Rich mode blocked any file with a line matching /^\[[^\]]+\]:\s+\S+/m,
so prose like "[Bug]: Backspace at the start ..." forced code-only
mode with the reference-style links banner.
Keep that regex as a pre-filter, then confirm a real definition before
blocking: either a line the rich-mode encoder would strip, or a
definition the production Tiptap/marked parser collects from the
encoded source. Lexer failures fail closed. Bodies over 50k code units
keep the existing block so the per-edit parser cost stays bounded, and
footnotes are checked first so footnote files skip the lex.
Fixes#19795
Co-authored-by: Neil <neil@stably.ai>
On Linux the desktop app installs its CLI as `orca-ide` (GNOME's screen
reader owns `orca`), but help text, typo suggestions, missing-subcommand
hints, and runtime-unavailable hints all printed `orca`, so copied commands
launched the screen reader. Resolve the invoked command name (Orca terminal
shim via ORCA_CLI_COMMAND, `orca-dev` for dev launchers, `orca-ide` for the
Electron-run-as-node app launcher, `orca` elsewhere) and render it in those
outputs. macOS/Windows and SSH Node CLIs keep `orca`.
Fixes#42682
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Kimi Code v2 writes state.json createdAt/updatedAt as epoch milliseconds,
but the parser read them with extractString, leaving session timestamps
null so session search --since dropped the session. Pass the raw values
to updateTimeline, which already normalizes ISO strings and numbers.
Refs #27918
Co-authored-by: szsip239 <88223778+szsip239@users.noreply.github.com>
Add an unbound-by-default `editor.toggleChangesView` keybinding that
switches the active editor tab between Edit and Changes, mirroring the
toolbar toggle. The shortcut shares the toolbar's Changes gate (now
`canToggleChangesModeForFile`), skips binary previews, claims the chord
so Monaco does not also run it, and flushes custom editor frames before
switching modes.
Fixes#11095
Co-authored-by: Neil <neil@stably.ai>
* fix(native-chat): open chats saved by journals that predate later tables
A journal last opened by v1.4.219 through v1.4.245 has no outcome tables. The
new store reads that file read-only and never adds them, so converting any of
its chats failed with "no such table: journal_outcome_epochs" and no chat opened.
The original is now opened as the old writable open would have shaped it:
every table this build's legacy open would create but the file lacks reads as
empty through a private TEMP table. The file is never written. The profile
copy's per-table version guard and the empty-chat table guard are no longer
needed and are removed.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(native-chat): keep refusing a schema-4 journal missing its record tables
Only tables the file's version could legitimately lack are presented as empty.
The PR-watch counter is read from the file itself, not a TEMP sequence table.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore: record the legacy source opener in the worker-only SQLite baseline
It runs only on the chat store's writer and reader worker threads, like the
other journal files listed there; the two files it replaced shrink.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(native-chat): import profile rows whose chat record the old store already removed
A receipt left behind by a removed chat failed the first-launch copy with
FOREIGN KEY constraint failed on every launch, so no chat could open. The copy
now keeps the original rows as saved, as a file copy would.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test: drop an assertion the store open already guarantees
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Copy, fork and continue-session context fed SerializeAddon's replay
stream into a cleaner that drops every CSI sequence, so gaps that the
serializer encodes as ESC[nC (cursor moves, tab stops, styled blanks)
vanished and adjacent words were joined. Read the visible xterm cells
instead with translateToString, joining soft-wrapped rows before trimming
so a printed space at a wrap boundary survives. Interior gaps and leading
indentation are kept exactly; trailing whitespace is trimmed per logical
line and blank rows below the output are dropped. Normal-buffer history
stays ahead of an active alternate screen.
Fixes#38888
Co-authored-by: Neil <neil@stably.ai>
undici's HTTP/1 writer calls socket.setTypeOfService before each write,
outside any fetch promise. On macOS the syscall fails with EINVAL once the
peer has reset the socket (VPN, Wi-Fi switch, proxy), so a network change
became an uncaught main-process exception and a modal error dialog. Treat
the TOS syscall as best effort, as Node already does on Windows; the write
then fails through undici's normal ECONNRESET path. Argument validation
errors still throw.
The test resets a real peer connection and calls setTypeOfService on the
reset socket, reproducing the native EINVAL on macOS.
Fixes#39579
Co-authored-by: Jinwoo-H <jinwoo0825@gmail.com>
Workspace cards, refresh sweeps, the status bar and the palette keyed and
fetched linked issues by number only, so a fork workspace linked to
fork#247 showed upstream#247 (the number-only lookup is upstream-first).
Derive the repository from the workspace's saved GitHub issue URL (or a
matching GitHub attachment in linkedItems), pass it through gh:issue /
github.issue as an optional ownerRepo, and include it in the issue cache
key. Number-only links keep the existing heuristic.
Fixes#24378
Co-authored-by: Neil <neil@stably.ai>
* Move environment store operations off the main thread
* Preserve CLI startup speed with the serialized environment owner
* Handle absent paired host labels in the producer regression fixture
* Preserve publication ownership and copy selected environment results
* Correct publication and WSL readiness regression fixtures
* Include environment store worker in deployed server artifacts
* Check fresh listing publication across an intervening store update
* Await store-backed results in new recovery coverage
* Retire failed page publications and retain CLI worker support entries
* Unpack environment store support modules for the installed CLI
* Align async store callers with their return contracts
* Avoid hardening work for absent environment store files
* Keep permission repair bounded to one hardening attempt
* Align store test fixtures and await pairing replacement
* Use the semantic store operation for external pairing fixtures
* Provide the live sender frame in the timed subscription fixture
* Share pending server lookups and settle retired phone requests
* Keep the relay error type import
* Keep status ownership and transport imports acyclic
* Await replacement pairing in status retirement cases
* Keep asynchronous server rows bound to their admitted pairing
* Keep inventory snapshots and queued delivery at one publication boundary
* Await saved-host and graph test setup through publication completion
* Finish async fixture types and remove foreground timing probe
* Await renderer surface transitions in process identity coverage
* Await nested graph publication fixtures through their callers
* Await restored pane graph publication before hook assertions
* Include imported store worker and sidebar files in web typecheck
* Prepare saved host storage in source-based WSL test rigs
* Await saved server lookups in the source retirement diagnostic
* Preserve newer WSL status across async saved-host reads
* Fence stopped probes when a newer WSL Ready arrives
* Keep saved WSL host selection covered when admission is off
* Check WSL probe ownership after async admission
* Preserve WSL stop authority without cancelling Ready persistence
* Prune removed WSL E2E real-time wait from flaky-test baseline
* Limit WSL callback invalidation to confirmed stops
* Retire pending WSL notices on every confirmed stop
* Close the WSL picker dialog explicitly between scenarios
* perf(layout): skip unused normalization reports in pane commands (#42615)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* Clarify cleanup pinning and ignored-file protection (#42616)
* test: remove 55 low-value suites and assertion driver (#42605)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
* fix(skills): skip Claude sync scratch folders during skill discovery (#42610)
Claude Code parks replaced skill copies in `.trash` and in-progress syncs
in `.staging` under its skills roots. Discovery walked into both, so each
synced skill showed up as duplicates. Prune those directory names in the
native file walk and in the generated WSL `find` command.
Co-authored-by: Neil <4138956+nwparker@users.noreply.github.com>
* test: remove 78 low-value suites and trim redundant cases (#42621)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
* Keep quit teardown within the source line budget
---------
Co-authored-by: Jinjing <6427696+AmethystLiang@users.noreply.github.com>
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Co-authored-by: OrcaWin <alpha-eng@stably.ai>
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
Co-authored-by: BingZ <zzb@gxsmjx.com>
Co-authored-by: Neil <4138956+nwparker@users.noreply.github.com>
The appearance methods first shipped in v1.4.251 and editor.readTypography
in v1.4.257, so a plugin requiring the old floors installed on Orca that
lacks them.
The untitled save dialog stripped `.md` and always re-appended it, so a
plugin-created untitled file such as `untitled-2.excalidraw.svg` was
saved as `architecture.md`. The dialog now splits the current name into
stem and (possibly compound) extension, shows that extension next to the
name input, strips it if the user re-types it, and saves extensionless
files as-is. The extension is held through the close animation so the
suffix label does not flicker.
Co-authored-by: Kien Le <122910950+kien-ship-it@users.noreply.github.com>
Co-authored-by: Neil <neil@stably.ai>
* feat(plugins): check host API calls against released Orca at validation
orca plugin validate and pnpm validate:plugin now read the literal host
method names a plugin's worker, panel and editor code calls, and report
methods this Orca does not have, proposed methods, wrong-surface calls,
and methods newer than engines.orca (unless guarded by orca.host.has).
Deprecated methods warn with a stable 'deprecated host method <name>'
prefix. Adds the Host API minor to first-Orca-release table.
* fix(plugins): avoid host-call validation false positives
Guarded proposed and wrong-surface calls warn instead of failing, and an orca.api.minor comparison counts as a floor guard. Only host.call/has receivers count, comments are stripped, files are stat-capped and read through their real path inside the plugin, root-relative script sources resolve, and each file is read once.
* fix(plugins): keep has-guarded host calls as errors before orca.host.has existed
* fix(plugins): fail only certain host-call problems and bound the scan
Floor and wrong-surface findings are warnings, since guards and per-surface reachability are not visible to a text scan; unknown and proposed methods without a literal has guard stay errors. Read attempts, references per file and queued files are capped, has?.() counts as a guard, and HTML script and comment scanning is linear.
* fix(plugins): treat any orca.host.has call as feature detection for unknown methods
* build(cli): ship the host-call validation module with the packaged CLI
* fix(browser): start a profile's extensions with its first page, not at launch
Every local browser profile attached its enabled extensions at startup, so
each profile's service workers ran even with no tab open. A profile now
starts its extensions when it gets its first page or when Settings acts on
it, and that page's first load waits (up to 3 s) until they have loaded.
* fix(browser): keep extension startup out of import cycles
The loopback callback used to answer "Signed in to Orca" as soon as it
received the authorization code, before the app exchanged it for a
session. A failed exchange left the app signed out while the browser
claimed success.
Keep the callback response open and answer it once connect finishes:
connected (200), cancelled/superseded (200, neutral), or failed (500
with the error code). A 45s fallback answers with
orca_cloud_auth_response_timeout if the app never responds.
Co-authored-by: Neil <4138956+nwparker@users.noreply.github.com>
Cancel only reached the lane token installed at spawn, so a Cancel during
the staged-context read or agent-env preparation was ignored and the agent
still ran. Register a pending lane token for the whole request: it settles
the request as canceled immediately and blocks the later spawn; a spawned
agent replaces it with its own kill token as before, and the latest
request on a lane wins.
Closes#10855
Co-authored-by: Neil <neil@stably.ai>
The toast shows the error plus the client's Orca/OS/shell details and asks
users to file an issue, but copying it meant drag-selecting the monospace
text. Add a copy icon beside dismiss that copies the error together with the
same environment footer the toast renders.
Move the generic { text, label } copy icon button out of the editor checks
panel into components/clipboard as CopyTextButton so the checks panel and
the terminal toast share it, and fold showClientEnvironment into the footer
so the copied text and the rendered footer share one guard.
Fixes#30061
Co-authored-by: Claude <noreply@anthropic.com>
Add `orca fill --value-stdin` so passwords and other secrets stay out of
argv and shell history. Fill drops one trailing newline from stdin (secret
readers end with one) and rejects empty stdin so a failed reader cannot
silently clear the field; `--value ''` still clears.
Share one flag-or-stdin reader with `orca computer type`/`set-value` and
the `--body-file -` reader. Computer commands keep their existing
semantics: no newline trimming, set-value still accepts empty stdin, and
`--text-stdin=true` style keeps reading stdin.
Fixes#29849
Co-authored-by: Neil <neil@stably.ai>
* fix(browser): enable offscreen painting for headless pages
* test(browser): verify hidden backend capture pixels in Electron
* test(browser): capture without forced viewport or clip
* test(browser): use current process-host API in capture proof
* test(browser): handle already-loaded capture fixture
* fix(browser): stop delivering unused offscreen paint frames and prove serve captures on Linux
* test: drop the serve screenshot spec that also passes without the fix
* fix(browser): keep extension observers on offscreen serve pages
---------
Co-authored-by: oliver-mee <102673257+oliver-mee@users.noreply.github.com>
* fix(updater): hold orcad's browser helper during every macOS install
* fix(updater): keep supervised serve self-updates out of the blocker-free park
* fix(updater): honor late park refusals and hold on any staged quit install
* test(orcad): run the install-park helper test off macOS
* fix(updater): commit the install-on-quit park with async writes
* fix(updater): make quit teardown wait for the install-on-quit park commit
* refactor(backend): extract live core startup modules
* refactor(backend): import orcad profile startup directly
* refactor(backend): address review — keep smoke load check covering activation, drop thin wrappers and import-graph test
* perf(layout): skip unused normalization reports in pane commands (#42615)
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
* Clarify cleanup pinning and ignored-file protection (#42616)
* test: remove 55 low-value suites and assertion driver (#42605)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
* fix(skills): skip Claude sync scratch folders during skill discovery (#42610)
Claude Code parks replaced skill copies in `.trash` and in-progress syncs
in `.staging` under its skills roots. Discovery walked into both, so each
synced skill showed up as duplicates. Prune those directory names in the
native file walk and in the generated WSL `find` command.
Co-authored-by: Neil <4138956+nwparker@users.noreply.github.com>
* test: remove 78 low-value suites and trim redundant cases (#42621)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
---------
Co-authored-by: Jinjing <6427696+AmethystLiang@users.noreply.github.com>
Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
Co-authored-by: OrcaWin <alpha-eng@stably.ai>
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
Co-authored-by: BingZ <zzb@gxsmjx.com>
Co-authored-by: Neil <4138956+nwparker@users.noreply.github.com>
Adds tab.moveLeft and tab.moveRight with no default bindings, so tabs can
be reordered from the keyboard once a user binds them in Settings >
Shortcuts. A move steps the focused group's active tab past its visible
neighbor (or the floating panel's tab when that panel owns the event),
stops at the group's edges, and never crosses into another group.
The same-group reorder drag-and-drop already ran is now shared as
reorderTabInGroup: main's layout owner when it owns layout, otherwise the
local reorder plus the web-runtime mirror. Drag behavior is unchanged.
Fixes#26109
* feat(plugins): list any workspace's agents with agents.list({ workspaceId })
* feat(plugins): version agents.list workspaceId as host API 1.10
An added param is new surface a plugin must be able to require, so specs
gain paramsSince and api.minor counts it. Regenerates the declarative
typings the first commit missed.
* fix(plugins): report runtime failure codes for agents.list workspaceId
The param is 1.10, after the runtime failure codes, so a call naming a
workspace reports unavailable for a disconnected host; calls without it
keep action_failed.
---------
Co-authored-by: Neil <neil@stably.ai>
An agent's click in a browser page moves window focus into that page's
guest and Orca's focus guard moves it back. A tab press held across that
bounce read it as the user leaving the window and was dropped.
Electron's executeJavaScript waits for the page to stop loading and never
settles if the page is destroyed first. Every navigation sent a grab
teardown even with grab off, so closing a loading tab left the request
without a reply ("reply was never sent").
The Default profile's trash button was disabled whenever the profile had
no import record (`disabled={!profile.source}`). Clearing the jar resets
`source` to null, so the button locked itself after one use, and users
who signed in directly in Orca could never clear the default partition.
The button is now always enabled, labeled ("Clear default cookies"
tooltip and aria-label), and asks for destructive confirmation before
wiping the default partition. Strings are added to every locale.
Co-authored-by: Neil <neil@stably.ai>
The helper asked Chromium for its full accessibility tree only for a
fixed list of bundle ids. Orca's own builds were missing, so they got
only the basic tree the role read turns on, where AXPress on web content
does nothing. Gecko apps such as Zotero and Firefox were missing too and
showed only the window frame.
Add Orca's bundle ids and any app that ships Gecko's Contents/MacOS/XUL.
Stop asking on every snapshot: each request restarted Chromium's 2 s
delay, so a click right after a snapshot hit a page that had not
switched yet. Ask on first contact and wait out the delay, then again
every 30 s, since turning VoiceOver off drops the mode. The list stays a
list: the mode also turns on screen-reader mode, which VS Code and its
forks react to.
Fixes#23233Fixes#24843
Co-authored-by: BeamNawapat <39422342+BeamNawapat@users.noreply.github.com>
* test(computer): run the Linux bridge against an unreachable accessibility bus
* fix(computer): report an unreachable AT-SPI bus instead of aborting the Linux bridge
* test(computer): read the bridge response past the registry banner
* test(computer): keep the healthy-bus check off the desktop's X display
* fix(computer-use): put the macOS pointer back after --restore-window clicks, scrolls and drags
HID gestures left the user's pointer at the target. Save it first, and after a
200 ms settle warp it back unless it no longer sits where the last posted event
left it, which means the user moved it. The HID idle counter can't tell: posting
to the HID tap resets it too.
* fix(computer-use): keep the pointer where the user moved it during a macOS HID gesture
* fix(computer-use): save the macOS pointer right before the first HID event
SQLite names the rowid result column after an INTEGER PRIMARY KEY column,
so the profile copy read pr_check_events rows with an undefined rowid and
failed to bind it. Alias the rowid so every profile table copies.