lms

Why we included this project

LM Studio covers the day-to-day work of running local models, but the desktop app is a GUI-first experience. lms puts the same operations in a terminal: list and search downloaded models, load one with GPU offload and context settings, unload it, start or stop the local inference server, and stream server logs when a request misbehaves. That makes it a natural fit for automation. You can start the server, load a model, and point scripts or API clients at the local endpoint without touching the window. The memory-estimate flag helps you check whether a model fits on your hardware before committing to a load. For teams juggling several models or wanting stable identifiers for repeatable API calls, it is a useful layer between LM Studio and their own tooling.