[Bug] Testing with selected apps

[Maxissc]

Description

Testing while selecting Discord and Google Chrome as selected apps

System Info

• OS: Windows-11-10.0.26200-SP0
• Python: 3.14.7
• PyTorch: 2.12.1+cpu
• CUDA available: False

Error

Traceback (most recent call last):
  File "C:\Users\Maxi\TalkTrack\app\main_window.py", line 445, in _on_test_mic_toggled
    self._start_system_monitor()
    ~~~~~~~~~~~~~~~~~~~~~~~~~~^^
  File "C:\Users\Maxi\TalkTrack\app\main_window.py", line 465, in _start_system_monitor
    monitor = ProcessAudioCapture(
        pids=pids,
    ...<2 lines>...
        enable_buffer=False,
    )
TypeError: ProcessAudioCapture.__init__() got an unexpected keyword argument 'enable_buffer'

Recent Log

2026-08-21 13:18:55,581 [INFO] app.recording.audio_capture: WASAPI loopback: Speakers (HyperX Cloud III S Wireless) [Loopback] (index=17, rate=96000, ch=2)
2026-08-21 13:19:18,006 [INFO] app.recording.mic_monitor: Mic monitor started on device 1
2026-08-21 13:19:18,007 [CRITICAL] talktrack: Uncaught exception:
Traceback (most recent call last):
  File "C:\Users\Maxi\TalkTrack\app\main_window.py", line 445, in _on_test_mic_toggled
    self._start_system_monitor()
    ~~~~~~~~~~~~~~~~~~~~~~~~~~^^
  File "C:\Users\Maxi\TalkTrack\app\main_window.py", line 465, in _start_system_monitor
    monitor = ProcessAudioCapture(
        pids=pids,
    ...<2 lines>...
        enable_buffer=False,
    )
TypeError: ProcessAudioCapture.__init__() got an unexpected keyword argument 'enable_buffer'

[ObscureAintSecure]

Confirmed and fixed in d0b93a3 — thanks for the report, the traceback had everything needed.

ProcessAudioCapture lost its enable_buffer parameter in 23147ca when capture moved to streaming disk writes, but this call site kept passing it. So Test Mic crashed for anyone in per-app mode with apps selected. The level-meter-only configuration is the sink=None default now, so the argument simply goes.

Worth saying plainly: nothing exercised that path, so the test suite stayed green the whole time the button was dead. I've added cover that calls _start_system_monitor against a stub self with ProcessAudioCapture autospec'd, so a signature change like this one fails the suite instead of shipping.

Unrelated to your crash, but flagging since you're on per-app mode with Discord and Chrome: Chrome should capture fine, but conferencing apps can opt out of per-process loopback and record as silence. Not your issue here, just adjacent.