V-Lucent companion floating above desktop Chat interface and controls 3D avatar reacting in real time
V-Lucent is a desktop-ready AI companion that blends a VRM avatar, speech-to-speech chat, and OpenClaw’s VPS brain into a single Tauri application. Windows builds are ready today (macOS/Linux later) with a slim Svelte front-end and Rust backend.
- 🪟 Transparent Overlay – Native desktop companion that floats with zero background chrome.
- 👁️ Screen Awareness – Optional
screen_vision_service.pyloop summarizes your live screen. - 🗣️ Speech-to-Speech – MediaRecorder ➜ Deepgram (STT) ➜ multi-provider LLM ➜ ElevenLabs (TTS) with viseme sync.
- 🎭 3D Emotional Presence – Three.js + @pixiv/three-vrm avatar with emotion-driven blendshapes and custom VRM upload.
- 🧠 Brain Switcher – Hot-swap between Ollama, OpenClaw, OpenRouter, and Groq at runtime.
- 🧳 Legacy Isolation – Heavy Docker/proxy experiments live under
legacy/so the active app stays lightweight.
V-Lucent/
├── local-companion/ # Active Tauri + Svelte 5 application
│ ├── src/ # Frontend (VRMRenderer, chat UI, settings)
│ ├── src-tauri/ # Rust backend (commands, proxy, vision services)
│ └── public/models/ # Default VRM assets
├── legacy/ # Archived Docker/proxy experiments (ignored by default build)
└── README.md
- Node.js 20+
- Rust toolchain (stable) +
cargo - Windows 10/11 with WebView2 runtime (see Tauri checklist)
# 1. Clone the universe git clone https://github.com/codacipher/v-lucent.git # 2. Enter the chamber cd v-lucent/local-companion # 3. Install dependencies npm install # 4. Launch V-Lucent npm run tauri dev
💡 Tip: The Svelte UI runs at
localhost:1420, while the Rust backend exposes helper APIs onlocalhost:3030.
| Component | Location | Details |
|---|---|---|
| LLM Provider | Settings → LLM | Switch between Ollama, OpenClaw, OpenRouter, Groq |
| Custom VRM | Camera Panel | Drag & drop .vrm; stored under %APPDATA%/local-companion |
| Audio Keys | Settings → Audio | Add Deepgram (STT) + ElevenLabs (TTS) API keys |
| Vision Assist | Settings → Visual | Toggle screen_vision_service.py for context capture |
- Frontend (Svelte 5):
VRMRenderer.sveltedrives the Three.js scene, visemes, and chat UI. - Backend (Rust/Tauri):
src-tauri/src/lib.rshandles LLM proxying, native screen capture, settings persistence, and helper daemons. - Speech Pipeline: MediaRecorder ➜ Deepgram ➜ chosen LLM ➜ ElevenLabs chunked playback for natural voice loops.
- Persistence:
%APPDATA%/OllamaGUI/settings.jsonfor secrets +%APPDATA%/local-companion/custom_vrmsfor avatars (both git-ignored).
npm run tauri devstreams both frontend and Rust logs—search for tags like[S2S],[STT][Deepgram],[Chat].- Reset to defaults by deleting
%APPDATA%/OllamaGUI/settings.jsonand relaunching. - VRM acting up? Clear
%APPDATA%/local-companion/custom_vrmsor upload a new file.
cd local-companion
npm run tauri buildBuild artifacts land in local-companion/src-tauri/target/release/. Bundle your own VRM/model assets or ship an installer/downloader for large binaries.
- v0.4 – Tauri 2 / Svelte 5 core engine
- v0.5 – Long-term memory (vector DB)
- v0.6 – Real-time viseme & lip-sync polish
- v0.7 – Multi-provider routing + proactive vision loop
- v0.8 – Twitch/YouTube live mode
- Fork the repo
- Create a feature branch:
git checkout -b feature/NewFeature - Commit your changes:
git commit -m "Add Feature" - Push to the branch:
git push origin feature/NewFeature - Open a Pull Request
Licensed under MIT • Built by CodaCipher
