Reduce main-thread CPU in AudioEngine capture path

[ysdede]

Context

• Audio capture drives AudioEngine.handleAudioChunk on the main thread, doing resample checks, energy + SMA, VAD (AudioSegmentProcessor), visualization summary, segment extraction, and subscriber callbacks.
• Capture is active even when model is not ready (v4 path), so this work runs continuously.

Evidence

• Main thread active time in trace: main_thread.total_dur_ms = 21958.8ms over 29.7s.
• Hot path locations: src/lib/audio/AudioEngine.ts (handleAudioChunk + updateVisualizationBuffer + notifyVisualizationUpdate), src/lib/audio/AudioSegmentProcessor.ts.

Hypothesis

• Doing VAD + visualization summary + segment extraction on the main thread contributes to high CPU during capture.

Actions

• Move VAD + energy computation into AudioWorklet or a worker and send only small metrics back.
• Split visualization summary into a worker or throttle it to <30fps.
• Guard processing when not recording or when UI is hidden.

Acceptance

• Main-thread CPU usage drops measurably when recording with model not ready.
• 30s trace shows lower main-thread active time and fewer long tasks.

[coderabbitai]

⚠️ Possible Duplicate Issue(s)

• Perf: Move resampling/VAD off main thread (AudioEngine onmessage spikes) #37

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!