mirror of
https://github.com/stablyai/orca.git
synced 2026-11-15 02:17:31 +00:00
de3da487fa6bbfde6f9e4b53fba7672fada5322c
16002
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Neil
|
51cb2f189c | refactor: organize native chat input and transcript owners (#37334) | ||
|
Neil
|
95d1e9127e |
fix(wsl-hosts): a stopped distro still lists all its projects after a restart (#37396)
* fix(wsl-hosts): a stopped distro still lists all its projects after a restart The desktop now keeps each WSL host's last project list (from the repo.list answers the window already receives) and seeds those rows at startup before any live host catalog is requested. The first live list replaces them, so projects deleted on the host disappear on that fetch. Loading the saved list is a desktop read only: no wsl.exe, no host RPC, no \\wsl.localhost access. The saved list is cleared when the host is removed. * fix(wsl-hosts): pin saved project list writes, removal, WSL gate and seed races - Exempt HostRepoCatalogSnapshotPersistence.environmentId from the id-only store write ratchet like the paired-server layout snapshot (the id names the host). - A live list applied before the startup seed now skips that host's seed, and a live answer requested before the seed may still drop seeded rows. - Tests now count saves, cover remove then re-add, check paired servers in storage and in the listing, and cover both seed/live-list orderings. |
||
|
Neil
|
6c6e3d93c3 |
Show a penguin icon for WSL host rows and the stopped-distro notice (#37389)
* WSL hosts: penguin mark for WSL host rows, and on the stopped-distro notice
Sidebar project headers and workspace rows that named a WSL host with a
text badge ("orca-demo (WSL)") now show a small penguin icon instead,
labelled and tooltipped "Runs on WSL: orca-demo (Linux)". The same mark
replaces the host chip in the jump palette. The "<distro> is stopped"
notice leads with the same penguin.
* Wait on the tooltip instead of sleeping in the WSL mark tests
* Redraw the WSL penguin as the classic mascot; focusable tooltip mark; muted token on the stopped notice icon
* Use the classic Linux mascot silhouette for the WSL penguin
* Tolerate unloaded runtime environments in the WSL host mark selector
|
||
|
Neil
|
499da04f10 |
fix(wsl-hosts): keep the e2e fake WSL connector out of release builds (#37378)
* fix(wsl-hosts): keep the e2e fake WSL connector out of release builds The fake WSL host connector was compiled into every build and honored whenever ORCA_E2E_FAKE_WSL_HOSTS was set. Gate main and preload on import.meta.env.MODE === 'e2e' (the mode every e2e build already uses), so other builds fold the branch away and drop the module, and mark its module-level schemas pure so none of it is retained. * fix(wsl-hosts): build the e2e fake's zod schemas lazily so release builds drop them Eager module-level zod chains survived tree-shaking even with a PURE annotation on the outer call, so the fake's schema construction still shipped in out/main/index.js. |
||
|
Neil
|
76320e58cf |
fix(wsl-hosts): remove a freshly installed server that fails its start check (#37388)
* fix(wsl-hosts): remove a freshly installed orcad that fails its start check A WSL connect that uploaded an orcad build whose candidate then failed its preflight or readiness check left the complete install in place, so every later connect reused the same broken bytes. deployOrcad now reports whether it uploaded the slot and whether the candidate's own start check failed, and the WSL deploy removes that slot under the same pins, liveness probe and GC claim the version GC uses (active, rollback target, the serving daemon's build, journaled transactions), so the next connect re-uploads it. * fix(wsl-hosts): fence the failed-install removal and keep a slot any process runs from Review follow-up. The pin check and the delete now run under the activation fence (a held fence keeps the copy), so a concurrent connect cannot preflight or launch the slot between them; it waits, finds the slot gone and re-uploads. Liveness now also scans the process table and the daemon pid files for anything running from the slot, because a candidate's stop ends only its server, not a terminal daemon it spawned there. * fix(wsl-hosts): match the resolved slot path in the slot process probe Node realpaths its entry, so a daemon behind a symlinked home or data root names the resolved slot. The probe now also matches $(cd slot && pwd -P)/ and answers UNKNOWN (keep) when it cannot resolve. * test(wsl-hosts): wait for sibling slots to run before proving this slot idle Replaces a fixed 100 ms sleep (flagged by test-no-real-time-wait) with an expect.poll until each sibling's own slot probes LIVE, so the DEAD answer for this slot is meaningful. |
||
  Kelvin Amoabaandagustin2606
|
47506f3f33 |
fix(github): link a branch's open PR instead of a newer closed one (#37452)
* fix(github): prefer the open PR when a branch has several The branch lookup asked GitHub for one PR across all states. GitHub sorts a branch's PRs newest first, so a closed PR opened after the live one shadowed it and got linked to the worktree on every checkout. Fetch a page of the branch's PRs, REST and the `gh pr list` fallback alike, and take the open one; with none open, keep the newest as before. * refactor(github): keep branch lookup casts on their own lines --------- Co-authored-by: agustin2606 <agustinvarela2606@gmail.com> |
||
 Kelvin Amoabaandrafael rotiroti
|
a444233103 |
fix(sidebar): keep folder project roots visible when hiding the default branch (#37444)
* fix(sidebar): keep folder roots visible when hiding default branch * test(sidebar): trim folder-root hide-default tests to distinct cases --------- Co-authored-by: rafael rotiroti <rotirotirafa@gmail.com> |
||
 Brennan Benson
|
1ce79608a4 |
Full access Claude chats stop asking for approval (#37246)
* fix(claude): Full access approves Claude's own permission asks in chats A Claude chat in Full access still showed Allow/Deny cards whenever Claude Code forwarded an ask it keeps under bypass mode (ask rules, ask-answering hooks, built-in shell safety checks). Approve those at once while the chat runs with Full access, read per request; questions and plan approval still go to the user. * fix(claude): Full access grant comes from the launch; Claude's mode reports restrict it in wire order Only a process Orca launched with Full access can auto-approve Claude's own asks; Own settings never does, whatever Claude reports. Claude's init/status mode reports, read from stdout before the SDK sees them, can suspend and restore that grant, and each permission request keeps the mode in effect at its own position in the stream, so a paused journal consumer or a read-ahead chunk cannot apply a report late or early. Questions, plan approval and requests Claude flags as needing the person's answer keep their card. * fix(claude): Full access keeps every request's own wire position and restores after an approved plan A permission request Orca never saw on stdout no longer inherits a later mode: each request keeps the first snapshot for its id until the SDK hands it over, including after a cancel and past any number of unread requests, and requests resent in initialize's reply are recorded where that reply arrived. Approving a plan in a Full access chat now reads Claude's mode in wire order, so the restore is sent even while the chat's presentation still lags behind the plan report. * test(claude): narrow contract-pin frames to permission requests with a type guard * test(claude): count the policy's stdout search work instead of timing it * test(claude): type the contract-pin frame guard with the fixture frame union * refactor(claude): let the launch permissions wrap the exit handler, keeping acquisition under its line limit |
||
|
Kelvin Amoaba
|
8458e704c4 |
fix(native-chat): show a terminal's unsent input in the chat composer and replace it on send (#36493)
* fix(native-chat): show a terminal's unsent input in the chat composer and replace it on send Switching a pane to chat view now seeds the composer with the text left unsent in Claude Code's or Codex's input box, and a chat send clears every line of it instead of one Ctrl+U. The clear is confirmed with the same typed-only read and kept under the 64-byte chunk Claude Code reads as a paste. * test(native-chat): drop the removed screen-text clear check from the background-rows test * fix(native-chat): clear a seeded multi-line draft before typing a Codex command from chat * test(native-chat): expect the typed command's clear argument in composer tests * fix(native-chat): confirm a seeded draft is cleared before typing a Codex command * fix(native-chat): find Claude's closing frame past dim suggestion rows A typed-only read stopped at the suggestion row, so a draft under Claude's hidden caret never reached chat. |
||
|
Brennan Benson
|
255401c2a5 |
Continue chats automatically when a usage limit resets (#36873)
* fix(native-chat): an open chat moves to the account you switch to before its next message A running chat's agent read its account only when it started, and nothing compared it with the account a start would use now, so open chats kept answering on the old account after a switch. The agent a chat runs now records the account its launch used, and the host's replace-before-the- next-send check replaces an idle one whose account is no longer what a start would use. A chat mid-reply finishes on the old account first. Claude chats follow; agents whose chats keep the account they were created on are never compared. A chat's model-list reads and saves now use the account it runs on, so a moved chat no longer writes its new account's list under its first account (STA-11281). Tracked in STA-11211. * fix(native-chat): declare the launch-account read for the runtime layer that wires the host * fix(claude): find custom chat history before managed-account resume * test(claude): assign the complete custom account home * feat(native-chat): offer account switching and hand-off at usage limits * fix: track the provider in account-list reload effects * fix(native-chat): target recovery actions at the chat host and transcript * fix(native-chat): verify account recovery for each chat on its host * refactor(native-chat): retain the existing image-support result * fix(native-chat): withdraw account recovery after failed eligibility reads * fix(native-chat): leave unavailable recovery facts absent * Align Japanese usage-limit recovery copy with catalog convention * Preserve complete chat context through usage-limit recovery * Keep Continue callback focus optional * Keep recovery fixtures and account effects type safe * Keep chat recovery on native sources and native accounts * Validate native recovery files after path admission * Use the optional cursor shape in recovery fixture * fix(native-chat): exclude source CLI family from limit handoff * Resolve recovery context in the owning workspace directory * Add the continue-at-usage-limit-reset setting and its chat row lines * Continue chats automatically after usage limits reset * Fix automatic continuation fairness, lifecycle and host settings * fix: bind automatic continuation withdrawal to accepted actions * fix: withdraw source auto continuation after handoff starts * Read unopened chats with the saved-record reader for usage-limit continue Unopened chats were read through the raw-row replay, which no longer sees chats saved as keyed records, so a usage-limited chat that was not open was never continued after its reset. Read them the way the rest of the app reads a chat without opening it. Also use the shared visible-tab check, keep the ordinary rewind result without a sequence as main does, and adapt two tests to the current queued-message and saved-status behavior. * Keep a usage-limited turn idle after any rewind restores it Rewind completion read the resulting conversation through the raw-row replay, which no longer sees chats saved as keyed records, and a forked rewind wrote its completion before the new history's records. So a limited turn that a rewind brought back was continued automatically after its reset. The completion now reads the saved result the way the app reopens a chat, and the forked rewind writes its completion after the new history in the same transaction. * Type the usage-limit auto-continue test mock without an assertion * Fix type errors from main's settings, failure-wording and refusal changes * Keep ordinary settings replies when the paired host changes mid-call, and list the cancel method in wire contracts Settings reads and writes that don't carry the automatic-continue choice go through the common runtime call again, so a reply that completes after the paired host is removed still merges, as before this change. Only a write of that choice stays pinned to the host it was made for. The structured method count and the cross-version method manifest now include the cancel method. * Keep a stopped turn's automatic continue withdrawn when its usage limit arrives later A person could Stop a running Codex turn, have the turn then end with a usage-limit failure, and still get an automatic "Continue where you left off." after the reset, because Stop had nothing to cancel yet. The journal write that first reveals a limited turn the latest personal Stop owns now saves the same exact cancellation in that transaction, and a Stop accepted just after the limit landed withdraws it too. Resume, deleting a queued card and restarting keep it withdrawn; manual Continue and later turns are unaffected. * Bound the automatic-continue sweep and keep notice rows from re-rendering The 10-second sweep now does nothing while the setting is off, republishes only chats with a scheduled continuation, reads at most three unopened chats per pass, and keeps going when one chat's saved history can't be read (that chat is retried after the others instead of stopping every chat on the host). Settings writes republish only when the switch changes. Working out the usage-limit cut no longer indexes the whole conversation on every streamed row. The chat's notice rows no longer re-render for status changes they don't show. Adds operation-count and render-count tests for these bounds. * Test that streamed journal commits after a Stop scan no history and build no command index Drives real Codex rows through the host at 20 and 200 history rows: each streamed commit runs the Stop's commit hook and the status projection's cut reader with no newest-turn scan, no render and no full command index; the next turn end scans once and renders nothing. * Leave a usage limit manual once its reset passed more than a day ago Turning automatic continue on later no longer sends into chats whose limit reset days ago. Eligibility, which the row's summary, the sweep and the pre-send recheck all read, now requires the reset to be within the last 24 hours; older limits show the usual manual actions and are never continued unasked. * Show only manual actions for a usage limit Orca can never continue A limit more than a day past its reset is never continued, but its row still showed the setting link, inviting people to turn on a switch that could not help that chat. One shared check now decides whether a cut can ever be scheduled; the status summary and the sweep's eligibility both use it, so past the bound the row keeps only Continue Now, Hand Off and Add Account. A row already on screen when the day passes drops the line on the shared minute clock; the host alone decides sends. * List every journal row kind in the Stop cut filter The type-aware lint requires exhaustive switches; epoch, tombstone, submission and queue rows still cannot reveal a cut. * Type the status-subscription mock instead of casting its handler The changed-code casting gate rejects type assertions; the mock now carries the real subscription signature, so the recorded handler is typed. * Open store-less conversations without the Stop withdrawal hook Acquisition-only opens have no record store, so a person's Stop followed by a limited turn end made the journal commit throw and drop the turn end. Install the withdrawal only when a store exists; without one there is no continuation to withdraw. Also narrow the status summary before reading it and type the sweep test's send mock. |
||
|
Brennan Benson
|
c1641a8d7d |
Show the open Orchestration panel as a floating card (#37412)
* Show the open orchestration panel as a floating card The open panel now floats in the closed tab's corner as a card as tall as its rows. In chats it stops above the composer (the transcript frame already ends there); in terminals it keeps the bottom quarter clear for the CLI's prompt and recent output. Past that it scrolls inside. * Keep the orchestration card's collapse button reachable in tiny panes A terminal split can be 50px high, leaving the card no room under the 75% limit, so its header and collapse button were clipped away. The card is now never shorter than its header. * Keep the orchestration card's collapse button inside narrow cards In a narrow split the card is 20px narrower than the pane and its title didn't truncate, so the collapse button was pushed past the card's edge. The title now truncates and the header's side padding gives way before the button, down to the 50px pane minimum. |
||
|
Brennan Benson
|
10a623b910 |
Launch an agent with a first prompt from its setup panel (#36659)
* Launch an agent with a first prompt from its setup panel The agent setup panel in the new-tab menu gets an optional prompt box above its Launch buttons. It is the chat composer in a compact launch layout: one line that grows to six, Attach as its only control, / listing the launch workspace's skills and commands and @ finding its files. Enter launches the panel's default way, either Launch button sends the prompt its own way, and with nothing typed both launch exactly as before. The prompt rides the existing launch prompt field and is kept in memory only until the launch. * Preserve launch prompt input across submission and refusal * Retain scoped composer submissions through launch ownership * Keep launch prompt tests deterministic during collection and persistence checks * Retire launch source drafts at committed qualified removals * Retain workspace launch setup through ordinary rich drafts * Preserve rich draft authority through accepted setup handoffs * Keep setup prompt moves within current rich draft memory * Create setup drafts on first input instead of on row render Rendering an agent row in the launch menu or search no longer creates a setup record. The row derives its scope from its workspace, host and agent without touching session state; the record is created by the first typed prompt, option change or pending attachment, matching main, which kept only authored setups. Catalog applies, readoptions and renames now visit authored setups only. Option listeners use the existing per-scope draft listeners and subscribe only while the setup popover is open. Also return the runtime-status files to main, which only carried leftovers from a removed hook. * Return group-delete draft revocation to main and repair launch tests Deleting a project group no longer retires in-flight setup submissions for its folders: folder keys are never derived again, so it changed nothing users could see. The group mutation, its removal state and the prompt-delivery type widening go back to main's content, with the test that exercised the revocation. Test fixtures catch up with the setup prompt composer: the TabBar window stubs gain event listeners, the picker inventory's mock store gains agent status, create-rejection handbacks expect inComposer, and the folder removal test expects a same-ID survivor's chip to stay, as on main. * Keep each search preset's setup options in its own slot Opening a search result's setup no longer writes its preset over the workspace's setup. As on main, each preset keeps its own options slot, now inside the workspace/host/agent setup, keyed by what the search parsed rather than by defaults; the plain row keeps its own slot. Only a real change saves, into that entry's slot, and creates the setup then. The typed prompt stays one per workspace, host and agent. * Fix the moved agent profiles import and merge duplicate imports Main moved the native chat agent profiles module under shared/native-chat, so the setup prompt composer imports it from there. Files that imported one module twice now import it once; nothing else changes. * Fix the web typecheck for the setup prompt changes The catalog continuity test no longer imports the main process's SSH target reassignment, which pulled the whole main graph into the web program; it builds the reassigned folder inline. Test fixtures and spies now match current types, the edit-session tests pass a composer ref as production does, and the setup panel again accepts callers that return no launch answer. The pending-chip owner filter drops an unused name. No shipped behavior changes. * Show the setup prompt's skill and file lists above the setup panel In the agent setup popover, the / and @ lists opened underneath the panel, so they could not be seen or clicked. A list opened inside a popover now takes the menu layer that selects and dropdowns already use above popovers; an ordinary chat's lists keep their container and layer. The prompt placeholder is translated in every locale, and folder draft continuity takes the fetched catalog itself instead of reading its folder list field. * Fit the setup prompt box to a placeholder that wraps With a longer translation, the empty setup prompt's placeholder wrapped to a second line that the one-line box cut off. The empty launch prompt now lets its placeholder's wrapped height count, so the box fits it; English still fits one line, and an ordinary chat's box is unchanged. * Type the submit test's pending sends as the boolean main now returns * Show the dictation button in the agent setup prompt * Update draft test handle for agent references |
||
|
Neil
|
2541edc98e |
feat(settings): import Kitty configs through a shared terminal config import pipeline (#37375)
* feat(settings): import Kitty configs through a shared terminal config import pipeline * fix(settings): harden Kitty include and value parsing * fix(settings): match Kitty globs linearly, skip read-only system configs, report non-family font selectors * refactor(settings): read current settings through a typed map when diffing imports * style: format Ghostty importer * fix(settings): type the import hook stub and let dialog primitives own their typography |
||
|
Neil
|
36f17a04a7 |
plugins: Box layout container for components panels (catalog 4) (#37420)
- Box (ui.catalog 4, since 1.4.248): direction, gap/padding/paddingX/ paddingY/margin on the gap scale, align, justify, wrap, grow, shrink, width/height/min/max (px up to 4096 or whole percents), columns, scroll, border (true or a tone), radius, background (theme surfaces or tone tints). Strict props: hex colours, CSS lengths and unknown props are dropped with a warning like other components. - Renderer maps props to theme tokens through data attributes; sizes reach inline style only as validated px or percents. - Typings, docs, skill references, gallery Layout tab; orca/testing finds Boxes and summarizes danger or warning Boxes. |
||
 Jinjing
|
81ec82b365 |
Rename Chat settings to Chat UI (#36613)
* Rename Chat settings section to Chat UI * Localize Chat UI settings titles * Update chat settings search fallback to Chat UI |
||
|
Neil
|
3fa4a059d0 |
feat(settings): import Ghostty themes, including light/dark pairs, as per-mode terminal themes (#37368)
* feat(settings): import Ghostty themes, including light/dark pairs, as per-mode terminal themes * fix(settings): keep Ghostty theme re-imports stable and explicit colors when a theme half is missing * fix(settings): reuse identical built-in terminal themes and label Ghostty variants * fix(settings): keep same-named Ghostty theme files apart, import theme divider colors, follow Ghostty lookup order for picked configs |
||
|
Brennan Benson
|
3aa633d389 |
Mention an agent from the orchestration panel with one click (#34836)
* Add scoped terminal identity acquisition and preserve crew source wakes
* Share raw crew identity inputs through scoped production readers
* Exercise crew generation retirement through actual source owners
* Retire crew inputs on consumed ownership and visibility writes
* Use typed crew mutation and fixture access
* fix: use owned crew styles and committed render counters
* Repair crew fixture and RPC type contracts
* Align crew fixtures with pane and mention contracts
* Count crew consumer render attempts in the fixture
* Localize crew panel labels and details
* Align crew integration fixtures and RPC parameter contracts
* Keep mention contracts within their compiler boundaries
* Encapsulate terminal identity mutations and captured reader contracts
* Seed terminal identity fixtures through canonical mutations
* Complete crew panel English catalog entries
* Fix terminal identity fixture return type
* Cover terminal identity lifecycle schedules
* Align terminal fixtures with canonical identity sources
* Repair runtime fixture source contracts
* Use checked database fixtures for federation contracts
* Update crew source inventories and exit fixtures
* Preserve crew setup outcomes at existing subscription owners
* Match managed restore fixture to its nullable async return type
* Keep absent setup observations distinct from undefined failures
* Drive pending Crew questions from existing metadata
* Preserve nested pairing repair outcomes for shared callers
* Route attached workers to explicit local recipients
* Type Crew pending-question measurement receipts
* Match SSH repair fixture target absence to store contract
* Repair composer fixture and consolidate crew English catalog
* Add finite crew subscription boundary candidates
* Isolate crew portal in terminal file drop fixture
* test: use checked property access for crew owner assertions
* Consolidate web runtime auth type imports
* test: type crew API legacy arguments and pending results
* Bound cold terminal layout construction and finite pane acquisition
* test: use public runtime identity in pane acquisition fixture
* Follow image reference module relocation in mention serializer
* Avoid repeated searches in cold workspace layout loading
* Verify cold loader fault identity and hook claimant effects
* Revert "Route attached workers to explicit local recipients"
This reverts commit
|
||
  Rayyan34andNeil
|
e1ff15ee91 |
feat(settings): preview each font in its own typeface in font pickers (#33254)
* feat(settings): preview each font in its own typeface in font pickers * perf(settings): skip offscreen font rows so opening the picker loads only visible fonts * fix(settings): keep symbol font names readable in font pickers * fix(settings): size skipped font rows by their content box --------- Co-authored-by: Neil <neil@stably.ai> |
||
|
Neil
|
adb2d2d2cd |
feat(settings): delete imported terminal themes from the theme picker (#37356)
* feat(settings): delete imported terminal themes from the theme picker * fix(settings): theme delete reads latest settings, undo keeps later choices, focus stays in picker * fix(settings): serialize theme deletes and keep later edits on undo * fix(settings): theme delete button visibility on a wrapper, not a Button restyle |
||
|
Brennan Benson
|
c7195a8d50 |
Show who each agent works with in a panel on its chat or terminal (#34891)
* Nest chat-UI orchestration workers under the agent that started them
Main's lineage projection only walked terminal panes, so workers running
in the chat UI never got a parent link and listed flat in the sidebar.
It now also projects chat workers, keyed by their visible chat row, and
resolves a chat creator or coordinator to that chat's visible row
instead of its handle or private mailbox pane key.
* Fix public worker lineage across assignment and row identities
* Complete terminal lineage provenance and bound active session selection
* Repair lineage test runtime boundaries and SQLite counters
* Provide read-only crew snapshots for each agent
* Batch public lineage verification after exact recency selection
* Derive crew relations once per meaningful host change
* Remove package manager lockfile churn from hook
* Show each pane’s agent crew with open and copy actions
* Use checked runtime values in lineage integration tests
* Preserve host-aware crew navigation remedies
* Document the trusted Dispatch SQL row boundary
* Keep lineage integration imports within their test programs
* Load native lineage fixtures before test execution
* test: complete lineage inputs in runtime and sidebar fixtures
* test: seed benchmark facts from active dispatches
* test: replace partial orchestration DB casts with typed fixtures
* test: complete stale-leaf mailbox fixture session input
* Derive durable crew parents from retained assignment proof
* fix(test): render worker lineage fixture with production list
* fix: preserve crew pane controls and summary identity
* refactor: extract crew row details presentation
* Add scoped terminal identity acquisition and preserve crew source wakes
* Share raw crew identity inputs through scoped production readers
* Exercise crew generation retirement through actual source owners
* Retire crew inputs on consumed ownership and visibility writes
* Use typed crew mutation and fixture access
* fix: use owned crew styles and committed render counters
* Repair crew fixture and RPC type contracts
* Count crew consumer render attempts in the fixture
* Localize crew panel labels and details
* Align crew integration fixtures and RPC parameter contracts
* Encapsulate terminal identity mutations and captured reader contracts
* Seed terminal identity fixtures through canonical mutations
* Complete crew panel English catalog entries
* Fix terminal identity fixture return type
* Cover terminal identity lifecycle schedules
* Align terminal fixtures with canonical identity sources
* Repair runtime fixture source contracts
* Use checked database fixtures for federation contracts
* Update crew source inventories and exit fixtures
* Preserve crew setup outcomes at existing subscription owners
* Match managed restore fixture to its nullable async return type
* Keep absent setup observations distinct from undefined failures
* Drive pending Crew questions from existing metadata
* Preserve nested pairing repair outcomes for shared callers
* Route attached workers to explicit local recipients
* Type Crew pending-question measurement receipts
* Match SSH repair fixture target absence to store contract
* Add finite crew subscription boundary candidates
* Isolate crew portal in terminal file drop fixture
* test: use checked property access for crew owner assertions
* Consolidate web runtime auth type imports
* test: type crew API legacy arguments and pending results
* Bound cold terminal layout construction and finite pane acquisition
* test: use public runtime identity in pane acquisition fixture
* Avoid repeated searches in cold workspace layout loading
* Verify cold loader fault identity and hook claimant effects
* Revert "Route attached workers to explicit local recipients"
This reverts commit
|
||
|
Neil
|
9d22b026de |
feat(settings): let the Ghostty import read a config file you pick (#37353)
* feat(settings): let the Ghostty import read a config file you pick * fix(settings): guard Ghostty file-pick races and non-regular files |
||
|
Neil
|
522c7b6b8c |
fix(fonts): list Windows fonts through Chromium instead of PowerShell and surface failures (#37369)
* fix(fonts): enumerate Windows fonts natively instead of through PowerShell Read the registered font files' name/fvar/STAT tables so a Group Policy block on PowerShell no longer collapses the picker to five fonts, and say so in the picker when the OS list is unreadable. * fix(fonts): harden sfnt parsing and Windows font file discovery * fix(fonts): follow STAT naming rules and bound Windows font reads * test(fonts): type platform timeout cases without assertions * refactor(fonts): list Windows fonts with Chromium's queryLocalFonts instead of a custom sfnt parser * fix(fonts): scope the local-fonts grant to the main window and keep the web client quiet |
||
 Neilandmmarabel
|
72099885dd |
fix(appearance): changed divider colors also paint workspace splits (#37342)
* fix(appearance): changed divider colors also paint tab-group splits Co-authored-by: mmarabel <mmarabel@users.noreply.github.com> * fix(appearance): follow the terminal divider slot and keep the value's case --------- Co-authored-by: mmarabel <mmarabel@users.noreply.github.com> |
||
|
Neil
|
e7575e5bed | fix(theme): ignored files in the explorer meet 4.5:1 in the built-in themes (#37355) | ||
 
|
ab0e81d341 |
fix(repo-icon): round project image icons on desktop to match mobile (#37383)
* fix(repo-icon): round project image icons on desktop to match mobile Mobile rounds project image icons with a 4px radius, but desktop draws them with square corners, so the same avatar looks different across clients and sits awkwardly beside the rounded sidebar controls. Apply the same 4px radius to the image in RepoIconGlyph so every desktop surface picks it up. Emoji and lucide icons are unchanged. Covered by a repo-icon test that failed before the change. * test: drop repo-icon radius test --------- Co-authored-by: Rongzheng Liu <tianlrz@users.noreply.github.com> Co-authored-by: Neil <neil@stably.ai> |
||
|
Neil
|
f8ce00682c |
Say why to move projects in the WSL move dialog (#37390)
* WSL move dialog: one line on why to move, and the way back * Capitalise Git in the WSL move why-line |
||
|
Neil
|
f358f420af |
fix(wsl): offer the existing host when adding a folder from a distro that already runs as one (#37377)
* fix(wsl): offer the existing host when adding a folder from a distro that already runs as one * fix(wsl): translate the existing-host offer, let a later drop win, keep the offer when the host fails to start - Translate existingTitle/existingWhy/existingAccept into every locale. - A drop onto the open dialog after "Add on host" now prefills the dropped path. - A failed host start keeps the offer so the user can retry or add on this computer. - Tests run the real saved-host read (flag off gives the run-as-host offer) and the offer-to-prefill wiring. |
||
 NeilandClaude
|
2bbf73adc4 |
fix(wsl-move): land on the workspace view after a move's window reload (#37376)
* fix(wsl-move): land on the workspace view after a move's window reload A move or "Use on this computer again" reloads the window from Settings. Settings cannot restore which page was open, so the reload opened General, and the saved view kept reopening Settings on later launches. Main now switches the saved view from Settings to the workspace view right before that reload. Co-Authored-By: Claude <noreply@anthropic.com> * fix(wsl-move): leave Settings even when the window is gone at reload The landing view now comes from the write hold whenever a reload is owed, so a window closed before the reload still relaunches on the workspace view. Co-Authored-By: Claude <noreply@anthropic.com> * fix(wsl-move): keep the reload going if the landing view cannot be saved Picking the landing view is now best-effort, so a failing store write no longer strands the write hold or skips the reload. Adds a wiring test that fails if the landing is not registered. Co-Authored-By: Claude <noreply@anthropic.com> --------- Co-authored-by: Claude <noreply@anthropic.com> |
||
|
Neil
|
3a00c74dbd |
test: remove the redraw wall-time race and shrink oversized scaling populations (#37385)
Deletes terminal-tail-redraw-wall-time: its result equality duplicates the randomized equivalence suite and its complexity claim is pinned by the join-counting scaling suite; it only added a 3x wall-clock race. Shrinks scaling suites whose bounds are exact or log-bounded formulas in n to the smallest population that still separates linear from superlinear work, and pairs a worktree-poll test's scans so it no longer waits out the 5 s scan timeout. |
||
|
Brennan BensonandClaude Opus 5.5
|
f5a36554ff |
fix(plugins): keep the plugin consent fingerprint Hermes-compatible so main's Mobile checks pass (#37370)
* fix(plugins): build the consent fingerprint without toSorted so the phone bundle passes Hermes checks Main's Mobile verify has failed since #37316 because the plugin consent fingerprint (bundled into the phone app via src/shared) used Array.prototype.toSorted, which Hermes lacks. flatMap already returns a fresh array, so sorting it in place keeps the same result. * test(plugins): pin the consent fingerprint's command-order independence Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com> |
||
|
Neil
|
d8a66e41c9 |
fix(wsl): name the stopped distro, not its raw hold code, wherever a held terminal is shown (#37380)
The terminal error banner and a paired host's held pane showed terminal_held_wsl_distro_*:<distro> as is; the banner now swaps either the code (older hosts) or the English sentence (newer hosts) for the localized copy, and drops the issue link for it. A held start that arrives as the sentence now gets the Start card too. The CLI names the distro for an older server's raw hold code and for an unavailable WSL folder. The banner's footer reports the shell whose spawn failed (wsl.exe, say) instead of the client's default shell. |
||
|
Neil
|
4f0aaccc5f | fix(repos): a catalog answer sent before an add no longer drops the project that add made (#37384) | ||
|
Jinjing
|
e6a9ce50e2 |
Preserve bounded tool evidence in chat handoffs (#36842)
* Preserve bounded tool evidence in chat handoffs - Keep recent conversation and useful tool failure context without exporting huge outputs - Safely summarize tool inputs and retain transcript order * Preserve tool call IDs and block order in handoffs * Preserve cross-message tool results in handoff context * Prioritize latest tool results in handoff context * Preserve bounded saved tool failures in handoff context * Avoid reflective tool preview reads |
||
|
Brennan Benson
|
4343dc212a |
fix(native-chat): report missing saved conversations once (#37189)
* fix(native-chat): report missing saved conversations once * fix(native-chat): keep failures of answered Claude sends and translate the missing-conversation row A Claude result settles its own send before the journal sees it, so a failed send with no echo yet lost its owner and its failure row. The result now carries the send it answered. Desktop and phone rows also show the missing saved conversation sentence in the reader's language. * fix(native-chat): keep failures of Claude results that name a user message A root failure result that names a user message answered accepted input, so it keeps its row even when that send's waiter was retired, in doubt or ambiguously written. Only a result with no turn, no started send and no named user message (the pre-init missing-resume capture) is left to the host's final startup row. This replaces the per-frame settlement carry, so the acquisition and sign-in tests return to main. Also ship the new missing-conversation sentence in the runtime English catalog. * fix(native-chat): leave only a starting Claude session's failure to the startup row A Claude failure result with no turn now writes no row only while its own session is still opening (initialize and resume catch-up). Once open succeeds, every failure result is reported as before, whatever send evidence it carries, so per-send ownership checks are removed. * fix(native-chat): fail a starting Claude session with the startup failure it reports A turnless failure result that names no user input or resumed turn, arriving while its own acquisition is still starting, now fails that acquisition's open (initialize and resume catch-up) with Claude's own diagnostic; only then is its provider row left to the host's final startup row. Results naming input, and every result once open succeeds, keep main's reporting. The missing-conversation classifier is shared by stderr and the result's errors, so the typed refusal ends in one attempt whichever arrives first; other reported startup failures stay retryable through the existing provider-request error. * fix(native-chat): keep a Claude startup result's row once another failure ended the start A Claude startup failure result was still taken by its acquisition's startup outcome after initialize, a transport fault or an unexpected exit had already failed that start, so its diagnostic was neither shown nor kept in the final failure. The outcome now declines once the open fails by another route or the exit is observed (synchronously, before buffered frames drain), so such a result keeps its row as on main. It still takes a repeated result it already accepted, and a result whose refusal the ending exit already carries; a Stop still owns its start. * fix(native-chat): report a Claude start's first ending cause from its open The startup outcome now records the first cause that ends a start: an accepted startup result, the raced initialize or catch-up rejecting, a transport fault or unexpected exit, or any other open failure. That cause rejects the pending race, open() refuses success after it, and the open rejects with it, so a hidden result is always the open's error and any other result keeps its row. * test(native-chat): check the startup rejection with a type guard instead of an assertion |
||
|
Neil
|
0fad93e2ad |
test: stop unit tests from waiting out real 5 s fallbacks; fix two flaky tests (#37362)
The ConPTY prelude mock delivered its DA1 query on a microtask, which reached only the physical owner's early onData listener (added in #36422); the startup watch subscribes after an await, never saw it, and every Windows-path spawn waited out CONPTY_STARTUP_DECISION_LIMIT_MS (5 s). Deliver it on a macrotask like real output. Also give two fixtures the readiness their production peers provide, and fix the counter-file race in structured-agent-shell-environment and the async sign-in marker race in claude-status-sign-in. |
||
|
Neil
|
cb031c4517 |
fix(plugins): merge disk writes into a custom editor's unsaved edits (#37343)
* fix(plugins): merge disk writes into a custom editor's unsaved edits A write to the file on disk while a custom editor tab had unsaved edits never reached the editor: the buffer kept the edits, Orca showed its Reload/Overwrite banner, and the session's merge never ran. Now a frame whose document session has merge announces it (editor.merges), and Orca sends it the new disk text with the text the edits started from (editor.document with ancestor). The session merges, posts the result, and the tab stays unsaved with it; the banner is skipped. Without merge the banner decides as before. * test(editor): an explicit save under the changed-on-disk banner keeps the edits * fix(plugins): make the outside merge exact, and narrow it to where it is - The host settles a merge only on the session's edit made against the merge's own document version, and always commits that result to the buffer (even when it equals the disk text); silence past the 1 s flush deadline, a remount or a throwing merge leave the banner and the suspended autosave in place. - Orca stamps exactly the disk snapshot the acknowledged merge used and records it as the next ancestor; a write that landed meanwhile puts the conflict back for its own merge. - The ancestor is the tab's disk text recorded while clean and advanced on acknowledged merges, and must match the tab's disk baseline. - Only a local file in one merging editor view, with no other tab on its buffer, takes this path; anything else gets the banner. - Frame events keep reaching their handlers after the merge capability. * fix(plugins): move nothing on a failed disk merge; advance the saved text on success A merge that throws now leaves the session's document, baseline and version untouched, so the user's next edit goes out against the old version: Orca refuses it and keeps the banner instead of taking it as the merge's acknowledgement. After an acknowledged merge the tab's saved text moves to the disk text it used, so typing back to the old ancestor no longer looks clean. * chore(plugins): regenerate the testing bundle on main |
||
|
Brennan Benson
|
d712485fe2 |
Restore main unit checks for plugin routing and prompt receipts (#37345)
* test: restore plugin routing and prompt receipt checks on main * test: flush pending receipts before releasing the prompt write * test: route screenshot CLI sources to the journeys that run them |
||
|
NeilandClaude
|
5846caf2ba |
plugins: editor.save.before hook (format on save, or refuse the save) (#37347)
* plugins: editor.save.before hook (format on save, or refuse the save) Explicit saves of text files from an editor tab run plugins' editor.save.before in main before the write (local, SSH, WSL and folder workspaces all save through that one renderer path). A plugin can return changed text or deny with a reason, shown in the save error. The buffer takes the hook's text only if it is unchanged since the hook started; edits made meanwhile stay dirty. Autosave, restart saves and bytes documents are not hooked, and with no live plugin listing the hook the save path adds no await. Co-Authored-By: Claude <noreply@anthropic.com> * test(plugins): editor.save.before text reaches the visible Monaco model The hook writes its text into the tab's draft, which reaches the model through the same content-sync bridge as Reload from disk. A model-level test covers the formatted text showing, the next keystroke editing it, and edits typed during the hook staying in the model. Co-Authored-By: Claude <noreply@anthropic.com> * plugins: editor.save.before review fixes: flush held typing, update the surviving view, keep the denial - After the hook answers, pending view text (rich Markdown's serialize timer) is flushed before deciding whether the buffer still holds the hook's input. - The hook's text goes to whichever view of the buffer record is still open, so closing the saving view keeps a sibling view in step with disk. - The save-request bridge passes Errors through whole, so the toast shows the plugin and its reason instead of a generic failure. Co-Authored-By: Claude <noreply@anthropic.com> * chore(plugins): regenerate after rebasing the save hook on main --------- Co-authored-by: Claude <noreply@anthropic.com> |
||
|
Brennan Benson
|
8ddc35b8e9 |
test: import the conversation view from its new location (#37346)
#37191 added an import of the old path after #37289 moved the module, breaking main's typecheck for every PR. |
||
|
Neil
|
60e25bde45 |
feat(plugins): make the editor template the rich-editor recipe (#37349)
orca plugin create --template editor wrote an inline-script HTML editor whose extension came from the folder name squashed to 12 letters. It now writes the recipe: editor/main.ts (the view), editor/sync.ts (the document session with a stand-in text merge), editor/sync.test.ts (a mockEditorHost test of an outside write merged during a save), a tsconfig on the generated typings, frame isolated, engines.orca >=1.4.248, and extension .<id>. It passes plugin check and plugin test as generated. |
||
|
Neil
|
999160b3db |
feat(plugins): commands[].replaces lets a mod take over a built-in action (#37316)
* plugins: commands[].replaces lets a worker command take over a built-in action A worker command with replaces: <action> (the declarative action list, engines >=1.4.248, needs main) runs in place of that built-in from its shortcut, plugin aliases and the sidebar menus; a failed or silent (2 s) worker falls back to Orca's own. The install screen names each replacement, the plugin page has a per-action switch (settings.pluginReplacementsOff), and when two plugins replace one action the one installed first wins and the other's log says so. Update review counts replaces as new access. * fix(plugins): narrow replaced action ids before looking up their titles * plugins: harden command replaces after review - Consent includes the replaced actions, so adding replaces to a dev folder needs a new approval; CLI approval and update review name each takeover. - Shortcuts main intercepts (Tasks, Workspace Board, history), search with a selection and floating maximize now go through the replacement too. - A worker that only starts after the 2 s fallback no longer runs the command as well: main refuses to start it past the deadline. - A workspace-only command does not replace an action without a workspace. - An update keeps the plugin's first install time, so it keeps its takeover. - Drop the commands.run promise until Orca offers it. * plugins: commands.run always runs the built-in, never a replacement A5b's commands.run dispatches with the 'plugin-host' source; the app dispatcher now builds handlers that run Orca's own action for it, so a plugin that replaces an action and calls commands.run cannot loop. Docs and schema describe wrapping the built-in with commands.run; regenerated. |
||
|
Neil
|
3e52963a9a |
docs(plugins): fix custom-editor authoring papercuts from dogfood (#37337)
- shot --act: say edits only dirty the tab; end with Cmd/Ctrl+S to save - manifest: compound suffixes like .sketch.json open by default - cli.md: list plugin check's exit codes, including plugin_check_failed - workflow: how to run and stop the dev app from a source checkout - plugin create --help: real descriptions for --template/--publisher/--id/--name - recipes: declaration file for vendored ESM libraries |
||
|
NeilandClaude
|
2fb3a0975e |
feat(plugins): workers can open files and run built-in actions (#37323)
* feat(plugins): add editor.open and commands.run for full-access workers Both ride the existing plugin view request channel (the one plugin open and plugin shot use) to the one main window's renderer; headless serve and a missing window answer unavailable. editor.open takes a path that stays inside the workspace (Git worktrees and folders) and opens it as a tab without switching what the person sees. commands.run takes only the built-in contributes.commands action list and dispatches Orca's own handler, never a plugin replacement. Both need workspace:write. workspace.list now also lists project-group folder workspaces. * fix(plugins): open folder workspaces and host-naming ids through Orca's window workspace.open of a project-group folder asks the window to show it (setActiveFolderWorkspace), since the runtime activates Git worktrees only. editor.open resolves the identity: ids workspace.list now returns. commands.run documents that it must call the raw built-in, never a replacement wrapper. * plugins: regenerate the testing bundle and skill guides on the rebased host actions Generated files only (lint-staged's oxlint refuses a commit with no lintable files). Co-Authored-By: Claude <noreply@anthropic.com> * chore(plugins): regenerate the testing bundle on main --------- Co-authored-by: Claude <noreply@anthropic.com> |
||
|
NeilandClaude
|
544577fe02 |
feat(plugins): workers can list and open workspaces, create and read terminals, launch agents (#37312)
* feat(plugins): let full-access workers act on workspaces, terminals and agents Host API minor 3 adds worker-only rows: workspace.list (workspace:read), workspace.open (workspace:write), terminal.create (terminal:send), terminal.read (terminal:read, at most 2,000 lines) and agent.launch (agent:launch). Each dispatches the runtime RPC method the orca CLI calls, in process with the CLI's identity, so it runs on the host that owns the workspace (local, WSL or SSH) and covers folder workspaces. The new capabilities arrive in 1.4.248; typings, docs and skill references are generated from the table. * fix(plugins): make workspace host actions safe to revoke, cancel and target - Each runtime effect re-checks the plugin's approval right before the runtime acts (a dispatcher beforeInvoke hook, after its own waits), and a worker's cancel or exit now aborts the call's signal through to the runtime. - agent.launch starts the agent's CLI in a background terminal and returns its handle at once; a prompt is delivered afterwards through the runtime's own reuse-terminal path, so the host call never waits for the composer. - workspace.list adds project-group folder workspaces and returns ids that name their host (identity: keys), so twins on two hosts are never ambiguous. - terminal.create passes presentation background; terminal.read counts a runtime-limited read as truncated. workspace.open refuses folders as unavailable until a window-side open lands. * plugins: keep both API extensions and host-call tracking under the line limits after rebasing on the hooks Worker-only status bar and workspace methods spread into PLUGIN_HOST_API through one module, the in-flight host call map moves into PluginWorkerHostCalls, and bindPluginRuntime builds the runtime RPC so main-process-plugins.ts keeps its size; regenerated docs and typings. Co-Authored-By: Claude <noreply@anthropic.com> * chore(plugins): regenerate after rebasing on the editor recipe and shots * chore(plugins): regenerate after rebasing on the plugin shot recipe --------- Co-authored-by: Claude <noreply@anthropic.com> |
||
|
Brennan BensonandClaude Opus 5.5
|
1783cf58b1 |
Show agent sign-in notices before the first chat message (#37191)
* Show explicit agent sign-out facts before sending a chat * Drop the Grok status process; keep listed models when Pi or Claude replies arrive late Grok goes back to one listing process with no Grok-specific sign-out guess; an agent-authored auth-required reply still reports notSignedIn. Pi reports notSignedIn only when it has no provider and nothing listable. Once Claude's required replies are in, the listing deadline returns them instead of failing. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> * Drop the runner tests' timeout bumps, which sat below the default Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> * Type the endpoint test's launch env as a string map Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> * Drive the signed-out Claude re-probe trigger in the scaling contract N and 10N distinct chats (distinct read claims, one account-scoped Claude fingerprint) read with waitForListing against the real probe and store on an injected clock: one listing before the held-reason TTL, exactly one more once it is due (held until every read has decided), none within the next TTL, each result carrying notSignedIn and models. Checked locally that it bites: making heldReasonDue ignore the TTL fails it (2 listings inside the TTL). Removing the in-flight dedup (the probeDue isListing guard plus the lister's early return) does not: store refresh then runs once per chat, but a newer flight replaces each queued one behind the two admission slots, so still one process per TTL. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> * Accept Claude control replies as main did, without a type field The shared reader required type 'control_response' and prefiltered on that text, dropping success frames main's model and slash-command parsers accepted (the released-phone wire test's fake listing). It now prefilters on request_id and keeps the structure limit and success-only reading. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> * Narrow the scaling contract's catalog result before reading models Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com> |
||
|
Brennan BensonandClaude Opus 5.5
|
06b8f7dc37 |
Keep Windows and Linux chat switching separate from worktree shortcuts (#37199)
* fix(shortcuts): give Windows and Linux chat switching a working default * test(shortcuts): verify the worktree and chat chords independently * fix(shortcuts): preserve agent newline keys while switching chat views * fix(shortcuts): use Ctrl+Shift+F11 for Windows and Linux chat switching The Ctrl+Alt+Shift+J default could not match non-Latin layouts or IME input, because Ctrl+Alt chords skip the physical-key fallback that protects AltGr characters. A function-key primary matches on every layout; Ctrl+Shift+J stays as the alias where Switch worktree doesn't claim it. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> * fix(shortcuts): keep the chat toggle's own default alias when its shortcuts are edited Saving any change to Switch between Chat UI and terminal on Windows and Linux kept its untouched Ctrl+Shift+J alias, which the conflict check then refused against the worktree palette's default, so replacing or removing the primary failed. The check now exempts a chord per binding when it is the action's own default and every other owner also holds it as a default, so non-default explicit claims are still refused. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> * test(shortcuts): expect the new Windows/Linux chat switch label The context menus show the toggle's primary binding, which is now Ctrl+Shift+F11 on Windows and Linux. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com> |
||
|
Brennan Benson
|
5d3dfe943e |
Tasks source summary, repo link and Filters row follow the list's actual source (#37226)
* Show the active GitHub list source in Tasks summary
* Tasks header, repo link and source row follow the list's resolved source
The Tasks summary could name the saved upstream project while the list came
from the fork, and the "Open in GitHub" link could open the other kind's
repository when the active kind's source was unresolved. A new search or an
Issues/PRs switch also blanked the source until the new fetch landed.
Compute the source summary once, after mode and item kind are known, from
the list's resolved sources (saved project identity no longer reaches the
formatter). The summary, the single-project repo link (strict active kind)
and the Filters source row share one display-only read: the current query's
entry, else the latest same-scope sibling while the query is fetchable.
Notices, errors, Retry, filter metadata and page rows still read only the
current query's entry. Projects, GitLab, Linear and Jira output is unchanged.
* Tasks source carries the shown entry across queries, read live
While a new search or Issues/PRs switch has no cached result yet, the Tasks
summary, repository link and Filters source row keep showing the source of
the entry they last showed for that project copy, read live from the cache,
instead of picking the newest cached sibling. A refresh that changes that
entry is followed, other queries' results never take over, a preference
change's eviction ends the claim, and a present unresolved result or a
too-large query makes no claim.
The display read now happens in the existing current-entry selection pass
(at most two cache reads per selected project, no cache scan), and the
summary re-translates when the language changes. Adds a SourceBar fixture
with a long fork slug at normal and narrow widths.
* Tasks source display perf contract measures each update on its own
Bound every measured update (another host's patch, a non-cache store write,
current and sibling fetches, a real preference eviction) at 10 and 100
selected rows and 50 or 500 cache entries: no cache enumeration, at most two
reads per row per selector call, a fixed selector-call budget equal across
sizes, no key rebuild on cache writes, linear row visits including pointer
upkeep, and reused output references for writes that change nothing shown.
* Use typed reads in the Tasks source display perf instrumentation
The static-analysis rule rejects Reflect.get; the cache and row proxies now
read by typed keys and count exactly the same reads and enumerations.
* Drop type assertions from the Tasks source display changes
Stage hooks return Object.assign(model, {...}) instead of casting the model
(same object, same writes, same type). The fixture and test stubs keep their
partial-model casts on one line each with a SAFETY reason, the row proxy uses
a type guard, and the filler work item is a typed fixture.
|
||
|
Neil
|
3b5eed8ca4 | docs(plugins): point the editor and status bar recipes at plugin shot (#37332) | ||
|
Brennan Benson
|
9ce35be4ff |
fix(native-chat): effort, speed and fast controls stay usable while the model list refreshes; dotted versions in id-derived model labels (#36870)
* fix(native-chat): running chats keep effort menus during model discovery; dotted versions in id-derived model labels
* fix(native-chat): options stay usable while a model list refreshes; dated preview ids keep their dashes
A running host listing no longer disables effort, speed or fast mode. Those
controls follow the data they draw from: before any list exists they show
the placeholder or a read-only saved effort, and once a list or the chat's
own answer is shown they stay usable through every refresh, for every agent.
The listing flag is now only the host's raw fact, which keeps a seed from
settling to the built-in list while discovery runs.
Id-derived labels no longer dot a zero-led part, so MM-DD preview dates
(gemini-2.5-flash-preview-05-20) stay dates.
* fix(native-chat): a model list refresh keeps the verified models and default effort already shown
While the host refreshes a Claude or Codex list, the chat keeps the model
ids an earlier finished listing verified that the refreshing rows still
contain, so a running Codex chat's CLI-default effort ("Medium") no longer
blanks from the pill and menu until the listing lands. Ids no finished
listing verified stay unverified, and a finished listing still replaces
the set.
* test(native-chat): type the listing-wait check without an assertion
* fix(model-labels): month-day dates after October stay undotted
* refactor(native-chat): listing-verified model ids get their own module
|
||
|
Brennan Benson
|
0bf96e0d7a |
Show a GitLab MR's pipeline result even when it has no jobs (#37177)
* Show failed pipelines and check suites without jobs * Use the parent pipeline as the GitLab checks verdict The MR's own pipeline status now leads the Source Control Checks panel, the MR dialog and the phone checks badge whenever the details reply carries it, even when the returned jobs are empty, partial or all passing. Job rows and counts stay a separate breakdown. The parent follows pipeline-level policy (required manual is pending, canceled and skipped are not failures, mergeable manual is neutral) from one shared rule, and it is a provider link rather than an expandable job, so it never shows a details error with Retry. A failed parent leads the triage strip and joins the Fix prompt by name and link. Drops the GitHub zero-run failed check-suite rows from this change; existing approval-required suite rows are unchanged. * Name the GitLab pipeline outcome and keep its verdict in one rule The host now settles the parent pipeline's verdict with the same rule and reply that produce the MR status, so a manual pipeline on older GitLab reads pending everywhere. The phone checks badge names the pipeline outcome (Pipeline passed/failed/pending) instead of an unlabeled one-item count; job wording without a parent is unchanged. The desktop parent row's status is localized, and a failed MR-details read still shows its error beside a kept parent with no jobs. * Keep Fix for failed GitLab jobs and treat waiting_for_callback as pending The triage strip shows again when a returned job failed under a passing parent pipeline, so its Fix stays available; a failed parent still leads. A waiting_for_callback pipeline now reads pending instead of neutral, in the MR status and the parent pipeline alike. * Mark the checks-panel action test fixture cast as safe * Pause the pipeline row spinner offscreen and under reduced motion |