* fix(server): keep the model selected by the provider
A live model setter can retain or resolve a different model than the requested ID. Read its effective selection before persisting manager state so an ignored selection cannot overwrite the actual model. Preserve explicit preference resets.
* fix(server): preserve completed model selections
Return the effective selection from the provider mutation so a separate status request cannot reject an already completed switch or leave configured and runtime models inconsistent.
---------
Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
* fix(server): preserve the selected model when importing sessions
Import preparation resolved a new-session catalog default and sent it as an explicit resume override. Keep import selections with the resumed provider and persist its reported model for subsequent reloads.
* fix(codex): allow the first reply after importing a session
Resolve collaboration settings again when default model and effort are applied so imported sessions send a complete first-turn request.
---------
Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
* fix(server): give agents their workspace id as PASEO_WORKSPACE_ID
Agents got PASEO_AGENT_ID and PASEO_AGENT_CWD but not their workspace id, so an
agent couldn't rename or archive its own workspace, or name it to create_agent,
when several workspaces share its cwd. The launch env now sets
PASEO_WORKSPACE_ID whenever the agent has a workspace, including on resume after
a daemon restart, where the id now comes from the stored record. OpenCode treats
it as a per-session key, so it doesn't force a dedicated server.
Refs #5098
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(server): keep the workspace when resuming an agent by handle
resume_agent_request matched the stored agent but dropped its workspace id, so
the resumed agent had no PASEO_WORKSPACE_ID and no workspace. Pass the matched
record's workspace id into the resume.
resumeAgentFromPersistence now resolves the workspace once and gives it to both
the launch env and the registered agent, so a caller that omits it no longer
erases the stored value on the next snapshot.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(server): exercise current daemon creation and permission contracts
The workspace resume regression shares this suite with creation checks that still waited for legacy status frames and permission listeners without an owned subscription. Verify the returned error snapshot and acquire event demand so the complete suite can validate workspace identity changes.
Co-authored-by: Darrell <DT@Terastar.biz>
---------
Co-authored-by: Darrell <DT@Terastar.biz>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
* feat(server): add an in-memory store for agent background work
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(claude): report background tasks as background work
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(server): track agent background work in the agent manager
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(protocol): add agent background work messages
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(server): serve and push agent background work
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(client): expose agent background work on agent handles
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(sdk): document agent background work
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(claude): cover background shells in background work end to end
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(client): resync background work subscriptions on every snapshot
subscribe() now lists the agent's background work each time the event
observation is established, so the listener starts with the current list
and catches up after a reconnect. A list answered after a newer update
is dropped.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(server): answer background work lists without loading dormant agents
Background work lives only in memory, so a stored agent that is not
loaded has none. The list handler now answers [] from storage instead of
resuming the agent, and still rejects unknown, internal, and archived
agents. A new test pins that old apps never receive
agent.background_work.update.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(server): read Claude background tasks defensively
The user's claude binary can differ from the SDK typings, so a task's
description is read as unknown and an empty or missing one becomes null.
Drop the monitor kind: Claude Monitor watches report as local_bash, so
kind is "shell" or "other" today. The store now copies only the wire
fields of each input.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(server): prove internal background work was emitted before checking globals
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(sdk): describe background work resync and feature gating
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(claude): cover background work reported by a retired query
The daemon stops reading a Claude query it no longer owns, so work that a
retired process still had queued cannot reach the background work list.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(client): let background work subscribers handle a failed list read
When the list read that follows each subscription snapshot failed, the
error was only logged and the caller got neither the list nor a failure.
subscribe() now takes an optional onError that receives it; without one
the error is still logged. The read is not retried, since the next
reconnect reads the list again.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(agents): show background work as active
Keep background activity visible after a foreground turn ends while preserving independent send and interrupt behavior.
* fix(agents): protect background work subscriptions
Keep internal work out of public updates and restore observation after a subscription failure.
* feat(agents): show Codex and OpenCode background work
Keep provider-owned shell activity visible after a turn ends while leaving foreground turn controls independent.
* feat(agents): stop individual background work
Wait for provider confirmation so stopping one background item leaves sibling work and foreground turn controls independent.
* refactor(agents): name background stop parameters
* test(website): isolate browser tests from live metadata
* test(agents): clarify background stop journeys
* fix(sdk): report background work subscription failures
Report setup failures so subscribers can replace an observation that the host released.
* fix(app): restore background work after reconnect
Wait for the host handshake before selecting the retained client so an offline pane starts observing when the connection returns.
* fix(agents): retain background shells across rewind
Preserve known live shell ownership when conversation history is rewound, and revoke it when the provider runtime exits.
* fix(website): launch test seeder across platforms
Invoke Wrangler through Node so local website test startup does not depend on an executable npm command shim.
* feat(agents): clarify background task details and tracks
Keep launch descriptions and commands distinct so background work remains identifiable, and allow composer tracks to scroll when space is limited.
* fix(agents): preserve active background work during archive
Keep background-running children out of bulk cleanup and warn before an individual archive. Preserve first taps on composer tracks while the keyboard is open.
* fix(codex): retain background terminals across rewind
Keep live terminal inventories attached to their originating threads so a fork does not hide or misroute their stop actions.
* fix(codex): scope background terminal identities to threads
Prevent commands on separate threads with the same provider process ID from sharing a stop target.
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Mohamed Boudra <boudra.moha@gmail.com>
* fix(server): retain archived parent provenance on detach
* fix(server): publish stored child detach updates on parent archive
Keep subscribed clients current when an unloaded child survives its parent archive. Narrow the contributor change to the missing update; archived-parent provenance remains a separate feature.
Co-authored-by: luke <luopeihuan@gmail.com>
* test(server): clean up the detach fixture static directory
Keep the test daemon static directory inside the fixture root so cleanup removes all test-owned directories.
* fix(server): retain runtime settings in stored agent updates
Parent archive publishes stored survivors through the manager. Preserve their saved runtime model and thinking settings so the update matches stored fetches and subscription filters.
---------
Co-authored-by: luke <luopeihuan@gmail.com>
Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
* fix(sidebar): clear the failed status once the error is read
Reading an errored agent already cleared its error attention, but the
state bucket also mapped lifecycle status "error" to failed. Lifecycle
only leaves "error" when the next turn starts, so the red dot and the
Failed group stuck until the agent was prompted again. The failed bucket
now follows unread error attention, the same way the green dot follows
unread finished attention; lifecycle status stays "error".
Because failed now depends on attention, a failure must replace unread
finished attention, otherwise a reload that fails on an unread agent
would show green.
* fix(sidebar): keep failed subagents red and let read failures be marked unread
The failed bucket now follows unread error attention, which broke three
places that relied on lifecycle "error" alone:
- Subagent track rows and provider-subagent tabs have no read state of
their own, so failed children lost their red mark. They now map the
child's lifecycle directly.
- A read errored workspace became "done" and offered Mark as unread, but
the daemon only accepted idle or closed agents. Marking it unread now
restores error attention.
- A reload that failed again while the lifecycle was already "error"
raised no new attention. A failed reload now raises error attention
directly instead of relying on a lifecycle transition.
An execution adopted during session initialization emitted turn_started before the manager could subscribe. Deliver the current turn to a new subscriber so the manager owns the run and routes Stop to OpenCode.
Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
Separate runtime release from explicit turn cancellation, drain close-time cleanup before retirement, and reject late query creation. Add regressions for closure, pending tools and permissions, and asynchronous launch.
Co-authored-by: babo-jeonguk <278435975+babo-jeonguk@users.noreply.github.com>
* fix(server): let a message steered into a running Claude turn be rewound to
A steered message never got a provider message id, because Claude takes it
into the running turn without echoing it back, so rewinding to it refused
with "Cannot rewind before the provider acknowledges the submitted prompt".
The steer admission now reports the uuid Claude minted for the message, and
a Claude conversation rewind to it forks at the entry Claude read it after.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(server): keep messages steered into a running Claude turn in the replayed timeline
Claude records a steered message as a queued_command attachment rather than a
user entry, and the history replay skipped it, so every such message, replies
from other agents included, vanished from the timeline after a daemon restart.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(claude): verify steer identities across lifecycle tests
Match the accepted-steer contract used by conversation rewind and verify that its identity names the message delivered to the provider.
Co-authored-by: tieo <65707274+tieo@users.noreply.github.com>
* fix(claude): release transcript streams after rewind lookup
Closing readline leaves a partially read transcript open after an early match. Destroy the owned stream on exit and cover repeated lookups with unread data.
---------
Co-authored-by: tieo <65707274+tieo@users.noreply.github.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>
* fix(server): bound every timeline item so one fetch cannot close the socket
A timeline page is one WebSocket message. Only three tool-call fields were
capped, so MCP results, file contents, errors and merged assistant rows could
make a 20-entry or even a 1-entry fetch pass the outbound high-water mark.
limitAgentTimelineItemContent is now one budgeted walk over any item: 64 KiB
for tool calls, 256 KiB for other items. Over budget, the largest strings are
cut to a shared length with a visible marker and oversized data URLs become a
marker. Fetches also apply it to merged assistant and reasoning rows.
* fix(server): keep short strings whole and typed arrays valid when limiting items
Strings up to 1 KiB are never cut, so a tool call keeps its callId and an
item limits to the same result twice. Arrays drop trailing elements
without inserting a marker element, which typed arrays such as todo items
would reject.
* fix(server): broadcast agent removals across logical sessions
* test(server): make subscription fixtures portable on Windows
The newly enabled integration suite assumed /bin/cat and a one-second reply deadline, so Windows CI failed before completing its subscription coverage. Use the running Node executable for echo terminals and let diagnostics finish their bounded external tool probes.
Co-authored-by: loganmeyer <50949542+loganmeyer@users.noreply.github.com>
---------
Co-authored-by: loganmeyer <50949542+loganmeyer@users.noreply.github.com>
Co-authored-by: paseo-bot[bot] <paseo-bot[bot]@users.noreply.github.com>
* Deliver a steer message that arrives while the previous turn is still starting
A steer send that reached an agent before its previous turn had a turn id
found no turn to steer and replaced the starting turn, so the earlier
message never reached the provider. Steering now waits for the starting
turn to be accepted (or fail) and steers the turn it became.
Fixes#6186
* Wake a waiting steer only once the starting turn has an id or its run has ended
A failed start used to wake the waiting steer before its run was cleared,
leaving the later send to replace a run that had already failed.
* Clear a failed start's run even when reporting the failure throws
A steer waiting on that start is released only when the run is cleared.
* Test that a failed start keeps a waiting steer until its run is cleared
* fix(server): steer Codex turns while native identification is pending
Wait for the accepted foreground turn to receive its native ID before steering it. Returning unavailable in this window replaced the turn even after the manager had waited for startup.
---------
Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
* feat(server): resume interrupted agents after host restart
Continue saved running conversations after graceful restart or crash, with host opt-out and a configurable continuation prompt. Preserve explicit launch overrides when reopening the same provider session.
Co-authored-by: coolwind <45729206+coolwindX@users.noreply.github.com>
Co-authored-by: suntianc <suntc8985@gmail.com>
* fix(app): translate restart recovery host settings
* test(app): synchronize sidebar drags and clean up added projects
Wait for the drag sensor click blocker to expire before the next user action. Resolve UI project keys to daemon IDs for cleanup so empty projects do not contaminate later sidebar tests.
* fix(app): tuck the resume prompt behind an edit sheet
Match the System prompt row: an Edit button left of the restart recovery
switch opens the prompt in a sheet, and is disabled while resume is off.
* fix(server): resume interrupted agents without holding startup
Startup now only chooses resume candidates before anything else reads or
repairs persisted running state. Reopening and prompting run in the
background after connections open, so a provider that is slow to reopen
holds back only its own agent. Finish observation waits for a pending
child's resume, so parents are not told the child was closed meanwhile.
* style(server): format restart resume e2e test
* test(server): wait for the fetch result before advancing the narrow refresh debounce
The test released the held fetch, flushed six microtasks, then advanced
the 1s refresh debounce. When the fetch result landed after that advance
(slow Windows runners), the debounce started late and vi.waitFor's
auto-advance reached it only at its real-time timeout. Wait on the
service's in-flight fetch count instead, as the sibling test does.
* test(server): retry stalled OpenCode fixture downloads within the test budget
npm abandons a stalled registry download only after its 300s default
fetch timeout, past the install's 180s budget. The killed install left
the fixture directory locked on Windows, so cleanup failed with EBUSY and
hid the timeout. Against a registry proxy that stalls the first tarball,
a default install hung 225s; with --fetch-timeout it retried and finished
in 16s.
---------
Co-authored-by: suntianc <suntc8985@gmail.com>
* refactor(plugins): make lifecycle hook timing explicit
Use one registration method and subject.operation.step names so plugin authors can read whether an operation waits. Preserve old registrations through a single SDK adapter and keep timing and payloads unchanged.
* fix(plugins): align hook notes and example fixtures with release contracts
Keep the changelog inside its versioned beta entry for app parsing. Run current plugin examples against the checkout version so their declared requirements remain valid.
* docs(plugins): align translated hook dependency guidance
* fix(plugins): align hook exports and release documentation
* feat(plugins): use lifecycle positions and await worktree initialization
* test(plugins): cover first-agent worktree initialization
* test(plugins): verify rollback against workspace catalog entries
* fix(app): keep autonomous turns distinct and readable
Hidden system prompts merged autonomous work into the previous response and reopened completed turns. Preserve canonical turn boundaries and expose their triggers, while making received messages attributable and expanded turn details readable in one step.
* fix(timeline): mount history at actual run boundaries
* fix(timeline): load and retain complete available turns
* fix(app): summarize actions and preserve conversation events
* test(timeline): verify whole-turn paging stays responsive
* fix(app): group reasoning according to thinking preference
* test(app): align history journeys with whole-turn loading
* feat(server): release idle agent runtimes with provider safety checks
Require provider permission before releasing resumable runtimes so foreground idleness does not discard tracked background work. Preserve the saved conversation and serialize prompt admission with runtime closure; allow hosts to disable collection.
* docs: translate idle runtime cleanup configuration and provider contract
* feat(server): collect runtimes after one idle hour with optional provider vetoes
Keep callers resident while local or remote finish notification intent remains outstanding. Daemon-owned work determines baseline eligibility; provider checks add optional vetoes. Own the idle clock and sweep lifecycle in one cleanup module while the manager retains guarded closure.
* feat(agents): control idle runtime release live from host settings
Let hosts retain runtimes without restarting the daemon. Disabling invalidates pending collection decisions while preserving idle clocks for re-enabling. ACP-owned terminals and pending adapter work veto collection until they finish.
* fix(settings): clarify idle agent release with an info tooltip
* fix(server): restore agent finish notifications after daemon restart
Finish observers lived only in memory, so a restarted daemon silently lost the wakeup parents relied on. Persist subscription intent through notification dispatch acknowledgment and reconcile local and remote children after bootstrap.
* fix(server): reject finish subscriptions for unsent remote prompts
Connection and snapshot preflight failures happen before remote prompt transmission. Reject those candidates while retaining ambiguous sends, and preserve the host identity in callbacks. Exercise local permission cycles and storage errors through MCP.
* fix(server): notify parents when remote finish observation is exhausted
Three failed observations or a removed registration consume the watch after the parent accepts dispatch. Report unknown child status, retain the captured outcome through delivery failures, and keep connecting attempts outside the failure budget.
* fix(server): describe exhausted child observation accurately
Remote RPC failures do not prove the host is offline. Report the failed observation and unknown child status, and give Windows integration polling more margin.
* fix(server): batch agent reports and completion notifications
Admit agent-originated messages immediately so a child can report and finish before one recipient-owned delivery. Keep human prompts direct, retain individual provenance through replay, and complete send receipts only after provider dispatch.
* test(server): wire queued delivery into bridge integration fixture
* test(server): resolve batching CLI fixture independently of cwd
Workspace CI runs from packages/server, while the original regression was verified from the repository root. Resolve the CLI entrypoint relative to this test file in both contexts.
* test(server): run batching CLI fixture with source exports
Server integration CI builds only the client dependency stack. Match the existing CLI subprocess fixtures so the send regression exercises workspace sources without requiring compiled server output.
* test(server): await persisted batch receipts within integration deadline
The provider-start callback precedes asynchronous receipt commits. Use the same explicit integration deadline as dispatch instead of the matcher default of one second on Windows CI.
Stop materializing and copying the entire conversation through lifecycle hooks. Keep the deprecated timeline field empty and direct plugins to bounded timeline reads.
* fix(server): run slash commands sent by another agent out of band again
* fix(omp): send /steer and /follow-up as prompts so an idle agent starts a run
* test(omp): verify idle follow-ups and parent notifications
Cover real daemon and MCP delivery, and keep the fake runtime tests independent of an installed OMP executable.
Co-authored-by: Hung Le <22406703+nhthung@users.noreply.github.com>
* fix(omp): report idle follow-up admission failures
Use the existing prompt lifecycle for an idle command so acknowledgements and late rejections settle the child and notify its caller. Keep busy commands on the streaming prompt path, and accept one runtime prompt input object.
---------
Co-authored-by: Hung Le <22406703+nhthung@users.noreply.github.com>
Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
* feat(server): let clients create internal agents
The daemon's own branch-name and commit-message generators run as
internal agents: never persisted, listed, flagged, announced, or seen by
plugin lifecycle hooks. Nothing outside the daemon could ask for one,
because the wire config schema strips `internal` and the agent.create
transform re-pins it. A plugin that runs a one-turn helper per event
therefore left an archived session in History every time.
Add an optional `internal` flag to `create_agent_request` (inherited by
`agent.create.request`), copy it onto the config in the session's create
path, and skip provider session persistence for it like the generators
do. The client SDK forwards the flag on both creation paths and gates it
once on `server_info.features.internalAgents`.
Two daemon paths assumed every agent is public and had to follow:
`resolveAgentIdentifier` now accepts a live agent by exact id, since the
id set it matched against came from listings that exclude internal
agents and left `wait_for_finish` answering "Agent not found"; and
archiving an internal agent now closes it without writing the record
that `registry.applySnapshot` used to leave behind, so `autoArchive`
does not accumulate hidden files on disk.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(server): keep internal agents whole through failure, archive, and wait
Review follow-ups for internal agents:
- A create that fails after the runtime is registered now closes the
internal agent before rethrowing, so a hidden runtime cannot outlive
the failed request. Public agents keep today's behaviour.
- Archiving an internal agent cascades to attached children like any
parent. cascadeArchiveChildren takes the parent record instead of
reading it back from storage, where an internal agent never is.
- The archived snapshot and last message of an internal agent stay
readable by exact id for ten minutes, so create-then-waitForFinish
returns the finished turn whether the wait lands before or after an
auto-archive. Storage plays that role for public agents.
- The internal-agent e2e scenarios live in the existing
create_agent_request lifecycle suite rather than a new file.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(server): cascade to live internal children and keep a failed internal create's worktree until it closes
Internal children never reach storage, so the parent archive cascade scans
live agents for them as well. A failed internal create now clears the
created id only after the runtime is archived, so worktree cleanup cannot
delete the directory under a runtime that would not close. Stored agent
payloads carry lastError, so a retained failed helper reports its error
through wait_for_finish instead of the generic message.
* feat(cli,mcp): expose internal agents and document --no-wait
paseo run gains --internal and create_agent gains internal, both forwarded
to the create request. --no-wait becomes the documented spelling for
returning immediately, matching paseo send; --background and --detach stay
as hidden aliases under COMPAT tags.
* test: follow the --no-wait rename and stub getRetiredInternalAgent
* feat: list internal agents on request and add internal workspaces
Internal agents stay hidden by default, but a caller can now ask for
them: `fetch_agents_request` takes `filter.includeInternal`, snapshots
carry `internal: true`, and `paseo ls --internal` shows them in an
INTERNAL column. The app never passes the flag.
Workspaces get the same flag. `workspace.create.request` with
`internal: true` persists a hidden workspace: listings, update
subscriptions, sync reads, empty-project counting, open-by-path, MCP
`list_workspaces`, and plugin workspace hooks all skip it unless the
caller opts in with `filter.includeInternal`. Every agent created inside
an internal workspace is internal and unpersisted; AgentManager applies
that from the workspace id before the plugin hook so creators cannot
opt out. Archive and shutdown enumerate internal agents so nothing
inside an internal workspace outlives it.
Surfaces: `paseo workspace create --internal`, `paseo workspace ls
--internal`, `paseo run --internal --new-workspace`, MCP
`create_workspace({ internal })`, SDK `workspaces.create({ internal })`
and `list({ filter: { includeInternal } })`. Clients gate on
`features.internalWorkspaces`.
Co-Authored-By: Claude <noreply@anthropic.com>
* feat(workspaces): add background visibility with inherited creation defaults
* fix(cli): resolve workspace callers on the selected daemon
* test(workspaces): keep caller integration coverage in server suite
* test(cli): exercise workspace callers through CLI integration harness
* test(mcp): remove obsolete execution flags from browser fixtures
---------
Co-authored-by: marvin-ambrozio <marvin@gustavo.eng.br>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Co-authored-by: Mohamed Boudra <boudra.moha@gmail.com>
* fix(agents): preserve creation when the first prompt fails
A rejected initial prompt must leave a created agent with a turn error instead of returning the user to a draft whose creation key is already committed. Preserve receipt replay without repeating the prompt, and bound fork history to 100,000 characters before it becomes an attachment.
* test(creation): isolate user agent counters from background naming
Workspace naming uses internal provider sessions. Counting and rejecting those sessions made creation assertions depend on installed provider discovery timing. Keep discovery within the test adapters and scope observations to user-created agents.
* test(creation): retry temporary directory cleanup on Windows
Windows CI completed the rejection assertions but failed teardown with EBUSY. Match the daemon test harness bounded removal retries so transient directory locks do not fail creation tests.
* fix(agents): record rejected prompts before their errors
Startup rejection recorded only the system error, leaving the client to append its local prompt after it. Record the rejected submission at the terminal failure boundary so stream and fetched history agree on prompt-before-error ordering.
* fix(agents): require acceptance for delivery notifications
Retain rejected human submissions before their errors without projecting rejected inter-agent prompts as completed deliveries. Reproduces and fixes the notification regression exposed by merge CI.
* test(agents): retain rejected submissions across retries
The existing browser retry test assumed terminally rejected prompts disappeared. Assert the intended prompt-before-error history and a separate accepted retry while preserving draft, attachment, and ambiguous-steering checks.
* test(creation): allow integration latency in raw RPC responses
The peer inherited a one-second polling timeout despite exercising Git and provider startup. Reproduce the timeout with delayed startup and wait within an explicit integration budget while retaining all response and identity assertions.
* feat(agents): show agent messages and completion notifications with sender identity
* fix(agents): preserve human envelope examples and attachment context
* fix(agents): separate message provenance from timeline presentation
* fix(server): resume with last live mode, not creation-time mode
Provider-initiated mode switches (ACP current_mode_update, in-session
commands like /bypass, permission-driven plan->acceptEdits transitions)
update currentModeId/lastModeId but never reach config.modeId. On
resume, buildConfigOverrides restored the stale creation-time mode and
applyConfiguredOverrides pushed it back via set_mode — silently
downgrading the session (e.g. bypass -> acceptEdits after restart).
- buildConfigOverrides + buildStoredAgentConfig prefer lastModeId
(the last observed live mode) over config.modeId
- applyObservedMode syncs config.modeId on every provider-reported
mode change (mode_changed, model_changed runtimeInfo, and the
refreshSessionState poll) so in-memory reloadAgentSession carries
the live mode as well
Generated with [Devin](https://devin.ai)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(server): assert config.modeId sync on mode_changed/model_changed events
Extends the config-drift stream test with a registry so both the
in-memory agent.config.modeId and the persisted record.config.modeId
are verified — covering the event paths Greptile flagged as untested.
Generated with [Devin](https://devin.ai)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(usage): stream reports for live agent accounts
Resolve account scope from the running harness environment without invoking lifecycle hooks. Keep the client aggregate result while delivering report updates as each fetch settles; the app presentation follows separately.
* fix(usage): retain account scope for open agent sessions
Capture the launch environment once per session, independently of lazy child-process startup or exit. Use a session UUID for registry identity. Update remaining global discovery fixtures to the scoped SDK contract.
* test(usage): update browser fixtures for scoped report streams
* feat(usage): let a usage card show window rows without pins
UsageCard takes a required pinnable prop. When false, a window row is a
plain bar: no Pressable, hover highlight, or pin glyph. The settings
Usage list passes true.
* feat(usage): show the agent's usage in the context window popover
Reports stream into the query cache as each update arrives, for the
host list and for a new per-agent query keyed by agent. The context
window popover fetches its agent's reports while open and renders one
unpinnable card per report as it lands, a loading sentence before the
first, the request error as a sentence, and nothing on hosts without
usage reports.
* feat(usage): hide the Refresh button on popover usage cards
UsageCard takes a required refreshable prop. The popover passes false
because its content takes no pointer events, so the button could not
be pressed; the card then shows its freshness inline. The settings
Usage list keeps the button.
* feat(usage): make the sidebar Usage summary opt-in
The sidebar Usage item starts hidden on every layout, not only on
compact ones, so resolveSidebarNavItems drops its compact input. While
the summary is off, the Usage list renders window rows without pins;
saved pins stay in preferences and return when it is turned on. Specs
that read the summary now turn it on first.
* docs(glossary): describe the popover usage card and the opt-in sidebar summary
* fix(usage): show a popover request error without retrying
A daemon error for an agent's usage, such as an unknown agent, does not
heal on retry, and reopening the popover fetches again. Retrying held
the loading sentence for about seven seconds before the error.
* test(usage): cover agent usage in the context window popover
The usage fixture records agentId, streams reports held behind
promises, and can switch host support between reloads. The meter spec
walks loading, streaming, a report problem, a request error and a host
without usage reports, on desktop and compact in both themes.
* fix(usage): keep the plain context window tooltip on hosts without usage
The popover widened to fit usage cards on every host. Hosts that cannot
report usage now keep the tooltip's default padding and width, as
before, and mount no usage section.
* fix(usage): let a report's Refresh replace it in the agent's list too
Refresh wrote the refreshed report only into the host's list, so a card
rendered from an agent's list kept the old report. Both lists now sit
under one host key and Refresh replaces the report in each that holds it.
* feat(usage): open the context window details in a sheet on compact screens
Wide screens keep the tooltip. On compact screens pressing the meter opens
a content-sized bottom sheet titled Context window. Both render one
ContextWindowDetails body; only the sheet's usage cards have Refresh,
since the tooltip cannot be pressed. Pins stay off in both.
* test(usage): cover the compact context window sheet
On the compact viewport the meter opens the sheet; its cards have Refresh,
a Refresh replaces the card, and Close dismisses the sheet. The glossary
names the sheet.
* feat(ui): give tooltips and the context window sheet accessible roles
Tooltip content is role=tooltip, and the context window sheet is a dialog
named Context window, so assistive tech and tests find them by role. The
agent usage message drops its test ID; it is found by its text.
* test(usage): reshape the context window and Usage switch specs as journeys
The context window spec runs one tooltip journey and one sheet journey per
theme, with no layout branches. Its fixture script, meter actions and
report builders move to a context-window helper, and it finds the meter,
tooltip, sheet and Refresh buttons by role. The theme comes from the
system colour scheme. The Usage switch test uses named helpers for the
Usage icon and the Summary in sidebar switch.
* fix(usage): say an agent's usage request failed instead of showing older reports
A failed request keeps whatever reports the query held: those from an
earlier open, or those that streamed in before the failure. The context
window details showed them as if they were the agent's current usage.
The request error now wins over held reports.
* test(usage): let a fixture clean up the context window spec's agent
The tooltip and sheet tests no longer wrap their steps in try/finally;
a spec-local fixture seeds the agent and removes it after the test.
* fix(usage): keep a refreshed report when the list it is in finishes
Pressing Refresh on one card while another source was still streaming
put the pre-refresh report back: the finished list replaced the query's
data with the copies the request had fetched when it started. A report
now keeps whichever copy was fetched later, both when a streamed copy
arrives and when the finished list replaces the streamed one.
* fix(usage): show only the login a resumed agent runs under now
An agent resumed under another login keeps its ID, so its usage query
kept the previous login's reports: reopening the context window showed
them while the new login's reports loaded, then both side by side until
the request finished. The agent's usage is now dropped as soon as its
tooltip or sheet closes, so each open loads the login the agent runs
under at that moment.
* fix(usage): stop a closed agent usage request from writing reports
Closing the context window details cancels the agent's usage request and
drops reports still on their way, so reopening after a resume under
another login sends a new request instead of joining the old one.
Resolve retained variants against the destination model before applying a model change. Publish the effective thinking choice and refresh the provider resume handle immediately so a reload cannot restore the old selection.
* fix(opencode): support v2 while preserving v1 provider behavior
* fix(opencode): target v2.0.7, keeping v1 support
OpenCode 2.0.4 removed the /api/health endpoint and the plugin
await-activation endpoint, and moved form reply/cancel under the session
namespace. The v2 adapter was pinned to the 2.0.3 SDK, so any 2.0.4+
binary failed at helper-server startup with an opaque HTTP 404.
- Pin @opencode/client to 2.0.7 and require v2 >= 2.0.4 at runtime
selection, with a clear error for older binaries.
- Use server.info for helper-server readiness.
- Move form reply/cancel/list to session.form and send the permission
decision field instead of reply.
- Replace plugin.awaitActivation with a plugin.list readiness wait. A cold
location registers its config-derived commands, skills, and providers
asynchronously; an empty inventory means it is still warming.
- Send session.command.name, which replaced command, and drop the removed
SkillInfo.slash field.
- Update the v2 test harness, unit tests, dependency lockfile, CI matrix,
and provider docs to 2.0.7.
* fix(opencode): target 2.0.10 and address v2 review findings
* test(opencode): colocate v2 suites with their modules
Greptile's review flagged that the v2 tests lived in top-level
opencode-agent.*.test.ts files and imported the readiness helper, v2
client, harness, and API types from a distance. Move each v2 suite beside
the module it describes and leave the top-level files to the v1 public
path:
- opencode/v2/agent.test.ts: session lifecycle, structured output,
resume, reconciliation, interrupt/replacement, helper isolation.
- opencode/v2/readiness.test.ts: location readiness deadline behavior.
- opencode/v2/permissions.test.ts: question and approval routing.
- opencode/v2/configuration.test.ts: permissionRules, feature flags, and
helper-isolation decisions, exercised directly through the module.
- opencode/v2/mapping.test.ts: model and mode normalization.
- opencode/runtime-client.test.ts: runtime version selection.
The v1 recovery, permission-action, and catalog-deadline suites stay in
place. Test counts are preserved, with deeper per-module coverage for
configuration.
* test(opencode): split v2 local E2E by behavior and add harness helpers
The re-review flagged the v2 scenarios in opencode-bridge.local.e2e.test.ts
for inline event filtering, permission-kind branching, and polling in test
bodies. Move stream, permission, and subagent coordination behind helpers
in opencode/test-utils/v2-local-e2e-helpers.ts, and split the session
scenarios into opencode-v2.local.e2e.test.ts so each file covers one
behavior:
- opencode-bridge.local.e2e.test.ts: process sharing, shell.env scoping,
the native tool bridge, and runtime version selection.
- opencode-v2.local.e2e.test.ts: session lifecycle, structured output,
streaming, approvals, MCP preapproval, stop/replacement, rewind and
subagents, commands, and skills/attachments.
Each helper waits on or answers the provider directly, so the tests state
domain intent and assert the collected result instead of polling or
branching inline. The CI-selected "versioned runtime" case stays in the
bridge file, and the live approvals and stop/replacement scenarios were
re-run green through the new helpers.
* fix(ci): preserve required server checks across OpenCode versions
Run v1 and v2 integration coverage sequentially so branch protection keeps its existing static check names.
* fix(agents): load startup history before registering sessions
A failed history stream while adding a runtime notice must close the unregistered session instead of leaving an initializing agent behind.
* fix(agents): limit startup history before buffering it
Apply the existing timeline content limits to each incoming history event so raw cumulative tool output is not retained until registration completes.
* fix(opencode): resolve Windows executables from platform packages
Recent v1 npm releases place opencode.exe in the architecture dependency rather than the CLI bin directory. Resolve that dependency from the CLI package while retaining support for the older layout.
* refactor(opencode): name public runtime option types
* test(opencode): let the spec own both runtime versions
Install each version in an isolated temporary directory and select it through provider configuration. Restore the CI workflow unchanged so version coverage is owned by the spec rather than workflow orchestration.
* test(opencode): launch native Windows binaries in isolated specs
Avoid npm command wrappers retaining locked executable files after helper shutdown during temporary installation cleanup.
* fix(opencode): preserve the legacy path when adding v2 support
Fall back to the existing client when version probing is inconclusive. Load v2 code and its bridge plugin only after positive v2 selection, keep v1 sessions undecorated, and restore original v1 discovery errors.
* fix(opencode): retry inconclusive version selection
A failed or unrecognized version probe previously cached the legacy fallback, pinning an upgraded binary to the v1 API until client recreation. Retry detection on the next operation while keeping the legacy path available.
* fix(opencode): release legacy runtime after inconclusive probe
Retryable version detection can clear the selected client after falling back to v1. Track whether the legacy client was selected so shutdown still releases its server without touching an unused v1 manager during a v2-only run.
* test(opencode): assert full runtime notice history
* fix(opencode): restore notice and retry behavior after rebase
---------
Co-authored-by: Jonathan karras <6307688+karrots@users.noreply.github.com>
Co-authored-by: Dhruv Singal <dhruvsingalabc@gmail.com>
* Clear the committed timeline when an agent refresh rehydrates from disk
Refreshing a loaded agent reloads it with `rehydrateFromDisk`, which wipes
the in-memory timeline so provider history can be re-streamed. It left the
committed timeline in place, and that is what `getTimelineRows` reads, so
every refresh appended a second full copy of the history: N refreshes, N
copies of every message.
Wipe both timelines on that path, the way the force-hydrate path already
does, so the re-streamed history replaces the old rows instead of stacking
on them.
* Replace the committed timeline from the replay instead of wiping it first
Deleting the committed rows on the reload path meant a history replay that
failed halfway left the agent with no transcript, and a retry appended a
fresh replay to whatever partial rows it had written.
Collect the whole replay before touching either store, then drop the rows a
previous hydration committed and write the new ones. A failed stream now
leaves the committed timeline exactly as it was, and this covers the replay
that runs when an agent is loaded after a daemon restart, which duplicated
the same way a refresh did.
---------
Co-authored-by: paseo-bot[bot] <266920839+paseo-bot[bot]@users.noreply.github.com>
Derive Pi thinking options from each model's thinkingLevelMap, read Pi's effective level back after create, resume, set_model and set_thinking_level, surface failed state reads instead of persisting the requested level, and show the thinking control for single-level models.
Closes#4382
Co-authored-by: therainisme <therainisme@qq.com>
Co-authored-by: trim21 <i@trim21.me>
* fix(agents): prevent OOM from cumulative tool history
Repeated progress updates retained complete historical copies of growing tool output. Retain projected items on ingestion and serve them for every history fetch, preserving source sequence coverage for catch-up. Keep intermediate payloads only in streaming; confine compatibility degradation to child transcripts and overwritten fork checkpoints.
* test(client): expect projected subagent capability in hello
* fix(app): report unexpected child history failures
Keep connection failures recoverable while forwarding unexpected fetch and reconciliation errors to diagnostics. Name observer arguments to keep server and agent identities explicit.
* test(agents): assert projected history after rewind
* docs(protocol): tag projected child compatibility gates
* fix(agents): restore archived Codex workspaces without losing focus
History selection must survive without cached agent records, and reading an archived Codex thread must not acquire an interactive writer. Keep workspace restore explicit, serialize native lifecycle transitions, and leave transcript lifetime with its owner.
* test(app): express Codex restore specs as user journeys
* test(app): finish restore fixture cleanup after teardown errors
* test(app): align restore coverage and await fixture readiness
Workspace restore leaves agents archived until explicit Unarchive. Keep that distinction covered for multiple agents. Compile the recovery fixture before its interaction test starts, and wait for sidebar agent hydration before asserting tab presence.
Restore unread attention on finished workspaces and preserve the reminder when leaving the current workspace. Cover desktop and compact read flows, and retain change-only file paths for deletion observation.
Co-authored-by: Edi Hasaj <34984294+edihasaj@users.noreply.github.com>
* fix(server): support OpenCode 2 in terminal activity plugin
OpenCode 2 rejects V1 hook-object plugins, so the installed
paseo-terminal-activity.js failed to load on every OpenCode 2
session. Ship one definition object that serves both generations:
OpenCode 2 loads id + setup() and ignores the 1.x server entrypoint,
OpenCode 1 loads the same object through server().
* fix(server): preserve terminal activity across plugin generations
* test(server): drain persistence before reload test cleanup
---------
Co-authored-by: Mohamed Boudra <boudra.moha@gmail.com>
* fix(server): keep agent reloads from overlapping session writers
Idle runtimes can still own their persisted session writer. Release the original runtime before resuming and retain failed cleanup for retry; a timeout must not authorize another writer.
* fix(server): honor closes queued during agent reload
Serialize public close with reload so a completed close cannot be undone by replacement registration. Keep archive cleanup inside its existing lifecycle operation.
* fix(server): filter empty ACP import sessions
ACP metadata probes can leave persisted sessions that later appear in import results. Replay available session history to exclude truly empty entries while retaining sessions whose history is unavailable, and close probe sessions before process teardown.
* fix(server): bound ACP session history loading
Keep import enrichment inside the provider listing deadline when session histories are inaccessible. Catalog abort cleanup now also waits briefly for an in-flight session creation so the resulting probe session can be closed.
* test(server): avoid wall-clock ACP assertion
Verify the history budget through the number of attempted protocol loads instead of elapsed process time, so scheduler delays cannot make the regression flaky.
* fix(server): allow slower ACP history loads
Give native session history more time to replay while retaining a shared enrichment budget below the provider listing deadline.
* fix(server): give ACP history replay enough time
Use the provider RPC timeout scale for native history loading, with a larger shared budget and provider-list deadline. Keep timeout coverage fast and deterministic with a controlled clock.
* fix(server): isolate generous ACP import deadlines
Keep the long listing deadline for the picker’s single-provider requests while preserving the short aggregate deadline for callers that fan out across providers. Restore deterministic coverage of unresolved history loads through fake timers and an injected clock boundary.
* fix(server): preserve ACP aggregate listings
Add backend-only exact-provider controls for the Paseo tool catalog across native and MCP-backed agents, including persisted configuration, session snapshots, public documentation, and regression coverage.\n\nCo-authored-by: Matt Cowger <matt@cowger.us>
* feat(app): make import session a first-class entry point
* feat(server): improve import session discovery resilience
Search up to 500 persisted candidates per provider and isolate listing failures behind eight-second deadlines so healthy results remain usable. Preserve native persisted title precedence while keeping discovery read-only.
Parsing now continues through each transcript so native titles and the true last prompt are retained; the old early exit made lastPromptPreview equal firstPromptPreview. A cheap record prefilter avoids JSON.parse for unrelated lines during the full scan.
* feat(app): searchable, grouped import session sheet with per-provider errors
The sheet now carries a debounced search field gated on the host's
importSessionSearch feature, grows its per-provider page on demand
(15 -> 45 -> 90 -> 200) instead of capping at the 15 most recent, and groups
unscoped rows by folder under the project name the app knows for that
directory. Worktrees of one project would otherwise all head up as the same
bare name, so a directory below its project root carries the path under that
root beside the name.
A workspace-scoped sheet says so and offers "Show all sessions". Importing a
row from another directory once that is on drops the workspace from the
request, which the daemon would otherwise reject, and routes the user to the
agent's own workspace rather than opening a dead tab in the current one. That
decision lives in resolveImportTarget: a scoped listing is trusted wholesale
because the daemon already matched it realpath-aware, which the client cannot.
A provider that fails now gets its own error row with a Retry that refetches
only that provider, sourced from both a failed request and the daemon's
providerErrors. React Query no longer retries these, so the header spinner
settles once every provider has answered or given up (#2512).
* test(e2e): import session flow with fixture Claude sessions
* fix(app): shorten import sheet scope copy so it fits at phone width
At 390px the scoped subtitle rendered "Sessions in this wor..." — the label,
its action, and the sheet's refresh and close buttons do not share one line at
phone width. The label already sits under the sheet title "Import session", so
"Sessions in this" was repeating context the header had: it becomes "This
workspace" and the action becomes "Show all". The unscoped label still names
the host, which nothing else on the sheet does.
* refactor(app): flat recency-ordered import list with per-row folder label
Grouping by folder buried recency: a busy directory's block of fifteen pushed
a two-minute-old session in a quiet directory below the fold, and the axis a
user is scanning on is when they last touched the session, not where it lives.
The unscoped sheet is now one list in the order aggregateSessionEntries already
produces, and each row carries its folder as a muted third line.
resolveDirectoryLabel stays — it is what keeps two worktrees of one project
apart — and formatDirectoryLabel renders its two parts as the single ellipsised
line a row has room for. A scoped sheet lists one directory, so its rows carry
no folder line at all.
* fix(app): close import sheet before refreshing sessions
Provider-list invalidation waited for every active query before the sheet could close. Mark the disabled picker queries stale after close and navigation so reopening refreshes without delaying the imported agent.
* fix(server): reuse workspaces for session imports
Host-wide imports should join the oldest active workspace for their directory, or restore its archived workspace, instead of minting duplicates. Rollback remains limited to workspace records created by the failed import.
* test(e2e): express the import session journey through a helper DSL
* test(e2e): align sidebar journeys with import entry point
Track each foreground run start as pending, started, or failed so a previous turn error cannot reject a newly accepted prompt. Includes a regression for resuming after a failed turn.
* Stream assistant text at a steady rate instead of in arrival lumps
Assistant text arrived in lumps and was painted that way. The daemon
coalesces one message per 60ms window, and that window carries however
many characters the model produced in it, which swings by an order of
magnitude within a turn. Painting each delta as it landed made the size
of those lumps visible.
Three changes:
- The agent stream coalescer is now leading + trailing. The first delta
after an idle window flushes synchronously instead of waiting a full
window, matching the invariant TerminalOutputCoalescer already
documents. The window stays at 60ms; shrinking it would raise message
rate on a loop that already contends with terminal frames and relay
encryption, without making the stream smoother.
- The client reducer queue commits on a frame instead of an arbitrary
48ms timer, with the timer kept as a ceiling so a hidden tab still
advances.
- Assistant and reasoning text are revealed at a rate derived from the
backlog rather than from arrival timing. Arrival changes the target,
not the rate. The store keeps the full text and only the rendered
slice is paced, so copy, selection, the chat outline, and scroll
geometry all stay consistent with what is on screen.
The leading flush means a burst's first chunk lands as its own timeline
row. The projection already merges contiguous same-turn rows and clients
read the projected timeline, so history is unchanged; the server tests
that assert on raw rows were updated for the extra row.
Measured on Expo web against a local dev daemon. Mock bursty-stream
provider, 6s samples, paint-on-arrival vs paced: frames that advanced
the text 55 -> 244, chars-per-frame CV 2.66 -> 1.44, gap p50 100ms ->
17ms. Real Claude Haiku agent over 52s: 84% of frames advance the text,
CV 1.29, gap p50 and p95 both 17ms (one frame at 60Hz).
Adds a bursty-stream mock model with seeded burst sizes, a gated browser
perf spec, and docs/agent-stream-performance.md.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* Address review: safe reveal boundaries and a testable seam
Two findings from the PR review.
Unicode boundaries. Slicing the revealed text at an arbitrary index could
strand a lone surrogate, which paints as a replacement glyph, or split an
emoji ZWJ sequence so a family briefly rendered as its members. Both were
visible for a frame or two mid-stream — the exact flicker the paced reveal
exists to remove. The rendered cut is now pulled back off surrogate pairs,
combining marks, variation selectors, skin tone modifiers, and ZWJ joins.
The reveal counter stays monotonic and only the painted boundary moves, so
a long cluster cannot stall the reveal.
Test placement. The hook test used JSDOM, @testing-library renderHook, and
monkey-patched requestAnimationFrame, all of which docs/testing.md rules
out; it prescribes fixing the missing seam instead. The reveal policy is
now a pure state machine in text-reveal.ts (begin/retarget/advance/
complete/visible) covered by plain unit tests with no environment, and the
hook is reduced to the frame-clock binding. Rendered behavior is left to
the browser spec, which no longer carries frame sampling and statistics in
its body — those moved to e2e/support/helpers/stream-smoothness.ts.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* Measure total painted text, not the tail message
The smoothness sampler read the length of the last assistant-message
element. A real turn emits many assistant messages, so that element keeps
changing identity and its length is not monotonic — handovers read as
resets, and a stream that was plainly moving on screen reported almost no
growth at all. Sampling the total across every assistant message fixes it.
This invalidated the numbers previously recorded for the real-agent run,
so the doc table is replaced with a fresh A/B taken with the corrected
sampler against a Claude Haiku agent: frames that advanced the text 6% ->
87%, chars-per-frame CV 4.11 -> 1.86, gap p50 317ms -> 17ms, p95 383ms ->
17ms. The paced column also reported zero Unicode defects across a stream
containing emoji, which exercises the reveal boundary clamp end to end.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* perf(agent-stream): harden paced text reveal
Use standards-compliant grapheme segmentation when available and preserve full arrivals on older runtimes. Cap reveal commits at 60Hz and measure the real browser rendering path so high-refresh displays stay smooth without rendering at hardware frame rate.
* Hold back trailing fragments split by arrival boundaries
The reveal boundary is not the only place a grapheme gets cut. The
daemon's coalescing window ends a delta wherever the window closes, so a
caught-up reveal can be handed half a cluster and paint it. Measured on a
mixed-script Haiku stream: five frames in ~350 ending in a single regional
indicator, i.e. half a flag on screen for a frame.
While the turn is streaming, a trailing fragment that cannot stand alone —
a dangling high surrogate, a trailing joiner, an odd regional indicator —
now waits for the rest of its cluster. Once the turn ends everything
received is painted, because nothing more is coming. Complements the
grapheme-boundary clamp, which handles the reveal cut; this handles the
arrival cut.
The same stream now measures one such frame in ~400 rather than five. The
remainder is a flush boundary that coincides with the end of an assistant
message, putting the two halves in different messages, which the client
cannot rejoin — that one belongs in the coalescer.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Co-authored-by: Mohamed Boudra <boudra.moha@gmail.com>
* fix(server): recover already-settled interrupts
Provider state can drift from Paseo’s tracked run. Treat an authoritative already-idle interrupt as settled and retire the captured run token so retrying can start new work without weakening ambiguous-cancellation protection.
* fix(codex): fence interruption during turn startup
Reserve foreground-start ownership before asynchronous provider setup. An interrupt can now cancel preflight before turn/start or wait for an issued start to become interruptible, preventing acknowledged cancellation from leaving unmanaged work.
The filesystem timeline store rewrote each agent's full transcript on every buffered update, causing write amplification that grew with conversation length. Return production timelines to runtime memory and provider-history hydration while preserving the dormant storage interface for a future intentional design.
A running autonomous turn was excluded from steer admission because only foreground ownership was considered. That silently degraded steering into interruption and could race a provider-initiated restart, dropping the prompt.\n\nUse the provider's canonical active-turn identity for admission so autonomous work receives steering on its existing stream without replacement.
* feat(agents): add active-turn steering
Let users choose whether an immediate prompt interrupts or joins the active turn. The daemon owns exact-turn admission, provider delivery, safe fallback, and canonical timeline identity while queueing remains client-side.
* fix(agents): stabilize steering history and replay QA
* fix(agents): preserve steering timeline order
* fix(agents): reconcile resumed provider history
* fix(agents): serialize steering with cancellation
* test(agents): prove Claude multi-tool steering
* test(agents): force explicit history hydration
* fix(agents): flush output before steer admission
* fix(subagents): preserve opened children on parent archive
Opening a managed child represents user ownership only while that client's tab remains open. Track that intent per client, detach surviving children during parent archive, and keep single and bulk tab-close behavior consistent.
* fix(subagents): mark generic tab openings
Centralize managed child tab ownership in agent navigation so list, command-center, deep-link, notification, and subagent-track openings all mark ownership before creating the tab. Use stable localized errors for lifecycle failures.
* fix(subagents): classify cold tab openings
Fetch uncached agents before deciding whether tab ownership must be recorded, so notification, desktop, and cold-route openings cannot bypass parent archive protection.
* fix(app): translate navigation fallback errors
* fix(navigation): allow offline workspace intents
* refactor(app): derive subagent ownership from open tabs
* fix(server): serialize subagent lifecycle ownership
* feat(providers): add native options and exact MCP grants
Keep provider policy in provider-owned schemas so unattended Hub runs can grant only requested tools from injected MCP servers. Preserve the native resolved sandbox policy when a workflow override is present, including configured writable roots.
* test(hub): verify real provider tool policy
* test(app): avoid pinning transient restart status
* feat(app): launch a terminal from the New workspace screen
New workspace only ever started a chat agent. Empty workspaces existed but
were reachable only by submitting an empty composer, which nothing advertised,
and terminals could not be started from this screen at all. People concluded
Paseo did not support terminals.
Adds a launch chip to the meta row, shaped from the same ComboboxTrigger as
its neighbours, that chooses between Chat and a terminal. Terminal profiles
mark where a typed prompt goes with a {{{prompt}}} sentinel in their command
or args, substituted client-side before create_terminal_request, so the daemon
learns nothing new and the protocol is untouched. A profile without the
sentinel takes no input: the composer goes read-only and shows the exact
command, so there is nothing to type wrong.
Prompt forms come from each CLI's own --help. claude, codex, and pi take a
trailing positional; opencode's positional is the project directory, so it
uses --prompt=. Profiles predating the sentinel adopt the right form on read,
keyed on the command's base name rather than the profile id, because profiles
created in settings get generated ids.
* fix(app): stop existing launchers passing the prompt sentinel literally
Adoption made every caller of resolveTerminalProfiles see sentinel-bearing
args, but only New workspace substituted them. Pinned targets and the
workspace terminal menu forwarded profile.args straight through, so launching
Claude from an existing workspace ran `claude "{{{prompt}}}"` with that text as
the prompt.
Every launcher now goes through resolveTerminalProfileLaunch. Surfaces with
nowhere to type pass an empty prompt, which drops the prompt-only arg and
gives back the bare command they had before.
Also from the same review:
- Adoption skipped when args lead with a subcommand. `opencode run` takes a
positional message rather than --prompt, and `codex login` takes no prompt at
all. Only the first arg is checked, so `pi --thinking high` still adopts.
- Settings edits the persisted list again, not the adopted view. It writes the
whole list back on any change, so resolving first baked read-time adoption
into the user's config the moment they reordered a row.
- A hidden draft no longer names a workspace. Text typed for a prompt-taking
profile stayed in state after switching to a read-only one and still seeded
the generated title and branch.
- Windows .cmd escaping now has a round-trip test: a real cmd.exe runs a real
.cmd shim that echoes its argv, covering metacharacters, an injection
attempt, and newlines. The existing tests only asserted the command line we
generate.
- Fixed an ENOTEMPTY flake in the terminal suite. kill() only signals, so the
shell was still writing into $ZDOTDIR while its directory was being removed.
Cleanup waits for exit instead.
* fix(app): keep the host and isolation accessible names stable
Wiring the new tooltips also replaced these two accessible names with the
tooltip copy, which broke `sidebar-project-grouping` at the host picker: its
selector is `getByRole("button", { name: "Host", exact: true })` and the button
had become "Choose the host". It timed out on both attempts in CI.
An accessible name identifies a control; a tooltip explains it. Conflating them
renamed a control that other surfaces address by name. The names go back to
what they were and the tooltips keep the "Choose the ..." copy.
* test(server): stop temp-dir teardown failing unrelated terminal tests
The ENOTEMPTY flake survived both earlier attempts. Waiting for the session to
exit shrank the window; retrying the remove did not close it either. The writes
come from processes zsh forks on the way out, which outlive the pty and answer
to nothing the test holds.
Teardown is now best-effort and logs what it left behind. A temp directory we
cannot remove is not a defect in the terminal code under test, and failing on it
reports the race against whichever assertion ran last, which is exactly how this
surfaced: as a flake in an unrelated zsh title test. 10 consecutive runs green.
Also replaces a cast in the Windows .cmd helper with a real narrow, so a change
to ResolvedTerminalCommand fails loudly there. Test files are excluded from
typecheck (tsconfig.server.json), so a cast in one is never checked at all.
* fix(app): keep terminal launch policy in its owning workflows
Profile adoption now parses known global option values before deciding whether argv already contains a positional. Workspace naming is derived inside the terminal workflow so hidden launch-only drafts cannot affect branch or title generation.
* fix(server): preserve cmd shim arguments through both parses
npm command shims forward %*, so metacharacters lose one escape layer before the target process receives argv. Exercise the production node-pty string command line against a forwarding shim, and only tolerate the known ENOTEMPTY teardown race.
* fix(server): stabilize Windows subprocess tests
* test(server): isolate Windows argv recorder output
* test(server): allow native watcher integration to settle
paseo-bot[bot]