Skip to content

Clients and providers

A client will send requests to LiteGate. The gateway will choose a destination allowed by configuration and policies, and the local engine or cloud provider will run the model.

The first delivery will focus on one OpenAI-compatible provider and Chat Completions, including streaming responses. Full compatibility with OpenAI Responses and Anthropic Messages belongs to a later phase.

LiteGate does not replace the inference engine: that engine must be installed and running separately. LiteGate will maintain the connection information.

API access depends on the provider’s credentials and permissions. A subscription to a chat product does not imply API access.

This guide will include verified connection recipes as the gateway is completed. In the meantime, you can validate the configuration.