🤖 Docker-based tool for managing Ollama models with backup and compose generation.

Language: English | 中文

OMO (Oh-My-Ollama / Ollama Model Organizer) is a Docker-based shell script for managing Ollama models. It provides essential model lifecycle operations including download, backup, restore, and removal, plus generates Docker Compose configurations for production deployments.

• Download Models: Install models from Ollama registry or HuggingFace repositories (only GGUF)
• List Models: Display installed models with detailed information
• Remove Models: Delete models individually or in batch
• Model Verification: Check model integrity after operations

• Complete Backup: Full model backup including manifests and blob files
• Integrity Verification: MD5 checksum validation for backup integrity
• Auto-Restore: --install automatically checks and restores from backup before downloading
• Manual Restore: Direct restore from backup (use only when needed)
• Batch Operations: Backup or restore multiple models

• Compose Generation: Create Docker Compose configurations
• Multi-Service Stack: Integrated Ollama, One-API, and Open-WebUI services
• GPU Support: Automatic NVIDIA GPU detection and configuration
• Production Ready: Proper volume management and networking

• Docker: Required for all operations
• nvidia gpu: for GPU acceleration

1. Clone the repository:

git clone https://github.com/LaiQE/omo.git
cd omo

2. Make the script executable:

chmod +x omo.sh

3. Edit the models list file:

# Edit models.list to add your desired models
# The repository includes a template with examples
vim models.list

# Show help
./omo.sh --help

# Install missing models from models.list
./omo.sh --install

# List installed models
./omo.sh --list

# Backup a specific model
./omo.sh --backup ollama:qwen2.5:7b-instruct

# Backup all models
./omo.sh --backup-all

# Restore from backup (manual, not recommended)
# Note: --install automatically restores from backup if available
./omo.sh --restore qwen2.5_7b-instruct

# Remove a model
./omo.sh --remove deepseek-r1:1.5b

# Remove all models
./omo.sh --remove-all

# Generate Docker Compose
./omo.sh --generate-compose

Create a models.list file with one model per line:

# Ollama registry models
ollama:deepseek-r1:1.5b
ollama:llama3.2:3b
ollama:qwen2.5:7b-instruct

# HuggingFace GGUF models
hf-gguf:hf.co/bartowski/Llama-3.2-1B-Instruct-GGUF:latest
hf-gguf:hf.co/MaziyarPanahi/gemma-3-1b-it-GGUF:q4_0

# Comments and empty lines are ignored
# Add your own models here

omo/
├── omo.sh              # Main script
├── models.list         # Model configuration
├── ollama/            # Ollama data directory
│   └── models/        # Model storage
├── backups/           # Model backups
└── docker-compose.yml # Generated compose file (optional)

Generate a complete Docker stack:

./omo.sh --generate-compose

This creates a docker-compose.yml with:

External References:

• Ollama - AI model runtime
• One-API - OpenAI API gateway
• Open-WebUI - Web interface

# Enable verbose logging
export VERBOSE="true"

# Use custom directories
./omo.sh --ollama-dir /custom/ollama --backup-dir /custom/backups

# 1. Check what models need to be downloaded
./omo.sh --check-only

# 2. Install missing models
./omo.sh --install

# 3. List all installed models
./omo.sh --list

# 4. Backup important models
./omo.sh --backup ollama:qwen2.5:7b-instruct

# 5. Generate Docker Compose for deployment
./omo.sh --generate-compose

# 6. Start the stack
docker-compose up -d

# Recommended workflow: Backup all models
./omo.sh --backup-all

# Install will automatically restore from backup if available
# This is the recommended way - no manual restore needed
./omo.sh --install

# Manual restore (only use when auto-restore doesn't work)
./omo.sh --restore qwen2.5_7b-instruct

# Force restore (overwrite existing)
./omo.sh --force --restore qwen2.5_7b-instruct

OMO includes basic error handling for:

• Network Issues: Clear error messages for download failures
• Docker Problems: Docker daemon status verification
• File Integrity: MD5 checksum validation for backups
• Model Conflicts: Confirmation prompts for destructive operations

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

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

• Ollama - Excellent LLM runtime platform
• Docker - Containerization platform

• Issues: GitHub Issues
• Discussions: GitHub Discussions

Author: Chain Lai
Repository: https://github.com/LaiQE/omo