Add maximum dictation duration guard

[wpflueger]

Summary

• Adds a configurable maxDurationSeconds setting (default: 60s) that auto-stops recording when reached, with an overlay notification
• Adds a server-side max_audio_frames safety limit (5000 frames ≈ 100s) in the backend as a backstop against runaway clients
• Prevents CUDA out-of-memory errors from unbounded audio accumulation during very long recordings

Test plan

• Hold the hotkey for >60s and verify recording auto-stops with "Maximum duration reached" overlay
• Change maxDurationSeconds in settings and verify the new limit applies
• Verify normal short dictation (<60s) is unaffected
• Verify backend logs warning when frame limit is hit

Closes #30

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Adds client- and server-side limits to prevent unbounded audio accumulation during dictation, reducing the risk of CUDA OOM errors from very long recordings.

Changes:

• Adds a client maxDurationSeconds audio setting (default 60s) that auto-stops recording and shows an overlay.
• Adds a backend max_audio_frames cap (default 5000) that stops collecting streamed frames as a safety backstop.
• Logs when the client/server duration/frame limits are hit.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
backend/src/parakey_backend/service.py	Enforces a maximum number of streamed audio frames before processing.
backend/src/parakey_backend/config.py	Introduces max_audio_frames config default for the backend safety limit.
apps/desktop/electron/settings.ts	Adds audio.maxDurationSeconds to settings schema and defaults.
apps/desktop/electron/main.ts	Implements the max-duration timer that auto-stops dictation and notifies the user.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

There are existing unit tests for DictationService.StreamAudio (see backend/tests/test_service.py), but nothing currently verifies the new max_audio_frames cap. Please add a test that streams more than max_audio_frames frames and asserts the engine only receives up to the configured limit (and optionally that a warning is logged).

[Copilot]

stopDictation() is async, but it’s being invoked from setTimeout without await/error handling. If it throws/rejects, this can surface as an unhandled promise rejection in the main process. Consider using void stopDictation().catch(...) (or similar) to make the intent explicit and ensure failures are logged/handled.

Suggested change

	stopDictation();
	void stopDictation().catch((error) => {
	const message =
	error instanceof Error ? error.message : "Error while stopping dictation after max duration.";
	sendToMain("backend:log", message);
	});

[Copilot]

max_audio_frames is added to BackendConfig but load_config_from_env() doesn’t populate it from an environment variable, so the safety limit can’t actually be configured in the default server path. Consider adding something like PARAKEY_MAX_AUDIO_FRAMES (with int parsing + validation) or explicitly documenting that it’s fixed at the default unless BackendConfig is constructed manually.