- Gracefully fall back to file-name summaries when staged diffs exceed
node/ssh execution maxBuffer limits, preventing generation failures.
- Split oversized diffs by file and allocate budget via water-filling,
ensuring single huge files do not starve smaller human changes.
- Clip truncated diff sections on line boundaries to avoid half-lines.
* fix(windows): stop main-thread PowerShell storm on env-store reads
Two changes fix the v1.4.52+ Windows performance regression (#4901 regression
against #4840) where 49 powershell.exe processes were spawned in 27 seconds
during load, saturating the Electron main thread and causing black terminals
and runtimeEnvironments:call timeouts.
Root cause: `readEnvironmentStore` calls `hardenExistingSecureFile` on every
read. The env-store parent directory's mtime churns constantly (every secure
write updates it), so the mtime-keyed idempotency cache never matched →
`bestEffortRestrictWindowsPath` (powershell, ~1-1.5s synchronous) fired on
every call. After #4901, the remote-runtime tab-sync loop reads the store
~2×/s, turning sporadic mtime misses into a continuous main-thread storm.
Fix 1 – path-cached directory hardening: add `hardenedDirectoryPathsThisProcess
(Set<string>)` that caches directory hardening by PATH for the process lifetime.
A directory's required ACL does not change when its mtime changes; only file
hardening retains the metadata-keyed cache so post-rename inode changes are
detected correctly.
Fix 2 – async ACL application: replace `execFileSync(powershell.exe, ...)` with
`execFile` (fire-and-forget). PowerShell cold-start is ~1-1.5s; the function is
already named `bestEffortRestrictWindowsPath` so async/optimistic caching is
correct. `applySecurePathRestriction` returns `true` optimistically on win32 so
the cache entry is written before the background process completes.
Tests: new regression tests verify the directory is hardened exactly once even
when its mtime changes between calls, that unchanged files are not re-hardened,
and that ACL application goes through async execFile (not execFileSync).
* fix(windows): apply credential-file ACL synchronously on write path
Follow-up rigor on the env-store PowerShell ACL storm fix (#5006). The
read-path storm fix (path-cached async directory hardening + async file
re-harden) is retained, but switching ALL ACL application to async opened a
narrow Windows-only security window: because writeFileSync({mode}) is a no-op
on Windows, writeSecureFile returned with the credential file still carrying
the parent directory's inherited (broader) ACL for the ~1-1.5s PowerShell
cold-start, affecting the e2ee keypair, device registry, and runtime env auth
store.
Fix: apply the credential FILE's ACL synchronously (execFileSync) on the
infrequent write path, before the atomic rename publishes it, and cache the
path as hardened only on confirmed success so a failed apply retries. Keep the
DIRECTORY hardening async + path-cached for the process lifetime (that is what
killed the #4901/#5006 main-thread storm). The read path's existing-file
re-harden stays async + metadata-cached (fires at most once per file, no storm).
Also:
- Document the dir-path cache process-lifetime known limitation (deleted+
recreated dir not re-hardened until restart).
- Remove the redundant double dir-cache write in writeSecureFile.
- Add docs/windows-secure-file-acl-hardening.md describing the sync-file/
async-dir model and a manual Windows e2e test plan (the cross-platform
Playwright harness runs on Linux and cannot reach the PowerShell path).
Tests (src/shared/secure-file.test.ts, 13 passing): credential file hardened
synchronously while dir stays async (no async file-ACL window); failed sync
file-ACL apply is not cached and retries; dir hardened exactly once across many
writes despite mtime churn; no PowerShell spawned on non-win32.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix: keep POSIX secure directory hardening metadata-aware
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Neil <4138956+nwparker@users.noreply.github.com>
* feat: add windows ssh relay base support
* feat: support windows ssh relay runtime services
* fix: default windows ssh pty cwd to user profile
* fix: support windows hosts over system ssh
* fix: preserve degraded windows relay native deps
* fix: gate windows shell args by relay platform
* fix: preserve windows relay fallback pipes
* test: align windows native deps relay fixture
* fix: build valid windows install lock command
* fix: address windows SSH relay review findings
Resolve correctness, efficiency, and reuse issues found reviewing the
Windows SSH native-host support:
- GC liveness on Windows now probes the actual named pipe (via node
net.connect against markers + deterministic candidates) instead of
substring-matching Win32_Process command lines, which could remove a
live relay dir. Reports ALIVE conservatively only when there is no
liveness signal at all (no markers and no seed pipes).
- Resolve the remote node path once per deploy and thread it through
install/repair/launch instead of re-resolving 3-7x.
- Replace the 200ms node -e poll loop with a single long-lived remote
wait process during Windows relay startup.
- Skip the no-op executable command on Windows in uploadRelay.
- Make the Windows fallback pipe name deterministic and recoverable
(drop the global counter), with an extra reconnect attempt.
- Normalize the prepended node bin dir to backslashes on Windows PATH.
- Batch the system-SSH Windows directory upload into a single streamed
JSON package instead of one ssh process per file.
- Extract relay endpoint/marker helpers into ssh-relay-endpoints.ts and
consolidate the PowerShell EncodedCommand encoding into the shared
powershell-command-encoding module.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* Support cancellation and timeouts in Windows port scanning
- Propagate the request AbortSignal and a 5-second timeout to both
PowerShell and netstat child processes during Windows port scanning.
- Avoid spawning the netstat fallback process if the port scan has
already been aborted.
- Wrap the .NET OSArchitecture check in a try/catch block during SSH
Windows platform detection to robustly fall back to environment
variables if needed.
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: Jinjing <6427696+AmethystLiang@users.noreply.github.com>
* fix(ssh): sync ~/.ssh/config on import and Manage open
importFromSshConfig only inserted unknown hosts and never updated an
existing target, so a rotated port (or any changed config field) was
silently ignored on re-import. Make it an upsert: config-sourced targets
are refreshed in place, genuinely new hosts are inserted, and the SSH
Manage pane auto-syncs when it opens so changes appear without a manual
Import click.
A new optional SshTarget.source ('ssh-config' | 'manual') protects
user-created and user-edited targets from being overwritten by sync —
editing a target in the UI detaches it from config-sync. Unchanged
targets are skipped so a repeat sync performs no disk write. Editing a
target now also carries cleared optional fields (identity file, proxy
command, jump host) so removing a config-derived value actually deletes
it instead of silently keeping the stale value through the partial merge.
Resolves item #1 of #4684 (item #2 was fixed in #4860).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(ssh): preserve legacy manual targets during config sync
Co-authored-by: Orca <help@stably.ai>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Jinwoo-H <jinwoo0825@gmail.com>
Co-authored-by: Orca <help@stably.ai>
* WIP: Changes before auto-review fixes
* Customize Source Control AI action recipes
- Add per-action CLI arguments for generation and launch flows so saved
recipes can select model flags without putting prompts in argv.
- Share the text-generation dialog between commit messages and hosted-review
details, with first-run defaults and repo/global recipe support.
- Launch fix-check agents directly from saved recipes and harden prompt/template
handling for invalid args, blank prompts, and inherited variables.
* Allow custom commands for source control action recipes
- Let text action recipes save and resolve the custom-command sentinel
- Add settings UI for custom command recipes and preserve per-action defaults
- Split large source-control dialogs and direct launch helpers into focused modules
- Keep launch actions from treating custom text agents as runnable TUI agents
* Add per-repo Source Control AI enablement, custom command, and save-targ
- Repo overrides now support `enabled` and `customAgentCommand`, letting
repositories opt in/out of Source Control AI independently and supply a
repo-scoped custom command that takes precedence over the global one.
- Recipe-save dialogs gained a save-target selector ("Don't save / Save for
this repo / Save as global default") replacing the old boolean checkbox,
routing saves through the new `saveSourceControlActionRecipe` helper in
`source-control-ai-recipe-save.ts`.
- `normalizeRepoSourceControlAiOverrides` now returns `undefined` for empty
objects and passes the `null` sentinel through the IPC/RPC layer so the
persistence layer can clear repo overrides cleanly.
- `resolveSourceControlLaunchPlatform` resolves the correct shell platform
for SSH and WSL worktrees so agent launch commands are built correctly.
- Settings UI gained `RepositorySourceControlAiEnablement` and
`RepositorySourceControlAiCustomCommand` rows; draft/label logic was
extracted into focused modules to stay within lint line limits.
* Extract action recipe defaults into own component and use id-prefixed wo
- Move action recipe draft state and UI out of CommitMessageAiPane into SourceControlAiActionRecipeDefaults and source-control-ai-action-recipe-draft.ts to respect the max-lines lint rule
- Use toRuntimeWorktreeSelector() across all runtime git RPC calls so the runtime can resolve worktrees by ID rather than path
- Fix SSH launch platform resolution to use the repo's connection when the newly created worktree isn't hydrated yet
- Add edit and delete handlers for PR conversation comments with confirmation dialog
- Use text-status-success design token instead of hardcoded text-emerald-500
* add more search keyword
* Rename "Enable Source Control AI defaults" to "Show Source Control AI ac
* fix test
* Remove unused imports and variable assignment in launch-work-item-direct
* Fix test mocks to use `mocks.store` instead of `storeState.value` for di
* Extract Source Control AI logic into focused modules with fix-checks dia
---------
Co-authored-by: Orca <help@stably.ai>
* feat: non-blocking worktree creation with in-tab progress
The Create Worktree modal stayed open with a spinning button for the full
create IPC (base-ref git fetch + `git worktree add`, ~10-15s on heavy
repos) and only dismissed once it resolved, so the user stared at a frozen
modal with no way to work elsewhere.
Run creation in the background instead. On submit the modal closes
immediately and an in-tab "Creating worktree…" panel shows live setup
status, wiring the previously-unused `createWorktree:progress` main->renderer
event via a per-creation correlation id. A sidebar row tracks each
in-flight create, the user can navigate to other worktrees or cancel while
it runs, and on success it swaps to the real worktree + terminal in one
frame. Failure shows the error in the panel with retry; remote/runtime
targets (no progress events) show an indeterminate spinner.
Pending creations live in a separate store map rather than a faked Worktree
row, so git-status, the tab model, persistence, and PTY spawning are
untouched. Only the composer quick-create path changes; other createWorktree
callers keep their synchronous behavior.
* refactor: present in-flight worktree creates as inline tabs and rows
Rework the two surfaces that show an in-flight create so each reads like
the real thing it stands in for.
The in-tab panel is now a faux tab: a tab strip carrying the new
worktree's name (the title) over a quiet top-left status line, instead of
a centered card with a step checklist. An in-flight create reads as a real
workspace tab whose content is loading, the title and status never
duplicate each other, and the handoff to the terminal stays a same-frame
swap. Failure shows the error inline with retry.
In the sidebar, a pending create now renders as an inline row under its
target repo group — where the worktree will land — replacing the separate
strip that pinned every in-flight create to the top of the list.
* fix: keep pending worktree rows visible without repo metadata
---------
Co-authored-by: Neil <4138956+nwparker@users.noreply.github.com>
* Suggest enabling local-main freshness when a new workspace finds it stale
Adds a "Keep Local Main Up to Date" suggestion path: when the setting is
off and a new workspace's local base branch is behind its remote, Orca
surfaces a one-time, dismissible toast nudging the user to enable it. The
toast is sticky (no auto-expire) so it can't be missed, with explicit
Turn On / Dismiss actions; dismissing (button, close X, or swipe) persists
localBaseRefSuggestionDismissed so the nudge — and its backend probe —
never runs again.
Also refactors the refresh logic so the advisory and mutating paths share
one fast-forward-safety evaluator, adds an SSH relay RPC for the ref
mutation, and fixes remote-tracking base parsing for fully-qualified refs.
Co-authored-by: Orca <help@stably.ai>
* fix: restore update-ref fast-forward for un-checked-out local base ref
The refactor that split refresh into evaluate + mutate dropped the
non-owner case: a local base branch checked out in no worktree was left
stale (return undefined) instead of fast-forwarded. Restore it across all
three layers — local evaluator/mutator, SSH evaluator, and relay handler
(which also removes the dead duplicated throw) — using the expected-old-OID
compare-and-swap form of update-ref so a concurrent ref move is a no-op.
The suggestion toast now also fires for this case.
Co-authored-by: Orca <help@stably.ai>
* refactor: restore resultBase spread in local-base-ref mutators
The evaluate/mutate split spelled out { baseRef, localBranch, status }
literally in the mutating paths; main used a resultBase spread. Restore
that pattern in both the local and SSH mutators — behavior-preserving,
collapses two identical skipped_error returns.
Co-authored-by: Orca <help@stably.ai>
---------
Co-authored-by: Orca <help@stably.ai>
- Remove `experimentalUnifiedNewTabLauncher` flag and gate; the unified
launcher is now always enabled
- Drop the settings toggle from ExperimentalPane and search index
- Update placeholder/aria text to "Open any file, URL, agent, ..."
- Restyle the input: borderless, flush with dropdown edge, no search icon
- Update test fixtures to include agent-detection state fields now
unconditionally read by TabBarCreateEntry
* Add browser import hint button to the browser toolbar
- Adds a dismissible "Import" button in the browser address bar that opens a popover with browser detection and one-click cookie import
- Persists the hint's dismissed state via `browserImportHintHidden` in `PersistedUIState` so the hint stays hidden after the user clicks "Hide Hint"
- Extracts platform-aware import source labels into `browser-cookie-import-sources.ts` and hint visibility logic into `browser-import-hint-visibility.ts` for isolated testing
* fix the onblur for the import popover
Creating a worktree from a cross-repository (fork) PR previously named the
local branch with the maintainer's branch prefix (e.g. `me/866`) and pushed to
origin instead of the contributor's fork, so maintainer edits never reached the
PR. Fork PRs now adopt the contributor's branch name (matching same-repo PRs)
and resolve a fork push target, with a non-blocking warning when the PR
disables maintainer edits and an indicator showing where a push will land.
- pr-start-point: return branchNameOverride/headSha/maintainerCanModify for
cross-repo PRs (previously only same-repo PRs received these)
- github client: surface maintainer_can_modify alongside the fork push target
- composer: warn (but still allow) when "Allow edits from maintainers" is off
- source control: show the fork push target (owner:branch) before pushing
- extract fork-remote cleanup and setup into dedicated, unit-tested modules
Adds unit suites for the cleanup multi-fork matrix, fork-remote setup,
push-target resolution, the warning decision, and the push-target label.
Note: pre-commit react-doctor hook bypassed — its warnings in useComposerState
are pre-existing (identical count on base) and not enforced by CI.
Co-authored-by: Neil <4138956+nwparker@users.noreply.github.com>
* fix(windows): stop synchronous powershell ACL spawns blocking the main thread
markEnvironmentUsed re-hardened the env-store on every runtime round-trip,
spawning powershell.exe ~5x synchronously (~1-1.5s each) and freezing the
Windows UI on launch and every keystroke. Throttle the non-security lastUsedAt
write (60s window; runtimeId changes bypass) and make path hardening idempotent
per process. ACL hardening is preserved; post-rename credential hardening still
always runs. ~14.0s -> ~6.8s launch; main thread 95.7% blocked -> 99.0% idle.
* Fix secure-file hardening cache retries
---------
Co-authored-by: Jinwoo Hong <73622457+Jinwoo-H@users.noreply.github.com>
- Remove `experimentalUnifiedNewTabLauncher` flag and gate; the unified
launcher is now always enabled
- Drop the settings toggle from ExperimentalPane and search index
- Update placeholder/aria text to "Open any file, URL, agent, ..."
- Restyle the input: borderless, flush with dropdown edge, no search icon
- Update test fixtures to include agent-detection state fields now
unconditionally read by TabBarCreateEntry
* Show Kimi Code subscription usage in the status bar
Add a read-only rate-limit fetcher for Kimi Code that reads the OAuth token from ~/.kimi-code/credentials (honoring KIMI_CODE_HOME) and queries Kimi's usages endpoint, surfacing the 5h session window and weekly quota in the existing status-bar rate-limit UI alongside Claude/Codex/Gemini. Gated on the kimi CLI being detected, with a settings toggle.
Read-only by design: the fetcher never refreshes or writes credentials — rotating Kimi's refresh token would log out the user's live CLI session — and never calls Kimi's completion endpoint. It only reads the existing token and the usages quota endpoint, mirroring how the CLI itself reads managed usage.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* Harden Kimi usage status handling
Preserve cached quota on transient Kimi credential/API failures and migrate the default status-bar item for existing users.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Jinwoo-H <jinwoo0825@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Substring detection mis-fired on titles like "opencode-blinker" or
"claude-scratch", labeling a Codex tab as OpenCode whenever the terminal
title fell back to the bare worktree directory name.
- Extract token-matching regexes into `agent-name-token-match.ts` with
a boundary guard that rejects path separators and hyphenated compounds
- Replace all `lower.includes(name)` calls in `agent-detection.ts` with
`titleHasAgentName()`; re-export `AGENT_NAMES` to keep existing importers working
- Allow `.exe/.cmd/.bat/.ps1` suffixes so Windows launcher process names
(e.g. `openclaude.exe`) still resolve correctly
- Update tests: cwd-path fragments now return null; real agent titles
that contain the name as a proper token still resolve correctly
Brennan BensonandOrca
Jinjing
Jinwoo Hong
Trevin ChowandNeil
KayleeandNeil
Lesley MurfinandJinwoo Hong
