Docker¶

Docker Run¶

docker build -t kiwi-voice .

docker run -d \
  --name kiwi-voice \
  --device /dev/snd \
  -v ./config.yaml:/app/config.yaml \
  -v ./.env:/app/.env \
  -v ./data:/app/data \
  -p 7789:7789 \
  kiwi-voice

Note

--device /dev/snd passes the sound devices to the container (Linux only). For audio output on macOS/Windows, use the Web Microphone instead.

Docker Compose¶

version: "3.8"
services:
  kiwi-voice:
    build: .
    restart: unless-stopped
    devices:
      - /dev/snd:/dev/snd
    volumes:
      - ./config.yaml:/app/config.yaml
      - ./.env:/app/.env
      - ./data:/app/data
    ports:
      - "7789:7789"
    environment:
      - KIWI_LANGUAGE=en
      - KIWI_TTS_PROVIDER=kokoro

With GPU (NVIDIA)¶

version: "3.8"
services:
  kiwi-voice:
    build: .
    restart: unless-stopped
    devices:
      - /dev/snd:/dev/snd
    volumes:
      - ./config.yaml:/app/config.yaml
      - ./.env:/app/.env
      - ./data:/app/data
    ports:
      - "7789:7789"
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    environment:
      - KIWI_LANGUAGE=en
      - KIWI_TTS_PROVIDER=kokoro

Requires the NVIDIA Container Toolkit.

Volumes¶

Mount	Purpose
`config.yaml`	Configuration
`.env`	Secrets
`data/`	Voice profiles, cached models

Dashboard Access¶

The dashboard is available at http://localhost:7789 (or your server IP).