* feat(android): anchor the Material type scale on the shared reading tokens
MaterialTheme was built with only a colorScheme, leaving HapiTypography's
body/code/caption tokens an island that never influenced Material
typography. Derive the app-wide Material type scale from those tokens so
Material widgets and direct token consumers share one type system: the
body token drives bodyLarge (the chat reading size) and the caption token
drives bodySmall (the metadata size), while every other slot keeps its
Material platform default so existing MaterialTheme.typography call sites
stay pixel-identical. Only size and leading come from the tokens; weight,
tracking and family stay Material's. HapiTheme now accepts a typography
argument (defaulting to the shared scale) and passes it through, and the
monospace code token stays a direct-consumer token because Material has
no code slot.
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* feat(android): render settings pickers as expressive choice dialogs
The theme and language pickers were the last bare AlertDialog + radio-row
surfaces in the app: plain stacked radio buttons with no visual anchor for
what each option means, and two focus targets per row (the row and the
RadioButton each carried their own click handler). Give small exclusive
pickers a shared expressive form: each option renders as a full-width
choice card whose selected state is a secondaryContainer fill plus a
trailing check, and options can carry a fixed preview swatch.
The theme modes show two-tone palette chips — follow-system spans the
light and dark palettes, OLED is outlined pure black, dark uses the dark
canvas rather than black so the two stay distinguishable at a glance.
Swatch colors are fixed values depicting the mode they name, not derived
from the active scheme, mirroring how the web color-theme cards hardcode
their previews: a "light" chip must stay light even inside a dark dialog.
Rows keep radio semantics through Modifier.selectable with Role.RadioButton
inside a selectableGroup, so TalkBack still announces the standard
"selected, n of m" picker phrasing while each option becomes a single
focus target. Tapping an option still applies it immediately. Language
options skip the swatch slot; the component stays generic over the value.
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
---------
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: HAPI <noreply@hapi.run>
Replace the single-column settings screen with a category directory
whose rows come from a declarative catalog (SettingsHubCatalog: one
entry = route + title + summary + icon + visibility lambda), so adding
a category later (Display, Chat, Voice, Machines) is a list element,
not layout surgery. Mirrors the web settings IA at the category level.
- General subpage (settings/general) inherits the uncategorized
sections: Appearance (theme + Material You), Language, About
(versions + hub health). Zero behavior change per item.
- The owner-only Insights section dissolves into hub-level rows:
Usage/Storage render only for the owner (same ns=default gate) and
route to their existing dashboards, matching the web's owner-only
nav categories (SettingsNav OWNER_ONLY_CATEGORIES).
- Flat settings/... route pattern with per-hub ViewModelHolder, same
as the existing usage/storage precedents; back from a subpage lands
on the hub. The hub's holder passes probeHealth=false so opening the
hub itself fires no /health request; the General holder keeps it.
- Previews: hub owner/non-owner gating + General body. JVM tests pin
catalog order, the owner gate, route uniqueness, and row summaries.
- Strings: settings_category_general(+_summary) in en/ru/zh; relocate
three zh keys (chat_tool_resolved, chat_plan_implement,
chat_plan_continue) that had drifted to the file tail and broke the
three-language key order (KEYSYNC) — pure moves, values untouched.
via [HAPI](https://hapi.run)
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: HAPI <noreply@hapi.run>
The ComposerAttachments harness left the constructor default
encodeDispatcher (Dispatchers.Default, a real thread pool) in place.
With two uploads in flight the second-launched one can finish encoding
first, reach FakeUploadApi.uploadFile first, and consume the scripted
results deque out of launch order — the queued failure gets stolen by
the upload meant to succeed, so the robbed chip never settles to its
scripted state and items.first {} hangs until runTest's dispatch
timeout (UncompletedCoroutinesError, intermittent under suite load).
Pin the encode hop to the scope's own dispatcher (the runTest scheduler
in every test), making upload arrival order equal launch order
deterministically; upload still hops dispatchers as in production. Add
a regression-lock test scripting two results over two concurrent adds.
Product code is untouched: upload() keys results by attachment id and
is arrival-order independent — the ordering dependency lives purely in
the test fake's deque.
via [HAPI](https://hapi.run)
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: HAPI <noreply@hapi.run>
* fix(codex): drop safe-yolo from Codex permission modes
Every Codex session runs on the native shared runtime (runCodex is
runSharedCodex), which rejects safe-yolo. The mode stayed in
CODEX_PERMISSION_MODES and six filters hid it again in the web picker,
composer, hub permission-mode route, hapi set-mode and both native
apps. API callers that validate against the flavor catalog (e.g. a
spawn with permissionMode: 'safe-yolo') still passed it through and
only failed inside the runner.
Remove it from the Codex catalog (shared, iOS, Android, fixture) and
delete the filters. autoReview is the native replacement. Safe Yolo
stays for Gemini, Kimi and Copilot.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(ios): Codex option list now has four modes
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* feat(android): add session date-range filter with activity-aware picker
- Pure SessionDateRangeFilter port of the web SessionList.tsx semantics:
strict YYYY-MM-DD parsing with calendar round-trip validation,
[start local midnight, end local midnight + 1 day) windows anchored to
the injected zone, session.updatedAt as the activity field
- M3 DateRangePicker dialog (opens on the selected start's month, Clear
resets without closing) plus a lightweight recent-activity legend —
declared divergence: M3 exposes no per-day decoration API, so the web's
in-cell "has session activity" emphasis surfaces beside the picker
instead; explicit Apply replaces the web's close-after-second-click
- Wired as a transient, unpersisted lens (web useState parity) applied
before the machine scope; the activity-day set and machine chips both
derive from the unfiltered list so filtering never hides alternatives
- Top-bar calendar action with active dot badge, active-range summary
row with clear, date-specific empty state; strings in en/zh-rCN/ru
(en-ru key order KEYSYNC clean)
- JVM tests: predicate window edges, DST spring-forward, zone-anchored
midnights, strict parsing, VM lens combination and pre-filter
activity-day derivation
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* fix(android): refresh session date filters after timezone changes
* Deduplicate Android sessions before applying date filters
---------
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: HAPI <noreply@hapi.run>
Co-authored-by: weishu <twsxtd@gmail.com>
* feat(android): add scratchlist row move, copy, and delete-confirm actions
Align the scratchlist entry card with the web client's per-row actions
(ScratchlistPanel inventory):
- Move up/down: adjacent swap, disabled at the list edges, local-only
like web v2 — no protocol request; the next refetch restores hub
`createdAt DESC` order (pure `ScratchlistEntryOps.moveEntry`).
- Copy: copies only entry.text via ClipboardManager (FullTextAction
precedent), swaps the icon to a check for 1500 ms on success, single
per-list copied state, silent no-op on failure.
- Delete confirm: texts strictly longer than 100 characters (web
SCRATCHLIST_CONFIRM_DELETE_THRESHOLD, untrimmed length) confirm via
AlertDialog before delete, both from the row and the editor sheet;
shorter notes delete directly.
New en/ru/zh-rCN strings ship in the same commit.
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* fix(android): guard saved scratchlist notes and restart copy feedback
* test(android): uniquely select scratchlist confirmation actions
* Fix Android scratchlist editor test targeting
---------
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: HAPI <noreply@hapi.run>
Co-authored-by: weishu <twsxtd@gmail.com>
* feat(android): remember new-session launch defaults per machine and agent
Port of the web launch-preference persistence
(`hapi:newSession:launchSettings:v1:<machine>:<agent>`): model, effort,
reasoning effort and — for shared-permission-mode flavors only — the
permission mode of a successfully spawned session are remembered per
(machineId, agentFlavor) dimension and prefill the next create form for
that dimension (fresh screen, agent switch, machine switch).
- Storage: NewSessionPrefsData gains a nested `launchSettings` map inside
the existing `hapi_prefs` DataStore blob; legacy blobs without the
field decode to empty defaults, corrupt shapes keep the pre-existing
whole-blob default degrade (DirectoryExpandPrefs contract).
- Capture happens at spawn success only (web saves on every
create-success path); grok renders the native select but its mode is
deliberately not remembered (web `usesSharedPermissionModeState`).
- Restore: resolvePreferredLaunchSettings validates claude's static model
and effort lists, drops `xhigh` from opencode's reasoning efforts and
coerces a removed permission mode to `default` — never a stale toggle;
codex's dynamic model passes through for the existing settled-catalog
reconciliation.
- A restored in-progress draft outranks the remembered defaults until the
user switches agent or machine (web `preserveRestoredDraftRef`).
- Machine switches now also reset effort/reasoning effort/shared mode to
the new dimension's remembered values (or sentinels), matching the web
restore effect that re-runs on machineId change.
Divergences from web: no `cursorSelectedBase` (no Android picker yet);
agy's static model presets are not re-validated on Android (no v1 picker,
the stored value is always `auto`).
Tests: prefs round-trip/cold-start, legacy-blob tolerance, corrupt blob
and corrupt-entry fallback (real temp-file DataStore, pure JVM); capture
gating and resolve coercion logic; VM prefill per dimension, machine and
agent switching, spawn-success capture, and draft precedence.
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* fix(android): preserve launch selections across machine heartbeats
---------
Co-authored-by: HAPI <noreply@hapi.run>
Co-authored-by: weishu <twsxtd@gmail.com>
* feat(android): settled permission verdict details
Port the web PermissionFooter verdict classification so a settled
permission row explains HOW a request was decided, not just that it was.
Behavior vs web (web/src/components/ToolCard/PermissionFooter.tsx:11-63):
- codex family sessions (flavor family + Codex/Gemini/OpenCode/Copilot/
Cursor tool prefixes): approved + decision approved_for_session shows
"Approved for session"; denied + decision abort shows "Aborted" with
the wire reason appended when present.
- claude sessions: mode acceptEdits shows "Approved: allow all edits"
(previously the raw mode id was appended); decision approved_for_session
or an allowedTools hit — plain tool name or the `Bash(command)`
qualifier built from the tool input command/cmd — shows
"Approved for session" (previously indistinguishable from a plain
approval).
- canceled now appends the wire reason (previously dropped).
- resolved/pending/denied unchanged in shape; denied keeps the error tone
and aborted shares it.
Deliberate divergence: web's own settled footer renders only a minimal
red reason box (PermissionFooter.tsx:207-220) — its formatPermissionSummary
classification is unreachable there. Android renders the classified
verdict line in the existing subdued row style instead, as intended.
Implementation: new pure JVM classifier
feature/chat/permissions/PermissionVerdict.kt (exhaustive status ->
PermissionVerdictLabel mapping, unit-tested in PermissionVerdictTest);
ToolCallBlockView.PermissionStateRow becomes a thin renderer over it.
Strings added in en/zh-rCN/ru (ru mirrors en key order; zh values from
web locales).
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* feat(android): persistent permission-mode chip + inactive gating
Item 2 — persistent permission-mode chip above the composer, mirroring
the web StatusBar mode span (StatusBar.tsx:253-265, 389-393):
- New PermissionModeChip composable + PermissionModeChipRule pure object:
resolve the raw wire mode, keep it only when the flavor offers it, and
hide `default` everywhere except copilot (model=auto sessions still
surface the chip). Unknown wire ids hide (no raw fallback).
- Tone mapping follows the web PERMISSION_TONE_CLASSES (StatusBar.tsx:
43-48): neutral=onSurfaceVariant, info=primary, warning=tertiary,
danger=error. Labels stay en catalog strings — web mode labels are
hardcoded English, not i18n.
- The chip consults the unfiltered flavor catalog: the concurrent-clients
Safe-Yolo carve-out only narrows the switcher, never the chip.
- Deliberate divergence: the web span is not clickable (its switcher is
the composer settings sheet / Shift+Tab). Android has no Shift+Tab, so
tapping the chip opens the session config sheet that hosts the
switcher — same surface as the top-bar gear, no new top-bar icon
(ChatScreen keeps its two-icons-max rule).
Item 4 — permission-mode gating aligned with web (SessionChat.tsx:
2079, 2138-2163; HappyComposer.tsx:536-537, 1544-1549, 1960-1995):
- Inactive sessions stay switchable, matching web: SessionChat passes
allowSendWhenInactive unconditionally, so configurationControlsDisabled
never trips for inactive threads and the offline notice stays
informational only (the send path auto-resumes).
- The permission section now disappears for copilot sessions under local
terminal control — web withholds onPermissionModeChange there and the
whole section then hides (HappyComposer.tsx:1595). Other flavors keep
switching while terminal-controlled.
- ChatViewModel.setPermissionMode gains the matching belt-and-suspenders
guard; the controlledByUser derivation is extracted into
terminalControlled() (controlledByUser && !concurrentClients, same as
SessionChat.tsx:996) and shared with buildConfigUi.
Tests: PermissionModeChipTest (6 cases — default/codex/copilot/pi/unknown/
safe-yolo), ChatViewModelInteractionTest +3 (withheld for terminal
copilot; survives terminal claude + shared-clients copilot; inactive
stays switchable). No new user-visible strings (chip label is catalog
en, contentDescription reuses chat_config_permission_mode).
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* test(android): derive the dsh chip expectation from the flavor catalog
The hardcoded "dsh offers no modes" assumption breaks on trees where dsh
gained the codex-style mode list — derive the expectation from the
catalog so the test pins the actual contract (the chip mirrors the
flavor's offered list).
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* feat(android): permission input semantics parity (editor inputType, URL confirm, grok auto gate)
Three behavior alignments with the web app's permission surfaces:
1. Editor inputType (web requestUserInput.ts:152-153, 191-195):
a required text question with `inputType === 'editor'` is always
answered and its note is pushed untrimmed (even the literal
"user_note: "); non-editor notes stay trimmed and optional.
Android: PermissionInputs.requestUserInputAnswerValues /
isRequestUserInputAnswered now take the question, gate on the
literal "editor" inputType; the footer's note field grows to
minLines = 2 to invite multi-line answers.
2. request_user_input URL confirmation (web requestUserInput.ts:46-70,
129-141; RequestUserInputFooter.tsx:142-152): the sign-in url only
survives when it parses as http/https; submitting requires the
CLI-synthesized __mcp_url_confirmation option (description === url)
to be selected, and submission opens the url before approving.
Android: parseRequestUserInput + normalizeRequestUserInputUrl +
isRequestUserInputUrlConfirmed; the footer enforces the confirm
guard, then opens the url via ACTION_VIEW. Divergences: the raw
wire string is compared (the CLI synthesizes the description from
the same value web would re-normalize, so they are equivalent for
every CLI-produced request), and the open-failure message is
platform-adapted (no popup blocker on Android).
3. Grok Auto permission gate (web NewSessionForm index.tsx + grok
models probe): when the machine reports
`autoPermissionModeSupported === false` for the chosen cwd, the Auto
option is disabled with an " (unavailable)" suffix, a hint line is
shown, and an existing Auto selection resets to Default. The gate is
grok-only (claude and other agents never gate) and tri-state — null
(unknown, probe pending or failed) never disables.
Android: GrokModelsResponse wire type (nullable field),
HapiApi.getMachineGrokModelsForCwd, NewSessionViewModel probes on
agent/cwd/machine changes (deduped, cancellable, existence-gated),
OptionItem gains `enabled`, the permission dropdown honors it and
renders the hint. In-session switching stays ungated, matching web.
New strings in en / zh-rCN / ru (KEYSYNC: en and ru key order
identical).
Tests: RequestUserInputTest covers editor/untrimmed-note/url
parsing/confirmation semantics; NewSessionViewModelTest covers gate
disable+reset+hint, tri-state never-gate cases, and agent/cwd gating;
HapiApiTest covers the cwd query param and nullable decoding.
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* fix(android): accept browser-valid sign-in URLs
---------
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: HAPI <noreply@hapi.run>
Co-authored-by: weishu <twsxtd@gmail.com>
* android/protocol: carry hasConversationContent through the session projection
The hub sends `hasConversationContent` on both `SessionSummary` and the
full `Session`; the sidebar's empty-stub hiding (web `SessionList.tsx`)
depends on it. Decode it on both wire models and project it through
`toSessionSummary` exactly like shared/src/sessionSummary.ts does
(`session.hasConversationContent ?? false`), so SSE upserts land with the
same signal REST lists already carry. Optional-with-default keeps older
hub payloads decodable.
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* android: session-list sidebar parity — agent-session dedup, stub hiding, relevance search
Port the web sidebar's list body to the Android session list:
- Collapse duplicate views of one agent session (resume/reopen chains)
into a single row — winner by active → selected → pin → recency —
keyed on `flavor:agentSessionId` (web `getSessionDedupKey`).
- Hide fresh auto-created "new session" stubs: inactive, conversation-
less, no title signal and no agent id (web `isSidebarEmptySessionStub`);
selected/active/pinned rows are never hidden.
- Live search box with the web's relevance ranking (field weights,
boundary + title-phrase bonuses, wildcard `*`/`?`, AND across terms)
and a distinct no-results state. Pinned rows stay contiguous while
searching.
Pure logic lives in `SessionListFilters` / `SessionListSearch` (Kotlin
objects, JVM-tested against the web reference cases); the ViewModel
pipeline is visible → dedup+stub → canonical sort → search filter+rank.
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* android: ru translations for session search parity
Upstream landed values-ru after this stack's merge base; new strings
need ru entries or :app:lintDebug fails with MissingTranslation.
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* fix(android): preserve global pins in session search
---------
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: HAPI <noreply@hapi.run>
Co-authored-by: weishu <twsxtd@gmail.com>
* feat(codex): expose Approve for me as a permission mode (#2016)
Add autoReview to the Codex permission catalog. It keeps the
workspace-write sandbox with on-request approvals and routes escalations
to Codex's native reviewer (approvals_reviewer = "auto_review"), the same
config as `hapi codex --approve-for-me`. It works at launch and mid-session
via thread/settings/update; explicit modes now reset the reviewer to user,
and sessions report autoReview from the native reviewer setting.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(native): expect Auto-review in the Codex permission lists (#2016)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(modes): label autoReview as Auto, matching Claude's auto mode (#2016)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* feat(android): settled permission verdict details
Port the web PermissionFooter verdict classification so a settled
permission row explains HOW a request was decided, not just that it was.
Behavior vs web (web/src/components/ToolCard/PermissionFooter.tsx:11-63):
- codex family sessions (flavor family + Codex/Gemini/OpenCode/Copilot/
Cursor tool prefixes): approved + decision approved_for_session shows
"Approved for session"; denied + decision abort shows "Aborted" with
the wire reason appended when present.
- claude sessions: mode acceptEdits shows "Approved: allow all edits"
(previously the raw mode id was appended); decision approved_for_session
or an allowedTools hit — plain tool name or the `Bash(command)`
qualifier built from the tool input command/cmd — shows
"Approved for session" (previously indistinguishable from a plain
approval).
- canceled now appends the wire reason (previously dropped).
- resolved/pending/denied unchanged in shape; denied keeps the error tone
and aborted shares it.
Deliberate divergence: web's own settled footer renders only a minimal
red reason box (PermissionFooter.tsx:207-220) — its formatPermissionSummary
classification is unreachable there. Android renders the classified
verdict line in the existing subdued row style instead, as intended.
Implementation: new pure JVM classifier
feature/chat/permissions/PermissionVerdict.kt (exhaustive status ->
PermissionVerdictLabel mapping, unit-tested in PermissionVerdictTest);
ToolCallBlockView.PermissionStateRow becomes a thin renderer over it.
Strings added in en/zh-rCN/ru (ru mirrors en key order; zh values from
web locales).
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* feat(android): persistent permission-mode chip + inactive gating
Item 2 — persistent permission-mode chip above the composer, mirroring
the web StatusBar mode span (StatusBar.tsx:253-265, 389-393):
- New PermissionModeChip composable + PermissionModeChipRule pure object:
resolve the raw wire mode, keep it only when the flavor offers it, and
hide `default` everywhere except copilot (model=auto sessions still
surface the chip). Unknown wire ids hide (no raw fallback).
- Tone mapping follows the web PERMISSION_TONE_CLASSES (StatusBar.tsx:
43-48): neutral=onSurfaceVariant, info=primary, warning=tertiary,
danger=error. Labels stay en catalog strings — web mode labels are
hardcoded English, not i18n.
- The chip consults the unfiltered flavor catalog: the concurrent-clients
Safe-Yolo carve-out only narrows the switcher, never the chip.
- Deliberate divergence: the web span is not clickable (its switcher is
the composer settings sheet / Shift+Tab). Android has no Shift+Tab, so
tapping the chip opens the session config sheet that hosts the
switcher — same surface as the top-bar gear, no new top-bar icon
(ChatScreen keeps its two-icons-max rule).
Item 4 — permission-mode gating aligned with web (SessionChat.tsx:
2079, 2138-2163; HappyComposer.tsx:536-537, 1544-1549, 1960-1995):
- Inactive sessions stay switchable, matching web: SessionChat passes
allowSendWhenInactive unconditionally, so configurationControlsDisabled
never trips for inactive threads and the offline notice stays
informational only (the send path auto-resumes).
- The permission section now disappears for copilot sessions under local
terminal control — web withholds onPermissionModeChange there and the
whole section then hides (HappyComposer.tsx:1595). Other flavors keep
switching while terminal-controlled.
- ChatViewModel.setPermissionMode gains the matching belt-and-suspenders
guard; the controlledByUser derivation is extracted into
terminalControlled() (controlledByUser && !concurrentClients, same as
SessionChat.tsx:996) and shared with buildConfigUi.
Tests: PermissionModeChipTest (6 cases — default/codex/copilot/pi/unknown/
safe-yolo), ChatViewModelInteractionTest +3 (withheld for terminal
copilot; survives terminal claude + shared-clients copilot; inactive
stays switchable). No new user-visible strings (chip label is catalog
en, contentDescription reuses chat_config_permission_mode).
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
* test(android): derive the dsh chip expectation from the flavor catalog
The hardcoded "dsh offers no modes" assumption breaks on trees where dsh
gained the codex-style mode list — derive the expectation from the
catalog so the test pins the actual contract (the chip mirrors the
flavor's offered list).
Co-Authored-By: Claude Code <noreply@anthropic.com>
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
---------
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: HAPI <noreply@hapi.run>
Port the web PermissionFooter verdict classification so a settled
permission row explains HOW a request was decided, not just that it was.
Behavior vs web (web/src/components/ToolCard/PermissionFooter.tsx:11-63):
- codex family sessions (flavor family + Codex/Gemini/OpenCode/Copilot/
Cursor tool prefixes): approved + decision approved_for_session shows
"Approved for session"; denied + decision abort shows "Aborted" with
the wire reason appended when present.
- claude sessions: mode acceptEdits shows "Approved: allow all edits"
(previously the raw mode id was appended); decision approved_for_session
or an allowedTools hit — plain tool name or the `Bash(command)`
qualifier built from the tool input command/cmd — shows
"Approved for session" (previously indistinguishable from a plain
approval).
- canceled now appends the wire reason (previously dropped).
- resolved/pending/denied unchanged in shape; denied keeps the error tone
and aborted shares it.
Deliberate divergence: web's own settled footer renders only a minimal
red reason box (PermissionFooter.tsx:207-220) — its formatPermissionSummary
classification is unreachable there. Android renders the classified
verdict line in the existing subdued row style instead, as intended.
Implementation: new pure JVM classifier
feature/chat/permissions/PermissionVerdict.kt (exhaustive status ->
PermissionVerdictLabel mapping, unit-tested in PermissionVerdictTest);
ToolCallBlockView.PermissionStateRow becomes a thin renderer over it.
Strings added in en/zh-rCN/ru (ru mirrors en key order; zh values from
web locales).
via [HAPI](https://hapi.run)
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: HAPI <noreply@hapi.run>
* feat(android): add Russian locale
Translate all 518 app strings into values-ru/strings.xml, register ru in
locales_config.xml, and add AppLanguage.RUSSIAN to the in-app language
picker.
* fix(android): use count-agnostic Russian forms for dynamic counts
The *_many strings used a genitive-plural noun after the number, which is
wrong for 2-4 (e.g. "2 инструментов"). The house style avoids <plurals>,
so phrase these labels with the noun first ("Инструменты: %1") instead:
tool groups, review findings, queued/new messages, agent steps, questions,
lines, chars, edits, usage requests, scratchlist count and attachment cap.
* fix(android): make the process-steps label count-safe
* fix(android): use a count-safe token label in the usage tooltip
---------
Co-authored-by: mrudy <mrudy@local>
Add current-proposal decoding, dedicated implementation requests and composer-only continuation on iOS and Android. Keep pending/errors across row recycling, reject stale actions and prevent duplicate submissions.
Include native regression tests and localized plan controls.
Render ExitPlanMode and exit_plan_mode Markdown from input.plan in iOS and Android conversations. Preserve approvals, raw source and diagnostics while hiding empty output placeholders and prewarming plan documents.
Add generated protocol fixtures and native regression coverage for long plans, live updates, recycling, themes and typography.
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.
* refactor(web): route create-form permission control through one native-select predicate
Extract usesNativePermissionSelect(flavor) (grok || codex-family, matching
the existing iOS/Android predicate of the same name) and route
PermissionField's select-vs-toggle gate through it instead of an inline
condition. Rename the codex-family-only state codexFamilyPermissionMode to
nativePermissionMode since it now backs a shared predicate, not just the
codex family. Behavior is unchanged: any stale sessionStorage draft written
under the old codexFamilyPermissionMode key has no value under the new key
and falls back to 'default', which only matters within a single browser
tab's lifetime.
* feat(web): let Claude pick a permission mode when creating a session
Claude was the only create-form flavor still on the global HAPI YOLO toggle
while grok and the codex family got the native permission select, so there was
no way to start a session in Plan Mode without creating it first and switching
the mode from the composer. usesNativePermissionSelect now gates the control
for claude as well, and the spawn body carries permissionMode (including
'default') instead of yolo, which is the shape the other native-select flavors
already send.
The stored hapi:newSession:yolo preference is bridged into the select rather
than dropped, but only for the flavors that have actually moved onto it
(LEGACY_YOLO_BRIDGE_AGENTS = codex, claude). copilot, gemini, kimi and opencode
moved earlier and settled on 'default'; re-enabling Yolo for them now would
widen permissions rather than migrate a preference. This narrows the
sessionStorage draft bridge too, which until now fired for the whole codex
family with no allow-list, so their draft restores yield 'default' instead of
'yolo' — same-tab-lifetime state only.
Claude and the codex family share one nativePermissionMode state and their mode
sets do not overlap, so the existing agent-change reset plus the flavor filters
in the draft loader and the stored launch settings are what keep a codex mode
out of a Claude spawn. Adds the regression test that pins it: pick a mode under
codex, switch to Claude, create, assert the payload carries 'default'.
* feat(ios): let Claude pick a permission mode when creating a session
Extend usesNativePermissionSelect to include claude alongside grok and the
codex family, matching the web change. buildSpawnRequest now derives both
yolo and permissionMode from that single predicate instead of two separate
local flags, so claude sends permissionMode (including 'default') and no
longer sends yolo. Unlike web, iOS carries no persistent YOLO preference
across sessions to migrate — the toggle only lives in the in-memory form or
a draft deleted on success — so there is no bridging logic to add here.
* feat(android): let Claude pick a permission mode when creating a session
Extend usesNativePermissionSelect to include claude alongside grok and the
codex family, matching the web and iOS changes. buildSpawnRequest derives
both yolo and permissionMode from that single predicate, so claude sends
permissionMode (including 'default') and no longer sends yolo. Unlike web,
Android has no persistent YOLO preference to migrate: the toggle only lives
in the form draft, which is deleted once a session is created.
The agent-switch test asserted claude renders the YOLO toggle; it now checks
the native select for claude and keeps the toggle assertion on cursor, which
still carries it.
Expedited WorkManager requests needed an FGS fallback on API 26-30,
which dragged in FOREGROUND_SERVICE + FOREGROUND_SERVICE_DATA_SYNC and
the Play Console foreground-service declaration that comes with them.
Not worth it: gate setExpedited on API 31+ (JobScheduler path, no FGS),
let 26-30 enqueue as plain work, delete the getForegroundInfo overrides
and the worker notification, and strip WorkManager's library-merged
FOREGROUND_SERVICE with tools:node=remove. Merged manifest verified
FGS-free.
* 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
With the row down to two lines the meta became the sole secondary line
and the project scan key, but it wore a label role: 11sp with 0.5sp
tracking on Android (stringy on path-like text), 12pt caption on iOS.
Promote it to bodySmall / footnote — title-to-meta contrast lands at
~1.3, matching the web sidebar's 14/12. Also move the AI summary
directly under the title (its prose continuation) so the meta closes
the row as a footer instead of splitting the two text blocks.
Rows went from three text lines to two but kept the 10dp vertical
padding, so the unchanged 20dp between items read as oversized gaps
around the now-shorter rows (device-measured 69px inter vs 17px intra,
1:4 — the old layout sat near 1:3). 8dp brings the proportion back
while keeping rows comfortably above the touch-target minimum.
Rows carried a machine-only line plus the full absolute path (prefix
noise, tail-truncated exactly where the information lives, machine
repeated under the filter chips). Replace both with a single meta line:
the last two segments of the worktree base path (session path fallback
— the web sidebar's group-name rule), the worktree name when present,
and the machine label only while several machines are known with no
machine filter active. The subtitle now carries the AI summary or
nothing; full paths stay in the session detail. Typical rows shrink
from three or four lines to two.
A hub where most sessions stay connected turns a green online dot into
noise: an indicator that is almost always in one state carries no
information, and a column of saturated green outshouts the markers that
matter (pending approval, unread). Align both natives with the web
sidebar semantics: no leading status dot, disconnected rows render at
half opacity, and a small green spinner appears after the title only
while a turn is in flight. Android's StatusIndicator is removed; iOS
keeps StatusDot for the chat header.
Four trailing icons left no room for the session title (device
feedback). The chat top bar now shows two: gear (config sheet) plus one
menu holding Session files and Scratchlist (with entry count) ahead of
the existing Rename/Reopen/Delete/Park entries. Drops the standalone
scratchlist badge buttons on both platforms.
With the default soft-input mode the AppCompat subdecor also resizes the
window for the IME, stacking a keyboard-sized gap on top of imePadding
(device-observed on the chat composer). SOFT_INPUT_ADJUST_NOTHING on
API 30+ leaves Compose as the single keyboard-inset owner; 26-29 keep
adjustResize because the ime() inset backport depends on it.
- decode fs.stat-derived epoch fields leniently (fractional mtimeMs from
real hubs broke machines/files decode and pinned the offline banner)
- offline banner: only a failed sessions fetch counts; machines/baseline
failures are advisory; live SSE emission clears it and seeds the unread
baseline (all-rows-unread gray dot cascade)
- session row: single weighted title (short names no longer truncated at
half width), unread dot moved beside the timestamp
- composer restyle: borderless input pill with inline mic, hand-drawn
vector glyphs replacing emoji icons, 42dp round actions, park-draft
relocated to the session overflow menu
Extraction sweep: every user-visible hardcoded English string in
android/app moved to values/strings.xml with feature-prefixed keys
(sessions_/chat_/files_/scratchlist_/pairing_/new_session_/tool_...);
values-zh-rCN/strings.xml translates all 448 keys, terminology aligned
with web/src/lib/locales/zh-CN.ts.
ViewModels stay string-free: transient notices became semantic sealed
types resolved at the UI layer (ChatNotice, ScratchlistNotice +
ScratchlistImportRejection, PairingError, DictationErrorKind,
SessionListError.DeleteFailed.stillActive); ViewModels that genuinely
compose display text take small Strings seams with English defaults
for JVM tests (FilesStrings, FileViewerStrings, NewSessionStrings);
toolCardPresentation gains a Resources parameter.
Language switching: LanguagePrefs gains SYSTEM (follow system, the new
default); Settings applies AppCompatDelegate.setApplicationLocales
immediately; MainActivity now extends AppCompatActivity over
Theme.AppCompat.DayNight.NoActionBar with autoStoreLocales +
android:localeConfig; FCM/WorkManager surfaces wrap the application
context via localizedForAppLanguage (per-app locales miss non-activity
contexts below API 33).
Verification: :app:assembleDebug green, :app:lintDebug 0 errors
(MissingTranslation clean), full JVM test gate green (176 app tests;
notice assertions updated to the semantic types).
