Disable recorder warm-up to prevent coreaudiod CPU spike #139

benjaminfrombe · 2026-01-09T09:28:44Z

Summary

Disable launch-time recorder warm-up that keeps a CoreAudio client alive
Keep sound-effect preloading lightweight (no engine start)
Add changeset for user-facing CPU fix

Testing

Manual: launch app, verify coreaudiod CPU remains low until first recording

Notes

The CPU spike appears immediately on app start due to recorder priming; disabling warm-up removes the idle CoreAudio IO thread.

Fixes #137

Warm-up priming keeps a CoreAudio client alive and can drive coreaudiod CPU to ~10% at app start. Remove the launch-time warm-up and keep sound effect preloading lightweight so the app stays idle until the first recording.

coderabbitai · 2026-01-09T09:28:54Z

📝 Walkthrough

Walkthrough

Recorder warm-up was disabled and recorder lifecycle handling was changed: RecordingClient now stops and destroys the AVAudioRecorder and delays before priming a new recorder; TranscriptionFeature startup no longer warms up the recorder; SoundEffect now tracks active plays and enables AVAudioEngine auto-shutdown.

Changes

Cohort / File(s)	Summary
Changeset Documentation `\.changeset/silent-brisk-vivid.md`	Adds a patch-level changeset entry documenting disabling recorder warm-up to avoid coreaudiod CPU spike.
Recorder Lifecycle Management `Hex/Clients/RecordingClient.swift`	Stop-and-destroy recorder on stopRecording, delete recordings, add ~200ms post-stop delay before priming a new recorder, enable metering only when ensuring readiness, add debug logs for creation/warm-up/cleanup, update priming/state messages.
Audio Playback & Engine Management `Hex/Clients/SoundEffect.swift`	Add playback tracking (`activePlays`, tokens), schedule buffer completions to decrement counts, helpers to handle playback end and stop engine if idle, enable AVAudioEngine auto-shutdown, remove immediate engine prepare in preload.
Feature Startup Orchestration `Hex/Features/Transcription/TranscriptionFeature.swift`	Remove recorder warm-up from startup effects (warmUpRecorderEffect removed); startup effects now exclude recorder warm-up.

Sequence Diagram(s)

sequenceDiagram
  actor User
  participant TranscriptionFeature
  participant RecordingClient
  participant CoreAudio as coreaudiod

  User->>TranscriptionFeature: trigger transcription start (hotkey)
  TranscriptionFeature->>RecordingClient: startMetering / startRecording
  RecordingClient->>coreaudiod: open AVAudioRecorder (create)
  RecordingClient-->>TranscriptionFeature: recorder ready

  User->>RecordingClient: stopRecording
  RecordingClient->>coreaudiod: recorder.stop() and deleteRecording()
  RecordingClient->>RecordingClient: wait 200ms
  RecordingClient->>coreaudiod: prime new recorder (create fresh instance)
  RecordingClient-->>TranscriptionFeature: primed for next session

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Poem

🐰 I used to warm up, hum and prance,
Now I hush the recorder, give CoreAudio a chance,
Engines nap softly, tokens counted true,
A quieter meadow — I hop back to you. 🎶

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Out of Scope Changes check	⚠️ Warning	Sound effect playback accounting changes (token tracking) are out of scope relative to issue #137's core focus on recorder warm-up and CPU spike mitigation.	Consider separating sound effect token tracking improvements into a distinct PR focused on preventing double-decrement accounting bugs in audio playback.
Docstring Coverage	⚠️ Warning	Docstring coverage is 11.76% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately describes the main objective of the PR: disabling recorder warm-up to prevent a coreaudiod CPU spike.
Linked Issues check	✅ Passed	The PR addresses the primary objective from issue #137 by disabling recorder warm-up at launch to prevent CoreAudio client persistence and the associated CPU spike, though it does not fully resolve the broader ~10% CPU leak issue.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

📝 Generate docstrings

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

coderabbitai

Actionable comments posted: 1

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

Hex/Features/Transcription/TranscriptionFeature.swift (1)
79-87: Stale comment references removed effect.

The comment on lines 81-83 still mentions "3) Priming the recorder for instant startup" but only two effects remain in the merge. Update the comment to reflect the current implementation.
📝 Suggested fix
       case .task:
-        // Starts two concurrent effects:
+        // Starts concurrent effects:
         // 1) Observing audio meter
         // 2) Monitoring hot key events
-        // 3) Priming the recorder for instant startup
         return .merge(
           startMeteringEffect(),
           startHotKeyMonitoringEffect()
         )

🤖 Fix all issues with AI agents

In @Hex/Clients/SoundEffect.swift:
- Around line 95-99: The stop(_ soundEffect: SoundEffect) method currently
decrements activePlays immediately which can double-decrement when the play()
completion handler later calls handlePlaybackEnded(); remove the activePlays
decrement from stop(_:) and let handlePlaybackEnded() be solely responsible for
decrementing activePlays, or alternatively implement per-sound play state (e.g.,
a Set or Bool flag keyed by SoundEffect in playerNodes) and guard against
decrementing twice by checking that flag in both stop(_:) and
handlePlaybackEnded() before modifying activePlays; update stop(_:) and
handlePlaybackEnded() accordingly to use that per-sound state if choosing the
second option.

📜 Review details

Configuration used: defaults

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between a90465c and ada8a02.

📒 Files selected for processing (4)

.changeset/silent-brisk-vivid.md
Hex/Clients/RecordingClient.swift
Hex/Clients/SoundEffect.swift
Hex/Features/Transcription/TranscriptionFeature.swift

🧰 Additional context used

📓 Path-based instructions (5)

**/*Feature.swift

📄 CodeRabbit inference engine (CLAUDE.md)

Use The Composable Architecture (TCA) for state management; implement features as TCA reducers

Files:

Hex/Features/Transcription/TranscriptionFeature.swift

**/{App,Transcription,Settings,History}Feature.swift

📄 CodeRabbit inference engine (CLAUDE.md)

Maintain the core features as AppFeature, TranscriptionFeature, SettingsFeature, and HistoryFeature

Files:

Hex/Features/Transcription/TranscriptionFeature.swift

**/*.swift

📄 CodeRabbit inference engine (.cursor/rules/composable-architecture-tree-navigation.mdc)

**/*.swift: Model presented child features in parent state using the Presents() macro with optional child state (non-nil presents, nil dismisses)
Expose child feature actions in the parent as PresentationAction<Child.Action> (e.g., case child(PresentationAction<Child.Action>))
Integrate child reducers into parents using .ifLet(.$child, action: .child) { ChildFeature() } to compose optional-driven navigation
Drive navigation by populating optional presentation state to present and setting it to nil to dismiss
In SwiftUI views, use @bindable var store: StoreOf and pass a binding store to SwiftUI modifiers like .sheet(item:), .popover(item:), .navigationDestination(item:) via $store.scope(state:action:)
Prefer a single enum-based destination over multiple optional @presents properties when a feature can navigate to multiple screens
Define a nested @Reducer enum Destination with cases for each child feature, hold @presents var destination: Destination.State?, and route actions via case destination(PresentationAction<Destination.Action>)
Compose destination enum into the parent with .ifLet(.$destination, action: .destination)
Scope SwiftUI bindings to specific destination cases using $store.scope(state: .destination?.case, action: .destination.case) with .sheet/.popover/.navigationDestination
For platforms earlier than iOS 16/macOS 13/tvOS 16/watchOS 9, use a NavigationLink extension that drives navigation from Binding<D?> with appropriate @available annotations
Dismiss from reducers using @dependency(.dismiss) and invoke await dismiss() inside an effect (.run) rather than directly in the reducer
Do not send additional actions after invoking dismiss() within the same effect

**/*.swift: Use Swift Composable Architecture (TCA) for state management, effects, and reducers throughout the macOS app
Target Swift 6 language features and compatibility across the codebase
Use Swift async/await for asynchronous code instead of callback-based patterns

Files:

Hex/Features/Transcription/TranscriptionFeature.swift
Hex/Clients/SoundEffect.swift
Hex/Clients/RecordingClient.swift

**/SoundEffect.swift

📄 CodeRabbit inference engine (CLAUDE.md)

Play audio feedback through SoundEffect utilities

Files:

Hex/Clients/SoundEffect.swift

**/RecordingClient.swift

📄 CodeRabbit inference engine (CLAUDE.md)

Use RecordingClient (AVAudioRecorder wrapper) for audio capture instead of direct AVAudioRecorder usage

Files:

Hex/Clients/RecordingClient.swift

🧠 Learnings (8)

📓 Common learnings

Learnt from: CR
Repo: kitlangton/Hex PR: 0
File: CLAUDE.md:0-0
Timestamp: 2025-09-10T12:14:22.927Z
Learning: Applies to **/RecordingClient.swift : Use RecordingClient (AVAudioRecorder wrapper) for audio capture instead of direct AVAudioRecorder usage

Learnt from: CR
Repo: kitlangton/Hex PR: 0
File: CLAUDE.md:0-0
Timestamp: 2025-09-10T12:14:22.927Z
Learning: Applies to **/TranscriptionClient.swift : Use WhisperKit via TranscriptionClient for ML transcription (tracking WhisperKit main branch)