A command-line utility to manage the Ollama cache, making it easier to maintain and move Ollama models between different storage locations. Useful for managing models across internal and external storage.

• List installed Ollama models with details about their storage location (internal/external cache)
• Toggle between internal and external Ollama cache directories using symbolic links
• Migrate model files between internal and external caches with:
  • SHA-256 checksum validation
  • Progress bars for large files
  • Metadata preservation
  • Blob file handling
• Remove models from either or both caches
• Pull models from Ollama.com
• Push models to Ollama.com (personal repositories only)
• Supports multiple model tags and versions

From PyPI:

pip install ollamautil

From source:

git clone https://github.com/yourusername/ollamautil.git
cd ollamautil
pip install -e .

Set the following environment variables before using:

# Required: Path to internal Ollama models directory
export OLLAMAUTIL_INTERNAL_DIR="/path/to/internal/ollama/models"

# Required: Path to external Ollama models directory
export OLLAMAUTIL_EXTERNAL_DIR="/path/to/external/ollama/models"

# Optional: List of files to ignore (defaults to [".DS_Store"])
export OLLAMAUTIL_FILE_IGNORE='[".DS_Store", "other_file_to_ignore"]'

Add these to your .bashrc, .zshrc, or equivalent to make them permanent.

Start the utility:

ollamautil

1. Copy - Migrate files between internal and external cache

   • Select models to copy
   • Choose direction (internal ↔ external)
   • Option to overwrite existing files
   • Automatic checksum validation

2. Toggle - Switch between internal and external cache

   • Shows current cache location
   • Option to migrate files before switching
   • Updates symbolic links automatically

3. Remove - Delete models from cache

   • Select models to remove
   • Remove from internal, external, or both
   • Confirmation required

4. Pull - Download models from Ollama.com

   • Select models to pull
   • Supports library and custom models
   • Shows download progress

5. Push - Upload models to Ollama.com

   • Supports personal repositories only
   • Format: username/model:tag

• prettytable - For displaying model tables
• tqdm - For progress bars during file operations
• ollama - For interacting with Ollama API

To contribute or modify:

1. Clone the repository
2. Install in development mode:

pip install -e .

3. Run tests:

python -m unittest tests/test_ollamautil.py

GPL-3.0 - See LICENSE.md for details