lazyllama
a smol tool for managing local models
A command-line tool that scans your computer for locally installed AI language models across HuggingFace, Ollama, and LM Studio and shows them all in one unified view with file sizes, storage locations, and metadata.
lazyllama is a small command-line tool for people who run AI language models on their own computer. If you use tools like HuggingFace, Ollama, or LM Studio to download and run models locally, the files can end up scattered across different folders on your machine. lazyllama scans all of those locations and gives you a unified view of every model you have installed, along with details like file sizes and where things are stored.
Once installed via pip, you run it by typing its name in a terminal. With no options it shows a compact summary. You can switch to a more detailed view that lists individual filenames, or a tree view that shows the folder hierarchy. There are also flags to pull out just the model directories, output a tab-separated list, or dump everything as raw JSON for use in scripts.
The JSON output is the most thorough option. It includes metadata like model name, source, size, file count, format, quantization level, and timestamps. This makes the tool useful not just for browsing but also for piping data into other tools or building your own scripts on top of it.
You can also point it at custom directories if your models live somewhere outside the standard locations that HuggingFace, Ollama, or LM Studio normally use. A debug mode shows which directories were scanned and how long the scan took.
Where it fits
- See every AI model installed across HuggingFace, Ollama, and LM Studio in a single unified list, with file sizes and storage paths.
- Export a full model inventory as JSON -- including name, source, size, format, quantization, and timestamps -- to use in scripts or pipe into other tools.
- Scan custom directories where your models live outside the default HuggingFace, Ollama, or LM Studio locations.