Compare commits

..
34 Commits
Author SHA1 Message Date
Mohamed Boudra b2ca56ee73 chore(release): cut 0.17.2 2026-11-22 12:46:24 +01:00
Mohamed Boudra 59f19e4e5e docs(changelog): add 0.17.2 release notes 2026-11-22 12:38:49 +01:00
Mohamed Boudra 884adbcc2f Archive the open workspace from the mobile workspace actions menu (#10507)
* feat(app): add Archive workspace to the mobile workspace actions menu

The compact header's actions menu had no way to archive the open workspace; it
was only reachable from the sidebar row or the keyboard shortcut. The route-keyed
archive logic behind the shortcut now lives in useActiveWorkspaceArchive so the
menu and the shortcut share one path (confirmation, undo toast, redirect).

* fix(app): redirect after archive from where you are now, not where the header was

The archive starts after the host check and confirmation. The header that began
it can be gone by then, and its last-rendered route would still match the
archived workspace, replacing the workspace you had since opened. Read the
current selection from the navigation owner when the archive starts.

* fix(app): skip the post-archive redirect when you have left workspace routes

The last-remembered workspace selection survives opening Settings, so an archive
that finished while Settings was open still redirected away from it. Read the
focused route instead, which is null on any non-workspace screen.
2026-11-22 18:25:55 +07:00
Mohamed Boudra 91247c8218 fix(app): keep turn timing inside the footer (#10510)
The Turn stats button narrows the timing column. On web, preserved spaces at soft wraps extend the timing text range past the button. Collapse wrapping spaces for the timing labels and include Turn stats in the existing hover stability check.
2026-11-22 18:01:18 +07:00
Mohamed Boudra 2d09c47672 Show provider session and turn metrics (#10496)
* feat(providers): let providers report session and turn metrics

Providers know things about a session and a turn that Paseo cannot derive:
timing, token flow, cache use, cost, the model that answered. They can now
report them as label/value rows that Paseo carries to clients unchanged, so
each provider decides which metrics make sense and how they read.

Session metrics live on the agent record and snapshot. Turn metrics are
stamped on the turn's footer row (last assistant message, else last tool
call) so they survive reload, pagination and a daemon restart, and so the
live turn_completed event and a later timeline fetch agree. Provider history
replay restores both.

Metrics are sanitized once where they enter the agent manager. The wire
schemas validate whole messages, so one malformed row from a provider would
otherwise make clients drop an entire snapshot, turn event or timeline page.

A live completed turn with no Model metric gets the agent's model, so every
provider shows model attribution before it is instrumented. Replayed turns do
not: they may have run on a different model than the agent has now.

Plugin providers publish session metrics behind a negotiated
`session.metrics` capability because an older daemon closes a connection
that sends an event type it does not know. Turn metrics ride an existing
event that older daemons strip, so they need no gate.

The wire additions are optional fields on existing schemas only.

* feat(providers): report session and turn metrics from OpenCode, OMP, Pi and Claude

Each provider now attaches metrics to turn_completed and emits
session_metrics_updated, live and when history is replayed.

- Live turns end on the adapter's clock. A provider's message timestamps
  can come from another clock.
- Tool time merges overlapping runs, so parallel tools count once.
- A total that is missing any part is hidden, so a partial total is
  never shown.
- Replayed history reports metrics only for turns that finished, which
  matches live, where failed and interrupted turns report none.
- Claude reports cost as a running total for each query(). A resumed
  query starts that total from a saved one, so a turn's cost is the
  change since the previous result of the same query. The first turn of
  a query shows no cost.
- Claude transcripts do not record result messages, so replayed Claude
  history has no turn metrics.
- Session metrics come from each provider's own cumulative counters. The
  cache hit is the last reply's or the last turn's, because a session-wide
  figure blends prompts from before and after compactions.

* feat(app): show provider session and turn metrics

Session metrics appear in the context window details, between the context summary and the
account usage cards. A provider-reported Cost row replaces the app's own session cost line.

Each completed turn with metrics gets a stats button in its footer: a hover card on web, a
press-opened tooltip on native wide screens, and a sheet on compact screens. The context window
sheet becomes a shared DetailsSheet so both use the same content-sized sheet.

Metrics ride on the row that hosts the turn's footer (the turn's last assistant message, else its
last tool call), the same row the daemon stamps. Live turn_completed events and timeline pages
both stamp that row, so a refetch agrees with what the live event showed.

The daemon updates session metrics without bumping updatedAt, like usage, so a stale buffered
upsert may still carry the newest metrics and is accepted for that field.

* fix(app): keep a turn's stats after the cached timeline reloads

The local timeline cache dropped turnMetrics when it saved a turn's footer row. After a page
reload the cached rows painted without the stats icon, and the catch-up fetch had no new
entries to stamp them from.

* fix(providers): keep imported session stats and never count a steered turn short

Importing a Pi, Oh My Pi or OpenCode v2 session dropped the stats the provider returned with its
history, so imported chats showed none until a forced replay. Import now carries them the way
replay does.

Replaying a Pi or Oh My Pi transcript read a prompt that steered a running turn as the start of a
new turn, so the steered turn's stats counted only the steps after the steer. A turn that follows
one ending on tools now reports no total, because the transcript reads the same as a stop and a
new prompt.

* fix(providers): keep stats through merged messages and plugin imports, start Pi first-token timing at the request

A message the timeline merges from several chunks keeps the stats stamped on its last chunk. Plugin session imports carry the stats the plugin replayed, through the same history reader the other providers use. Steers queued together after an unclear turn stay unclear instead of attributing the final reply to the wrong turn. Pi measures time to first token from the request, not the first reply header.

* fix(providers): carry a turn's stats onto the message the timeline merges from text chunks

The previous change put the carry-over in the reasoning merge. The assistant message merge dropped the stats of its last chunk, so an imported reply of two text blocks lost its Turn stats button.

* fix(providers): carry a turn's stats onto the tool card the timeline merges from updates

An imported turn that ends on a tool call stamps the stats on the tool's last update. The tool merge kept the first update's fields, so the stats were lost.

* fix(opencode): replay no session stats for a session that has none

An empty OpenCode session replayed a session stats event with no metrics, which broke the empty-history check in the versioned runtime test.
2026-11-22 16:43:56 +07:00
Mohamed Boudra 52377f595a Enable sidebar usage summary by default and move source toggles to a dedicated Sources list (#10491)
* feat(usage): enable sidebar usage summary by default and add dedicated Sources list

Show the sidebar Usage summary by default instead of requiring opt-in in
Appearance settings, while honoring explicit off preferences and continuing
to hide the row when no enabled source has pinned windows.

Let usage sources declare optional provider hints (`providers` on the
registration and `provider` on discovered accounts) so the daemon can mark
reports `defaultEnabled: false` when their hinted providers are neither
enabled nor installed, keeping usage sources decoupled from agent providers
while avoiding clutter from leftover credentials.

Move per-source show/hide controls out of individual card headers into a
dedicated `Sources` submenu inside Usage Settings (with tri-state
`cardVisibility` preferences migrating legacy `cardsOff`) so disabled
sources stay out of the main card list.

* fix(usage): resolve provider hints per report, refresh on forceRefresh, and clarify inactive copy

Resolve each usage report's provider activity hint in parallel with its
fetch and enforce a real timeout deadline in `isProviderActiveForUsage` so
a slow provider probe never holds back unrelated reports or hangs past the
deadline on non-abortable plugin RPCs.

Give the availability hint a 5-minute freshness TTL and re-probe when
`forceRefresh` is requested, retaining the fresher probe result across
subsequent normal reads until a newer snapshot or TTL expiry supersedes it.

Rename the default-off Sources sublabel from "Provider not installed" to
"Provider inactive" across all locales so it accurately covers both
disabled and unavailable providers.

* fix(usage): share provider availability reading across snapshot publishes and usage probes

Record both `ready` and `unavailable` global snapshot publications into
`usageActivityCache` so `isProviderActiveForUsage` reads one timestamped
availability cache instead of relying on `fetchedAt` (which unavailable
snapshot entries do not carry).

* fix(app): add MenuCheckboxItem so Usage Sources toggles expose checked state to screen readers

Introduce `MenuCheckboxItem` / `DropdownMenuCheckboxItem` backed by
`accessibilityRole="checkbox"` so toggleable menu rows expose their checked
state in Chromium's accessibility tree and native screen readers while
leaving existing `MenuItem` roles unchanged.

* fix(app): keep compact menu sheet rows in Tab order and activate on Space

Use `tabIndex={0}` for `MenuRow` in `sheet` presentation (leaving
popover roving arrow-key navigation at `tabIndex={-1}`) and handle Space
in web sheet rows since React Native Web only activates Space for
`role="button"`, not `menuitem` or `checkbox`.

* test(usage): include provider hint in claude-usage-source keychain and stores assertions
2026-11-22 16:43:11 +07:00
Mohamed Boudra 9cb521dda2 perf(ci): speed up server tests, Playwright shards, desktop/website E2E, and apt cache (#10504)
* perf(ci): cut PR CI wall-clock time across server tests, Playwright shards, and workflow setup

* fix(ci): guard empty apt cache copy, use 100% Vitest workers in CI, revert Playwright to 2 workers with built supervisor
2026-11-22 16:42:52 +07:00
Mohamed Boudra 234ddf8dd7 Simplify setup across supported provider docs (#10502)
* docs(providers): simplify setup across supported provider pages

Lead each supported provider page with the official CLI/protocol Paseo uses and the simplest default setup path, removing config.json snippets from Getting started on Antigravity and Oh My Pi. Fixes the outdated Settings path on Codex and keeps all Chinese translations in sync.

* docs(providers): address review feedback on Antigravity, Codex, Muse, and Pi
2026-11-22 15:22:12 +07:00
Mohamed Boudra 3914893aa1 Troubleshoot providers from their settings page, and tidy provider settings (#10487)
* feat(providers): add docs links, sign-in facts and a tidier provider page

A user whose Antigravity models looked stale had no way to find out why
without asking in chat. agy 1.5.14 picks its sign-in from the
environment: with SSH_CONNECTION, SSH_CLIENT or SSH_TTY set it reads a
token file, otherwise the OS keyring
(google-antigravity/antigravity-cli#1336). The daemon is often started
outside SSH while the user tests agy over SSH, so the two can be signed
in to different accounts and list different models.

- The Antigravity status reports whether agy runs in an SSH session, and
  the diagnostic report for plugin providers now includes the resolved
  command, path and version like built-in providers do.
- Providers carry an optional docsUrl (protocol field, plugin SDK
  registration, built-in definitions). The provider page links it from
  Troubleshooting, and the Antigravity page documents the sign-in
  mismatch.
- Provider and project settings get a breadcrumb title. Refresh moves to
  the Models header, the model list scrolls inside a fixed height with a
  full-width search, and section headers keep one spacing whether or not
  they hold an action.
- A switch no longer makes its row taller than the track, and every
  settings row is at least as tall as one with a switch, so a switch row
  and a text row line up.

* fix(providers): report Antigravity's sign-in when ACP does not start

A sign-in problem can be what stops agy's ACP server from starting, so
the diagnostic report now shows the SSH session rows in that case too.

The plugin provider diagnostic test no longer compares the daemon PATH
and shell rows, which differ per host (Windows reports ComSpec).

* fix(server): show the same provider version in the report and Installation

The plugin diagnostic read the version by calling the provider's
inspect itself with a 10s deadline, while the Installation row reads it
through the installation service with a 20s deadline and a cache. A
version check between 10 and 20 seconds showed in Installation and as
an error in the report.

The snapshot manager now hands getDiagnostic an inspectInstallation
that reads through the installation service, so both show one answer
and agy --version runs once for both.

* feat(app): give rows that open something the same height as other rows

Detail rows and collapsible row headers reserved a button's height so a
list where only some rows held a switch kept one height. Now that every
settings row is at least as tall as a switch row, that reserve only made
rows like Diagnostic report and Docs taller than their neighbours.

* test(app): compare rows that open something with text and switch rows

* feat(app): size the provider icon to the page title

* fix(app): scale the provider title icon with the interface size
2026-11-22 15:13:31 +07:00
Mohamed BoudraandClaude ede181d39a Refresh enabled provider catalogues in the background every 30 minutes (#10481)
* Refresh enabled provider catalogues in the background every 30 minutes

A long-running daemon cached discovered models until a manual Refresh or a
restart, so a server-side model rollout stayed invisible. The daemon now
refreshes each enabled provider about every 30 minutes, one provider per
staggered slot. A recent refresh pushes the next one back. Background
refreshes join discovery already in flight, publish only changed
catalogues, and keep a ready catalogue when a probe fails.

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

* Keep background refresh slots independent and its probes soft

Ticks no longer wait for a refresh to settle, so one slow provider cannot
delay the others. A provider with a background refresh still running is
skipped. The due check uses a fixed one-second timer slack instead of half
a slot, so a manual refresh always pushes the next probe a full interval.
Background probes pass force: false so OpenCode keeps its shared server.
A result that is not ready keeps a ready catalogue, which covers plugin
status probes that report a failed startup as unavailable.

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

* Keep a ready catalogue when a background refresh returns a partial one

Claude falls back to its curated list without discovery-gated models when
model discovery fails or times out. That result is ready, so a background
refresh published it and Mythos left the picker. Providers now mark such a
catalogue incomplete, and a background refresh keeps the ready catalogue it
would have replaced. A manual refresh still publishes it.

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

* Kill OpenCode servers whose startup is pending at shutdown

A probe still allocating its port when the daemon stopped spawned its
OpenCode server after shutdown returned, and the process outlived the
daemon. Background refresh makes that overlap likelier. Shutdown now waits
for pending startups and kills the generations they produce.

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

* Stop running probes before provider shutdown

A probe checking availability when the daemon stopped could still go on to
discovery and start an OpenCode server after provider cleanup returned.
The manager now aborts every running probe through its refresh signal when
it stops, and shutdown waits up to ten seconds for them to settle before it
shuts down provider clients.

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

* Refuse OpenCode spawns from operations begun before shutdown

A rotation accepted before shutdown waited for the old server to stop and
then started a replacement after cleanup returned. The server manager now
counts shutdowns; every operation that can start a server records the
epoch it began in, and the spawn step refuses once a shutdown has passed.
Operations begun after shutdown still start, since the manager is a
process-wide singleton.

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

* Keep manual refresh precedence over concurrent background refreshes

* Avoid eager provider snapshot downloads for unobserved directories

---------

Co-authored-by: Claude <noreply@anthropic.com>
2026-11-22 14:50:02 +07:00
Mohamed Boudra c28aa2570c docs(website): refresh sponsor page visitor and Discord member counts (#10498)
Homepage and docs visitors use the site-wide Plausible figure for the 30 complete days to 2026-11-21 (107,693), rounded down to 100,000. Discord members come from the live guild count (2,678). The README figure is unchanged because GitHub only exposes 14 days of traffic.
2026-11-22 14:29:02 +07:00
Mohamed Boudra 08ad7c7271 Let agents steer or interrupt another agent's running turn, and tell them how their message was delivered (#10483)
* feat(server): let agents queue, steer, or interrupt messages to other agents

An agent sender's `steer` used to mean "wait for the turn to end", unlike a
person's steer. Agents now get the same three choices a person has, plus a
named default: `queue` waits for the turn end, `steer` joins the running turn
(interrupting when the provider cannot steer), and `interrupt` replaces it.
A steer or interrupt carries the messages queued ahead of it, so delivery
order still equals admission order.

The receipt returns the applied behavior and guidance built from whether the
recipient is running and whether its provider can steer. A queued message's
guidance tells the sender to follow up with a short steer or interrupt
instead of resending, because a resent message is read twice.

Older clients send `steer` for every default agent send, so the daemon maps
their agent `steer` to `queue` when the client lacks the
`agent_active_turn_behaviors` capability. New clients talking to an older
daemon refuse `steer` rather than silently queueing it.

* fix(server): keep a steer's carried messages when the turn ends first

A steer that reached the provider after the turn ended made the manager
throw a plain error, and delivery failed the whole batch, including the
queued messages the steer carried. The manager now throws
AgentTurnChangedError, and delivery sends the batch again, into the next
turn or as a new one.

Also report the applied activeTurnBehavior in `paseo send --no-wait`
output, and expect the new hello capability in the client test.

* docs(skills): warn that an unread steer loses the messages it carried

* docs(skills): shorten the send_agent_prompt guidance

Keep what a sender acts on: leave activeTurnBehavior out, the three
values, read the guidance, follow up instead of resending, and the
steer loss case. Drop delivery facts the sender cannot act on.

* fix(server): stop the send_agent_prompt description promising an unread steer survives
2026-11-22 14:01:15 +07:00
Mohamed Boudra b2ecec8719 fix(opencode): stop adding a "This chat uses OpenCode v2" row to chats (#10490)
The notice flagged chats moved to the then-new v2 runtime. v2 has been
stable, so the row is noise. Previously saved openCodeRuntimeNotices
metadata is left in place and no longer read.
2026-11-22 13:44:56 +07:00
paseo-bot[bot] fa37fd20ca Make the recursive watcher inventory test deterministic (#10462)
* test(server): make recursive watcher inventory test deterministic

Deliver the file creation notification explicitly so the inventory check does not depend on the native watcher becoming ready. Hold scheduled scans until the test checks the inventory and require reconciliation to report the deletion.

* test(server): wait for the watcher callback without advancing audits

---------

Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
2026-11-22 06:07:40 +00:00
ff5a83bc5a Keep stale slash menu choices from archiving agents (#10484)
* fix(app): ignore slash menu picks made before the menu filters

On web the cursor reaches the autocomplete after paint. A Tab or Enter
pressed right after typing could pick from the unfiltered list, where
/exit is the default highlight, and archive the agent. Skip the pick when
the typed command no longer matches the query the menu was built for.

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

* fix(app): limit the stale slash menu guard to key presses

Clicks also pass the live input snapshot, so the guard could drop a
click on a visible option while the cursor update was still pending.
A click picks what the user sees. Check for a stale menu only when Tab
or Enter selects.

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

---------

Co-authored-by: 3ae3ae <ogyshjsm@naver.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
Co-authored-by: Bae Seongmin <83870598+3ae3ae@users.noreply.github.com>
2026-11-22 05:50:48 +00:00
Mohamed Boudra 9cb0625cf5 Start an edited New workspace after a failed creation, and unify composer errors (#10482)
* fix(app): allow edited New workspace retries and unify composer errors

New workspace and the workspace setup dialog reused the mount's raw draft
or creation ID as the workspace.create idempotencyKey. When workspace
provisioning failed before readiness, the daemon kept that key's request
fingerprint on disk, so editing isolation, project, branch, prompt, or
model on the same screen failed with workspace_request_key_conflict.

Deriving the workspace idempotencyKey with draftCreationKey(id, request)
matches #9534 for agent drafts: unchanged retries (double-click,
reconnect, or resending as-is) keep one key, while edited retries start a
new workspace creation without requiring daemon or protocol changes.

Also move composer submission errors into one <Alert variant="error"
size="sm" /> slot owned by <Composer> (directly above the input, and in
ComposerFullscreenFooter when fullscreen) so New workspace, Workspace
draft tab, Workspace setup dialog, and active chat share the composer's
outer and inner rails without duplicate error toasts.

* fix(app): include launch target in creation key and toast background failures

Include launchTarget when deriving the New workspace creation key so
switching between blank Chat and a promptless terminal profile after a
pre-ready failure starts a fresh creation instead of replaying the failed
receipt.

Also show toast.error when workspace creation fails after the user has
left New workspace or closed WorkspaceSetupDialog, while keeping the
single composer alert with no duplicate toast when the form is still
visible.

* fix(app): avoid duplicate toast when handed-off agent startup fails

Once New workspace hands a ready workspace to its draft tab, the draft
tab awaits agentCreation.result and displays startup errors in its own
composer alert. Returning early after handoff prevents New workspace's
catch handler from firing a duplicate app-wide toast.
2026-11-22 12:28:10 +07:00
Mohamed Boudra 03adc4d4ba Keep working host connections from flapping on busy heartbeats and back off background probes (#10395)
* fix(app): stop busy-heartbeat adaptive switches and back off online probes

Hosts with multiple saved connections still replaced healthy relay links
during active use: every 120 seconds the runtime compared an idle alternate
probe against the single latest heartbeat RTT on the busy active socket
using a fixed 40 ms threshold, then tore down the live socket before
connecting the replacement (leaving the host in connecting for ~2 s).

Use the 60-second rolling minimum of active heartbeats as the baseline,
require at least max(40 ms, 35% of baseline RTT) improvement to switch,
exponentially back off inactive online probes from 2m to 30m while the
active connection keeps winning (resetting on candidate win, reconnect,
connection list changes, or foreground verify), and connect the replacement
before disposing the live client on adaptive switches.

Closes #9149

* fix(app): preserve active listener, recheck candidate, and refresh caches on make-before-break upgrade

Keep installed active-client status listeners bound to `this.activeClient`
rather than `switchRequestVersion` so a pending or failed speculative
upgrade never mutes disconnect events on the active socket. Recheck that
the target connection still exists and matches before and after disposing
the old client so removing or editing an endpoint during handshake aborts
the upgrade. Track the last online `connectionEpoch` in `HostRuntimeStore`
so make-before-break upgrades that stay `online` still trigger reconnect
cache invalidation and history sync generation bumps.
2026-11-22 11:27:27 +07:00
Mohamed Boudra 4b429c0030 Prevent zombie processes after plugin startup and reload (#10470)
* fix(server): prevent zombie children after plugin builds

Retiring a compiler thread closed its libuv loop while esbuild still owned a native child. Run the compiler in a subprocess so the daemon can reap it when the idle compiler disconnects.

* test(ci): check plugin compiler reaping on macOS
2026-11-22 11:11:59 +07:00
Mohamed Boudra 20f217b8cb fix(acp): finish Antigravity turns after recurring schedules (#10469)
Antigravity returns end_turn while a recurring schedule stays in progress. Exclude cron registration from the provider's deferred-turn blockers so waiting for a tool completion cannot strand the foreground turn.
2026-11-22 11:11:55 +07:00
Mohamed Boudra 984e2ea5e7 Lead the 0.17.1 changelog with Antigravity and keep instructions out of release notes (#10466)
* docs(changelog): lead 0.17.1 with Antigravity and cut instructions from the entry

The 0.17.1 entry ranked Antigravity seventh and carried setup steps, click
paths, migration notes, and edge-case qualifiers. Each highlight is now one
sentence that announces the change, and Antigravity leads. Two details that
left their highlight keep their own bullet: plugin worktree management moves
to Plugins and green quota bars move to Changed.

docs/release.md allowed highlights up to three sentences and exempted them
from the conciseness rules, which is how the instructions got in. The rules
now cover highlights too, permit a second sentence only for a breaking
requirement, and put the headline change first whatever its ordering group.

* docs(changelog): keep shortened 0.17.1 bullets within what shipped

Three bullets lost a clause that limited their claim: provider updates exist
only for CLIs with an update command, the scroll fix leaves out HTML previews
on iOS, and Git status reloads when files changed during the disconnect. Each
bullet now names what it covers instead of carrying an exception clause, and
the release guide says to shorten that way.

The Antigravity requirement links the setup page, which has the `acp`
argument a custom command needs.

* docs(changelog): scope the agent message and quota forecast highlights

An explicit interrupt from another agent still replaces the running turn, and
Usage shows a run-out time only for a quota on pace to run out before it
resets. Both highlights now say so in one sentence, and the release guide uses
the agent message highlight as its example of keeping that distinction.

Rewind in Command Center applies to the latest turn, so the bullet names it
again.

* docs(antigravity): explain the Hub model error after the CLI upgrade

The 0.17.1 highlight used to carry this recovery step. The highlight now links
the setup page, so the step lives there in both languages.
2026-11-22 10:14:31 +07:00
paseo-ai[bot] 133b078ac4 fix: update lockfile signatures and Nix hash [skip ci] 2026-11-21 22:46:04 +00:00
Mohamed Boudra 6ea095cb49 chore(release): reconcile v0.17.1 (#10457) 2026-11-22 05:36:59 +07:00
Mohamed Boudra f7e2ab4076 Keep the full history when reopening a long Antigravity agent after a restart (#10428)
* fix(acp): keep the whole history when an agent streams it after session/load

Reopening a long Antigravity agent after a daemon restart lost the end of its
timeline. When the client reads slowly, agy answers session/load part-way
through the history and keeps streaming the rest, with no end marker. The
replay window closed after a fixed five quiet periods (1 s), and every history
update processed after that was discarded as stale.

The window now closes only once a quiet period passes without a history
update. Command, configuration, mode and usage updates no longer count as
activity, so they cannot hold the session open. When the quiet period expires,
the loop reads pending input once before the silence is trusted, because a
stalled event loop reaches the timer with history still unread in the pipe.

The wait has no upper bound: any bound is another way to truncate, and the
agent goes silent once its history is sent.

* fix(acp): stop waiting for history replay when the connection closes

With the replay wait no longer capped, closing the provider connection while a
session was still replaying waited for the whole history, so a daemon stop
during a long reopen could run past its deadline. The connection marks itself
closed and then awaits in-flight opens, but the replay wait only checked the
runtime's own closing flag. The wait now also ends when the connection closes.

* fix(acp): only extend replay on history updates and fail open on mid-replay crash

* fix(server): bound shutdown flush so provider shutdown cancels in-flight reopens

* fix(acp): keep load replay open while a multi-read wire record is in flight
2026-11-22 04:01:42 +07:00
paseo-bot[bot] 2445adecdf fix(app): finish custom model saves when the daemon confirms them (#10432)
The daemon already schedules catalog discovery when provider configuration changes. Waiting on a second discovery kept the Add sheet open after the model was saved.

Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
2026-11-21 18:33:44 +00:00
paseo-bot[bot] 0c02cfdfb1 fix(app): keep the composer focus hint from covering the placeholder (#6200)
* fix(app): keep the composer focus hint off the placeholder

The web composer draws its focus shortcut hint over the input's top-right
corner. When the placeholder and the hint did not fit on one line, as in
French at a phone width or with the longer desktop placeholder, the hint
was drawn on top of the end of the placeholder. The input now reserves
the hint's width as right padding while the hint shows, so the
placeholder wraps before it.

Closes #6191

* test(app): wait for the focus hint padding before checking overlap

* Keep focus hint padding in sync with composer layout

Reveal the shortcut after the input reserves space for its label. Remeasure draft height when padding changes the available text width.

---------

Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
2026-11-21 15:56:48 +00:00
paseo-ai[bot] 2098eb787c fix: update lockfile signatures and Nix hash [skip ci] 2026-11-21 13:55:05 +00:00
Mohamed Boudra 9c4106561a Refresh the Paseo Hub website (#10309)
* feat(website): refresh Paseo Hub page and add homepage Hub entry points

* chore(website): align Hub automation source copy and team visibility wording

* chore(website): simplify homepage Hub flow layout and remove highlight cards

* feat(website): unify page frame, rails, section rhythm, and hairline grids across home and /hub

* chore(website): remove forced eyebrow numbering and redundant section eyebrows

* feat(website): use Plus Jakarta Sans for display titles

* style(website): remove hairline rails from header and hero bands

* feat(website): derive landing page plugin count from published registry

* feat(website): replace Hub Discord demo with animated Slack mockup

* style(website): tighten display-scale Paseo Hub lockup gap

* website: interactive source showcase on /hub automations section

* website: prevent font-swap wrap jank and remove left bar on /hub source pills

* fix(website): align Hub examples and docs with service behavior

* fix(website): keep navigation and plugin carousels within the page frame

* fix(website): address Hub refresh review findings

* fix(website): preserve focus and labels in Hub showcase tabs

* fix(website): show the selected Hub preview immediately

* fix(website): keep Hub pricing tooltips inside their cards

* fix(website): wrap Hub source labels with larger text

* fix(website): scale Hub preview height with browser text

* fix(website): wrap long identifiers in Hub previews

* fix(website): wrap result chips in the Hub schedule preview

* fix(website): keep the Discord preview thread title readable

* fix(website): let Hub source descriptions wrap

* feat(website): link Hub previews to their trigger guides

Keep the self-hosting guides aligned with the Hub documentation already on main.

* fix(website): keep Hub previews readable at large font sizes

Let preview rows stack when the frame is too narrow for the avatar, padding, and text. Check the rendered text bounds at twice the default font size.
2026-11-21 20:47:56 +07:00
Mohamed Boudra c1c29fcb56 Deliver queued agent messages in order, so an interrupt never leaves a stale steer behind (#10404)
* fix(server): deliver an agent's messages in the order they were queued

An agent's interrupt was dispatched alone, ahead of the steers queued before
it. Those steers stayed queued and started a turn of their own after the
interrupt's turn ended, so a stale instruction ran after the newer message
that countermanded it.

The queue now keeps one ordered list per recipient and only ever dispatches
its front. An interrupt releases at admission and takes the older messages
along in the same replacing prompt. A person's prompt takes the queue along
the same way. Stop holds the queue for the recipient's next prompt instead
of releasing it into a new turn, and does not drop it: the queue also
carries finish notifications.

A quiet window now closes after ten seconds at most, so steady arrivals
cannot delay a waiting message without bound.

* fix(server): keep a prompt's place in line until the provider admits it

The queue removed a message when it handed the prompt to the agent
manager, before the provider had admitted it. In that gap the order was
no longer the queue's to decide:

- a person's prompt or a later interrupt sent while the provider was
  still starting a turn could pass a prompt that was admitted first;
- Stop during that start no longer saw the interrupt waiting to replace
  the turn, so the interrupt started a turn after the Stop;
- queued messages that rode a person's prompt were settled when the
  prompt was handed over, so a turn the provider rejected dropped them.

A message now stays in its recipient's list until the provider admits
the prompt that carries it, fails it, or the message is withdrawn. Every
prompt takes its place at admission, one operation acts on a recipient
at a time, and each operation covers only what was admitted before it
got in line. Stop holds every message still in the list and clears the
interrupt's urgency, so a held interrupt waits like a steer.

A prompt is a command only when it starts with a slash and a name with
no second slash, matching how providers read it. A prompt that begins
with a path now carries the queue like any other text. An interrupt the
provider runs as a command beside the turn takes no queued message
along, because it replaces nothing.

Test wiring that builds the MCP catalog or stubs the agent manager now
supplies the delivery module and the stop signal the server requires.

* fix(server): keep queue decisions made before a Stop or shutdown from going stale

Three races in the window where the provider is still starting or
stopping a turn:

- An interrupt admitted while a polling queue step was already in line
  got no step of its own. A person's steer sent next took the interrupt
  along and steered the turn the interrupt should have replaced. An
  interrupt now always reserves its place at admission.
- A dispatch in flight when Stop arrived checked the queue's phase to
  learn it was void. A message admitted before the provider answered
  moved the queue out of `held`, so the old dispatch became valid again:
  an interrupt from before the Stop started a turn alone, and a batch
  whose starting turn Stop canceled was rejected instead of kept. Each
  dispatch now records the Stops its recipient had seen when it was
  decided and is void once that count changes.
- Shutdown drained only what was queued when it began, so messages a
  rejected person's turn returned afterwards were left to the retry
  timer. The drain now runs after the operation in flight and reads the
  queue then.

* fix(server): void queue steps that were reserved before a Stop

A queue step that was already in line when Stop arrived checked for a
Stop only from the moment it ran. When an interrupt arrived while the
provider was still stopping, the queue left `held`, and the old step
started a turn with the held messages. The new interrupt then replaced
that turn in a second one.

A step now watches for a Stop from the moment it reserves its place and
does nothing once one has arrived. The held messages go with the next
prompt admitted after the Stop, in one turn.

The interleaving test gains a variant with a provider that is slow to
answer. It checks, for every five-step sequence, that no queue dispatch
after a Stop carries only messages from before it.

* fix(server): hold the queue again when the message that reopened it is withdrawn

After a Stop, a message admitted later reopens the queue and takes the
held messages along. When that message was withdrawn before its quiet
window closed, for example a change-request update whose watch was
cancelled, the window still fired and the held messages started a turn
on their own.

A Stop now marks the messages it finds. When a withdrawal leaves only
marked messages, the queue is held again and waits for the next prompt.

* docs(server): state that cancel_agent with no turn to stop holds nothing

Stop holds the queue only when it stops a running or starting turn.
On an idle recipient cancel_agent reports that it canceled no run, and
a queued steer keeps its quiet window. The lifecycle doc says so and a
daemon-level test pins it.

* docs(server): state when a queued message counts as delivered

A message leaves the queue when the agent manager acknowledges the start
of the turn that carries it. A turn that fails after that point is not
delivered again, and a start that is not acknowledged within 60 seconds
returns the messages a person's prompt carried.
2026-11-21 20:41:28 +07:00
Mohamed Boudra 1e30f9e046 Polish Usage quota bars, labels, and forecasts (#10397)
* fix(usage): clarify quota labels and show weekly forecasts sooner

* fix(usage): show the exhaustion time and keep exhausted bars red

* fix(usage): hide forecast text for quotas on track

* fix(usage): label quota times by local calendar date

* test(usage): keep the exhaustion clock test stable at midnight
2026-11-21 20:08:09 +07:00
paseo-bot[bot] 6946d0a429 fix(workspace): publish checkout status after restore (#10406)
A client that read checkout status while the worktree was absent can keep that response after Restore. Send the current status to subscribed clients before publishing the recovered workspace.

Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
2026-11-21 12:39:02 +00:00
paseo-bot[bot] 344da5d088 Let every split divider drag after panes are added to a resized row (#5996)
* fix(app): let every split divider drag after panes are added to a resized row

Divider drags store sizes per group id, apart from the layout tree. A group
keeps its id when panes are added, so after resizing a row and splitting it
again the stored list had fewer entries than panes, and dividers past its end
could not move. Use the dragged sizes only while they have one entry per pane.

Closes #5893

* refactor(app): choose split group sizes at the call site

* test(app): take the page from the divider locator in the split resize spec

* test(app): name the split resize helper arguments

---------

Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
2026-11-21 11:56:09 +00:00
Mohamed Boudra 01aa8b3406 Verify releases on isolated release branches and harden shipping (#10396)
* chore(release): verify releases on isolated release branches and harden shipping

High-traffic merges on main cancel more than half of main CI runs, so waiting on main CI repeatedly stalls releases. Releases also ran duplicate post-tag CI, lacked a release-branch Desktop Packages trigger, hit unretried Apple notarytool polling errors and a finalize-rollout directory bug on macOS retries, and relied on a non-idempotent npm publish chain and conflict-prone reconcile cherry-picks.

This splits release execution into release:prep:* (reversible version commit without tagging, pushed once to release/** for isolated CI + Desktop Packages) and release:ship[:beta] (idempotent npm publishing with retry, propagation checks, automatic beta dist-tag sync on stable, and release/*-only tag push), adds conflict-free main reconciliation, and extends macOS xcrun retries to notarytool submit and wait.

* fix(release): address review findings on Docker retries, tag push, and reconciliation

Isolate manual Docker publish dispatches in their own non-cancelling concurrency group, verify HEAD matches origin/<release-branch> before release:ship publishes or tags (and remove one-shot combined scripts that bypassed release-branch checks), drop --refs from peeled annotated-tag lookups in push-current-release-tag.mjs, retry lockfile sync in reconcile-release.mjs when package-lock.json version lags package.json, and document cutting from pre-existing tags.

* fix(release): preserve TTY OTP prompts in npm publish and keep newer beta changelogs

Inherit TTY stdin/stdout for interactive npm publish and dist-tag add (plus --otp passthrough) while keeping stderr piped for retry classification, and keep newer beta or promoted stable changelog entries intact when an older beta tag is reconciled.
2026-11-21 18:32:08 +07:00
Mohamed Boudraandtabulius-ux d816b94c54 Show workspace archive progress across sessions (#10391)
Share the archive marker across connections so clients and plugins see when a workspace is being archived.

Closes #7698

Co-authored-by: tabulius-ux <252143044+tabulius-ux@users.noreply.github.com>
2026-11-21 17:08:42 +07:00
Mohamed Boudraandtabulius-ux 4c0bf97dbf Honor the working directory when resuming Codex (#10390)
Codex loads configuration and starts MCP servers while restoring the thread. Send the session directory during resume so those steps use the requested workspace.

Closes #7699

Co-authored-by: tabulius-ux <252143044+tabulius-ux@users.noreply.github.com>
2026-11-21 17:08:23 +07:00
386 changed files with 21341 additions and 3428 deletions
+38 -9
View File
@@ -2,7 +2,7 @@ name: CI
on:
push:
branches: [main]
branches: [main, "release/**"]
# Every base branch, so stacked PRs and PRs into any base get the same checks.
pull_request:
merge_group:
@@ -20,6 +20,8 @@ env:
# CI does not use the CUDA execution provider, and the onnxruntime-node
# postinstall download from NuGet is large enough to make npm ci flaky.
ONNXRUNTIME_NODE_INSTALL: skip
NPM_CONFIG_AUDIT: "false"
NPM_CONFIG_FUND: "false"
jobs:
changes:
@@ -56,7 +58,7 @@ jobs:
filters: .github/ci-paths.yml
- name: Validate CI contracts
run: node --test scripts/ci-workflow.test.mjs scripts/daemon-launch-contract.test.mjs scripts/changelog-utils.test.mjs scripts/sync-fdroid-changelogs.test.mjs scripts/sync-release-notes-from-changelog.test.mjs scripts/upload-release-assets.test.mjs
run: node --test scripts/ci-workflow.test.mjs scripts/daemon-launch-contract.test.mjs scripts/changelog-utils.test.mjs scripts/release-version-utils.test.mjs scripts/publish-release-packages.test.mjs scripts/push-current-release-tag.test.mjs scripts/reconcile-release.test.mjs scripts/sync-fdroid-changelogs.test.mjs scripts/sync-release-notes-from-changelog.test.mjs scripts/upload-release-assets.test.mjs
quality:
name: quality
@@ -165,8 +167,29 @@ jobs:
- name: Run website tests
run: npx vitest run packages/website/src --bail=1
# WebKit needs about 180 system packages (125 MB) from the Ubuntu mirror. The mirror
# sometimes serves them at under 100 kB/s, which turns a 1-minute step into 25 minutes.
# Keep the downloaded .deb files in the Actions cache so apt installs them without the mirror.
- name: Restore website browser system packages
uses: actions/cache@v4
with:
path: ~/website-browser-debs
key: website-browser-debs-${{ runner.os }}-${{ hashFiles('package-lock.json') }}
restore-keys: website-browser-debs-${{ runner.os }}-
- name: Install website test browsers
run: npx playwright install --with-deps chromium webkit
run: |
mkdir -p ~/website-browser-debs
shopt -s nullglob
cached_debs=(~/website-browser-debs/*.deb)
if (( ${#cached_debs[@]} > 0 )); then
sudo cp "${cached_debs[@]}" /var/cache/apt/archives/
fi
npx playwright install --with-deps chromium webkit
archive_debs=(/var/cache/apt/archives/*.deb)
if (( ${#archive_debs[@]} > 0 )); then
cp "${archive_debs[@]}" ~/website-browser-debs/
fi
- name: Verify website mockups
run: npm run test:e2e --workspace=@getpaseo/website
@@ -278,8 +301,8 @@ jobs:
# native-recursive.ts serves darwin and win32, but windows-latest above is the only
# runner that exercises it. macOS — the platform the reported watcher bug was on —
# otherwise has no server-test coverage. Scoped to the observer/watcher suites to
# bound runtime.
# otherwise has no server-test coverage. Keep this job scoped to regressions
# reported on macOS.
server-tests-macos:
name: server-tests (macos-latest, file observation)
needs: changes
@@ -307,6 +330,12 @@ jobs:
src/server/workspace-git-service.observation.integration.test.ts \
--bail=1
- name: Verify plugin compiler children are reaped
run: |
npm run test:unit --workspace=@getpaseo/server -- \
src/server/plugins/index.posix.test.ts \
-t "reaps compiler children" --bail=1
desktop-tests-ubuntu:
name: desktop-tests (ubuntu-latest)
needs: changes
@@ -460,9 +489,9 @@ jobs:
OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
E2E_WORKERS: "2"
# File groups are weighted by observed reporter durations.
# Share of the test list per shard, rebalanced from the Playwright step
# minutes of five eight-shard runs so the slowest and fastest shards meet.
PWTEST_SHARD_WEIGHTS: "85:94:117:128:91:103:127:85"
# Share of the test list per shard, rebalanced from main and PR eight-shard
# durations so the slowest and fastest shards meet.
PWTEST_SHARD_WEIGHTS: "90:114:100:104:108:112:112:90"
- name: Upload test artifacts
uses: actions/upload-artifact@v4
@@ -576,7 +605,7 @@ jobs:
# Source tests do not read PASEO_CLI_TEST_SHARD. Run them once, on the
# shortest E2E shard, after its runner has built the server stack.
- name: Run CLI source tests
if: ${{ env.PASEO_CLI_TEST_SHARD == '3' }}
if: ${{ env.PASEO_CLI_TEST_SHARD == '2' }}
run: npm run test:unit --workspace=@getpaseo/cli
# The dependency watchers write build outputs, so run them after the parallel CLI tests.
+3 -2
View File
@@ -2,7 +2,7 @@ name: Desktop Packages
on:
push:
branches: [main]
branches: [main, "release/**"]
# Packaging leaves ordinary PR CI, but a PR that changes what gets packaged
# or how it is smoked still proves the real artifacts before merge.
pull_request:
@@ -11,10 +11,11 @@ on:
- "plugins/**"
- "nix/**"
- ".github/workflows/desktop-packages.yml"
workflow_dispatch:
concurrency:
group: desktop-packages-${{ github.ref }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
cancel-in-progress: true
permissions:
contents: read
+6 -27
View File
@@ -170,33 +170,11 @@ jobs:
run: |
set -euo pipefail
export PASEO_REAL_XCRUN="$(command -v xcrun)"
stapler_bin="$(mktemp -d)"
trap 'rm -rf "$stapler_bin"' EXIT
cat > "$stapler_bin/xcrun" <<'STAPLER'
#!/bin/bash
if [[ "${1:-}" != stapler || "${2:-}" != staple ]]; then
exec "$PASEO_REAL_XCRUN" "$@"
fi
output="$(mktemp)"
trap 'rm -f "$output"' EXIT
for attempt in 1 2 3 4; do
if "$PASEO_REAL_XCRUN" "$@" >"$output" 2>&1; then
cat "$output"
exit 0
else
status=$?
fi
cat "$output"
if [[ "$status" != 68 || "$attempt" == 4 ]] ||
! grep -Fq 'NSURLErrorDomain Code=-1005' "$output"; then
exit "$status"
fi
echo "::warning::Apple staple connection lost; retrying the same signed app ($attempt/4)" >&2
sleep 30
done
STAPLER
chmod +x "$stapler_bin/xcrun"
export PATH="$stapler_bin:$PATH"
xcrun_bin="$(mktemp -d)"
trap 'rm -rf "$xcrun_bin"' EXIT
cp scripts/macos-xcrun-retry.sh "$xcrun_bin/xcrun"
chmod +x "$xcrun_bin/xcrun"
export PATH="$xcrun_bin:$PATH"
build_args=(-- --publish never --mac --${{ matrix.electron_arch }})
build_args+=("-c.publish.releaseType=$RELEASE_TYPE")
build_args+=("-c.publish.channel=$RELEASE_CHANNEL")
@@ -502,6 +480,7 @@ jobs:
shell: bash
run: |
set -euo pipefail
mkdir -p release-manifests
cd release-manifests
manifests_dir="$PWD/final"
mkdir -p "$manifests_dir"
+2 -2
View File
@@ -29,8 +29,8 @@ on:
- "true"
concurrency:
group: docker-${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: false
group: docker-${{ github.workflow }}-${{ github.event_name == 'workflow_dispatch' && inputs.publish == 'true' && format('publish-{0}', inputs.paseo_version || github.run_id) || github.ref }}
cancel-in-progress: ${{ !startsWith(github.ref, 'refs/tags/') && !(github.event_name == 'workflow_dispatch' && inputs.publish == 'true') }}
env:
REGISTRY: ghcr.io
+1
View File
@@ -3,6 +3,7 @@ name: Nix
on:
push:
branches: [main]
workflow_dispatch:
permissions:
contents: read
+60 -28
View File
@@ -1,70 +1,101 @@
# Changelog
## 0.17.0 - 2026-11-21
## 0.17.2 - 2026-11-22
- **Use existing worktrees:** Start a workspace in a worktree you already have, or create one from a branch or pull request. Plugins can manage worktree creation, removal, and restoration. When you branch off a fork PR or MR, choose Run setup to approve workspace automation. ([#10190](https://github.com/getpaseo/paseo/pull/10190), [#10278](https://github.com/getpaseo/paseo/pull/10278))
- **Update providers in Paseo:** See each provider's status, installed version, and available updates in the provider list. Open a provider's settings from Settings, setup, or the model picker to update supported CLIs and watch their output. ([#9568](https://github.com/getpaseo/paseo/pull/9568), [#9661](https://github.com/getpaseo/paseo/pull/9661))
- **Provider quota forecasts:** Usage estimates when a quota will run out or how much will remain at reset, based on your current pace. The forecast appears inline with the reset time. ([#10289](https://github.com/getpaseo/paseo/pull/10289), [#10385](https://github.com/getpaseo/paseo/pull/10385))
- **Watch PRs and MRs:** Ask an agent to watch a GitHub, GitLab, Gitea, Forgejo, or Codeberg change request for updates to checks, reviews, or comments. Stop a watch from Background work. ([#10148](https://github.com/getpaseo/paseo/pull/10148), [#10279](https://github.com/getpaseo/paseo/pull/10279), [#10280](https://github.com/getpaseo/paseo/pull/10280))
- **Turn and session stats:** View per-turn and session timing, token, cache, and cost metrics for Claude, OpenCode v2, Pi, Oh My Pi, and plugin providers. ([#10496](https://github.com/getpaseo/paseo/pull/10496), [#10510](https://github.com/getpaseo/paseo/pull/10510))
- **Automatic model list refresh:** Enabled providers refresh their model lists in the background every 30 minutes. ([#10481](https://github.com/getpaseo/paseo/pull/10481))
- **Sidebar usage by default:** The sidebar usage summary is enabled by default, sources for inactive providers start hidden, and source toggles move to a Sources list. ([#10491](https://github.com/getpaseo/paseo/pull/10491))
- **Agents can steer turns:** Agents can steer another agent's running turn alongside queueing or interrupting it, and receive delivery guidance in the response. ([#10483](https://github.com/getpaseo/paseo/pull/10483))
### Added
- Added Archive workspace to the mobile workspace actions menu ([#10507](https://github.com/getpaseo/paseo/pull/10507))
- Added provider docs links, fuller diagnostic reports, and a scrollable Models list with inline refresh to provider settings ([#10487](https://github.com/getpaseo/paseo/pull/10487))
### Changed
- Removed the "This chat uses OpenCode v2." info row from OpenCode v2 chats ([#10490](https://github.com/getpaseo/paseo/pull/10490))
### Fixed
- Fixed pressing Tab or Enter in the slash menu running a stale `/exit` choice and archiving the agent ([#10484](https://github.com/getpaseo/paseo/pull/10484) by [@3ae3ae](https://github.com/3ae3ae))
- Fixed busy host connections dropping to reconnect when multiple connections are saved ([#10395](https://github.com/getpaseo/paseo/pull/10395))
- Fixed retrying an edited New workspace form hitting a conflict error after a failed start, and unified composer error alerts ([#10482](https://github.com/getpaseo/paseo/pull/10482))
- Fixed Antigravity turns staying running after creating a recurring schedule ([#10469](https://github.com/getpaseo/paseo/pull/10469))
- Fixed daemon startup and plugin reloads leaving zombie child processes for local directory plugins ([#10470](https://github.com/getpaseo/paseo/pull/10470))
## 0.17.1 - 2026-11-21
- **Antigravity approvals and history:** Choose a permission mode, answer tool approvals, and import or reopen sessions. Requires [Antigravity CLI 1.5.0 or later](https://paseo.sh/docs/supported-providers/antigravity). ([#10193](https://github.com/getpaseo/paseo/pull/10193))
- **Use existing worktrees:** Start a workspace in a worktree you already have, or create one from a branch or pull request. ([#10190](https://github.com/getpaseo/paseo/pull/10190))
- **Update providers in Paseo:** See each provider's status and installed version, and update supported CLIs from the app. ([#9568](https://github.com/getpaseo/paseo/pull/9568), [#9661](https://github.com/getpaseo/paseo/pull/9661))
- **Provider quota forecasts:** Usage shows when your current pace will use up a quota before it resets. ([#10289](https://github.com/getpaseo/paseo/pull/10289), [#10385](https://github.com/getpaseo/paseo/pull/10385))
- **Watch PRs and MRs:** Ask an agent to watch a pull or merge request for new checks, reviews, and comments. ([#10148](https://github.com/getpaseo/paseo/pull/10148), [#10279](https://github.com/getpaseo/paseo/pull/10279), [#10280](https://github.com/getpaseo/paseo/pull/10280))
- **Heartbeats in Background work:** See and stop an agent's heartbeats next to its background tasks. ([#10097](https://github.com/getpaseo/paseo/pull/10097), [#10216](https://github.com/getpaseo/paseo/pull/10216))
- **Agent messages wait their turn:** Reports and notifications from other agents arrive when the current turn ends instead of interrupting it. ([#9757](https://github.com/getpaseo/paseo/pull/9757))
- **Antigravity approvals and history:** Choose a permission mode, respond to tool approvals, and import or reopen sessions. Use Antigravity CLI 1.5.0 or later on the daemon host, and include `acp` if you override its command. In Hub, reselect the model and thinking level if a saved model ID includes an old effort suffix. ([#10193](https://github.com/getpaseo/paseo/pull/10193))
- **More goose models:** Select models from all providers configured in goose. Existing goose settings and sessions carry over to the bundled plugin. ([#9719](https://github.com/getpaseo/paseo/pull/9719) by [@i386](https://github.com/i386))
- **Agent messages wait their turn:** By default, messages from other agents wait for the current turn to end. ([#9757](https://github.com/getpaseo/paseo/pull/9757), [#10404](https://github.com/getpaseo/paseo/pull/10404))
- **More goose models:** Select models from every provider configured in goose. ([#9719](https://github.com/getpaseo/paseo/pull/9719) by [@i386](https://github.com/i386))
- **Plugin agent instructions:** Plugins can add their own instructions to agents. ([#9781](https://github.com/getpaseo/paseo/pull/9781))
### Added
- Added OpenRouter credit balances and API key spending to Usage and `paseo usage` ([#10269](https://github.com/getpaseo/paseo/pull/10269))
- Added Antigravity account quotas to Usage ([#10232](https://github.com/getpaseo/paseo/pull/10232) by [@luoph](https://github.com/luoph), [@RRD-dev](https://github.com/RRD-dev))
- Added Antigravity skill suggestions when you type `/` inside a prompt ([#10383](https://github.com/getpaseo/paseo/pull/10383))
- Added bundled orchestration skills to Antigravity on hosts where Paseo skills are installed ([#10273](https://github.com/getpaseo/paseo/pull/10273))
- Added Antigravity skills to `/` suggestions ([#10383](https://github.com/getpaseo/paseo/pull/10383))
- Added bundled orchestration skills to Antigravity ([#10273](https://github.com/getpaseo/paseo/pull/10273))
- Added shortcuts to open the model picker and cycle thinking levels with Cmd/Ctrl+/ and Cmd/Ctrl+Shift+/ ([#10147](https://github.com/getpaseo/paseo/pull/10147))
- Added agent settings to Command Center search before the first message ([#10182](https://github.com/getpaseo/paseo/pull/10182) by [@cleiter](https://github.com/cleiter))
- Added Rewind to Command Center for the latest turn ([#10144](https://github.com/getpaseo/paseo/pull/10144))
- Added a separate Ultracode toggle for eligible Claude models ([#10150](https://github.com/getpaseo/paseo/pull/10150) by [@fivetaku](https://github.com/fivetaku))
- Added a separate Ultracode toggle for Claude ([#10150](https://github.com/getpaseo/paseo/pull/10150) by [@fivetaku](https://github.com/fivetaku))
- Added Flex speed for Codex ([#9809](https://github.com/getpaseo/paseo/pull/9809) by [@qjqqyy](https://github.com/qjqqyy))
- Added syntax highlighting for shell scripts and Dockerfiles ([#10178](https://github.com/getpaseo/paseo/pull/10178))
- Added project listing on connected daemons for Hub sessions with `hub.execute` ([#10392](https://github.com/getpaseo/paseo/pull/10392))
- Added project listing on connected daemons for Hub sessions ([#10392](https://github.com/getpaseo/paseo/pull/10392))
- Added labels and workspace IDs to `paseo ls --json` ([#10076](https://github.com/getpaseo/paseo/pull/10076))
### Changed
- Changed Appearance content width to Default (820 px), Wide (1,200 px), and Full presets that resize open chats, the composer, and Markdown previews together ([#10231](https://github.com/getpaseo/paseo/pull/10231))
- Changed Antigravity tool cards to show shell commands and output, file diffs, and Paseo tool details ([#10379](https://github.com/getpaseo/paseo/pull/10379))
- Changed content width to Default, Wide, and Full presets that also resize open chats and Markdown previews ([#10231](https://github.com/getpaseo/paseo/pull/10231))
- Changed Antigravity tool cards to show shell commands, output, and file diffs ([#10379](https://github.com/getpaseo/paseo/pull/10379))
- Changed Clone from GitHub to show matching private repositories first ([#10264](https://github.com/getpaseo/paseo/pull/10264))
- Changed worktree teardown to keep the worktree when the command exceeds one minute or the `worktree.teardownTimeoutMs` limit in `paseo.json` ([#9852](https://github.com/getpaseo/paseo/pull/9852) by [@ragokan](https://github.com/ragokan))
- Changed workspace navigation to acknowledge taps promptly and initially load fewer offscreen chat rows on mobile ([#9887](https://github.com/getpaseo/paseo/pull/9887))
- Changed workspace Git status to reuse its snapshot after a short reconnect when the checkout has not changed ([#10142](https://github.com/getpaseo/paseo/pull/10142) by [@VladimirGVP](https://github.com/VladimirGVP))
- Changed healthy quota bars in Usage to green ([#10397](https://github.com/getpaseo/paseo/pull/10397))
- Changed worktree teardown to keep the worktree when the command times out ([#9852](https://github.com/getpaseo/paseo/pull/9852) by [@ragokan](https://github.com/ragokan))
- Changed workspace navigation on mobile to acknowledge taps sooner ([#9887](https://github.com/getpaseo/paseo/pull/9887))
- Changed workspace Git status to skip its reload after a short reconnect with no file changes ([#10142](https://github.com/getpaseo/paseo/pull/10142) by [@VladimirGVP](https://github.com/VladimirGVP))
- Changed `browser_new_tab` to ask agents to choose a Desktop when several are connected ([#10189](https://github.com/getpaseo/paseo/pull/10189))
### Fixed
- Fixed an online host's workspaces staying missing after the app evicts its cached directory ([#10318](https://github.com/getpaseo/paseo/pull/10318))
- Fixed chat history timing out when many open chats refresh at once after reconnecting ([#10368](https://github.com/getpaseo/paseo/pull/10368))
- Fixed an online host's workspaces going missing ([#10318](https://github.com/getpaseo/paseo/pull/10318))
- Fixed chat history timing out when many open chats refresh after reconnecting ([#10368](https://github.com/getpaseo/paseo/pull/10368))
- Fixed slow reopening of long Claude chats while the daemon is busy ([#10369](https://github.com/getpaseo/paseo/pull/10369))
- Fixed images from a live reply waiting for images in another chat's history to load ([#10371](https://github.com/getpaseo/paseo/pull/10371))
- Fixed agent recovery failing and idle agents showing errors after stopping a foreground daemon with Ctrl+C on macOS or Linux ([#10381](https://github.com/getpaseo/paseo/pull/10381))
- Fixed agents failing to recover after stopping a foreground daemon with Ctrl+C ([#10381](https://github.com/getpaseo/paseo/pull/10381))
- Fixed long Antigravity chats losing the end of their history after a daemon restart ([#10428](https://github.com/getpaseo/paseo/pull/10428))
- Fixed Antigravity turns ending while a background command is still running ([#10310](https://github.com/getpaseo/paseo/pull/10310))
- Fixed the Add model sheet staying open after saving a custom model ([#10432](https://github.com/getpaseo/paseo/pull/10432))
- Fixed slash suggestions failing when a provider reports duplicate command names ([#10275](https://github.com/getpaseo/paseo/pull/10275))
- Fixed resumed Codex sessions using their previous working directory for configuration and MCP servers ([#10390](https://github.com/getpaseo/paseo/pull/10390) by [@tabulius-ux](https://github.com/tabulius-ux))
- Fixed stale Codex subagent statuses after reopening or rewinding the parent chat ([#10171](https://github.com/getpaseo/paseo/pull/10171) by [@AnalyticETH](https://github.com/AnalyticETH))
- Fixed Usage showing the previous Claude account's quota after switching accounts ([#10236](https://github.com/getpaseo/paseo/pull/10236))
- Fixed the sidebar usage row dropping period labels such as `5h` and `wk` when multiple usage windows are pinned ([#10274](https://github.com/getpaseo/paseo/pull/10274))
- Fixed Changes failing to recognize a checkout immediately after restoring its worktree ([#10224](https://github.com/getpaseo/paseo/pull/10224), [#10304](https://github.com/getpaseo/paseo/pull/10304))
- Fixed the sidebar usage row dropping period labels when several usage windows are pinned ([#10274](https://github.com/getpaseo/paseo/pull/10274))
- Fixed Changes failing to recognize a checkout immediately after restoring its worktree ([#10224](https://github.com/getpaseo/paseo/pull/10224), [#10304](https://github.com/getpaseo/paseo/pull/10304), [#10406](https://github.com/getpaseo/paseo/pull/10406))
- Fixed split-pane dividers stopping or lagging during drags ([#5996](https://github.com/getpaseo/paseo/pull/5996))
- Fixed double-click word selection and triple-click line selection in diffs ([#10252](https://github.com/getpaseo/paseo/pull/10252) by [@pedrovieira](https://github.com/pedrovieira))
- Fixed the composer and chat column losing alignment when Chat outline is enabled ([#10277](https://github.com/getpaseo/paseo/pull/10277))
- Fixed the web composer's focus shortcut hint overlapping the placeholder ([#6200](https://github.com/getpaseo/paseo/pull/6200))
- Fixed the archive shortcut failing when the active workspace's sidebar group is collapsed ([#10356](https://github.com/getpaseo/paseo/pull/10356) by [@hiennd-ctrl](https://github.com/hiennd-ctrl), [@neihyud](https://github.com/neihyud))
- Fixed other sessions and plugins seeing a workspace as active while another session is archiving it ([#10391](https://github.com/getpaseo/paseo/pull/10391) by [@tabulius-ux](https://github.com/tabulius-ux))
- Fixed Android composer buttons requiring a second tap after typing or zooming a chat image ([#10057](https://github.com/getpaseo/paseo/pull/10057), [#10308](https://github.com/getpaseo/paseo/pull/10308))
- Fixed swiping a terminal in an Android browser opening the keyboard when it was closed ([#10225](https://github.com/getpaseo/paseo/pull/10225) by [@tonylt](https://github.com/tonylt))
- Fixed swiping a terminal in an Android browser opening the keyboard ([#10225](https://github.com/getpaseo/paseo/pull/10225) by [@tonylt](https://github.com/tonylt))
- Fixed Collapse turns hiding a finished OMP or Pi reply when an idle provider notice appears ([#10135](https://github.com/getpaseo/paseo/pull/10135))
- Fixed repeated finish notifications while an agent responds to background tasks and its earlier finish is unread ([#10200](https://github.com/getpaseo/paseo/pull/10200))
- Fixed repeated finish notifications while an agent responds to background tasks ([#10200](https://github.com/getpaseo/paseo/pull/10200))
- Fixed an empty chat losing the prompt and start of a reply after reconnecting mid-reply ([#9723](https://github.com/getpaseo/paseo/pull/9723) by [@ragokan](https://github.com/ragokan))
- Fixed changes between Claude thinking levels stopping its background work ([#10008](https://github.com/getpaseo/paseo/pull/10008) by [@gilad-rubin](https://github.com/gilad-rubin))
- Fixed changing the Claude thinking level stopping its background work ([#10008](https://github.com/getpaseo/paseo/pull/10008) by [@gilad-rubin](https://github.com/gilad-rubin))
- Fixed OpenCode turns staying running after a shared OpenCode service stops ([#10187](https://github.com/getpaseo/paseo/pull/10187))
- Fixed OMP denying an approved tool call when session or grant choices were offered ([#10126](https://github.com/getpaseo/paseo/pull/10126) by [@dsingal0](https://github.com/dsingal0))
- Fixed a Codex message stopped before its turn starts showing an error instead of canceled ([#9931](https://github.com/getpaseo/paseo/pull/9931) by [@3ae3ae](https://github.com/3ae3ae))
- Fixed Changes failing to open when Git metadata watching reaches its limit ([#10188](https://github.com/getpaseo/paseo/pull/10188))
- Fixed copying selected reply text into plain-text targets with unwanted Markdown syntax ([#10079](https://github.com/getpaseo/paseo/pull/10079) by [@danberindei](https://github.com/danberindei))
- Fixed file views losing their scroll position when switching Preview and Source, except HTML previews on iOS ([#10195](https://github.com/getpaseo/paseo/pull/10195) by [@apple-ouyang](https://github.com/apple-ouyang))
- Fixed Markdown previews and source views losing their scroll position when switching between them ([#10195](https://github.com/getpaseo/paseo/pull/10195) by [@apple-ouyang](https://github.com/apple-ouyang))
- Fixed file tabs in split panes showing or focusing the wrong file ([#9888](https://github.com/getpaseo/paseo/pull/9888) by [@KosmoCHE](https://github.com/KosmoCHE))
- Fixed touch-screen submenus closing as soon as they open ([#10175](https://github.com/getpaseo/paseo/pull/10175) by [@christiantwu](https://github.com/christiantwu))
- Fixed dictation failing after another chat that used the microphone closes ([#10185](https://github.com/getpaseo/paseo/pull/10185) by [@Brandon168](https://github.com/Brandon168))
@@ -74,13 +105,14 @@
- Fixed Mermaid flowcharts with `@` in labels showing source instead of the diagram ([#10165](https://github.com/getpaseo/paseo/pull/10165) by [@monarchmaisuriya](https://github.com/monarchmaisuriya))
- Fixed hidden background workspaces triggering the dock badge and favicon ([#9983](https://github.com/getpaseo/paseo/pull/9983))
- Fixed desktop sidebar navigation crowding out workspaces ([#9903](https://github.com/getpaseo/paseo/pull/9903) by [@pentatonic-alex](https://github.com/pentatonic-alex))
- Fixed historical closed PR badges appearing on a repository's default branch ([#10053](https://github.com/getpaseo/paseo/pull/10053))
- Fixed closed PR badges appearing on a repository's default branch ([#10053](https://github.com/getpaseo/paseo/pull/10053))
- Fixed plugin agents failing to reopen after their provider closes the session ([#10153](https://github.com/getpaseo/paseo/pull/10153))
- Fixed plugin screens and settings failing to use plugin state hooks ([#10049](https://github.com/getpaseo/paseo/pull/10049))
### Plugins
- Added a `focusHost` navigation option for plugins to select the destination host in the sidebar when opening a workspace or agent ([#5574](https://github.com/getpaseo/paseo/pull/5574) by [@discwl](https://github.com/discwl))
- Added plugin control of worktree creation, removal, and restoration ([#10278](https://github.com/getpaseo/paseo/pull/10278))
- Added a `focusHost` navigation option that selects the destination host in the sidebar ([#5574](https://github.com/getpaseo/paseo/pull/5574) by [@discwl](https://github.com/discwl))
## 0.16.2 - 2026-11-18
+58 -26
View File
@@ -1,62 +1,93 @@
# Changelog
## 0.17.0 - 2026-11-21
## 0.17.2 - 2026-11-22
- **使用现有 worktree**:在已有的 worktree 中启动工作区,或从分支或 PR 创建 worktree。插件可以管理 worktree 的创建、移除和恢复。从 fork 的 PR 或 MR 创建分支时,选择“运行设置”以批准工作区自动化。([#10190](https://github.com/getpaseo/paseo/pull/10190), [#10278](https://github.com/getpaseo/paseo/pull/10278))
- **在 Paseo 中更新 Provider**:在 Provider 列表中查看各 Provider 的状态、已安装版本和可用更新。从设置、初次配置或模型选择器打开 Provider 设置,即可更新受支持的 CLI 并查看输出。([#9568](https://github.com/getpaseo/paseo/pull/9568), [#9661](https://github.com/getpaseo/paseo/pull/9661))
- **Provider 配额预测**:“使用情况”页面根据当前使用速度估算配额何时用完,或重置时还会剩下多少。预测结果显示在重置时间同一行中。([#10289](https://github.com/getpaseo/paseo/pull/10289), [#10385](https://github.com/getpaseo/paseo/pull/10385))
- **监视 PR 和 MR**:让 Agent 监视 GitHub、GitLab、Gitea、Forgejo 或 Codeberg 上的 PR 或 MR,跟踪检查、审查或评论的变化。你可以在后台任务中停止监视。([#10148](https://github.com/getpaseo/paseo/pull/10148), [#10279](https://github.com/getpaseo/paseo/pull/10279), [#10280](https://github.com/getpaseo/paseo/pull/10280))
- **回合和会话统计**:查看 Claude、OpenCode v2、Pi、Oh My Pi 和插件 Provider 的单轮及会话耗时、token、缓存和费用指标。([#10496](https://github.com/getpaseo/paseo/pull/10496), [#10510](https://github.com/getpaseo/paseo/pull/10510))
- **自动刷新模型列表**:已启用的 Provider 每 30 分钟在后台刷新一次模型列表。([#10481](https://github.com/getpaseo/paseo/pull/10481))
- **默认显示侧边栏用量**:侧边栏用量摘要现在默认开启,未启用 Provider 的用量来源默认隐藏,来源开关移至“来源”列表。([#10491](https://github.com/getpaseo/paseo/pull/10491))
- **Agent 可引导进行中的轮次**:Agent 除了排队或中断外,现在可以引导另一个 Agent 正在运行的轮次,并在响应中收到投递指引。([#10483](https://github.com/getpaseo/paseo/pull/10483))
### 新增
- 在移动端工作区操作菜单中新增“归档工作区” ([#10507](https://github.com/getpaseo/paseo/pull/10507))
- 在 Provider 设置中新增文档链接、更完整的诊断报告,以及带内联刷新按钮的可滚动 Models 列表 ([#10487](https://github.com/getpaseo/paseo/pull/10487))
### 变更
- 移除 OpenCode v2 对话中的“This chat uses OpenCode v2.”提示行 ([#10490](https://github.com/getpaseo/paseo/pull/10490))
### 修复
- 修复在斜杠菜单中按 Tab 或 Enter 时误触发过期的 `/exit` 选项并归档 Agent 的问题 ([#10484](https://github.com/getpaseo/paseo/pull/10484),贡献者:[@3ae3ae](https://github.com/3ae3ae))
- 修复保存了多个主机连接时,繁忙的正常连接会断开重连的问题 ([#10395](https://github.com/getpaseo/paseo/pull/10395))
- 修复新建工作区启动失败后,修改表单再重试时出现冲突错误的问题,并统一了 Composer 错误提示 ([#10482](https://github.com/getpaseo/paseo/pull/10482))
- 修复 Antigravity 创建循环定时任务后轮次一直停留在运行状态的问题 ([#10469](https://github.com/getpaseo/paseo/pull/10469))
- 修复启动 daemon 和重载本地目录插件后遗留僵尸子进程的问题 ([#10470](https://github.com/getpaseo/paseo/pull/10470))
## 0.17.1 - 2026-11-21
- **Antigravity 审批和历史记录**:选择权限模式、回复工具审批,以及导入或重新打开会话。需要 [Antigravity CLI 1.5.0 或更高版本](https://paseo.sh/zh-CN/docs/supported-providers/antigravity)。([#10193](https://github.com/getpaseo/paseo/pull/10193))
- **使用现有 worktree**:在已有的 worktree 中启动工作区,或从分支或 PR 创建 worktree。([#10190](https://github.com/getpaseo/paseo/pull/10190))
- **在 Paseo 中更新 Provider**:查看各 Provider 的状态和已安装版本,并在应用内更新受支持的 CLI。([#9568](https://github.com/getpaseo/paseo/pull/9568), [#9661](https://github.com/getpaseo/paseo/pull/9661))
- **Provider 配额预测**:“使用情况”页面会显示配额按当前使用速度将在重置前何时用完。([#10289](https://github.com/getpaseo/paseo/pull/10289), [#10385](https://github.com/getpaseo/paseo/pull/10385))
- **监视 PR 和 MR**:让 Agent 监视 PR 或 MR 的检查、审查和评论更新。([#10148](https://github.com/getpaseo/paseo/pull/10148), [#10279](https://github.com/getpaseo/paseo/pull/10279), [#10280](https://github.com/getpaseo/paseo/pull/10280))
- **在后台任务中查看 heartbeat**:在 Agent 的后台任务旁查看和停止 heartbeat。([#10097](https://github.com/getpaseo/paseo/pull/10097), [#10216](https://github.com/getpaseo/paseo/pull/10216))
- **Agent 消息排队等待**:其他 Agent 的报告和通知会在当前轮次结束后到达,不会打断当前轮次。([#9757](https://github.com/getpaseo/paseo/pull/9757))
- **Antigravity 审批和历史记录**:选择权限模式、回复工具审批,以及导入或重新打开会话。请在 daemon 主机上使用 Antigravity CLI 1.5.0 或更新版本,如果覆盖了启动命令请包含 `acp`。如果 Hub 中保存的模型 ID 含有旧的推理强度后缀,请重新选择模型和思考级别。([#10193](https://github.com/getpaseo/paseo/pull/10193))
- **更多 goose 模型**:从 goose 中已配置的所有 Provider 选择模型。现有 goose 设置和会话会迁移到内置插件。([#9719](https://github.com/getpaseo/paseo/pull/9719),贡献者:[@i386](https://github.com/i386))
- **Agent 消息排队等待**:默认情况下,其他 Agent 的消息会等当前轮次结束后再送达。([#9757](https://github.com/getpaseo/paseo/pull/9757), [#10404](https://github.com/getpaseo/paseo/pull/10404))
- **更多 goose 模型**:从 goose 中已配置的所有 Provider 选择模型。([#9719](https://github.com/getpaseo/paseo/pull/9719),贡献者:[@i386](https://github.com/i386))
- **插件 Agent 指令**:插件可以向 Agent 添加自己的指令。([#9781](https://github.com/getpaseo/paseo/pull/9781))
### 新增
- 在“使用情况”页面和 `paseo usage` 中新增 OpenRouter 余额和 API 密钥支出 ([#10269](https://github.com/getpaseo/paseo/pull/10269))
- 在“使用情况”页面新增 Antigravity 账号配额 ([#10232](https://github.com/getpaseo/paseo/pull/10232),贡献者:[@luoph](https://github.com/luoph), [@RRD-dev](https://github.com/RRD-dev))
- 在提示词中输入 `/` 时,新增 Antigravity 技能建议 ([#10383](https://github.com/getpaseo/paseo/pull/10383))
- 在已安装 Paseo 技能的主机上,为 Antigravity 新增内置编排技能 ([#10273](https://github.com/getpaseo/paseo/pull/10273))
- 在 `/` 建议中新增 Antigravity 技能 ([#10383](https://github.com/getpaseo/paseo/pull/10383))
- 为 Antigravity 新增内置编排技能 ([#10273](https://github.com/getpaseo/paseo/pull/10273))
- 新增 Cmd/Ctrl+/ 和 Cmd/Ctrl+Shift+/ 快捷键,用于打开模型选择器和切换思考级别 ([#10147](https://github.com/getpaseo/paseo/pull/10147))
- 新增在发送第一条消息前通过命令中心搜索 Agent 设置 ([#10182](https://github.com/getpaseo/paseo/pull/10182),贡献者:[@cleiter](https://github.com/cleiter))
- 新增在命令中心回退最近轮次的操作 ([#10144](https://github.com/getpaseo/paseo/pull/10144))
- 新增符合条件的 Claude 模型的独立 Ultracode 开关 ([#10150](https://github.com/getpaseo/paseo/pull/10150),贡献者:[@fivetaku](https://github.com/fivetaku))
- 新增 Claude 的独立 Ultracode 开关 ([#10150](https://github.com/getpaseo/paseo/pull/10150),贡献者:[@fivetaku](https://github.com/fivetaku))
- 新增 Codex Flex 速度选项 ([#9809](https://github.com/getpaseo/paseo/pull/9809),贡献者:[@qjqqyy](https://github.com/qjqqyy))
- 新增 shell 脚本和 Dockerfile 的语法高亮 ([#10178](https://github.com/getpaseo/paseo/pull/10178))
- 为具有 `hub.execute` 权限的 Hub 会话新增列出已连接 daemon 活跃项目的能力 ([#10392](https://github.com/getpaseo/paseo/pull/10392))
- 为 Hub 会话新增列出已连接 daemon 项目的能力 ([#10392](https://github.com/getpaseo/paseo/pull/10392))
- 新增 `paseo ls --json` 输出的标签和工作区 ID ([#10076](https://github.com/getpaseo/paseo/pull/10076))
### 变更
- 将“外观”中的内容宽度改为默认(820 px)、宽版(1,200 px)和全宽预设,同时调整已打开对话、Composer 和 Markdown 预览的宽度 ([#10231](https://github.com/getpaseo/paseo/pull/10231))
- Antigravity 工具卡片现在显示 shell 命令及输出、文件 diff 和 Paseo 工具详情 ([#10379](https://github.com/getpaseo/paseo/pull/10379))
- 将内容宽度改为默认、宽版和全宽预设,并同时调整已打开对话和 Markdown 预览的宽度 ([#10231](https://github.com/getpaseo/paseo/pull/10231))
- Antigravity 工具卡片现在显示 shell 命令、输出和文件 diff ([#10379](https://github.com/getpaseo/paseo/pull/10379))
- 从 GitHub 克隆时,优先显示匹配的私有仓库 ([#10264](https://github.com/getpaseo/paseo/pull/10264))
- worktree 清理命令在超过默认的一分钟或 `paseo.json` 中 `worktree.teardownTimeoutMs` 限制时保留 worktree ([#9852](https://github.com/getpaseo/paseo/pull/9852),贡献者:[@ragokan](https://github.com/ragokan))
- 工作区导航会立即响应点击,并在手机端初始加载时减少屏幕外的对话行数 ([#9887](https://github.com/getpaseo/paseo/pull/9887))
- 工作区 Git 状态在短暂重连且检出的内容未更改时复用之前的快照 ([#10142](https://github.com/getpaseo/paseo/pull/10142),贡献者:[@VladimirGVP](https://github.com/VladimirGVP))
- “使用情况”页面中健康的配额进度条改为绿色 ([#10397](https://github.com/getpaseo/paseo/pull/10397))
- worktree 清理命令超时时保留 worktree ([#9852](https://github.com/getpaseo/paseo/pull/9852),贡献者:[@ragokan](https://github.com/ragokan))
- 手机端工作区导航更快响应点击 ([#9887](https://github.com/getpaseo/paseo/pull/9887))
- 短暂重连且文件未更改时,工作区 Git 状态不再重新加载 ([#10142](https://github.com/getpaseo/paseo/pull/10142),贡献者:[@VladimirGVP](https://github.com/VladimirGVP))
- 连接多个 Desktop 时,`browser_new_tab` 会让 Agent 选择打开标签页的 Desktop ([#10189](https://github.com/getpaseo/paseo/pull/10189))
### 修复
- 修复应用清除主机目录缓存后,在线主机的工作区一直不显示的问题 ([#10318](https://github.com/getpaseo/paseo/pull/10318))
- 修复在线主机的工作区不显示的问题 ([#10318](https://github.com/getpaseo/paseo/pull/10318))
- 修复重新连接后多个已打开的对话同时刷新时,历史记录加载超时的问题 ([#10368](https://github.com/getpaseo/paseo/pull/10368))
- 修复 daemon 繁忙时,重新打开较长的 Claude 对话速度慢的问题 ([#10369](https://github.com/getpaseo/paseo/pull/10369))
- 修复当前回复中的图片等待另一个对话的历史图片加载的问题 ([#10371](https://github.com/getpaseo/paseo/pull/10371))
- 修复在 macOS 或 Linux 上用 Ctrl+C 停止前台 daemon 后,Agent 恢复失败且空闲 Agent 显示错误的问题 ([#10381](https://github.com/getpaseo/paseo/pull/10381))
- 修复用 Ctrl+C 停止前台 daemon 后 Agent 恢复失败的问题 ([#10381](https://github.com/getpaseo/paseo/pull/10381))
- 修复 daemon 重启后较长的 Antigravity 对话丢失末尾历史记录的问题 ([#10428](https://github.com/getpaseo/paseo/pull/10428))
- 修复 Antigravity 后台命令仍在运行时轮次就已结束的问题 ([#10310](https://github.com/getpaseo/paseo/pull/10310))
- 修复保存自定义模型后“添加模型”弹窗一直保持打开的问题 ([#10432](https://github.com/getpaseo/paseo/pull/10432))
- 修复 Provider 返回重复命令名称时,斜杠建议列表出错的问题 ([#10275](https://github.com/getpaseo/paseo/pull/10275))
- 修复恢复的 Codex 会话的配置和 MCP 服务器仍沿用旧工作目录的问题 ([#10390](https://github.com/getpaseo/paseo/pull/10390),贡献者:[@tabulius-ux](https://github.com/tabulius-ux))
- 修复重新打开或回退父对话后,Codex 子 Agent 状态未更新的问题 ([#10171](https://github.com/getpaseo/paseo/pull/10171),贡献者:[@AnalyticETH](https://github.com/AnalyticETH))
- 修复切换 Claude 账号后,“使用情况”页面仍显示上一个账号配额的问题 ([#10236](https://github.com/getpaseo/paseo/pull/10236))
- 修复固定多个用量窗口时,侧边栏用量行隐藏 `5h` 和 `wk` 等周期标签的问题 ([#10274](https://github.com/getpaseo/paseo/pull/10274))
- 修复恢复 worktree 后,“更改”无法立即识别检出目录的问题 ([#10224](https://github.com/getpaseo/paseo/pull/10224), [#10304](https://github.com/getpaseo/paseo/pull/10304))
- 修复固定多个用量窗口时,侧边栏用量行隐藏周期标签的问题 ([#10274](https://github.com/getpaseo/paseo/pull/10274))
- 修复恢复 worktree 后,“更改”无法立即识别检出目录的问题 ([#10224](https://github.com/getpaseo/paseo/pull/10224), [#10304](https://github.com/getpaseo/paseo/pull/10304), [#10406](https://github.com/getpaseo/paseo/pull/10406))
- 修复拖动分屏分隔条时卡住或跟手缓慢的问题 ([#5996](https://github.com/getpaseo/paseo/pull/5996))
- 修复 diff 中双击选词和三击选行的问题 ([#10252](https://github.com/getpaseo/paseo/pull/10252),贡献者:[@pedrovieira](https://github.com/pedrovieira))
- 修复启用聊天大纲时,Composer 与对话栏对不齐的问题 ([#10277](https://github.com/getpaseo/paseo/pull/10277))
- 修复网页版 Composer 的聚焦快捷键提示遮挡占位符文字的问题 ([#6200](https://github.com/getpaseo/paseo/pull/6200))
- 修复当前工作区所在的侧边栏分组折叠时,归档快捷键失效的问题 ([#10356](https://github.com/getpaseo/paseo/pull/10356),贡献者:[@hiennd-ctrl](https://github.com/hiennd-ctrl), [@neihyud](https://github.com/neihyud))
- 修复一个会话正在归档工作区时,其他会话和插件仍将其显示为活跃状态的问题 ([#10391](https://github.com/getpaseo/paseo/pull/10391),贡献者:[@tabulius-ux](https://github.com/tabulius-ux))
- 修复在 Android 上输入文字或缩放对话图片后,首次点击 Composer 按钮无效的问题 ([#10057](https://github.com/getpaseo/paseo/pull/10057), [#10308](https://github.com/getpaseo/paseo/pull/10308))
- 修复在 Android 浏览器中滑动终端时弹出已关闭的键盘的问题 ([#10225](https://github.com/getpaseo/paseo/pull/10225),贡献者:[@tonylt](https://github.com/tonylt))
- 修复在 Android 浏览器中滑动终端时弹出键盘的问题 ([#10225](https://github.com/getpaseo/paseo/pull/10225),贡献者:[@tonylt](https://github.com/tonylt))
- 修复启用“折叠轮次”时,空闲的 Provider 通知导致已完成的 OMP 或 Pi 回复被隐藏的问题 ([#10135](https://github.com/getpaseo/paseo/pull/10135))
- 修复 Agent 响应后台任务时,在之前的完成状态未读的情况下重复发送完成通知的问题 ([#10200](https://github.com/getpaseo/paseo/pull/10200))
- 修复 Agent 响应后台任务时重复发送完成通知的问题 ([#10200](https://github.com/getpaseo/paseo/pull/10200))
- 修复空对话在回复中途重连后丢失提示词和回复开头的问题 ([#9723](https://github.com/getpaseo/paseo/pull/9723),贡献者:[@ragokan](https://github.com/ragokan))
- 修复切换 Claude 思考级别时停止后台任务的问题 ([#10008](https://github.com/getpaseo/paseo/pull/10008),贡献者:[@gilad-rubin](https://github.com/gilad-rubin))
- 修复共享 OpenCode 服务停止后轮次一直显示为运行中的问题 ([#10187](https://github.com/getpaseo/paseo/pull/10187))
@@ -64,7 +95,7 @@
- 修复 Codex 消息在轮次开始前停止时显示错误而非已取消的问题 ([#9931](https://github.com/getpaseo/paseo/pull/9931),贡献者:[@3ae3ae](https://github.com/3ae3ae))
- 修复 Git 元数据监听达到上限时无法打开“更改”的问题 ([#10188](https://github.com/getpaseo/paseo/pull/10188))
- 修复将选中的回复复制到纯文本目标时出现多余 Markdown 语法的问题 ([#10079](https://github.com/getpaseo/paseo/pull/10079),贡献者:[@danberindei](https://github.com/danberindei))
- 修复在预览和源码之间切换时文件视图丢失滚动位置的问题,iOS 的 HTML 预览除外 ([#10195](https://github.com/getpaseo/paseo/pull/10195),贡献者:[@apple-ouyang](https://github.com/apple-ouyang))
- 修复在预览和源码之间切换时 Markdown 预览和源码视图丢失滚动位置的问题 ([#10195](https://github.com/getpaseo/paseo/pull/10195),贡献者:[@apple-ouyang](https://github.com/apple-ouyang))
- 修复拆分窗格中的文件标签页显示或聚焦错误文件的问题 ([#9888](https://github.com/getpaseo/paseo/pull/9888),贡献者:[@KosmoCHE](https://github.com/KosmoCHE))
- 修复触摸屏上的子菜单打开后立即关闭的问题 ([#10175](https://github.com/getpaseo/paseo/pull/10175),贡献者:[@christiantwu](https://github.com/christiantwu))
- 修复关闭另一个使用过麦克风的对话后无法听写的问题 ([#10185](https://github.com/getpaseo/paseo/pull/10185),贡献者:[@Brandon168](https://github.com/Brandon168))
@@ -74,13 +105,14 @@
- 修复 Mermaid 流程图标签含有 `@` 时显示源码而非图表的问题 ([#10165](https://github.com/getpaseo/paseo/pull/10165),贡献者:[@monarchmaisuriya](https://github.com/monarchmaisuriya))
- 修复隐藏的后台工作区触发程序坞徽标和网站图标状态的问题 ([#9983](https://github.com/getpaseo/paseo/pull/9983))
- 修复桌面侧边栏导航占满工作区列表空间的问题 ([#9903](https://github.com/getpaseo/paseo/pull/9903),贡献者:[@pentatonic-alex](https://github.com/pentatonic-alex))
- 修复仓库默认分支显示历史已关闭 PR 徽标的问题 ([#10053](https://github.com/getpaseo/paseo/pull/10053))
- 修复仓库默认分支显示已关闭 PR 徽标的问题 ([#10053](https://github.com/getpaseo/paseo/pull/10053))
- 修复 Provider 关闭会话后插件 Agent 无法重新打开的问题 ([#10153](https://github.com/getpaseo/paseo/pull/10153))
- 修复插件页面和设置无法使用插件状态钩子的问题 ([#10049](https://github.com/getpaseo/paseo/pull/10049))
### 插件
- 新增插件导航 `focusHost` 选项,打开工作区或 Agent 时在侧边栏选中目标主机 ([#5574](https://github.com/getpaseo/paseo/pull/5574),贡献者:[@discwl](https://github.com/discwl))
- 新增插件对 worktree 创建、移除和恢复的管理能力 ([#10278](https://github.com/getpaseo/paseo/pull/10278))
- 新增插件导航 `focusHost` 选项,用于在侧边栏选中目标主机 ([#5574](https://github.com/getpaseo/paseo/pull/5574),贡献者:[@discwl](https://github.com/discwl))
## 0.16.2 - 2026-11-18
+21 -5
View File
@@ -150,13 +150,29 @@ Users can also detach an existing subagent from the subagents track. Detach is d
`notifyOnFinish` defaults to `true` for agent-scoped creation and prompt follow-ups because most delegated work needs to report back to the creating agent. Set it to `false` for explicit reporting or no completion callback.
Agent-originated messages and automatic notifications share a recipient-daemon delivery module. A queued receipt acknowledges admission before provider dispatch. Idle recipients retain the two-second quiet window. While a recipient has a running or starting turn (including compaction), messages stay queued across quiet windows and are delivered together after that turn settles, including failure or cancellation. A ready batch is released when the daemon observes the recipient settle; the two-second timer also retries it. Agent callers using `steer`, including the default, wait for the turn to end; only explicit `interrupt` replaces a running turn. An explicit interrupt or a command the provider accepts out of band runs before earlier queued messages. Those messages stay queued and are delivered after the running or replacement turn ends. Human prompts dispatch directly with their existing behavior. Conflicting session-mode choices retain separate ordered dispatches.
Agent-originated messages and automatic notifications share a recipient-daemon delivery module. A queued receipt acknowledges admission before provider dispatch. The module keeps one ordered queue per recipient, and the recipient reads its prompts in the order the daemon admitted them. Nothing overtakes an older queued message. A prompt takes its place when the daemon admits it, even while the provider is still starting an earlier turn, and a message stays in the queue until the turn that carries it starts. The turn has started when the agent manager acknowledges the run start; a turn that fails after that point is not delivered again.
The pending message window is in memory. Clean shutdown flushes it before closing agents; an abrupt
crash can lose queued prompts. A durable send receipt remains pending until actual dispatch
An agent sender picks `queue`, `steer`, or `interrupt` with `activeTurnBehavior`. `steer` and `interrupt` mean what they mean for a person. A person's send has no `queue`. Each kind of prompt is released this way:
- **`queue`** (the agent default) waits. An idle recipient gets the queue after two quiet seconds; steady arrivals cannot hold that window open past ten seconds. While the recipient has a running or starting turn (including compaction), the queue waits and is delivered together when that turn ends on its own or fails. The daemon releases it when it observes the recipient settle; a two-second timer also retries. Finish notifications and change-request updates always queue.
- **`steer`** waits for neither the quiet window nor the turn. It is dispatched at admission as one prompt that enters the running turn: the messages queued before it, then the steer. A provider that cannot steer, or that refuses this steer, gets the prompt as a replacing turn instead, the same as a person's steer.
- **`interrupt`** is dispatched at admission the same way, as one prompt that replaces the running turn. An older queued message cannot start a turn after the newer `interrupt` that countermands it.
- **A person's prompt** (app, CLI, top-level MCP) is dispatched directly with its existing steer or interrupt behavior and takes the queue along the same way: the messages admitted before it first, then the person's text, in one provider prompt. Messages admitted after it follow it. The person's text keeps its own transcript row. When the turn does not start, the messages return to the queue in their place. That covers a start the provider rejects and a start that is not acknowledged within 60 seconds; a provider that starts the turn after those 60 seconds receives the messages twice.
- **Stop** (the stop button, Escape, `cancel_agent`, a Hub interrupt) starts nothing. Messages admitted before the Stop are held, not dropped, including an interrupt whose turn had not started yet. They go ahead of the recipient's next prompt, in that prompt's turn: a person's prompt, or an agent message or notification admitted after the Stop. A held steer or interrupt no longer enters the turn; it waits like a queued message. A queue step or dispatch decided before the Stop is void, and the queue is held again when the message that reopened it is withdrawn. A replacing prompt, a reload, and a rewind are not a Stop. Neither is `cancel_agent` on a recipient with no running or starting turn: it stops nothing, reports that it canceled no run, and a queued message keeps its quiet window.
The queued receipt carries the applied behavior and guidance text for the sender. The text comes from two facts known at admission: whether the recipient has a running turn, and whether its provider can steer (`AgentManager.canSteer`, which is true when the session offers `steerActiveTurn`). It does not wait for the outcome. Guidance for a message queued behind a running turn tells the sender to send a short `steer` or `interrupt` follow-up instead of resending, because the follow-up carries the queued message ahead of it and a resend would be read twice. The MCP tool, the WebSocket `send_agent_message_response`, the CLI, and the connected-host path all relay this one text.
Older CLIs and connected hosts send an explicit `steer` for every default agent send. The daemon reads an agent sender's `steer` as `queue` unless the client advertises the `agent_active_turn_behaviors` capability. A new client talking to a daemon without `server_info.features.agentActiveTurnBehaviors` sends no behavior for `queue` and refuses a `steer`.
A command the provider accepts out of band is the one exception to admission order: it runs alone without ending the turn. A queued command runs ahead of older messages, and a `steer` or `interrupt` that runs this way takes no queued message along. A person's command leaves the queue where it is. A prompt is a command when it starts with a slash and a name that has no second slash, so `/compact` is one and `/tmp/build.log is stale` is plain text.
One provider prompt has one session mode. A batch released by a steer, an interrupt, or a person's prompt uses the latest explicit choice in it, the person's own first. Queued messages with conflicting choices keep separate ordered dispatches.
The pending message window is in memory. Clean shutdown flushes it, held messages included, before
closing agents; an abrupt crash can lose queued prompts. A durable send receipt remains pending until actual dispatch
acknowledgment, so a lost window never becomes a completed receipt. Completion waits cover queued
admission within their original timeout and cancellation scope, and return permission checkpoints
while messages are waiting.
admission within their original timeout and cancellation scope, and return at permission
checkpoints and when a Stop holds the queue.
Finish notification intent has a separate daemon-owned lifetime. The finish-subscriptions module
persists caller, child, host, and ownership policy through an injectable store; its JSON adapter uses
+1
View File
@@ -102,6 +102,7 @@ Each agent is stored as a separate JSON file, grouped by project directory.
| `config` | `SerializableConfig?` | Agent session configuration (see below) |
| `runtimeInfo` | `RuntimeInfo?` | Live runtime state (see below) |
| `features` | `AgentFeature[]?` | Provider-reported features (toggles/selects) |
| `sessionMetrics` | `AgentMetric[]?` | Provider-reported session metrics, already formatted for display; the latest report replaces the list |
| `persistence` | `PersistenceHandle?` | Handle for resuming sessions |
| `lastError` | `string?` (nullable) | Last error message, if any |
| `requiresAttention` | `boolean?` | Whether the agent needs user attention |
+2 -2
View File
@@ -78,7 +78,7 @@ In the composer, the first glyph's ink starts on the text column's edge, and the
One gate, `resolveControlSizing`, picks touch or pointer sizing: touch when the layout is compact or touch is the primary input. An iPad with the sidebar docked is wide and touch-sized; a desktop is pointer-sized; a phone is both compact and touch. The primary input comes from `(pointer: coarse)` (`hasTouchPrimaryPointer` in `hooks/pointer-capabilities.ts`), not from whether a touch screen exists: WebKit keeps touch primary on an iPad with a trackpad attached, and Chromium reports a fine primary pointer for a laptop with a touch screen that its trackpad drives. The fact is read once at startup: attaching a pointer to a tablet does not shrink its controls, and a 2-in-1 launched with a pointer keeps pointer sizes after it switches to tablet mode until the app reloads. Stylesheets take sizes from `bySizing({ touch, pointer })`, which resolves to a breakpoint value, so a caller never sees which sizing it got. Rows, icon buttons and icon slots take theirs from `styles/density.ts`, which uses the same gate. An `{ xs, md }` literal for a hit target or glyph is wrong: on an iPad it yields the pointer size.
Buttons are pills (`borderRadius.full`) with medium-weight labels; an icon-only button is a circle. `<Switch>` follows the same two sizes: a 34×20 track by default and in `sm`, 44×26 where controls are touch-sized, and it reserves its size's control height so a row holding one lines up with a row holding a button. Buttons, text fields, dropdown triggers, and `<SegmentedControl>` (`packages/app/src/components/ui/segmented-control.tsx`) share the size contract — the same size name means the same height and label size on every control kind, so controls in one row line up. Segments run tighter than a button, because the gap between segments already reads as padding. The selected segment is a `surface3` fill with `foreground` text, not an inverted one — inverting it inside thin chrome puts a white slab in the toolbar. A settings row sizes the controls in its trailing slot itself, default in a card and `sm` when nested (`packages/app/src/components/ui/control-size.tsx`), so don't pass a size to a button, dropdown or field placed in a row. Code that computes with a control's height outside a stylesheet uses `controlHeightForLayout`. Never shrink a control's font or padding locally to fit a context — if the context needs a smaller control, the size tier is missing or the wrong one is in use.
Buttons are pills (`borderRadius.full`) with medium-weight labels; an icon-only button is a circle. `<Switch>` follows the same two sizes: a 34×20 track by default and in `sm`, 44×26 where controls are touch-sized, and it takes presses over its size's control height without adding that height to its row. Buttons, text fields, dropdown triggers, and `<SegmentedControl>` (`packages/app/src/components/ui/segmented-control.tsx`) share the size contract — the same size name means the same height and label size on every control kind, so controls in one row line up. Segments run tighter than a button, because the gap between segments already reads as padding. The selected segment is a `surface3` fill with `foreground` text, not an inverted one — inverting it inside thin chrome puts a white slab in the toolbar. A settings row sizes the controls in its trailing slot itself, default in a card and `sm` when nested (`packages/app/src/components/ui/control-size.tsx`), so don't pass a size to a button, dropdown or field placed in a row. Code that computes with a control's height outside a stylesheet uses `controlHeightForLayout`. Never shrink a control's font or padding locally to fit a context — if the context needs a smaller control, the size tier is missing or the wrong one is in use.
A `<Pressable>` wrapping a `<Text>` is a sixth variant. It is wrong. `<Button>` accepts `style`, `textStyle`, `leftIcon`, `disabled`, `size`, and `variant`.
@@ -134,7 +134,7 @@ A section or group explains itself through the `info` prop on `<SettingsSection>
Cards inside a section sit closer than sections. Rows inside a card touch — only the divider separates them. The rhythm is page → spacious; section → spacious; card → tight.
Rows have generous vertical padding: roughly 16px of content plus 16px of vertical padding for settings rows, 8–12px for sidebar list items where many rows must fit. A row that opens something (`<SettingsDetailRow>`, a collapsible row's header) is at least as tall as one holding a default control, so a list where only some of those rows carry a switch keeps one row height. Other rows take their content's height. Compressing rows below the established density to fit more on the screen is wrong. Too many rows means more cards or more sections, not smaller rows.
Rows have generous vertical padding: roughly 16px of content plus 16px of vertical padding for settings rows, 8–12px for sidebar list items where many rows must fit. Every row is at least as tall as one holding a switch, so text rows, switch rows and rows that open something (`<SettingsDetailRow>`, a collapsible row's header) line up. A row holding a button or a select is taller. Compressing rows below the established density to fit more on the screen is wrong. Too many rows means more cards or more sections, not smaller rows.
Rows revealed under a `<SettingsCollapsibleRow>` are nested rows, one step smaller than card rows: 8px of vertical padding, `fontSize.sm` text, `sm` controls, and one fixed height so rows with and without a trailing button line up. The collapsible row sets this for everything under it (`packages/app/src/components/settings/row-density.tsx`). Never pass a density to a row, and never pad or resize a nested row by hand. Settings rows that share a menu surface with menu rows, such as Usage's options, go inside `<MenuSettingsRows>`, which puts their labels on the menu rows' label rail (`menuRowContentInset`).
+1 -1
View File
@@ -57,7 +57,7 @@ Authoritative terminology. UI label wins. Don't invent synonyms; use what's here
- **Agent controls** — Provider, model, mode, thinking, and provider-feature controls for an agent or draft agent. Code: `AgentControls` / `DraftAgentControls` (`packages/app/src/composer/agent-controls/index.tsx`). Forbidden: "Agent status bar".
- **Composer footer** — Optional area rendered below the composer input but still inside the keyboard-shifted composer layout. Code: `Composer.footer` (`packages/app/src/composer/index.tsx`).
- **Composer track** — A contextual lane at the foot of a pane, by the composer input. Specific tracks use the `<thing> track` form: **Queue track**, **Subagents track**, **Tasks track**. The subagents and tasks tracks are pills floating over the transcript, each opening a panel with its rows; the queue track is a lane inside the composer itself. Code: queue track inside `Composer` (`packages/app/src/composer/index.tsx`), pill and panel frame in `packages/app/src/composer/tracks.tsx`, `SubagentsTrack` (`packages/app/src/subagents/track.tsx`), `AgentTaskList` (`packages/app/src/composer/task-list/index.tsx`).
- **Steer** — A follow-up admitted into an agent's already-running turn instead of interrupting it. Interrupting the turn discards steers the provider has not read yet; there is no take-back short of that, and no provider offers one. The stop button and Escape both interrupt. A steer the provider never read keeps its transcript row, matching Codex. Sending while the agent waits on a permission denies that request first, so the message reaches the same turn instead of stalling behind it. UI: "Send and steer". Code: `steerActiveTurn` (`packages/server/src/server/agent/agent-sdk-types.ts`). Don't confuse with **Queue track**, which holds messages client-side until the turn ends and never enters the running turn. Forbidden: "Delivery", "Queued prompt".
- **Steer** — A follow-up admitted into an agent's already-running turn instead of interrupting it. Interrupting the turn discards steers the provider has not read yet; there is no take-back short of that, and no provider offers one. The stop button and Escape both interrupt. A steer the provider never read keeps its transcript row, matching Codex. Sending while the agent waits on a permission denies that request first, so the message reaches the same turn instead of stalling behind it. UI: "Send and steer". Code: `steerActiveTurn` (`packages/server/src/server/agent/agent-sdk-types.ts`). Don't confuse with **Queue track**, which holds messages client-side until the turn ends and never enters the running turn. An agent sender's `activeTurnBehavior: "steer"` (CLI `--behavior steer`) means the same as a person's steer. Agent senders default to `queue` instead: the daemon holds the message until the running turn ends, as the queue track does for a person. Forbidden: "Delivery", "Queued prompt".
- **Subagent** — User-facing term for an agent session related to a parent agent session. Use **subagents** in UI copy and docs. Internal daemon/provider plumbing may say "child agent" or `child_session`, especially for provider-managed imports; do not surface "child agent" as a product term.
- **Attachment tray** — The selected-attachments row inside the composer input, above the text input. Code: `renderAttachmentTray` (`packages/app/src/composer/index.tsx`). Forbidden: "Attachment bar".
- **Fork** — Starting a NEW agent session seeded with a copy of another session's curated chat history. The history rides along as a `chat_history` composer attachment on a draft; the fork does **not** resume or branch the provider-side session, and the source session keeps running untouched. Two targets, one surface, one RPC (`agent.fork_context`, gated on the `agentForkContext` host feature). UI: the fork menu lives in the assistant turn footer and offers "Fork in a new tab" / "Fork in a new workspace". On a **completed** turn it sits beside the copy button and pins a boundary to that turn, so the fork stops where the turn ended. On the **in-flight** turn it sits beside the progress loader and elapsed timer, and omits the boundary entirely — `selectForkContextRows` projects the whole timeline when no boundary is given, so the fork captures everything **up to now**, including the response still streaming. That omission is what makes mid-run forking work; pinning a boundary there would silently drop the live response. Forbidden on provider-subagent (read-only) panes. Code: `useForkAgent` (`packages/app/src/hooks/use-fork-agent.ts`), `selectForkContextRows` (`packages/server/src/server/agent/activity-curator.ts`). Don't confuse with: git fork (unrelated; Paseo has no such concept), **Worktree** (filesystem isolation, not history copying).
+3 -2
View File
@@ -72,8 +72,9 @@ hashes, never prompts, environment values, or credentials.
Before messaging an archived workspace, call `workspace.recovery.inspect.request`, then
`workspace.recovery.restore.request` and await success. The native message handler unarchives the
agent and loads its persisted provider session. `activeTurnBehavior: "steer"` uses the provider's
native steering behavior, including Paseo's existing behavior when that provider cannot steer.
agent and loads its persisted provider session. A Hub message is a direct send (no `sourceAgentId`), so
`activeTurnBehavior: "steer"` uses the provider's native steering and interrupts the running turn
when that provider cannot steer. The daemon rejects `queue` on a direct send.
Execution completion and arrival-specific output authority remain Hub responsibilities.
The older `hub.execution.*` RPCs remain accepted for existing clients. Their execution ownership
+23 -2
View File
@@ -221,11 +221,32 @@ Plugin callbacks follow the same contract; see [plugin providers](plugins.md#con
`ProviderSnapshotManager` owns one refresh deadline per provider. The deadline starts before the
availability check and covers that check plus the complete catalog probe. Providers that make
multiple catalog requests must not apply this deadline separately to each request. The manager
aborts the shared refresh signal at the deadline. Providers name active catalog operations and
aborts the shared refresh signal at the deadline, and when it stops. Shutdown waits for running
probes to settle before it shuts down provider clients, so a probe cannot start provider work,
such as an OpenCode server, after its client has cleaned up. Providers name active catalog operations and
finish subprocess, server, or session cleanup before rejecting. Timeout errors list the operations
that were still active when the deadline expired.
Catalogue results stay cached by identity until explicit refresh or a change to that provider's configuration.
Catalogue results stay cached by identity until a refresh or a change to that provider's configuration.
Providers change their model lists on the server without a CLI update, so the daemon refreshes
each enabled provider about every 30 minutes (`ProviderRefreshSchedule`). One timer ticks
`30 min / enabled providers` apart, and each tick starts at most one refresh: the provider refreshed
longest ago, if that was a full interval ago. Ticks do not wait for refreshes, so a provider that
probes many workspace targets does not delay the others, and provider CLIs never start in a burst.
Any forced refresh, configuration change, or newly enabled provider counts as a refresh and pushes
that provider's next one back a full interval.
The background refresh takes the settings refresh path for one provider, with these differences:
- It joins discovery already in flight.
- It passes `force: false`, because providers treat `force` as a requested hard refresh. OpenCode
rotates its shared server on it.
- An unchanged catalogue keeps the published record and its `fetchedAt`, so idle clients receive
no update.
- A result that is not ready, or a catalogue the provider marks `incomplete`, keeps a ready
catalogue. Plugin status probes report a failed startup as unavailable, so that counts as
transient. Set `ProviderCatalog.incomplete` when discovery fails and you fall back to a partial
list, as Claude does when model discovery times out. A manual refresh still publishes all three.
Keys are resolved on each read so project configuration can select a different cached catalogue.
Selector opening may read a loading or stale query, but does not force provider probing.
+187 -111
View File
@@ -6,21 +6,26 @@ All workspaces share one version and release together.
A release has exactly two steps. The agent does the first, the user authorizes the second.
**Preparation** (reversible — agent does this):
**Preparation** (reversible — agent does this on the release branch):
- format, lint, typecheck all green
- resolve the [release base](#release-base-and-source), get the [release inventory](#release-inventory) approved when one applies, and confirm CI for the resulting release source is green
- resolve the [release base](#release-base-and-source) and get the [release inventory](#release-inventory) approved when one applies
- classify the diff from the previous stable to the release source as patch or minor, then show the
target version and rationale to the user
- draft the changelog, show it to the user, wait for review
- run the pre-release sanity check, surface findings to the user
- draft `CHANGELOG.md` and `CHANGELOG.zh-CN.md`, show the draft to the user, and wait for review
- run the pre-release sanity check (stable only) and surface findings to the user
- commit the approved changelog and any requested ACP update, run format, lint, and typecheck, then
run `npm run release:prep:<mode>` to create the `chore(release): cut <version>` commit without a tag
- push the `release/vX.Y.Z` or `release/vX.Y.Z-beta.N` branch and confirm its `CI` and
`Desktop Packages` runs pass (never wait on `main` CI)
**Go-ahead** (user says "go ahead"):
- commit the approved release inputs locally on the release branch
- run the release, which publishes npm and pushes the release branch and tag
- run `npm run release:ship` (stable) or `npm run release:ship:beta` (beta), which publishes npm
with retry and registry visibility checks, advances the `beta` dist-tag on stable when newer, and
pushes the tag on the already-verified commit
- create the release heartbeat immediately and babysit it to completion
- [reconcile `main`](#reconcile-main) with what shipped
- open the [reconcile `main`](#reconcile-main) PR right away so its CI runs in parallel with release
builds, then delete the remote release branch once the PR merges
Rules that apply to both steps:
@@ -87,29 +92,42 @@ An explicit user instruction can select another base.
1. Fetch `origin` with tags and resolve the base to one commit.
2. Create the release workspace through Paseo MCP on a new branch from the base, named
`release/vX.Y.Z` or `release/vX.Y.Z-beta.N` for the version being cut. Run the release from
this workspace: `release:push` pushes `HEAD` to the current branch.
`release/vX.Y.Z` or `release/vX.Y.Z-beta.N` for the version being cut. `release:push` refuses
`main` and any branch outside `release/*`.
3. Cherry-pick the approved commits in their `main` order with `git cherry-pick -x`, so each
commit records the `main` commit it came from. Include the prerequisites the inventory listed.
Resolve conflicts within the approved changes and report how. A resolution that needs new
behavior is a scope addition: surface it for selection.
4. Confirm CI for the release source. When the source is a commit whose CI already passed (the
captured `origin/main`, or the beta tag being promoted), use that run and watch pending CI to
completion. Otherwise push the release branch and run `gh workflow run ci.yml --ref <branch>`,
because CI push runs target only `main`.
5. Keep the changelog, any explicitly requested ACP catalog update, lockfile preparation, and the
version commit local until the go-ahead. The release command pushes the release branch and tag
as one batch, and the tag push starts the release workflows. Then run
`gh workflow run ci.yml --ref <branch>` so CI covers the complete release commit.
6. [Reconcile `main`](#reconcile-main).
4. Commit the approved `CHANGELOG.md`, `CHANGELOG.zh-CN.md`, and any explicitly requested ACP
catalog update on the release branch. Run `npm run format`, `npm run lint`, and
`npm run typecheck`, then run `npm run release:prep:<mode>`. This checks the packages, bumps all
workspace versions, generates stable F-Droid changelogs, updates `package-lock.json`, and creates
`chore(release): cut <version>` without a tag.
5. Push the release branch (`git push -u origin HEAD`). Pushing any `release/**` branch triggers
`CI` (all 21 jobs) and `Desktop Packages` (Linux user-namespace smoke) in their own concurrency
groups, isolated from `main` traffic. Never wait on `main` CI, because `main` cancels in-progress
runs on every merge. Wait for both runs on the release branch commit to pass before shipping.
Pushing a `release/**` branch without a `v*` tag triggers no publish workflows.
6. After go-ahead, run `npm run release:ship` (stable) or `npm run release:ship:beta` (beta). This
publishes to npm and pushes the `v*` tag on the already-verified commit. Because the release
branch is already pushed at `HEAD`, no second CI run starts and only the tag release workflows
run.
7. Start [Reconcile `main`](#reconcile-main) immediately so its PR checks overlap with desktop and
EAS builds.
When promoting or branching from a tag created before `release:prep:*`, `release:ship`, and the
`release/**` workflow triggers existed (for example `v0.17.0-beta.1`), either cherry-pick the
release-tooling commit onto the release branch (leaving product code unchanged) or follow the
older tag's scripts (`version:all:*`, manual `gh workflow run ci.yml --ref <branch>`,
`release:publish`, and `release:push`) until the next tag carries the new tooling.
A stable hotfix can take the version a patch beta line would promote to, such as `0.13.3`
shipping while `0.13.3-beta.2` soaks. Never republish or reuse a shipped version. Continue the line
as the next unused patch beta, `0.13.4-beta.1`, from the tested beta tag. No `version:all:*` mode
computes that from a prerelease, so follow the [manual steps](#manual-step-by-step) with
`npm version 0.13.4-beta.1 --include-workspace-root --message "chore(release): cut %s"` and
`npm run release:publish:beta`. Record the tested beta tag as its source so its soak still counts
for the unchanged content.
computes that from a prerelease, so run `npm run release:check`, bump with
`npm version 0.13.4-beta.1 --include-workspace-root --no-git-tag-version && git commit -m "chore(release): cut 0.13.4-beta.1"`,
push the release branch for `CI` and `Desktop Packages`, then ship with `npm run release:ship:beta`.
Record the tested beta tag as its source so its soak still counts for the unchanged content.
## Release inventory
@@ -137,19 +155,31 @@ For a promotion, every item defaults to defer.
## Reconcile main
After the release push, open a PR into `main` that brings in what shipped and keeps newer
development on `main`:
Right after the tag push (in parallel with desktop and EAS builds), open a PR into `main` from a
fresh branch off `origin/main` named `reconcile/vX.Y.Z` or `reconcile/vX.Y.Z-beta.N`:
- The shipped `CHANGELOG.md` and `CHANGELOG.zh-CN.md` entries, exactly as tagged. A beta or
promotion entry replaces the series entry on `main`. A stable hotfix on an older line goes
below any newer entries.
- The shipped version in every `package.json` and `package-lock.json`, unless `main` already
carries a newer version, plus the F-Droid changelog files a stable release generates.
- No claims about unshipped work: a shipped entry describes only the tagged content, and notes
`main` holds for changes that have not shipped stay out of it.
```bash
git fetch origin --tags
git checkout -b reconcile/vX.Y.Z origin/main
npm run release:reconcile -- --tag vX.Y.Z
```
Delete the release branch once the PR merges. The tag is the record of shipped source; Paseo
maintains no release track besides `main`.
`scripts/reconcile-release.mjs` updates the working tree without cherry-picking stale lockfile
hunks from the release base:
- Copies the tagged `CHANGELOG.md` and `CHANGELOG.zh-CN.md` entries verbatim. A beta or promotion
entry replaces the series beta entry on `main`. A stable hotfix on an older line is placed below
any newer entries.
- Copies the generated F-Droid changelog files from the tag for stable releases.
- Updates every `package.json` version and regenerates `package-lock.json` on top of current `main`
only when the shipped version is newer than `main`'s version.
- Leaves unshipped work on `main` untouched and makes no claims about unshipped work in the
changelog entry.
Commit, open the PR into `main`, and merge once its checks pass. After the reconcile PR merges and
the release completes, delete the remote release branch (`git push origin --delete release/vX.Y.Z`
or `release/vX.Y.Z-beta.N`). The tag is the record of shipped source; Paseo maintains no release
track besides `main`.
## ACP catalog updates
@@ -189,35 +219,42 @@ as described in **Fixing a failed release build**.
## Standard release (stable)
Before running any stable release command:
Before preparing a stable release on `release/vX.Y.Z`:
- Make sure the release source passed CI, the approved release inputs are committed locally on the release branch, and the working tree is clean.
- **Run `npm run format`, `npm run lint`, and `npm run typecheck` and commit any resulting changes BEFORE you start any `release:*` command.** `release:check` runs `npm install --workspaces --include-workspace-root` as part of `release:prepare`, which can mutate `package-lock.json` (e.g. churning `"dev": true` markers on optional deps). The next step, `version:all:*`, runs `npm version` which aborts when the working tree is dirty. If this happens mid-flight you have to commit the lockfile churn before retrying — and the pre-commit format hook will reject a lockfile-only commit because oxfmt internally skips `package-lock.json` while lefthook's glob still matches it. Avoid the whole mess by running format/lint/typecheck first, then `release:prepare` once on its own to absorb any lockfile churn into a normal commit, then start the release.
- Do not use a release command as a substitute for checking whether the current commit is actually ready.
- Commit the approved `CHANGELOG.md`, `CHANGELOG.zh-CN.md`, and any requested ACP update.
- Run `npm run format`, `npm run lint`, and `npm run typecheck`, commit any resulting changes, and
confirm the working tree is clean.
- Do not use a release command as a substitute for checking whether the current commit is ready.
```bash
# Run exactly one, matching the approved decision:
npm run release:patch
npm run release:minor
# 1. Preparation (reversible): run checks and commit the version bump without tagging.
npm run release:prep:patch
# or:
npm run release:prep:minor
# 2. Push the release branch and wait for CI + Desktop Packages on this commit.
git push -u origin HEAD
# 3. After go-ahead and green release-branch checks: publish npm and push the tag.
npm run release:ship
```
This bumps the version across all workspaces, runs checks, publishes to npm, and pushes the branch + tag. The tag push triggers `Desktop Release`, `Android APK Release`, `Docker`, and `Release Notes Sync` on GitHub Actions. The workflows create the GitHub Release as a draft while builds and release-note sync run. EAS picks up the same tag via the EAS GitHub app and starts the iOS + Android store builds in parallel (see "Mobile builds (EAS)" below) — there is no mobile-release workflow under `.github/workflows`.
`release:ship` runs `scripts/publish-release-packages.mjs` and `scripts/push-current-release-tag.mjs`:
After the stable release succeeds, check `npm view @getpaseo/cli dist-tags`. When the new stable
is newer than `beta`, move `beta` to it for every published package. This changes dist-tags only;
do not republish the packages:
- Publishes all seven `@getpaseo/*` packages in dependency order, skips versions already on npm,
retries transient 429/5xx/network errors, and waits until each version is visible on npm before
continuing.
- Checks `npm view @getpaseo/cli dist-tags` and moves npm's `beta` dist-tag to the new stable
version across all seven packages whenever the stable version is newer than `beta` (leaving
`beta` untouched during a stable hotfix on an older line).
- Refuses non-`release/*` branches, tags `HEAD` with `vX.Y.Z`, and pushes the tag. Because `HEAD`
is already on the release branch, no second CI run starts.
```bash
PASEO_VERSION=$(node -p "require('./package.json').version")
for package in highlight relay protocol client plugin server cli; do
npm dist-tag add "@getpaseo/$package@$PASEO_VERSION" beta
done
```
The tag push triggers `Desktop Release`, `Android APK Release`, `Docker`, and `Release Notes Sync` on GitHub Actions. The workflows create the GitHub Release as a draft while builds and release-note sync run. EAS picks up the same tag via the EAS GitHub app and starts the iOS + Android store builds in parallel (see "Mobile builds (EAS)" below) — there is no mobile-release workflow under `.github/workflows`.
Verify `latest` resolves to `PASEO_VERSION`, and `beta` resolves to `PASEO_VERSION` or the
active newer beta, before considering the stable release complete.
Verify `npm view @getpaseo/cli dist-tags` shows `latest` at the shipped version, and `beta` at the shipped version or an active newer beta, before considering the stable release complete.
The Docker workflow builds images from the checked-out source tree on pull requests and on `main` as non-publishing checks. Stable `vX.Y.Z` tag pushes publish `ghcr.io/getpaseo/paseo:X.Y.Z` and `ghcr.io/getpaseo/paseo:latest`; beta `vX.Y.Z-beta.N` tag pushes publish only `ghcr.io/getpaseo/paseo:X.Y.Z-beta.N` and never move `latest`.
The Docker workflow builds images from the checked-out source tree on `main` as non-publishing checks (cancelling superseded `main` runs). Stable `vX.Y.Z` tag pushes publish `ghcr.io/getpaseo/paseo:X.Y.Z` and `ghcr.io/getpaseo/paseo:latest`; beta `vX.Y.Z-beta.N` tag pushes publish only `ghcr.io/getpaseo/paseo:X.Y.Z-beta.N` and never move `latest`.
The production relay is the Elixir service in [getpaseo/paseo-relay](https://github.com/getpaseo/paseo-relay), with its own deployment process. Paseo releases and pushes to this repository do not deploy it. The Cloudflare relay code and workflow in this repository are legacy and are not used in production.
@@ -228,29 +265,43 @@ The production relay is the Elixir service in [getpaseo/paseo-relay](https://git
```bash
npm run typecheck # Verify the exact commit you intend to release
npm run release:check # Typecheck, build, dry-run pack
# Run exactly one approved version command:
# Run exactly one approved version command (commits without tagging):
npm run version:all:patch
npm run version:all:minor
npm run release:publish # Publish to npm
npm run release:push # Push HEAD + tag (triggers CI workflows)
# Then move npm's beta dist-tag to this stable version when it is newer than beta (see above).
git push -u origin HEAD # Trigger CI + Desktop Packages on release/vX.Y.Z
# Wait for CI and Desktop Packages on this branch to pass, then after go-ahead:
npm run release:publish # Publish to npm (idempotent, with propagation + beta dist-tag sync)
npm run release:push # Create tag vX.Y.Z on HEAD and push (triggers release workflows)
```
## Beta flow
```bash
npm run release:beta:patch # Start the next patch beta line
npm run release:beta:minor # Start the next minor beta line
# ... test desktop and APK prerelease assets from GitHub Releases ...
npm run release:beta:next # Optional: cut X.Y.Z-beta.2, beta.3, ...
npm run release:promote # Promote X.Y.Z-beta.N to stable X.Y.Z
# 1. Preparation (reversible) on release/vX.Y.Z-beta.N:
npm run release:prep:beta:patch # Start the next patch beta line
# or:
npm run release:prep:beta:minor # Start the next minor beta line
# or for follow-up betas:
npm run release:prep:beta:next # Cut X.Y.Z-beta.2, beta.3, ...
# 2. Push the release branch and wait for CI + Desktop Packages to pass:
git push -u origin HEAD
# 3. After go-ahead:
npm run release:ship:beta
# Later, after soak, promote on release/vX.Y.Z:
npm run release:prep:promote
git push -u origin HEAD
# Wait for CI + Desktop Packages to pass, then after go-ahead:
npm run release:ship
```
- Beta tags are published GitHub prereleases like `v0.1.41-beta.1`
- Betas publish npm packages with `--tag beta`, so `npm install @getpaseo/cli@beta` opts in while plain `npm install @getpaseo/cli` stays on `latest`
- The user-facing beta instructions, including the public TestFlight join link for the `Paseo Beta` group, live in `public-docs/updates.md` and its `zh-CN` pair. Update them when a beta path is added or removed.
- Betas publish desktop assets and APKs for testing. They also build iOS, upload it to TestFlight, add it to the `Paseo Beta` external group, and submit it for Beta App Review. They do not submit mobile builds to the production stores.
- `release:promote` creates a fresh stable tag like `v0.1.41`; the final release never reuses the beta tag
- `release:prep:promote` + `release:ship` creates a fresh stable tag like `v0.1.41`; the final release never reuses the beta tag
- Desktop assets now come from the Electron package at `packages/desktop`
- Require the Linux artifact CI checks with both restricted and usable user namespaces to pass before publication; see [packaged desktop smoke](testing.md#packaged-desktop-smoke). Keep the installed-package and AppImage checks together.
- Beta releases use Electron's `beta` update channel. Users on the stable channel only receive stable releases; users on the beta channel receive beta releases and the final stable release when it is published.
@@ -278,17 +329,17 @@ Drafts do not appear in GitHub's releases feed. Updater clients continue to see
### Default behavior
`npm run release:patch` or `npm run release:minor` → tag push → 36h ramp. No extra action needed.
`npm run release:prep:patch` or `npm run release:prep:minor` → push release branch → green checks → `npm run release:ship` → tag push → 36h ramp. No extra action needed.
The `rollout_hours` input on `desktop-release.yml` is **only read on `workflow_dispatch`** — tag-push runs always default to 36. To get any other rollout duration on a fresh release, use the post-publish flip below.
### Instant-admit release (rollout_hours=0)
`npm run release:patch` and `npm run release:minor` tag-push, and tag-push runs always stamp 36 hours. To admit everyone immediately (hotfix, doc-only change, a release you want out fast), cut the release normally, wait for it to publish, then flip the manifests:
Tag-push runs from `npm run release:ship` always stamp 36 hours. To admit everyone immediately (hotfix, doc-only change, a release you want out fast), ship the release normally, wait for it to publish, then flip the manifests:
```bash
# 1. Cut and publish (36h ramp from tag push).
npm run release:patch
# 1. Ship the verified release branch (36h ramp from tag push).
npm run release:ship
# 2. Wait for Desktop Release to publish the GitHub Release.
gh run watch <desktop-release-run-id> --exit-status
@@ -339,7 +390,7 @@ gh workflow run desktop-release.yml \
-f rollout_hours=6
```
This does **not** apply to fresh releases cut via `npm run release:patch` or `npm run release:minor` — those paths always tag-push and stamp 36. For a fresh release with a custom ramp, cut normally and then dispatch `desktop-rollout.yml` (same pattern as the instant-admit flow above, with your chosen `rollout_hours`).
This does **not** apply to fresh releases shipped via `npm run release:ship` — tag-push runs always stamp 36. For a fresh release with a custom ramp, ship normally and then dispatch `desktop-rollout.yml` (same pattern as the instant-admit flow above, with your chosen `rollout_hours`).
### Releasing during an active rollout
@@ -449,10 +500,11 @@ that resumes the release in the current conversation. Create it automatically
with `create_heartbeat`. The heartbeat owns the release until it either reaches
the completion checklist or finds a failure that needs new user authority.
Each heartbeat checks the release tag commit, all GitHub Actions runs for the
release branch and tag, npm dist-tags, the GitHub Release body and assets,
desktop updater manifests, the published Docker image, and the applicable EAS
workflow. Inspect the GitHub Release itself and confirm that the macOS, Linux,
Each heartbeat checks the release tag commit, `CI` and `Desktop Packages` for the release branch,
every tag-triggered GitHub Actions run (`Desktop Release`, `Android APK Release`, `Docker`,
`Release Notes Sync`), npm dist-tags, the GitHub Release body and assets, desktop updater
manifests, the published Docker image, the applicable EAS workflow, the reconcile PR, and remote
release branch cleanup. Inspect the GitHub Release itself and confirm that the macOS, Linux,
Windows, and Android APK assets are present along with the channel manifests
(`latest-mac.yml`, `latest-linux.yml`, and `latest.yml` for stable;
`beta-mac.yml`, `beta-linux.yml`, and `beta.yml` for beta).
@@ -461,8 +513,9 @@ For stable releases, also confirm every required mobile build, external
TestFlight distribution, Beta App Review submission, and store submission job
for the release commit. For betas, confirm
the beta EAS workflow completed its TestFlight distribution and Beta App Review
path. Delete the heartbeat only after every applicable checklist item passes,
then report the release as shipped.
path. Once the reconcile PR merges and every applicable checklist item passes,
delete the remote release branch (`git push origin --delete release/vX.Y.Z` or
`release/vX.Y.Z-beta.N`), delete the heartbeat, and report the release as shipped.
Pattern:
@@ -474,7 +527,7 @@ Pattern:
"timezone": "UTC",
"maxRuns": 120,
"expiresIn": "24h",
"prompt": "Resume the vX.Y.Z release babysit for commit <sha>. Check npm tags; every GitHub Actions run for the release branch and tag; the published GitHub Release body, expected desktop/APK assets, and channel manifests; the Docker image; and the matching EAS workflow. Completion requires every applicable checklist item. For stable, require build_ios, distribute_ios (Paseo Beta external distribution and Beta App Review submission), submit_ios_for_review, build_android, and submit_android to succeed. For beta, require the beta TestFlight distribution and Beta App Review path. If work is pending, wait for the next heartbeat. If a failure can be retried safely for the same version, follow the failed-release procedure; otherwise report the blocker. When every applicable completion-checklist item passes, delete THIS heartbeat, report shipped, and stop.",
"prompt": "Resume the vX.Y.Z release babysit for commit <sha>. Check npm tags; CI and Desktop Packages on release/vX.Y.Z; every tag-triggered GitHub Actions run; the published GitHub Release body, expected desktop/APK assets, and channel manifests; the Docker image; the matching EAS workflow; and the reconcile PR into main. Completion requires every applicable checklist item. For stable, require build_ios, distribute_ios (Paseo Beta external distribution and Beta App Review submission), submit_ios_for_review, build_android, and submit_android to succeed. For beta, require the beta TestFlight distribution and Beta App Review path. If work is pending, wait for the next heartbeat. If a failure can be retried safely for the same version, follow the failed-release procedure; otherwise report the blocker. When every applicable completion-checklist item passes and the reconcile PR is merged, delete origin/release/vX.Y.Z, delete THIS heartbeat, report shipped, and stop.",
}
```
@@ -576,12 +629,16 @@ intentionally unavailable to desktop updater clients.
## Notes
- `version:all:*` bumps root + syncs workspace versions and `@getpaseo/*` dependency versions
- The npm `version` lifecycle regenerates F-Droid changelog files from `CHANGELOG.md` for stable releases only (`npm run fdroid:changelogs`) and stages them, so the release tag carries them. Betas are a no-op. A stable run **aborts the release** if `CHANGELOG.md` has no entry for the version being cut — commit the changelog entry first. See [docs/android.md](android.md) for why these files are generated per ABI.
- `release:prepare` refreshes workspace `node_modules` links to prevent stale types
- `npm run dev:desktop` and `npm run build:desktop` target the Electron desktop package in `packages/desktop`
- If `release:publish` partially fails, re-run it — npm skips already-published versions
- If `release:publish:beta` partially fails, re-run it — npm skips already-published versions and keeps prereleases off `latest` because every publish uses `--tag beta`
- `version:all:*` bumps root, syncs workspace versions and `@getpaseo/*` dependency versions, and
commits `chore(release): cut <version>` without creating a tag. `release:push` creates and pushes
the tag atomically after release-branch CI passes.
- `release:push` refuses `main` and any branch outside `release/*`.
- The npm `version` lifecycle regenerates F-Droid changelog files from `CHANGELOG.md` for stable releases only (`npm run fdroid:changelogs`) and stages them, so the release commit and tag carry them. Betas are a no-op. A stable run **aborts the release** if `CHANGELOG.md` has no entry for the version being cut — commit the changelog entry first. See [docs/android.md](android.md) for why these files are generated per ABI.
- `release:prepare` refreshes workspace `node_modules` links to prevent stale types.
- `npm run dev:desktop` and `npm run build:desktop` target the Electron desktop package in `packages/desktop`.
- `release:publish` and `release:publish:beta` run `scripts/publish-release-packages.mjs`: if a run
fails partway through, re-running skips versions already published on npm, retries transient
registry errors, and waits for each version to become visible on npm before continuing.
- The website uses GitHub's latest published release API for download links, so published beta prereleases do not replace the stable download target.
## Changelog format
@@ -602,6 +659,7 @@ No prefix (`v`), no extra text. `Release Notes Sync` matches the `## X.Y.Z` (or
- `CHANGELOG.md` holds stable releases and at most one entry for the current beta series.
- The first beta of a version inserts a top entry like `## 0.1.60-beta.1 - YYYY-MM-DD`.
- Each subsequent beta rewrites that entry in place: the heading becomes the new beta number and date, and the body describes the full change from the previous stable release to this beta. Never add a second beta entry.
- The entry announces changes and leaves setup steps, click paths, migration notes, configuration keys, hedges and edge cases to product docs. The headline change is the first highlight. [Changelog conciseness](#changelog-conciseness) and [Changelog ordering](#changelog-ordering) hold the rules.
- The entry is a list of changes, not of PRs. When a later beta reverts, reworks, or follows up on something an earlier beta added, rewrite the bullet to the current behavior or delete it. A stable user never saw the earlier beta behavior, so fixes to it are not "Fixed" bullets.
- Anything that already shipped to stable through a patch release on an older line stays out of the beta entry.
- Stable promotion renames the entry to `## 0.1.60 - YYYY-MM-DD` after a final review.
@@ -657,14 +715,30 @@ claims a general improvement, rewrite it with the concrete change.
## Changelog conciseness
Every section bullet must be scannable at a glance. The changelog is not release documentation — it's a list. [Highlights](#changelog-highlights) get more room; every other bullet follows these limits.
An entry announces what changed. Product docs own setup, navigation, migration and edge cases, and the PR link leads to the rest. These rules apply to highlights and section bullets.
- **One sentence per bullet, max.** If a bullet contains two sentences, the second one is doing work that belongs in product docs, not the changelog. Cut it.
- **No instructions.** Cut setup steps, click paths, configuration keys, flags and migration notes. "Open a provider's settings from Settings, setup, or the model picker" is a click path. "Existing goose settings and sessions carry over to the bundled plugin" is a migration note.
- **One warning for a breaking requirement.** When the update stops something working until the user acts, state the requirement in one short sentence after the announcement, and link the product doc that has the steps: "Requires [Antigravity CLI 1.5.0 or later](https://paseo.sh/docs/supported-providers/antigravity)." Nothing else earns a second sentence.
- **No hedges or exception clauses.** Cut "when needed", "for eligible models", "on hosts where…", "except…", "with a hint shown when…" and "across common install shapes". Keep the condition that triggers a fixed failure, as in `Fixed <failure> when <condition>`.
- **The shorter bullet is still true.** When a change reaches only some providers, platforms or file types, name what it covers in the sentence's own words: "update supported CLIs" where one provider has no update command, "Markdown previews and source views" where HTML previews on iOS are left out, "by default" where an explicit interrupt still replaces the turn. Do not widen the claim to remove a clause.
- **State the outcome and stop.** Say what the user gets. Leave out queues, snapshots, caches, timeouts, LAN/WAN topology, TLS handshakes and IPC.
Each row is from the first draft of 0.17.1:
| Avoid | Write |
| --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------- |
| Open a provider's settings from Settings, setup, or the model picker to update supported CLIs and watch their output | Update supported CLIs from the app |
| Use Antigravity CLI 1.5.0 or later on the daemon host, and include `acp` if you override its command | Requires Antigravity CLI 1.5.0 or later |
| Reports and notifications from other agents arrive in order when the current turn ends instead of interrupting it. An interrupt or your own prompt carries earlier queued messages in the same turn, and Stop holds queued messages until the next prompt | By default, messages from other agents wait for the current turn to end |
| Changed worktree teardown to keep the worktree when the command exceeds one minute or the `worktree.teardownTimeoutMs` limit in `paseo.json` | Changed worktree teardown to keep the worktree when the command times out |
| Fixed file views losing their scroll position when switching Preview and Source, except HTML previews on iOS | Fixed Markdown previews and source views losing their scroll position when switching between them |
Section bullets also follow these limits:
- **One sentence per bullet, max.** A second sentence is documentation. Cut it.
- **No trailing periods.** Bullets are list items, not prose. Drop the period at the end of every bullet, including the period inside any bolded lead-in. `**Configurable terminal scrollback**` not `**Configurable terminal scrollback.**`.
- **One line per bullet.** If a bullet wraps to three lines in a narrow column, it's too long.
- **Split bullets that pack multiple distinct changes.** If a bullet uses "and", "plus", a comma list, or an em-dash to chain several independent improvements, break them into separate bullets — even when they share a theme or author. One bullet = one user-facing change.
- **Trim qualifying clauses.** Drop "with a hint shown when…", "matching the CLI's behaviour", "across common install shapes". If the detail doesn't change whether a user cares, cut it.
- **Stop after identifying the change.** Do not explain LAN/WAN topology, TLS handshakes, IPC, or other architecture in a changelog bullet. Put necessary background in product docs.
- **Attribution follows the split.** When you split a dense bullet, move each PR/author to the bullet it belongs to. Never duplicate the same PR across multiple bullets.
## Changelog highlights
@@ -677,7 +751,7 @@ reader who stops after the highlights should have missed nothing they would ask
```markdown
## 0.16.0 - 2026-11-20
- **Undo workspace archive:** Archiving a workspace in the app now shows a "Workspace archived" toast for six seconds. Undo, or Cmd/Ctrl+Z, restores the workspace with its tabs and agents. ([#8904](https://github.com/getpaseo/paseo/pull/8904))
- **Undo workspace archive:** Restore an archived workspace with Undo or Cmd/Ctrl+Z. ([#8904](https://github.com/getpaseo/paseo/pull/8904))
### Added
```
@@ -688,17 +762,19 @@ reader who stops after the highlights should have missed nothing they would ask
- **The lead-in is not a place or something Paseo already had.** A device, screen or area names
where the change landed. A feature name that was true of the last release hides the news: Paseo
already had keyboard shortcuts, so the lead-in says they now work on mobile.
- **The first sentence says the plain thing.** Write what a user would tell a friend: Paseo now
- **One short sentence says the plain thing.** Write what a user would tell a friend: Paseo now
shows the background tasks an agent is running. Leave out the conditions and mechanics that make
it true.
it true. A second sentence is only for a
[breaking requirement](#changelog-conciseness).
- **Describe the new behavior, not the old one.** Say what is faster or better now, and do not
dwell on how slow or broken it was before. A number is welcome as a short aside after the new
behavior is stated: seconds, a percentage, a multiple.
- **Do not list what a sentence already covers.** After "the same shortcuts as desktop", naming
five of them adds nothing. Keep a list only when it sets the scope, such as the providers a
feature works with.
- **The description runs one to three sentences.** Sentences end with periods, and the attribution
follows the last one. [Changelog wording](#changelog-wording) applies in full.
- **Sentences end with periods, and the attribution follows the last one.**
[Changelog wording](#changelog-wording) and [Changelog conciseness](#changelog-conciseness)
apply.
- **A highlight replaces the section bullets it covers.** A PR appears once in an entry, so its
attribution moves to the highlight.
- **A highlight is something a user would name.** Everything else stays a section bullet. A patch
@@ -707,13 +783,13 @@ reader who stops after the highlights should have missed nothing they would ask
of related fixes, becomes a highlight only when it passes the same test as a feature, and then
it is written as what works now: `**Fewer reconnects on return:**`.
| Avoid | Write |
| ------------------------------------------------------------------------------------------ | ---------------------------------------------------------------------------------------------------------------- |
| `**iPad and touch screens:**` | `**Touch-sized tablet controls:**` |
| `**Hardware keyboard shortcuts:**` | `**Keyboard shortcuts on mobile:**` |
| When an agent ends its turn while shells it started are still running, a pill lists them | Paseo now shows the background tasks an agent is running |
| A diff no longer waits 4 to 40 seconds behind worktree scans | Opening Changes shows the file tree right away, and diffs load about 6 times sooner on hosts with many worktrees |
| The same shortcuts as desktop: command center, file picker, new workspace, workspace jumps | The same shortcuts as desktop, including your custom bindings |
| Avoid | Write |
| ------------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------- |
| `**iPad and touch screens:**` | `**Touch-sized tablet controls:**` |
| `**Hardware keyboard shortcuts:**` | `**Keyboard shortcuts on mobile:**` |
| When an agent ends its turn while shells it started are still running, a pill lists them | Paseo now shows the background tasks an agent is running |
| A diff no longer waits 4 to 40 seconds behind worktree scans | Opening Changes shows the file tree right away, and diffs load about 6 times sooner |
| The same shortcuts as desktop: command center, file picker, new workspace, workspace jumps | The same shortcuts as desktop, including your custom bindings |
Before showing a draft, read only the bold lead-ins top to bottom. Each one must answer "what is
new?" without its description, and each must outrank the one below it.
@@ -763,6 +839,9 @@ question.
for them: `Undo workspace archive`, `Fewer reconnects on return`. Fixes are routine, so one
that earned a highlight still follows everything new or improved.
- **The headline change goes first, whatever its group.** It is the change the release is announced
with. 0.17.1 leads with `Antigravity approvals and history`, which needs one provider. The groups
rank everything after it.
- **Reach decides before kind.** A new capability that needs a hardware keyboard is in group 3.
- **Rank the whole list in one pass.** The order you collected the pull requests in, the alphabet
and the order of the sections below are not rankings.
@@ -802,13 +881,12 @@ Every entry covers `previous stable release → release source`: first beta, lat
### Beta release
- [ ] The release base and changes match the [release base table](#release-base-and-source) and any approved [inventory](#release-inventory), and CI for the release source is green
- [ ] The release base and changes match the [release base table](#release-base-and-source) and any approved [inventory](#release-inventory)
- [ ] Every PR in the release range has been opened, and its full description and every linked issue have been read before drafting the changelog
- [ ] Fold this beta into the series' single `CHANGELOG.md` entry (heading `## X.Y.Z-beta.N - YYYY-MM-DD`, previous stable → release source), review it against the changelog policy, get approval, and commit it with its [`CHANGELOG.zh-CN.md` translation](#changelog-translations) before cutting the release
- [ ] Fold this beta into the series' single `CHANGELOG.md` entry (heading `## X.Y.Z-beta.N - YYYY-MM-DD`, previous stable → release source), review it against the changelog policy, get approval, and commit it with its [`CHANGELOG.zh-CN.md` translation](#changelog-translations) before running `release:prep:*`
- [ ] The diff from the previous stable to the resolved release source is classified as patch or minor, with the target version and rationale approved
- [ ] The changelog, ACP, lockfile, and version commits stayed local until the approved release command pushed the release branch and tag
- [ ] `npm run release:beta:patch`, `npm run release:beta:minor`, or `npm run release:beta:next` completes successfully
- [ ] Every GitHub Actions run for the complete release commit and tag is green
- [ ] `npm run release:prep:beta:patch`, `release:prep:beta:minor`, or `release:prep:beta:next` created the version commit on `release/vX.Y.Z-beta.N` without a tag, and the pushed branch passed `CI` and `Desktop Packages` before shipping (never relying on `main` CI)
- [ ] After go-ahead, `npm run release:ship:beta` published npm and pushed the `vX.Y.Z-beta.N` tag without triggering a second CI run
- [ ] npm shows the version under the `beta` dist-tag, not `latest`
- [ ] The GitHub prerelease was published only after the three beta manifests were uploaded, and it has the changelog body and every expected macOS, Linux, Windows, and Android APK asset
- [ ] GitHub `Desktop Release` workflow for the `v*-beta.N` tag is green
@@ -817,23 +895,21 @@ Every entry covers `previous stable release → release source`: first beta, lat
- [ ] GitHub `Docker` workflow is green and the versioned beta image is published without moving `latest`
- [ ] GitHub `Release Notes Sync` mirrored the beta entry into the prerelease body
- [ ] EAS `Release iOS Beta` completed its build, TestFlight distribution, external beta group, and Beta App Review path
- [ ] The [reconcile PR](#reconcile-main) brought the shipped changelog entries and version into `main`
- [ ] The [reconcile PR](#reconcile-main) brought the shipped changelog entries and version into `main`, and `origin/release/vX.Y.Z-beta.N` was deleted
- [ ] The release heartbeat was created after the tag push and deleted only after every item above passed
### Stable release (or promotion)
- [ ] Run the pre-release sanity check (see above) and address any findings
- [ ] The diff from the previous stable to the resolved release source is classified as patch or minor, with the target version and rationale approved
- [ ] The release base and changes match the [release base table](#release-base-and-source) and any approved [inventory](#release-inventory), and CI for the release source is green
- [ ] The release base and changes match the [release base table](#release-base-and-source) and any approved [inventory](#release-inventory)
- [ ] A promotion ships the tested beta's product content after its soak, unless the user explicitly overrode either default
- [ ] Every PR in the release range has been opened, and its full description and every linked issue have been read before drafting the changelog
- [ ] Ensure the approved release inputs are committed locally and the git worktree is clean before running any release command
- [ ] Ensure local `npm run typecheck` passes on that exact commit before running any release command
- [ ] Update `CHANGELOG.md` with user-facing release notes (features, fixes — not refactors). Promotion renames the series' `## X.Y.Z-beta.N` entry to `## X.Y.Z - YYYY-MM-DD` and re-checks it covers the full release. Commit the matching [`CHANGELOG.zh-CN.md` entry](#changelog-translations) with it
- [ ] Verify the changelog heading follows strict `## X.Y.Z - YYYY-MM-DD` format
- [ ] The changelog, ACP, lockfile, and version commits stayed local until the approved release command pushed the release branch and tag
- [ ] `npm run release:patch`, `npm run release:minor`, or `npm run release:promote` completes successfully
- [ ] Every GitHub Actions run for the complete release commit and tag is green
- [ ] Ensure local format, lint, and typecheck pass and the working tree is clean before running `release:prep:*`
- [ ] `npm run release:prep:patch`, `release:prep:minor`, or `release:prep:promote` created the version commit on `release/vX.Y.Z` without a tag, and the pushed branch passed `CI` and `Desktop Packages` before shipping (never relying on `main` CI)
- [ ] After go-ahead, `npm run release:ship` published npm, updated dist-tags, and pushed the `vX.Y.Z` tag without triggering a second CI run
- [ ] `latest` resolves to the new stable, and `beta` resolves to it too unless an active newer beta holds `beta`
- [ ] The GitHub Release was published only after the three stable manifests were uploaded, and it has the changelog body and every expected macOS, Linux, Windows, and Android APK asset
- [ ] GitHub `Desktop Release` workflow for the `v*` tag is green
@@ -848,5 +924,5 @@ Every entry covers `previous stable release → release source`: first beta, lat
- [ ] EAS iOS `submit_ios_for_review` succeeds, putting the build into App Store review
- [ ] EAS Android `build_android` completes for the same tag
- [ ] EAS Android `submit_android` succeeds, putting the build on its Play Store track
- [ ] The [reconcile PR](#reconcile-main) brought the shipped changelog entries and version into `main`
- [ ] The [reconcile PR](#reconcile-main) brought the shipped changelog entries, F-Droid changelogs, and version into `main`, and `origin/release/vX.Y.Z` was deleted
- [ ] The release heartbeat was created after the tag push and deleted only after every item above passed
@@ -1,3 +0,0 @@
- Use existing worktrees: Start a workspace in a worktree you already have, or create one from a branch or pull request. Plugins can manage worktree creation, removal, and restoration. When you branch off a fork PR or MR, choose Run setup to approve workspace automation.
Full notes: https://paseo.sh/changelog
@@ -1,3 +0,0 @@
- Use existing worktrees: Start a workspace in a worktree you already have, or create one from a branch or pull request. Plugins can manage worktree creation, removal, and restoration. When you branch off a fork PR or MR, choose Run setup to approve workspace automation.
Full notes: https://paseo.sh/changelog
@@ -1,3 +0,0 @@
- Use existing worktrees: Start a workspace in a worktree you already have, or create one from a branch or pull request. Plugins can manage worktree creation, removal, and restoration. When you branch off a fork PR or MR, choose Run setup to approve workspace automation.
Full notes: https://paseo.sh/changelog
@@ -1,3 +0,0 @@
- Use existing worktrees: Start a workspace in a worktree you already have, or create one from a branch or pull request. Plugins can manage worktree creation, removal, and restoration. When you branch off a fork PR or MR, choose Run setup to approve workspace automation.
Full notes: https://paseo.sh/changelog
@@ -0,0 +1,5 @@
- Antigravity approvals and history: Choose a permission mode, answer tool approvals, and import or reopen sessions. Requires Antigravity CLI 1.5.0 or later.
- Use existing worktrees: Start a workspace in a worktree you already have, or create one from a branch or pull request.
- Update providers in Paseo: See each provider's status and installed version, and update supported CLIs from the app.
Full notes: https://paseo.sh/changelog
@@ -0,0 +1,5 @@
- Antigravity approvals and history: Choose a permission mode, answer tool approvals, and import or reopen sessions. Requires Antigravity CLI 1.5.0 or later.
- Use existing worktrees: Start a workspace in a worktree you already have, or create one from a branch or pull request.
- Update providers in Paseo: See each provider's status and installed version, and update supported CLIs from the app.
Full notes: https://paseo.sh/changelog
@@ -0,0 +1,5 @@
- Antigravity approvals and history: Choose a permission mode, answer tool approvals, and import or reopen sessions. Requires Antigravity CLI 1.5.0 or later.
- Use existing worktrees: Start a workspace in a worktree you already have, or create one from a branch or pull request.
- Update providers in Paseo: See each provider's status and installed version, and update supported CLIs from the app.
Full notes: https://paseo.sh/changelog
@@ -0,0 +1,5 @@
- Antigravity approvals and history: Choose a permission mode, answer tool approvals, and import or reopen sessions. Requires Antigravity CLI 1.5.0 or later.
- Use existing worktrees: Start a workspace in a worktree you already have, or create one from a branch or pull request.
- Update providers in Paseo: See each provider's status and installed version, and update supported CLIs from the app.
Full notes: https://paseo.sh/changelog
@@ -0,0 +1,5 @@
- Turn and session stats: View per-turn and session timing, token, cache, and cost metrics for Claude, OpenCode v2, Pi, Oh My Pi, and plugin providers.
- Automatic model list refresh: Enabled providers refresh their model lists in the background every 30 minutes.
- Sidebar usage by default: The sidebar usage summary is enabled by default, sources for inactive providers start hidden, and source toggles move to a Sources list.
Full notes: https://paseo.sh/changelog
@@ -0,0 +1,5 @@
- Turn and session stats: View per-turn and session timing, token, cache, and cost metrics for Claude, OpenCode v2, Pi, Oh My Pi, and plugin providers.
- Automatic model list refresh: Enabled providers refresh their model lists in the background every 30 minutes.
- Sidebar usage by default: The sidebar usage summary is enabled by default, sources for inactive providers start hidden, and source toggles move to a Sources list.
Full notes: https://paseo.sh/changelog
@@ -0,0 +1,5 @@
- Turn and session stats: View per-turn and session timing, token, cache, and cost metrics for Claude, OpenCode v2, Pi, Oh My Pi, and plugin providers.
- Automatic model list refresh: Enabled providers refresh their model lists in the background every 30 minutes.
- Sidebar usage by default: The sidebar usage summary is enabled by default, sources for inactive providers start hidden, and source toggles move to a Sources list.
Full notes: https://paseo.sh/changelog
@@ -0,0 +1,5 @@
- Turn and session stats: View per-turn and session timing, token, cache, and cost metrics for Claude, OpenCode v2, Pi, Oh My Pi, and plugin providers.
- Automatic model list refresh: Enabled providers refresh their model lists in the background every 30 minutes.
- Sidebar usage by default: The sidebar usage summary is enabled by default, sources for inactive providers start hidden, and source toggles move to a Sources list.
Full notes: https://paseo.sh/changelog
+1 -1
View File
@@ -1 +1 @@
sha256-gUBXG6rGxAAMUDgSXQw6GzkJ8U0xEoL9X/kZkLM4uxg=
sha256-vtzoUMnKaAdBFd/hhnk9p2WwHGgL70RXh26vkkyRIH0=
+38 -28
View File
@@ -1,12 +1,12 @@
{
"name": "paseo",
"version": "0.17.0",
"version": "0.17.2",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "paseo",
"version": "0.17.0",
"version": "0.17.2",
"hasInstallScript": true,
"license": "Apache-2.0",
"workspaces": [
@@ -6651,6 +6651,15 @@
"url": "https://github.com/sponsors/ayuhito"
}
},
"node_modules/@fontsource-variable/plus-jakarta-sans": {
"version": "5.3.0",
"resolved": "https://registry.npmjs.org/@fontsource-variable/plus-jakarta-sans/-/plus-jakarta-sans-5.3.0.tgz",
"integrity": "sha512-/l/4r0yyWK9JzAlmA0LiYgGgmJe/Gswt4jTJEzr5QhJfwMbvJZDmYWyW0M4X7yCK69BajMVlV/Lx8g7WDc1+sw==",
"license": "OFL-1.1",
"funding": {
"url": "https://github.com/sponsors/ayuhito"
}
},
"node_modules/@getpaseo/app": {
"resolved": "packages/app",
"link": true
@@ -37468,7 +37477,7 @@
},
"packages/app": {
"name": "@getpaseo/app",
"version": "0.17.0",
"version": "0.17.2",
"dependencies": {
"@codemirror/commands": "6.10.4",
"@codemirror/language": "6.12.4",
@@ -39638,13 +39647,13 @@
},
"packages/cli": {
"name": "@getpaseo/cli",
"version": "0.17.0",
"version": "0.17.2",
"license": "Apache-2.0",
"dependencies": {
"@clack/prompts": "^1.0.0",
"@getpaseo/client": "0.17.0",
"@getpaseo/protocol": "0.17.0",
"@getpaseo/server": "0.17.0",
"@getpaseo/client": "0.17.2",
"@getpaseo/protocol": "0.17.2",
"@getpaseo/server": "0.17.2",
"chalk": "^5.3.0",
"commander": "^12.0.0",
"mime-types": "^2.1.35",
@@ -39891,11 +39900,11 @@
},
"packages/client": {
"name": "@getpaseo/client",
"version": "0.17.0",
"version": "0.17.2",
"license": "Apache-2.0",
"dependencies": {
"@getpaseo/protocol": "0.17.0",
"@getpaseo/relay": "0.17.0",
"@getpaseo/protocol": "0.17.2",
"@getpaseo/relay": "0.17.2",
"ws": "^8.20.0",
"zod": "^4.4.3"
},
@@ -39907,7 +39916,7 @@
},
"packages/desktop": {
"name": "@getpaseo/desktop",
"version": "0.17.0",
"version": "0.17.2",
"license": "Apache-2.0",
"dependencies": {
"@getpaseo/cli": "*",
@@ -40150,7 +40159,7 @@
},
"packages/expo-two-way-audio": {
"name": "@getpaseo/expo-two-way-audio",
"version": "0.17.0",
"version": "0.17.2",
"license": "MIT",
"devDependencies": {
"@types/jest": "^29.5.14",
@@ -40196,7 +40205,7 @@
},
"packages/highlight": {
"name": "@getpaseo/highlight",
"version": "0.17.0",
"version": "0.17.2",
"license": "Apache-2.0",
"dependencies": {
"@codemirror/language": "6.12.4",
@@ -40226,7 +40235,7 @@
},
"packages/plugin": {
"name": "@getpaseo/plugin",
"version": "0.17.0",
"version": "0.17.2",
"license": "Apache-2.0",
"dependencies": {
"@agentclientprotocol/sdk": "^1.4.0",
@@ -40234,8 +40243,8 @@
"use-sync-external-store": "^1.6.0"
},
"devDependencies": {
"@getpaseo/client": "0.17.0",
"@getpaseo/protocol": "0.17.0",
"@getpaseo/client": "0.17.2",
"@getpaseo/protocol": "0.17.2",
"@types/node": "^20.9.0",
"@types/react": "~19.2.0",
"react": "19.2.3",
@@ -40245,8 +40254,8 @@
"zod": "^4.4.3"
},
"peerDependencies": {
"@getpaseo/client": "0.17.0",
"@getpaseo/protocol": "0.17.0",
"@getpaseo/client": "0.17.2",
"@getpaseo/protocol": "0.17.2",
"react": ">=19.1.0 <20",
"react-native": ">=0.81.5",
"zod": "^4.4.3"
@@ -40268,7 +40277,7 @@
},
"packages/protocol": {
"name": "@getpaseo/protocol",
"version": "0.17.0",
"version": "0.17.2",
"license": "Apache-2.0",
"dependencies": {
"semver": "^7.8.5",
@@ -40296,7 +40305,7 @@
},
"packages/relay": {
"name": "@getpaseo/relay",
"version": "0.17.0",
"version": "0.17.2",
"license": "Apache-2.0",
"dependencies": {
"@noble/ciphers": "^2.4.0",
@@ -40518,17 +40527,17 @@
},
"packages/server": {
"name": "@getpaseo/server",
"version": "0.17.0",
"version": "0.17.2",
"license": "Apache-2.0",
"dependencies": {
"@agentclientprotocol/sdk": "^0.17.1",
"@anthropic-ai/claude-agent-sdk": "0.3.326",
"@anthropic-ai/sdk": "^0.104.2",
"@getpaseo/client": "0.17.0",
"@getpaseo/highlight": "0.17.0",
"@getpaseo/plugin": "0.17.0",
"@getpaseo/protocol": "0.17.0",
"@getpaseo/relay": "0.17.0",
"@getpaseo/client": "0.17.2",
"@getpaseo/highlight": "0.17.2",
"@getpaseo/plugin": "0.17.2",
"@getpaseo/protocol": "0.17.2",
"@getpaseo/relay": "0.17.2",
"@isaacs/ttlcache": "^2.1.4",
"@modelcontextprotocol/sdk": "1.29.0",
"@opencode-ai/sdk": "1.14.46",
@@ -41475,11 +41484,12 @@
},
"packages/website": {
"name": "@getpaseo/website",
"version": "0.17.0",
"version": "0.17.2",
"dependencies": {
"@cloudflare/vite-plugin": "^1.29.1",
"@cloudflare/workers-types": "^4.20260317.1",
"@fontsource-variable/geist": "^5.3.0",
"@fontsource-variable/plus-jakarta-sans": "^5.3.0",
"@getpaseo/protocol": "*",
"@tanstack/react-router": "^1.170.16",
"@tanstack/react-start": "^1.168.26",
@@ -43085,7 +43095,7 @@
},
"plugins": {
"name": "@getpaseo/builtin-plugins",
"version": "0.17.0",
"version": "0.17.2",
"dependencies": {
"@getpaseo/plugin": "*",
"zod": "^4.4.3"
+16 -13
View File
@@ -1,6 +1,6 @@
{
"name": "paseo",
"version": "0.17.0",
"version": "0.17.2",
"private": true,
"description": "Paseo: voice-controlled development environment for local AI coding agents",
"keywords": [
@@ -110,19 +110,22 @@
"version:all:beta:next": "node scripts/set-release-version.mjs --mode beta-next",
"version:all:promote": "node scripts/set-release-version.mjs --mode promote",
"release:check": "npm run release:prepare && npm run typecheck --workspace=@getpaseo/highlight && npm run typecheck --workspace=@getpaseo/relay && npm run typecheck --workspace=@getpaseo/protocol && npm run build:client:clean && npm run typecheck --workspace=@getpaseo/client && npm run typecheck --workspace=@getpaseo/plugin && npm run build:server:clean && npm run typecheck --workspace=@getpaseo/server && npm run typecheck --workspace=@getpaseo/cli && npm pack --dry-run --workspace=@getpaseo/highlight && npm pack --dry-run --workspace=@getpaseo/relay && npm pack --dry-run --workspace=@getpaseo/protocol && npm pack --dry-run --workspace=@getpaseo/client && npm pack --dry-run --workspace=@getpaseo/plugin && npm pack --dry-run --workspace=@getpaseo/server && npm pack --dry-run --workspace=@getpaseo/cli",
"release:publish:dry-run": "npm publish --dry-run --workspace=@getpaseo/highlight --access public && npm publish --dry-run --workspace=@getpaseo/relay --access public && npm publish --dry-run --workspace=@getpaseo/protocol --access public && npm publish --dry-run --workspace=@getpaseo/client --access public && npm publish --dry-run --workspace=@getpaseo/plugin --access public && npm publish --dry-run --workspace=@getpaseo/server --access public && npm publish --dry-run --workspace=@getpaseo/cli --access public",
"release:publish:beta:dry-run": "npm publish --dry-run --workspace=@getpaseo/highlight --access public --tag beta && npm publish --dry-run --workspace=@getpaseo/relay --access public --tag beta && npm publish --dry-run --workspace=@getpaseo/protocol --access public --tag beta && npm publish --dry-run --workspace=@getpaseo/client --access public --tag beta && npm publish --dry-run --workspace=@getpaseo/plugin --access public --tag beta && npm publish --dry-run --workspace=@getpaseo/server --access public --tag beta && npm publish --dry-run --workspace=@getpaseo/cli --access public --tag beta",
"release:publish": "npm publish --workspace=@getpaseo/highlight --access public && npm publish --workspace=@getpaseo/relay --access public && npm publish --workspace=@getpaseo/protocol --access public && npm publish --workspace=@getpaseo/client --access public && npm publish --workspace=@getpaseo/plugin --access public && npm publish --workspace=@getpaseo/server --access public && npm publish --workspace=@getpaseo/cli --access public",
"release:publish:beta": "npm publish --workspace=@getpaseo/highlight --access public --tag beta && npm publish --workspace=@getpaseo/relay --access public --tag beta && npm publish --workspace=@getpaseo/protocol --access public --tag beta && npm publish --workspace=@getpaseo/client --access public --tag beta && npm publish --workspace=@getpaseo/plugin --access public --tag beta && npm publish --workspace=@getpaseo/server --access public --tag beta && npm publish --workspace=@getpaseo/cli --access public --tag beta",
"release:prep:patch": "npm run release:check && npm run version:all:patch",
"release:prep:minor": "npm run release:check && npm run version:all:minor",
"release:prep:major": "npm run release:check && npm run version:all:major",
"release:prep:beta:patch": "npm run release:check && npm run version:all:beta:patch",
"release:prep:beta:minor": "npm run release:check && npm run version:all:beta:minor",
"release:prep:beta:major": "npm run release:check && npm run version:all:beta:major",
"release:prep:beta:next": "npm run release:check && npm run version:all:beta:next",
"release:prep:promote": "npm run release:check && npm run version:all:promote",
"release:publish:dry-run": "node scripts/publish-release-packages.mjs --dry-run",
"release:publish:beta:dry-run": "node scripts/publish-release-packages.mjs --beta --dry-run",
"release:publish": "node scripts/publish-release-packages.mjs",
"release:publish:beta": "node scripts/publish-release-packages.mjs --beta",
"release:push": "node scripts/push-current-release-tag.mjs",
"release:beta:patch": "npm run release:check && npm run version:all:beta:patch && npm run release:publish:beta && npm run release:push",
"release:beta:minor": "npm run release:check && npm run version:all:beta:minor && npm run release:publish:beta && npm run release:push",
"release:beta:major": "npm run release:check && npm run version:all:beta:major && npm run release:publish:beta && npm run release:push",
"release:beta:next": "npm run release:check && npm run version:all:beta:next && npm run release:publish:beta && npm run release:push",
"release:promote": "npm run release:check && npm run version:all:promote && npm run release:publish && npm run release:push",
"release:patch": "npm run release:check && npm run version:all:patch && npm run release:publish && npm run release:push",
"release:minor": "npm run release:check && npm run version:all:minor && npm run release:publish && npm run release:push",
"release:major": "npm run release:check && npm run version:all:major && npm run release:publish && npm run release:push"
"release:ship": "node scripts/push-current-release-tag.mjs --check && npm run release:publish && npm run release:push",
"release:ship:beta": "node scripts/push-current-release-tag.mjs --check && npm run release:publish:beta && npm run release:push",
"release:reconcile": "node scripts/reconcile-release.mjs"
},
"devDependencies": {
"@types/ws": "^8.5.14",
@@ -430,7 +430,8 @@ async function expectFooterWithinWidth(page: Page, width: number): Promise<void>
const label = page.getByRole("button", { name: /^Worked for/ }).last();
await expect(label).toBeVisible();
await expectTimingInsideFooter(page);
const controls = footer.getByRole("button", { name: /Copy turn|Fork chat/ });
await expect(footer.getByRole("button", { name: "Turn stats" })).toBeVisible();
const controls = footer.getByRole("button", { name: /Copy turn|Fork chat|Turn stats/ });
const before = await controls.evaluateAll((elements) =>
elements.map((element) => {
const rect = element.getBoundingClientRect();
@@ -59,6 +59,35 @@ async function selectClientSlashCommand(page: Page, query: string, label: string
await input.press("Enter");
}
async function typeBeforeMenuFilters(
page: Page,
input: { text: string; key: "Tab" | "Enter" },
): Promise<void> {
await composerLocator(page).evaluate(async (element, { text, key }) => {
if (!(element instanceof HTMLTextAreaElement)) {
throw new Error("Composer input is not a textarea");
}
const setValue = Object.getOwnPropertyDescriptor(HTMLTextAreaElement.prototype, "value")?.set;
if (!setValue) throw new Error("Textarea value setter is unavailable");
setValue.call(element, text);
element.setSelectionRange(text.length, text.length);
element.dispatchEvent(new Event("input", { bubbles: true }));
await Promise.resolve();
element.dispatchEvent(new KeyboardEvent("keydown", { key, bubbles: true, cancelable: true }));
await new Promise((resolve) => requestAnimationFrame(() => setTimeout(resolve, 0)));
}, input);
}
async function showExitSuggestion(page: Page, query: string): Promise<void> {
await composerLocator(page).fill(query);
await expect(page.getByText("/exit", { exact: true }).first()).toBeVisible();
}
async function expectOpenAgentWithInput(page: Page, agentId: string, text: string): Promise<void> {
await expect(composerLocator(page)).toHaveValue(text);
await expectWorkspaceTabVisible(page, agentId);
}
async function expectAgentArchivedInSessions(page: Page, title: string): Promise<void> {
await openSessions(page);
await expectSessionRowArchived(page, title);
@@ -131,6 +160,16 @@ test.describe("Client slash commands", () => {
);
});
for (const key of ["Tab", "Enter"] as const) {
test(`${key} ignores an exit suggestion for an earlier query`, async ({ page }) => {
await withOpenReadyMockAgent(page, { title: "Stale slash command" }, async ({ agentId }) => {
await showExitSuggestion(page, "/e");
await typeBeforeMenuFilters(page, { text: "/co", key });
await expectOpenAgentWithInput(page, agentId, "/co");
});
});
}
test("slash clear replaces the active agent with a matching draft", async ({ page }) => {
await withOpenReadyMockAgent(
page,
@@ -0,0 +1,143 @@
import type { Locator } from "@playwright/test";
import { expect, test, type Page } from "../support/fixtures";
import { openAgentRoute, seedMockAgentWorkspace } from "../support/helpers/mock-agent";
const APP_SETTINGS_KEY = "@paseo:app-settings";
async function useLanguage(page: Page, language: string): Promise<void> {
await page.addInitScript(
({ key, language: value }) => {
localStorage.setItem(key, JSON.stringify({ language: value }));
},
{ key: APP_SETTINGS_KEY, language },
);
}
/**
* Lays the placeholder out the way the textarea does, in a mirror element with
* the same box and font, and reports whether any of its lines reach the hint.
*/
async function placeholderOverlaps(input: Locator, hint: Locator): Promise<boolean> {
const hintBox = await hint.boundingBox();
if (!hintBox) throw new Error("Focus hint has no layout box");
return input.evaluate((element, box) => {
const textarea = element as HTMLTextAreaElement;
const style = getComputedStyle(textarea);
const rect = textarea.getBoundingClientRect();
const mirror = document.createElement("div");
for (const property of [
"boxSizing",
"paddingTop",
"paddingRight",
"paddingBottom",
"paddingLeft",
"borderTopWidth",
"borderRightWidth",
"borderBottomWidth",
"borderLeftWidth",
"fontFamily",
"fontSize",
"fontWeight",
"letterSpacing",
"lineHeight",
"textIndent",
"wordSpacing",
] as const) {
mirror.style[property] = style[property];
}
Object.assign(mirror.style, {
position: "fixed",
left: `${rect.left}px`,
top: `${rect.top}px`,
width: `${rect.width}px`,
borderStyle: "solid",
borderColor: "transparent",
whiteSpace: "pre-wrap",
overflowWrap: "break-word",
visibility: "hidden",
});
const text = document.createTextNode(textarea.placeholder);
mirror.appendChild(text);
document.body.appendChild(mirror);
const range = document.createRange();
range.selectNodeContents(text);
const lines = Array.from(range.getClientRects());
mirror.remove();
return lines.some(
(line) =>
line.right > box.x &&
line.left < box.x + box.width &&
line.bottom > box.y &&
line.top < box.y + box.height,
);
}, hintBox);
}
async function recordFirstVisibleHint(page: Page): Promise<void> {
await page.addInitScript(() => {
function sample() {
const root = document.querySelector('[data-testid="message-input-root"]');
const input = root?.querySelector("textarea");
const hint = Array.from(root?.querySelectorAll("div") ?? []).find(
(element) => !element.childElementCount && element.textContent?.endsWith("pour saisir"),
);
if (!input || !hint || Number(getComputedStyle(hint).opacity) === 0) {
requestAnimationFrame(sample);
return;
}
const inputRight = input.getBoundingClientRect().right;
const textRight = inputRight - parseFloat(getComputedStyle(input).paddingRight);
const hintLeft = hint.getBoundingClientRect().left;
Reflect.set(window, "__firstVisibleHintHasSpace", textRight <= hintLeft + 1);
}
requestAnimationFrame(sample);
});
}
async function expectFirstVisibleHintHasSpace(page: Page): Promise<void> {
await expect
.poll(() => page.evaluate(() => Reflect.get(window, "__firstVisibleHintHasSpace")))
.toBe(true);
}
const FRENCH_LAYOUTS = [
{ name: "phone width", width: 390, placeholder: "Message, @fichiers, /commandes" },
{
name: "desktop width",
width: 1024,
placeholder:
"Écrivez à l’agent, mentionnez des @fichiers ou utilisez les /commandes et /skills",
},
];
for (const layout of FRENCH_LAYOUTS) {
test.describe(`French composer at ${layout.name}`, () => {
test.use({ viewport: { width: layout.width, height: 844 } });
test("the focus hint does not cover the placeholder", async ({ page }) => {
const agent = await seedMockAgentWorkspace({
repoPrefix: "composer-focus-hint-",
title: "Composer focus hint",
});
try {
await useLanguage(page, "fr");
await recordFirstVisibleHint(page);
await openAgentRoute(page, agent);
const input = page.getByRole("textbox", { name: "Écrire à l’agent…" }).first();
await expect(input).toBeVisible();
await expect(input).toHaveAttribute("placeholder", layout.placeholder);
await input.blur();
const hint = page.getByText(/pour saisir$/).first();
await expect(hint).toBeVisible();
// The input takes the hint's width on the render after the hint lays out.
await expect.poll(() => placeholderOverlaps(input, hint)).toBe(false);
await expectFirstVisibleHintHasSpace(page);
} finally {
await agent.cleanup();
}
});
});
}
@@ -18,6 +18,7 @@ import {
expectComposerFullscreenControls,
sendComposerWithButton,
queueComposerWithButton,
fillComposerDraft,
configureComposerQueueSend,
activateRetainedComposerTab,
switchComposerTabWithKeyboard,
@@ -450,6 +451,32 @@ test("fullscreen Send keeps the configured queue behavior while an agent runs",
}
});
test("editing a queued draft restores its original height with the focus hint visible", async ({
page,
}) => {
await configureComposerQueueSend(page);
await resizeComposerViewport(page, "portrait");
const agent = await startRunningMockAgent(page, {
prefix: "composer-queued-draft-height-",
model: "one-minute-stream",
prompt: "Stay running while a draft is queued.",
});
try {
const draft = "Please review the latest changes.\nPlease review the latest changes.";
await fillComposerDraft(page, draft);
const originalHeight = await composerHeight(page);
await queueComposerWithButton(page);
await expect(composerLocator(page)).toHaveValue("");
await composerLocator(page).blur();
await expect(page.getByText(/ to focus$/)).toBeVisible();
await page.getByRole("button", { name: "Edit queued message", exact: true }).click();
await expect(composerLocator(page)).toHaveValue(draft);
await expectComposerHeight(page, originalHeight);
} finally {
await agent.cleanup();
}
});
test("image paste stays attached to the active editor across fullscreen entry and exit", async ({
page,
}) => {
@@ -1,4 +1,21 @@
import { execFileSync } from "node:child_process";
import { mkdtempSync, rmSync } from "node:fs";
import { tmpdir } from "node:os";
import path from "node:path";
import type { createCreationScenario } from "../support/helpers/creation";
import { gotoAppShell } from "../support/helpers/app";
import { fillComposerDraft } from "../support/helpers/composer";
import {
connectNewWorkspaceDaemonClient,
openNewWorkspaceComposer,
selectWorkspaceIsolation,
} from "../support/helpers/new-workspace";
import {
seedTerminalProfiles,
selectLaunchOption,
submitTerminalLaunch,
} from "../support/helpers/new-workspace-launch";
import { waitForSidebarHydration } from "../support/helpers/workspace-ui";
import { expect } from "../support/fixtures";
import { test } from "../support/creation-fixtures";
@@ -116,12 +133,14 @@ for (const scenario of [
test(`retrying failed agent initialization preserves its workspace${scenario.suffix}`, async ({
creation,
startup,
page,
}) => {
await creation.openWorkspaceForm("local");
await creation.submitPrompt("Retry this workspace and agent together.", "Create");
await creation.expectAgentStillStarting();
await startup.fail();
await creation.expectStartupFailure();
await expect(page.getByTestId("app-toast")).toHaveCount(0);
await creation.expectOneCreatedWorkspace();
await scenario.prepare(creation);
await creation.submitPrompt("Retry this workspace and agent together.");
@@ -180,6 +199,105 @@ test("an edited agent draft starts after its first start failed", async ({
await expect(page.getByText(/request_key_conflict/)).toHaveCount(0);
});
test("editing a failed New workspace submission starts without a request key conflict", async ({
page,
}) => {
// An empty git repository naturally rejects worktree creation before workspace_ready.
const repoDir = mkdtempSync(path.join(tmpdir(), "paseo-empty-git-ws-"));
execFileSync("git", ["init"], { cwd: repoDir, stdio: "ignore" });
const client = await connectNewWorkspaceDaemonClient();
let projectId: string | null = null;
try {
const seeded = await client.createWorkspace({
source: { kind: "directory", path: repoDir },
});
if (!seeded.workspace) throw new Error(seeded.error ?? "Failed to seed empty git project");
projectId = seeded.workspace.projectId;
const listed = await client.listProjects();
const project = listed.projects.find((entry) => entry.projectId === projectId);
if (!project?.projectKey) throw new Error("Seeded project has no projectKey");
await gotoAppShell(page);
await waitForSidebarHydration(page);
await openNewWorkspaceComposer(page, {
projectKey: project.projectKey,
projectDisplayName: seeded.workspace.projectDisplayName,
});
await selectWorkspaceIsolation(page, "worktree");
await fillComposerDraft(page, "Start original worktree task.");
await page.getByRole("button", { name: "Create", exact: true }).click();
const composerError = page.getByTestId("composer-error").filter({ visible: true }).first();
await expect(composerError).toContainText("Unable to resolve repository default branch");
// The failure renders once in the composer alert, not duplicated in a top toast.
await expect(page.getByText("Unable to resolve repository default branch")).toHaveCount(1);
await selectWorkspaceIsolation(page, "local");
await fillComposerDraft(page, "hi");
await page.getByRole("button", { name: "Create", exact: true }).click();
await expect(page).toHaveURL(/\/workspace\//);
await expect(page.getByText(/request_key_conflict/)).toHaveCount(0);
await expect(
page.getByTestId("user-message").filter({ hasText: "hi", visible: true }),
).toBeVisible();
} finally {
if (projectId) await client.removeProject(projectId).catch(() => undefined);
await client.close();
rmSync(repoDir, { recursive: true, force: true });
}
});
test("switching only the launch target after a failed New workspace creation starts a new workspace", async ({
creation,
workspaceProvision,
page,
}) => {
const profiles = await seedTerminalProfiles([
{
id: "promptless-shell",
name: "Promptless Shell",
command: "sh",
args: ["-c", "echo ready; sleep 5"],
promptArg: "none",
},
]);
try {
await creation.openWorkspaceForm("local");
await workspaceProvision.fail();
await page.getByRole("button", { name: "Create", exact: true }).click();
await expect(
page.getByTestId("composer-error").filter({ visible: true }).first(),
).toContainText("Workspace provisioning failed for test");
await selectLaunchOption(page, "promptless-shell");
await submitTerminalLaunch(page);
await creation.expectOneCreatedWorkspace();
await expect(
page.locator('[data-testid^="workspace-tab-terminal_"]').filter({ visible: true }).first(),
).toBeVisible();
} finally {
await profiles.restore();
}
});
test("failing New workspace creation after switching workspaces shows an error toast", async ({
creation,
workspaceProvision,
page,
}) => {
await creation.openWorkspaceForm("local");
await workspaceProvision.hold();
await page.getByRole("button", { name: "Create", exact: true }).click();
await page.locator('[data-testid^="sidebar-workspace-row-"]').first().click();
await expect(page).toHaveURL(/\/workspace\//);
await workspaceProvision.fail();
await expect(page.getByTestId("app-toast")).toContainText(
"Workspace provisioning failed for test",
);
});
for (const entry of [
{
name: "agent draft",
@@ -1,3 +1,5 @@
import { mkdir, rm, writeFile } from "node:fs/promises";
import { dirname } from "node:path";
import { expect, test } from "../support/fixtures";
import { gotoAppShell, openSettings } from "../support/helpers/app";
import {
@@ -49,8 +51,10 @@ interface ProviderConfigClient {
}): Promise<unknown>;
}
test("adds and removes a custom model while the provider fails", async ({ page }) => {
test("adds and removes a custom model while the provider fails", async ({ page }, testInfo) => {
test.setTimeout(120_000);
const discoveryGate = testInfo.outputPath("hold-provider-discovery");
await mkdir(dirname(discoveryGate), { recursive: true });
const client = await connectDaemonClient<ProviderConfigClient>({
clientIdPrefix: "custom-model-failing-e2e",
});
@@ -60,7 +64,14 @@ test("adds and removes a custom model while the provider fails", async ({ page }
[FAILING_PROVIDER.id]: {
extends: "acp",
label: FAILING_PROVIDER.name,
command: ["node", "-e", "process.exit(1)"],
command: [
"node",
"-e",
'const { existsSync } = require("node:fs");' +
"const exitIfReleased = () => { if (!existsSync(process.argv[1])) process.exit(1); };" +
"exitIfReleased(); setInterval(exitIfReleased, 25);",
discoveryGate,
],
},
},
});
@@ -76,9 +87,15 @@ test("adds and removes a custom model while the provider fails", async ({ page }
await expect(page.getByRole("textbox", { name: "Search models" })).toHaveCount(0);
});
await addCustomModel(page, shortModelId);
await test.step("saving closes the Add sheet while provider discovery is still pending", async () => {
await writeFile(discoveryGate, "");
await addCustomModel(page, shortModelId);
});
await rm(discoveryGate);
await expect(page.getByTestId("provider-error")).toBeVisible({ timeout: 30_000 });
await removeListedCustomModel(page, shortModelId);
} finally {
await rm(discoveryGate, { force: true });
await client
.patchDaemonConfig({ removeProviders: [FAILING_PROVIDER.id] })
.catch(() => undefined);
@@ -118,7 +118,8 @@ async function exerciseProviderSettingsStack(page: Page) {
await page.getByRole("button", { name: "Refresh", exact: true }).click();
await expectProviderSettingsVisible(page);
await expect(page.getByText("Checked just now", { exact: true })).toBeVisible({
// The Models header puts the model count before it: "12 · Checked just now".
await expect(page.getByText(/Checked just now$/)).toBeVisible({
timeout: 30_000,
});
}
@@ -0,0 +1,101 @@
import { expect, type Locator, type Page } from "@playwright/test";
import { test } from "../support/fixtures";
import { runWorkspaceActionFromCommandCenter } from "../support/helpers/command-center-workspace-actions";
import { clickNewChat, gotoWorkspace } from "../support/helpers/launcher";
import { seedWorkspace } from "../support/helpers/seed-client";
import { waitForWorkspaceTabsVisible } from "../support/helpers/workspace-tabs";
const DRAG_DISTANCE = -40;
function splitDividers(page: Page): Locator {
return page.getByTestId("workspace-split-resize-handle").filter({ visible: true });
}
async function dividerX(divider: Locator): Promise<number> {
const box = await divider.boundingBox();
if (!box) throw new Error("Divider has no bounding box");
return box.x;
}
async function dragDivider(divider: Locator, deltaX: number): Promise<void> {
const page = divider.page();
const box = await divider.boundingBox();
if (!box) throw new Error("Divider has no bounding box");
const x = box.x + box.width / 2;
const y = box.y + box.height / 2;
await page.mouse.move(x, y);
await page.mouse.down();
await page.mouse.move(x + deltaX / 2, y, { steps: 4 });
await page.mouse.move(x + deltaX, y, { steps: 4 });
await page.mouse.up();
}
async function expectDividerFollowsDrag({
page,
index,
}: {
page: Page;
index: number;
}): Promise<void> {
const divider = splitDividers(page).nth(index);
const before = await dividerX(divider);
await dragDivider(divider, DRAG_DISTANCE);
await expect
.poll(async () => (await dividerX(divider)) - before, {
message: `divider ${index + 1} should follow the drag`,
})
.toBeLessThan(DRAG_DISTANCE * 0.75);
}
async function splitRight({
page,
expectedDividers,
}: {
page: Page;
expectedDividers: number;
}): Promise<void> {
await runWorkspaceActionFromCommandCenter(page, "Split pane right");
await expect(splitDividers(page)).toHaveCount(expectedDividers);
}
test.describe("Split pane resize", () => {
test("every divider drags in a four-pane row", async ({ page }) => {
await page.setViewportSize({ width: 1600, height: 900 });
const workspace = await seedWorkspace({ repoPrefix: "split-resize-fresh-" });
try {
await gotoWorkspace(page, workspace.workspaceId);
await waitForWorkspaceTabsVisible(page);
await clickNewChat(page);
await splitRight({ page, expectedDividers: 1 });
await splitRight({ page, expectedDividers: 2 });
await splitRight({ page, expectedDividers: 3 });
for (const index of [0, 1, 2]) {
await expectDividerFollowsDrag({ page, index });
}
} finally {
await workspace.cleanup();
}
});
test("every divider drags after panes are added to a resized row", async ({ page }) => {
await page.setViewportSize({ width: 1600, height: 900 });
const workspace = await seedWorkspace({ repoPrefix: "split-resize-grown-" });
try {
await gotoWorkspace(page, workspace.workspaceId);
await waitForWorkspaceTabsVisible(page);
await clickNewChat(page);
await splitRight({ page, expectedDividers: 1 });
await expectDividerFollowsDrag({ page, index: 0 });
await splitRight({ page, expectedDividers: 2 });
await splitRight({ page, expectedDividers: 3 });
for (const index of [0, 1, 2]) {
await expectDividerFollowsDrag({ page, index });
}
} finally {
await workspace.cleanup();
}
});
});
+138 -19
View File
@@ -34,6 +34,7 @@ import {
openUsageFromItem,
pinRow,
openUsageOptions,
openUsageSources,
closeUsageOptions,
pickUsageHost,
usageModalHost,
@@ -250,16 +251,15 @@ test.describe("usage modal", () => {
await expect(alphaHandle).toHaveCSS("opacity", "1");
await alphaHandle.blur();
// The off card collapses to its header, which still names the account.
await setUsageCardOn(page, "beta plan (work)", false);
const work = page.getByTestId("usage-report-beta:work");
await expect(work.getByText("12% used")).toHaveCount(0);
await expect(work.getByText("work", { exact: true })).toBeVisible();
// Turning a source off in the dedicated Sources list removes its card and sidebar summary.
await setUsageCardOn(page, "beta:work", false);
await expectUsageCardOrder(page, ["alpha:a", "beta:home"]);
await expect(page.getByTestId("usage-report-beta:work")).toHaveCount(0);
await expectPinnedUsage(page, ["31% Weekly", "20% Weekly"]);
const stopRecording = await recordUsageCardPositionsAfterRelease(page);
await dragUsageCardAbove(page, "beta:home", "alpha:a");
await expectUsageCardOrder(page, ["beta:home", "alpha:a", "beta:work"]);
await expectUsageCardOrder(page, ["beta:home", "alpha:a"]);
await expectPinnedUsage(page, ["20% Weekly", "31% Weekly"]);
// The cards are in their new slots from the first frame after the drop.
const framesAfterDrop = await stopRecording();
@@ -270,10 +270,114 @@ test.describe("usage modal", () => {
await page.reload();
await expectPinnedUsage(page, ["20% Weekly", "31% Weekly"]);
await openUsageFromItem(page);
await expectUsageCardOrder(page, ["beta:home", "alpha:a", "beta:work"]);
await expect(page.getByTestId("usage-report-beta:work").getByText("12% used")).toHaveCount(0);
await expectUsageCardOrder(page, ["beta:home", "alpha:a"]);
await expect(page.getByTestId("usage-report-beta:work")).toHaveCount(0);
});
for (const [size, viewport] of Object.entries({
desktop: { width: 1440, height: 900 },
compact: { width: 390, height: 844 },
})) {
test(`dedicated Sources list manages default-off sources without clutter (${size})`, async ({
page,
}) => {
test.setTimeout(120_000);
const [claude, codex] = claudeAndCodexReports();
const fetchedAt = twoHoursAgo();
await installUsageReportsFixture(page, {
lists: [
() => [
claude!,
codex!,
{
id: "copilot:default",
account: {},
fetchedAt,
sourceId: "copilot",
sourceLabel: "GitHub Copilot",
defaultEnabled: false,
report: {
status: "unavailable",
problem: { kind: "rejected", status: 403 },
},
},
{
id: "cursor:default",
account: { label: "dev@example.com" },
fetchedAt,
sourceId: "cursor",
sourceLabel: "Cursor",
defaultEnabled: false,
report: {
status: "available",
windows: [
{ id: "monthly", label: "Monthly", shortLabel: "mo", summary: true, usedPct: 18 },
],
},
},
],
],
});
await page.setViewportSize(viewport);
await gotoAppShell(page);
if (size === "compact") await openCompactSidebar(page);
// Only active sources appear in the sidebar summary and modal cards.
await expectPinnedUsage(page, ["31% 5h", "54% wk", "7% 5h", "12% wk"]);
await openUsageFromItem(page);
await expectUsageCardOrder(page, ["claude:default", "codex:default"]);
await expect(page.getByTestId("usage-report-copilot:default")).toHaveCount(0);
await expect(page.getByTestId("usage-report-cursor:default")).toHaveCount(0);
await qaScreenshot(page, `usage-sources-${size}-cards`);
await openUsageOptions(page);
await expect(page.getByTestId("usage-options-sources")).toContainText("2 of 4");
await qaScreenshot(page, `usage-sources-${size}-settings`);
if (size === "compact") {
// Compact sheets expose menu rows in the browser Tab order so keyboard users can open
// Sources and toggle checkboxes without a pointer.
await page.getByTestId("usage-options-sources").focus();
await expect(page.getByTestId("usage-options-sources")).toBeFocused();
await expect(page.getByTestId("usage-options-sources")).toHaveAttribute("tabindex", "0");
await page.keyboard.press("Enter");
} else {
await openUsageSources(page);
}
const claudeToggle = page.getByRole("checkbox", { name: /^Claude/ });
const copilotToggle = page.getByRole("checkbox", { name: /^GitHub Copilot/ });
const cursorToggle = page.getByRole("checkbox", { name: /^Cursor/ });
await expect(claudeToggle).toBeChecked();
await expect(copilotToggle).not.toBeChecked();
await expect(copilotToggle).toContainText("Provider inactive");
await qaScreenshot(page, `usage-sources-${size}-list`);
// Opting into a default-off source brings its card and summary window in immediately.
if (size === "compact") {
await cursorToggle.focus();
await expect(cursorToggle).toBeFocused();
await expect(cursorToggle).toHaveAttribute("tabindex", "0");
await page.keyboard.press("Space");
} else {
await cursorToggle.click();
}
await expect(cursorToggle).toBeChecked();
await closeUsageOptions(page);
await expectUsageCardOrder(page, ["claude:default", "codex:default", "cursor:default"]);
// Hiding every source shows a quiet empty state with a shortcut back to Settings.
await openUsageSources(page);
await page.getByTestId("usage-source-toggle-claude:default").click();
await page.getByTestId("usage-source-toggle-codex:default").click();
await page.getByTestId("usage-source-toggle-cursor:default").click();
await closeUsageOptions(page);
await expect(page.getByTestId("usage-all-hidden")).toBeVisible();
await expect(page.getByTestId("usage-choose-sources")).toBeVisible();
await qaScreenshot(page, `usage-sources-${size}-empty`);
await page.getByTestId("usage-choose-sources").click();
await expect(page.getByTestId("usage-options-sources")).toContainText("0 of 4");
});
}
test("refreshes one report from its card", async ({ page }) => {
test.setTimeout(120_000);
const serverId = getServerId();
@@ -836,6 +940,8 @@ test("usage percentages state their meaning and follow the selected language", a
}
});
const exhaustionClock = /^runs out (?:[A-Z][a-z]{2} )?\d{1,2}:\d{2}(?: [AP]M)?$/;
test("open usage clocks advance without refreshing or reopening", async ({ page }) => {
test.setTimeout(120_000);
const now = new Date();
@@ -853,12 +959,16 @@ test("open usage clocks advance without refreshing or reopening", async ({ page
const card = page.getByTestId("usage-report-login-journey:account");
await expect(card.getByTestId("usage-freshness")).toHaveText("Updated just now");
await expect(card.getByText("resets in 30m", { exact: false }).first()).toBeVisible();
const forecast = card.getByText(exhaustionClock);
await expect(forecast).toBeVisible();
const forecastTime = (await forecast.innerText()).replace(/^[^0-9]+/, "");
// The shared minute clock is wall-aligned, and the daemon timestamps use real time.
// After two browser minutes the latest displayed tick is one or two minutes old.
await advanceUsageClock(page, 120_000);
await expect(card.getByTestId("usage-freshness")).toHaveText(/^Updated [12]m ago$/);
await expect(card.getByText(/resets in (28|29)m/).first()).toBeVisible();
await expect(card.getByText(/runs out in (28|29)m/)).toBeVisible();
// The weekday drops off at local midnight, but the forecast time stays fixed.
await expect(forecast).toHaveText(new RegExp(` ${forecastTime}$`));
await expect(card.getByTestId("provider-usage-balance-credits-reset")).toHaveText(
/^resets in (28|29)m\s·\s/,
);
@@ -867,7 +977,7 @@ test("open usage clocks advance without refreshing or reopening", async ({ page
}
});
test("Refresh all keeps quota visible and restores the forecast after a timeout", async ({
test("Refresh all keeps quota visible after a timeout and shows only reset details when back on track", async ({
page,
}) => {
const reading = new Date();
@@ -875,7 +985,7 @@ test("Refresh all keeps quota visible and restores the forecast after a timeout"
try {
const transport = await installDaemonWebSocketGate(page);
await openCompactUsageWithClock(page, reading);
await expectQuota(page, "40% used", /runs out in /);
await expectQuota(page, "40% used", exhaustionClock);
await timeOutUsageRefresh(page, transport);
await expectQuota(page, "40% used", null);
@@ -884,7 +994,7 @@ test("Refresh all keeps quota visible and restores the forecast after a timeout"
await login.setReport(usageReading(reading, 10));
transport.releaseHeldClientRequest();
await refreshAllUsage(page);
await expectQuota(page, "10% used", "~50% left at reset");
await expectQuota(page, "10% used", null);
await expectUsageRefreshError(page).toBeHidden();
} finally {
await login.cleanup();
@@ -897,23 +1007,29 @@ test("a card refresh timeout is visible alongside a newer quota forecast", async
try {
const transport = await installDaemonWebSocketGate(page);
await openCompactUsageWithClock(page, reading);
await expectQuota(page, "40% used", /runs out in /);
await expectQuota(page, "40% used", exhaustionClock);
await delayCardRefreshResponse(page, transport);
await login.setReport(usageReading(reading, 10));
await login.setReport(usageReading(reading, 45));
await refreshAllUsage(page);
await expectQuota(page, "10% used", "~50% left at reset");
await expectQuota(page, "45% used", exhaustionClock);
await advanceUsageClock(page, 61_000);
await expectUsageRefreshError(page).toBeVisible();
await expectQuota(page, "10% used", "~50% left at reset");
await expect(
usageModal(page)
.getByTestId("usage-report-login-journey:account")
.getByTestId("usage-refresh-error"),
).toHaveText("Unable to refresh usage");
await expectQuota(page, "45% used", exhaustionClock);
} finally {
await login.cleanup();
}
});
function expectUsageRefreshError(page: Page) {
return expect(usageModal(page).getByText("Unable to refresh usage", { exact: true }));
return expect(
usageModal(page).getByRole("alert").getByText("Unable to refresh usage", { exact: true }),
);
}
async function delayCardRefreshResponse(
@@ -947,7 +1063,10 @@ async function expectQuota(page: Page, amount: string, forecast: string | RegExp
await expect(usage.getByText(amount, { exact: true })).toBeVisible();
await expect(usage.getByText(/^resets in /)).toBeVisible();
if (forecast === null) {
await expect(usage.getByText(/^runs out in /)).toBeHidden();
await expect(usage.getByRole("checkbox", { name: /^Pin Claude Session/ })).toHaveAccessibleName(
/^Pin Claude Session, \d+% used · resets in [^·]+ · [^·]+$/,
);
await expect(usage.getByText(/^runs out /)).toBeHidden();
return;
}
await expect(usage.getByText(forecast)).toBeVisible();
@@ -291,12 +291,12 @@ test.describe("Usage item", () => {
await gotoAppShell(page);
const screen = page.getByTestId(`usage-host-${serverId}`);
await test.step("a fresh device shows default windows once the Usage item is on", async () => {
// The Usage item starts off on every layout; turning it on stores the choice.
await expectNoUsageItem(page);
await test.step("a fresh device shows default windows in the sidebar automatically", async () => {
// The Usage item starts on by default on every layout.
await expectPinnedUsage(page, ["31% 5h", "54% wk", "7% 5h", "12% wk"]);
await qaScreenshot(page, "desktop-footer-defaults", { kind: "footer" });
await openSidebarNavSettings(page);
await expectFooterItemSetting(page, "usage", false);
await setFooterItemVisible(page, "usage", true);
await expectFooterItemSetting(page, "usage", true);
await qaScreenshot(page, "desktop-settings-sidebar-footer");
await page.setViewportSize(COMPACT);
await expectFooterItemSetting(page, "usage", true);
@@ -308,7 +308,6 @@ test.describe("Usage item", () => {
await page.setViewportSize(WIDE);
await gotoAppShell(page);
await expectPinnedUsage(page, ["31% 5h", "54% wk", "7% 5h", "12% wk"]);
await qaScreenshot(page, "desktop-footer-defaults", { kind: "footer" });
await openUsageFromItem(page);
await qaScreenshot(page, "default-pins-card");
await expect(pinRow(screen, "Claude", "Session")).toBeChecked();
@@ -478,7 +477,7 @@ test("without summary data the footer drops the Usage item and keeps the Usage i
await openUsageFromIcon(page);
});
test("the Usage Settings switch turns on the sidebar summary and the pins with it", async ({
test("the Usage Settings switch turns the sidebar summary and pins off and back on", async ({
page,
}) => {
const serverId = getServerId();
@@ -488,8 +487,17 @@ test("the Usage Settings switch turns on the sidebar summary and the pins with i
const screen = page.getByTestId(`usage-host-${serverId}`);
const claude = screen.getByTestId("usage-report-claude:default");
await test.step("off by default: no Usage item, and rows that do not pin", async () => {
await openUsageFromIcon(page);
await test.step("on by default: the Usage item and pins appear", async () => {
await expectPinnedUsage(page, ["31% 5h", "54% wk", "7% 5h", "12% wk"]);
await openUsageFromItem(page);
await expect(pinRow(screen, "Claude", "Weekly")).toBeChecked();
await togglePin(screen, "Claude", "Weekly");
await expectPinnedUsage(page, ["31% 5h", "7% 5h", "12% wk"]);
await qaScreenshot(page, "usage-summary-on");
});
await test.step("off: no Usage item, and rows that do not pin", async () => {
await setSummaryInSidebar(page, false);
await expectNoUsageItem(page);
await expect(claude.getByText("Session", { exact: true })).toBeVisible({ timeout: 10_000 });
await expectUnpinnableRows(claude);
@@ -498,19 +506,7 @@ test("the Usage Settings switch turns on the sidebar summary and the pins with i
await qaScreenshot(page, "usage-summary-off");
});
await test.step("on: the Usage item and the pins appear", async () => {
await setSummaryInSidebar(page, true);
await expectPinnedUsage(page, ["31% 5h", "54% wk", "7% 5h", "12% wk"]);
await expect(pinRow(screen, "Claude", "Weekly")).toBeChecked();
await togglePin(screen, "Claude", "Weekly");
await expectPinnedUsage(page, ["31% 5h", "7% 5h", "12% wk"]);
await qaScreenshot(page, "usage-summary-on");
});
await test.step("off then on again, after a reload, brings back the saved pins", async () => {
await setSummaryInSidebar(page, false);
await expectNoUsageItem(page);
await expectUnpinnableRows(claude);
await test.step("after a reload, turning it back on brings back the saved pins", async () => {
await page.reload();
await openUsageFromIcon(page);
await expectSummaryInSidebar(page, false);
@@ -554,6 +550,7 @@ test("released hosts supply source logos through the client conversion", async (
test("hidden usage sends no requests until opened", async ({ page }) => {
test.setTimeout(180_000);
const usage = await installUsageReportsFixture(page, { lists: [() => claudeAndCodexReports()] });
await seedSidebarFooterPreferences(page, [{ key: "usage", visible: false }]);
await page.setViewportSize(WIDE);
await gotoAppShell(page);
await expectNoUsageItem(page);
@@ -569,10 +566,9 @@ test("hidden usage sends no requests until opened", async ({ page }) => {
expect(usage.listRequests()).toHaveLength(1);
});
test("visible usage summary requests reports on load", async ({ page }) => {
test("visible usage summary requests reports on load by default", async ({ page }) => {
test.setTimeout(180_000);
const usage = await installUsageReportsFixture(page, { lists: [() => claudeAndCodexReports()] });
await seedSidebarFooterPreferences(page, [{ key: "usage", visible: true }]);
await page.setViewportSize(WIDE);
await gotoAppShell(page);
await expectPinnedUsage(page, ["31% 5h", "54% wk", "7% 5h", "12% wk"]);
@@ -15,6 +15,11 @@ import { delayBrowserAgentCreatedStatus } from "./helpers/new-workspace";
export const test = base.extend<{
startup: { release(): Promise<void>; fail(): Promise<void> };
workspaceProvision: {
hold(): Promise<void>;
fail(): Promise<void>;
release(): Promise<void>;
};
promptRejection: void;
composerMetadata: { set(patch: Record<string, string>): Promise<void> };
creation: Awaited<ReturnType<typeof createCreationScenario>>;
@@ -70,6 +75,57 @@ export default function contribute(server) {
await client.close();
}
},
workspaceProvision: async ({ page }, provide) => {
void page;
const client = await connectNewWorkspaceDaemonClient();
const plugin = await copyPluginExample("lifecycle-logger");
const gate = path.join(plugin.directory, "workspace-provision.txt");
const holdFile = path.join(plugin.directory, "workspace-provision-hold.txt");
const failFile = path.join(plugin.directory, "workspace-provision-fail.txt");
const previous = await client.getDaemonConfig();
try {
await writeFile(
path.join(plugin.directory, "index.server.ts"),
`
import { readFile, rm } from "node:fs/promises";
export default function contribute(server) {
return server.before("workspace.create", async ({ request }) => {
const gate = ${JSON.stringify(gate)};
let command = await readFile(gate, "utf8").catch(() => "release");
while (command === "hold") {
await new Promise(resolve => setTimeout(resolve, 20));
command = await readFile(gate, "utf8").catch(() => "release");
}
if (command === "fail") {
await rm(gate, { force: true });
throw new Error("Workspace provisioning failed for test");
}
return request;
});
}
`,
);
await client.patchDaemonConfig({ pluginsEnabled: true });
await client.installDirectoryPlugin(plugin.directory);
await provide({
hold: async () => {
await writeFile(holdFile, "hold");
await rename(holdFile, gate);
},
fail: async () => {
await writeFile(failFile, "fail");
await rename(failFile, gate);
},
release: () => rm(gate, { force: true }),
});
} finally {
await rm(gate, { force: true });
await client.removePlugin("lifecycle-logger");
await client.patchDaemonConfig({ pluginsEnabled: previous.config.pluginsEnabled });
await plugin.cleanup();
await client.close();
}
},
promptRejection: async ({ page }, provide) => {
void page;
const client = await connectNewWorkspaceDaemonClient();
@@ -19,8 +19,6 @@ export async function addCustomModel(page: Page, id: string): Promise<void> {
.getByRole("button", { name: "Add", exact: true })
.click();
await expect(page.getByRole("button", { name: `Remove ${id}`, exact: true })).toHaveCount(1);
// Config updates before provider refresh finishes and dismisses the add sheet.
// Resizing or interacting with the model list must wait for that save to finish.
await expect(page.getByTestId("add-custom-model-sheet")).toBeHidden();
}
@@ -6,6 +6,7 @@ import {
type ChildProcess,
type SpawnOptions,
} from "node:child_process";
import { existsSync } from "node:fs";
import { mkdtemp, readFile, rm, writeFile } from "node:fs/promises";
import net from "node:net";
import { tmpdir } from "node:os";
@@ -229,7 +230,11 @@ export async function startIsolatedHostDaemon(
stdio: ["ignore", "ignore", "pipe"],
detached: false,
};
const spawned = publishedPackageRoot
const builtSupervisor = path.join(serverDir, "dist/scripts/supervisor-entrypoint.js");
const useBuiltSupervisor = Boolean(
publishedPackageRoot || (process.env.CI && existsSync(builtSupervisor)),
);
const spawned = useBuiltSupervisor
? spawn(process.execPath, ["dist/scripts/supervisor-entrypoint.js"], spawnOptions)
: spawnTsx("scripts/supervisor-entrypoint.ts", ["--dev"], spawnOptions);
+3 -1
View File
@@ -567,7 +567,9 @@ export async function expectProviderPageLoaded(page: Page, providerName: string)
await expect(page.getByTestId("provider-detail-title")).toHaveText(providerName, {
timeout: 30_000,
});
await expect(page.getByText("Provider status")).toBeVisible({ timeout: 30_000 });
await expect(page.getByRole("button", { name: "Diagnostic report" })).toBeVisible({
timeout: 30_000,
});
}
export async function expectProviderNotOnHost(page: Page): Promise<void> {
@@ -160,11 +160,23 @@ export async function expectUsageCardOrder(page: Page, reportIds: string[]): Pro
.toEqual(reportIds);
}
/** Turns a card on or off from its header switch, named by its source and account. */
export async function setUsageCardOn(page: Page, name: string, on: boolean): Promise<void> {
const cardSwitch = usageModal(page).getByRole("switch", { name: `Show ${name}`, exact: true });
await cardSwitch.click();
await expect(cardSwitch).toBeChecked({ checked: on });
/** Turns a source on or off from the Sources page in Usage Settings, by report id. */
export async function setUsageCardOn(page: Page, reportId: string, on: boolean): Promise<void> {
await openUsageSources(page);
const item = visible(page, `usage-source-toggle-${reportId}`);
const checked = (await item.getAttribute("aria-checked")) === "true";
if (checked !== on) {
await item.click();
await expect(item).toHaveAttribute("aria-checked", String(on));
}
await closeUsageOptions(page);
}
/** Opens the Sources submenu inside Usage Settings (flyout on wide, in-place page on compact). */
export async function openUsageSources(page: Page): Promise<void> {
await openUsageOptions(page);
await visible(page, "usage-options-sources").click();
await expect(page.locator('[data-testid^="usage-source-toggle-"]:visible').first()).toBeVisible();
}
/** Drags a card by its handle until it sits above another card, with the mouse. */
@@ -245,18 +257,14 @@ export function pinRow(scope: Locator, source: string, window: string): Locator
/**
* A report card whose window rows do not pin: no pin toggle, no pin glyph, and nothing focusable
* but the header's buttons and on/off switch, so no row presses or highlights on hover.
* but the header's buttons, so no row presses or highlights on hover.
*/
export async function expectUnpinnableRows(card: Locator): Promise<void> {
await expect(card.getByRole("checkbox")).toHaveCount(0);
await expect(card.locator('[data-testid^="usage-pin-"]')).toHaveCount(0);
const focusable = await card
.locator("[tabindex]")
.evaluateAll(
(nodes) =>
nodes.filter((node) => !["button", "switch"].includes(node.getAttribute("role") ?? ""))
.length,
);
.evaluateAll((nodes) => nodes.filter((node) => node.getAttribute("role") !== "button").length);
expect(focusable).toBe(0);
}
@@ -388,6 +396,7 @@ export async function closeUsageOptions(page: Page): Promise<void> {
await waitForSettledPosition(usageOptionsSurface(page));
await page.mouse.click(0, 0);
await expect(page.getByTestId("usage-display-used")).toHaveCount(0);
await expect(page.locator('[data-testid^="usage-source-toggle-"]:visible')).toHaveCount(0);
}
export async function showUsageAs(page: Page, displayAs: "used" | "remaining") {
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@getpaseo/app",
"version": "0.17.0",
"version": "0.17.2",
"private": true,
"main": "index.ts",
"scripts": {
+19
View File
@@ -29,6 +29,25 @@ export function agentCommandsQueryRoot(serverId: string) {
return [AGENT_COMMANDS_QUERY_ROOT, serverId] as const;
}
function queryCommandsCwd(query: { queryKey: readonly unknown[]; meta?: Record<string, unknown> }) {
const isDraft = query.queryKey[2] === "draft";
const cwd = query.meta?.agentCommandCwd ?? (isDraft ? query.queryKey[4] : undefined);
return typeof cwd === "string" ? normalizeAgentCommandsCwd(cwd) : undefined;
}
/** Marks every catalog for a directory stale. Only observed catalogs refetch; the rest refetch
* when next mounted. Used when a provider change is announced without its new snapshot. */
export function invalidateCommandsForDirectory(input: {
queryClient: QueryClient;
serverId: string;
cwd: string | null;
}): Promise<void> {
return input.queryClient.invalidateQueries({
queryKey: agentCommandsQueryRoot(input.serverId),
predicate: (query) => !input.cwd || queryCommandsCwd(query) === input.cwd,
});
}
/** Snapshot announcements contain provider metadata, not command catalogs. Ignore initial
* discovery and unrelated providers; explicit refresh still invalidates every catalog. */
export function invalidateCommandsForProviderChange(input: {
@@ -393,6 +393,39 @@ describe("layoutStream", () => {
expect(findLayoutItem(layout, thinking.id).toolSequence).toBe("last");
});
it.each(["web", "android"] as const)(
"carries turn metrics from the footer's source row on %s",
(platform) => {
const metrics = [{ label: "Model", value: "opus" }];
const earlier = { ...assistantMessage("a1", 2), turnMetrics: metrics };
const latest = { ...toolCall("tool-2", 5), turnMetrics: [{ label: "Steps", value: "2" }] };
const layout = layoutFor({
platform,
tail: [userMessage("u1", 1), earlier, toolCall("tool-1", 3), userMessage("u2", 4), latest],
});
expect(findLayoutItem(layout, "tool-1").completedFooter?.turnMetrics).toBe(metrics);
expect(layout.auxiliaryTurnFooter?.turnMetrics).toBe(latest.turnMetrics);
},
);
it("lays out the footer again when its source row gains turn metrics", () => {
const assistant = assistantMessage("a1", 2);
const trailing = toolCall("tool-2", 4);
const tail = [userMessage("u1", 1), assistant, toolCall("tool-1", 3), trailing];
const next = userMessage("u2", 5);
const before = layoutFor({ platform: "android", tail: [...tail, next] });
const metrics = [{ label: "Model", value: "opus" }];
const stamped = [...tail];
stamped[1] = { ...assistant, turnMetrics: metrics };
const after = layoutFor({ platform: "android", tail: [...stamped, next] });
expect(findLayoutItem(before, trailing.id).completedFooter?.turnMetrics).toBeUndefined();
expect(findLayoutItem(after, trailing.id).completedFooter?.turnMetrics).toBe(metrics);
expect(findLayoutItem(after, next.id)).toBe(findLayoutItem(before, next.id));
});
it("keeps bottom and inline footer ownership mutually exclusive", () => {
const assistant = assistantMessage("a1", 2);
const layout = layoutFor({
+8
View File
@@ -1,3 +1,4 @@
import type { AgentMetric } from "@getpaseo/protocol/agent-types";
import type { TurnTiming } from "@/timeline/turn-time";
import type { StreamItem } from "@/types/stream";
import { getAssistantMessageSpacing, getGapBetweenStreamItems } from "./spacing";
@@ -10,6 +11,7 @@ export interface TurnFooterHost {
itemId: string;
items: StreamItem[];
timing?: TurnTiming;
turnMetrics?: AgentMetric[];
startIndex: number;
}
@@ -69,10 +71,15 @@ function createTurnFooterHost(input: {
index: number;
timingByItemId: Map<string, TurnTiming>;
}): TurnFooterHost {
const turnMetrics =
input.item.kind === "assistant_message" || input.item.kind === "tool_call"
? input.item.turnMetrics
: undefined;
return {
itemId: input.item.id,
items: input.items,
timing: input.timingByItemId.get(input.item.id),
...(turnMetrics ? { turnMetrics } : {}),
startIndex: input.index,
};
}
@@ -247,6 +254,7 @@ function areTurnFooterHostsEqual(
return (
left.itemId === right.itemId &&
left.timing === right.timing &&
left.turnMetrics === right.turnMetrics &&
left.startIndex === right.startIndex &&
left.items === right.items
);
@@ -1,6 +1,10 @@
import React, { act } from "react";
import { createRoot, type Root } from "react-dom/client";
import { afterEach, beforeEach, expect, it, vi } from "vitest";
import { PortalProvider } from "@gorhom/portal";
import type { AgentMetric } from "@getpaseo/protocol/agent-types";
import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
import { page } from "vitest/browser";
import { FloatingPanelPortalHost } from "@/components/ui/floating-panel-portal";
import { i18n } from "@/i18n/i18next";
import type { TurnTiming } from "@/timeline/turn-time";
import type { StreamItem } from "@/types/stream";
@@ -92,3 +96,57 @@ it.each([
"Friday 11:11 PM": 1,
});
});
const COMPLETED_TIMING: TurnTiming = { completedAt: FRIDAY_NIGHT, durationMs: 26_000 };
describe("turn stats", () => {
const metrics: AgentMetric[] = [
{ label: "Model", value: "opus" },
{ label: "Cost", value: "$0.42", tone: "warning" },
];
function renderFooterWithMetrics(turnMetrics?: AgentMetric[]): void {
act(() =>
root.render(
<PortalProvider>
<CompletedTurnFooterRow
strategy={createNativeStreamStrategy()}
items={[reply]}
timing={COMPLETED_TIMING}
startIndex={0}
supportsTimelineCursor={false}
turnMetrics={turnMetrics}
/>
<FloatingPanelPortalHost />
</PortalProvider>,
),
);
}
it("has no trigger when the provider reported nothing for the turn", () => {
renderFooterWithMetrics(undefined);
expect(page.getByRole("button", { name: "Turn stats" }).elements()).toHaveLength(0);
});
it("shows the turn's metrics while the pointer is on the trigger", async () => {
renderFooterWithMetrics(metrics);
await page.getByRole("button", { name: "Turn stats" }).hover();
const card = page.getByRole("dialog", { name: "Turn stats" });
await expect.element(card).toBeVisible();
await expect.element(card.getByText("Model")).toBeVisible();
await expect.element(card.getByText("opus")).toBeVisible();
await expect.element(card.getByText("$0.42")).toBeVisible();
});
it("keeps the footer's height when the trigger appears", () => {
renderFooterWithMetrics(undefined);
const without = host.getBoundingClientRect().height;
renderFooterWithMetrics(metrics);
expect(host.getBoundingClientRect().height).toBe(without);
});
});
@@ -1,3 +1,4 @@
import type { AgentMetric } from "@getpaseo/protocol/agent-types";
import React, { memo, useCallback, useMemo, type ReactNode } from "react";
import { View } from "react-native";
import { StyleSheet, withUnistyles } from "react-native-unistyles";
@@ -81,6 +82,7 @@ export const TurnFooter = memo(function TurnFooter({
items={host.items}
timing={host.timing}
startIndex={host.startIndex}
turnMetrics={host.turnMetrics}
supportsTimelineCursor={supportsTimelineCursor}
onForkAssistantTurn={onForkAssistantTurn}
/>
@@ -92,6 +94,7 @@ export const CompletedTurnFooterRow = memo(function CompletedTurnFooterRow({
items,
timing,
startIndex,
turnMetrics,
supportsTimelineCursor,
onForkAssistantTurn,
}: {
@@ -99,6 +102,7 @@ export const CompletedTurnFooterRow = memo(function CompletedTurnFooterRow({
items: StreamItem[];
timing?: TurnTiming;
startIndex: number;
turnMetrics?: AgentMetric[];
supportsTimelineCursor: boolean;
onForkAssistantTurn?: AssistantTurnForkHandler;
}) {
@@ -109,6 +113,7 @@ export const CompletedTurnFooterRow = memo(function CompletedTurnFooterRow({
items={items}
timing={timing}
startIndex={startIndex}
turnMetrics={turnMetrics}
supportsTimelineCursor={supportsTimelineCursor}
onForkAssistantTurn={onForkAssistantTurn}
/>
@@ -165,6 +170,7 @@ function CompletedTurnFooter({
items,
timing,
startIndex,
turnMetrics,
supportsTimelineCursor,
onForkAssistantTurn,
}: {
@@ -172,6 +178,7 @@ function CompletedTurnFooter({
items: StreamItem[];
timing?: TurnTiming;
startIndex: number;
turnMetrics?: AgentMetric[];
supportsTimelineCursor: boolean;
onForkAssistantTurn?: AssistantTurnForkHandler;
}) {
@@ -205,6 +212,7 @@ function CompletedTurnFooter({
completedAt={timing?.completedAt}
durationMs={timing?.durationMs}
onFork={boundary && onForkAssistantTurn ? handleFork : undefined}
metrics={turnMetrics}
/>
</View>
);
+1
View File
@@ -188,6 +188,7 @@ function renderStreamItemWithTurnFooter(input: {
items={footerHost.items}
timing={footerHost.timing}
startIndex={footerHost.startIndex}
turnMetrics={footerHost.turnMetrics}
supportsTimelineCursor={input.supportsTimelineCursor}
onForkAssistantTurn={input.onForkAssistantTurn}
/>
@@ -0,0 +1,65 @@
import React, { act } from "react";
import { createRoot, type Root } from "react-dom/client";
import { QueryClient, QueryClientProvider } from "@tanstack/react-query";
import type { AgentMetric } from "@getpaseo/protocol/agent-types";
import { afterEach, beforeEach, expect, it, vi } from "vitest";
import { page } from "vitest/browser";
import { i18n } from "@/i18n/i18next";
import { ContextWindowDetails } from "./context-window-details";
const CONTEXT = { percentage: 40, usedTokens: 80_000, maxTokens: 200_000 };
let root: Root;
let host: HTMLDivElement;
beforeEach(async () => {
await i18n.changeLanguage("en");
vi.stubGlobal("React", React);
vi.stubGlobal("IS_REACT_ACT_ENVIRONMENT", true);
host = document.createElement("div");
document.body.appendChild(host);
root = createRoot(host);
});
afterEach(() => {
act(() => root.unmount());
host.remove();
vi.unstubAllGlobals();
});
function renderDetails(sessionMetrics: AgentMetric[]): void {
act(() =>
root.render(
<QueryClientProvider client={new QueryClient()}>
<ContextWindowDetails
serverId="server"
agentId="agent"
context={CONTEXT}
sessionCost="$1.20"
sessionMetrics={sessionMetrics}
showTitle
refreshable={false}
/>
</QueryClientProvider>,
),
);
}
it("shows what the provider reports about the session", async () => {
renderDetails([
{ label: "Model", value: "opus" },
{ label: "Requests", value: "12" },
]);
await expect.element(page.getByText("Model")).toBeVisible();
await expect.element(page.getByText("opus")).toBeVisible();
await expect.element(page.getByText("Requests")).toBeVisible();
await expect.element(page.getByText("Session cost $1.20")).toBeVisible();
});
it("leaves the cost to the provider when it reports one", async () => {
renderDetails([{ label: "Cost", value: "$1.25" }]);
await expect.element(page.getByText("$1.25")).toBeVisible();
expect(page.getByText("Session cost $1.20").elements()).toHaveLength(0);
});
@@ -1,8 +1,10 @@
import { Text, View } from "react-native";
import { StyleSheet } from "react-native-unistyles";
import { useTranslation } from "react-i18next";
import type { AgentMetric } from "@getpaseo/protocol/agent-types";
import { AgentUsage } from "@/usage";
import { formatTokenCount } from "./context-window-meter.utils";
import { MetricsList } from "./metrics-list";
interface ContextWindowDetailsProps {
serverId: string;
@@ -13,6 +15,8 @@ interface ContextWindowDetailsProps {
maxTokens: number;
} | null;
sessionCost: string | null;
/** What the provider reports about the session. A reported cost replaces `sessionCost`. */
sessionMetrics: AgentMetric[];
/** The hover card and tooltip title themselves; the sheet's header carries the title instead. */
showTitle: boolean;
/** Whether the usage cards have a Refresh button; without one they show their freshness. */
@@ -20,18 +24,22 @@ interface ContextWindowDetailsProps {
}
/**
* What the context window meter opens: how full the window is, then the usage of the agent's
* account. The desktop hover card, the native wide tooltip, and the compact sheet all render it.
* What the context window meter opens: how full the window is, what the provider reports about the
* session, then the usage of the agent's account. The desktop hover card, the native wide tooltip, and the compact sheet all render it.
*/
export function ContextWindowDetails({
serverId,
agentId,
context,
sessionCost,
sessionMetrics,
showTitle,
refreshable,
}: ContextWindowDetailsProps) {
const { t } = useTranslation();
const providerReportsCost = sessionMetrics.some(
(metric) => metric.label.toLowerCase() === "cost",
);
return (
<>
<View style={styles.summary}>
@@ -51,10 +59,11 @@ export function ContextWindowDetails({
) : (
<Text style={styles.detail}>{t("contextWindow.noData")}</Text>
)}
{sessionCost ? (
{sessionCost && !providerReportsCost ? (
<Text style={styles.detail}>{t("contextWindow.sessionCost", { cost: sessionCost })}</Text>
) : null}
</View>
{sessionMetrics.length > 0 ? <MetricsList metrics={sessionMetrics} /> : null}
<AgentUsage serverId={serverId} agentId={agentId} refreshable={refreshable} />
</>
);
@@ -13,11 +13,12 @@ import { HoverCard, HoverCardContent, HoverCardTrigger } from "@/components/ui/h
import { Tooltip, TooltipContent, TooltipTrigger } from "@/components/ui/tooltip";
import { useIsCompactFormFactor } from "@/constants/layout";
import { isNative } from "@/constants/platform";
import type { AgentMetric } from "@getpaseo/protocol/agent-types";
import { useHostReportsUsage } from "@/usage";
import type { Theme } from "@/styles/theme";
import { TOOLBAR_DISC_SIZE, type ToolbarRing } from "@/components/ui/control-geometry";
import { ContextWindowDetails } from "./context-window-details";
import { ContextWindowSheet } from "./context-window-sheet";
import { DetailsSheet } from "./details-sheet";
import { formatContextPercentage } from "./context-window-meter.utils";
interface ContextWindowMeterProps {
@@ -26,11 +27,14 @@ interface ContextWindowMeterProps {
maxTokens: number | null;
usedTokens: number | null;
totalCostUsd?: number | null;
/** What the provider reports about the session, shown in the details. */
sessionMetrics?: AgentMetric[];
showPercentage?: boolean;
/** The ring's box and stroke when it sits in an icon toolbar. */
toolbarRing?: ToolbarRing;
}
const EMPTY_METRICS: AgentMetric[] = [];
const SVG_SIZE = 14;
const USAGE_POPOVER_WIDTH = 300;
const COMPACT_SVG_SIZE = 12;
@@ -156,13 +160,16 @@ export function ContextWindowMeter({
maxTokens,
usedTokens,
totalCostUsd,
sessionMetrics = EMPTY_METRICS,
showPercentage = false,
toolbarRing,
}: ContextWindowMeterProps) {
const { t, i18n } = useTranslation();
const { width } = useWindowDimensions();
// Usage cards need a wider popover; without them it keeps the plain tooltip shape.
// Usage cards and session metrics need a wider popover; without them it keeps the plain tooltip
// shape.
const showsUsage = useHostReportsUsage(serverId);
const isWide = showsUsage || sessionMetrics.length > 0;
const popoverWidth = Math.min(USAGE_POPOVER_WIDTH, width - 24);
// Compact screens open the details in a sheet, which can hold a pressable Refresh.
const isCompact = useIsCompactFormFactor();
@@ -226,21 +233,22 @@ export function ContextWindowMeter({
{ring}
{percentageLabel}
</Pressable>
<ContextWindowSheet open={isSheetOpen} onClose={closeSheet}>
<DetailsSheet open={isSheetOpen} onClose={closeSheet} title={t("contextWindow.title")}>
<ContextWindowDetails
serverId={serverId}
agentId={agentId}
context={context}
sessionCost={formattedSessionCost}
sessionMetrics={sessionMetrics}
showTitle={false}
refreshable
/>
</ContextWindowSheet>
</DetailsSheet>
</>
);
}
const popoverStyle = showsUsage
const popoverStyle = isWide
? [styles.usagePopover, { width: popoverWidth }]
: styles.plainPopover;
@@ -264,7 +272,7 @@ export function ContextWindowMeter({
side="top"
align="center"
offset={8}
maxWidth={showsUsage ? popoverWidth : undefined}
maxWidth={isWide ? popoverWidth : undefined}
style={popoverStyle}
testID="context-window-meter-tooltip"
>
@@ -273,6 +281,7 @@ export function ContextWindowMeter({
agentId={agentId}
context={context}
sessionCost={formattedSessionCost}
sessionMetrics={sessionMetrics}
showTitle
refreshable={false}
/>
@@ -307,6 +316,7 @@ export function ContextWindowMeter({
agentId={agentId}
context={context}
sessionCost={formattedSessionCost}
sessionMetrics={sessionMetrics}
showTitle
refreshable
/>
@@ -1,7 +1,6 @@
import { useMemo, type ReactNode } from "react";
import { View, useWindowDimensions } from "react-native";
import { StyleSheet, withUnistyles } from "react-native-unistyles";
import { useTranslation } from "react-i18next";
import { BottomSheetScrollView } from "@gorhom/bottom-sheet";
import { useSafeAreaInsets } from "react-native-safe-area-context";
import { SheetHeaderView, SHEET_HORIZONTAL_PADDING_SCALE } from "@/components/adaptive-modal-sheet";
@@ -23,23 +22,24 @@ const ThemedBottomSheetModal = withUnistyles(IsolatedBottomSheetModal, (theme) =
}));
/**
* The context window meter's details on compact screens, in a sheet as tall as its content.
* Gorhom mounts the content only while the sheet is presented, so whatever it fetches stops once
* the sheet closes.
* What a hover card or tooltip shows on wide screens, shown on compact screens in a sheet as tall
* as its content. Gorhom mounts the content only while the sheet is presented, so whatever it
* fetches stops once the sheet closes.
*/
export function ContextWindowSheet({
export function DetailsSheet({
open,
onClose,
title,
children,
}: {
open: boolean;
onClose: () => void;
title: string;
children: ReactNode;
}) {
const { t } = useTranslation();
const { height } = useWindowDimensions();
const insets = useSafeAreaInsets();
const header = useMemo(() => ({ title: t("contextWindow.title") }), [t]);
const header = useMemo(() => ({ title }), [title]);
const scrollContentStyle = useMemo(() => ({ paddingBottom: insets.bottom }), [insets.bottom]);
const { sheetRef, handleSheetChange, handleSheetDismiss } = useIsolatedBottomSheetVisibility({
visible: open,
@@ -50,7 +50,8 @@ export function ContextWindowSheet({
<ThemedBottomSheetModal
ref={sheetRef}
contextBridge={null}
// Content-sized rather than fixed snap points: the usage cards stream in one at a time.
// Content-sized rather than fixed snap points: a few rows of details, or usage cards that
// stream in one at a time.
enableDynamicSizing
maxDynamicContentSize={height * 0.8}
onChange={handleSheetChange}
+14 -2
View File
@@ -3,6 +3,7 @@ import { MessagePreview, MessageSender } from "@/components/user-message";
import { useStableEvent } from "@/hooks/use-stable-event";
import { getAgentAttachmentPillContent } from "@/attachments/attachment-pill-content";
import { AssistantForkMenu, type AssistantForkTarget } from "@/components/assistant-fork-menu";
import { TurnMetricsButton } from "@/components/turn-metrics";
import { AttachmentLightbox, type ImageLightboxSource } from "@/components/attachment-lightbox";
import {
AttachmentFrame,
@@ -36,7 +37,11 @@ import { useToolCallExpanded } from "@/tool-calls/expansion-context";
import { useTargetAgentTitle } from "@/tool-calls/use-target-agent-title";
import type { DaemonClient } from "@getpaseo/client/internal/daemon-client";
import type { AgentMessage } from "@getpaseo/protocol/agent-message";
import type { AgentCapabilityFlags, ToolCallDetail } from "@getpaseo/protocol/agent-types";
import type {
AgentCapabilityFlags,
AgentMetric,
ToolCallDetail,
} from "@getpaseo/protocol/agent-types";
import type { AgentAttachment } from "@getpaseo/protocol/messages";
import MaskedView from "@react-native-masked-view/masked-view";
import {
@@ -560,6 +565,8 @@ interface AssistantTurnFooterProps {
completedAt?: Date;
durationMs?: number | null;
onFork?: (target: AssistantForkTarget) => Promise<void> | void;
/** What the provider reported about the turn. Without any, the footer has no stats button. */
metrics?: AgentMetric[];
}
const assistantTurnFooterStylesheet = StyleSheet.create((theme) => ({
@@ -596,6 +603,8 @@ const assistantTurnFooterStylesheet = StyleSheet.create((theme) => ({
flexShrink: 0,
color: theme.colors.foregroundMuted,
fontSize: STREAM_METADATA_FONT_SIZE,
// React Native Web preserves trailing spaces at line wraps by default.
...(isWeb ? { whiteSpace: "normal" } : {}),
},
overlappingLabel: {
marginLeft: "-100%",
@@ -617,6 +626,7 @@ export const AssistantTurnFooter = memo(function AssistantTurnFooter({
completedAt,
durationMs,
onFork,
metrics,
}: AssistantTurnFooterProps) {
const { t } = useTranslation();
const [hovered, setHovered] = useState(false);
@@ -666,10 +676,11 @@ export const AssistantTurnFooter = memo(function AssistantTurnFooter({
[onFork],
);
const canFork = Boolean(onFork);
const hasMetrics = metrics !== undefined && metrics.length > 0;
return (
<View style={assistantTurnFooterStylesheet.container}>
{getContent || canFork ? (
{getContent || canFork || hasMetrics ? (
<View style={assistantTurnFooterStylesheet.actions}>
{getContent ? (
<TurnCopyButton
@@ -678,6 +689,7 @@ export const AssistantTurnFooter = memo(function AssistantTurnFooter({
/>
) : null}
{canFork ? <AssistantForkMenu onFork={handleFork} /> : null}
{hasMetrics ? <TurnMetricsButton metrics={metrics} /> : null}
</View>
) : null}
{primaryLabel ? (
@@ -0,0 +1,72 @@
import { memo } from "react";
import { Text, View } from "react-native";
import { StyleSheet } from "react-native-unistyles";
import type { AgentMetric } from "@getpaseo/protocol/agent-types";
/**
* Label and value rows a provider reports for a session or a turn. The provider owns the words and
* the tone; the list only lines them up.
*/
export const MetricsList = memo(function MetricsList({
metrics,
testID,
}: {
metrics: AgentMetric[];
testID?: string;
}) {
return (
<View style={styles.list} testID={testID}>
{metrics.map((metric) => (
<View key={metric.label} style={styles.row}>
<Text style={styles.label} numberOfLines={1}>
{metric.label}
</Text>
<Text style={[styles.value, toneStyle(metric.tone)]} numberOfLines={1}>
{metric.value}
</Text>
</View>
))}
</View>
);
});
function toneStyle(tone: AgentMetric["tone"]) {
switch (tone) {
case "ok":
return styles.ok;
case "warning":
return styles.warning;
case "danger":
return styles.danger;
default:
return null;
}
}
const styles = StyleSheet.create((theme) => ({
list: {
gap: theme.spacing[1.5],
},
row: {
flexDirection: "row",
alignItems: "baseline",
gap: theme.spacing[6],
},
label: {
flexShrink: 1,
color: theme.colors.foregroundMuted,
fontSize: theme.fontSize.sm,
lineHeight: theme.fontSize.sm * 1.4,
},
value: {
flex: 1,
textAlign: "right",
color: theme.colors.foreground,
fontSize: theme.fontSize.sm,
lineHeight: theme.fontSize.sm * 1.4,
fontVariant: ["tabular-nums"],
},
ok: { color: theme.colors.statusSuccess },
warning: { color: theme.colors.statusWarning },
danger: { color: theme.colors.statusDanger },
}));
@@ -1,13 +1,15 @@
import { AppearanceText as Text } from "@/appearance/text";
import { ChevronRight } from "lucide-react-native";
import { useMemo, type ReactNode } from "react";
import { Platform, ScrollView, View } from "react-native";
import { Platform, ScrollView, View, type PressableStateCallbackType } from "react-native";
import { KeyboardAwareScrollView } from "react-native-keyboard-controller";
import { useSafeAreaInsets } from "react-native-safe-area-context";
import { StyleSheet } from "react-native-unistyles";
import { StyleSheet, withUnistyles } from "react-native-unistyles";
import { BackHeader } from "@/components/headers/back-header";
import { MenuHeader } from "@/components/headers/menu-header";
import { PressHighlight } from "@/components/ui/press-highlight";
import { useIsCompactFormFactor } from "@/constants/layout";
import { SPACING } from "@/styles/theme";
import { ICON_SIZE, SPACING, type Theme } from "@/styles/theme";
const PageScrollView = Platform.OS === "android" ? KeyboardAwareScrollView : ScrollView;
@@ -58,14 +60,7 @@ export function PageLayout({
testID={testID}
>
<View style={width === "wide" ? styles.contentWide : styles.content}>
{showTitle ? (
<View style={styles.titleRow} testID={titleTestID}>
<Text style={styles.title} testID="page-title">
{title}
</Text>
{actions}
</View>
) : null}
{showTitle ? <PageTitle title={title} actions={actions} testID={titleTestID} /> : null}
{children}
</View>
</PageScrollView>
@@ -73,6 +68,77 @@ export function PageLayout({
);
}
/** The page that holds this one. Its name is the first part of the title and goes back to it. */
export interface PageTitleParent {
label: string;
/** What the press does, such as "Back to providers". */
accessibilityLabel: string;
onPress: () => void;
testID?: string;
}
interface PageTitleProps {
parent?: PageTitleParent;
/** Drawn before the title, such as the icon of the thing the page is about. */
icon?: ReactNode;
title?: string;
/** A control that belongs to the title, drawn right after it. */
adornment?: ReactNode;
/** Controls for the whole page, drawn at the trailing edge. */
actions?: ReactNode;
testID?: string;
titleTestID?: string;
}
/**
* The heading of a desktop page. A page that opens from another page, such as one provider from
* Providers, names that page first: "Providers > Antigravity".
*/
export function PageTitle({
parent,
icon,
title,
adornment,
actions,
testID,
titleTestID = "page-title",
}: PageTitleProps) {
return (
<View style={styles.titleRow} testID={testID}>
<View style={styles.titleParts}>
{parent ? (
<PressHighlight
accessibilityRole="button"
accessibilityLabel={parent.accessibilityLabel}
onPress={parent.onPress}
testID={parent.testID}
>
{({ hovered }: PressableStateCallbackType & { hovered?: boolean }) => (
<Text style={[styles.title, styles.parent, hovered && styles.parentHovered]}>
{parent.label}
</Text>
)}
</PressHighlight>
) : null}
{parent && title !== undefined ? (
<ThemedChevronRight size={ICON_SIZE.lg} uniProps={separatorColor} />
) : null}
{icon}
{title !== undefined ? (
<Text style={[styles.title, styles.titleLeaf]} numberOfLines={1} testID={titleTestID}>
{title}
</Text>
) : null}
{adornment}
</View>
{actions}
</View>
);
}
const ThemedChevronRight = withUnistyles(ChevronRight);
const separatorColor = (theme: Theme) => ({ color: theme.colors.foregroundMuted });
const styles = StyleSheet.create((theme) => ({
container: {
flex: 1,
@@ -102,11 +168,27 @@ const styles = StyleSheet.create((theme) => ({
gap: theme.spacing[4],
marginBottom: theme.spacing[6],
},
title: {
titleParts: {
// Line up with the section titles, which sit inset from their cards.
marginLeft: theme.spacing[1],
flexDirection: "row",
alignItems: "center",
gap: theme.spacing[2],
flexShrink: 1,
minWidth: 0,
},
title: {
fontSize: theme.fontSize["4xl"],
fontWeight: theme.fontWeight.medium,
color: theme.colors.foreground,
},
titleLeaf: {
flexShrink: 1,
},
parent: {
color: theme.colors.foregroundMuted,
},
parentHovered: {
color: theme.colors.foreground,
},
}));
@@ -2,7 +2,9 @@ import { ArrowLeft } from "lucide-react-native";
import type { ReactNode } from "react";
import { Text, View } from "react-native";
import { StyleSheet } from "react-native-unistyles";
import { PageTitle, type PageTitleParent } from "@/components/page-layout";
import { Button } from "@/components/ui/button";
import { useIsCompactFormFactor } from "@/constants/layout";
/**
* "Back to …" above a settings detail page on wide layouts, where the page has no header of its
@@ -32,32 +34,43 @@ export function SettingsBackLink({
}
/**
* The top of a settings detail page: "Back to …" when the page has no header of its own, then the
* icon and name the page is about, with an optional action after the name.
* The top of a settings detail page: the icon and name of the thing the page is about, with an
* optional action after the name. On wide layouts it is the page's title and names the parent page
* first, which goes back to it. On compact layouts the back header already names the parent.
*/
export function SettingsDetailHeader({
back,
parent,
icon,
title,
trailing,
titleTestID,
}: {
back?: { label: string; onPress: () => void; testID?: string };
icon: ReactNode;
title: string;
parent: PageTitleParent;
icon?: ReactNode;
/** Absent while the page has nothing to name, such as a project that no longer exists. */
title?: string;
trailing?: ReactNode;
titleTestID?: string;
}) {
const isCompact = useIsCompactFormFactor();
if (!isCompact)
return (
<PageTitle
parent={parent}
icon={icon}
title={title}
adornment={trailing}
titleTestID={titleTestID}
/>
);
if (title === undefined) return null;
return (
<View style={styles.header}>
{back ? <SettingsBackLink {...back} /> : null}
<View style={styles.titleRow}>
{icon}
<Text style={styles.title} numberOfLines={1} testID={titleTestID}>
{title}
</Text>
{trailing}
</View>
<View style={styles.titleRow}>
{icon}
<Text style={styles.title} numberOfLines={1} testID={titleTestID}>
{title}
</Text>
{trailing}
</View>
);
}
@@ -67,15 +80,13 @@ const styles = StyleSheet.create((theme) => ({
alignSelf: "flex-start",
paddingHorizontal: 0,
},
header: {
gap: theme.spacing[2],
marginBottom: theme.spacing[4],
},
titleRow: {
marginTop: theme.spacing[2],
// Line up with the section titles, which sit inset from their cards.
marginLeft: theme.spacing[1],
flexDirection: "row",
alignItems: "center",
gap: theme.spacing[3],
marginBottom: theme.spacing[4],
},
title: {
color: theme.colors.foreground,
@@ -0,0 +1,22 @@
import type { ReactNode } from "react";
import { View } from "react-native";
import { StyleSheet } from "react-native-unistyles";
/**
* The action at a heading's end. It adds no height, so a heading with a button is as tall as one
* without and both sit the same distance from their cards. The button stays centered on the title
* and its hit area grows past the heading.
*/
export function HeadingTrailing({ children }: { children: ReactNode }) {
if (children === null || children === undefined || children === false) return null;
return <View style={styles.trailing}>{children}</View>;
}
const styles = StyleSheet.create({
trailing: {
height: 0,
justifyContent: "center",
// A long trailing gives way to the title.
flexShrink: 1,
},
});
@@ -1,6 +1,7 @@
import { useMemo, type ReactNode } from "react";
import { Text, View, type StyleProp, type ViewStyle } from "react-native";
import { StyleSheet } from "react-native-unistyles";
import { HeadingTrailing } from "./heading-trailing";
import { SettingsInfoTip } from "./settings-info-tip";
interface SettingsGroupProps {
@@ -40,7 +41,7 @@ export function SettingsGroup({
/>
) : null}
</View>
{trailing}
<HeadingTrailing>{trailing}</HeadingTrailing>
</View>
{children}
</View>
@@ -3,6 +3,7 @@ import { useMemo, type ReactNode } from "react";
import { View, type StyleProp, type ViewStyle } from "react-native";
import { StyleSheet } from "react-native-unistyles";
import { ExternalLink, type ExternalLinkProps } from "@/components/ui/external-link";
import { HeadingTrailing } from "./heading-trailing";
import { SettingsInfoTip } from "./settings-info-tip";
import { settingsStyles } from "@/styles/settings";
@@ -67,7 +68,7 @@ export function SettingsSection({
<ExternalLink {...link} />
</View>
) : null}
{trailing}
<HeadingTrailing>{trailing}</HeadingTrailing>
</View>
<View style={styles.content}>{children}</View>
</View>
@@ -7,6 +7,7 @@ import {
SettingsCollapsibleRow,
SettingsDetailRow,
SettingsRow,
SettingsSwitch,
} from "@/components/settings";
let root: Root | null = null;
@@ -20,6 +21,7 @@ afterEach(() => {
});
function openNothing() {}
function toggleNothing() {}
function mountRows() {
container = document.createElement("div");
@@ -33,6 +35,13 @@ function mountRows() {
Download
</Button>
</SettingsRow>
<SettingsRow label="Text only" testID="text-row" />
<SettingsSwitch
label="Toggles"
value={false}
onValueChange={toggleNothing}
testID="switch-row"
/>
<SettingsDetailRow label="Opens a page" onPress={openNothing} testID="detail-row" />
<SettingsCollapsibleRow label="Group" defaultExpanded testID="group">
<SettingsRow label="Nested with a button" testID="nested-with-button">
@@ -74,11 +83,13 @@ describe("settings row density", () => {
expect(heightOf("group-toggle")).toBeGreaterThan(heightOf("nested-without-button"));
});
it("keeps a row that opens something as tall as a row holding a control", () => {
it("lines up text, switch, and opening rows below a row holding a button", () => {
mountRows();
expect(heightOf("detail-row")).toBe(heightOf("card-row"));
expect(heightOf("group-toggle")).toBe(heightOf("card-row"));
expect(heightOf("switch-row")).toBe(heightOf("text-row"));
expect(heightOf("detail-row")).toBe(heightOf("text-row"));
expect(heightOf("group-toggle")).toBe(heightOf("text-row"));
expect(heightOf("card-row")).toBeGreaterThan(heightOf("text-row"));
});
it("sizes nested text and controls one step below card rows", () => {
@@ -27,11 +27,6 @@ export function MenuSettingsRows({ children }: { children: ReactNode }) {
export interface SettingsRowLayout {
row: StyleProp<ViewStyle>;
/**
* A row as tall as one holding a control of `controlSize`, whether or not it holds one. In a menu
* it keeps the menu's row height.
*/
controlRow: StyleProp<ViewStyle>;
title: StyleProp<TextStyle>;
hint: StyleProp<TextStyle>;
/** The size of the controls in the row's trailing slot. */
@@ -43,7 +38,6 @@ function layoutFor(placement: RowPlacement): SettingsRowLayout {
case "nested":
return {
row: [settingsStyles.row, styles.nestedRow],
controlRow: [settingsStyles.row, styles.nestedRow],
title: [settingsStyles.rowTitle, styles.nestedText],
hint: [settingsStyles.rowHint, styles.nestedText],
controlSize: "sm",
@@ -51,7 +45,6 @@ function layoutFor(placement: RowPlacement): SettingsRowLayout {
case "menu":
return {
row: [settingsStyles.row, styles.menuRow],
controlRow: [settingsStyles.row, styles.menuRow],
title: settingsStyles.rowTitle,
hint: settingsStyles.rowHint,
controlSize: "md",
@@ -59,7 +52,6 @@ function layoutFor(placement: RowPlacement): SettingsRowLayout {
case "card":
return {
row: settingsStyles.row,
controlRow: [settingsStyles.row, styles.cardControlRow],
title: settingsStyles.rowTitle,
hint: settingsStyles.rowHint,
controlSize: "md",
@@ -74,9 +66,6 @@ export function useSettingsRowLayout(): SettingsRowLayout {
}
const styles = StyleSheet.create((theme) => ({
cardControlRow: {
minHeight: mapLayout(CONTROL_HEIGHTS.md, (height) => height + theme.spacing[4] * 2),
},
// Every nested row is as tall as one holding an `sm` control, so rows with and without a
// trailing button line up.
nestedRow: {
@@ -86,7 +86,7 @@ export function SettingsRowHeader({
const handlePointerEnter = useCallback(() => setPointerOver(true), []);
const handlePointerLeave = useCallback(() => setPointerOver(false), []);
const hovered = pointerOver && !disabled;
const rowStyle = useMemo(() => [layout.controlRow, hovered && styles.hovered], [hovered, layout]);
const rowStyle = useMemo(() => [layout.row, hovered && styles.hovered], [hovered, layout]);
const chevronStyle = useMemo(
() => [styles.chevron, expanded && styles.chevronExpanded],
[expanded],
@@ -462,8 +462,12 @@ function SplitNodeView({
groupId ? state.splitSizesByWorkspace[workspaceKey]?.[groupId] : undefined,
);
const groupChildren = node.kind === "group" ? node.group.children : EMPTY_SPLIT_NODES;
// Dragged sizes live per group id, apart from the layout tree, and a group keeps its id when
// panes are added or removed. They only apply while they still have one entry per child;
// otherwise the dividers past the end of the list could not move.
const layoutGroupSizes = node.kind === "group" ? node.group.sizes : EMPTY_SPLIT_SIZES;
const groupSizes =
storedGroupSizes ?? (node.kind === "group" ? node.group.sizes : EMPTY_SPLIT_SIZES);
storedGroupSizes?.length === groupChildren.length ? storedGroupSizes : layoutGroupSizes;
const visibleFlex = useMemo(
() => resolveVisibleGroupFlex(groupChildren, groupSizes, maximizedPaneId),
[groupChildren, groupSizes, maximizedPaneId],
@@ -0,0 +1,148 @@
import { memo, useCallback, useState } from "react";
import { useTranslation } from "react-i18next";
import { Text, View } from "react-native";
import { ChartNoAxesColumn } from "lucide-react-native";
import { StyleSheet, withUnistyles } from "react-native-unistyles";
import type { AgentMetric } from "@getpaseo/protocol/agent-types";
import {
HoverCard,
HoverCardContent,
HoverCardTrigger,
useHoverCardOpen,
} from "@/components/ui/hover-card";
import { PressHighlight } from "@/components/ui/press-highlight";
import { Tooltip, TooltipContent, TooltipTrigger } from "@/components/ui/tooltip";
import { useIsCompactFormFactor } from "@/constants/layout";
import { isNative } from "@/constants/platform";
import { ICON_SIZE, type Theme } from "@/styles/theme";
import { DetailsSheet } from "./details-sheet";
import { MetricsList } from "./metrics-list";
const ThemedChart = withUnistyles(ChartNoAxesColumn);
const foregroundColorMapping = (theme: Theme) => ({ color: theme.colors.foreground });
const foregroundMutedColorMapping = (theme: Theme) => ({ color: theme.colors.foregroundMuted });
/**
* The turn footer's stats button: what the provider reported about this turn, in a hover card on
* web, a tooltip opened by press on native wide screens, and a sheet on compact screens.
*/
export const TurnMetricsButton = memo(function TurnMetricsButton({
metrics,
}: {
metrics: AgentMetric[];
}) {
const isCompact = useIsCompactFormFactor();
if (isCompact) return <TurnMetricsSheetButton metrics={metrics} />;
if (isNative) return <TurnMetricsTooltipButton metrics={metrics} />;
return <TurnMetricsHoverCardButton metrics={metrics} />;
});
function TurnMetricsSheetButton({ metrics }: { metrics: AgentMetric[] }) {
const { t } = useTranslation();
const [open, setOpen] = useState(false);
const openSheet = useCallback(() => setOpen(true), []);
const closeSheet = useCallback(() => setOpen(false), []);
return (
<>
<PressHighlight
style={styles.trigger}
accessibilityRole="button"
accessibilityLabel={t("message.actions.turnStats")}
onPress={openSheet}
testID="turn-metrics-trigger"
>
{({ hovered }) => <TurnMetricsIcon active={open || Boolean(hovered)} />}
</PressHighlight>
<DetailsSheet open={open} onClose={closeSheet} title={t("message.actions.turnStats")}>
<MetricsList metrics={metrics} />
</DetailsSheet>
</>
);
}
function TurnMetricsTooltipButton({ metrics }: { metrics: AgentMetric[] }) {
const { t } = useTranslation();
const [open, setOpen] = useState(false);
return (
<Tooltip open={open} onOpenChange={setOpen} delayDuration={0} enabledOnDesktop enabledOnMobile>
<TooltipTrigger asChild triggerRefProp="ref">
<PressHighlight
style={styles.trigger}
accessibilityRole="button"
accessibilityLabel={t("message.actions.turnStats")}
testID="turn-metrics-trigger"
>
{({ hovered }) => <TurnMetricsIcon active={open || Boolean(hovered)} />}
</PressHighlight>
</TooltipTrigger>
<TooltipContent side="top" align="start" offset={8} style={styles.popover}>
<TurnMetricsDetails metrics={metrics} />
</TooltipContent>
</Tooltip>
);
}
function TurnMetricsHoverCardButton({ metrics }: { metrics: AgentMetric[] }) {
const { t } = useTranslation();
return (
<HoverCard>
<HoverCardTrigger focusable accessibilityLabel={t("message.actions.turnStats")}>
<View style={styles.trigger} testID="turn-metrics-trigger">
<HoverCardTurnMetricsIcon />
</View>
</HoverCardTrigger>
<HoverCardContent
placement="top"
alignment="start"
offset={8}
role="dialog"
accessibilityLabel={t("message.actions.turnStats")}
testID="turn-metrics-details"
style={styles.popover}
>
<TurnMetricsDetails metrics={metrics} />
</HoverCardContent>
</HoverCard>
);
}
function HoverCardTurnMetricsIcon() {
return <TurnMetricsIcon active={useHoverCardOpen()} />;
}
function TurnMetricsIcon({ active }: { active: boolean }) {
return (
<ThemedChart
size={ICON_SIZE.sm}
uniProps={active ? foregroundColorMapping : foregroundMutedColorMapping}
/>
);
}
function TurnMetricsDetails({ metrics }: { metrics: AgentMetric[] }) {
const { t } = useTranslation();
return (
<>
<Text style={styles.title}>{t("message.actions.turnStats")}</Text>
<MetricsList metrics={metrics} />
</>
);
}
const styles = StyleSheet.create((theme) => ({
// Matches the Copy and Fork buttons beside it, so the glyphs share one row and one spacing.
trigger: {
alignSelf: "center",
padding: theme.spacing[1],
borderRadius: theme.borderRadius.md,
},
popover: {
minWidth: 220,
padding: theme.spacing[3],
gap: theme.spacing[2],
},
title: {
color: theme.colors.foreground,
fontSize: theme.fontSize.base,
},
}));
@@ -302,23 +302,23 @@ describe("control geometry", () => {
trackHeight: 20,
thumbSize: 16,
thumbTravel: 14,
reservedHeight: 32,
hitHeight: 32,
});
expect(switchGeometryForLayout("md", true)).toEqual({
trackWidth: 44,
trackHeight: 26,
thumbSize: 22,
thumbTravel: 18,
reservedHeight: 44,
hitHeight: 44,
});
// Tight chrome keeps the pointer-sized track on every layout and reserves its row's height.
// Tight chrome keeps the pointer-sized track on every layout and takes presses over its row's control height.
expect(switchGeometryForLayout("sm", true)).toMatchObject({
trackWidth: 34,
reservedHeight: 32,
hitHeight: 32,
});
expect(switchGeometryForLayout("sm", false)).toMatchObject({
trackWidth: 34,
reservedHeight: 28,
hitHeight: 28,
});
});
@@ -223,8 +223,11 @@ export interface SwitchGeometry {
thumbSize: number;
/** How far the thumb slides between off and on. */
thumbTravel: number;
/** The control height the switch reserves, so its row lines up with one holding a button. */
reservedHeight: number;
/**
* The height of the area that takes the press: the control height of the switch's size. The
* switch itself is only as tall as its track, so it never makes a row taller.
*/
hitHeight: number;
}
function switchSize(trackWidth: number, trackHeight: number, thumbSize: number) {
@@ -235,6 +238,15 @@ function switchSize(trackWidth: number, trackHeight: number, thumbSize: number)
const SWITCH_POINTER = switchSize(34, 20, 16);
const SWITCH_TOUCH = switchSize(44, 26, 22);
/**
* The height of a default switch. Exported so a row can be as tall as one holding a switch whether
* or not it holds one. A breakpoint value, so it only works inside a Unistyles stylesheet.
*/
export const SWITCH_HEIGHT = bySizing({
touch: SWITCH_TOUCH.trackHeight,
pointer: SWITCH_POINTER.trackHeight,
});
/**
* A switch follows the same two sizes as every other control. Its thumb travel feeds an
* animation, so the size is resolved to plain numbers here instead of a stylesheet breakpoint.
@@ -242,7 +254,7 @@ const SWITCH_TOUCH = switchSize(44, 26, 22);
export function switchGeometryForLayout(size: ControlSize, isCompact: boolean): SwitchGeometry {
const touchSized = controlSizingForLayout(isCompact) === "touch";
const track = size === "md" && touchSized ? SWITCH_TOUCH : SWITCH_POINTER;
return { ...track, reservedHeight: controlHeightForLayout(size, isCompact) };
return { ...track, hitHeight: controlHeightForLayout(size, isCompact) };
}
function fieldLineHeight(fontSize: number): number {
@@ -1,5 +1,6 @@
import { useCallback, type ReactElement } from "react";
import {
MenuCheckboxItem,
MenuHint,
MenuItem,
MenuLabel,
@@ -23,11 +24,16 @@ import {
export { MenuRoot as DropdownMenu };
export { MenuTrigger as DropdownMenuTrigger };
export { MenuItem as DropdownMenuItem };
export { MenuCheckboxItem as DropdownMenuCheckboxItem };
export { MenuLabel as DropdownMenuLabel };
export { MenuSeparator as DropdownMenuSeparator };
export { MenuHint as DropdownMenuHint };
export type { MenuTriggerProps as DropdownMenuTriggerProps };
export type { ActionStatus, MenuItemProps as DropdownMenuItemProps } from "@/components/ui/menu";
export type {
ActionStatus,
MenuCheckboxItemProps as DropdownMenuCheckboxItemProps,
MenuItemProps as DropdownMenuItemProps,
} from "@/components/ui/menu";
export type { MenuPageDefinition } from "@/components/ui/menu";
export { MenuSubTrigger as DropdownMenuSubTrigger } from "@/components/ui/menu";
@@ -267,6 +267,11 @@ function WebHoverCard({
);
}
/** Whether the enclosing card is open, so its trigger can look active while it is. */
export function useHoverCardOpen(): boolean {
return useContext(HoverCardContext)?.open ?? false;
}
export function HoverCardTrigger({
children,
focusable = false,
@@ -19,6 +19,7 @@ export {
} from "./menu-surface";
export { MenuSubTrigger } from "./menu-sub";
export {
MenuCheckboxItem,
MenuHint,
MenuItem,
MenuLabel,
@@ -26,6 +27,7 @@ export {
MenuTextField,
menuRowContentInset,
type ActionStatus,
type MenuCheckboxItemProps,
type MenuItemProps,
} from "./menu-item";
export { MENU_ITEM_HEIGHT } from "./menu-geometry";
@@ -18,6 +18,7 @@ import { Check, CheckCircle } from "lucide-react-native";
import { AdaptiveTextInput } from "@/components/adaptive-modal-sheet";
import { LoadingSpinner } from "@/components/ui/loading-spinner";
import { Tooltip, TooltipContent, TooltipTrigger } from "@/components/ui/tooltip";
import { isWeb } from "@/constants/platform";
import type { Theme } from "@/styles/theme";
import { MenuDepthProvider, useMenuContext } from "./menu-context";
import { MENU_ITEM_HEIGHT } from "./menu-geometry";
@@ -221,6 +222,36 @@ function resolveItemLabel(input: {
return children;
}
function resolveTrailingContent(input: {
trailing: ReactElement | null | undefined;
showSelectedCheck: boolean;
selected: boolean | undefined;
}): ReactElement | null {
if (input.trailing !== undefined) return input.trailing;
if (!input.showSelectedCheck && input.selected) {
return <ThemedCheck size={16} uniProps={mutedMapping} />;
}
return null;
}
function useSheetSpaceKeyProps(
enabled: boolean,
onPress: () => void,
): { onKeyDown?: (event: { key?: string; preventDefault?: () => void }) => void } {
return useMemo(() => {
if (!isWeb || !enabled) return {};
return {
// React Native Web's Pressable activates Space only for role="button", not menuitem/checkbox.
onKeyDown: (event: { key?: string; preventDefault?: () => void }) => {
if (event.key === " " || event.key === "Spacebar") {
event.preventDefault?.();
onPress();
}
},
};
}, [enabled, onPress]);
}
export interface MenuItemProps {
description?: string;
onSelect?: () => void;
@@ -253,7 +284,37 @@ export interface MenuItemProps {
tooltip?: string;
}
export function MenuItem({
export function MenuItem(props: PropsWithChildren<MenuItemProps>): ReactElement {
return <MenuRow {...props} accessibilityRole="menuitem" />;
}
export interface MenuCheckboxItemProps extends Omit<MenuItemProps, "selected" | "onSelect"> {
checked: boolean;
onToggle: () => void;
}
/**
* A toggleable menu row that stays open on press by default and exposes `checkbox` semantics so
* screen readers announce its checked state.
*/
export function MenuCheckboxItem({
checked,
onToggle,
closeOnSelect = false,
...rest
}: PropsWithChildren<MenuCheckboxItemProps>): ReactElement {
return (
<MenuRow
{...rest}
selected={checked}
onSelect={onToggle}
closeOnSelect={closeOnSelect}
accessibilityRole="checkbox"
/>
);
}
function MenuRow({
children,
description,
onSelect,
@@ -272,8 +333,11 @@ export function MenuItem({
closeOnSelect = true,
testID,
tooltip,
}: PropsWithChildren<MenuItemProps>): ReactElement {
const { selectItem } = useMenuContext("MenuItem");
accessibilityRole,
}: PropsWithChildren<
MenuItemProps & { accessibilityRole: "menuitem" | "checkbox" }
>): ReactElement {
const { selectItem, presentation } = useMenuContext("MenuItem");
const isPending = status === "pending" || loading;
const isSuccess = status === "success";
const isDisabled = disabled || isPending || isSuccess;
@@ -286,14 +350,14 @@ export function MenuItem({
const label = resolveItemLabel({ children, isPending, isSuccess, pendingLabel, successLabel });
const trailingContent =
trailing ??
(!showSelectedCheck && selected ? <ThemedCheck size={16} uniProps={mutedMapping} /> : null);
const trailingContent = resolveTrailingContent({ trailing, showSelectedCheck, selected });
const handleItemPress = useCallback(() => {
if (isDisabled) return;
selectItem(onSelect, closeOnSelect);
}, [isDisabled, selectItem, onSelect, closeOnSelect]);
const isSheet = presentation === "sheet";
const sheetKeyProps = useSheetSpaceKeyProps(isSheet, handleItemPress);
// A row that draws a check has to say so as well: a multi-select page is a list of things that
// are on or off, and the check is the only thing telling a sighted user which. Rows that answer
@@ -336,11 +400,13 @@ export function MenuItem({
const content = (
<Pressable
{...sheetKeyProps}
testID={testID}
accessibilityRole="menuitem"
accessibilityRole={accessibilityRole}
accessibilityState={accessibilityState}
aria-checked={selected}
tabIndex={-1}
// Popovers manage focus via arrow keys; compact sheets rely on the browser's Tab order.
tabIndex={isSheet ? 0 : -1}
dataSet={itemDataSet}
disabled={isDisabled}
onPress={handleItemPress}
@@ -25,6 +25,8 @@ export interface SearchFieldProps {
/** Falls back to `placeholder`, which already names the field. */
accessibilityLabel?: string;
clearAccessibilityLabel: string;
/** Fill the container, for a search that heads one list. Defaults to the filter rail's width. */
fullWidth?: boolean;
testID?: string;
clearTestID?: string;
}
@@ -42,6 +44,7 @@ export function SearchField({
placeholder,
accessibilityLabel,
clearAccessibilityLabel,
fullWidth = false,
testID,
clearTestID,
}: SearchFieldProps): ReactElement {
@@ -55,7 +58,9 @@ export function SearchField({
}, [onChangeText]);
return (
<View style={[styles.field, isFocused && styles.fieldFocused]}>
<View
style={[styles.field, fullWidth && styles.fieldFullWidth, isFocused && styles.fieldFocused]}
>
<ThemedSearch size={14} uniProps={mutedColorMapping} />
<ThemedTextInput
testID={testID}
@@ -103,6 +108,9 @@ const styles = StyleSheet.create((theme) => ({
borderWidth: theme.borderWidth[1],
borderColor: theme.colors.border,
},
fieldFullWidth: {
maxWidth: "100%",
},
fieldFocused: {
borderColor: theme.colors.borderAccent,
backgroundColor: theme.colors.surface2,
+7 -2
View File
@@ -123,11 +123,16 @@ export function Switch({
const pressableStyle = useMemo(
() => [
styles.switchControl,
{ minHeight: geometry.reservedHeight },
// The press area is a full control tall and grows out of the track; the layout only sees
// the track.
{
minHeight: geometry.hitHeight,
marginVertical: (geometry.trackHeight - geometry.hitHeight) / 2,
},
disabled ? styles.disabled : null,
style,
],
[disabled, geometry.reservedHeight, style],
[disabled, geometry.hitHeight, geometry.trackHeight, style],
);
return (
@@ -12,6 +12,7 @@ import { Composer } from "@/composer";
import { ProjectIconView } from "@/components/project-icon-view";
import { ICON_SIZE } from "@/styles/theme";
import { useToast } from "@/contexts/toast-context";
import { draftCreationKey } from "@/composer/draft/creation-key";
import { useAgentInputDraft } from "@/composer/draft/input-draft";
import { useProjectIcon } from "@/projects/icons";
import { useHostRuntimeClient, useHostRuntimeIsConnected } from "@/runtime/host-runtime";
@@ -96,13 +97,16 @@ async function callWorkspaceCreation({
worktreeSlug: string;
input: { cwd: string; agent?: CreateWorkspaceRequestOptions["agent"] };
}) {
return connectedClient.createWorkspace({
idempotencyKey: creationId,
const request: Omit<CreateWorkspaceRequestOptions, "idempotencyKey"> = {
agent: input.agent,
source:
creationMethod === "create_worktree"
? { kind: "worktree", cwd: input.cwd, worktreeSlug }
: { kind: "directory", path: input.cwd },
};
return connectedClient.createWorkspace({
...request,
idempotencyKey: draftCreationKey(creationId, request),
});
}
@@ -363,7 +367,9 @@ export function WorkspaceSetupDialog() {
} catch (error) {
const message = toErrorMessage(error);
setErrorMessage(message);
toast.error(message);
if (!getIsStillActive()) {
toast.error(message);
}
} finally {
if (getIsStillActive()) {
setPendingAction(null);
@@ -455,11 +461,10 @@ export function WorkspaceSetupDialog() {
commandDraft={composerState?.commandDraft}
agentControls={agentControlsWithDisabled}
inputWrapperStyle={styles.composerInputWrapper}
error={errorMessage}
/>
</KeyboardTranslateView>
</FileDropZone>
{errorMessage ? <Text style={styles.errorText}>{errorMessage}</Text> : null}
</AdaptiveModalSheet>
);
}
@@ -482,11 +487,6 @@ const styles = StyleSheet.create((theme) => ({
marginHorizontal: -theme.spacing[6],
marginVertical: -theme.spacing[2],
},
errorText: {
fontSize: theme.fontSize.base,
color: theme.colors.destructive,
lineHeight: 20,
},
composerSlot: {
paddingBottom: theme.spacing[4],
},
@@ -8,6 +8,21 @@ const request = {
clientMessageId: "draft-1:initial-message",
};
const workspaceRequest = {
source: {
kind: "worktree" as const,
cwd: "/repo",
projectId: "prj_1",
worktreeSlug: "clever-fox",
},
firstAgentContext: { prompt: "Start the original task.", attachments: [] },
agent: {
config: { provider: "codex", cwd: "/repo", model: "gpt-6" },
initialPrompt: "Start the original task.",
clientMessageId: "draft-1:initial-message",
},
};
describe("draftCreationKey", () => {
it("keeps the key for a retry of the same request, whatever its key order", () => {
const reordered = {
@@ -27,4 +42,52 @@ describe("draftCreationKey", () => {
expect(key).not.toBe(draftCreationKey("draft-1", request));
expect(key.startsWith("draft-1:")).toBe(true);
});
it.each([
{
name: "isolation",
edited: {
...workspaceRequest,
source: { kind: "directory" as const, path: "/repo", projectId: "prj_1" },
},
},
{
name: "project",
edited: {
...workspaceRequest,
source: { ...workspaceRequest.source, projectId: "prj_2", cwd: "/other" },
},
},
{
name: "prompt",
edited: {
...workspaceRequest,
firstAgentContext: { prompt: "hi", attachments: [] },
agent: { ...workspaceRequest.agent, initialPrompt: "hi" },
},
},
])(
"gives a New workspace request with edited $name a new key on the same draft",
({ edited }) => {
const key = draftCreationKey("draft-1", edited);
expect(key).not.toBe(draftCreationKey("draft-1", workspaceRequest));
expect(key.startsWith("draft-1:")).toBe(true);
},
);
it("gives a New workspace request with a changed launch target a new key even when the wire request matches", () => {
const emptyRequest = {
source: { kind: "directory" as const, path: "/repo", projectId: "prj_1" },
};
const chatKey = draftCreationKey("draft-1", {
request: emptyRequest,
launchTarget: { kind: "chat" },
});
const terminalKey = draftCreationKey("draft-1", {
request: emptyRequest,
launchTarget: { kind: "terminal", profileId: "shell" },
});
expect(terminalKey).not.toBe(chatKey);
expect(terminalKey.startsWith("draft-1:")).toBe(true);
});
});
@@ -1,7 +1,7 @@
import type { ComposerOptionControls } from "@/agent-controls/option-controls";
import { useCallback, useEffect, useMemo, useRef } from "react";
import type { JsonValue } from "@getpaseo/protocol/agent-types";
import { Keyboard, StyleSheet as RNStyleSheet, Text, View } from "react-native";
import { Keyboard, StyleSheet as RNStyleSheet, View } from "react-native";
import { useTranslation } from "react-i18next";
import { StyleSheet } from "react-native-unistyles";
import { CHAT_COLUMN_RAIL_DATASET } from "@/chat-rail";
@@ -536,16 +536,6 @@ export function WorkspaceDraftComposer({
);
return (
<View style={animatedStaticStyles.inputAreaWrapper} onLayout={onInputAreaLayout}>
{formErrorMessage ? (
<View
style={styles.formErrorRow}
dataSet={isComposerFullscreen ? undefined : CHAT_COLUMN_RAIL_DATASET}
>
<View style={styles.formError}>
<Text style={styles.errorText}>{formErrorMessage}</Text>
</View>
</View>
) : null}
{importPillPress && !isComposerFullscreen ? (
<View style={styles.importPillRow} dataSet={CHAT_COLUMN_RAIL_DATASET}>
<View style={styles.importPillContent}>
@@ -577,6 +567,7 @@ export function WorkspaceDraftComposer({
commandDraft={composerState.commandDraft}
agentControls={composerAgentControls}
isCompactLayout={isCompactComposerLayout}
error={formErrorMessage || null}
/>
</View>
);
@@ -592,11 +583,6 @@ const animatedStaticStyles = RNStyleSheet.create({
});
const styles = StyleSheet.create((theme) => ({
formErrorRow: {
width: "100%",
paddingHorizontal: theme.spacing[4],
alignItems: "center",
},
importPillRow: {
width: "100%",
paddingHorizontal: theme.spacing[4],
@@ -615,18 +601,4 @@ const styles = StyleSheet.create((theme) => ({
maxWidth: theme.contentMaxWidth,
flexDirection: "row",
},
formError: {
width: "100%",
maxWidth: theme.contentMaxWidth,
marginBottom: theme.spacing[2],
paddingHorizontal: theme.spacing[3],
paddingVertical: theme.spacing[2],
borderRadius: theme.borderRadius.md,
backgroundColor: theme.colors.surface2,
borderWidth: 1,
borderColor: theme.colors.destructive,
},
errorText: {
color: theme.colors.destructive,
},
}));
+14 -9
View File
@@ -8,6 +8,7 @@ import { createStore, type StoreApi } from "zustand/vanilla";
import { useStore } from "zustand";
import { getHostRuntimeStore } from "@/runtime/host-runtime";
import { LoadingSpinner } from "@/components/ui/loading-spinner";
import { Alert } from "@/components/ui/alert";
import {
View,
Pressable,
@@ -146,6 +147,7 @@ import { RenderProfile } from "@/utils/render-profiler";
import { AfterPaintPublication } from "@/composer/after-paint-publication";
import { isWeb, isNative } from "@/constants/platform";
import type { ForgeSearchItem } from "@getpaseo/protocol/messages";
import type { AgentMetric } from "@getpaseo/protocol/agent-types";
import type {
AttachmentMetadata,
ComposerAttachment,
@@ -294,6 +296,8 @@ function buildRealtimeVoiceButtonStyle(pressed: boolean, voiceButtonDisabled: bo
);
}
const EMPTY_METRICS: AgentMetric[] = [];
function buildAgentStateSelector(serverId: string, agentId: string) {
return (state: ReturnType<typeof useSessionStore.getState>) => {
const agent = state.sessions[serverId]?.agents?.get(agentId) ?? null;
@@ -302,6 +306,7 @@ function buildAgentStateSelector(serverId: string, agentId: string) {
contextWindowMaxTokens: agent?.lastUsage?.contextWindowMaxTokens ?? null,
contextWindowUsedTokens: agent?.lastUsage?.contextWindowUsedTokens ?? null,
totalCostUsd: agent?.lastUsage?.totalCostUsd ?? null,
sessionMetrics: agent?.sessionMetrics ?? EMPTY_METRICS,
model: agent?.model ?? null,
};
};
@@ -1016,6 +1021,8 @@ interface ComposerProps {
* `inputMode`.
*/
editingSession?: ComposerEditingSession;
/** Caller-owned submission error rendered in the composer's shared error slot. */
error?: string | null;
}
const EMPTY_ARRAY: readonly QueuedMessage[] = [];
@@ -1319,6 +1326,7 @@ function ComposerContentImpl({
isCompactLayout: isCompactLayoutOverride,
inputMode = "chat",
editingSession: providedEditingSession,
error: externalError,
readOnly = false,
submitLabel,
placeholder,
@@ -2221,6 +2229,7 @@ function ComposerContentImpl({
maxTokens={agentState.contextWindowMaxTokens}
usedTokens={agentState.contextWindowUsedTokens}
totalCostUsd={agentState.totalCostUsd}
sessionMetrics={agentState.sessionMetrics}
toolbarRing={TOOLBAR_RINGS[controlSizing]}
/>
</View>
@@ -2232,6 +2241,7 @@ function ComposerContentImpl({
agentState.contextWindowMaxTokens,
agentState.contextWindowUsedTokens,
agentState.totalCostUsd,
agentState.sessionMetrics,
controlSizing,
],
);
@@ -2544,14 +2554,13 @@ function ComposerContentImpl({
// Focusing the composer on a touch screen raises the on-screen keyboard over the conversation.
const messageInputAutoFocus = autoFocus && isDesktopWebBreakpoint && hasFinePointer;
const submitLoadingPressHandler = isAgentRunning ? handleCancelAgent : undefined;
const displayError = externalError || sendError;
const sendErrorNode = useMemo(
() =>
sendError ? (
<Text accessibilityRole="alert" style={styles.sendErrorText}>
{sendError}
</Text>
displayError ? (
<Alert variant="error" size="sm" description={displayError} testID="composer-error" />
) : null,
[sendError],
[displayError],
);
const githubEmptyText = githubSearchResultsQuery.isFetching
? t("composer.github.searching")
@@ -2837,10 +2846,6 @@ const styles = StyleSheet.create((theme: Theme) => ({
queueSendButton: {
backgroundColor: theme.colors.accent,
},
sendErrorText: {
color: theme.colors.palette.red[500],
fontSize: theme.fontSize.base,
},
})) as unknown as Record<string, object>;
const ThemedAttachmentSpinner = withUnistyles(LoadingSpinner);
@@ -117,10 +117,11 @@ export function useComposerHeight({
useEffect(() => {
const source = textareaRef.current;
if (!source || typeof ResizeObserver === "undefined") return;
let previousWidth = source.clientWidth;
const observer = new ResizeObserver(() => {
const nextWidth = source.clientWidth;
if (Math.abs(nextWidth - previousWidth) < 1) return;
let previousWidth: number | undefined;
const observer = new ResizeObserver(([entry]) => {
// Hint padding changes the width text wraps at even when clientWidth stays fixed.
const nextWidth = entry.contentRect.width;
if (nextWidth === previousWidth) return;
previousWidth = nextWidth;
measure(paramsRef.current.getText());
});
+39 -24
View File
@@ -677,24 +677,6 @@ function MessageInputOverlay({
return null;
}
function FocusHint({
visible,
focusInputKeys,
label,
}: {
visible: boolean;
focusInputKeys: ShortcutChord | null | undefined;
label: string;
}) {
const { isActiveComposer } = useComposerKeyboardScope();
if (!isActiveComposer || !visible || !focusInputKeys || !label.trim()) return null;
return (
<Text style={styles.focusHintText} pointerEvents="none">
{label}
</Text>
);
}
interface ComposerTextSurfaceProps {
readOnly: boolean;
value: string;
@@ -745,6 +727,32 @@ function ComposerTextSurface(props: ComposerTextSurfaceProps): React.ReactElemen
onRenderedLinesChange(event.nativeEvent.lines.length),
[onRenderedLinesChange],
);
const { isActiveComposer } = useComposerKeyboardScope();
const { focusHintLabel, focusHintVisible, focusInputKeys } = props;
const [focusHintLayout, setFocusHintLayout] = useState<{ label: string; width: number }>();
const handleFocusHintLayout = useCallback(
(event: LayoutChangeEvent) => {
setFocusHintLayout({ label: focusHintLabel, width: event.nativeEvent.layout.width });
},
[focusHintLabel],
);
const shouldShowFocusHint =
isActiveComposer &&
focusHintVisible &&
Boolean(focusInputKeys) &&
focusHintLabel.trim().length > 0;
const focusHintWidth =
shouldShowFocusHint && focusHintLayout?.label === focusHintLabel
? focusHintLayout.width
: undefined;
// Make room for the hint in the same render that reveals it.
const textInputStyle = useMemo(
() =>
focusHintWidth === undefined
? props.textInputStyle
: [props.textInputStyle, { paddingRight: focusHintWidth }],
[focusHintWidth, props.textInputStyle],
);
if (props.readOnly) {
return (
<View style={styles.textInputScrollWrapper}>
@@ -775,7 +783,7 @@ function ComposerTextSurface(props: ComposerTextSurfaceProps): React.ReactElemen
accessibilityLabel={props.accessibilityLabel}
onFocus={props.onFocus}
onBlur={props.onBlur}
style={props.textInputStyle}
style={textInputStyle}
multiline
scrollEnabled={props.scrollEnabled}
editable={props.editable}
@@ -785,11 +793,16 @@ function ComposerTextSurface(props: ComposerTextSurfaceProps): React.ReactElemen
onPasteError={props.onPasteError}
autoFocus={props.autoFocus}
/>
<FocusHint
visible={props.focusHintVisible}
focusInputKeys={props.focusInputKeys}
label={props.focusHintLabel}
/>
{shouldShowFocusHint ? (
<Text
key={focusHintLabel}
style={[styles.focusHintText, focusHintWidth === undefined && styles.unmeasuredFocusHint]}
pointerEvents="none"
onLayout={handleFocusHintLayout}
>
{focusHintLabel}
</Text>
) : null}
</View>
);
}
@@ -2460,10 +2473,12 @@ const styles = StyleSheet.create((theme: Theme) => ({
position: "absolute",
top: 0,
right: 0,
paddingLeft: theme.spacing[2],
fontSize: theme.fontSize.sm,
color: theme.colors.foregroundMuted,
opacity: 0.5,
},
unmeasuredFocusHint: { opacity: 0 },
textInput: {
// Preserve the controls when an ancestor constrains an overlong draft.
flexShrink: 1,
@@ -637,7 +637,7 @@ it("stores and expands identical content once across directory associations", as
});
it("coalesces missing bodies and keeps a newer pushed association when an older pull completes", async () => {
const { QueryClient } = await import("@tanstack/react-query");
const { QueryClient, QueryObserver } = await import("@tanstack/react-query");
const { applyProvidersSnapshotUpdate } = await import("./push-router");
const { fetchProvidersSnapshot, providersSnapshotQueryKey } =
await import("./providers-snapshot");
@@ -653,6 +653,16 @@ it("coalesces missing bodies and keeps a newer pushed association when an older
fetchedAt: { pi: fetchedAt },
requestId: hash,
});
// Announcements only download for directories a screen is showing.
const unsubscribers = ["/a", "/b", "/c"].map((cwd) => {
const queryKey = providersSnapshotQueryKey(serverId, cwd);
queryClient.setQueryData(queryKey, snapshot("seed"));
return new QueryObserver(queryClient, {
queryKey,
queryFn: async () => snapshot("seed"),
staleTime: Infinity,
}).subscribe(() => {});
});
let transfers = 0;
const client = {
async getProvidersSnapshot() {
@@ -776,6 +786,7 @@ it("coalesces missing bodies and keeps a newer pushed association when an older
entries: [{ models: [{ id: "shared" }] }],
});
} finally {
for (const unsubscribe of unsubscribers) unsubscribe();
queryClient.clear();
}
});
@@ -1,5 +1,5 @@
import { createRequire } from "node:module";
import { QueryClient, CancelledError } from "@tanstack/react-query";
import { QueryClient, CancelledError, QueryObserver } from "@tanstack/react-query";
import type { ProviderSnapshotEntry } from "@getpaseo/protocol/agent-types";
import type { GetProvidersSnapshotResponseMessage } from "@getpaseo/protocol/messages";
import { compactProviderSnapshot } from "@getpaseo/protocol/provider-snapshot-codec";
@@ -307,6 +307,14 @@ describe("provider snapshot publication identity", () => {
snapshotHash: "next-hash",
compactSnapshot: compactProviderSnapshot([{ ...entries[0]!, label: "Updated provider" }]),
};
// Announcements only publish into a directory a screen is showing.
const shownKey = providersSnapshotQueryKey(serverId, snapshot.cwd);
queryClient.setQueryData(shownKey, snapshot);
const unsubscribe = new QueryObserver(queryClient, {
queryKey: shownKey,
queryFn: async () => snapshot,
staleTime: Infinity,
}).subscribe(() => {});
try {
for (const body of [snapshot, changed, changed]) {
await applyProvidersSnapshotUpdate({
@@ -325,6 +333,7 @@ describe("provider snapshot publication identity", () => {
expect(published.entries[0]!.models).toBe(canonical.entries[0]!.models);
}
} finally {
unsubscribe();
queryClient.clear();
}
});
+168 -1
View File
@@ -8,8 +8,12 @@ import { checkoutDiffQueryKey } from "@/git/query-keys";
import { buildTerminalsQueryKey } from "@/screens/workspace/terminals/state";
import { daemonConfigQueryKey } from "@/data/daemon-config";
import { daemonPairingOfferQueryKey } from "@/data/daemon-pairing";
import { providersSnapshotQueryKey } from "@/data/providers-snapshot";
import { providersSnapshotQuery, providersSnapshotQueryKey } from "@/data/providers-snapshot";
import { replicaQueryOptions } from "@/data/query";
import { createProviderSnapshotCache } from "@/data/provider-snapshot-cache";
import { draftAgentCommandsQueryKey } from "@/agent-commands/cache";
import {
applyProvidersSnapshotUpdate,
checkoutDiffPushRoute,
invalidateServerDataQueriesAfterReconnect,
mountServerDataPushRouter,
@@ -594,6 +598,13 @@ describe("server data push router", () => {
const serverId = "server-1";
const pairingOfferKey = daemonPairingOfferQueryKey(serverId);
queryClient.setQueryData(pairingOfferKey, { relayEnabled: true, url: "https://pairing" });
const snapshotKey = providersSnapshotQueryKey(serverId);
queryClient.setQueryData(snapshotKey, { entries: [], generatedAt: "seed", requestId: "seed" });
const unsubscribeSnapshot = new QueryObserver(queryClient, {
queryKey: snapshotKey,
queryFn: async () => ({ entries: [], generatedAt: "seed", requestId: "seed" }),
staleTime: Infinity,
}).subscribe(() => undefined);
const unmount = mountServerDataPushRouter({ client: fake.client, queryClient, serverId });
fake.emit(providerUpdate("2026-01-01T00:00:00.000Z"));
@@ -622,6 +633,7 @@ describe("server data push router", () => {
generatedAt: "2026-01-01T00:00:00.000Z",
requestId: "providers_snapshot_update",
});
unsubscribeSnapshot();
});
it.each(["base", "all"] as const)("routes %s diff snapshots and updates", (mode) => {
@@ -982,3 +994,158 @@ describe("server data push router", () => {
expect(queryClient.getQueryState(otherProviderKey)?.isInvalidated).toBe(false);
});
});
describe("applyProvidersSnapshotUpdate by observation", () => {
const serverId = "observed-host";
const cwd = "/repo";
function createStorage() {
const values = new Map<string, string>();
return {
values,
async getItem(key: string) {
return values.get(key) ?? null;
},
async setItem(key: string, value: string) {
values.set(key, value);
},
async removeItem(key: string) {
values.delete(key);
},
async getAllKeys() {
return [...values.keys()];
},
async multiGet(keys: readonly string[]) {
return keys.map((key) => [key, values.get(key) ?? null] as const);
},
async multiRemove(keys: readonly string[]) {
for (const key of keys) values.delete(key);
},
};
}
function snapshotFor(label: string) {
return {
requestId: label,
cwd,
entries: [
{
provider: "codex",
status: "ready",
enabled: true,
label,
models: [{ provider: "codex", id: label, label, thinkingOptions: [] }],
},
],
generatedAt: "2026-01-01T00:00:00.000Z",
} as never;
}
function setup() {
const storage = createStorage();
const cache = createProviderSnapshotCache(storage);
const queryClient = new QueryClient({ defaultOptions: { queries: { retry: false } } });
const calls: unknown[] = [];
let current = "v1";
const client = {
async getProvidersSnapshot(options?: unknown) {
calls.push(options);
return snapshotFor(current);
},
};
const observe = () =>
new QueryObserver(
queryClient,
replicaQueryOptions(
providersSnapshotQuery({ serverId, cwd, client: () => client, queryClient, cache }),
),
);
const announce = () =>
applyProvidersSnapshotUpdate({
serverId,
queryClient,
cache,
client,
message: {
type: "providers_snapshot_update",
payload: {
cwd,
entries: [],
snapshotHash: "hash-v2",
generatedAt: "2026-01-01T00:00:01Z",
},
} as never,
});
return {
queryClient,
calls,
storage,
observe,
announce,
setCurrent: (value: string) => {
current = value;
},
};
}
it("does nothing for a directory this client never opened", async () => {
const { queryClient, calls, storage, announce } = setup();
await announce();
expect(calls).toEqual([]);
expect(
queryClient.getQueryCache().find({ queryKey: providersSnapshotQueryKey(serverId, cwd) }),
).toBeUndefined();
expect(storage.values.size).toBe(0);
queryClient.clear();
});
it("marks an unobserved cached directory stale without downloading and refetches it on mount", async () => {
const { queryClient, calls, observe, announce, setCurrent } = setup();
const queryKey = providersSnapshotQueryKey(serverId, cwd);
const commandsKey = draftAgentCommandsQueryKey({
serverId,
draftConfig: { provider: "codex", cwd },
});
queryClient.setQueryData(queryKey, snapshotFor("v1"));
queryClient.setQueryData(commandsKey, [{ name: "review" }]);
setCurrent("v2");
await announce();
expect(calls).toEqual([]);
expect(queryClient.getQueryState(queryKey)?.isInvalidated).toBe(true);
expect(queryClient.getQueryState(commandsKey)?.isInvalidated).toBe(true);
const observer = observe();
const unsubscribe = observer.subscribe(() => {});
await queryClient.getQueryCache().find({ queryKey })?.promise;
expect(calls).toHaveLength(1);
expect(
(queryClient.getQueryData(queryKey) as { entries: Array<{ label: string }> }).entries[0]!
.label,
).toBe("v2");
unsubscribe();
queryClient.clear();
});
it("keeps the reference-aware fetch for an observed directory", async () => {
const { queryClient, calls, observe, announce } = setup();
const queryKey = providersSnapshotQueryKey(serverId, cwd);
const observer = observe();
const unsubscribe = observer.subscribe(() => {});
await queryClient.getQueryCache().find({ queryKey })?.promise;
expect(calls).toHaveLength(1);
await announce();
// The announced hash is not resident, so exactly one content transfer is fetched and
// published to the observed query.
expect(calls).toHaveLength(2);
expect(queryClient.getQueryState(queryKey)?.isInvalidated).toBe(false);
unsubscribe();
queryClient.clear();
});
});
+18 -1
View File
@@ -8,7 +8,10 @@ import type {
MutableDaemonConfig,
SessionOutboundMessage,
} from "@getpaseo/protocol/messages";
import { invalidateCommandsForProviderChange } from "@/agent-commands/cache";
import {
invalidateCommandsForDirectory,
invalidateCommandsForProviderChange,
} from "@/agent-commands/cache";
import { shareCheckoutDiff } from "@/git/diff-sharing";
import { orderCheckoutDiffFiles } from "@/git/diff-order";
import { daemonConfigQueryKey } from "@/data/daemon-config";
@@ -184,6 +187,20 @@ export async function applyProvidersSnapshotUpdate(input: {
const snapshot = { ...input.message.payload, requestId: "providers_snapshot_update" };
const cwd = normalizeProvidersSnapshotCwd(snapshot.cwd);
const queryKey = providersSnapshotQueryKey(input.serverId, cwd);
const cached = input.queryClient.getQueryCache().find({ queryKey, exact: true });
// Only a screen showing this directory is worth a download. A cached but unobserved
// directory refetches on its next mount; one this client never opened needs nothing.
if (!cached) return;
if (!cached.isActive()) {
await input.queryClient.cancelQueries({ queryKey, exact: true });
await input.queryClient.invalidateQueries({ queryKey, exact: true, refetchType: "none" });
await invalidateCommandsForDirectory({
queryClient: input.queryClient,
serverId: input.serverId,
cwd,
});
return;
}
const previous = input.queryClient.getQueryData<{ entries: ProviderSnapshotEntry[] }>(queryKey);
let announcement: typeof snapshot | undefined = snapshot;
void input.queryClient.cancelQueries({ queryKey, exact: true });
+50
View File
@@ -0,0 +1,50 @@
import { QueryClient, QueryObserver } from "@tanstack/react-query";
import { describe, expect, it } from "vitest";
import { replicaQueryOptions } from "./query";
function setup() {
const queryClient = new QueryClient();
const queryKey = ["replica", "value"] as const;
let fetches = 0;
const options = replicaQueryOptions({
queryKey,
pushEvent: "value_update",
queryFn: async () => {
fetches++;
return `fetch-${fetches}`;
},
});
return { queryClient, queryKey, options, fetchCount: () => fetches };
}
describe("replicaQueryOptions mount behavior", () => {
it("does not refetch on mount while the cached value is not invalidated", async () => {
const { queryClient, queryKey, options, fetchCount } = setup();
queryClient.setQueryData(queryKey, "cached");
const observer = new QueryObserver(queryClient, options);
const unsubscribe = observer.subscribe(() => {});
await Promise.resolve();
expect(fetchCount()).toBe(0);
expect(observer.getCurrentResult().data).toBe("cached");
unsubscribe();
queryClient.clear();
});
it("refetches on mount when the cached value was invalidated while unobserved", async () => {
const { queryClient, queryKey, options, fetchCount } = setup();
queryClient.setQueryData(queryKey, "cached");
await queryClient.invalidateQueries({ queryKey, refetchType: "none" });
expect(fetchCount()).toBe(0);
const observer = new QueryObserver(queryClient, options);
const unsubscribe = observer.subscribe(() => {});
await queryClient.getQueryCache().find({ queryKey })?.promise;
expect(fetchCount()).toBe(1);
expect(observer.getCurrentResult().data).toBe("fetch-1");
unsubscribe();
queryClient.clear();
});
});
+3 -1
View File
@@ -86,7 +86,9 @@ export function replicaQueryOptions<
},
},
queryFn: options.queryFn ?? skipToken,
refetchOnMount: false,
// Pushes keep a replica fresh; a mount only refetches what an invalidation marked stale
// while nothing observed it.
refetchOnMount: (query) => query.state.isInvalidated,
refetchOnReconnect: false,
refetchOnWindowFocus: false,
staleTime: Infinity,
@@ -0,0 +1,91 @@
import { useCallback } from "react";
import {
useActiveWorkspaceSelection,
type ActiveWorkspaceSelection,
} from "@/stores/navigation-active-workspace-store";
import { getFocusedWorkspaceSelection } from "@/navigation/workspace-route-navigation";
import { useSessionStore } from "@/stores/session-store";
import { useWorkspaceFields } from "@/stores/session-store-hooks";
import { resolveWorkspaceMapKeyByIdentity } from "@/utils/workspace-identity";
import { redirectIfArchivingActiveWorkspace } from "@/utils/sidebar-workspace-archive-redirect";
import { useWorkspaceArchive } from "@/workspace/use-workspace-archive";
import { isWorkspaceListed, type WorkspaceArchiveTarget } from "@/workspace/workspace-archive";
// The redirect compares descriptor ids; the route's id is not guaranteed to equal one (see
// use-global-workspace-pin-action.ts), so resolve it through the session store.
function readActiveDescriptorSelection(): ActiveWorkspaceSelection | null {
const selection = getFocusedWorkspaceSelection();
if (!selection) {
return null;
}
const workspaces = useSessionStore.getState().sessions[selection.serverId]?.workspaces;
const key = resolveWorkspaceMapKeyByIdentity({
workspaces,
workspaceId: selection.workspaceId,
});
const descriptor = key ? workspaces?.get(key) : undefined;
return descriptor ? { serverId: selection.serverId, workspaceId: descriptor.id } : null;
}
export interface ActiveWorkspaceArchive {
/** False when no workspace is routed, or the routed one is already archiving. */
canArchive: boolean;
/** Starts the archive for the routed workspace. Returns false when nothing was started. */
archive: () => boolean;
}
/**
* Archives the workspace the route points at. Keyed on the route selection rather than a rendered
* row, so it works from the keyboard shortcut and from the compact header menu alike.
*/
export function useActiveWorkspaceArchive(): ActiveWorkspaceArchive {
const selection = useActiveWorkspaceSelection();
const serverId = selection?.serverId ?? null;
const routeWorkspaceId = selection?.workspaceId ?? null;
// `id` is the descriptor id, not the route id; see use-global-workspace-pin-action.ts.
const fields = useWorkspaceFields(serverId, routeWorkspaceId, (workspace) => ({
id: workspace.id,
workspaceKind: workspace.workspaceKind,
workspaceDirectory: workspace.workspaceDirectory,
name: workspace.name,
archivingAt: workspace.archivingAt,
diffStat: workspace.diffStat,
}));
// The archive starts after the host answers and you confirm, by which time the header that began
// it may be gone (a resize swaps it, and on native switching workspaces unmounts it), and you may
// have left workspace routes. So the redirect reads the focused route now, not this render's.
const redirectAfterArchive = useCallback((workspace: WorkspaceArchiveTarget) => {
redirectIfArchivingActiveWorkspace({
...workspace,
activeWorkspaceSelection: readActiveDescriptorSelection(),
});
}, []);
const archiveController = useWorkspaceArchive({
serverId: serverId ?? "",
workspaceId: fields?.id ?? "",
workspaceKind: fields?.workspaceKind ?? "directory",
workspaceDirectory: fields?.workspaceDirectory ?? "",
name: fields?.name ?? "",
diffStat: fields?.diffStat,
onArchiveStarted: redirectAfterArchive,
});
const canArchive = fields !== null && fields.archivingAt === null;
const archive = useCallback(() => {
if (!canArchive || !serverId) {
return false;
}
// Archiving hides its target immediately. Read the store in case this runs again before
// React renders that update.
if (!isWorkspaceListed({ serverId, workspaceId: fields.id })) {
return false;
}
archiveController.archive();
return true;
}, [archiveController, canArchive, fields, serverId]);
return { canArchive, archive };
}
@@ -110,6 +110,10 @@ function resolveAgentAutocompleteSnapshot(input: {
};
}
function isSameSlashCommand(a: SlashCommandRange | null, b: SlashCommandRange | null): boolean {
return a?.start === b?.start && a?.query === b?.query;
}
interface DirectorySuggestionEntry {
path: string;
kind: "file" | "directory";
@@ -746,9 +750,25 @@ export function useAgentAutocomplete(input: UseAgentAutocompleteInput): AgentAut
);
const selectOptionFromKeyPress = useCallback(
(option: AutocompleteOption, event?: AgentAutocompleteKeyPressEvent) =>
onSelectOption(option, event?.input),
[onSelectOption],
(option: AutocompleteOption, event?: AgentAutocompleteKeyPressEvent) => {
// The cursor reaches the menu after paint, so a key can land while the options still
// belong to an older query. Never pick from that list. A click picks what the user sees.
if (
mode === "command" &&
event &&
!isSameSlashCommand(
findActiveSlashCommand({
text: event.input.text,
cursorIndex: event.input.selection.start,
}),
activeSlashCommand,
)
) {
return;
}
onSelectOption(option, event?.input);
},
[activeSlashCommand, mode, onSelectOption],
);
const { selectedIndex, onKeyPress } = useAutocomplete({
@@ -1,11 +1,6 @@
import { useCallback } from "react";
import { useActiveWorkspaceArchive } from "@/hooks/use-active-workspace-archive";
import { useKeyboardActionHandler } from "@/hooks/use-keyboard-action-handler";
import type { KeyboardActionId } from "@/keyboard/keyboard-action-dispatcher";
import { useActiveWorkspaceSelection } from "@/stores/navigation-active-workspace-store";
import { useWorkspaceFields } from "@/stores/session-store-hooks";
import { redirectIfArchivingActiveWorkspace } from "@/utils/sidebar-workspace-archive-redirect";
import { useWorkspaceArchive } from "@/workspace/use-workspace-archive";
import { isWorkspaceListed, type WorkspaceArchiveTarget } from "@/workspace/workspace-archive";
const WORKSPACE_ARCHIVE_ACTIONS: readonly KeyboardActionId[] = ["workspace.archive"];
@@ -14,64 +9,13 @@ const WORKSPACE_ARCHIVE_ACTIONS: readonly KeyboardActionId[] = ["workspace.archi
// the pin shortcut (see use-global-workspace-pin-action.ts), it is one registration keyed on the
// active route selection.
export function useGlobalWorkspaceArchiveAction() {
const selection = useActiveWorkspaceSelection();
const serverId = selection?.serverId ?? null;
const routeWorkspaceId = selection?.workspaceId ?? null;
// `id` is the descriptor id, not the route id; see use-global-workspace-pin-action.ts.
const fields = useWorkspaceFields(serverId, routeWorkspaceId, (workspace) => ({
id: workspace.id,
workspaceKind: workspace.workspaceKind,
workspaceDirectory: workspace.workspaceDirectory,
name: workspace.name,
archivingAt: workspace.archivingAt,
diffStat: workspace.diffStat,
}));
// useWorkspaceArchive calls the latest render's callback, so `activeWorkspaceId` is whatever is
// routed when the archive starts, after any confirmation. It is the descriptor id, which is what
// the redirect compares against.
const activeWorkspaceId = fields?.id ?? null;
const redirectAfterArchive = useCallback(
(workspace: WorkspaceArchiveTarget) => {
redirectIfArchivingActiveWorkspace({
...workspace,
activeWorkspaceSelection:
serverId && activeWorkspaceId ? { serverId, workspaceId: activeWorkspaceId } : null,
});
},
[activeWorkspaceId, serverId],
);
const archiveController = useWorkspaceArchive({
serverId: serverId ?? "",
workspaceId: fields?.id ?? "",
workspaceKind: fields?.workspaceKind ?? "directory",
workspaceDirectory: fields?.workspaceDirectory ?? "",
name: fields?.name ?? "",
diffStat: fields?.diffStat,
onArchiveStarted: redirectAfterArchive,
});
const canArchive = fields !== null && fields.archivingAt === null;
const handle = useCallback(() => {
if (!canArchive || !serverId) {
return false;
}
// Archiving hides its target immediately. Read the store in case the shortcut
// runs again before React renders that update.
if (!isWorkspaceListed({ serverId, workspaceId: fields.id })) {
return false;
}
archiveController.archive();
return true;
}, [archiveController, canArchive, fields, serverId]);
const { canArchive, archive } = useActiveWorkspaceArchive();
useKeyboardActionHandler({
handlerId: "workspace-archive-global",
actions: WORKSPACE_ARCHIVE_ACTIONS,
enabled: canArchive,
priority: 0,
handle,
handle: archive,
});
}
@@ -352,6 +352,19 @@ describe("applyProvidersSnapshotUpdate", () => {
queryClient = new QueryClient();
});
// Announcements only download for directories a screen is showing, so these tests mount one.
function showSnapshot(cwd?: string) {
const queryKey = providersSnapshotQueryKey(serverId, cwd);
if (queryClient.getQueryData(queryKey) === undefined) {
queryClient.setQueryData(queryKey, providersSnapshot([]));
}
return new QueryObserver(queryClient, {
queryKey,
queryFn: async () => providersSnapshot([]),
staleTime: Infinity,
}).subscribe(() => {});
}
function updateMessage(entries: ProviderSnapshotEntry[], cwd?: string): ProvidersSnapshotUpdate {
return {
type: "providers_snapshot_update",
@@ -364,6 +377,7 @@ describe("applyProvidersSnapshotUpdate", () => {
}
it("refreshes active commands only when their provider changes after discovery", async () => {
showSnapshot("/repo-a");
const key = sessionAgentCommandsQueryKey({ serverId, agentId: "agent-1" });
let requests = 0;
const observer = new QueryObserver(queryClient, {
@@ -447,6 +461,7 @@ describe("applyProvidersSnapshotUpdate", () => {
});
it("retries a failed command discovery when its provider first becomes ready", async () => {
showSnapshot("/repo-a");
const key = sessionAgentCommandsQueryKey({ serverId, agentId: "agent-1" });
let ready = false;
let requests = 0;
@@ -493,6 +508,7 @@ describe("applyProvidersSnapshotUpdate", () => {
providersSnapshotQueryKey(serverId, "/repo-a"),
providersSnapshot([codexEntry("ready", [readyCodexModel])]),
);
showSnapshot("/repo-a");
const send = (label: string) =>
applyProvidersSnapshotUpdate({
client: createClient(),
@@ -511,6 +527,7 @@ describe("applyProvidersSnapshotUpdate", () => {
});
it("routes updates to the home query cache when the message carries no cwd", async () => {
showSnapshot();
await applyProvidersSnapshotUpdate({
client: createClient(),
serverId,
@@ -527,6 +544,7 @@ describe("applyProvidersSnapshotUpdate", () => {
it("routes workspace updates to the matching scope without touching siblings", async () => {
queryClient.setQueryData(providersSnapshotQueryKey(serverId, "/repo-b"), providersSnapshot([]));
showSnapshot("/repo-a");
await applyProvidersSnapshotUpdate({
client: createClient(),
@@ -546,6 +564,7 @@ describe("applyProvidersSnapshotUpdate", () => {
});
it("persists compact push updates", async () => {
showSnapshot("/repo-a");
const entries = [codexEntry("ready", [readyCodexModel])];
const compactSnapshot = compactProviderSnapshot(entries);
const cache = createCache();
@@ -579,6 +598,7 @@ describe("applyProvidersSnapshotUpdate", () => {
providersSnapshotQueryKey(serverId, workspaceCwd),
providersSnapshot([codexEntry("loading")]),
);
showSnapshot(workspaceCwd);
await applyProvidersSnapshotUpdate({
client: createClient(),
@@ -606,6 +626,7 @@ describe("applyProvidersSnapshotUpdate", () => {
providersSnapshotQueryKey(serverId, "/repo-a"),
providersSnapshot([codexEntry("ready")]),
);
showSnapshot("/repo-a");
await applyProvidersSnapshotUpdate({
client: createClient(),
@@ -331,7 +331,7 @@ describe("loadAppSettingsFromStorage", () => {
pins: null,
serverId: null,
cardOrder: [],
cardsOff: [],
cardVisibility: {},
});
const deps = makeDeps({
@@ -357,7 +357,7 @@ describe("loadAppSettingsFromStorage", () => {
],
serverId: "server",
cardOrder: [],
cardsOff: [],
cardVisibility: {},
});
});
@@ -374,7 +374,7 @@ describe("loadAppSettingsFromStorage", () => {
pins: [],
serverId: "server",
cardOrder: [],
cardsOff: [],
cardVisibility: {},
});
});
@@ -396,7 +396,7 @@ describe("loadAppSettingsFromStorage", () => {
pins: [{ sourceId: "claude", windowId: "weekly" }],
serverId: null,
cardOrder: [],
cardsOff: [],
cardVisibility: {},
});
});
+9 -5
View File
@@ -349,6 +349,7 @@ export const ar: TranslationResources = {
copyTurn: "نسخ بدوره",
copyMessage: "انسخ الرسالة",
forkMenu: "تفريع الرسالة",
turnStats: "إحصاءات الدور",
forkInNewTab: "تفريع في تبويب جديد",
forkInNewWorkspace: "تفريع في مساحة عمل جديدة",
forkUnavailable: "حدّث المضيف لاستخدام هذا.",
@@ -2264,7 +2265,11 @@ export const ar: TranslationResources = {
title: "استخدام الخطة",
refresh: "تحديث",
reorderCard: "إعادة ترتيب {{source}}",
showCard: "إظهار {{source}}",
sources: "المصادر",
sourcesCount: "{{shown}} من {{total}}",
providerInactive: "المزود غير نشط",
allSourcesHidden: "لا توجد مصادر استخدام معروضة",
chooseSources: "اختيار المصادر",
refreshAll: "تحديث الكل",
refreshing: "جارٍ التحديث...",
loading: "جارٍ تحميل بيانات الاستخدام...",
@@ -2323,11 +2328,9 @@ export const ar: TranslationResources = {
resetsIn: "تتم إعادة التعيين خلال {{duration}}",
resetsAtWeekday: "{{weekday}} {{time}}",
resetsAtDate: "{{date}}، {{time}}",
projectedLeftAtReset: "نحو {{percentage}} متبقية عند إعادة التعيين",
limitReached: "تم بلوغ الحد",
mayHaveRunOut: "ربما نفدت الحصة",
runsOutNow: "ينفد الآن",
runsOutIn: "ينفد خلال {{duration}}",
runsOutAt: "ينفد عند {{time}}",
justNow: "الآن",
daysAgo: "قبل {{value}} يومًا",
daysAgo_zero: "قبل {{value}} يوم",
@@ -3115,6 +3118,7 @@ export const ar: TranslationResources = {
status: "حالة المزود",
checked: "تم الفحص {{time}}",
diagnosticReport: "تقرير التشخيص",
docs: "الوثائق",
dangerZone: "منطقة الخطر",
removeHint: "يحذف هذا المزود من config.json",
},
@@ -3160,7 +3164,7 @@ export const ar: TranslationResources = {
noSearchMatches: "لا توجد نماذج تطابق بحثك",
noneDetected: "لم يتم اكتشاف أي نماذج",
noCustom: "لا توجد نماذج مخصصة",
discovered: "اكتشف",
title: "النماذج",
custom: "نماذج مخصصة",
updated: "تم تحديث{{time}}",
},
+9 -5
View File
@@ -348,6 +348,7 @@ export const en = {
copyTurn: "Copy turn",
copyMessage: "Copy message",
forkMenu: "Fork chat from here",
turnStats: "Turn stats",
forkInNewTab: "Fork in a new tab",
forkInNewWorkspace: "Fork in a new workspace",
forkUnavailable: "Update the host to use this.",
@@ -2292,7 +2293,11 @@ export const en = {
title: "Plan usage",
refresh: "Refresh",
reorderCard: "Reorder {{source}}",
showCard: "Show {{source}}",
sources: "Sources",
sourcesCount: "{{shown}} of {{total}}",
providerInactive: "Provider inactive",
allSourcesHidden: "No usage sources shown",
chooseSources: "Choose sources",
refreshAll: "Refresh all",
refreshing: "Refreshing...",
loading: "Loading usage...",
@@ -2333,11 +2338,9 @@ export const en = {
resetsIn: "resets in {{duration}}",
resetsAtWeekday: "{{weekday}} {{time}}",
resetsAtDate: "{{date}}, {{time}}",
projectedLeftAtReset: "~{{percentage}} left at reset",
limitReached: "Limit reached",
mayHaveRunOut: "May have run out",
runsOutNow: "runs out now",
runsOutIn: "runs out in {{duration}}",
runsOutAt: "runs out {{time}}",
justNow: "just now",
daysAgo: "{{value}}d ago",
hoursAgo: "{{value}}h ago",
@@ -3213,6 +3216,7 @@ export const en = {
status: "Provider status",
checked: "Checked {{time}}",
diagnosticReport: "Diagnostic report",
docs: "Docs",
dangerZone: "Danger zone",
removeHint: "Deletes this provider from config.json",
},
@@ -3258,7 +3262,7 @@ export const en = {
noSearchMatches: "No models match your search",
noneDetected: "No models detected",
noCustom: "No custom models",
discovered: "Discovered",
title: "Models",
custom: "Custom models",
updated: "Updated {{time}}",
},
+9 -5
View File
@@ -354,6 +354,7 @@ export const es: TranslationResources = {
copyTurn: "Copiar turno",
copyMessage: "Copiar mensaje",
forkMenu: "Bifurcar mensaje",
turnStats: "Estadísticas del turno",
forkInNewTab: "Bifurcar en una pestaña nueva",
forkInNewWorkspace: "Bifurcar en un espacio de trabajo nuevo",
forkUnavailable: "Actualiza el host para usar esto.",
@@ -2313,7 +2314,11 @@ export const es: TranslationResources = {
title: "Uso del plan",
refresh: "Actualizar",
reorderCard: "Reordenar {{source}}",
showCard: "Mostrar {{source}}",
sources: "Fuentes",
sourcesCount: "{{shown}} de {{total}}",
providerInactive: "Proveedor inactivo",
allSourcesHidden: "No hay fuentes de uso visibles",
chooseSources: "Elegir fuentes",
refreshAll: "Actualizar todo",
refreshing: "Actualizando...",
loading: "Cargando uso...",
@@ -2354,11 +2359,9 @@ export const es: TranslationResources = {
resetsIn: "se restablece en {{duration}}",
resetsAtWeekday: "{{weekday}} {{time}}",
resetsAtDate: "{{date}}, {{time}}",
projectedLeftAtReset: "~{{percentage}} restante al restablecerse",
limitReached: "Límite alcanzado",
mayHaveRunOut: "Puede que ya se haya agotado",
runsOutNow: "se agota ahora",
runsOutIn: "se agota en {{duration}}",
runsOutAt: "agotamiento previsto {{time}}",
justNow: "ahora mismo",
daysAgo: "hace {{value}} d",
hoursAgo: "hace {{value}} h",
@@ -3143,6 +3146,7 @@ export const es: TranslationResources = {
status: "Estado del proveedor",
checked: "Comprobado {{time}}",
diagnosticReport: "Informe de diagnóstico",
docs: "Documentación",
dangerZone: "Zona de peligro",
removeHint: "Elimina este proveedor de config.json",
},
@@ -3188,7 +3192,7 @@ export const es: TranslationResources = {
noSearchMatches: "Ningún modelo coincide con tu búsqueda",
noneDetected: "No se detectaron modelos",
noCustom: "No hay modelos personalizados",
discovered: "descubierto",
title: "Modelos",
custom: "Modelos personalizados",
updated: "{{time}}actualizado",
},
+9 -5
View File
@@ -354,6 +354,7 @@ export const fr: TranslationResources = {
copyTurn: "Copier l’échange",
copyMessage: "Copier le message",
forkMenu: "Forker la conversation à partir d’ici",
turnStats: "Statistiques de l’échange",
forkInNewTab: "Forker dans un nouvel onglet",
forkInNewWorkspace: "Forker dans un nouvel espace de travail",
forkUnavailable: "Mettez à jour l’hôte pour utiliser cette fonction.",
@@ -2314,7 +2315,11 @@ export const fr: TranslationResources = {
title: "Utilisation du forfait",
refresh: "Actualiser",
reorderCard: "Réorganiser {{source}}",
showCard: "Afficher {{source}}",
sources: "Sources",
sourcesCount: "{{shown}} sur {{total}}",
providerInactive: "Fournisseur inactif",
allSourcesHidden: "Aucune source d'utilisation affichée",
chooseSources: "Choisir les sources",
refreshAll: "Tout actualiser",
refreshing: "Actualisation...",
loading: "Chargement de l’utilisation...",
@@ -2355,11 +2360,9 @@ export const fr: TranslationResources = {
resetsIn: "se réinitialise dans {{duration}}",
resetsAtWeekday: "{{weekday}} {{time}}",
resetsAtDate: "{{date}}, {{time}}",
projectedLeftAtReset: "~{{percentage}} restant à la réinitialisation",
limitReached: "Limite atteinte",
mayHaveRunOut: "Le quota est peut-être épuisé",
runsOutNow: "s’épuise maintenant",
runsOutIn: "s’épuise dans {{duration}}",
runsOutAt: "épuisement prévu {{time}}",
justNow: "à l’instant",
daysAgo: "il y a {{value}} j",
hoursAgo: "il y a {{value}} h",
@@ -3149,6 +3152,7 @@ export const fr: TranslationResources = {
status: "État du fournisseur",
checked: "Vérifié {{time}}",
diagnosticReport: "Rapport de diagnostic",
docs: "Documentation",
dangerZone: "Zone de danger",
removeHint: "Supprime ce fournisseur de config.json",
},
@@ -3195,7 +3199,7 @@ export const fr: TranslationResources = {
noSearchMatches: "Aucun modèle ne correspond à votre recherche",
noneDetected: "Aucun modèle détecté",
noCustom: "Aucun modèle personnalisé",
discovered: "Détectés",
title: "Modèles",
custom: "Modèles personnalisés",
updated: "Mis à jour {{time}}",
},
+9 -5
View File
@@ -354,6 +354,7 @@ export const ja: TranslationResources = {
copyTurn: "ターンをコピー",
copyMessage: "メッセージをコピー",
forkMenu: "メッセージをフォーク",
turnStats: "ターンの統計",
forkInNewTab: "新しいタブにフォーク",
forkInNewWorkspace: "新しいワークスペースにフォーク",
forkUnavailable: "これを使用するにはホストを更新してください。",
@@ -2287,7 +2288,11 @@ export const ja: TranslationResources = {
title: "プラン使用量",
refresh: "更新",
reorderCard: "{{source}} を並べ替え",
showCard: "{{source}} を表示",
sources: "ソース",
sourcesCount: "{{shown}} / {{total}}",
providerInactive: "プロバイダー無効",
allSourcesHidden: "表示中の使用量ソースはありません",
chooseSources: "ソースを選択",
refreshAll: "すべて更新",
refreshing: "更新中...",
loading: "使用状況を読み込み中...",
@@ -2328,11 +2333,9 @@ export const ja: TranslationResources = {
resetsIn: "{{duration}}後にリセット",
resetsAtWeekday: "{{weekday}} {{time}}",
resetsAtDate: "{{date}} {{time}}",
projectedLeftAtReset: "リセット時に約{{percentage}}残る見込み",
limitReached: "上限に達しました",
mayHaveRunOut: "上限に達した可能性があります",
runsOutNow: "まもなく上限に到達",
runsOutIn: "{{duration}}後に上限に到達",
runsOutAt: "{{time}}に上限に到達",
justNow: "たった今",
daysAgo: "{{value}}日前",
hoursAgo: "{{value}}時間前",
@@ -3111,6 +3114,7 @@ export const ja: TranslationResources = {
status: "プロバイダーの状態",
checked: "{{time}}に確認",
diagnosticReport: "診断レポート",
docs: "ドキュメント",
dangerZone: "危険な操作",
removeHint: "このプロバイダーを config.json から削除します",
},
@@ -3156,7 +3160,7 @@ export const ja: TranslationResources = {
noSearchMatches: "検索に一致するモデルがありません",
noneDetected: "モデルが検出されませんでした",
noCustom: "カスタムモデルはありません",
discovered: "検出済み",
title: "モデル",
custom: "カスタムモデル",
updated: "{{time}}に更新",
},
+9 -5
View File
@@ -350,6 +350,7 @@ export const ko: TranslationResources = {
copyTurn: "턴 복사",
copyMessage: "메시지 복사",
forkMenu: "여기에서 채팅 분기",
turnStats: "턴 통계",
forkInNewTab: "새 탭으로 분기",
forkInNewWorkspace: "새 워크스페이스로 분기",
forkUnavailable: "이를 사용하려면 호스트를 업데이트하세요.",
@@ -2275,7 +2276,11 @@ export const ko: TranslationResources = {
title: "플랜 사용량",
refresh: "새로 고침",
reorderCard: "{{source}} 순서 변경",
showCard: "{{source}} 표시",
sources: "소스",
sourcesCount: "{{shown}} / {{total}}",
providerInactive: "제공업체 비활성",
allSourcesHidden: "표시된 사용량 소스가 없습니다",
chooseSources: "소스 선택",
refreshAll: "모두 새로 고침",
refreshing: "새로 고치는 중...",
loading: "사용량 불러오는 중...",
@@ -2316,11 +2321,9 @@ export const ko: TranslationResources = {
resetsIn: "{{duration}} 후 재설정",
resetsAtWeekday: "{{weekday}} {{time}}",
resetsAtDate: "{{date}} {{time}}",
projectedLeftAtReset: "초기화 시 약 {{percentage}} 남을 예정",
limitReached: "한도에 도달함",
mayHaveRunOut: "한도를 소진했을 수 있음",
runsOutNow: "지금 한도에 도달",
runsOutIn: "{{duration}} 후 한도에 도달",
runsOutAt: "{{time}}에 한도에 도달",
justNow: "방금",
daysAgo: "{{value}}일 전",
hoursAgo: "{{value}}시간 전",
@@ -3096,6 +3099,7 @@ export const ko: TranslationResources = {
status: "프로바이더 상태",
checked: "{{time}} 확인됨",
diagnosticReport: "진단 보고서",
docs: "문서",
dangerZone: "위험 구역",
removeHint: "config.json에서 이 프로바이더를 삭제합니다",
},
@@ -3142,7 +3146,7 @@ export const ko: TranslationResources = {
noSearchMatches: "검색과 일치하는 모델이 없습니다",
noneDetected: "감지된 모델이 없습니다",
noCustom: "사용자 지정 모델 없음",
discovered: "발견됨",
title: "모델",
custom: "사용자 지정 모델",
updated: "{{time}} 업데이트됨",
},
+9 -5
View File
@@ -353,6 +353,7 @@ export const ptBR: TranslationResources = {
copyTurn: "Copiar turno",
copyMessage: "Copiar mensagem",
forkMenu: "Bifurcar mensagem",
turnStats: "Estatísticas do turno",
forkInNewTab: "Bifurcar em uma nova aba",
forkInNewWorkspace: "Bifurcar em um novo workspace",
forkUnavailable: "Atualize o host para usar isto.",
@@ -2298,7 +2299,11 @@ export const ptBR: TranslationResources = {
title: "Uso do plano",
refresh: "Atualizar",
reorderCard: "Reordenar {{source}}",
showCard: "Mostrar {{source}}",
sources: "Fontes",
sourcesCount: "{{shown}} de {{total}}",
providerInactive: "Provedor inativo",
allSourcesHidden: "Nenhuma fonte de uso exibida",
chooseSources: "Escolher fontes",
refreshAll: "Atualizar tudo",
refreshing: "Atualizando...",
loading: "Carregando uso...",
@@ -2339,11 +2344,9 @@ export const ptBR: TranslationResources = {
resetsIn: "redefine em {{duration}}",
resetsAtWeekday: "{{weekday}} {{time}}",
resetsAtDate: "{{date}}, {{time}}",
projectedLeftAtReset: "~{{percentage}} restará na renovação",
limitReached: "Limite atingido",
mayHaveRunOut: "A cota pode ter se esgotado",
runsOutNow: "esgota agora",
runsOutIn: "esgota em {{duration}}",
runsOutAt: "esgotamento previsto para {{time}}",
justNow: "agora mesmo",
daysAgo: "há {{value}} d",
hoursAgo: "há {{value}} h",
@@ -3124,6 +3127,7 @@ export const ptBR: TranslationResources = {
status: "Status do provedor",
checked: "Verificado {{time}}",
diagnosticReport: "Relatório de diagnóstico",
docs: "Documentação",
dangerZone: "Zona de perigo",
removeHint: "Exclui este provedor do config.json",
},
@@ -3169,7 +3173,7 @@ export const ptBR: TranslationResources = {
noSearchMatches: "Nenhum modelo corresponde à sua busca",
noneDetected: "Nenhum modelo detectado",
noCustom: "Nenhum modelo personalizado",
discovered: "Descobertos",
title: "Modelos",
custom: "Modelos personalizados",
updated: "Atualizado {{time}}",
},
+9 -5
View File
@@ -353,6 +353,7 @@ export const ru: TranslationResources = {
copyTurn: "Скопировать ответ",
copyMessage: "Копировать сообщение",
forkMenu: "Форкнуть чат отсюда",
turnStats: "Статистика ответа",
forkInNewTab: "Создать форк в новой вкладке",
forkInNewWorkspace: "Создать форк в новом рабочем пространстве",
forkUnavailable: "Обновите хост, чтобы использовать эту функцию.",
@@ -2300,7 +2301,11 @@ export const ru: TranslationResources = {
title: "Использование плана",
refresh: "Обновить",
reorderCard: "Переместить {{source}}",
showCard: "Показывать {{source}}",
sources: "Источники",
sourcesCount: "{{shown}} из {{total}}",
providerInactive: "Провайдер неактивен",
allSourcesHidden: "Источники использования скрыты",
chooseSources: "Выбрать источники",
refreshAll: "Обновить всё",
refreshing: "Обновление...",
loading: "Загрузка данных об использовании...",
@@ -2341,11 +2346,9 @@ export const ru: TranslationResources = {
resetsIn: "сброс через {{duration}}",
resetsAtWeekday: "{{weekday}} {{time}}",
resetsAtDate: "{{date}}, {{time}}",
projectedLeftAtReset: "К сбросу останется ~{{percentage}}",
limitReached: "Лимит исчерпан",
mayHaveRunOut: "Лимит мог уже исчерпаться",
runsOutNow: "закончится сейчас",
runsOutIn: "закончится через {{duration}}",
runsOutAt: "закончится {{time}}",
justNow: "только что",
daysAgo: "{{value}} дн. назад",
hoursAgo: "{{value}} ч назад",
@@ -3135,6 +3138,7 @@ export const ru: TranslationResources = {
status: "Состояние провайдера",
checked: "Проверено {{time}}",
diagnosticReport: "Диагностический отчёт",
docs: "Документация",
dangerZone: "Опасная зона",
removeHint: "Удаляет этого провайдера из config.json",
},
@@ -3181,7 +3185,7 @@ export const ru: TranslationResources = {
noSearchMatches: "Ни одна модель не соответствует вашему запросу",
noneDetected: "Модели не обнаружены",
noCustom: "Нет пользовательских моделей",
discovered: "Обнаруженные модели",
title: "Модели",
custom: "Пользовательские модели",
updated: "Обновлено {{time}}",
},

Some files were not shown because too many files have changed in this diff Show More