Speech & Voice
- Local ASR (SenseVoice engine) — added an on-device speech-to-text engine (sherpa-wasm) for
private, offline voice transcription, alongside the existing Whisper (Local) engine. - Cloud ASR (gemini-web) — added a zero token speech-to-text engine with gemini-web
- Push-to-talk dictation — hold the record hotkey (default: Right Alt) to record and release to stop; transcribes
straight into the chat input. - Voice record UI — new record/stop button placed side-by-side with send, with a live waveform in the footer and a
stop icon while recording. Button styling now matches the send button. - Browser mic + web speech provider — added a speech web-provider abstraction layer with browser microphone support,
enabling zero-cost transcription via the browser. - Dedicated Speech settings tab — moved STT/TTS configuration into its own Speech tab under Settings.
- Engine label — renamed the local engine label to "Whisper (Local)" for clarity.
Providers
- Requesty support — added Requesty as an OpenAI-compatible provider (#14), with the correct base URL now shown in the
model dialog.
Sessions & UI
- Rename sessions — double-click a session name (in the sidebar or the chat header) to rename it inline.
- First-run setup — refreshed the first-run flow with web-auth setup.
Fixes
- Fixed Gemini web STT returning an empty transcript on streamed audio clips.
- Fixed Gemini web-provider images not displaying in the chat UI.