45 lines
1.6 KiB
Markdown
45 lines
1.6 KiB
Markdown
# 2026-02-08 - GLaDOS Voice Bot Setup
|
|
|
|
## What Happened
|
|
|
|
- Discussed GLaDOS project structure - it's a multi-agent voice assistant with:
|
|
- Parakeet ASR for speech recognition (excellent, low latency)
|
|
- Kokoro TTS with custom GLaDOS voice
|
|
- Autonomy loop for proactive speech (no wake words)
|
|
- Subagents (emotion, memory, news, weather) that write to shared slots
|
|
- FastVLM for vision-based scene understanding
|
|
- MCP tools for extensibility (Home Assistant, system info, etc.)
|
|
|
|
- Corey wants a Discord voice bot that's always available in a voice channel
|
|
- Existing bot at `/home/admin/.openclaw/workspace/discord-voice-bot/` was clunky
|
|
- Decision: Clean restart with simpler architecture
|
|
|
|
## Discord Voice Bot - New Setup
|
|
|
|
**Location:** `/home/admin/.openclaw/workspace/discord-voice-bot/`
|
|
|
|
**Configuration:**
|
|
- `config.yaml`: Discord token, channel ID 1468627455656067074 (#coding voice)
|
|
- STT: Wyoming Whisper on 192.168.0.17:10300
|
|
- LLM: Ollama at 192.168.0.17:11434 with `qwen3-coder-next:cloud`
|
|
- TTS: HTTP endpoint at localhost:5050 with "glados" voice
|
|
|
|
**Commands:**
|
|
- `!join` - Join voice channel (auto-joins on startup)
|
|
- `!leave` - Leave voice channel
|
|
- `!test [text]` - Test TTS
|
|
|
|
**Key Files:**
|
|
- `main.py` - Bot implementation using discord.py, wyoming protocol
|
|
- `config.yaml` - Configuration
|
|
- `requirements.txt` - Dependencies: discord.py, requests, numpy, wyoming
|
|
|
|
**What Changed:**
|
|
- Removed GLaDOS direct dependencies (simpler integration)
|
|
- Uses existing services (Whisper TTS, Ollama, HTTP TTS)
|
|
- Auto-joins voice channel on startup
|
|
- Clean async architecture
|
|
|
|
## Pending
|
|
- Test the bot in production
|