* Show workspace and branch identity below the chat composer
* Use chat context for composer workspace identity
* Read composer hosts from unified tab records
* Keep composer behavior fixtures focused on input flows
* fix: reuse guarded workspace reader in chat footer
* fix(chat): preserve folder identity and composer focus ring
* Show machine and workspace identity below the chat composer
* Avoid unrelated host focus invalidating composer identity
* Correct composer context and runtime fixture types
* Isolate machine-name lookup in composer flow fixtures
* Simplify composer submission bindings
* Keep the composer session options result together
* Match the composer machine icon to the sidebar
* Keep composer types and session options in their owning modules
Paired clients now publish the chat actually selected on the desktop as the active tab; saved layouts, CLI-created chats, closes and restarts keep it consistent. Carries community PR #34526.
Co-authored-by: Jurassic Knights <jurassicknights@gmail.com>
When Orca's local-account runtime is WSL on Windows, the Antigravity usage
meter still ran the host `agy`, so it showed the wrong account or "not
installed". Route the probe through runWslProcess for the selected distro
(never booting a stopped one), key the unsupported-CLI latch per runtime,
tag readings with the runtime as provenance, and drop another runtime's
last reading instead of keeping it as stale. Settings changes retarget
the fetcher and reread usage.
A WSL target without a distro now names the default through
resolveWslDefaultDistro (registry DefaultDistribution) rather than
listWslDistrosAsync()[0], which skips docker-desktop, falls back to
registry order, and caches the list for the process lifetime.
Ported from #31542 onto current main: reuses main's antigravity fetch
generation (#34165) instead of adding a second one, follows moved WSL and
provider-account-services paths, and keeps only the tests that cover the
new behavior.
Co-authored-by: CAVIN <35207870+zzz163519@users.noreply.github.com>
Co-authored-by: Jinwoo-H <jinwoo0825@gmail.com>
* fix(codex): remove Orca-created pre-trust entries when a worktree is removed
Records which [projects."<path>"] tables Orca's preflight trust created in a
managed Codex config (with the exact bytes written), and on worktree removal
deletes only those, inside the existing trust-config mutation queue. Removal
paths await the purge so a removal is never reported before the entry is gone.
Ported from #24803 (commits 88d5469c, c1ffc5a4, 11b0bd74, f85606b3) onto
current main: conflicts resolved, stray re-export wrappers dropped, and tests
trimmed to AGENTS.md rules (ordering/timing, private-state, Cartesian and
lower-layer duplicate cases removed).
Co-authored-by: Neil <neil@stably.ai>
* fix(antigravity): drop Orca-added trustedWorkspaces entries on worktree removal
Generalizes the Codex pre-trust ownership ledger into a small factory
(agent-trust-ownership-ledger.ts) and uses it for agy: when Orca appends a
worktree path to ~/.gemini/antigravity-cli/settings.json trustedWorkspaces, it
records the exact entry, and worktree removal removes only that recorded entry.
Entries that already existed before Orca's grant are never recorded or removed.
* fix(trust): keep folder-workspace trust on session removal; isolate trust test userData
Only checkout-deleting removal paths drop Orca-created agent trust now.
Folder-workspace sessions and forgotten orphans leave the directory on
disk, so removing one must not strip the grant siblings still rely on.
Also point the execution-host trust test at a temp userData so the new
ownership ledger never writes to the real Orca config, and make the e2e
removeWorktreeMetadataAndHistory stub return its Promise.
---------
Co-authored-by: ggbdpq <ggbdpq@gmail.com>
* fix(antigravity): switch agy accounts through the Linux Secret Service keyring
agy stores its Linux login with go-keyring's Secret Service provider
(service=gemini, username=antigravity, login collection or default alias)
and falls back to its credential file without a session bus, in containers,
SSH and WSL. Add a secret-tool/gdbus adapter that mirrors that collection
choice, refuses locked keyrings with a clear error, and keep the file store
for headless hosts.
Fixes#32818 (STA-11911)
* test(antigravity): avoid casts in Linux keyring adapter test
Share one language-aware unicodeHighlight builder across the file editor,
the standalone diff viewer, and combined diff sections. Markdown skips
ambiguous-character highlighting so full-width punctuation no longer
triggers the warning banner; invisible-character detection stays on
everywhere. includeStrings is pinned to true because Monaco otherwise
resolves it to false in trusted workspaces.
Fixes#33860
Co-authored-by: QTom <qt@truesightai.com>
Co-authored-by: Neil <4138956+nwparker@users.noreply.github.com>
An ACP agent's TMPDIR sat inside the data folder. On macOS that path made
a socket such as datacloud-mcp-dataAgentKit-antigravityide.sock 131 bytes
long, over the 103-byte limit, so child programs could not bind it.
When a socket under the data-folder root would not fit, use a short
per-user folder in /tmp (mode 0700, owner checked via ensureOwnedPrivateDir,
keyed by a hash of the data root so Orca copies never wipe each other), and
clear the old data root once. Roots that fit, and Windows, keep the current
behavior.
Ported from #33948 onto current main (acp-private-temp now uses
launchPreparationFailure; daemon-socket-endpoint-path moved under
daemon/endpoint/publication; the launch-parity test it touched was removed).
Co-authored-by: Manan Chataut <thefrikidude@thefrikidude.local>
Co-authored-by: Jinwoo-H <jinwoo0825@gmail.com>
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
* feat(cli): select and remove any saved agent account from `orca account`
`orca account select` and `orca account rm` accepted only OpenCode and
Devin. They now cover every provider with saved accounts — Claude, Codex,
OpenCode, Devin and Antigravity — through the runtime methods the
status-bar switcher and Settings already use (accounts.selectClaude,
selectCodex/selectCodexForTarget, selectData, antigravitySelect and the
matching remove methods). No runtime or wire change.
`--account` takes an id, an email/label (case-insensitive) or `system`.
A name shared by several accounts is refused with their ids rather than
guessing which subscription to bill. `system` is refused for Antigravity
(no system default), for Claude from the WSL bridge (the RPC has no
target and would reset the Windows host), and always for `rm`.
`orca account list` now prints each account's id and a `system` row,
accepts `--agent` for every provider, and adds the Antigravity block when
the host can read it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(cli): assert the host refusal for Antigravity accounts from WSL
The host rejects every client-selected WSL target for Antigravity account
management, reads included, so the test now mirrors that refusal and checks
that no select call is made instead of asserting an impossible success path.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(cli): keep only the account-selection tests that guard this change
Moves the guards into account.test.ts (the handler layer that owns them) and
drops layer-duplicating, copy-pinning, Cartesian and mock-only cases per
AGENTS.md test rules; un-exports resolveAccountId (test-only export).
---------
Co-authored-by: Maxime AUBURTIN <m@hellomax.io>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(runtime): fail agent readiness fast when the program reports a sign-in block
agy shows its login menu when signed out and reports OSC 7501
state=blocked:kind=auth. tui-idle treated that as pending, so worker-start
waited out its whole readiness timeout and reported a vague
'Agent did not become ready'. Rank a fresh program sign-in block as a
blocked verdict (agent-sign-in-prompt) so worker-start fails at once with
Agent startup blocked: agent-sign-in-prompt and never types the task into
the login menu.
Fixes#37056 (STA-13300)
* fix(runtime): require an agy sign-in block to persist before failing readiness
agy flashes its sign-in splash at startup and then signs itself in (#28348), so a fresh OSC 7501 blocked kind=auth is not yet final. Hold the verdict pending for 3s from the first auth report (repeats keep the first stamp; idle/working supersede it) and recheck when the window ends.
* feat(github): filter Tasks PRs by state and by who, like GitLab MRs
GitHub PRs in Tasks get the same two chip groups as GitLab MRs: state (Open, Merged, Closed, All) and who (Everyone, Mine, Reviewing). The chips read and write the search query, so labels and typed text survive a click and nothing new is saved. A combination equal to one of the old presets is applied as that preset, so saved views, prefetch and the phone see the same strings as before.
Tracked in STA-14084.
* fix(github): keep right-click on Open saving the open-PRs default view
* fix(github): wrap the PR state and who chips as two groups
* feat(native-chat): show details on link cards for any repo on a known host
A chat link card for a pull request or issue in a repository other than the
workspace's own showed only its name, because the lookup answered only for the
workspace's configured remotes. Previews now read any repository on a host those
remotes use, with that host's sign-in, as the common pattern does. Other lookups
keep the exact-repository check, and an unknown host still gets no credentials.
* fix(native-chat): open link cards for other repos on the chat repo's host
The hover card still admitted only the chat repo's own links, so the wider
lookup was never reached. Links to any repo on a host the chat repo uses now
open the GitHub card. When an older host or a repo the sign-in can't read
leaves it unanswered, the card falls back to the web preview, or just the
link's name, as such a link showed before, instead of a failure.
* fix(native-chat): keep another repo's web preview until its details arrive
A link card for another repo showed the GitHub skeleton first and only fell back
once the lookup failed, so older hosts got two loading stages and a change of
shape. It now shows the web preview it always showed until the details arrive,
and keeps it when none do.
* fix(native-chat): show just the name for another repo's link while its details load
With web previews off, a link card for another repo showed the GitHub skeleton
and collapsed to the name when no details came. It now shows the name from the
start, so nothing collapses when an older host or an unreadable repo answers
with nothing.
* feat(native-chat): show a card for pull requests a chat turn opened
When a turn's own command runs gh pr create and prints the new URL, the chat
shows an "Opened orca#3" card under that turn with the PR's state, title and
changed lines, read live from GitHub. The eye-slash control hides the card kind
and a toast links to the new Settings switch.
* fix(native-chat): satisfy lint in the opened pull request card
* fix(native-chat): keep opened pull request cards through tool runs and failed rereads
The card now rides a turn's work run, a clipped create output is always read
whole (retried when the chat is shown, never while hidden), older Codex JSON
output is decoded, rereads keep the last answer and restart past an in-flight
read, failing lookups back off, and a failed hide save says so.
* fix(native-chat): detect pull requests Codex opens through PowerShell, and wait for clipped output
Codex on Windows runs commands as pwsh -Command '...', which detection now
unwraps (and cmd /c), with tests on Claude's and Codex's recorded calls. A
clipped create output is decided from its whole text, each message is scanned
once, rereads keep their bytes counted, a merged card rereads after an edit,
and the header names another repo's pull request in full.
* fix(native-chat): retry each clipped pull request output until its host answers
Each create output cut short on the wire keeps its own backoff while the chat is
visible; a host whose support is unknown stays pending, and only one that says
it can't give outputs whole is decided from the head. PowerShell options and
backtick escapes are read as PowerShell does.
* fix(native-chat): read PowerShell create commands in PowerShell's own quoting
Decoding a tool call's script moves to its own module. PowerShell launches are
lexed with its quoting (single quotes keep backticks literal, double quotes and
bare text escape with them, here-strings and comments are text) and launch
options follow the common pattern's tables; what can't be read whole opens
nothing. A cancelled retry no longer records a failure.
* fix(native-chat): decode a launch wrapper inside a PowerShell tool call
* fix(native-chat): read Codex's escaped quotes in a Windows PowerShell command
Codex on Windows writes a double quote inside a double-quoted -Command as \",
so a create with an apostrophe in its title was skipped. Tests use the three
calls recorded from Codex on Windows.
* fix(native-chat): read a PowerShell launch line in the quoting of what ran it
Codex joins argv in POSIX quoting even on Windows, splitting one word into
differently quoted pieces, so a script ending in a backslash or holding both
$ and an apostrophe showed no card. The launch line now goes through the same
POSIX word decoder as the login-shell preview, and a PowerShell tool's own
line through PowerShell's quoting.
* fix(native-chat): find a PowerShell or cmd launch anywhere in a command list
A Bash call such as `pwsh -Command 'gh pr create --fill' && echo done` lost its
card once launch lines were read as POSIX words, and Codex's joined
`cmd.exe /c 'script'` never had one. Each command in a POSIX list is now read on
its own, through one launch reader shared with argv, while Node's Windows
`cmd.exe /d /s /c "script"` form keeps cmd's own quoting. PowerShell lexing
moves to its own module.
* fix(native-chat): unwrap a PowerShell or cmd launch only when it is the whole line
Rebuilding a command list around a nested launch could pull one out of a comment
or heredoc into a false card, and an unreadable launch hid a real create beside
it. As the common pattern does, only a line that is one launch is unwrapped;
any other line is analyzed as written. A PowerShell tool's line and Node's cmd
form must likewise be a single launch.
* fix(native-chat): unwrap cmd and PowerShell-tool lines only when they are one launch
Node's cmd form now unwraps only a /c script that is one quoted string with no
quote inside, since cmd escapes none, and a PowerShell tool's line ends at a
bare line break. Text after either launch can no longer pass for its script.
* fix(native-chat): read a PowerShell line continuation as a word break
A backtick before a line break in a PowerShell tool's line joined the next line
into the current word, so an argument on the next line could pass for a
command after the launch.
* fix(native-chat): unwrap only the launch shapes agents are recorded writing
Hand-written readers for cmd launches and for a launch nested in a PowerShell
tool's script kept letting printed text pass for a create. Only argv, a shell's
whole -c line and Codex's Windows pwsh -Command line are unwrapped now; argv is
never space-joined, and detection reads a shell line as whole POSIX words rather
than through the preview's lenient prefix. cmd launches and nested launches are
intended misses until an agent is recorded using them.
* fix(native-chat): read an escaped separator as text when finding a create
An escaped \; stayed in the analyzed commands and passed for a separator, so
printf '%s\n' \; gh pr create showed a card though only printf ran. Escaped
characters are now text, like quoted strings and comments.
* fix(native-chat): remove a line continuation when finding a create
Blanking escapes also turned a backslash-newline into a word, so a create on
the line after a continuation no longer started a command and lost its card.
A continuation is now removed, as the shell removes it.
* fix(native-chat): treat only backslash-LF as a line continuation
Bash removes a backslash before LF but not before CRLF, where the backslash
escapes CR and LF still ends the command. Removing both lost the card for a
create on the line after a CRLF continuation.
* fix(editors): launch Open In editors by id from settings
* test(editors): drop settings casts by narrowing the shell store type
* fix(editors): import shell-open types from their moved path
* test(editors): type the shell test settings instead of casting
* fix(web): confirm before a pairing link replaces a saved server
* test(web): drop the stored-environment cast by narrowing the pairing decision input
* i18n(web): translate the replace-saved-server prompt
The tooling reorganization moved run-typecheck-projects-in-parallel.mjs under config/scripts/typecheck/project-execution/, leaving the mobile package's typecheck and typecheck:tests scripts pointing at a missing file.
* fix(omp): launch OMP chats with their saved model and accept OMP effort
OMP's Chat UI model option names only listed provider/id values, so a
worker started with an OMP role or provider/model:level selector failed
before its first message. OMP resolves those on its launch line, so the
chat's saved model now rides 'omp acp --model', and a new session counts
it as applied. OMP's launch catalog gains its thinking levels, so
--effort is accepted and applied (--thinking in a terminal, the session's
thinking option in a chat).
Tracked in STA-13562.
* fix(omp): keep OMP role chats working after a restart
OMP's ACP model option takes listed provider/id values only, and a load restores the conversation's own model, so a role-started chat failed with "Unknown ACP model" on its first send after an Orca restart.
- New chats keep the saved model on OMP's launch line; resumed chats no longer pass it, so a renamed role cannot stop a good conversation from opening.
- When OMP's model option refuses a value as unknown, Orca applies it through OMP's own /model, the resolver --model uses, as a hidden send-time command. It counts as applied only on OMP's "Model set to" reply and a normal end; cancel or timeout retires the process.
- The command's reply and title notices never reach the chat, its status or unread state.
- Proof from the launch line or /model is retired when OMP changes its model or thinking itself; ordinary model-option proof keeps its existing rule.
- A refusal names the agent and the next step; on OMP older than 18.1.9, whose /model takes ids only, it also names the release to update to, carried on the modelUnavailable fact.
Tracked in STA-13562.
Co-authored-by: Narry <3047793+NarryG@users.noreply.github.com>
* fix(omp): keep role steers in order and say when OMP has no model
- A role steer that OMP's model option refuses now holds its place among queued steers: answers ahead of it are cancelled as they start, and a Stop or teardown while it waits withdraws it without sending /model or the message.
- OMP's "No model selected" (an unresolved launch role, or signed out) is reported as its own failure, "OMP has no model selected. Check your model selection and OMP sign-in, then send again.", instead of a sign-in failure; the next send applies the selection again.
- A cancelled model command reads as not sent, and an option write settles once.
Tracked in STA-13562.
Co-authored-by: Narry <3047793+NarryG@users.noreply.github.com>
* fix(omp): keep OMP's no-model text in the log and test a late /model answer
Tracked in STA-13562.
Co-authored-by: Narry <3047793+NarryG@users.noreply.github.com>
* fix(omp): a Stop still withdraws a waiting role steer after main's turn changes
Main now clears stopAt when a turn finishes, so reserveSteer counts Stops
instead of reading stopAt. Refusal wording moves to acpPromptRejection to
keep the turns file within its line budget; behaviour is unchanged.
Co-authored-by: Narry <3047793+NarryG@users.noreply.github.com>
* fix(omp): a Stop withdraws a role message anywhere in its preparation
A message that falls back to OMP's /model now counts as a held steer from
before its option writes until it is admitted. A Stop in that window
withdraws it unsent, even with no turn running; ordinary messages keep
their Stop behaviour. Restores the launch-role no-resend test main's file
removal dropped.
Co-authored-by: Narry <3047793+NarryG@users.noreply.github.com>
* fix(omp): a Stop before OMP refuses /model reads as a withdrawn message
A Stop the role message already took outranks OMP's own refusal of the
model command, so the message reads as cancelled and never sent rather
than as a model failure. Pins that a Stop with no turn running is still
refused while an idle send's first model-option write is pending: the hold
starts when the send falls back, not before its option writes (the host's
Stop aborts the send's signal there).
Co-authored-by: Narry <3047793+NarryG@users.noreply.github.com>
* test(omp): put the cast's safety note on the line above the cast
* test(omp): cover the new failure copy on phone and desktop, and OMP's thinking levels
Tests the no-model sentence and the update hint in English and one
translation on each surface, every level omp --thinking takes, and that a
missing model classifies as the agent's own refusal.
* test(omp): import the launch module once in the catalog test
---------
Co-authored-by: Narry <3047793+NarryG@users.noreply.github.com>
* Keep a lifetime owner for claimed worktree preparation
* Record preparation ownership loss before retirement
* Type the preparation lock-write fixture without an assertion
* Revoke preparation cleanup when ownership changes
* Show OpenCode choice questions in chat
* Preserve form replies for omitted fields and cover rendered questions
* Keep cross-process form tests out of renderer typecheck
* feat(chat): shared contract for forking a chat from a response
Adds the create-time fork source, the host fork capability, the agent fork
level, the forked-from record relationship and the forked-from divider row.
* feat(chat): advertise each agent's fork level and a chat's fork support
* feat(chat): carry a fork source through chat launch, reload and retry
A launch with forkFrom {sessionId, throughItemId} keeps it on the create call,
the saved launch record, reload, retry and its holders, sends no new-chat
options, and is refused before create on a host that does not advertise forks.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(chat): fork a native chat from a finished response on desktop
Finished root responses offer "Fork from this response" where the host and
agent can fork: disabled while support is unknown or a fork is opening, opened
in a new tab right after the source, with one toast naming the agent on
failure. The forked-from row draws as a divider linking to its source.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(mobile): fork a chat from a response in the message actions sheet
The sheet offers "Fork from this response" with the desktop's eligibility and
host capability, creates the fork through the phone's chat create, and draws
the forked-from divider with a link to the source chat.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(chat): a fork create replays only as the same fork
* Add native conversation copy adapters
* Create native chat forks with derived source markers
* fix(chat): offer a fork only on rows that show the agent's own message
A displayed row can be a tool call or a status projected as a plan while
reading as a response; both clients now check the journal row's own kind.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(chat): open phone and paired-desktop forks right after their source
The phone names its source tab as the create's placement and the host takes
that tab's group; a paired desktop sends the address captured at the click
when the fork has no tab of its own yet.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(chat): float a mid-turn response's fork action instead of adding a line
The action on a response the turn continues past hangs from a line of no
height, as the reply controls hang into the gap, so tool-heavy turns keep
their spacing.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(chat): follow the chat's own fork verdict on both clients
The options read already says whether the installed agent forks; an observed
no hides the action, an unanswered read keeps it disabled, and yes enables it.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Preserve native copy boundaries and rewind behavior
* Preserve imported fork tips and serialize create replays
* fix(mobile): open a fork's source through the session's own chat opener
The forked-from link now asks the controller's existing opener for the source
by history id, so it follows that opener's host, selection and reveal rules
instead of navigating on its own.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Project fork markers from touched journal groups
* Retain fork group restoration for delayed readers
* fix(chat): keep a mid-turn response's fork in its own right gutter
The floated action covered the start of the next row. It now sits beside the
response's last line, in a gutter as tall as the prose, so it adds no line and
never reaches the following row.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Preserve native fork chain boundaries and durable streamed copies
* Place phone forks beside their source in window publications
* fix(tabs): adopt a host's new tab where the host placed it
First adoption put every unseen tab at the end of its group, so a fork the host
published after its source landed last. A new tab now follows its nearest host
predecessor when that tab is here; tabs already adopted keep the user's order.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(chat): reserve the fork gutter on every root response from the first frame
The gutter came and went with the fork offer, so prose reflowed when a turn
finished or support resolved. A pane that draws forks now keeps it on all root
responses, and the fork sits only there, not also beside Copy.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Return Codex copies without automatic provider renaming
* Resolve chat placement once at the current layout owner
* Preserve chat publication when saved opens are skipped
* Retry chat opens when queued placement changes
* Bound stale chat placement retries
* test: expect the OpenCode reader's imported fork tip
* perf(chat): keep the fork offer stable while a chat streams
Every stream commit made a new forkable-id set, so the fork context changed and
every offered response re-rendered. The set is reused while its members hold,
and rows get an action object that changes only with support or pending.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix: keep the fork params and Claude fork worker inside main's new lint limits
* Test fork copy and source refresh scaling
* fix: merge duplicate imports and stringify the OpenCode fork tip id for main's lint rules
* Keep fork placement behind session tab projection boundaries
* test: annotate the Claude copy test's hoisted account homes instead of asserting
* fix: add the fork action and forked-from strings to the English catalogs
* test: give fixtures the fork copy their agents declare; add the forked-from string to the boot catalog
* test: keep registry fixtures consistent with the fork copy their agents declare
* fix(orcad): ship the Claude fork worker with the server; update tests for fork fields
* Test saved fork selections through import and first send
* Preserve ordinary chat publication while saving placement
* Show fork beside the copy button on phone chat responses
Long-press opens text selection on every current phone, so the actions sheet that held fork never opened. The fork button now sits next to each response's copy button.
* Keep the phone fork slot fixed and offer it on every forkable response
The fork slot is now held by the row's own kind, as desktop's gutter is, so finishing a turn or learning the host's support never moves the timestamp. Responses with no copyable text (a visual only) still get fork, and each button re-renders only when its own offer changes.
* Follow main's moved runtime modules in fork code and tests
* test: build the phone response-actions fixture without type casts
* Let the phone read the host's agents so its fork action can turn on
Phone QA found the fork button always disabled: the phone asks the host which agents can fork, and the host refused that read from paired phones.
* Retry the phone's agent-list read, and read again after a reconnect
A read lost to a dropped connection or a timeout left Fork disabled until the chat was reopened.
* Own the phone's agent-list retry timer in its own effect
CI's effect-cleanup check could not see a timer set inside the read's callback; each retry is now a state change with an effect that sets and clears one timer.
* Give each phone connection its own agent-list retries
A count left over from an earlier connection could leave a new one with no retries, so one failed read kept Fork disabled after the transport moved.
* Scope the phone's agent-list retries to each spell of discovery
When the host's fork support dropped and came back on the same connection, the earlier failures carried over, so one more failure got no retry. Each stretch of discovery now owns its reads, failures and retry timer.
* Keep the forked-from source link reachable by screen readers
The divider's separator role and row label hid the source chat button inside it; the rules are now decorative and the text and button read normally.
* Keep fork names owned by adopted chat records
* Use the fork icon for Fork from this response
* Show the phone meta actions only when a row has one
An empty actions group still shifted the time by the icons' edge margin on rows with no copy, fork or rewind.
* fix(chat): put Fork beside Copy in the response's hover row
The fork button sat alone in a right gutter far from the other row actions.
It now joins the response's existing hover controls next to Copy, where those
controls show, and the gutter is gone.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(chat): keep a finished turn's actions on its last words and hold the Fork slot
A turn whose last prose was followed by tool or thought rows showed no Copy
and no Fork anywhere until it was collapsed. The actions now belong to the
turn's last prose row, decided when the transcript's rows are built, and keep
the gap below it whole so they never cover the next row. Root responses also
keep an empty Fork slot when not offered, so resolving support never moves
the row's other actions.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(chat): give a turn's actions only to a reply with words
An image-only reply after the turn's text took the response actions but could
not draw them, so the text lost Copy and Fork. Slot construction now picks the
turn's last reply by the same words the row draws and copies.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(chat): let an image-only response hold a turn's actions when its text is folded
A collapsed turn that ends in an image hides its earlier text, so no drawn row
could show the response actions and Fork vanished. Each turn's actions now go
to its last drawn response with words, else its last with images. Copy and its
context-menu target appear only where there are words to copy.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Fix Claude fork cuts after local commands
* Extract the conversation copy input contract
* Preserve caller results at Claude fork boundaries
* test: check the copied fork for the wake answer without a type assertion
---------
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* perf(native-chat): one chat's status change wakes only that chat
The renderer status feed copied every retained summary and called every
listener on each one-session update, and the status bridge walked every
open tab. Adopt the common keyed-entity pattern: update the session in
place, wake only readers registered for that session, and give the bridge
a session-to-tabs index. Option state now reads only the runtime
permission answer, so a reply preview no longer re-renders the chat.
* fix(native-chat): keep the phone's permission projection and prune stale retry members
Restore projectRuntimePermission for the phone app, ignore a malformed
conversation-name list as the names fold does, drop superseded bridge
members from the retry set, key resume-run ids by the run's start, and
update the continuation test mock.
* perf(native-chat): retry and resume-run checks look only at the touched host and chats
Keep the bridge's retry set per host so another host's update never scans
it, and check only the resume-run entries an update touched.
* test: give the launch create-message feed fake the per-chat subscription
* test: drop call-count assertions from the bridge contact-loss test
Keep the behavior it guards: no throw, revocation applied, later listeners
still notified, and the reconnect snapshot projects.
* perf(native-chat): find a transcript by session id without walking the whole history
Claude lookups probe the chat's cwd bucket, then each project bucket once,
stopping at the first file, instead of reading every folder under projects/.
Codex and omp walks stop at the first match. Session-id discovery for a
transcript that has not appeared backs off to 30 s after its first minute.
* test: drop the launch transcript work-count test the bucket probe replaces
* fix(native-chat): let a cwd hint only fall back, keep its exact encoding, and honor a late cancel
* fix(native-chat): report a cancelled Claude probe instead of a miss; expect the launch cwd hint
* test: spy the walk the transcript resolver now uses in the OMP metadata test
* fix(native-chat): keep known file paths on the fast retry cadence; only id discovery backs off
* test: drop work-count assertions from the adoption budget and OMP metadata tests
* perf(native-chat): a bare tag like Vec<T> no longer makes a streaming reply re-render whole
Only a tag the renderer keeps, left open at the end of its block, can wrap later
blocks; every other tag is escaped or unwrapped. Whole-document mode is now
re-derived on each append instead of latched.
* fix(native-chat): only HTML rendered as text skips whole-reply rendering
Tag counting could not prove a split reply renders like the whole one. Keep the
whole-document rule for every raw HTML node, and skip only nodes the unknown-HTML
transform replaces with text, which never reach the HTML parser. Cuts restart when
that policy changes.
* test(native-chat): drop the render-count test for Vec<T> replies
Main's test rules ban work-count tests; the remount test still covers the
user-visible behavior and the numbers stay in the PR description.
* Count paired-phone chat sends, launches and tab opens in usage stats
The host already knows when a request comes from a paired phone, so it
records the phone's chat sends, agent launches and opened tabs with the
same opt-out as the desktop's own counts. The desktop chat box tags its
sends as desktop.
* Count the views a phone opens and skip replayed sends
A phone picks chat or terminal view on its own, so the opened count now
comes from the stream it opens (chat transcript or full terminal), not
from tab activation. Replayed sends no longer count twice, and phone
sends report an unknown runtime.
* Count a phone's view only once its stream is open
A terminal counts after its snapshot reaches the phone, and a chat
transcript the host cannot read is not counted.
* Tag phone launches where the terminal launch is built
Keeps the launch and runtime identity files within their line budget:
the existing pane-agent reader is public instead of wrapped, and the
phone launch tag is applied where the launch reaches the terminal.
* Keep the phone launch tag when a launch creates a workspace
* Use type guards instead of casts in phone usage counts
* Never let a phone view count end the stream it observes
The agent is read only for a view not yet counted today, inside a guard.
Launch tests that check a caller's own label use a desktop caller, and a
phone launch that creates a workspace is checked to stay tagged mobile.
A command statusLine replaces agy's '? for shortcuts' hint row, so the
composer_ready rule never matched and readiness held whenever OSC 7501
could not decide (no fresh program status). Accept the composer when up to
three status rows sit below it, no busy hint is among them, and no braille
spinner row sits above it (agy's only turn marker in that layout).
Fixes#34707 (STA-12555)
Orca sets a synthetic SSH_CONNECTION so Codex forwards copies over OSC 52
for remote viewers. Codex's shell tool inherited it, so agy launched from
Codex treated the local host as remote and ignored its keyring login.
The wrapper now blanks only Orca's sentinel in Codex's shell environment
policy; genuine SSH metadata is untouched.
Fixes#37723 (STA-13596)
Hook CLI detection now reads the persisted Windows PATH like terminal launch
does, and an explicit enable while already enabled reconciles hooks so
`orca agent hooks on` repairs a wiped ~/.gemini/config/hooks.json.
Fixes#17672 (STA-6166)
* feat(native-chat): holding Cmd/Ctrl turns Send into what Cmd/Ctrl+Enter will do
While the chord's modifier is held alone in a busy chat's composer, Send
becomes a labelled Steer, Queue or Steer queued button, and with an empty
composer the queued card it would steer is highlighted. Releasing the key,
pressing any other key, leaving the composer or the window clears it.
* fix(native-chat): the held-Cmd preview promises only what Cmd/Ctrl+Enter will do
The queue offers its head-card steer only while that card can go (not
being edited, no steer already on its way), Shift or Alt held first and an
open picker show no preview, and a draft Send can't take keeps its button
and reason. The card highlight follows the composer's Steer queued button
in CSS, dropping the state and callback that carried it.
* fix(native-chat): only the composer holding a preview listens for window blur
Every busy chat's composer added window and document listeners; now only
the one showing a held-Cmd preview does, and a hold that outlasts a Cmd+Tab
never shows.
* fix(native-chat): use the Button's own size for the held-Cmd pill, and clear its timer in cleanup
* fix(native-chat): no held-Cmd preview while an IME composes
The IME owns Enter during composition, so the chord shows nothing then. The
listener-count test goes: the work it guarded no longer exists.
* test(native-chat): clean up between held-Cmd preview tests
* fix(native-chat): the held-Cmd delay timer lives in its own effect
The key listener records whether the modifier is down; a separate effect
owns the delay's timer and its cleanup, as React Doctor requires.
* fix(native-chat): a double-click on the held-Cmd button acts once
Sending a draft can turn the button into Steer queued before the second
press, which would steer a queued card the person never chose.
* fix(native-chat): the held-Cmd button is a pill as tall as Send
* fix(native-chat): Cmd/Ctrl+Enter sends the first queued card after a Stop, as steering does mid-turn
* test(native-chat): pin which composers send the head card on an empty Cmd/Ctrl+Enter; goal mode never does
* test(native-chat): the send-state resolver takes only the transport fields it reads, so its test needs no cast
* fix(plugins): say that full-access plugins can also control Orca
Worker plugins can use Orca's local runtime API, so the consent text now
mentions terminals and agents alongside files, network and programs.
* fix(plugins): keep the glossary term in the Japanese full-access warning
* test(e2e): expect the updated plugin full-access text
* test(e2e): expect the updated full-access text in the plugin demo spec
`ssh -G` prints `identitiesonly no` when ssh_config leaves it unset, so the
resolved value always won and a manual target's own IdentitiesOnly was ignored.
Manual targets now use IdentitiesOnly when either the target or ssh_config
enables it; ssh_config-backed targets are unchanged.
Co-authored-by: Martin Vlach <mr.huge@seznam.cz>
* fix(terminal): keep cold-restored scrollback in a tab hidden at relaunch (STA-14063)
When the window predeclares its pane serializer, main skipped seeding its
host copy of the terminal, including for cold-restore payloads. A tab hidden
at relaunch paints the restored scrollback, but when the fresh shell's prompt
arrives while it is hidden, revealing the tab repaints from main's host copy,
which held only that prompt, so the pane came back blank.
Seed the host copy from a cold-restore payload even while the window owns
serialization; the window's xterm holds only this same payload, so it has no
richer history to prefer. Reattach snapshots and replays keep the old rule.
* test(terminal): verify cold restore through recovery snapshot
* feat(mobile): make Markdown easier to read on the phone
Line height 19 → 21 for body, list and quote text (desktop's 1.6 prose
leading), h1 18/25 and h2 16/23 so heading levels differ, an 8pt gap
between list items, a hairline border on inline code, and denser table
cells (8×4). Existing theme tokens only.
* fix(mobile): announce GitHub task boxes as read-only checkboxes
Screen readers now hear each task marker as a checkbox with its checked
state, disabled because the text can't be edited there.
* fix(mobile): size table columns for the denser cell padding
* fix(mobile): use ARIA props for the task box so pages announce its state
react-native-web ignores accessibilityState, so the box read as unchecked
and became a tab stop. role, aria-checked and aria-disabled reach the DOM
and native alike; focusable={false} keeps it out of the tab order.
* feat(native-chat): let users hide the per-turn changed files card
Adds a Changed files summary setting under Settings > Chat > Behavior (on by
default) and a Hide button on the card that turns it off everywhere, with a
toast linking back to the setting.
* fix(native-chat): confirm Hide only after the setting saves
Show Hidden only once the setting is saved and report a failed save instead;
keep the card header one line tall; cover saved settings without the key.
* test(native-chat): import getDefaultSettings from shared constants
* feat(native-chat): reveal Hide on card hover and move the setting to Appearance
The card's Hide control is now an icon with a Hide tooltip that appears on
hover or keyboard focus, and the Changed files summary switch lives in
Settings > Chat UI > Appearance, where Reset restores it.
* fix(native-chat): put the Hide reveal classes on a wrapper, not the Button
Button owns its effects; the hover and keyboard-focus reveal now lives on a
wrapper span, as other row actions do.
* fix(checks): show a check's state in its icon tooltip, not beside it
The status word repeated what the icon already shows. The icon now carries
the state as its tooltip and its accessible name.
* fix(checks): draw the status icon from the same state its label names
A finished check without a conclusion spun as pending while its tooltip said
Neutral. Both now come from one conclusion, matching the header counts.
* feat(checks): show when each check ran and how long it took
Check rows in the right-sidebar Checks panel now show how long a
finished check took and how long ago it finished (or started, while it
runs), with exact start and finish times on hover. GitHub check runs,
legacy commit statuses and GitLab jobs carry provider run times; rows
without times show nothing.
A GitLab pipeline that ran on a commit the merge request has since moved
past says so. The GitHub rollup read keys its 60s cache by the expected
head, so a push never serves the previous head's checks.
* test(checks): bound the clock work of check time labels
* perf(checks): rebuild check time text only when its data changes
Clock ticks now move only each row's age; exact times, duration and copy
are built once per data or language change. Move the run-time mapping
tests to the mapping layer to keep the client test file within budget.
* fix(checks): annotate the gh pr checks reply instead of casting it
* fix(checks): say when a running check started; key the rollup cache by a query comment
A running row reads 'started 5m ago' so its age isn't read as when it
ran. The expected head now rides in a GraphQL comment, which servers
ignore but gh's cache key includes. Add the new copy to the English
catalog and test behavior rather than formatted copy.
* fix(checks): let check times truncate before names in a narrow sidebar
At the narrowest sidebar the time text never shrank, squeezing check
names to nothing and pushing status labels out of their rows. The time
now sits beside the name, gives up width first, and the name keeps a
minimum width.
* fix(checks): rebuild check time copy when translation catalogs change
* fix(checks): reserve room for a running check's growing start age
* test(checks): keep REST check-run times covered through the schema
* fix(checks): read the checks rollup without gh's disk cache
gh's cache key ignores the PR head, so after a push it answered with the
previous head's checks, and it never evicts expired entries. Orca's own
head-keyed checks cache and read coalescing already bound the requests.
* fix(checks): hide check times in rows too narrow to show them
Below a 17rem row the time text is hidden, so the narrowest sidebar keeps
main's name, status and link layout; wider rows truncate the time before
the name, which keeps a 36px minimum.
* fix(checks): truncate a check's duration before its age
The age answers when a check ran, so in a tight row the duration now
gives up its width first.
* test(checks): type the rollup fixture
* fix(checks): show check times only in rows with room for them
Between a 17rem and 19rem row the times squeezed names to their minimum
and clipped; hold them back until 19rem, which the default 350px sidebar
clears.
* fix(checks): keep a minimum name width only while check times show
Rows too narrow for times, including the narrower folder-workspace parent
PR rows, now lay out exactly as before.
* fix(checks): show a running check's elapsed time instead of its start age
'started 10m ago' reserved so much room that running check names shrank
to a few characters at the default sidebar width; beside a running status
the elapsed time ('10m') says the same thing in a quarter of the space.
* test(checks): use a type import for the mocked module