Truss

Truss by Baseten is a Python framework and CLI for packaging machine learning models as production API endpoints. You define serving logic in Python and Truss handles containerization, dependency management, and GPU configuration, then deploys to Baseten or any environment. It works with models from vLLM, SGLang, TensorRT-LLM, Transformers, Diffusers, PyTorch, and TensorFlow. Released under the MIT license.