语音对话AI

# Pull and run
docker run -d --gpus all \
  -p 8998:8998 \
  -e HF_TOKEN=your_token \
  --name personaplex \
  neosun/personaplex:latest

# Access Web UI
open http://localhost:8998

version: '3.8'
services:
  personaplex:
    image: neosun/personaplex:latest
    container_name: personaplex
    ports:
      - "8998:8998"
    environment:
      - HF_TOKEN=${HF_TOKEN}
    volumes:
      - ~/.cache/huggingface:/root/.cache/huggingface
      - /tmp/personaplex:/tmp/personaplex
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    restart: unless-stopped

# Set token and start
export HF_TOKEN=your_huggingface_token
docker-compose up -d

# Clone repository
git clone https://github.com/neosun100/personaplex.git
cd personaplex

# Set HF token
export HF_TOKEN=your_huggingface_token

# Start (auto-selects GPU with lowest memory usage)
./start.sh

# Use specific GPU
export NVIDIA_VISIBLE_DEVICES=2
docker-compose up -d

# Or in docker-compose.yml
deploy:
  resources:
    reservations:
      devices:
        - driver: nvidia
          device_ids: ['2']
          capabilities: [gpu]

You are a wise and friendly teacher. Answer questions or provide advice in a clear and engaging way.

You work for First Neuron Bank which is a bank and your name is Alexis Kim. Information: The customer's transaction for $1,200 at Home Depot was declined. Verify customer identity.

You enjoy having a good conversation.

curl -X POST http://localhost:8998/api/offline \
  -F "file=@input.wav" \
  -F "voice_prompt=NATF2.pt" \
  -F "text_prompt=You are a helpful assistant." \
  -o output.wav

{
  "mcpServers": {
    "personaplex": {
      "command": "docker",
      "args": ["exec", "-i", "personaplex", "python", "-m", "app.mcp_server"]
    }
  }
}

personaplex/
├── app/
│   ├── server.py          # FastAPI server
│   ├── mcp_server.py      # MCP server
│   └── templates/         # Web UI
├── moshi/                  # Core model package
├── client/                 # Original React client
├── assets/                 # Test files
├── Dockerfile             # All-in-One image
├── docker-compose.yml     # Compose config
├── start.sh               # One-click start
└── MCP_GUIDE.md           # MCP documentation