OpenGateLLM is an open-source API gateway for generative AI. It offers a single OpenAI-compatible API behind which requests are routed to model providers and inference servers such as vLLM, Mistral, OpenAI and Text Embeddings Inference. Organisations use it to govern access to language models centrally, with roles, users and organisations, per-user budgets, rate limiting, inference monitoring and tracking of environmental footprint. It is installed as server software on your own infrastructure; the repository includes a Docker Compose example, among other things.
The project is developed under France's Direction interministérielle du numérique (DINUM); the MIT licence names DINUM as the copyright holder, and the documentation describes it as 'powered by French Government'. The code is freely available on GitHub. Because you run the gateway yourself, requests and data stay under your own control, and according to the README the gateway does not store chat history. Note that, according to its makers, the API is still in beta and major breaking changes may follow.