A recognizer that loaded before the Silero VAD file existed kept cutting text
only at stop until its next restart. When the VAD lands, the speech service now
unloads that recognizer once no dictation uses it and warms it again, so pause
cuts start on first use. A running dictation is never interrupted.
The adapter crossed the 300-line cap after #40069 and main's lint fails. The same receipt check ran twice in it; one shared predicate brings the file back under the cap with no behaviour change.
* Fork Codex replies from the current history after rewind
* Strengthen Codex fork preparation failure coverage
* Preserve Codex fork compatibility and original chat guidance
* Explain when Codex cannot start for a fork
The GPU stall watch gave the whole model load one 25 s budget, so a cold disk
or busy machine restarted a healthy load on the CPU and could miss the 60 s
start limit. The speech process now reports each load step, and the watch
restarts on it; a step that hangs for 25 s still moves to the CPU.
* fix(agent-status): show Codex's mid-turn async question as waiting for the user
* fix(agent-status): keep Codex's async question waiting on paired hosts under its working title
The phone's PTY-title renewal turned the waiting row back into working, because Codex keeps its working title while the question waits.
* fix(agent-status): clear Codex's open async questions when the user sends a plain message
Orca's chat card answers an async question as a quoted message, not Codex's reply wrapper, so the pane stayed waiting until the turn ended.
* refactor(agent-status): move the legacy status test seed out of listener-state
* Reuse Claude CLI titles for default chat naming
* Preserve independent naming while fencing Claude title reuse
* Keep optional title account lookup from blocking chat naming
* Clarify deferred preparation in naming tests
* Bind first chat naming to its ready handover
* Exercise title cleanup through a real queued command
* Use the timestamped operation fixture for compaction
* test: exercise naming fallback through Send now
* test: exercise terminal naming fallback after final starts
* Keep Claude background tasks linked to their subagent across restarts
* Keep task ownership scoped to each Claude invocation
* Keep partial task events scoped to their invocation
* Distinguish pending task facts from admitted invocations
* Wait for saved task ownership before interpreting Claude events
* Verify journal binding order and saved task liveness
* Keep owned task updates from reopening Claude root turns
* Exercise owned historical notifications with provider identity
* Show AskUserQuestion option previews on desktop and phone
Claude can attach a preview (a mockup or snippet) to each option of a
single-select question. Orca dropped it, so users chose between bare labels.
Options now carry the preview through the journal, terminal asks and the
phone, and the question card shows the selected option's preview as plain
monospace text. Questions with previews confirm explicitly instead of
auto-advancing, so the preview stays readable.
* Keep phone previews bounded and digit focus on the named option
Phone previews get a capped height with their own vertical scroll, and
opening one never scrolls past its top. A digit always focuses its row
before the chosen-row check, so Enter confirms the option the user named.
Tighter preview line height, a stable phone row root, and a persistence
test whose expected previews come from the fixed input.
* Narrow the reopened question with a type guard instead of a cast
* perf(native-chat): project structured chat tabs per workspace bucket
The shared chat-tab list was keyed on the whole unifiedTabsByWorktree
identity, so every tab write anywhere (terminal title batches included)
rescanned all tabs and handed every consumer a new list. Rebuild only the
buckets whose identity changed and return the previous list while no chat
tab changed.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* perf(native-chat): reconcile chat status membership per changed tab
The status bridge rebuilt every member and host group and captured every
chat on any tab-map or owner-input change, and recaptured every chat on any
launch notification. Membership now updates only the tabs whose identity
changed; owner-input changes re-resolve only unstamped tabs, the only ones
whose host depends on them; launch notifications name their session so only
its tabs are recaptured (unaddressed ones still recapture all). The sleep
and launch-directory companions of one projection land in one store write.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(ratchets): guard whole tab-map triggers and hot-map listeners
Add a count ratchet for identity comparisons of the whole
unifiedTabsByWorktree/tabsByWorktree map in renderer code (counts may only
fall), wired into pnpm lint, run-ratchets discovery and changed-check
selection. Extend the advisory no-whole-hot-map-selector rule to inline
store listeners that compare a whole hot map and to effect bodies that use
one whole from getState().
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* revert(lint): drop the hot-map listener/effect extension
The guard stays scoped to the repaired seams; no repository-wide rule change.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(native-chat): capture chat status on semantic tab changes; address tombstone retirement
Focus, preview and order-only tab writes no longer recapture a chat's
status: membership and the projection fences compare the fields the
projection reads. Absent-tombstone retirement now names the sessions it
retired, so no launch producer sends an unaddressed notification.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(store): keyed-record transition seam for tab maps and companion records
Writers build next tab buckets and keyed companion records through
replaceTabBuckets / replaceRecordEntries (or name the keys of a record they
already built), which record exactly which keys and tab IDs changed against
the produced record. Readers ask readTabMapChanges / readRecordChanges for
an addressed delta; any unannotated link reads as a full resync. Adds
subscribeTabTransitions and a store-observer installer registry. The
chat-tab list and the status bridge now consume addressed deltas and keep
the whole-list walk only as the cold path.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(store): annotate every shared tab-map writer through the seam
Shared tabsByWorktree/unifiedTabsByWorktree writers (tab, editor, browser,
worktree, ssh, terminal create/close/navigation/title-batch/reconnect/
hydration, web-session mirroring and layout frames) now build or annotate
their next tab buckets and companion records with the exact workspace and
tab IDs they changed; full hydration, full frame builds and reconnect
inventories are named full resyncs. Writers that copied every bucket for
one tab (preview promotion, dirty-file preview clear, worktree removal
filters, sleeping-record identity rename) now keep unchanged buckets.
Replace the renderer-wide tab-map comparison ratchet with a narrow seam
guard: the repaired consumers may walk a whole tab map only in named cold
paths, and the pinned writer modules may not build raw seam records.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(store): name seam fields once; type seam records; record moved respawn writes
Writers stop repeating tracked field names (centralized in
seam-record-fields.ts), the seam takes a named KeyedSnapshot record type
instead of casts, and the layout-writer baseline records the respawn
helpers moved unchanged out of worktree-layout-lifecycle.ts.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(store): drop seam field cast; record the seam field list in the layout-writer baseline
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(native-chat): keep chat-tab list in source workspace order (C3/F1)
Every source workspace keeps a projection slot, even without chats, and the
seam reports when a write added or deleted a workspace key or moved a
surviving bucket, so the incremental list never reorders first-match
consumers; those transitions take the cold rebuild.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(native-chat): keep projection and retirement debt awake (C2, F3)
A member is owed from attach or semantic change, members without a host
feed keep their own debt, and every reconciliation drains existing owed or
retired work even when no new member changed, including a closed last chat
whose feed is gone.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(store): stop inferring host-mirror companion IDs from visible tabs (C1)
A host retraction can delete PTY or layout records for tabs neither
inventory lists, so the mirror funnel marks its tab-keyed records full
resync instead of advertising an incomplete addressed set.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(ratchets): close indirect paths in the tab-map seam guard (C4)
The guard now rejects ordinary callers of the bridge cold path, value
imports the repaired consumers have not reviewed, and raw seam records built
into a local binding without a seam helper annotating it.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test(native-chat): drive real tab actions through the status bridge (F2)
Close, session rebind with a late old-session tick, and a legacy owner
change, through createUnifiedTab/setTabLabel/closeUnifiedTab, so the
addressed reconciliation path is exercised.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore: exempt seam field-name constants from the layout-writer ratchet; drop dead setters (F4)
The seam's SEAM_*_FIELDS lists name fields without writing them, so the
ratchet skips them and the raised baseline row is gone (renderer total 569,
main 571). setStructuredChatSleptAt and the two launch-directory setters
lost their last callers to the batched companion write.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(store): Map provenance and source-ordered change keys in the seam (PR A addenda)
replaceMapEntries gives Map-typed fields (runtimeStatusByEnvironmentId) the
same weak provenance and readRecordChanges protocol. Change keys compose
forward through a chain, oldest write first, so added keys come out in
source order (index keys first for records, as Object.keys orders them). A
surviving key deleted and re-added, in one write or across a chain, and a
record builder that moves a surviving key, read as a 'reordered' full
resync.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(store): writers declare key order; seam reports ordered additions (N1c, N1a)
The seam no longer scans both records to rediscover order on every
annotation. replace* helpers know insertion order by construction; annotate*
takes a declared order ('appended' by default, 'reordered' for a builder
that moves keys, read as a full resync), verified only under test.
Addressed changes now report keys, source-ordered added keys and deleted
keys, so a reader holding its own key ranks places changes in O(changed).
Worktree removal declares its editor-transfer records reordered.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* perf(native-chat): scope the chat-tab projection to chat-bearing workspaces (N2c)
The projection keeps chat subsets only for workspaces that have chats, plus
per-workspace source ranks that change only when a workspace is added or
deleted. A focus, label or title write costs O(that workspace); the
ordered list is materialized lazily, only for full-list readers, and keeps
its identity while no chat changed. Workspace adds and deletes are
addressed through the ranks instead of a cold rebuild.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(ratchets): resolve aliases and require annotation before publication (C4)
Any reference to a cold helper or Object enumerator outside its named
callers counts, including aliases and callbacks; writer values are followed
through local aliases and must be annotated before the write that
publishes them.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test(native-chat): scoped hydration keeps the warm chat list equal to a cold read (D1)
Drives the real scoped hydrateTabsSession on early/later chats of one
session: source order, list order and first-match precedence match a cold
rebuild, including when the selection order differs from hydrated order.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* perf(native-chat): cache the chat-workspace order across chat edits (D1 item 5)
The lineage keeps chat-bearing workspaces in rank order and only inserts or
removes one when it gains its first or loses its last chat; a chat object
change re-flattens the list without re-sorting. Full resyncs rebuild it.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(ratchets): the seam guard recognizes only construction helpers and cold marks (D1)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* refactor(store): addressed provenance only from constructing helpers (D1)
Remove annotateRecordReplacement/annotateTabBucketReplacement and their
caller-declared order: only replace*/recordDraft/tabBucketDraft, which
perform the construction, derive changed keys, additions and deletions from
their operations (delete-then-reinsert reads as reordered); opaque
reconstructions use markFullResync. Every former annotate caller is
migrated: ordinary local writes build their records once through helpers
or drafts with today's operations and order; scoped tab and terminal
hydration, the reconnect inventory, host-mirror application and layout
flush are full resyncs. Change readers move to keyed-record-change-reads.ts.
Renderer layout-writer total 567 (main 571).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* test(native-chat): type the hydration fixture without a cast
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(store): seam record writes define own entries, even __proto__ (N5c)
replaceRecordEntries and drafts assigned through the inherited __proto__
setter, so a tab id '__proto__' lost its own PTY, layout and startup
entries. Both now define an own data property, as the computed-key literals
they replaced did; contract test included.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* perf(terminal): build each createTab record once (N4c)
createTab took ensureGroup's and insertSplitGroup's whole records and the
orphan cleanup's seam records, then rebuilt them through the seam. It now
takes the per-worktree group and split results and gives the orphan
cleanup only the bucket it reads, so every seam record is constructed once
with its ordered operations.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(ratchets): record createTab's orphan-cleanup input view (renderer total 570, main 571)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* refactor(terminal): split orphan cleanup so createTab cleans only companion caches (N6c)
buildOrphanTerminalCompanionCleanupPatch holds the per-tab cache, tab-bar
order and selection cleanup; buildOrphanTerminalCleanupPatch adds the seam
record deletes on top, unchanged for reconciliation. createTab calls only
the companion part with real state and deletes the seam entries through its
single construction, so the stub input view and its layout-writer baseline
rise are gone (terminal-tab-creation.ts back to 10).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* perf(terminal): filter the orphan cleanup's kept bucket once (N7c)
The full cleanup filters the workspace's tabs once and hands the kept
bucket to the companion part, which takes its replacement active tab from
it; createTab's companion-only call still makes its single find pass.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* fix(tabs): keep removed worktrees' group entries deleted on removal
The second emptied-group collapse in removeWorktreeLayout also visited
removed worktrees and wrote their deleted groupsByWorktree entry back as
an empty list, a delete-then-set the seam read as a reordered full resync.
It now visits only surviving workspaces, so removal stays addressed.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(tabs): closeEmptyGroup collapses through the shared builder again
collapseGroupLayout takes the record builder, and the renderer passes the
seam's replaceRecordEntries, so the per-worktree collapse publishes with
provenance without new counted layout writes (tabs-group-actions back to 12).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(tabs): reconciliation names its draft fields once (2 + 4 <= 9)
The written-field sets come from SEAM_ field-name constants instead of
string literals, so the reconciliation split no longer raises the combined
layout-writer count (tabs-reconciliation 2, tabs-reconciliation-batch 4,
against the original 9).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* chore(tabs): publish reconciliation drafts with explicit field writes (N9c)
Revert the SEAM_ field-name constants: the drafts are held under their own
names and finishReconciliationDrafts publishes each written draft as an
explicit unifiedTabsByWorktree/groupsByWorktree/activeGroupIdByWorktree/
layoutByWorktree property, so the counter sees every real write and the
split totals 2 + 4 against the original 9, with no exemption.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(dictation): show pending dots while the engine holds speech it hasn't committed
The VAD pause segmenter already knows when a phrase is open. The recognizer
now posts speech-pending when a phrase opens and clears it once that phrase's
text is posted, and the desktop pill shows three dots after the text while it
is on. Only pause-cut on-device streams send it, so cloud, the paired web
client and phones keep today's pill.
* fix(speech): keep a lone click from turning the pending dots on
* fix(speech): let a slow on-device finish run while it still returns text
The stop timeout counted 60 s from Finish however much audio the engine still had
to decode, so a CPU model on a busy machine failed a dictation it was still
transcribing. It now fails only after 60 s without text for that dictation.
* fix(speech): count any engine answer as progress while a dictation finishes
Error rates come from Orca's speech bench (transcribe-cpp 0.4.1, Metal, pinned GGUFs):
LibriSpeech for English, 5 FLEURS dev clips for de, es, fr, ja, ko, pt, ru, tr and zh.
speechModelQuality turns them into the Voice settings word scale.
* Source Control lists Codex models from app-server model/list instead of debug models
The Codex text-action picker read 'codex debug models', a dump of Codex's whole
catalog with prompt templates (40-90 KB per model). Once it passed the 4 MiB
output cap discovery failed and the picker kept Orca's built-in list, and the
dump's hidden entries were selectable. List models with app-server model/list
(hidden rows excluded), over Source Control's own launch, lock and teardown,
with the same 4 MiB input budget and 60 s deadline.
* Bound Codex model discovery input in bytes, key its cache case-insensitively on Windows, and stop the app-server only when it ignores stdin EOF
- Count raw bytes from both streams against the 4 MiB budget.
- Read CODEX_HOME (and other account dirs) case-insensitively on Windows for the host cache key.
- Skip the post-listing kill when the app-server already exited; fail waiting listing requests once discovery settles.
- Report an incomplete model list as such.
- Tests: fake-timer grace and successor admission after a successful listing, method-not-found and unreadable-page fallbacks.
* Key the Codex discovery cache on the env spelling Windows spawn uses
Windows env names ignore case and Node's spawn takes the lexicographically first spelling, so
CODEX_HOME wins over codex_home; the cache key now picks the same one.
* Narrow Codex app-server records with the provider record guard instead of type assertions
* Read Source Control action model picks under the repo's host key, as the picker saves them
Local repos resolved generation under the 'unknown' host key because their connection id is undefined, so the saved pick was skipped and the agent default ran.
* Key Source Control model picks by connection until the repo record loads, and test with a real local repo
* Key a native Codex listing's cache on process.env when no agent env was resolved
* fix(native-chat): apply Windows environment overrides consistently
Use the same case-aware composition for native chat launches, preflight,
model discovery and naming. Canonicalize configured Windows names before
account readers and preserve account pins and POSIX case distinctions.
Tracked in #39075.
Co-authored-by: 김태겸 <oharapass@gmail.com>
Co-authored-by: Vitalii Kulish <mancunianez@gmail.com>
* test(native-chat): cover effective Windows catalog changes
* test(native-chat): retain child environment test import after sync
* test(native-chat): verify Windows child environment precedence
---------
Co-authored-by: 김태겸 <oharapass@gmail.com>
Co-authored-by: Vitalii Kulish <mancunianez@gmail.com>
* fix(native-chat): show OpenCode subagents and keep a chat open while they run
OpenCode reports each subagent as a child session when the client opts in.
Orca now opts in, files each child's tool calls and replies under its own
subagent, and keeps the subagent open until the child reports that it
completed, failed or was interrupted, including after the parent turn ends.
Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>
* fix(native-chat): keep a subagent's tool context when its rows are read back from the journal
Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>
* fix(native-chat): a Stop also ends running OpenCode subagents, and each subagent can be stopped from the strip
Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>
* fix(native-chat): only a chat Stop closes OpenCode subagents, an early subagent Stop stays available, and the close request has a deadline
Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>
* fix(native-chat): a subagent Stop that the subagent confirms while the request is out reads stopped
Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>
* refactor(acp): close child sessions from the agent connection
Keeps the session runtime under the line limit after the main merge.
* test(acp): read a child command through a type guard
* fix(native-chat): a subagent that reports its ending reads stopped even when the close request fails
A child Stop's session/close now asks the agent to drop the call on timeout instead of withdrawing the write.
* test(native-chat): OpenCode subagent calls group with other agents' subagent calls
---------
Co-authored-by: SeaBee <85791528+seabee33@users.noreply.github.com>
* Save a pasted workspace directory as the path it names
A workspace directory with a leading space (" /Users/me/dev") was saved as
typed. Readers treat it as relative, so Orca resolved it inside every
registered repository and created an empty " /Users/me/dev" tree in each.
Normalize the saved value once where settings are written and loaded, as
the leading ~ fix did: leading whitespace before an absolute or home path is
dropped, while relative paths and trailing characters stay exact. A value
repaired on load keeps its old spelling in history so checkouts under the
old root stay visible. The settings field shows the saved value.
Co-authored-by: Eng. Mahmoud Bek ElNemer <ElNeMeR70@users.noreply.github.com>
* Expand ~\ only where it is a home path; drop type casts
* Classify the padded path with the host's own path rules
On macOS and Linux a leading backslash is part of a relative name, so only
Windows treats \work as absolute.
---------
Co-authored-by: Eng. Mahmoud Bek ElNemer <ElNeMeR70@users.noreply.github.com>
* Show the setup-script choice instead of greying out Create
With a project's setup script set to "Ask every time", Create stayed
disabled until the user picked Run or Skip, but that choice only renders
inside the collapsed Advanced section, so nothing explained the greyed-out
button. Keep the choice out of the disabled gate (as the gate's own comment
already intends) and have Create open Advanced to show it.
* Annotate the partial submit fixture cast
* Bring the setup choice into view when Create needs it
Create now scrolls the Run/Skip choice into the dialog's view and focuses
it on every attempt, including when Advanced is already open, and the
late-discovered setup path uses the same reveal. A submit that only jumps
to an existing pending creation no longer asks for a setup choice.
* Reveal the setup choice once per Create attempt
The choice clears the reveal request after scrolling and focusing, so a
later remount (for example after switching projects) cannot pull focus
without a new Create.
* Drop a pending setup reveal once the user has chosen
Choosing Skip in the moment before the reveal fired could still move focus
onto Run, so a later Enter switched the choice.
Offline models write out the speech before each pause while recording goes on.
Short noises and slow window-padded decodes are held off the pause cuts, a VAD
error falls back to bounded chunks, and slow phrase decodes are judged per
dictation against the time spoken. The VAD download carries legacy files and
fetches after any GGUF model.
* Improve project group ordering across hosts
* fix(project-groups): restore each group order after a failed drop
A deleted group made the aggregate rollback stop early, leaving other groups at
the ranks from the failed drop. Restore each group independently and log the
swallowed single-update failures again.
* refactor(main): give project group RPC methods their own module
Keeps repo.ts under the 300-line cap as the project group surface grows.
* fix(dictation): join Chinese and Japanese phrases without a space, and space Korean ones
The shared transcript fold joined every phrase with a space, so the pill and
phone replies showed Chinese and Japanese text with spaces the inserted text
doesn't have. It now joins with the same rule as insertion. That rule treated
Hangul like Chinese and Japanese, so Korean phrases were inserted with no space
between sentences; Korean spaces words, so Hangul now spaces like Latin text.
* fix(dictation): don't space after Chinese and Japanese full-width punctuation
* fix(dictation): join retried phrases with the same spacing rule
* fix(dictation): escape the spacing rule's non-ASCII marks for the orcad bundle
The shared fold now pulls the spacing rule into the orcad server bundle, which
must stay ASCII. A rich-draft test pinned Hangul joining without a space; it
now expects the Korean spacing.
Stopping a CPU dictation closed the renderer's audio port while frames still
waited behind a decode in the speech worker; the close ended the stream at once
and the queued frames were dropped. A close after 'end' now leaves them to
decode. A model slower than 30 s to load also lost the start of the dictation,
since the renderer kept only 30 s of startup audio; it now keeps as much as the
host waits for the load.
* feat(speech): spell dictionary terms in dictation, with repo terms and Whisper prompt
* test(speech): pin host repo terms for phone dictation
* fix(speech): keep symbol terms whole, accept null setup params, resolve SSH repos
* refactor(speech): key transcript words once and type phone terms at the reply schema
* refactor(speech): merge speech-types imports in the speech IPC
* feat(runtime): tell phones when the desktop refreshes a PR
* fix(runtime): name every checkout a coalesced PR read stood for
* fix(runtime): don't count a read that couldn't load checks as a checks change
transcribe-cpp is now the only on-device engine on every platform, so
the sherpa-onnx packages, their packaging (extraResources, the darwin
libonnxruntime prune, the native-variant assert and the Linux libstdc++
exemption), the sherpa recognizer in the speech process and the
per-model engine choice all go. Hotwords are written as plain terms.
* perf(github): answer phone PR lookups from the desktop's last result
* fix(github): order last PR outcomes by lookup start sequence
* fix(github): keep reused phone PR outcomes out of the refresh observers
* fix(github): reuse a fallback-found PR only for the same fallback number
* feat(speech): move the speech catalog to transcribe-cpp models
Parakeet v2/v3, Whisper tiny and SenseVoice now download their GGUF
builds, which run on transcribe-cpp everywhere, Intel Macs included (on
the CPU). A model still holding its old sherpa-onnx download shows
"Update to the faster version (<size>)"; updating replaces the old
files, and dictating before then says the model needs that update.
The five models with no GGUF build are retired and their users moved
over: the Japanese Parakeet, the Chinese-English Zipformer and the
Paraformer to SenseVoice, and the Korean and Chinese streaming
Zipformers to the new Nemotron 3.5 (28 languages). Nemotron's locale
codes (ko-KR) are matched from the ISO language setting.
* feat(speech): offer the speech model update once after the engine change
A card in the notification corner tells people whose selected on-device
model needs the new download, or whose retired model was replaced, and
starts the same download as Settings > Voice. After a replacement it
offers to delete the retired model's files.
* feat(speech): move retired Korean and Chinese models to SenseVoice and offer Whisper Large v3 Turbo
On transcribe-cpp 0.4.1 with Metal, SenseVoice beat Nemotron 3.5 on Korean
and Chinese FLEURS clips, so the retired Korean and Chinese streaming
Zipformers now move to SenseVoice. Whisper Large v3 Turbo (Q8_0, 886 MB)
joins the catalog as the most accurate choice for Japanese, Korean and
Chinese. Nemotron 3.5 stays, described by the languages it does best.
* feat(mobile): keep dictating after leaving the session, with a floating pill
* feat(mobile): translate the dictation pill and its toasts
* fix(mobile): type away words without claiming the terminal, and keep the pill above screens
* fix(mobile): float the pill above bottom bars, reopen the owner's own screen, and tell a refused start wherever you are
* fix(mobile): float the dictation pill above the source control bar
* fix(mobile): keep pill state out of the host context and reopen the starting screen first
* refactor(mobile): subscribe the pill only while it shows and trim the session dictation hook
* fix(mobile): key bar clearances by symbol
* feat(mobile): save the chat when the app leaves the foreground
A swipe-kill from the app switcher can end the app before the page is
hidden: iOS fires visibilitychange only at background, and Android Recents
leaves the activity resumed. The shell now sends an appInactive frame on
iOS inactive and Android window blur, to pages that declare they read it,
and the page saves any chat rows the store doesn't hold yet.
* fix(mobile): keep the leave callback out of render
* fix(mobile): notice a changed middle row before skipping a chat save
* perf(mobile): show accounts from the last snapshot
* fix(mobile): never count a stored accounts snapshot as live
* test(mobile): pin when accounts usage counts as live
* fix(mobile): date accounts reads by when they were sent and block switches until live
* fix(mobile): show accounts usage only from this screen's own read, drop replies from before a reconnect
* feat(settings): add a Use GPU switch for on-device dictation
On by default (Auto: the GPU when it works, the CPU when it fails).
Turning it off keeps the transcribe-cpp speech process on the CPU, for
machines whose GPU driver freezes or restarts them. The row shows when
the selected model runs on transcribe-cpp.
* fix(speech): reload the warm speech model when Use GPU changes
A loaded recognizer kept the GPU choice it started with, so turning Use GPU
off left the next dictations on the GPU until the model unloaded.
* fix(speech): reload the speech model after a dictation when Use GPU changed during it
A recognizer still in use when Use GPU changed stayed loaded afterwards with
the old choice. It now unloads as soon as its last dictation ends, and the
warm-up loads it again with the new choice.
The pill showed only the current partial and went back to Listening after
each final, so pause-committed phrases from transcribe-cpp never appeared.
The host's transcript fold moves to src/shared and the desktop pill uses it
too: committed text in the foreground, tentative text muted.