STING-CE Fresh Installation Guide

Prerequisites

  1. Docker Desktop installed and running.
  2. Python 3.8+ installed.
  3. Internet connectivity for downloading Docker images.
  4. Hugging Face account with API token (optional but recommended).
  5. At least 20GB free disk space (10GB for models, 10GB for Docker images).

Installation Steps

Step 1: Clone the Repository

git clone https://github.com/your-repo/STING-CE.git
cd STING-CE/STING

Step 2: Download LLM Models (REQUIRED)

Before running the installer, you must download the LLM models:

# For testing/development (recommended - ~5GB)
./download_small_models.sh

# OR for production use (~15GB)
./download_optimized_models.sh

Models will be downloaded to: ~/Downloads/llm_models/

Step 3: Set Hugging Face Token (Optional)

export HF_TOKEN="your_huggingface_token_here"

Step 4: Run the Installer

./install_sting.sh

What the Installer Does

  1. Checks Prerequisites

    • Verifies Docker is running.
    • Checks network connectivity.
    • Verifies LLM models are pre-downloaded.

  2. Builds Docker Images

    • Creates base images for all services.
    • Configures networking and volumes.

  3. Starts Core Services

    • PostgreSQL database.
    • Vault for secrets management.
    • Kratos for authentication.
    • Frontend and backend services.

  4. Configures LLM Services

    • On macOS: Starts native Metal-accelerated service.
    • On Linux: Starts Docker-based LLM services.
    • Uses pre-downloaded models (no download during install).

Troubleshooting

Network Issues

If you see DNS resolution errors:

  1. Check your internet connection.
  2. Check Docker’s DNS settings.
  3. Restart Docker Desktop.
  4. Try using a different DNS server.

Model Download Issues

If models fail to download:

  1. Ensure you have a valid HF_TOKEN.
  2. Check disk space in ~/Downloads/.
  3. Try downloading models manually.
  4. Check firewall/proxy settings.

Installation Hangs

If installation appears stuck:

  1. Check ~/.sting-ce/logs/manage_sting.log.
  2. Ensure models were pre-downloaded.
  3. Check Docker container logs.
  4. Verify no conflicting services on required ports.

Post-Installation

After successful installation:

  1. Access the frontend at: https://localhost:8443.
  2. Register a new account.
  3. Check service status: ./manage_sting.sh status.

Required Ports

Ensure these ports are available:

  • 8443: Frontend
  • 5050: Backend API
  • 5432: PostgreSQL
  • 8200: Vault
  • 8080: LLM Gateway
  • 8081: Chatbot Service
  • 8086: Native LLM Service (macOS)

Default Model Configuration

The system now defaults to TinyLlama for better compatibility:

  • Model: TinyLlama-1.1B-Chat
  • Path: ~/Downloads/llm_models/TinyLlama-1.1B-Chat
  • Size: ~2.2GB

You can change models after installation using:

./sting-llm load <model_name>

Last updated: