CasAInit - Local AI Stack Bootstrapper for Arch Linux

CasAInit is a fully self-contained, POSIX-compliant shell bootstrapper that installs, configures, and integrates a complete local AI stack on Arch Linux systems.

Features

• Fully Automated: Single command installation of entire AI stack
• POSIX Compliant: Works with dash/sh, no bash dependencies
• Hardware Aware: Automatically detects CPU, RAM, GPU, and VRAM to install optimal models
• Self-Repairing: Built-in fix command to repair permissions and configurations
• Service Management: Unified interface for managing all AI services

Supported Tools

1. Ollama - Local LLM runtime (binary + systemd)
2. OpenCode - Code-focused AI assistant (binary + systemd)
3. Piper - Text-to-speech engine (binary + systemd)
4. OpenWebUI - Web interface for AI models (Docker container)
5. OpenDiffusion - Image generation (Docker container)
6. Continue - VS Code AI extension
7. MCP Tools - Additional AI tools from registry

Quick Start

One-line install (curl | sh safe):

curl -sL https://raw.githubusercontent.com/casapps/casainit/main/casainit | sh

Or download and run:

wget https://raw.githubusercontent.com/casapps/casainit/main/casainit
chmod +x casainit
./casainit

Commands

casainit                           # Install everything (default)
casainit status                    # Show installation status
casainit gpu                       # Show GPU information
casainit service start             # Start all services
casainit service restart ollama    # Restart specific service
casainit models remote             # List available models
casainit models installed          # Show installed models
casainit fix                       # Repair permissions and configs
casainit update                    # Update all tools
casainit uninstall                 # Remove CasAInit (preserves models)
casainit uninstall --remove-data   # Remove everything including models

Flags

• --raw - Disable colors and emojis (for scripting)
• --interactive / -i - Run in interactive mode (prompt for options)
• --yes / -y - Assume yes to all prompts (default)
• --remove-data - Remove all models and configurations during uninstall
• --help / -h - Show help message
• --version / -v - Show version
• --type=TYPE - Filter by type (ollama,tts,diffusion)

System Requirements

• Arch Linux
• 8GB+ RAM recommended
• 50GB+ free disk space
• Optional: NVIDIA/AMD GPU for acceleration

File Layout

$HOME/.local/bin/              # Binaries
$HOME/.local/share/casaicli/   # Archives and extracted tools
$HOME/.config/casaicli/        # Configurations
$HOME/.local/log/casaicli/     # Logs
/var/lib/casaicli/models/      # Shared model storage

Model Selection

CasaICLI automatically scores your system and installs appropriate models:

• Score = (CPU cores × 0.5) + RAM + VRAM
• Tier 1 (48+): High-end models (CodeLlama 13B, etc.)
• Tier 2 (32-47): Mid-range models (Mistral 7B, etc.)
• Tier 3 (<32): Lightweight models (Llama2 7B, etc.)

Configuration

All settings stored in ~/.config/casaicli/settings.env:

SYSTEM_CPU_CORES="8"
SYSTEM_RAM_GB="16"
SYSTEM_VRAM_GB="6"
SYSTEM_GPU_TYPE="nvidia"
INSTALLED_ollama="yes"
PORT_ollama="11434"

Service Management

Binary services use systemd:

systemctl --user status casaicli-ollama

Container services use Docker:

docker ps | grep casaicli

Troubleshooting

Fix permissions:

casainit fix

Check logs:

tail -f ~/.local/log/casainit/*.log

Reinstall specific tool:

casainit uninstall ollama
casainit install ollama

License

MIT License - See LICENSE file for details

Contributing

Contributions welcome! Please ensure:

• POSIX compliance (test with dash)
• No bashisms
• Follow existing code style
• Test on fresh Arch Linux install