Account-switch restart used the shared rule only for Claude; it now uses it
for every agent. Thanks to JustinCMR (#37304) for finding the account-switch
and cold-restore gaps.

* fix(terminal): keep the launch recipe of a window-started agent in main

A window spawn hands main its recipe, but main kept only the token and agent,
so restarts that main runs (workspace sleep capture, closed-worker reopen)
found no recipe and resumed bare. Keep it with the launch authority, the same
gate that admits the token.

* fix(orchestration): reopen a closed worker with the recipe it launched with

A closed worker's pane record goes away with its tab, so orchestration send
reopened it bare (#37693). The worker row now keeps the recipe Orca launched
the agent with, keeps it when a later capture of the same launch arrives
without one (main loses it on a daemon reattach), and a reopen prefers it over
a pane record that may hold another launch of the conversation.

Adapted from #37757 by Pr1p, without a new column.

* fix(terminal): keep a slept pane's earlier recipe only for the same launch

A host sleep capture without a recipe reused the pane's earlier record for the
same conversation. That is right after a daemon reattach, when Orca's launch
still runs but main lost its recipe, and wrong for an agent typed into the
shell that resumed the conversation: it never ran that launch.

* fix(agent-session): restore a paired pane with the recipe its host recorded

A client restoring a pane on a paired host sends only its recorded Arguments,
and the host filled the rest from its current settings. A saved restore now
replays the host's own record for that pane, or, with none at all, resumes
bare and keeps a bypass only when this host's hooks saw it reported. History
and other user-chosen resumes keep current settings.

* fix(terminal): replay a saved note only for the launch main still holds

A note belongs to the agent in a pane only while main holds that launch's
token: main admits it at spawn and retires it when the launch command
finishes, and a daemon reattach restores neither. Worker, host sleep and
hook/quit captures now take a note only from that attested launch, and no
longer copy an earlier record's note by agent or conversation match, which a
typed resume in the same shell also satisfies. With no attested launch the
unrecorded rule applies.

Replaces the inferred orcaLaunched / recipeLostPaneKeys flags and drops the
mock-only reopen and fixture-echo worker tests.

* fix(agent-status): take a pane's note from the launch main attests

The window carried a pane's saved note into each new record by agent and
conversation match, so an agent typed into the shell that resumed the same
conversation inherited an earlier launch's bypass. Main now sends the note of
the launch it still holds with each status, and the window records only that,
its own launch registration, or, for a restored replay of the very status a
record was written from, that record's note. /clear and compact keep the note
because main still holds the same launch.

* fix(orchestration): reopen a worker with its own recipe or none

A worker without its own note took the note of whichever pane record held the
same conversation. Its reopen now always passes a complete recipe: its own
note, or the unrecorded rule. The unrecorded bypass is also scoped to Claude,
whose hooks are the ones that report a root agent's permission mode.

* fix(agent-session): restore a paired pane from its source pane, never current settings

A saved restore on a paired host looked up the destination pane, which a Wake
opens as a new tab, and fell back to the host's current settings whenever the
client sent its partial Arguments. The client now names the source pane (to
hosts advertising agent-session.restore-source-pane.v1), and the host replays
its note for that pane or, with none, resumes bare keeping only a bypass its
own hooks saw there. Older hosts keep today's behavior.

* fix(terminal): re-arm a launch across a main restart while its agent still runs

The daemon outlives main, so it now keeps main's admitted launch (token and
note) and drops it when that launch's command finishes. On reattach main
re-arms the launch only when the same agent still runs in the foreground:
nothing can be typed into a daemon PTY while main is down, so a running agent
there is the one the unretired note launched. A shell foreground clears the
daemon's note; an unknown read, including every Windows read, declines.

Older daemons keep no note, so restore declines there as before.

* fix(terminal): the daemon retires a launch note when its command finishes

A launch can finish while main is down, so main's own retirement never sees it and a
hand-typed agent in that pane would later pass the reattach foreground check.

* fix(agent-status): a fresh Claude start never inherits the last process's permission mode

A hand-typed resume of the same conversation may run without the bypass the previous
process had, so only /clear and compact carry the mode forward.

* fix(agent-session): name a restore's source pane in the host's own keys

A mirrored pane's key carries the desktop tab prefix, so the host's exact lookup missed
its saved note.

* fix(terminal): retire a daemon launch note by its token from the one retirement owner

A delayed reattach result or a late retire could clear a newer launch's note. Retires now
name the token they retire, the reattach checks its PTY before clearing, and re-arming the
same launch keeps the daemon's completion scanner.

* fix(agent-status): only the launch owner's statement changes a pane's note

Main now states the attested note, or null for none, on every hook status. An OSC or
display update makes no statement, so after a completed turn it no longer erases an Orca
launch's note; a typed resume still drops it through main's null.

* fix(terminal): an attach leaves alone a launch main already holds

The renderer attaches a pane the CLI just launched before its agent starts, so the
attach saw the shell and retired the new launch's daemon note.

* fix(agent-status): a paired client keeps a pane's note only on its host owner's word

The host's session-tabs frame now says whether the Orca launch it admitted still runs on
the published incarnation, and the client applies that fact whichever writer wins the
status. With an owner's word the conversation no longer decides; an older host that says
nothing keeps the same-conversation rule.

* refactor(agent-status): one decision picks a pane's note for every reader

The live entry, quit capture, provider-session updates, account restart and cold
restore now ask the same owner-aware decision. An owner's none is final and, when it
turns, ends the pane's registry entry and saved note, locally and from a paired host's
frame. Cold restore and account restart take that answer for a live invocation and read
the saved record only when none is live. Inlines single-use wrappers.

* refactor(agent-status): a paired retirement test reaches the cold-restore command; trim comments

Replaces the projection and continuity tests with one transition: host frame retires the
launch, the next OSC keeps no note, and the cold-restore command has no old bypass.

* fix(agent-status): an owner's none retires the pane's matched launch registry entry

* chore: satisfy changed-code gates (type guards, SAFETY notes, fixed test clock)

* fix(agent-launch): a teams launch keeps its note; a legacy leaf id restores without throwing

* refactor(daemon): keep the session launch note on the process base so session.ts stays in budget

* fix(agent-launch): a restarted terminal main can't vouch for keeps main's earlier launch rule

* fix(agent-launch): read the pane's current terminal; a finished launch settles an undecided pane

* fix(agent-launch): undecided and held launches end with their process

A finished command, a fresh launch or a replacement process settles an undecided pane;
a newer attach or finish supersedes a foreground read still in flight; a legacy relay
reattach without an incarnation stays undecided; a held launch applies only to the
process it was recorded for.

* refactor(terminal): watch and arm a created terminal's launch in one call

Keeps terminal-create.ts within the 300-line budget after main's growth.

* refactor(terminal): keep main's launch evidence in one immutable record

Replace the undecided flag, the reattach check map and the re-arm's writes
to main's orchestration launch fields with one launchEvidence object
(unknown, gone for an incarnation, or held with its recipe), owned by
launch-evidence.ts.

- A fresh admitted launch is held; any other fresh process is gone.
- A finished command or confirmed exit is gone, sticky for its process:
  no reattach, legacy or noted, can revive it.
- A replacement retires its predecessor's evidence without deciding the
  newcomer, which may still prove its own launch through a reattach.
- A reattach note is held only after a foreground read sees its agent;
  the evidence object fences that read.
- A gone launch clears its recipe and blocks both refill sites; a
  deliberate sleep snapshot keeps its recipe past the launch it captured.
- launchToken, launchIncarnationId and launchAgent writes match main.

* fix(terminal): keep launch evidence through a first incarnation and sleep snapshots

- Learning a process's first incarnation binds its held or gone evidence
  to it, so a later attach can neither revive a finished launch nor drop
  a running one; a known X->Y replacement is fenced as before and now also
  clears the predecessor's recipe.
- A status from no Orca launch leaves a sleep snapshot of its session for
  the wake and never takes that snapshot's recipe into a live or quit
  record.

* fix(terminal): record a runtime split's fresh shell as running no launch

* chore: rerun CI on main with the typecheck fix
2026-11-09 19:06:43 -05:00
Kelvin Amoaba 258d3b375a perf(mobile): brotli-compress the page build (#40532) 2026-11-10 00:03:30 +00:00
Neil c0d7a90729 refactor: break runtime import cycles; check import/no-cycle at full depth (#39732)
Extract shared state/helpers into leaf modules (pure moves, re-exports kept)
so import/no-cycle can run at full depth with type-only imports ignored.
120 src + 8 mobile cycles -> 0. 10 suppressions remain on the browserManager
singleton hub, which needs DI to break.
2026-11-09 15:45:32 -08:00
Kelvin Amoaba 4cd3088380 feat(settings): one-click dictation setup with a model suggested for your language (#40462) 2026-11-09 23:39:32 +00:00
Jinjing 6df5a57077 Guide agent installation and sign-in from chat (#40208)
* wip

* add treanslation

* wip

* fix the renderer import

* fix static analysis

* wip

* wip
2026-11-09 23:33:19 +00:00
Jinjing 3e86c5ac73 fix(i18n): translate sparse preset validation failures (#40277)
* fix(i18n): translate sparse preset validation failures

* fix: classify sparse directory validation by type
2026-11-09 15:12:50 -08:00
Brennan Benson 50ecb96893 Offer only the speeds the model lists (#40257)
A speed the chat reports that no model lists, in practice Flex from the
user's own Codex config, is no longer added to the Speed choices. It
still runs and is still sent. A chat with no model of its own keeps its
saved speed when some model lists it, so Standard can still turn it off.
2026-11-09 15:08:48 -08:00
Brennan Benson 9174df298d fix(claude): isolate new launches from parent sessions (#40281)
* fix(claude): isolate independent launches from parent session environments

* test(claude): use supported account environment fixture
2026-11-09 15:06:28 -08:00
Kelvin Amoaba 1f63f18a45 feat(browser): simplify the toolbar to one Annotate button and one menu (#40501)
Grab, Annotate and Draw become one Annotate split button. DevTools, open
in default browser, hard reload, find and extensions move into the menu,
with the profile list, new profile and cookie import under Profile. Zoom,
viewport size, password offers and file actions show as chips inside a
flat address field, and the Import hint moves to the new tab page. The row
drops from 50 to 40 px and the fold from nine stages to two.
2026-11-09 23:03:58 +00:00
Kelvin AmoabaandEng. Mahmoud Bek ElNemer bd8138b231 fix(mobile): keep the selected GitHub project when switching projects (#39964)
* fix(mobile): keep the selected GitHub project when switching projects

Picking another GitHub Project on the phone snapped back to the first one.
A remembered desktop Board view made the pick throw before it committed,
and a committed pick wrote picker rows (title, summary) into settings,
which the host's strict schema refused, so the next hydration restored
the old project.

Commit only identity fields, ask for a view when the remembered one is
not a table, keep the candidate project's views out of the active list
until it commits, drop late view replies from an earlier pick, and key
Browse rows by project identity.

Co-authored-by: Eng. Mahmoud Bek ElNemer <tt21913@gmail.com>

* fix(mobile): keep github.com implicit in project refs and error on projects with no table view

---------

Co-authored-by: Eng. Mahmoud Bek ElNemer <tt21913@gmail.com>
2026-11-09 22:53:54 +00:00
Kelvin Amoaba f9a10a75f1 feat(native-chat): bring back the New divider behind a setting (#40490)
Restores the line above messages that arrived while you were away, the scroll to it on return and the screen-reader announcement, all removed in #38044. A new Chat > Behavior switch, "Mark new messages when returning to a chat" (on by default), turns it off; off tracks nothing.
2026-11-09 22:16:18 +00:00
Jinwoo Hong bd56be7146 feat(orchestration): mail an agent on a paired server with send --on; its reply comes back by itself (#40106)
* feat(orchestration): mail agents on a paired server with send --on, replies come back by themselves

* fix(orchestration): gate --on on the carrying Orca, ack retried deliveries, refuse payloads and federated workers

* chore(orchestration): annotate peer-mail row casts and drop a wall-clock wait in its test

* chore(orchestration): name mail-between-servers codes server_mail_unsupported and mail_requires_server_pairing

* docs(cli): list --on in the orchestration reply usage

* fix(orchestration): name --environment, the flag on main, in agent mail refusals and the guide

* chore(orchestration): mark the two deliberate agent-kind branches mail between servers needs

* docs(cli): leave --on out of the reply usage; it is accepted only to refuse it

* fix(orchestration): keep peer conversations to peer ids, survive resets and re-pairs, keep pulling on a registry error

- A server takes only peerconv_ conversation ids, so a paired computer cannot name another Orca's dispatch in the relay.
- A replayed reply after a mail reset is acknowledged, not retried forever; reset tasks keeps replies waiting to be pulled.
- A re-paired server starts a new conversation instead of refusing every later send and reply.
- A failed registry read keeps the pull loop retrying.
- Drop the two ast-grep-ignore lines that only #40193's rule needed.

* fix(orchestration): worker-start --on runs on the caller's own Orca, like send --on

An inherited ORCA_ENVIRONMENT sent worker-start --on to that server, so the
--on command the agent refusal suggests did not work. One rule now keeps
send, reply and worker-start with --on on the caller's own Orca.

* fix(orchestration): refuse reply --on on both sides of a conversation between servers

The side that opened the conversation still took --on when it named the right
server; the reply usage says --on is accepted only to refuse it.
2026-11-09 17:00:11 -05:00
Brennan BensonandClaude Opus 5.5 b7ac7e5178 perf(native-chat): keep streaming chats from re-projecting their whole transcript (#40251)
* perf(native-chat): keep streaming chats from re-projecting their whole transcript

Each stage of the streaming transcript pipeline (reducer merge, question
pass and receipts, message projection, desktop list fold) keeps its
previous input and output and re-derives only revised or appended rows,
falling back to the existing full projection whenever another row could
change. Desktop folds through the phone's fold units, now shared, and live
trims batch by a bounded slack so appends keep row positions.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* perf(native-chat): reuse appended subagent rows and batched revisions

Appended rows in chats with subagent rows extend their turn chunk when
earlier turn inputs are unchanged; batch merges and multi-row revisions
do one pass each; rows without journal positions keep in-place reuse.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* perf(native-chat): look up subagent chunks once per appended batch

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* test(native-chat): pin the live trim slack

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-11-09 13:46:36 -08:00
Kelvin Amoaba 854928969e feat(mobile): add Shift, Ctrl and Alt latches and opt-in agent keys to the terminal key row (#38776)
* feat(mobile): smart rail keys for the terminal agent's state, with Shift/Ctrl/Alt latches

* fix(mobile): cancel key presses only when the key row really moves

* fix(mobile): keep Smart rail on through Reset Defaults and finish two-write keys

* fix(mobile): only a finger scrolling the key row cancels its presses

* fix(mobile): hold the smart group until the tap's press lands, and drop clicks on keys that moved

* refactor(mobile): simplify the smart rail touch hook and layout read

* chore(mobile): allowlist the Shift, Ctrl and Alt key names

* feat(mobile): Codex async question keys on the smart rail

* fix(mobile): drop the Codex Always key, whose key differs by approval

Codex binds "don't ask again" to p for a command-prefix rule and a for session and file approvals (0.136 through 0.166). The permission hook doesn't say which one is on screen.
2026-11-09 21:19:11 +00:00
Kelvin Amoaba 1899653040 feat(settings): manage dictation vocabulary in Voice settings, per user and per repo (#40471)
* feat(settings): add a Vocabulary group to Voice settings for personal and repo terms

* feat(speech): use the repo name as a dictation hint without recasing one-word names

* fix(settings): use Input and Button styling as the design system defines it

* fix(settings): chain quick repo-term edits, report a host that drops them, keep the hint off phone replies

* fix(settings): key a pending repo-term write by host too

* refactor(settings): catch a dropped repo-term write in updateRepo and keep the name hint out of stored terms

* fix(settings): one vocabulary chip per term with its aliases, and no repo suffix on the scope tab

* fix(settings): skip a bare vocabulary term that already has an entry
2026-11-09 21:18:14 +00:00
Kelvin Amoaba 81e5e5a9ef feat(settings): redesign Voice settings around a setup card and a rated model list (#40452) 2026-11-09 20:46:25 +00:00
Kelvin Amoaba acf3aead78 fix(agent-status): show Codex's mid-turn async question as waiting for the user (#39087)
* fix(agent-status): show Codex's mid-turn async question as waiting for the user

* fix(agent-status): keep Codex's async question waiting on paired hosts under its working title

The phone's PTY-title renewal turned the waiting row back into working, because Codex keeps its working title while the question waits.

* fix(agent-status): clear Codex's open async questions when the user sends a plain message

Orca's chat card answers an async question as a quoted message, not Codex's reply wrapper, so the pane stayed waiting until the turn ended.

* refactor(agent-status): move the legacy status test seed out of listener-state
2026-11-09 20:42:22 +00:00
Brennan Benson c804008e4f Show friendly advice for Codex turn failures (#39959)
* Show friendly Codex turn failure advice

* Match terminal failure callback to status row type

* Keep terminal diagnostics in Details and clarify unknown failure advice

* Preserve usage-limit diagnostic metadata

* Validate custom host advice in phone renderer fixture

* Check categorized compaction failures and retained diagnostics

* Complete localized model-unavailable notices

* test(mobile): keep the failure notice helpers main's test cleanup removed
2026-11-09 12:34:12 -08:00
Brennan Benson 3ee128db1a Keep Claude background tasks attached to their subagent after restart (#39958)
* Keep Claude background tasks linked to their subagent across restarts

* Keep task ownership scoped to each Claude invocation

* Keep partial task events scoped to their invocation

* Distinguish pending task facts from admitted invocations

* Wait for saved task ownership before interpreting Claude events

* Verify journal binding order and saved task liveness

* Keep owned task updates from reopening Claude root turns

* Exercise owned historical notifications with provider identity
2026-11-09 12:33:19 -08:00
Brennan Benson b602376e8b Show AskUserQuestion option previews on desktop and phone (#40109)
* Show AskUserQuestion option previews on desktop and phone

Claude can attach a preview (a mockup or snippet) to each option of a
single-select question. Orca dropped it, so users chose between bare labels.
Options now carry the preview through the journal, terminal asks and the
phone, and the question card shows the selected option's preview as plain
monospace text. Questions with previews confirm explicitly instead of
auto-advancing, so the preview stays readable.

* Keep phone previews bounded and digit focus on the named option

Phone previews get a capped height with their own vertical scroll, and
opening one never scrolls past its top. A digit always focuses its row
before the chosen-row check, so Enter confirms the option the user named.
Tighter preview line height, a stable phone row root, and a persistence
test whose expected previews come from the fixed input.

* Narrow the reopened question with a type guard instead of a cast
2026-11-09 12:31:52 -08:00
Brennan BensonandClaude Opus 5.5 376cf18634 perf(native-chat): addressed tab-map transitions for chat membership (#40098)
* perf(native-chat): project structured chat tabs per workspace bucket

The shared chat-tab list was keyed on the whole unifiedTabsByWorktree
identity, so every tab write anywhere (terminal title batches included)
rescanned all tabs and handed every consumer a new list. Rebuild only the
buckets whose identity changed and return the previous list while no chat
tab changed.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* perf(native-chat): reconcile chat status membership per changed tab

The status bridge rebuilt every member and host group and captured every
chat on any tab-map or owner-input change, and recaptured every chat on any
launch notification. Membership now updates only the tabs whose identity
changed; owner-input changes re-resolve only unstamped tabs, the only ones
whose host depends on them; launch notifications name their session so only
its tabs are recaptured (unaddressed ones still recapture all). The sleep
and launch-directory companions of one projection land in one store write.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* chore(ratchets): guard whole tab-map triggers and hot-map listeners

Add a count ratchet for identity comparisons of the whole
unifiedTabsByWorktree/tabsByWorktree map in renderer code (counts may only
fall), wired into pnpm lint, run-ratchets discovery and changed-check
selection. Extend the advisory no-whole-hot-map-selector rule to inline
store listeners that compare a whole hot map and to effect bodies that use
one whole from getState().

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* revert(lint): drop the hot-map listener/effect extension

The guard stays scoped to the repaired seams; no repository-wide rule change.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* fix(native-chat): capture chat status on semantic tab changes; address tombstone retirement

Focus, preview and order-only tab writes no longer recapture a chat's
status: membership and the projection fences compare the fields the
projection reads. Absent-tombstone retirement now names the sessions it
retired, so no launch producer sends an unaddressed notification.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* feat(store): keyed-record transition seam for tab maps and companion records

Writers build next tab buckets and keyed companion records through
replaceTabBuckets / replaceRecordEntries (or name the keys of a record they
already built), which record exactly which keys and tab IDs changed against
the produced record. Readers ask readTabMapChanges / readRecordChanges for
an addressed delta; any unannotated link reads as a full resync. Adds
subscribeTabTransitions and a store-observer installer registry. The
chat-tab list and the status bridge now consume addressed deltas and keep
the whole-list walk only as the cold path.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* feat(store): annotate every shared tab-map writer through the seam

Shared tabsByWorktree/unifiedTabsByWorktree writers (tab, editor, browser,
worktree, ssh, terminal create/close/navigation/title-batch/reconnect/
hydration, web-session mirroring and layout frames) now build or annotate
their next tab buckets and companion records with the exact workspace and
tab IDs they changed; full hydration, full frame builds and reconnect
inventories are named full resyncs. Writers that copied every bucket for
one tab (preview promotion, dirty-file preview clear, worktree removal
filters, sleeping-record identity rename) now keep unchanged buckets.

Replace the renderer-wide tab-map comparison ratchet with a narrow seam
guard: the repaired consumers may walk a whole tab map only in named cold
paths, and the pinned writer modules may not build raw seam records.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* chore(store): name seam fields once; type seam records; record moved respawn writes

Writers stop repeating tracked field names (centralized in
seam-record-fields.ts), the seam takes a named KeyedSnapshot record type
instead of casts, and the layout-writer baseline records the respawn
helpers moved unchanged out of worktree-layout-lifecycle.ts.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* chore(store): drop seam field cast; record the seam field list in the layout-writer baseline

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* fix(native-chat): keep chat-tab list in source workspace order (C3/F1)

Every source workspace keeps a projection slot, even without chats, and the
seam reports when a write added or deleted a workspace key or moved a
surviving bucket, so the incremental list never reorders first-match
consumers; those transitions take the cold rebuild.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* fix(native-chat): keep projection and retirement debt awake (C2, F3)

A member is owed from attach or semantic change, members without a host
feed keep their own debt, and every reconciliation drains existing owed or
retired work even when no new member changed, including a closed last chat
whose feed is gone.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* fix(store): stop inferring host-mirror companion IDs from visible tabs (C1)

A host retraction can delete PTY or layout records for tabs neither
inventory lists, so the mirror funnel marks its tab-keyed records full
resync instead of advertising an incomplete addressed set.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* chore(ratchets): close indirect paths in the tab-map seam guard (C4)

The guard now rejects ordinary callers of the bridge cold path, value
imports the repaired consumers have not reviewed, and raw seam records built
into a local binding without a seam helper annotating it.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* test(native-chat): drive real tab actions through the status bridge (F2)

Close, session rebind with a late old-session tick, and a legacy owner
change, through createUnifiedTab/setTabLabel/closeUnifiedTab, so the
addressed reconciliation path is exercised.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* chore: exempt seam field-name constants from the layout-writer ratchet; drop dead setters (F4)

The seam's SEAM_*_FIELDS lists name fields without writing them, so the
ratchet skips them and the raised baseline row is gone (renderer total 569,
main 571). setStructuredChatSleptAt and the two launch-directory setters
lost their last callers to the batched companion write.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* feat(store): Map provenance and source-ordered change keys in the seam (PR A addenda)

replaceMapEntries gives Map-typed fields (runtimeStatusByEnvironmentId) the
same weak provenance and readRecordChanges protocol. Change keys compose
forward through a chain, oldest write first, so added keys come out in
source order (index keys first for records, as Object.keys orders them). A
surviving key deleted and re-added, in one write or across a chain, and a
record builder that moves a surviving key, read as a 'reordered' full
resync.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* fix(store): writers declare key order; seam reports ordered additions (N1c, N1a)

The seam no longer scans both records to rediscover order on every
annotation. replace* helpers know insertion order by construction; annotate*
takes a declared order ('appended' by default, 'reordered' for a builder
that moves keys, read as a full resync), verified only under test.
Addressed changes now report keys, source-ordered added keys and deleted
keys, so a reader holding its own key ranks places changes in O(changed).
Worktree removal declares its editor-transfer records reordered.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* perf(native-chat): scope the chat-tab projection to chat-bearing workspaces (N2c)

The projection keeps chat subsets only for workspaces that have chats, plus
per-workspace source ranks that change only when a workspace is added or
deleted. A focus, label or title write costs O(that workspace); the
ordered list is materialized lazily, only for full-list readers, and keeps
its identity while no chat changed. Workspace adds and deletes are
addressed through the ranks instead of a cold rebuild.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* chore(ratchets): resolve aliases and require annotation before publication (C4)

Any reference to a cold helper or Object enumerator outside its named
callers counts, including aliases and callbacks; writer values are followed
through local aliases and must be annotated before the write that
publishes them.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* test(native-chat): scoped hydration keeps the warm chat list equal to a cold read (D1)

Drives the real scoped hydrateTabsSession on early/later chats of one
session: source order, list order and first-match precedence match a cold
rebuild, including when the selection order differs from hydrated order.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* perf(native-chat): cache the chat-workspace order across chat edits (D1 item 5)

The lineage keeps chat-bearing workspaces in rank order and only inserts or
removes one when it gains its first or loses its last chat; a chat object
change re-flattens the list without re-sorting. Full resyncs rebuild it.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* chore(ratchets): the seam guard recognizes only construction helpers and cold marks (D1)

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* refactor(store): addressed provenance only from constructing helpers (D1)

Remove annotateRecordReplacement/annotateTabBucketReplacement and their
caller-declared order: only replace*/recordDraft/tabBucketDraft, which
perform the construction, derive changed keys, additions and deletions from
their operations (delete-then-reinsert reads as reordered); opaque
reconstructions use markFullResync. Every former annotate caller is
migrated: ordinary local writes build their records once through helpers
or drafts with today's operations and order; scoped tab and terminal
hydration, the reconnect inventory, host-mirror application and layout
flush are full resyncs. Change readers move to keyed-record-change-reads.ts.
Renderer layout-writer total 567 (main 571).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* test(native-chat): type the hydration fixture without a cast

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* fix(store): seam record writes define own entries, even __proto__ (N5c)

replaceRecordEntries and drafts assigned through the inherited __proto__
setter, so a tab id '__proto__' lost its own PTY, layout and startup
entries. Both now define an own data property, as the computed-key literals
they replaced did; contract test included.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* perf(terminal): build each createTab record once (N4c)

createTab took ensureGroup's and insertSplitGroup's whole records and the
orphan cleanup's seam records, then rebuilt them through the seam. It now
takes the per-worktree group and split results and gives the orphan
cleanup only the bucket it reads, so every seam record is constructed once
with its ordered operations.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* chore(ratchets): record createTab's orphan-cleanup input view (renderer total 570, main 571)

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* refactor(terminal): split orphan cleanup so createTab cleans only companion caches (N6c)

buildOrphanTerminalCompanionCleanupPatch holds the per-tab cache, tab-bar
order and selection cleanup; buildOrphanTerminalCleanupPatch adds the seam
record deletes on top, unchanged for reconciliation. createTab calls only
the companion part with real state and deletes the seam entries through its
single construction, so the stub input view and its layout-writer baseline
rise are gone (terminal-tab-creation.ts back to 10).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* perf(terminal): filter the orphan cleanup's kept bucket once (N7c)

The full cleanup filters the workspace's tabs once and hands the kept
bucket to the companion part, which takes its replacement active tab from
it; createTab's companion-only call still makes its single find pass.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* fix(tabs): keep removed worktrees' group entries deleted on removal

The second emptied-group collapse in removeWorktreeLayout also visited
removed worktrees and wrote their deleted groupsByWorktree entry back as
an empty list, a delete-then-set the seam read as a reordered full resync.
It now visits only surviving workspaces, so removal stays addressed.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* chore(tabs): closeEmptyGroup collapses through the shared builder again

collapseGroupLayout takes the record builder, and the renderer passes the
seam's replaceRecordEntries, so the per-worktree collapse publishes with
provenance without new counted layout writes (tabs-group-actions back to 12).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* chore(tabs): reconciliation names its draft fields once (2 + 4 <= 9)

The written-field sets come from SEAM_ field-name constants instead of
string literals, so the reconciliation split no longer raises the combined
layout-writer count (tabs-reconciliation 2, tabs-reconciliation-batch 4,
against the original 9).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* chore(tabs): publish reconciliation drafts with explicit field writes (N9c)

Revert the SEAM_ field-name constants: the drafts are held under their own
names and finishReconciliationDrafts publishes each written draft as an
explicit unifiedTabsByWorktree/groupsByWorktree/activeGroupIdByWorktree/
layoutByWorktree property, so the counter sees every real write and the
split totals 2 + 4 against the original 9, with no exemption.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-11-09 12:19:13 -08:00
Kelvin Amoaba 98bc1acf16 feat(dictation): pulse three dots in the pill while spoken words are still on their way (#40443)
* feat(dictation): show pending dots while the engine holds speech it hasn't committed

The VAD pause segmenter already knows when a phrase is open. The recognizer
now posts speech-pending when a phrase opens and clears it once that phrase's
text is posted, and the desktop pill shows three dots after the text while it
is on. Only pause-cut on-device streams send it, so cloud, the paired web
client and phones keep today's pill.

* fix(speech): keep a lone click from turning the pending dots on
2026-11-09 20:14:48 +00:00
Kelvin Amoaba 1d92f6fcdb fix(speech): let a slow on-device finish run while the engine still answers (#40460)
* fix(speech): let a slow on-device finish run while it still returns text

The stop timeout counted 60 s from Finish however much audio the engine still had
to decode, so a CPU model on a busy machine failed a dictation it was still
transcribing. It now fails only after 60 s without text for that dictation.

* fix(speech): count any engine answer as progress while a dictation finishes
2026-11-09 20:09:25 +00:00
Kelvin Amoaba 035de0a70c feat(speech): record each model's languages, licence and measured error rates in the catalog (#40431)
Error rates come from Orca's speech bench (transcribe-cpp 0.4.1, Metal, pinned GGUFs):
LibriSpeech for English, 5 FLEURS dev clips for de, es, fr, ja, ko, pt, ru, tr and zh.
speechModelQuality turns them into the Voice settings word scale.
2026-11-09 20:08:43 +00:00
Brennan Benson 90a0748576 Source Control's Codex model menu lists your account's models even when Codex's catalog is large (#40117)
* Source Control lists Codex models from app-server model/list instead of debug models

The Codex text-action picker read 'codex debug models', a dump of Codex's whole
catalog with prompt templates (40-90 KB per model). Once it passed the 4 MiB
output cap discovery failed and the picker kept Orca's built-in list, and the
dump's hidden entries were selectable. List models with app-server model/list
(hidden rows excluded), over Source Control's own launch, lock and teardown,
with the same 4 MiB input budget and 60 s deadline.

* Bound Codex model discovery input in bytes, key its cache case-insensitively on Windows, and stop the app-server only when it ignores stdin EOF

- Count raw bytes from both streams against the 4 MiB budget.
- Read CODEX_HOME (and other account dirs) case-insensitively on Windows for the host cache key.
- Skip the post-listing kill when the app-server already exited; fail waiting listing requests once discovery settles.
- Report an incomplete model list as such.
- Tests: fake-timer grace and successor admission after a successful listing, method-not-found and unreadable-page fallbacks.

* Key the Codex discovery cache on the env spelling Windows spawn uses

Windows env names ignore case and Node's spawn takes the lexicographically first spelling, so
CODEX_HOME wins over codex_home; the cache key now picks the same one.

* Narrow Codex app-server records with the provider record guard instead of type assertions

* Read Source Control action model picks under the repo's host key, as the picker saves them

Local repos resolved generation under the 'unknown' host key because their connection id is undefined, so the saved pick was skipped and the agent default ran.

* Key Source Control model picks by connection until the repo record loads, and test with a real local repo

* Key a native Codex listing's cache on process.env when no agent env was resolved
2026-11-09 12:08:24 -08:00
bfa941aa06 fix(native-chat): honor Windows environment overrides consistently (#40301)
* fix(native-chat): apply Windows environment overrides consistently

Use the same case-aware composition for native chat launches, preflight,
model discovery and naming. Canonicalize configured Windows names before
account readers and preserve account pins and POSIX case distinctions.

Tracked in #39075.

Co-authored-by: 김태겸 <oharapass@gmail.com>
Co-authored-by: Vitalii Kulish <mancunianez@gmail.com>

* test(native-chat): cover effective Windows catalog changes

* test(native-chat): retain child environment test import after sync

* test(native-chat): verify Windows child environment precedence

---------

Co-authored-by: 김태겸 <oharapass@gmail.com>
Co-authored-by: Vitalii Kulish <mancunianez@gmail.com>
2026-11-09 12:04:26 -08:00
Brennan BensonandSeaBee 55fb5584c0 OpenCode subagents show in Chat UI and keep a chat open while they run (#40141)
* fix(native-chat): show OpenCode subagents and keep a chat open while they run

OpenCode reports each subagent as a child session when the client opts in.
Orca now opts in, files each child's tool calls and replies under its own
subagent, and keeps the subagent open until the child reports that it
completed, failed or was interrupted, including after the parent turn ends.

Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>

* fix(native-chat): keep a subagent's tool context when its rows are read back from the journal

Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>

* fix(native-chat): a Stop also ends running OpenCode subagents, and each subagent can be stopped from the strip

Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>

* fix(native-chat): only a chat Stop closes OpenCode subagents, an early subagent Stop stays available, and the close request has a deadline

Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>

* fix(native-chat): a subagent Stop that the subagent confirms while the request is out reads stopped

Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>

* refactor(acp): close child sessions from the agent connection

Keeps the session runtime under the line limit after the main merge.

* test(acp): read a child command through a type guard

* fix(native-chat): a subagent that reports its ending reads stopped even when the close request fails

A child Stop's session/close now asks the agent to drop the call on timeout instead of withdrawing the write.

* test(native-chat): OpenCode subagent calls group with other agents' subagent calls

---------

Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>
2026-11-09 12:03:55 -08:00
Brennan BensonandEng. Mahmoud Bek ElNemer 0f89de5276 Save a pasted workspace directory as the path it names (#40347)
* Save a pasted workspace directory as the path it names

A workspace directory with a leading space (" /Users/me/dev") was saved as
typed. Readers treat it as relative, so Orca resolved it inside every
registered repository and created an empty " /Users/me/dev" tree in each.
Normalize the saved value once where settings are written and loaded, as
the leading ~ fix did: leading whitespace before an absolute or home path is
dropped, while relative paths and trailing characters stay exact. A value
repaired on load keeps its old spelling in history so checkouts under the
old root stay visible. The settings field shows the saved value.

Co-authored-by: Eng. Mahmoud Bek ElNemer <ElNeMeR70@users.noreply.github.com>

* Expand ~\ only where it is a home path; drop type casts

* Classify the padded path with the host's own path rules

On macOS and Linux a leading backslash is part of a relative name, so only
Windows treats \work as absolute.

---------

Co-authored-by: Eng. Mahmoud Bek ElNemer <ElNeMeR70@users.noreply.github.com>
2026-11-09 12:03:33 -08:00
OrcaWinandOrcaWin cff5a151fc test: remove low-value CLI, browser and E2E checks (#40466)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-09 11:58:10 -08:00
OrcaWinandOrcaWin 0680d9753e test: remove low-value browser, CLI and budget checks (#40459)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-09 11:44:48 -08:00
Jinjing 8ef4680142 Improve project group ordering across hosts (#39816)
* Improve project group ordering across hosts

* fix(project-groups): restore each group order after a failed drop

A deleted group made the aggregate rollback stop early, leaving other groups at
the ranks from the failed drop. Restore each group independently and log the
swallowed single-update failures again.

* refactor(main): give project group RPC methods their own module

Keeps repo.ts under the 300-line cap as the project group surface grows.
2026-11-09 11:40:51 -08:00
Kelvin Amoaba 1bf2276bdd fix(dictation): match the pill's Chinese and Japanese phrases to the inserted text, and space Korean ones (#40455)
* fix(dictation): join Chinese and Japanese phrases without a space, and space Korean ones

The shared transcript fold joined every phrase with a space, so the pill and
phone replies showed Chinese and Japanese text with spaces the inserted text
doesn't have. It now joins with the same rule as insertion. That rule treated
Hangul like Chinese and Japanese, so Korean phrases were inserted with no space
between sentences; Korean spaces words, so Hangul now spaces like Latin text.

* fix(dictation): don't space after Chinese and Japanese full-width punctuation

* fix(dictation): join retried phrases with the same spacing rule

* fix(dictation): escape the spacing rule's non-ASCII marks for the orcad bundle

The shared fold now pulls the spacing rule into the orcad server bundle, which
must stay ASCII. A rich-draft test pinned Hangul joining without a space; it
now expects the Korean spacing.
2026-11-09 19:39:34 +00:00
Kelvin Amoaba 3aee704b8b fix(speech): keep words spoken while the model loads (#40454)
Stopping a CPU dictation closed the renderer's audio port while frames still
waited behind a decode in the speech worker; the close ended the stream at once
and the queued frames were dropped. A close after 'end' now leaves them to
decode. A model slower than 30 s to load also lost the start of the dictation,
since the renderer kept only 30 s of startup audio; it now keeps as much as the
host waits for the load.
2026-11-09 19:39:23 +00:00
Kelvin Amoaba 58733086c9 feat(speech): spell dictated code terms with a personal and per-repo dictionary (#40450)
* feat(speech): spell dictionary terms in dictation, with repo terms and Whisper prompt

* test(speech): pin host repo terms for phone dictation

* fix(speech): keep symbol terms whole, accept null setup params, resolve SSH repos

* refactor(speech): key transcript words once and type phone terms at the reply schema

* refactor(speech): merge speech-types imports in the speech IPC
2026-11-09 19:22:46 +00:00
Kelvin Amoaba f126d6ef37 feat(runtime): tell phones when the desktop refreshes a PR (#40438)
* feat(runtime): tell phones when the desktop refreshes a PR

* fix(runtime): name every checkout a coalesced PR read stood for

* fix(runtime): don't count a read that couldn't load checks as a checks change
2026-11-09 19:16:14 +00:00
Kelvin Amoaba d291670a01 refactor(speech): remove sherpa-onnx (#39473)
transcribe-cpp is now the only on-device engine on every platform, so
the sherpa-onnx packages, their packaging (extraResources, the darwin
libonnxruntime prune, the native-variant assert and the Linux libstdc++
exemption), the sherpa recognizer in the speech process and the
per-model engine choice all go. Hotwords are written as plain terms.
2026-11-09 18:59:10 +00:00
Kelvin Amoaba fd46321d85 feat(speech): move the speech models to transcribe-cpp and offer an update for old downloads (#39468)
* feat(speech): move the speech catalog to transcribe-cpp models

Parakeet v2/v3, Whisper tiny and SenseVoice now download their GGUF
builds, which run on transcribe-cpp everywhere, Intel Macs included (on
the CPU). A model still holding its old sherpa-onnx download shows
"Update to the faster version (<size>)"; updating replaces the old
files, and dictating before then says the model needs that update.

The five models with no GGUF build are retired and their users moved
over: the Japanese Parakeet, the Chinese-English Zipformer and the
Paraformer to SenseVoice, and the Korean and Chinese streaming
Zipformers to the new Nemotron 3.5 (28 languages). Nemotron's locale
codes (ko-KR) are matched from the ISO language setting.

* feat(speech): offer the speech model update once after the engine change

A card in the notification corner tells people whose selected on-device
model needs the new download, or whose retired model was replaced, and
starts the same download as Settings > Voice. After a replacement it
offers to delete the retired model's files.

* feat(speech): move retired Korean and Chinese models to SenseVoice and offer Whisper Large v3 Turbo

On transcribe-cpp 0.4.1 with Metal, SenseVoice beat Nemotron 3.5 on Korean
and Chinese FLEURS clips, so the retired Korean and Chinese streaming
Zipformers now move to SenseVoice. Whisper Large v3 Turbo (Q8_0, 886 MB)
joins the catalog as the most accurate choice for Japanese, Korean and
Chinese. Nemotron 3.5 stays, described by the languages it does best.
2026-11-09 18:00:13 +00:00
OrcaWinandOrcaWin fc379afda7 test: remove low-value locale, automation and settings cases (#40437)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-09 09:49:58 -08:00
Kelvin Amoaba 16e21424a7 feat(dictation): show committed and in-flight words in the dictation pill (#40428)
The pill showed only the current partial and went back to Listening after
each final, so pause-committed phrases from transcribe-cpp never appeared.
The host's transcript fold moves to src/shared and the desktop pill uses it
too: committed text in the foreground, tentative text muted.
2026-11-09 17:16:31 +00:00
Kelvin Amoaba 9a45e8bdfb feat(speech): dictate through any OpenAI-compatible transcription service (#40160)
Add an OpenAI-compatible model whose base URL, model and optional key
are set up in a Voice settings dialog and sealed together in ~/.orca,
so a key is never sent to a host it wasn't saved for. It reaches
OpenAI, Groq, Mistral and local servers, and replaces the unused
userModels setting.
2026-11-09 16:57:24 +00:00
Kelvin Amoaba 0b8e94aa5c feat(speech): run GGUF models on transcribe-cpp with a CPU fallback (#39326)
* feat(speech): run GGUF models on transcribe-cpp, falling back to the CPU

The speech host child can now load a second engine. A model whose files
are GGUF runs on transcribe-cpp (Metal on Apple silicon, Vulkan on
Windows and Linux, CPU otherwise); everything else stays on sherpa-onnx,
which remains the only engine on Intel Macs.

A GPU that fails to load, errors, stalls for 25 s or crashes the process
moves the dictation to a CPU-only process (TRANSCRIBE_BACKENDS=cpu) that
replays the audio the GPU had not decoded yet, so no words are lost or
repeated. A GPU that crashed or would not load is remembered per
transcribe-cpp version and OS release, and voice.speechGpu = 'off' keeps
it on the CPU.

Packaging copies transcribe-cpp, koffi and the target's native package
into resources/node_modules outside asar, fails when a target's package
is missing, and ships the MSVC runtime beside transcribe.dll on Windows.

No catalog model is GGUF yet, so nothing runs on transcribe-cpp until
the catalog change that follows.

* fix(speech): keep the speech queue running after a failed task, and drop audio frames after a discard

* fix(speech): keep the engine's file reads off the main thread's hot paths

The engine lookup and the GPU failure marker are read once per process
instead of on every dictation start and mobile catalog request, the marker
is written asynchronously, and the child reads the hotwords file with
fs/promises. The baseline moves the existing engine presence checks from
stt-worker-paths to the shared package table and adds the once-per-process
marker reads.
2026-11-09 16:37:03 +00:00
OrcaWinandOrcaWin 23588ed935 test: remove low-value browser, store and adapter suites (#40399)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-09 07:54:05 -08:00
OrcaWinandOrcaWin 3ce8d75ae9 test: remove low-value agent-hook and row-projection coverage (#40395)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-09 07:35:12 -08:00
Neil 9a21a515d3 fix(vm-recipes): hide an in-progress VM checkout in every window, the phone and the jump palette (#40373)
* fix(vm-recipes): hide an in-progress VM checkout in every window, the phone and the jump palette

A VM create's checkout was hidden only in the sidebar of the window that
started the create. Windows now publish their pending VM checkouts to the
main process, which shares every window's set with all windows and
filters the paired phone's host workspace list and workspace context.
The jump palette and the sidebar read the shared set, and a grouped VM
project no longer draws an empty-project placeholder while its create
is pending. A window's set is dropped when it closes, crashes or reloads.

* test(project-groups): reset leaked runtime environment revisions between files
2026-11-09 06:51:21 -08:00