ANNOUNCEMENT: SMG has graduated from LightSeek Foundation

[lightseek-bot]

Today SMG (Shepherd Model Gateway), an engine-agnostic gateway for LLM inference, has officially graduated from LightSeek Foundation and moves to a new independent home:
https://github.com/smg-project/smg
SMG was the first project incubated by the LightSeek Foundation and was recently accepted into the @PyTorch Ecosystem. It has now reached the point where its next stage should be driven by the broader open-source community.
Why now?
Over the past few months — supporting Day 0 launches for leading open models and maintaining inference engines day to day — it became clear how much of the frontend and gateway layer is genuinely shared across engines. Some of those shared components are already running in production today, and maintainers from TokenSpeed and @vllm_project will keep building them with SMG. Rather than every project maintaining these pieces independently, they're better built once, together.
Model provider protocols have also kept evolving beyond the traditional OpenAI-compatible API, which makes this a good moment to revisit SMG's architecture for the next generation of model interfaces.
Fundamentally this move is about governance. SMG is built by contributors across several companies and inference projects, and it should stay that way: community-driven, maintained across the ecosystem, owned by a broader open-source community. Shared infrastructure makes everyone stronger.
Whether you maintain an inference engine, serve models at scale, or care about where AI serving infrastructure goes next, there's room for you here.

[lightseek-bot]

https://x.com/lightseekorg/status/2083236252728397926
https://x.com/vllm_project/status/2083270762610540817
https://x.com/PyTorch/status/2083276366838222933