What's New
- Optional cloud ASR support — Two new transcription providers alongside existing local models: AssemblyAI (upload + poll with custom spelling) and ElevenLabs Scribe (single-shot with keyterm boosting). Opt in via Settings > Transcription. Audio is sent only when a cloud model is explicitly selected and an API key is entered. (#240, closes #238)
Security & Privacy
- Ephemeral URLSession (no disk caching of audio or transcripts)
- SSRF validation on AssemblyAI upload URLs
- API keys stored in macOS Keychain
- Error types exclude response bodies to prevent key leakage in logs
- Provider-specific privacy notices displayed in Settings
Other
- Updated README demo video (#242)
Contributors
Thanks to @alessandro-lac for the detailed feature request and to @Newarr for the implementation.