locodellm

locodellm is a small toolkit for experimenting with local large language models. It wraps onnxruntime-genai behind a simple session API to load an ONNX model directory and generate text, supports multi-turn conversations, and ships helpers to build tiny dummy models for testing. A minimal command line interface is also provided.

Install

pip install -e .

Quick start

from locodellm.session import create_session

session = create_session("path/to/model")
session.generate("Once upon a time")
print(session.text)

Contents

• Examples
• Command Line Interface
• API
  • locodellm.session
  • generate
  • locodellm.bench
  • locodellm.test_models