Echo v1.0.3

Fixes

• Hotkey setup now works. The global chord could fail to register on first launch — before Accessibility permission was granted — and never recovered without an app relaunch. It now re-arms automatically once permission is granted, and the Settings recorder shows a clear "Accessibility required" prompt with a fix action instead of silently failing.
• The hotkey is now a single app-wide chord with one editor in Settings → General (was a stale per-profile recorder).
• .paste / .none output modes no longer leak speaker audio.
• Audio frames and utterance markers can no longer reach Gemini out of order.
• Fixed keycode 0 (the A key) being rejected as an unset binding.

Internal

• Removed dead code from the Gemini-only refactor: HotkeyManager, the KeyboardShortcuts/Defaults dependencies, and the unused Silero VAD model + onnxruntime. The download is now ~0.8MB, down from 12MB.

Echo v1.0.2

Highlights

• Gemini-only: dropped OpenAI Realtime and Grok Voice providers. Gemini Live is the sole backend.
• 2-key chord push-to-talk: hold ⌥ to keep the WSS session alive, hold ` to talk. Release ` ends the utterance and triggers the response; release ⌥ tears the session down. Works from any focused app.
• Lower latency: server VAD off + TURN_INCLUDES_ONLY_ACTIVITY + thinkingBudget=0. Connect / setup / first-audio-out timings now logged.
• Profile migration: legacy persisted profiles still decode (provider/mode/vad fields ignored).

Install

1. Download Echo.dmg, open, drag Echo.app to Applications.
2. First launch will request Microphone + Accessibility permissions.

Echo v1.0.1

Remove auto-open settings on startup

• Settings and onboarding now accessible only via menu bar or dock context menu
• Menu bar apps should not open windows automatically

Echo v1.0.0

Initial release with speaker audio fixes

• Fix speaker volume: move voice processing to per-session to prevent system audio ducking
• Add engine interrupt recovery: auto-restart audio engine after system interruptions

Echo v0.1.0

First public release.

Install

1. Download Echo.dmg below.
2. Open it, drag Echo → Applications.
3. Launch — accept the microphone permission prompt.
4. Open Settings (menu-bar icon → Open Settings…) → paste your API key.

Features

• Push-to-talk to Gemini Live (gemini-3.1-flash-live-preview), OpenAI Realtime, or xAI Grok.
• Notch-engulfing HUD with live waveform and per-session cost.
• AVAudioEngine prewarmed at launch with voice processing AEC; mic muted during assistant playback.
• Optional client-side Silero v5 VAD (ONNX, on-device).
• Keychain-stored API keys, fresh-per-press session lifecycle.
• Clipboard-paste output mode (auto-⌘V via Accessibility, falls back to clipboard-only).

Build SHA

5e03598