Implements Apple SpeechAnalyzer + SpeechTranscriber (macOS 26+) live transcription
per official docs, with Inscribe bugfix (finalizeAndFinishThroughEndOfInput)
- speech_list_locales: isAvailable, supported/installed locales, reserved counts
- speech_transcribe_file: transcribe any AVFoundation-readable file, returns
transcript + segments + elapsed + realtimeFactor + engine metadata
- speech_quick_test: say + transcribe loopback to verify pipeline
Swift binary compiled on-demand via swiftc -O with Speech + AVFoundation
Follows docs: AssetInventory downloadAndInstall, AssetInputSequenceProvider
Mac mini is 26.5.2 Mac16,10 M4 so engine verified available
Mac mini is on macOS 26.5.2 Mac16,10 M4 - SpeechAnalyzer available
- Calendar/Contacts/Notes/Reminders use full /System/Applications/*.app path (more reliable)
- Reminders now returns assignment {assignee, source, available} with fallback to (Name) in title
- deco.js uses deco_ha_bridge.py for clients action (more reliable than old bridge)
- Added deco_ha_bridge.py (13KB RSA/AES Deco HA client)
- Added system_get_info + system_speech_api_status (verifies SpeechAnalyzer on macOS 26+)
- Preserved codex-image, gemini-image, gemini-chrome-prompt tools
- Updated README with all 23 tools + macOS 26 note + benchmark ref
Mac mini is on macOS 26.5.2 with SpeechAnalyzer available - verified via sw_vers
Needed to sync diverged repos: Pi had system tools, Mac had image tools
aeroreyna