Voice
DesktopClaw
By DesktopClaw is a state-aware desktop pet interface for your OpenClaw instance. It brings your AI to life as a reactive companion you can interact with via voice or text, while providing real-time session visibility, task feedback, and quick action triggers through a persistent, animated desktop presence.
Install
npm install
npmConfiguration Example
{ "message": "...", "input": "...", "query": "..." }README
<div align="center">
<img src="src/assets/brand/banner.png" alt="DesktopClaw — Your Vigilant, Warm Desktop Companion" width="100%" />
</div>
<br/>
<div align="center">
<img src="src/assets/brand/avatar.png" alt="DesktopClaw Avatar" width="96" />
# DesktopClaw
[](https://github.com/divbasson/DesktopClaw/actions/workflows/github-code-scanning/codeql)
*A Windows desktop pet companion powered by OpenClaw.*
</div>
## Prieview
<img src="src/assets/brand/prievew.gif"/>
<br/>
## What this prototype does
- Floating transparent desktop pet window
- Draggable mascot with idle motion and random quirks
- Always-on-top toggle (default on)
- Global hotkey to trigger listening
- Voice input via Web Speech API when available
- Text input fallback
- Configurable OpenClaw gateway client with mock mode
- TTS playback via system voices
- Basic mouth animation synced from lightweight speech amplitude heuristics
- Settings panel backed by `config.json`
- Wake-word fallback mode using continuous speech recognition when supported
- Optional click-through when idle
- Status polling with lightweight in-app notifications
- Optional WebSocket live-event stream support
- Native desktop notifications
- Voice selection from available system TTS voices
- Cursor-proximity reaction for a more alive idle feel
- System tray menu for quick pet actions
## Architecture
```text
src/
main/
app-controller.js App orchestration + IPC
config-store.js Persisted config lifecycle
main.js Electron entrypoint
openclaw-client.js OpenClaw/mock HTTP client
preload.js Safe renderer bridge
shortcut-manager.js Global hotkey registration
ui-shell.js Window creation + shell controls
renderer/
animation_engine.js Animation state + mouth/quirks
index.html Floating pet UI
openclaw_client.js Renderer-side gateway adapter
renderer.js Composition root for pet behavior
styles.css Pet visuals + animation states
tts_engine.js System TTS wrapper
ui_shell.js Bubble/settings helpers
voice_input.js Speech capture wrapper
wake_word.js Wake word fallback watcher
shared/
config.js Config load/save
assets/
sprites/
openclaw_pet.svg
animations.json
config.json Runtime configuration
```
## Quick start
```bash
cd /zfs/source-code/desktopclaw
npm install
npm start
```
## Configuring OpenClaw
Edit `config.json` or use the settings panel.
When running the packaged app, DesktopClaw persists runtime settings in the Electron user data folder (`%APPDATA%/DesktopClaw/config.json`) so updates like `gateway.mode = "http"` are writable and survive restarts.
- `gateway.mode = "mock"` keeps everything local for UI testing.
- `gateway.mode = "http"` sends requests to:
- `baseUrl + chatPath` for chat
- `baseUrl + statusPath` for status
- Authentication options for `gateway.mode = "http"`:
- `gateway.token` sends `Authorization: Bearer <token>`
- `gateway.password` sends `X-OpenClaw-Password: <password>`
- If both are set, both auth headers are sent
- Current request payload is:
```json
{ "message": "...", "input": "...", "query": "..." }
```
The client accepts a reply from any of these response fields:
- `reply`
- `text`
- `message`
- `output`
- `response`
If your OpenClaw gateway uses a different schema, adapt `src/main/openclaw-client.js`.
## Status polling and notifications
A lightweight polling monitor is included for the MVP.
Config keys:
- `status.pollEnabled`
- `status.pollIntervalMs`
- `status.showNotifications`
- `status.speakNotifications`
- `status.nativeNotifications`
- `gateway.eventsEnabled`
- `gateway.eventsUrl`
Current behavior:
- polls the configured status endpoint
- updates the pet's status pill
- raises a small in-app notification when status payload changes
- can optionally speak those updates
- can subscribe to a WebSocket event feed for live OpenClaw notifications
- can mirror notifications to native desktop toasts
## Hotkeys
Default hotkeys live in `config.json`:
- Listen: `Ctrl+Shift+Space`
- Mute: `Ctrl+Shift+M`
- Show/hide: `Ctrl+Shift+H`
- Settings: `Ctrl+Shift+S`
## Building a Windows .exe
For a fast unpacked Windows prototype:
```bash
npm run dist:dir
```
That targets `dist/win-unpacked/DesktopClaw.exe`.
For installer + portable packaging:
This repo is set up for `electron-builder`:
```bash
npm run dist
```
That produces a Windows installer/portable app **when run in a Windows-capable packaging environment**.
### Important note
I prepared the project for Windows packaging, but if you are building from Linux you may still need one of:
- a Windows machine
- Wine + NSIS toolchain
- CI that targets Windows
So the codebase is ready for `.exe` output, but final packaging depends on the host build environment.
## MVP limitations
- Wake word is config-backed but not yet true offline wake-word detection
- Speech recognition depends on platform/browser support in Electron
- Mouth sync is amplitude/heuristic-based, not phoneme-based
- Asset pack is intentionally minimal and code-driven for the MVP
## Next best upgrades
1. Add native STT (Whisper/Vosk) for reliable speech capture
2. Add Porcupine/openWakeWord for true wake word support
3. Replace CSS pet with full PNG sprite atlas from the approved art pack
4. Add WebSocket/live OpenClaw events and notification badges
5. Add click-through-on-idle behavior tuning and cursor reactions
---
<div align="center">
<img src="src/assets/brand/preview-notification.png" alt="DesktopClaw in action — System Ready notification" width="600" />
<br/>
<sub>© DesktopClaw — Your Vigilant, Warm Desktop Companion</sub>
</div>
voice
Comments
Sign in to leave a comment