Local AI Stack

A streamlined Docker Compose stack for running a local AI environment with Ollama, Kokoro TTS, and Open WebUI.

🚀 Features

• Ollama: Run large language models (like Llama 3.2) locally.
• Kokoro TTS: High-quality, fast text-to-speech engine.
• Open WebUI: A comprehensive web interface for interacting with your AI models (currently optional/commented out).
• Timezone Support: Pre-configured for America/Los_Angeles.

🛠️ Prerequisites

• Docker
• Docker Compose
• (Optional) NVIDIA GPU with NVIDIA Container Toolkit for hardware acceleration.

📥 Setup

1. Clone the repository:

   git clone <repository-url>
   cd ai-stack

2. Configure Environment: Copy the example environment file and adjust values as needed.

   cp .env.example .env

   Note: Ensure you set your HF_TOKEN if using Open WebUI.

3. Start the Stack:

   docker compose up -d

🔌 Service Details

Service	Host Port	Internal Port	Description
Ollama	11434	11434	LLM Backend
Kokoro TTS	8880	8880	Text-to-Speech API
Open WebUI	3000	8080	Web Interface (Optional)

Ollama Model Initialization

The stack is configured to pull the model specified by LLM_MODEL (default: llama3.2) on startup if using the ollama-init service.

🏎️ GPU Acceleration (NVIDIA)

To enable GPU support for Ollama, uncomment the deploy section in docker-compose.yaml:

# ...
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
# ...

📅 Timezone

All services are configured to use the America/Los_Angeles timezone by default. You can change this in the environment section of each service in docker-compose.yaml.