The previous cache was a synchronizedMap(LinkedHashMap(access-order=true)).
Access-order LRU rewrites the linked list on get(), so every cache hit
took the global monitor — at 700 concurrent relay reconnects, the lock
serialized every dispatcher worker through a single critical section.
Switch to ConcurrentHashMap so reads are lock-free. Amethyst's steady
state is well under maxEntries (~750 distinct hosts vs cap of 2000), so
strict LRU was never going to evict anything and the access-order
machinery was pure contention.
Bump positive TTL from 5min to 24h: relay and CDN IPs change on the
order of days, and we are not a recursive resolver — there's no
correctness reason to honor authoritative TTLs. A 5min cycle made us
re-resolve all 700 relays every 5 minutes.
Also add a leader-side cache re-check after acquiring inflight
ownership. If a peer leader refreshed the entry while we were claiming
the slot, we skip getaddrinfo entirely.
Eviction is now a cheap on-demand sweep of expired entries when the
map exceeds maxEntries; in normal use it never runs.
A Nostr feed can touch hundreds of distinct hosts (Blossom servers,
relays, NIP-05 domains, image proxies). The 256-entry cap was small
enough that active users would churn the LRU and pay extra getaddrinfo
calls. 2000 still costs <100KB of heap.
Avoids paying the getaddrinfo tax on every HTTP call to the same
host. Adds a single-flight, LRU+TTL DNS resolver wired into both the
media and relay OkHttp clients via a process-wide shared instance, so
resolutions cross-cut images, relays, and NIP-05 lookups.
- Per-host coalescing: N concurrent lookups for the same host share one
upstream call.
- Different hosts proceed in parallel — no global lock around the
upstream resolver.
- Negative cache (10s) prevents hammering on typos / dead hosts.
- Positive cache (5m) survives a feed scroll.
Casting LocalContext.current to NestActivity tripped the Jetpack
Activity Compose lint rule ContextCastToActivity, failing
:amethyst:lintPlayBenchmark in CI.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Wire the local user's avatar into the existing speaking-ring animation so
the host/speaker sees the same green ring + amplitude glow on their own
cell that remote speakers already get.
Ground truth is the broadcaster's `publisher.send(opus)` success: the new
`onLevel` callback fires only after a frame actually leaves on the wire,
computed from the raw PCM peak (shared `peakAmplitude` util used by both
the player decode loop and the broadcaster capture loop). This means the
animation reflects what the relay sees, not any UI-side mute / role
state that could be stale or buggy — if frames are flowing, the ring
plays.
Plumbing: `NestsSpeaker.startBroadcasting` gains an optional `onLevel`
that the moq-lite + IETF broadcasters both honour, the reconnecting
wrapper replays it on every session reissue (alongside the existing
mute-intent replay), and `NestViewModel.startBroadcast` hands it a
lambda that calls the existing `onSpeakerActivity` / `onAudioLevel`
plumbing keyed on the local pubkey. The 250 ms speaking-timeout fades
the ring naturally when the user mutes or stops broadcasting.
Three more leave paths now fully release the speaker session and
listener (and with them AudioRecord + the system mic indicator)
before the Activity destruction queue eats viewModelScope:
- PIP overlay's Leave action — pipReceiver invokes a cleanup
callback set by NestActivityBody, then finish().
- PIP swipe-dismiss — NestActivity.onStop() invokes the same
callback when in PIP and isFinishing. Non-PIP onStop stays a
no-op so backgrounding doesn't tear down.
- Host's Just leave — calls viewModel.leave() before onLeave().
Unlike Close the Room, this still leaves the kind-30312
meeting space open for other users.
The cleanup callback is registered via DisposableEffect in
NestActivityBody so it's auto-cleared when the composable leaves.
The host-leave confirmation dialog's Close the Room button only called
finish(), relying on VM.onCleared() to release the AudioRecord. That
runs late in the destroy lifecycle, so the system mic-in-use indicator
stayed lit while the activity was queued for destruction.
Adds NestViewModel.leave(), which mirrors onCleared() (sets closed,
runs both teardowns with finalCleanup=true so closes route through
cleanupScope/GlobalScope and survive Activity destruction). Wires it
into the Close the Room callback only — the Just Leave and non-host
Leave paths are unchanged per request.
After tapping Leave the Stage, the bar lingered because it was gated
solely on isOnStageMe — a value derived from the presence aggregator.
That aggregator only updates after a presence event signs, broadcasts
to relays, and loops back through LocalCache. With the recent
broadcast / mute traffic contending on the signer plus the 500 ms
mute debounce in NestPresencePublisher, the round-trip ran long
enough that the user perceived Leave the Stage as not firing.
AND the visibility gate with ui.onStageNow (the local intent flag,
which flips synchronously on setOnStage(false)) so the tap hides the
bar on the next frame. Host demotion still hides it via the existing
isOnStageMe path.
The kebab on a NoteCompose card opens the comprehensive
NoteDropDownMenu (Follow/Unfollow, Copy text/pubkey/note id, Share,
Edit, Broadcast, Timestamp, Pin, Bookmark, Delete/Report) — not the
smaller LongPressToQuickAction popup. The previous nests commit
wired the kebab to the wrong menu.
Replace the manual ClickableBox + VerticalDotsIcon + showQuickAction
plumbing with the existing MoreOptionsButton composable, which
already wraps the icon and triggers NoteDropDownMenu for both card
variants. Long-press still opens NoteQuickActionMenu via the outer
LongPressToQuickAction wrapper, matching NoteCompose's two-tier
gesture model exactly.
Drops the onMore parameter threading on ObserveAndRenderSpace,
RenderLiveSpacesThumb, and SpaceHostAndReactions — MoreOptionsButton
is self-contained, so the trailing optional parameter is no longer
needed.
Long-press on the cards already opened the standard
NoteQuickActionMenu, but the affordance was hidden — especially on
the compact ENDED card which has no visible interactive elements.
Add an explicit MoreVert (VerticalDotsIcon) tap target that opens
the same menu:
- ENDED compact card: 40dp icon on the trailing edge of the row.
- LIVE/SCHEDULED full card: 35dp icon in the existing
SpaceHostAndReactions row, after Like and Zap.
Both wire through the showQuickAction lambda from
LongPressToQuickAction, so the icon and the long-press gesture
trigger the exact same menu (share, copy, broadcast, report,
block, delete) — no second popup-state plumbing.
ObserveAndRenderSpace, RenderLiveSpacesThumb, and
SpaceHostAndReactions gain an optional `onMore: (() -> Unit)?`
parameter so other callers (e.g. existing live-stream surfaces
that reuse SpaceHostAndReactions) keep their current rendering
unchanged when they don't pass one.
The previous commit placed the controls strip as a sibling row right
after StageGrid. Visually it read as a separate band, not as part of
the stage card.
Adds a bottomBar slot to StageGrid that renders inside the same
Surface, below the grid, sharing the card's tonal background and
horizontal padding. StageControlsBar drops its own horizontal padding
since the card supplies it, and keeps a top inset so it has breathing
room from the avatars.
Replace the flat NestsScreen list with three explicit sections so the
audio-room surface separates actionable rooms from historic ones:
- LIVE: ordered by follows-participating DESC, total-participants DESC,
createdAt DESC. PRIVATE rooms continue to live in the LIVE bucket
since they're audible once a join token is granted.
- SCHEDULED: ordered by `starts` ASC (soonest first). PLANNED rooms
now have their own bucket instead of collapsing into ENDED.
- ENDED: ordered by createdAt DESC (proxy for "ended at"), capped to
the last 7 d, and rendered with a compact one-line card variant
(square thumb + name + host + "Ended Xh ago"). Tapping still
routes through the read-only lobby so recordings remain reachable.
The DAL exposes `NestsFeedFilter.bucketOf()` so the screen can walk
the pre-sorted feed once and inject sticky-header sections without
re-sorting on each recomposition.
Cards now wrap in `LongPressToQuickAction`, giving share / copy /
delete / report / block / broadcast on long press — matching the
NoteCompose interaction model and giving hosts a path to delete
(kind:5) ended rooms without entering the room first.
NestActionBar held the speaker controls (Talk / MicMute / Leave the
Stage) alongside the audience controls (hand-raise / react / leave
room), making the bar uncomfortably wide on phones once a user was
promoted.
Splits responsibilities:
- NestActionBar now only handles connection state in the start
cluster (Connect / Connecting chip / Reconnecting chip / empty)
plus the existing end cluster.
- StageControlsBar is a new strip rendered under the Stage card,
gated only on isOnStage, holding the full broadcast state machine.
isOnStage is the single visibility signal — a stable derived value
from the kind-30312 role + presence onstage flag — so connection
blips, broadcast state churn, mute toggles, and permission flows
never make the new bar appear or disappear. It only flips on a real
promote/demote or the user tapping Leave the Stage.
`AudioTrackPlayer` was constructing its `AudioTrack` with
`USAGE_VOICE_COMMUNICATION` + `CONTENT_TYPE_SPEECH` so the room would
behave like a phone call (volume rocker controls call volume, ducks
notifications). The `AudioTrack` then renders on `STREAM_VOICE_CALL`,
which Android only services audibly while the device is in
`MODE_IN_COMMUNICATION`. Nests never drives `AudioManager.mode` (only
NIP-100's `CallAudioManager` does), so on most devices the playback
either dropped to the earpiece at near-zero volume or produced no audio
at all — making rooms appear silent on both phones.
Fix: route through `USAGE_MEDIA` + `CONTENT_TYPE_SPEECH` (=
`STREAM_MUSIC`) so audio-room playback comes out of the loudspeaker by
default and the volume rocker controls it any time. Same approach
Twitter/X Spaces and Clubhouse take for hands-free audio rooms. Echo
cancellation still works on the capture side via
`MediaRecorder.AudioSource.VOICE_COMMUNICATION` regardless of the
playback usage.
Also update `NestForegroundService.requestAudioFocus` to request focus
under the matching `USAGE_MEDIA` attributes so the focus claim and the
playback attributes line up.
MicMuteToggle already covers 'stop sending audio while staying on stage'
with sample-accurate resume, and LeaveStageButton covers full teardown.
StopBroadcastButton drew a Mic icon next to MicMuteToggle's Mic icon,
reading visually as two adjacent mic buttons.
ScriptedSpeaker publishes startCount > 0 from inside startBroadcasting,
so polling startCount alone races the wrapper pump's subsequent
`if (desiredMuted) handle.setMuted(true)` step. Under load (full suite
run on CI) the assertion fired before the pump replayed mute intent,
producing setMutedCalls=0. Wait for the post-condition the assertion
checks instead.
Audit-driven UI fixes that don't fit the perf bucket:
* NestActionBar — wire StopBroadcastButton between MicMuteToggle and
LeaveStageButton while broadcasting. The doc table at the top of
the file already advertised [Mute] [Stop] [Leave the Stage] but the
button was defined and never used; speakers had to leave the stage
to stop sending audio.
* NestActivityContent — render an explicit loading spinner while the
kind-30312 resolves and a clear "this room can't be opened" page
with a Back button when the event is missing service/endpoint.
Replaces the previous black-screen dead end.
* HandRaiseQueueSection — wrap filter+sort in remember so a heartbeat
doesn't rebuild the queue on every recompose, swap to LazyColumn
so a long queue doesn't render every row every frame, and show
UsernameDisplay instead of the raw 8-char hex prefix.
* CreateNestSheet schedule picker — compute the timezone offset at
the picked instant rather than at Instant.now(); the previous
variant was off by one hour for rooms scheduled across a DST
transition.
* NestViewModelFactory — pass OkHttpNestsClient's httpClient via
named parameter so the (String) -> OkHttpClient function reference
doesn't get bound to the Long callTimeoutMs slot. Fixes the
pre-existing :amethyst:compilePlayDebugKotlin failure on the
branch base.
Adds three strings (loading, unjoinable title/body, back).
https://claude.ai/code/session_01DMeCvWyBYVVVPez2hwqCs4
NestBridge holds a process-singleton AccountViewModel reference so
NestActivity (separately tasked) can read the active account without
serialising a NostrSigner through an Intent extra. The bridge's doc
promised it'd be cleared on logout / account switch but no call site
existed — so the audio-room activity could pick up a stale
AccountViewModel from the previous session and sign with the old
key after an account swap.
Wire NestBridge.clear() into both transition points in
AccountSessionManager: switchUserSync (always) and logOff (only when
the current account is being torn down).
https://claude.ai/code/session_01DMeCvWyBYVVVPez2hwqCs4
Feed thumbnails were reusing NestRoomFilterAssemblerSubscription to
flip the LIVE/ENDED badge based on the freshest kind-10312 presence.
That assembler subscribes to chat + presence + reactions (kinds 1311,
10312, 7) — so a feed of N rooms paid each popular room's full
chat-history pull per outbox relay just to render a status pill.
Add NestRoomLivenessAssembler, scoped to a single kinds=[10312],
#a=[room], limit=1 filter per relay, and switch the feed thumbnail
to it. The full chat / reaction / admin command REQ stays gated
behind the join flow.
https://claude.ai/code/session_01DMeCvWyBYVVVPez2hwqCs4
Audit-driven fixes for the audio-room reactions overlay:
1. RoomReactionsAggregator now dedups by event id. The previous code
appended every event to a flat list, but LocalCache.observeEvents
re-emits the full matching list on every cache mutation — so an
N-reaction window grew quadratically and the overlay rendered the
same emoji once per replay. Keyed by event id collapses re-emits.
2. RoomPresenceAggregator gains applyOrNull that returns null when an
incoming heartbeat is older-or-equal to the cached presence; the VM
skips the StateFlow write in that case. In a 200-peer room, every
replay used to copy a 200-entry map plus run an O(N) StateFlow
equality check 200 times per emission. Now it's one-and-skip.
3. ReactionsEvictionTicker only ticks while the aggregator is non-empty.
Once the last reaction expires the loop self-cancels until the next
reaction lands — a quiet room costs no scheduled work.
4. REACTION_WINDOW_SEC dropped 30s -> 10s. Reactions are about what
the speaker is saying right now; a 10s window keeps the overlay
timely instead of bleeding into the next paragraph.
5. SpeakerReactionOverlay drives a per-chip lifecycle animation: each
chip drifts upward ~16dp and fades over the 10s window. A fresh
reaction (same emoji) restarts the chip's animation. The
AnimatedVisibility outer entry/exit still smooths first-arrival and
final disappearance.
Tests: added a re-emit dedup case to RoomReactionsStateTest plus an
end-to-end replay assertion in NestViewModelTest. Existing tests
updated to use unique event ids.
https://claude.ai/code/session_01DMeCvWyBYVVVPez2hwqCs4
- NestsTokenResponse.parse catch list now includes IllegalStateException
so a malformed escape from a misbehaving auth server surfaces as
NestsException instead of crashing the listener.
- OkHttpNestsClient gains a configurable callTimeoutMs (default 90 s)
enforcing an upper bound on the entire mintToken round-trip including
retries. Without it a stalled server suspends the reconnect
orchestrator indefinitely (the orchestrator's openOnce step parks on
mintToken). 90 s leaves headroom over the worst-case 63 s 429
retry chain documented in MAX_RATE_LIMIT_RETRIES.
- parseEndpoint tightens IPv6 bracket check from `closeBracket > 0` to
`> 1`, rejecting the empty `[]` literal.
- NestBroadcaster + NestMoqLiteBroadcaster gain an `onTerminalFailure`
callback that fires once when the consecutive-send-error guard bails.
MoqLiteNestsSpeaker wires this to flip the speaker state to Failed,
giving ReconnectingNestsSpeaker the signal it needs to recycle the
session — without this hook the broadcaster bailed silently and the
outward speaker state stayed on Broadcasting forever.
Adds regression test:
- onTerminalFailure_fires_once_after_consecutive_send_failures
225 tests pass, 0 failures. Android target compiles clean.
A fresh audit caught a real race in the round-1 subscribe() rewrite plus
several distinct bugs in the Android MediaCodec layer.
**MoqLiteSession.subscribe() race regression** — the launched collector
that reads the SubscribeResponse and watches for peer FIN ran cleanup
(remove from subscriptionsBySubscribeId, close frames) before subscribe()
itself reached the post-await registration. If the publisher FIN'd
immediately after sending Ok, the collector exited first against an
empty map; subscribe() then registered the subscription, leaving frames
in the map with no live collector to ever close it on transport drop.
Consumer hung forever — exactly the failure mode round-1 fix#3 was
supposed to prevent. Fix: pre-register the subscription BEFORE launching
the collector. Drop unused `ok` field from ListenerSubscription.
**MoqLiteNestsSpeaker.startBroadcasting publisher leak** — if
session.publish() succeeded but broadcaster.start() then threw (mic
permission denied, AudioRecord allocation failure), the publisher was
never closed and stayed registered as session.activePublisher,
permanently blocking subsequent startBroadcasting calls.
**runCatching{suspend close} swallowing CancellationException** —
MoqLiteNestsSpeaker.close, MoqLiteBroadcastHandle.close,
MoqLiteNestsListener.close all wrapped suspending closes in
runCatching, breaking structured cancellation when the parent scope
cancelled teardown. Replaced with explicit cancel-rethrowing try/catch.
**MediaCodecOpusDecoder ArrayList<Short> boxing on every audio frame**
— at 50 fps × 960 samples × N speakers ≈ 48 000 boxed Short
allocations/sec/speaker on the audio hot path. Rewritten to write
directly into a pre-sized ShortArray via ShortBuffer.get(dst, off, len).
**MediaCodecOpusEncoder bugs**
- KEY_AAC_PROFILE = AACObjectLC was being set on an audio/opus
encoder. Meaningless for Opus; stricter Codec2 stacks on Android
13+ reject the configure() call with IllegalArgumentException and
surface as DeviceUnavailable. Removed.
- The drain loop's INFO_OUTPUT_FORMAT_CHANGED branch had no progress
guard. A buggy encoder re-emitting FORMAT_CHANGED without producing
output would busy-spin against the 10 ms dequeue timeout. Now
absorbed at most once per encode call. Same guard added to the
decoder.
Adds regression test:
- frames_flow_completes_when_peer_FINs_immediately_after_Ok
224 tests pass, 0 failures. Android target compiles clean.
Claude