Alice AI Server

Local AI infrastructure for the Alice home assistant system.

🪟 Windows or Linux?

Windows is recommended if you're not comfortable with Linux. All services work on Windows via Docker Desktop.

Guide	For
WINDOWS-SETUP.md	👈 Start here - GUI-friendly, beginner-friendly
Setup for Linux	Advanced users comfortable with terminal

Overview

This repo contains everything needed to set up a local AI server that powers:

Local LLM (Ollama) - Fast responses, zero API costs
Local STT (Faster-Whisper) - Speech-to-text
Local TTS (Piper + voice clone) - Text-to-speech
Local Image Gen (ComfyUI) - Image generation
Local Vision (LLaVA) - Camera/image analysis

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                         HOUSE                                    │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│   ┌─────────┐    ┌─────────┐    ┌─────────┐    ┌─────────┐     │
│   │Voice PE │    │Voice PE │    │Voice PE │    │Voice PE │     │
│   │ Office  │    │ Kitchen │    │ Bedroom │    │ Living  │     │
│   └────┬────┘    └────┬────┘    └────┬────┘    └────┬────┘     │
│        │              │              │              │           │
│        └──────────────┴──────────────┴──────────────┘           │
│                              │                                   │
│                              ▼                                   │
│                    ┌─────────────────┐                          │
│                    │ Home Assistant  │                          │
│                    │   (Whisper)     │                          │
│                    └────────┬────────┘                          │
│                             │                                    │
│              ┌──────────────┴──────────────┐                    │
│              ▼                              ▼                    │
│    ┌─────────────────┐            ┌─────────────────┐          │
│    │  Alice (Main)   │            │   AI Server     │          │
│    │    OpenClaw     │◄──────────►│   (Ubuntu)      │          │
│    │   Gateway       │            │                 │          │
│    │                 │            │ • Ollama (LLM)  │          │
│    │ • Claude API    │            │ • Whisper (STT) │          │
│    │ • ElevenLabs    │            │ • Piper (TTS)   │          │
│    │ • Tools/Skills  │            │ • ComfyUI       │          │
│    └────────┬────────┘            │ • LLaVA         │          │
│             │                      └─────────────────┘          │
│             ▼                                                    │
│    ┌─────────────────┐                                          │
│    │ Google Speakers │ (TTS output / Music)                     │
│    └─────────────────┘                                          │
│                                                                  │
└─────────────────────────────────────────────────────────────────┘

Hardware Requirements

AI Server (Recommended):

GPU: NVIDIA RTX 4090 / 3090 (24GB VRAM) or RTX 4080 (16GB)
RAM: 32GB+ (64GB ideal)
Storage: 1TB+ NVMe SSD
CPU: Modern 8+ core (AMD Ryzen / Intel i7+)

VRAM Usage Estimates:

Service	VRAM
Ollama (Llama 3.3 70B Q4)	~40GB
Ollama (Llama 3.3 8B)	~6GB
Ollama (Qwen 2.5 32B Q4)	~20GB
Faster-Whisper large-v3	~3GB
ComfyUI (SDXL)	~8GB
LLaVA 13B	~10GB

Quick Start

1. Fresh Ubuntu Server Install

# Download Ubuntu Server 22.04 LTS
# Create bootable USB with Rufus/Balena Etcher
# Install with minimal options, enable SSH

2. Run Setup Script

# SSH into server
ssh user@ai-server.local

# Clone this repo
git clone https://github.com/andyritchie/alice-ai-server.git
cd alice-ai-server

# Run setup (installs everything)
chmod +x setup.sh
./setup.sh

3. Configure OpenClaw

Copy the config template and update with your server IP:

cp configs/openclaw-ai-server.json ~/.openclaw/config.json

Directory Structure

alice-ai-server/
├── README.md                 # This file
├── setup.sh                  # Main setup script
├── docker-compose.yml        # All services
├── configs/
│   ├── openclaw-ai-server.json   # OpenClaw config template
│   ├── ollama-models.txt         # Models to pull
│   └── piper-voices.txt          # TTS voices to download
├── scripts/
│   ├── install-nvidia.sh         # NVIDIA driver install
│   ├── install-docker.sh         # Docker setup
│   ├── install-ollama.sh         # Ollama setup
│   ├── install-whisper.sh        # Faster-Whisper setup
│   ├── install-piper.sh          # Piper TTS setup
│   ├── install-comfyui.sh        # ComfyUI setup
│   └── test-all.sh               # Test all services
├── voice-pe/
│   ├── README.md                 # Voice PE setup guide
│   ├── open-voice-pe.yaml        # Custom firmware config
│   └── alice-wakeword/           # Wake word model
├── home-assistant/
│   ├── README.md                 # HA integration guide
│   └── automations.yaml          # Voice command automations
└── docs/
    ├── ARCHITECTURE.md           # Detailed architecture
    ├── TROUBLESHOOTING.md        # Common issues
    └── MODELS.md                 # Model recommendations

Services

Ollama (Local LLM)

Port: 11434
URL: http://ai-server:11434
Models: Llama 3.3, Qwen 2.5, Mistral, CodeLlama

Faster-Whisper (STT)

Port: 8080
URL: http://ai-server:8080
Model: large-v3-turbo

Piper (TTS)

Port: 5000
URL: http://ai-server:5000
Voices: en_US-amy-medium (or custom clone)

ComfyUI (Image Gen)

Port: 8188
URL: http://ai-server:8188
Models: SDXL, Flux

Open WebUI (Chat Interface)

Port: 3000
URL: http://ai-server:3000
Purpose: Web chat with local models

OpenClaw Integration

Enable HTTP API

Add to your OpenClaw config:

{
  "gateway": {
    "http": {
      "endpoints": {
        "chatCompletions": { "enabled": true }
      }
    }
  }
}

Model Routing

Configure OpenClaw to route based on task:

{
  "agents": {
    "main": {
      "models": {
        "default": "anthropic/claude-sonnet-4-20250514",
        "fast": "ollama/llama3.3:8b",
        "local": "ollama/qwen2.5:32b"
      }
    }
  }
}

Webhook for Voice Commands

{
  "hooks": {
    "enabled": true,
    "token": "your-secret-token",
    "path": "/hooks"
  }
}

Voice PE Integration

See voice-pe/README.md for:

Flashing Open Voice PE firmware
Custom "Alice" wake word setup
Connecting to OpenClaw

Home Assistant Integration

See home-assistant/README.md for:

Assist pipeline configuration
Media player routing
Voice command automations

Monitoring

Access Portainer for container management:

URL: http://ai-server:9000

Costs Saved (Monthly)

Service	Cloud Cost	Local Cost
ElevenLabs TTS	$22/mo	$0
Whisper API	~$5/mo	$0
Simple LLM queries	~$20/mo	$0
Image generation	~$10/mo	$0
Total	~$57/mo	$0

Plus: Faster responses, full privacy, no rate limits.

Documentation

Doc	Description
DEPENDENCIES.md	Full checklist of requirements
docs/MODELS.md	Which models for which tasks
docs/DOWNLOADING-MODELS.md	How to download all models
docs/COMFYUI.md	Image generation setup
docs/BROWSER-CONTROL.md	Browser automation
docs/WEB-SERVICES.md	Hosting dashboards & websites
docs/VOICE-CLONING.md	Local voice clone (replace ElevenLabs)
voice-pe/README.md	Voice PE + wake word setup

Roadmap

Quick Reference

Service URLs (after setup)

Service	URL	Purpose
Open WebUI	http://ai-server:3000	Chat with local LLMs
ComfyUI	http://ai-server:8188	Image generation
Portainer	http://ai-server:9000	Container management
Ollama API	http://ai-server:11434	LLM inference
Whisper API	http://ai-server:8080	Speech-to-text
Piper	http://ai-server:10200	Text-to-speech
XTTS	http://ai-server:8020	Voice cloning TTS

Useful Commands

# Check GPU
nvidia-smi

# Check containers
docker compose ps

# View logs
docker compose logs -f ollama

# Pull a new model
docker exec ollama ollama pull llama3.2:8b

# Test Ollama
curl http://localhost:11434/api/generate -d '{"model":"llama3.2:8b","prompt":"hi"}'

# Restart a service
docker compose restart whisper

License

MIT

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
apps		apps
configs		configs
docs		docs
scripts		scripts
voice-pe		voice-pe
DEPENDENCIES.md		DEPENDENCIES.md
README.md		README.md
WINDOWS-SETUP.md		WINDOWS-SETUP.md
docker-compose.apps.yml		docker-compose.apps.yml
docker-compose.yml		docker-compose.yml
setup.sh		setup.sh

Folders and files

Latest commit

History

Repository files navigation

Alice AI Server

🪟 Windows or Linux?

Overview

Architecture

Hardware Requirements

Quick Start

1. Fresh Ubuntu Server Install

2. Run Setup Script

3. Configure OpenClaw

Directory Structure

Services

Ollama (Local LLM)

Faster-Whisper (STT)

Piper (TTS)

ComfyUI (Image Gen)

Open WebUI (Chat Interface)

OpenClaw Integration

Enable HTTP API

Model Routing

Webhook for Voice Commands

Voice PE Integration

Home Assistant Integration

Monitoring

Costs Saved (Monthly)

Documentation

Roadmap

Quick Reference

Service URLs (after setup)

Useful Commands

License

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages