ChatLlama

A multi-chat HTTP server powered by Llama 3 8B with persistent memory and web interface.

Features

• Web-based chat interface with markdown rendering
• Multiple concurrent chat sessions
• Persistent SQLite database for chat history
• Memory system tracking facts, experiences, and topics
• Server-sent events for real-time streaming responses
• Docker-first deployment with embedded model

Quick Start (Docker)

Prerequisites

• Docker and Docker Compose
• 8GB+ available RAM
• 10GB+ free disk space

Running with Docker

1. Clone the repository:

git clone https://github.com/ByteBaker/ChatLlama
cd ChatLlama

2. Start the application:

docker-compose up --build

The model (~5GB) will be automatically downloaded during the first build. This may take a few minutes depending on your connection.

3. Access the application:

http://localhost:8000

Configuration

Set custom port via environment variable:

PORT=3000 docker-compose up --build

Manual Installation

Prerequisites

• Python 3.11+
• 8GB+ available RAM

Setup

1. Install dependencies:

pip install -r requirements.txt

2. Download the model:

python src/utils/download_model.py

3. Run the server:

python src/main.py

File Structure

ChatLlama/
├── src/
│   ├── main.py              # Server entry point
│   ├── config.py            # Model configuration
│   ├── chat_server.py       # Chat logic and memory system
│   ├── http_handler.py      # HTTP request handling
│   ├── index.html           # Web interface
│   ├── script.js            # Frontend JavaScript
│   ├── styles.css           # Interface styling
│   └── utils/
│       └── download_model.py # Model download utility
├── data/                    # SQLite database (created at runtime)
├── docker-compose.yml       # Docker configuration
├── Dockerfile              # Container build instructions
└── requirements.txt        # Python dependencies

Memory System

The server automatically categorizes conversation content into:

• Facts: Concrete information and data points
• Experiences: Personal stories and events
• Topics: Subject areas and themes discussed

Memory statistics are included in all chat responses and can be queried independently.

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.