v0.2.0-beta.12

This beta improves recording controls and keeps ScreenCaptureKit internal audio capture recoverable across display-idle sleep and OS-initiated stream stops.

Added

• Add settings for speaker diarization, custom summary prompts, transcription throttling, quit confirmation, and native completion notifications.
• Add structured ScreenCaptureKit diagnostics for stream errors, captured bytes, elapsed duration, restart reasons, and recovery state.

Fixed

• Prevent display-idle sleep while internal audio recording is active.
• Recreate the same ScreenCaptureKit capture path after stream stops, sidecar exits, raw-byte stalls, or a silent tail after valid audio.
• Preserve every captured RAW segment and insert silence only for the measured recovery gap.
• Wait for the display capture source to return without consuming the recovery budget, while limiting consecutive non-display restart failures.
• Avoid repeated silent-tail restarts until the recreated stream has captured new non-silent audio.
• Use Seam Setup as the DMG volume name to avoid macOS mount-path conflicts.

Tests

• ./scripts/oss_preflight.sh
• uv run python tests/test_system_capture_failures.py (27 tests)
• uv run python tests/test_recorder_preflight.py (11 tests)
• uv run python tests/test_recording_pipeline_failures.py (6 tests)
• PYTHONPATH=. uv run python tests/test_audio_leveling.py (3 tests)
• uv run python tests/test_streaming_generation.py (3 tests)
• cargo test --manifest-path gui/src-tauri/Cargo.toml -- --nocapture (3 tests)
• Objective-C sidecar compile with -Wall -Wextra -Werror
• Target PR: #82