Installation

1. LM Studio and its server

Install LM Studio, then start its local server, either from the app's Developer tab or with the lms CLI:

lms server start

The server listens on http://localhost:1234 by default. If it isn't running, every request from this gem raises a RubyLLM::Error that tells you to start it, rather than a raw connection failure.

2. A model

The gem talks to whatever models you have downloaded. Get one from the app's search, or with lms get:

lms get qwen/qwen3-4b                                    # a chat model
lms get https://huggingface.co/LiquidAI/d1-omni-600M-GGUF  # a decision model, for judgments
lms ls                                                   # what you have

Use the identifiers lms ls shows (qwen/qwen3-4b, d1-omni-600m) as model ids in Ruby — not OpenAI model names.

3. The gem

Add it to your Gemfile:

gem 'ruby_llm-providers-lms', require: 'ruby_llm/providers/lms'

and bundle install. Requiring ruby_llm/providers/lms registers the :lms provider with RubyLLM; there is nothing else to set up.

Requirements

Component	Version
Ruby	3.3 or newer (CI runs 3.3, 3.4 and 4.0)
RubyLLM	2.1.1 or newer
LM Studio	any recent build; 0.4.26 or newer for judgments

See Development for the exact versions the documentation was checked against.