Commit Graph
20927 Commits
Author SHA1 Message Date
Kelvin Amoaba 500a5b5e80 fix(native-chat): stop parsing every chat item's key on each journal commit (#40874) 2026-11-10 21:53:49 +00:00
Brennan Benson 4a21d4f4e4 fix(settings): fit Chat/Terminal tiles and shortcut labels in the narrower Settings column (#40869)
#40249 narrowed Settings to 784px, which left the default-surface section
below its two-column breakpoint (tiles always stacked) and squeezed shortcut
names to a few characters beside their badges.
2026-11-10 13:47:58 -08:00
Kelvin Amoaba abff3a41f5 fix(mobile): serve the phone page after a Linux package update replaces a running Orca (#40863)
* fix(mobile): serve the phone page from a Linux install updated in place

Package managers (pacman, apt, dnf) replace Orca's files while the old
process keeps running. The host cached the manifest for the life of the
process and read assets through Electron's launch-time asar header, so
every read of the replaced files failed with asset_changed until Orca
restarted. Read from app.asar.unpacked and re-read the manifest when
manifest.json changes on disk.

Fixes STA-14385

* fix(mobile): check the page manifest off the main thread, and retry a mismatch

Read and stat the manifest with fs/promises and keep the probed root, so
bundle RPCs add no sync file calls. Keep only passing hash verdicts: a
package install writes files one at a time, so a mismatch may clear.

* fix(mobile): key the page manifest cache on its bytes

Inode, size and mtime can all repeat when ext4 rewrites a same-length
manifest within one clock tick.

* fix(mobile): read a newer in-place install's page manifest

A release written in place may add schema-v1 manifest fields the running
host predates. Read it with a stripping schema so the host serves the
fields it supports instead of reporting no bundle until restart.

* refactor(mobile): derive the installed-manifest reader from the writer's shapes
2026-11-10 21:18:51 +00:00
Brennan Benson 8b1906a797 fix: register repositories before optional hosted metadata (#40050)
* fix: keep hosted metadata off repository registration

* test: update repository metadata ownership fixtures

* test: explain nested import runtime fixture cast

* fix: preserve explicit setup progress and metadata ownership

* fix: settle project links from the prospective saved row

* Keep explicit project choices stable after alias enrichment

* Keep project identity stable when opening repository settings
2026-11-10 13:15:49 -08:00
Brennan Benson 6d3d8245b2 Clean up folder Source Control cards and add Remove from Parent (#40850)
* wip(folder-sc): card header cleanup and Remove from Parent in card menu

Work in progress handed to a Codex session; see ~/orca-qa/folder-sc-card/handoff-codex.md.

* test(source-control): cover collapsed folder card identity

* Keep folder card controls usable while status is unavailable

* Retain branch before card body mounts and allow conflict rows to grow

* Preserve source control label size and folder card parent actions
2026-11-10 12:58:29 -08:00
Kelvin Amoaba d5759e67a2 fix(browser): keep unsent page notes across restarts and ask before closing a tab that has them (#40851)
* fix(browser): keep unsent page notes across restarts and ask before closing a tab that has them

Unsent browser notes lived only in the renderer store and hydration reset them, so a reload, restart or tab close lost them. Each saved browser page now carries its unsent notes, hydration restores them, and every close path (tab strip, keyboard, CLI, phone) asks before discarding them. A page script's window.close() can't discard them.

* fix(browser): keep tabs with unsent notes open on bulk close, like pinned tabs
2026-11-10 20:48:45 +00:00
Jinwoo Hong 88880fca7a feat(orchestration): a paired Orca as a Run's coordinator on the server it placed work on (inert) (#40717)
* feat(orchestration): a paired Orca as a Run's coordinator on the server it placed work on (inert)

The server half of remote-owned workers. A paired Orca's call names one of its
Runs; the server admits it as a peer caller kind, writes that Run at frame
receipt, and holds it to the one access check. Its workers' mail and state
queue home under a sequence no reset reuses. Nothing sends the envelope until
the coordinator side ships, so this changes nothing today.

* fix(orchestration): reserve the peer address, keep conversation kinds apart, queue a correction once with its result

* fix(orchestration): hold back a corrected Task's change whatever its latest attempt; rotate pulls in a stable order per database

* chore(orchestration): typed field read, cast rationales, a wake test on fake timers

* fix(orchestration): a peer's receipts per conversation, its sends only into its Run, a page that fits every conversation, pending adopted Tasks keep their gate

* fix(orchestration): a peer's follow-up places where its worker ran

* fix(orchestration): a peer's current/new-child start gets the --on placement refusal
2026-11-10 14:56:02 -05:00
Kelvin Amoaba 6843d8fd08 fix(browser): stop a stalled tab capture keeping the Mac's display awake for hours (#40847)
* fix(browser): hold a page drawable until its capture settles

A stalled capture stays pending until the page draws and keeps macOS display
sleep off ("Capturing"); dropping the paint hold at our timeout left it there.

Fixes #24946

* fix(browser): keep the debugger attached while a CDP capture is pending

* refactor(browser): share one settle-scoped hold helper for paint and debugger leases
2026-11-10 19:54:49 +00:00
Kelvin AmoabaandWooseong Kim 3a8daa2418 fix(terminal): let a paste made while the terminal starts wait for it like typed keys (#40840)
Co-authored-by: Wooseong Kim <2222333+innocarpe@users.noreply.github.com>
2026-11-10 19:50:17 +00:00
6de7387637 fix(editor): keep Mermaid diagram labels readable in dark themes with pastel chart colours (#40843)
* fix(editor): keep Mermaid labels readable on pastel dark theme palettes

In dark mode the Mermaid config fills nodes, actors and boxes with the
theme's --chart-3/4/5 tokens while every label uses --foreground. Themes
with pastel chart tokens (e.g. Catppuccin Mocha) put light text on light
fills, so sequence actors, class boxes and similar shapes became
unreadable (contrast ~1.06:1). The flowchart contrast pass only covers
flowcharts.

Tint each palette fill toward the background just until the foreground
reaches WCAG 4.5:1. Palettes that already read (Orca's default dark
blues) are unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(editor): tint only Mermaid fills under foreground labels

Darkening the palette seeds also darkened the git graph branch boxes and
venn colours, which Mermaid pairs with its own label colours, and the
active Gantt border. Keep the seeds and tint mainBkg, clusterBkg,
requirementBackground and the Gantt task fills instead.

* fix(editor): pin Mermaid fills only when the theme colour needs tinting

An always-set mainBkg overrode a diagram's own primaryColor in every dark
theme. Pin the tinted fills only when the chart colour fails, and cover
git tag pills too.

* refactor(editor): reuse readableOn for Mermaid fill contrast

Drop the hand-rolled hex mix loop and the extra luminance exports; move
hexToRgb next to rgbaToHex so terminal appearance and Mermaid share it.

* fix(editor): draw clickable Mermaid Gantt labels in the foreground colour

Mermaid's default dark-blue clickable label reads only on light bars; on
tinted or dark bars, and beside a bar on a dark page, it disappears.

* fix(editor): tint Mermaid railroad terminal and comment fills too

---------

Co-authored-by: Pro100v <2681328+Pro100v@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 19:41:13 +00:00
Kelvin Amoaba 83e36f8484 fix(agent-status): keep a finished Codex turn done while its background terminal runs (#40839) 2026-11-10 19:28:43 +00:00
Kelvin AmoabaandManuel Meurer 21de0ae3f9 fix(agent-launch): select a focused CLI chat launch in the window (#40828)
`orca agent launch --focus` and `worktree create --agent --activate` left a
chat launch unselected: main marked the chat active only in its tab listing,
which the window follows solely on its own focus intent (#5435). Send the
existing focusEditorTab for that chat; when its tab is not listed yet, record
the focus intent so the listing selects it unless the user moved on.

Co-authored-by: Manuel Meurer <manuel@meurer.io>
2026-11-10 19:14:44 +00:00
Kelvin Amoaba 34d98debdf fix(mobile): read and resume session history on the server a workspace runs on (#40829) 2026-11-10 19:11:30 +00:00
Brennan Benson 15f7cb194f Avoid repeated tab searches for terminal notifications (#40598)
* perf: avoid repeated notification membership scans

* fix: use the canonical tab transition reader

* test: type optional notification membership fixtures

* test: preserve optional membership type during replacement
2026-11-10 11:09:08 -08:00
Kelvin Amoaba a29fcfe720 fix(linear): refuse relation writes when the duplicate lookup omits pageInfo (#40832) 2026-11-10 19:07:30 +00:00
Kelvin Amoaba 5444524a2a fix(editor): show SVGs without a fixed width in the rich markdown editor (#40825)
The image wrapper was inline-block, so it sized itself from the image's
natural width. An SVG with a viewBox but no absolute width has none, so
the wrapper and the image collapsed to 0x0. Keep the wrapper inline so
the image sizes against the paragraph, as it does in the preview.
2026-11-10 19:06:06 +00:00
Wooseong KimandKelvin Amoaba dcefac8938 fix(agent-status): accept OpenCode 2 hooks after OSC reports opencode (#40759)
* fix(agent-status): accept OpenCode 2 hooks after OSC reports opencode

OpenCode 2's OSC app id is opencode, so the status plugin's opencode2 prompt was treated as a nested CLI and dropped. Branch rename never saw the first prompt. A different OpenCode process still does not own the pane.

* fix(agent-status): keep OpenCode 2 on rows its OSC 7501 calls opencode

OSC 7501 names every OpenCode `opencode`, so after an OpenCode 2 turn ended
its OSC idle relabeled the row and dropped the session it resumes from.
Read the OSC claim as the OpenCode 2 row's generation, and replace the
identity unit tests with one server test of the real report order.

---------

Co-authored-by: Kelvin Amoaba <97001695+AmoabaKelvin@users.noreply.github.com>
2026-11-10 19:01:05 +00:00
Brennan Benson bd44dba7ae Reuse orchestration facts and read pane context directly (#40564)
* Cache orchestration source facts by dispatch and address

* Declare keyed projection services and use lifecycle transaction in regression

* Keep empty orchestration reads lightweight without hiding inserts

* Preserve pane selection and bound unattended projection invalidations

* Use persisted lifecycle inputs in projection regression scenarios

* Own exact completion eligibility at the SQLite source

* Keep public status source fixture deterministic

* Coalesce pane candidates and preserve source refresh ownership

* Align projection fixtures and annotations with static checks

* Remove obsolete identity implementation-count comparison

* Expand each addressed terminal family once per snapshot

* Maintain addressed dispatch winners in source order indexes

* Keep terminal identity rows separate from public orchestration status

* Match optional public status reader to constructor dependency

* Complete orchestration status and fixture database contracts
2026-11-10 10:45:17 -08:00
Brennan Benson 714242cdce Release terminal warning and editor discard history when owners retire (#40725)
* Retire terminal notices and pending discard state with their lifetime

* Preserve the DOM window in terminal notice lifecycle fixtures

* test: exercise fresh terminal eligibility and stale edit output

* Release notice contributions at accepted terminal retirement

* Keep terminal notice delivery independent of store initialization

* Assert the public workspace removal failure result

* Keep buffered terminal retirement within delivery modules

* Initialize canonical terminal maps in workspace slice fixtures

* Expose terminal lifecycle fixture actions after commit
2026-11-10 10:44:32 -08:00
Brennan Benson 7d52a73873 Update paired workspace snapshots without copying unrelated records (#40553)
* Update paired layout snapshots by workspace

* Retain immutable paired snapshot values at intake
2026-11-10 10:37:06 -08:00
Brennan Benson cbb21bd2a8 Stop rebuilding unused dashboard launch choices (#40666)
* perf(dashboard): count launch workspaces during enumeration

* perf: remove unused dashboard launch projection

* fix: scope history catalog retries to changed hosts

* test: use complete history resume fixtures and prune removed baseline

* test: complete history preview timestamp fixture
2026-11-10 10:35:30 -08:00
Wooseong KimandKelvin Amoaba 80a8c28c86 fix(linear): reject connections that omit pagination evidence (#40784)
* fix(linear): reject connections that omit pagination evidence

A list or activity page without a boolean hasNextPage was reported as a finished read. Callers that stop when hasMore is false then treated a dropped page as the full set. Fail that page with linear_partial instead.

* fix(linear): require pagination evidence in every Linear list walker

The desktop issue list, orca linear list, project and custom view issue
lists, project, custom view and project team lists still read a missing
pageInfo as the last page. Route them through the same check, and drop
tests that pass without the fix.

---------

Co-authored-by: Kelvin Amoaba <97001695+AmoabaKelvin@users.noreply.github.com>
2026-11-10 18:21:56 +00:00
buf0-bot[bot] aef89c7fca Update README downloads badge (#40817)
Co-authored-by: buf0-bot[bot] <252831055+buf0-bot[bot]@users.noreply.github.com>
2026-11-10 18:21:44 +00:00
Vitalii Kulishandfactory-droid[bot] 436bef6d13 fix(chat): show readable Cloud Code Assist error messages (#40765)
* fix(chat): show readable Cloud Code Assist error messages

Extract Antigravity's nested API error message before diagnostic truncation,
while preserving existing fallbacks and leaving terminal output unchanged.

Fixes stablyai/orca#40762

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>

* revert: drop unrelated redaction fixture edits from merge

---------

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
2026-11-10 18:19:23 +00:00
Brennan Benson 5af55dc999 Show phone option previews as thumbnails with a full-height viewer (#40569)
* Show phone option preview thumbnails with a full-height viewer

* Use typed renderer host queries in preview interaction coverage

* Register phone preview viewer labels in the mobile catalog

* Keep phone preview thumbnails within the compact frame

* Count supplementary wide characters in phone option previews

* Keep the readable option preview subscribed to theme changes

* Correct mobile preview test typing and positional page identity

* Explain the immutable answer identity used to reset preview scrolling

* Preserve native preview text and measure thumbnail bounds

* fix(mobile): measure preview text at its intrinsic web width

* chore(mobile): document the web-only intrinsic width boundary
2026-11-10 10:16:22 -08:00
Brennan Benson cec9a6627d Enforce native Claude chat identity at launch boundaries (#40548) 2026-11-10 10:14:23 -08:00
Brennan Benson fb5a8a0860 Run setup and archive scripts for folder projects (#40034)
* feat(workspaces): run setup and archive scripts for folder projects

Folder projects get the setup and archive scripts git projects have:
creating a workspace runs setup in a Setup tab, and removing one runs
archive first. The script sees the workspace's name and id, so
`orca worktree create` inside it attaches each new worktree to that
workspace, however the workspace was created.

- Folder creates (desktop IPC and runtime) prepare setup through the
  shared setup prep; runtime creates start it with the worktree terminal
  startup, so Setup-before-agent ordering matches.
- Runner files live in <folder>/.orca, one fresh file per launch, removed
  with the workspace.
- Only the host whose catalog holds the folder reads its orca.yaml or runs
  its scripts; a setup problem becomes a warning, never a failed create.
- Shared orca.yaml setup with no approval from an older client is skipped;
  ask/inherit never refuses a folder create.
- No orca.yaml default tabs for folders, since every workspace shares the
  directory.
- New capability worktree.folder-setup.v1 gates the Settings section on
  runtime hosts; the composer, Settings and trust prompt reuse the
  existing setup surfaces.

* i18n: add folder workspace hooks strings

* chore: drop the runner's sync-fs baseline and an array cast

* fix(workspaces): close review gaps in folder project setup

- Whole-project trust sends an explicit run for folder creates the
  policy would run, since it carries no content hash.
- Folder setup skips file access for a skipped setup and never starts a
  stopped WSL distro for a create nobody started; removal does the same.
- Runtime folder creates return the trust mismatch, so the window re-asks.
- The composer hides folder setup controls on hosts without
  worktree.folder-setup.v1 and keeps their old explicit skip.
- Hook checks read a folder only on the host that owns it.
- Git archive hooks get the removed workspace's id; WSL hooks blank
  inherited pane identity too.
- Runner files are written through the drive path WSL needs for its
  drive mounts, and kept out of git status with .orca/.gitignore.
- The CLI's older-host warning covers --agent creates and --json output.

* fix(workspaces): wait for the folder setup capability and read orca.yaml once

Create now waits for a runtime host's folder-setup answer instead of
sending Skip while the probe is pending, and runtime setup reuses the
parsed orca.yaml when merging effective hooks.

* fix(workspaces): keep a failed folder setup probe unknown and retry it on create

* fix(workspaces): share folder setup capability answers with the composer

* fix(workspaces): re-ask the host for folder setup support on every create

* fix(settings): read orca.yaml for folder projects and use workspace copy in hook rows

QA found Settings reporting no orca.yaml for folder projects while the
composer ran that script, and the inner hook rows still said worktree.

* fix(folder-workspaces): open one Setup tab for an instant folder create

* fix(workspaces): run setup when a created workspace has only editor tabs

* perf(folder-workspaces): per-workspace runner dir, bounded setup answers, one archive config read

* fix(folder-workspaces): retain setup capability for active consumers

* fix(folder-workspaces): match hook search to workspace descriptions

* fix(i18n): drop obsolete hook search descriptions

* test(folder-workspaces): follow relocated setup trust modules
2026-11-10 10:12:12 -08:00
Kelvin Amoaba 12e809c755 feat(mobile): show only the open tab, with a tab count and a sheet of every tab (#40592)
* feat(mobile): show the open tab as one pill with a tabs sheet

The phone copied desktop's tab strip, which fits about three tabs cut to
~95 pt and hides a tab that needs you behind an edge dot. The bar now shows
only the tab you're in, at full width, with the tab count. Tapping it opens
a sheet of every tab grouped by what needs you, with each tab's state and
last activity; swiping it steps to the next tab. The + sheet shows agents as
a grid above Terminal, Browser and Markdown note, and the header names the
branch instead of the tab count.

* feat(mobile): pair the tab count with + and open tab actions from the pill

The count and its chevron inside the pill read like a menu and hid the
"another tab needs you" dot. The count now shares a capsule with +, the
way phone browsers put the tab-count button beside new tab, and tapping
the pill opens the tab's own actions.

* feat(mobile): give the tab count its own button beside +

The shared capsule made each half a small target. The count is now a
round nav button like +, lined up under the header's buttons.

* feat(mobile): draw the tab bar as the strip's square cells

The rounded pill and round buttons read as a different control from the
strip people know. The open tab is now the strip's raised, underlined cell
from the screen edge, with the count and + as square cells in the same
row, which also gives them 46x44 pt targets.

* fix(mobile): keep a hairline after the tab bar's last cell in the top bar
2026-11-10 15:32:40 +00:00
Neil b07b3d318a refactor(renderer): group startup platform and runtime owners (#40777)
* refactor(renderer): group startup platform and runtime owners

* refactor(renderer): reconcile imports and retained validation metadata

* test: clear runtime pairing revisions between repo generation cases
2026-11-10 06:30:52 -08:00
Neil d1466f6017 refactor: finish terminal runtime and typography owner imports (#40761)
* refactor: finish terminal and Mobile defining-owner imports

* fix: retain completion coordinator type binding

* refactor: consolidate direct terminal tree imports

* refactor(renderer): import typography from defining owners
2026-11-10 04:51:57 -08:00
buf0-bot[bot] 2c6f84cac0 Update README downloads badge (#40763)
Co-authored-by: buf0-bot[bot] <252831055+buf0-bot[bot]@users.noreply.github.com>
2026-11-10 12:25:22 +00:00
Neil 811973ea8a refactor(renderer): group file and document feature owners (#40745)
* refactor(renderer): group file and document feature owners

* refactor(renderer): remove orphaned editor recovery registration suite

* refactor(renderer): colocate file explorer tree result contract
2026-11-10 03:46:03 -08:00
Neil 9ef5e48f09 Plugins: add editor.readTypography and editor.typography.changed (#40736)
* feat(plugins): add editor.readTypography and editor.typography.changed

Plugins that render code next to Orca's editor had to open the private
profile database and reimplement Orca's font rules to match it. Host API 1.6
adds editor.readTypography (no capability; worker, panel and editor frames):
the font family and CSS stack, size with editor zoom, weight, line height in
px, ligatures, word wrap and minimap that the file editor resolves now.
editor.typography.changed carries the same values after a settings write or
an editor zoom changes them; it is the first event that needs no capability.

The pure resolution (editor typography, editor zoom, monospace font stack)
moves to src/shared so main and the renderer use one implementation.

* fix(plugins): keep editor typography resolution from throwing in main

A hand-edited profile with a non-string font family made the shared resolver
throw, which the main-process watcher would surface at startup and on every
settings write. Font values are now type-checked, the watcher catches and
logs, and it reads live settings so a write without listeners cannot leave the
event payload stale. Over-long font families are cut to the schema bound
instead of failing every read. Regenerates the testing runtime.
2026-11-10 03:38:47 -08:00
adbccb69a8 fix(terminal): keep remote-viewer clipboard env on command-less creates (#40550)
* fix(terminal): keep remote-viewer clipboard env on command-less creates (#39852)

A paired viewer's plain terminal reaches the host renderer's tab-create
bridge with spawn env (ORCA_REMOTE_VIEWER_CLIPBOARD, PI_CLIPBOARD_OSC52)
but no startup command. The bridge queued pendingStartup only when a
command existed, dropping the env, so the spawned shell never saw the
remote-viewer clipboard marker and copied text stayed on the host
clipboard (#39852).

- Queue the startup whenever the request carries launch material
  (env/envToDelete/launchConfig/resume), not only a command.
- PtyPaneStartup.command becomes optional for env-only startups; spawn
  options already tolerate a missing command.
- Keep the plain-shell pre-bind input hold for startups that plan no
  post-bind work, so an env-only startup types as a plain shell does.

Co-Authored-By: Claude Code <noreply@anthropic.com>

* test(e2e): layout oracle marks only panes the user can see

A hidden tab with a queued startup is laid out at opacity 0 so it spawns at
real geometry. The marker check counted it as visible because it has a size,
then waited for output that the hidden-delivery gate correctly withholds.
Paired-client creates now queue spawn env, so their background tabs hit this.

---------

Co-authored-by: Chen <zwq19980411@gmail.com>
Co-authored-by: Claude Code <noreply@anthropic.com>
2026-11-10 03:30:15 -08:00
NeilandVyacheslav Scherbinin fedb28bbc4 test(ssh): clear SSH session holds between connection test files (#40673)
* test(ssh): isolate connection manager session holds

* test(ssh): reset both session-hold stores in the shared SSH connection fixture

The fixture reset now clears credential declines as well as user disconnects,
so every suite built on it starts without a stale hold from an earlier file in
the same worker. The per-suite resets it supersedes are removed.

Fixes #40291

---------

Co-authored-by: Vyacheslav Scherbinin <gentslava@mail.ru>
2026-11-10 03:02:59 -08:00
Neil dc9a7abfc1 refactor(mobile): group host catalog credentials and descriptors (#40739)
* refactor(mobile): group host catalog credentials descriptors and connection history

* refactor(mobile): import RPC sender and result types from defining owners

* refactor(mobile): colocate push delivery with registration operations

* refactor(mobile): point push recordings at delivery owner
2026-11-10 02:52:14 -08:00
Neil 7c01847447 refactor(renderer): group appearance and terminal feature owners (#40719)
* refactor(renderer): group appearance and terminal owners and use defining imports

* refactor(renderer): remove redundant later module requests
2026-11-10 02:13:08 -08:00
Neil 33a4b473bc fix(notifications): stop repeating bell notifications while an agent waits (#40718)
* fix(notifications): ring the synthetic waiting BEL once per prompt, not per hook restatement

Codex restates waiting on every tool call while a request_user_input_async
question stays open; each restatement wrote a permission title plus BEL, so
Orca posted a terminal-bell OS notification every few seconds for the whole
turn. Keep reasserting the title but ring only once per distinct prompt per
needs-input stretch.

* fix(notifications): keep ringing for repeat approvals during an open async question

Codex's PermissionRequest hook carries no request id, so a per-stretch prompt
set would silence a second identical approval while an async question stays
open. Ring whenever the prompt changes; only async questions, which Codex keeps
open across the turn, stay rung for the whole stretch.
2026-11-10 01:38:45 -08:00
Jinjing 8a10dd371c test: run persisted profile WAL fixture on Node (#40715) 2026-11-10 01:14:56 -08:00
Neil 50e265e6af refactor(mobile): finish native chat owner and test lifecycle cleanup (#40710) 2026-11-10 01:11:28 -08:00
Neil 92a08cdd1d feat(attention): record why a finished agent turn left its workspace read (#40703)
An agent row can turn bold while its workspace never becomes unread, and every
path that drops the finish on the way to the unread flag was silent, so the
next occurrence could not be traced. Each drop site now leaves one
`agent_finish_dropped` breadcrumb (mirrored to main.trace.ndjson) with its
cause, workspace and pane:

- paired-host mirror: a restored, unconfirmed done, or a done skipped as a
  host replay;
- hook lane: a pane it does not consider live, completion notifications off
  for the pane, a new turn end swallowed because an earlier hook end used up
  the turn with no working state in between, a pane pruned during the quiet
  window, or a held turn-end revision that expired;
- dispatcher: a fresh active hook status, a superseded snapshot, an
  attention-policy rejection, or a viewed pane.

No behaviour changes.

Refs #40340
2026-11-10 00:38:10 -08:00
JinjingandClaude 115976acca feat(dictation): offer the speech model update in the dictation pill (#40686)
* feat(dictation): offer the speech model update in the dictation pill

Moves the speech model update out of the launch notification stack and into
the dictation pill. Pressing the dictation shortcut now checks the selected
model before any mic capture; when the speech engine needs a download, the
pill offers it (notice, downloading, done, failed) instead of recording.

Co-Authored-By: Claude <noreply@anthropic.com>

* fix(dictation): address review on the setup pill check

- Drop a stale press before opening the setup pill, not only before recording.
- Swallow cancelDownload rejections.
- Don't cache a failed catalog load, and don't mark a model up to date from an empty catalog.

Co-Authored-By: Claude <noreply@anthropic.com>

* fix(dictation): cancel a pill download closed before its first progress

Co-Authored-By: Claude <noreply@anthropic.com>

* fix(dictation): make the pre-record model check a cancellable pending start

- A toggle, Escape, stop or controller teardown during the model check now
  drops the pending start, and the check rechecks current voice settings
  before opening the mic.
- A press mid-download keeps the setup pill with its progress; only a model
  seen ready is cached as ready.
- A cancelled or failed update no longer deletes the old model files, and the
  renderer keeps offering the setup target it showed until the model is ready.
- The check is prefetched off the startup path and in-flight checks are shared,
  so the first press rarely waits on main.

---------

Co-authored-by: Claude <noreply@anthropic.com>
2026-11-10 00:34:36 -08:00
Neil 2ceaff4869 refactor(renderer): group workspace creation and launch owners (#40688)
* refactor(renderer): group workspace creation and launch owners

* refactor(renderer): combine concrete workspace naming imports
2026-11-10 00:31:11 -08:00
Neil d076ad1a39 fix(serve): timestamp headless diagnostics and log client connections and restart cause (#40694)
Investigating #39166, the reporter could not line up host logs with daemon or
kernel logs, nor tell which client was attached or whether a pause came from
the current start.

- `orca serve` and orcad stamp every stderr diagnostic with an ISO time.
- The runtime logs each authenticated client connect and disconnect (device
  id, scope, transport, connection id; never the client's credential).
- The automatic-work pause log now says whether this start followed a clean
  exit, an unclean one or none, and when the pause began, so a carried-over
  pause is not mistaken for a fresh crash.

Refs #40328
2026-11-10 00:23:04 -08:00
Neil e5a1d2ce02 fix(web): turn off the floating workspace in the browser client and say why (#40699)
The floating workspace's terminals and Markdown notes live on the user's own
computer, and a browser tab connected to `orca serve` has none, so in the web
client its terminal and Markdown actions did nothing.

The web client's stored settings now always read the floating workspace as
off, even if an older build saved it as on, so its button and panel do not
appear. The settings toggle is disabled with an explanation. Anything that
would reveal the panel (for example "open beside", an activity row or "Edit
file in Orca" for keybindings) shows that explanation instead of doing
nothing.

Fixes #40268
2026-11-10 00:19:20 -08:00
NeilandClaude 6798ebcd40 fix(daemon): gate reportFocusLost by daemon protocol and bump to v56 (#39821)
#38730 added the reportFocusLost request inside v55, so v55 daemons from
1.4.251 reject it and an app that adopts such a daemon never replaces it.
Bump to v56 so updated apps start a daemon that is known to accept the
request, never ask pre-v55 daemons, and ask v55 daemons once per daemon
incarnation instead of once per adapter, so a replacement daemon at the
same endpoint is not treated as unsupported.

Co-authored-by: Claude <noreply@anthropic.com>
2026-11-10 00:16:37 -08:00
Neil 482136e352 fix(native-chat): end chat streams that stop delivering instead of leaving them silent (#40691)
A remote chat view could sit on a stream that looked healthy but carried no
frames, so replies never rendered and sends stayed on "Sending…" until the tab
was reopened. Two paths left a stream open after delivery had stopped:

- The host dropped a chat subscriber whose frame failed to send, but never
  ended its stream or logged why. It now logs the failure and ends the stream
  once, so the client resubscribes from its cursor.
- The desktop discards frames from a retired transport generation, but a
  rebuilt control connection advanced the generation without closing the
  streams opened under the old one. Retiring a generation now closes those
  streams, tells their windows (which resubscribe) and logs how many.

The client's reconnect wait now resets only after a stream has stayed up for
10 s, so a host that ends a stream right after its first frames cannot drive a
resubscribe every 750 ms.

Refs #39926
2026-11-10 00:12:08 -08:00
Neil a9db2a6571 fix(session-tabs): retire a mirrored pane once the host has closed its tab (#40692)
A mirrored pane whose leaf lost its PTY binding (for example after its agent
exited) is resolved by pane key. Once the host had closed the tab, recovery
never treated the host's answer as proof the pane was gone, so it re-injected a
pending row into every frame and the closed tab stayed in the tab strip and the
saved session forever.

- Host: an exited PTY record outlives its tab, so terminal.resolvePane kept
  answering "disconnected" for a closed tab's pane. It now answers
  terminal_not_found when no leaf holds the pane, its PTY is not connected
  and the tab is gone from the host's workspace session.
- Client: unbound panes now follow the same rules as bound ones. A close
  record for the pane retires it at once; otherwise two authoritative absences
  (the frame omits the pane and the host cannot resolve it, or resolves it only
  as a disconnected leftover) retire it. Any sign the host still owns the pane
  restarts the count.

Refs #9194
2026-11-10 00:03:27 -08:00
Jinwoo Hong 643411cfa3 refactor(orchestration): one access declaration and check on every method (#40646)
Every orchestration method is now defined with defineOrchestrationMethod and
declares its access: kind (reads, writes or localOnly), the caller field, the
target fields with their subject kinds, whether it reads a collection, and an
optional named rule. The registry refuses an orchestration method whose
handler was not built by those factories, so no method runs without the check.

The existing per-method authorization moves out of handler preambles into
named rules, in the same order: reply (helper answerer or the Run's current
consumer), taskCreate, taskList, taskUpdate, dispatch, gateCreate, gateResolve
and gateList. The check runs where the handler ran, inside the mutation
executor's invoke, keeping each handler's sync or async error contract. An
observed replay of an accepted reply or send runs its declared replay hook
before any rule, so an accepted call is never authorized again. The
session-caller binding reads the declared caller field instead of its own
table, and send --on is carried by a wrapper inside the check.

No behaviour change: a table over every method, recorded on main, pins each
call's outcome for four callers.
2026-11-10 02:44:38 -05:00
Jinjing 225bea6923 fix(native-chat): preview issue links that are pull requests (#40674)
The strict target check added in #38514 rejected a preview answer whose type
differed from the request, so every /issues/N link to a pull request (all bare
#N references in chat) showed "Preview unavailable.". Previews now match on
repository and number and accept either type; other callers stay strict.

Also log the failure class of a swallowed owner/repo lookup error.
2026-11-09 23:37:48 -08:00