Use one native app-server for terminal, Web and phone clients while retaining the existing CLI and Runner lifecycle.
Synchronize native queues, permissions, question history and steering state; preserve explicit permission precedence and per-turn usage models. Resume inactive clear commands through Runner and reject independent child cold resumes.
Add shared-runtime regression tests, generated protocol fixtures and lifecycle documentation.
Add viewport-driven paging with layout acknowledgements, bounded retries, cancellation gates, and epoch-safe history retention.
Preserve transcript anchors and expansion state, fix tool-group identity collisions, and serialize Android history coordination on Main.
Reduce per-scroll composition and layout work; add native regression tests, CI coverage, and profiling guidance.
* feat(shared): steer capability gates and live steered signal schemas
- STEERING_SUPPORTED_FLAVORS / isSteeringSupportedForSession gate which
agents can deliver queued messages into the active turn (pi, codex,
cursor ACP; legacy stream-json cursor excluded)
- AgentState.steeringActive, DecryptedMessage.steered and
messages-consumed live signal (never persisted by the hub)
* feat(cli): queue reservations and steered messages-consumed option
- MessageQueue2 gains takeByLocalId/restoreReservation/
beginReservationDispatch/commitReservation so an async steer can reserve
a queued row without racing the main loop's turn/start drain
- emitMessagesConsumed accepts steered: true to mark mid-turn delivery
* feat(codex): mid-turn steer via app-server turn/steer (#888)
- CodexAppServerClient.steerTurn + TurnSteerParams/Response types
- CodexRemoteLauncher registers the steer-queued-message RPC handler:
reserves the queued row, validates it against the active turn (no
control commands, matching mode hash), injects via turn/steer with an
epoch guard that invalidates in-flight steers on abort/cleanup
- steeringActive agent state tracks the active-turn window
- hub syncEngine gate opens to codex; messages-consumed relays steered
* feat(web): Steered badge and steer gating for codex sessions
- HappyUserMessage shows a ↳ Steered badge fed by the live
messages-consumed steered signal, preserved across server echoes and
refetches (mergeMessages carries the optimistic marker)
- SessionChat gates canSteer via isSteeringSupportedForSession instead of
the pi-only check
- clearStaleQueuedStatus normalizes a queued status on an invoked message
- fix(web): drop duplicate showSessionSummaryInChat in markdown test
(upstream typecheck breakage)
* fix(codex,shared): address bot findings on steer gate and ambiguous turn/steer
- STEERING_SUPPORTED_FLAVORS / isSteeringSupportedForSession advertise
codex and pi only; cursor joins when its soft-steer handler lands (#1609)
- turn/steer now splits dispatch (stdin accepted) from completion (turn
finished): the hub RPC acks once dispatch succeeds — never on the
concurrent turn's completion, which can exceed the 30s RPC window
- queue row commits only after the turn settles; a rejected/aborted steer
restores the row so the message still delivers via turn/start, and a
dispatched steer is never restored (no duplicate delivery)
- steer carries clientUserMessageId (echoed as userMessage.clientId) so
ambiguous transport failures can reconcile the thread later
- client tests cover dispatch/complete split and stdin-write failure
* fix(codex): reconcile dispatched steers before restoring; align error copy
- A dispatched turn/steer whose completion fails (disconnect / protocol
error) is now reconciled via thread/read by clientUserMessageId before
the queued row is restored — the instruction is only re-delivered by
turn/start when the thread never received it
- Reconcile targets the pinned steer thread, not whichever turn is
current when completion fails
- syncEngine unsupported-flavor error now matches the capability gate
(Pi and Codex only until the cursor handler lands)
- launcher tests cover steer success (ack on dispatch), reconcile-accepted
and reconcile-rejected outcomes
* fix(codex): consume the row at dispatch; drop background reconcile
- The hub RPC acks and the queue row is consumed as soon as stdin accepts
turn/steer; completion is background-only logging. A dispatched steer is
never restored, so the same localId cannot be re-delivered via turn/start
after the caller was told the steer succeeded
- Dispatch failure (stdin write error) still restores the row and reports
failure
- steer.completed rejection is always handled (no unhandled rejection on
the dispatch-failure path)
- tests updated: completion failure after dispatch keeps the row consumed;
dispatch failure restores it
* fix(codex): distinguish definite rejection from indeterminate completion
- Transport-level failures (timeout, abort, disconnect, spawn, protocol)
carry an indeterminate marker; explicit JSON-RPC error responses do not
- After a dispatched steer, turn completion resolves → commit + consumed;
a definite app-server rejection restores the row (instruction was never
accepted, so turn/start cannot duplicate it); an indeterminate outcome
leaves the row reserved so it can never be delivered twice
- Completion handling registers before awaiting dispatch so the
dispatch-failure path cannot leak an unhandled rejection
- client/launcher tests cover explicit rejection (restore), indeterminate
outcome (row stays reserved) and dispatch failure
* fix(codex): reconcile indeterminate steers instead of a permanent reservation
- After an indeterminate completion (disconnect/protocol), reconcile the
thread by clientUserMessageId immediately: accepted → commit + consumed,
provably rejected → restore, still unreadable → keep the reservation and
retry from the main-loop top on later passes (post-reconnect)
- A row never sits in dispatching forever: the hub cannot stamp it invoked
while the instruction may never have been accepted
- tests: indeterminate keeps reserved while thread unreadable; accepted
reconciliation consumes; rejected path restores
* fix(codex): accept all thread item shapes; retry reconcile; ack through abort
- Reconcile matcher accepts userMessage/user_message with clientId/
client_id, matching the shapes the thread parser supports — an accepted
steer can no longer be misclassified as rejected
- A pending reconciliation schedules a wakeLoop retry, so a temporary
app-server outage cannot strand the reservation behind waitForTurnOrRecovery
- The success-path ACK no longer checks the steer epoch: the hub already
reported steered on dispatch, so commit + messages-consumed must reach
it even when an abort resets the queue in between
* fix(codex): reinit reconnected app-server; keep reconcile retries alive
- thread/read after a disconnect auto-connects a fresh app-server, which
must be initialized before any request — reconcile now ensures
connect + initialize (isConnected getter added to the client)
- every still-unknown loop-top reconciliation schedules the next retry,
so recovery without external traffic is eventually observed
- launcher mock gains isConnected
* fix(codex): timer-driven reconciliation; init tracking; abort-safe ACK
- Reconciliation runs on a self-rescheduling 1s timer independent of the
main loop (wakes it too), so idle loops and waitForTurnOrRecovery still
observe app-server recovery; abort clears nothing implicitly — the ACK
path commits and consumes even when the reservation was cancelled
- Absence of a durable client id is ambiguous: unmatched reads stay
'unknown' and keep retrying instead of restoring the row
- CodexAppServerClient tracks initialized state (reset on disconnect/exit)
so ensureAppServerInitialized re-initializes a fresh process before
thread/read; initialize failures leave the flag false for the next retry
- tests: accepted reconciliation via scheduled timer, indeterminate
keeps reserved, explicit rejection restores
* fix(codex): bind reconciliation to the launcher lifecycle
- runSteerReconciliation clears any armed retry timer on entry and never
installs a second one, so loop-top and timer-driven passes cannot
multiply
- shuttingDown is set when the main loop ends: timers are cleared and the
pending map is dropped, so an unresolved steer can never respawn an
app-server after cleanup (remote-to-local switch included)
* fix(codex): report steered only after app-server acceptance
- The handler now awaits steer.completed (the inject-acceptance response):
an explicit JSON-RPC rejection surfaces as failed and restores the row
for the normal turn/start path instead of a false steered
- Transport failure after dispatch reports 'Steer outcome is being
reconciled' and keeps the row reserved while the timer-driven thread
reconciliation runs
- dispatch-failure path also swallows the paired completion rejection
* fix(steer): tri-state cancel, clear-safe reservations, bounded acceptance wait
- MessageQueue2.cancelByLocalId returns 'in-flight' for a dispatching
steer reservation: the hub neither deletes the row nor stamps invoked_at
(new CancelMessageResponse 'busy' status; web restores the optimistic
row); pushIsolateAndClear and reset/close share cancelReservations so
/clear-style commands cannot have a rejected steer resurrect a discarded
prompt
- turn/steer acceptance wait bounded at 25s (< hub 30s RPC timeout): a
lost response is indeterminate and funnels into thread reconciliation
instead of stranding the reservation
- tests updated for the tri-state cancel contract
* fix(codex,web): busy-aware edit flow; bound reconciliation reads
- QueuedMessagesBar edit flow treats a 'busy' cancel as unsuccessful: it
never prefills the composer when the row is inside an async steer, so a
second client cannot send a duplicate
- reconcileSteerByClientId bounds thread/read with a 5s timeout so a
connected-but-silent app-server cannot hold the reservation in-flight
indefinitely
* fix(steer): inFlight-dominated cancel acks; bounded reconciliation
- hub cancel-queued-message acks check inFlight before removed: a stale
duplicate socket reporting removed can no longer delete the durable row
while another socket is dispatching the steer
- reconciliation entries expire after 60s and mark delivered: after the
rejection window, a dispatched steer that the app-server never proved
(client ids dropped on restart) is committed instead of polling
thread/read forever
- pre-dispatch failures (abort before write included) never enter
reconciliation — they restore the row and report failure
* fix(steer): persist indeterminate outcomes without replay
* fix(steer): make ambiguous delivery restart-safe
* fix(steer): recover crash-held rows and preserve retry dedup
* fix(steer): ack retries and bound stdin dispatch
* fix(steer): reconcile indeterminate dispatches and serialize retries
* fix(codex): classify stdin callback failures as indeterminate
* fix(steer): recheck indeterminate cancels after ACK
* fix(steer): close retry and abort races
* fix(steer): serialize live retries and abort admission
* fix(steer): distinguish live dispatching from unknown
* fix(steer): keep ACK failures held and reconcile busy cancel
* fix(steer): distinguish held cancel from removal
* fix(store): combine schema v24 migrations
* fix(store): reserve schema v25 for steer delivery state
* fix(steer): keep held cancel state and notify requeue
* fix(steer): release explicitly cancelled unknown reservations
* fix(codex): reject cancelled reservations before native steer
* fix(codex): make reservation restore atomic with state
* fix(codex): terminate abandoned transport writes
* fix(steer): own abandoned app-server lifecycle and consume races
* fix(codex): confirm dispatch and recover abandoned turns
* test(codex): mock abandoned transport callback
* fix(codex): clear visible turn state on transport loss
* fix(steer): claim retries and cover native delivery state
* fix(native): preserve indeterminate state on Android hydration
* fix(steer): make retry claims single-winner
* fix(steer): serialize concurrent retry claims
* fix(socket): tolerate missing steer-state ACK callbacks
* fix(native): serialize retry operations
* docs(web): document unknown steer delivery and retry controls
* fix(steer): handle retry failures and abort-before-connect
* fix(steer): reinitialize after transport loss and finish iOS retry errors
* fix(steer): preserve indeterminate rows across reconnect gaps
* test(web): mock indeterminate queued recovery state
* fix(steer): recover consumed ACK tombstones
* fix(steer): expose consumed cancel tombstones
Gradle/Firebase: firebase-messaging + work-runtime-ktx in the catalog and
:app; com.google.gms.google-services applied CONDITIONALLY (only when
app/google-services.json exists) so the repo builds green without any
Firebase config; committed google-services.json.example + README section
(CI-injected official builds / self-build drop-in / v1.x runtime
FirebaseOptions path); real config gitignored. push/PushBinding.kt is the
availability seam: no Firebase -> every push path no-ops.
Registration (:core:data push/DeviceRegistrar): stable DataStore UUID
deviceId; POST /api/devices/register {token, platform:'phone', deviceId}
fanned out to EVERY paired hub on app start, on pairing (roster addition),
and on onNewToken; transient failures retry via a per-hub WorkManager
unique work item; best-effort DELETE on sign-out while the hub's JWT still
works.
Service (fcm/HapiFirebaseMessagingService): data-only contract v1 decoding
in :core:data push/PushPayload — channels permission_requests(HIGH) /
ready / task_notifications (created on app start), type-<sessionId>
coalescing tags, severity accent colors, notifySummary-driven ready
bodies, unknown type/contractVersion degrade to plain title/body (default
channel, no actions). Suppress-when-open: foreground + that session's
chat composed -> skip (SSE already shows it). Tap -> internal MainActivity
intent route (no public URI) -> existing navigation opens the chat.
Actions: permission-request Allow/Deny and ready/task RemoteInput Reply +
Dismiss -> NotificationActionReceiver -> expedited CoroutineWorkers
(approve/deny {} bodies, reply {text, localId}) through on-demand
HubSessions built from stored credentials (HapiWorkerFactory +
Configuration.Provider + on-demand WorkManager init — no HubGraph needed
in background); notification updates pending -> done / "Already handled"
(404 request-gone) / inactive / failed. Multi-hub: payload has no hub URL,
so workers try the ACTIVE hub first, then other paired hubs on 404
session-miss (single-hub users always hit first try).
Hub check: android.priority=HIGH is already set unconditionally for every
FCM message (hub/src/fcm/fcmService.ts) — no hub change needed.
Tests (34 new, :core:data): payload keys/severities/unknown contract
version, channel routing, suppress-when-open; registrar fan-out /
addition-only / retry-on-transient / null-token no-op via fake seams;
action wire bodies + Bearer header + multi-hub resolution through a real
HubSession against two MockWebServers. Full gate green with AND without
google-services.json (protocol 227, data 182, app 96 tests; debug +
release/R8/lintVital assemble).
Composer attachment tray wired end to end: "+" bottom sheet (photo
library / camera / files), upload-on-pick against POST upload
(JSON+base64), per-chip uploading -> ready/failed states with retry and
best-effort delete-on-remove, AttachmentMetadata riding SendMessageRequest
and the optimistic row so user bubbles thumbnail instantly, and a
UserTextBlockView upgrade decoding wire previewUrl data URLs (web-sent
messages thumbnail too).
Mobile-data compression policy (differs from web, which uploads
originals): recompressible images over 4 MB downscale to 2048 px JPEG
q85 (filename swaps to .jpg); GIF/SVG/non-images keep originals; hard
50 MB reject with a notice, plus a capped read guarding unknown-size
picks. previewUrl embeds a <=512 px JPEG thumb instead of the web's
full-size data URL to keep send bodies small.
Simplifications noted in KDoc: attachments do not persist in drafts v1
(holder onCleared discards un-sent uploads after best-effort deletes);
inactive sessions fail the upload chip until a text send auto-resumes.
Scheduled sends guard the wire constraint (scheduledAt excludes
attachments) with a loud check.
Seam: ChatSessionApi now extends AttachmentUploadApi (HapiApi methods
gain override); camera captures use a FileProvider cache scratch,
rememberSaveable across rotation.
Tests: pure policy decisions (plan/sample-size/filenames/data URLs),
controller state machine incl. mid-upload removal orphan cleanup,
VM send/retry/refusal flows with metadata assertions, MockWebServer
wire shapes for upload + delete. Full gate green (protocol/data/app
tests + assembleDebug).
Interaction layer turning the read-only chat into a working remote control:
- Composer: multiline input bar with optimistic sends (appendOptimistic ->
POST -> status settle), queue-by-default delivery with a long-press
Send&Steer intent while a turn is active, abort button during thinking,
tap-to-retry on failed rows, per-session drafts (DataStore, hub-scoped
keys), attachment chip seam for M4.
- session_inactive (409) recovery: one POST /resume then retry; a
superseding session id seeds the new window, migrates the draft and
emits SessionSuperseded for renavigation (web resolveSessionId parity).
- Queued bar: uninvoked sends with Cancel (DELETE; invoked-race ingests the
authoritative row as sent), Edit (cancel + composer prefill, draft-kept
guard) and Steer (POST steer; invoked answers reconcile a missed consume).
reconcileQueuedState now runs on chat open and on session-pipe gap.
- Permission actions: flavor-exact bodies mirroring PermissionFooter.tsx --
claude {} / allowTools / mode:acceptEdits, codex-family decision:
approved / approved_for_session / abort -- plus AskUserQuestion flat
answers and request_user_input nested answers forms; optimistic
Resolving/AlreadyHandled overrides settled by the agentState patch.
- Session config sheet: catalog-driven permission-mode picker, claude
static model/effort catalogs (ported to :core:protocol catalog), codex
models via GET /codex-models (new HapiApi endpoint + wire types) with
per-model reasoning efforts; optimistic detail updates rolled back to
server truth on error.
- Lifecycle: ProcessLifecycleOwner -> SseEngine.setLifecycleForeground +
POST /api/visibility per subscription (VisibilityReporter fed by the new
SyncTargets.onHandshake hook); the global SSE pipe moved from the session
list VM to HubGraph lifetime (GlobalSsePipe) so queued/consumed
bookkeeping and list badges stay fresh while a chat is open.
- Tests: VM-level interaction suite (optimistic send/failure/retry,
inactive-resume both id paths, cancel invoked-race, steer, exact
approve/deny body JSON incl. both answers formats, config optimistic +
rollback, drafts) + GlobalSsePipe tests; full gate green (554 tests,
assembleDebug).
:core:protocol window/ — pure state machine ported function-for-function from
web/src/lib/message-window-store.ts + messages.ts: merge-by-(id|localId) with
optimistic echo reconciliation, position ordering (invokedAt ?? createdAt, seq,
ASCII id tie-break), trim-preserving-queued with the codex agent-run budget,
epoch reset handling, latest-replace with request-baseline identity
preservation, consumed/cancelled/queued-reconcile transitions, tail/history
modes, hydrate/persist shapes. MessageRetention ports the null-decision tree
of normalizeDecryptedMessage (dedup with the B-M2a pipeline port flagged).
:core:data store/ — per-session MessageWindowStore (StateFlow + Mutex,
single-flight tail sync with trailing drain, fetchOlder with epoch-reset
resync, SSE ingest hooks, optimistic sends, queued-state reconciliation,
seedFrom for resume id changes) + WindowSnapshots (atomic JSON files, LRU 10;
JsonSnapshotStore dedup TODO) + MessageWindowStores registry. Minimal
MessagesApi interface (sealed MessagesQuery) extracted over the two message
endpoints, implemented by HapiApi.
Gate: PaginationFixtureTest replays all 11 shared/fixtures/pagination scripts
against the real store — requests, older-load outcomes, reconcile candidates
and the final window projection all exact — 11/11 green; plus targeted
concurrency/snapshot/seed unit tests. :app:assembleDebug green.
:core:data app.hapi.data.sse — the /api/events transport per
docs/api/client-contract/sse.md (reference web/src/hooks/useSSE.ts):
- SseConnection: SseTransport seam (Connected/Event/Failure flow) with the
okhttp-sse implementation on a dedicated client (readTimeout=0, no cache,
own dispatcher), buildEventsUrl, Last-Event-ID header on resume, and an
acceptEncodingIdentity fallback flag for the gzip escape hatch.
- ReconnectPolicy: normative constants + pure backoff schedule (immediate
first retry, 1s..30s exponential, 300s ceiling after 8 attempts, 0..500ms
injected jitter).
- SseEngine: per-key (global / session:<id>) connection loops — handshake
gate on connection-changed{status:connected} with ok/gap resume verdict,
per-key cursors advanced only after the downstream hand-off
(at-least-once), 10s connect deadline, 90s staleness watchdog ticking 10s,
one silent 401 re-auth per cycle costing no backoff attempt, background
retry deferral + 45s foreground stale check; all timing via injected
delay/clock for virtual-time tests.
- SyncEventRouter: 13-type union fan-out to the SyncTargets seam (M2 wires
stores), handshake gap -> requestFullResync, Unknown ignored.
Tests (32): virtual-time engine suite (fake transport + turbine), policy
schedule with seeded jitter, router mapping, and MockWebServer integration
through the real okhttp transport — including proof that gzip SSE frames
surface incrementally (flush-per-event body withheld behind a throttle).
Gzip also verified against a live local hub (Content-Encoding: gzip,
handshake decoded instantly, heartbeat +30.0s mid-stream).
weishu