Added 90+ new tools in various categories and made the models self sufficient to make more tools themselves. Upto 260 tools rn.
Fixed the download button not working on the website, created a one line installer, turned out pretty good, happy with that 😎
Ship
Changes requested
Lumi — Complete Voice Assistant
Core System
Wake Word Detection — Detects “hey lumi”, “ok lumi”, “yo lumi”, “lumi” using energy-based detection + quick Whisper verification
Speech-to-Text — Whisper (GPU-accelerated, large-v3)
LLM Integration — Supports OpenAI, Anthropic, Google, OpenRouter, HackClub, or custom endpoints
Text-to-Speech — Piper with sentence-streaming (British female voice)
Desktop Overlay — 600x220 animated particles at bottom center (invisible → golden waveform → blue orb → golden swirl)
System Tray — Color-coded icon + right-click menu
263 Built-in Tools — PC control, file management, web, media, system info, etc.
Installers & Launchers
install.bat — Full terminal installer (9 steps):
Python 3.11 check/install
Install directory picker
Git clone
Virtual environment
Dependencies
Voice model download
API key + provider configuration
Wake word training option
Finalization
launch.bat — Starts Lumi as tray app (no terminal window)
launch_voice.bat — Starts with terminal for live output
Wake Word Training System
record_wake_samples.py — Records 20 positive + 10 negative samples
configs/lumi.yaml — Training config for livekit-wakeword
core/wake_word.py — Dual-mode detector (trained model + energy fallback)
Supports “hey lumi”, “ok lumi”, “yo lumi”, “lumi”
Documentation
Updated README.md with wake word section
Training instructions for custom model
Configuration reference
Troubleshooting guide
What Works Now
✅ Install via install.bat
✅ Launch via launch.bat
✅ Wake word detection (energy-based + verification)
✅ All 263 tools
✅ Desktop overlay
✅ System tray
✅ Voice conversations
✅ API key configuration
✅ Provider selection (6 options)
Ready for Custom Training
When you install espeak-ng manually:
pip install livekit-wakeword[listener,train,eval,export]
python record_wake_samples.py
livekit-wakeword run configs/lumi.yaml
Current status: Fully functional. You can use it right now with launch.bat. The wake word detection works immediately (no training needed), and all tools are operational.
