ModelDock

Key features

• Live model catalog with an offline bundled fallback
• Install single models or whole categories
• Integrity verification and cache management
• Clean Architecture — pure domain and ports layers, no I/O
• Config via TOML or MODELDOCK_* environment variables

1. SDK + CLI

   modeldock / Typer

2. Core

   lifecycle + model manager

3. Ports

   RuntimePort protocol

4. Adapters

   Ollama · LM Studio · llama.cpp

1. 1The catalog is scraped from ollama.com/library and cached locally for 24 hours, with a bundled catalog as an offline fallback.
2. 2md.search, md.recommend and md.info help pick a model by name, capability or task.
3. 3md.install, md.update, md.remove and md.verify manage what is on disk.
4. 4Concrete runtimes implement a RuntimePort protocol, so the core never depends on a specific backend.

• Python 3.9+
• Typer
• Ollama
• LM Studio
• llama.cpp